Package ganeti :: Module cmdlib
[hide private]
[frames] | no frames]

Source Code for Module ganeti.cmdlib

    1  # 
    2  # 
    3   
    4  # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012 Google Inc. 
    5  # 
    6  # This program is free software; you can redistribute it and/or modify 
    7  # it under the terms of the GNU General Public License as published by 
    8  # the Free Software Foundation; either version 2 of the License, or 
    9  # (at your option) any later version. 
   10  # 
   11  # This program is distributed in the hope that it will be useful, but 
   12  # WITHOUT ANY WARRANTY; without even the implied warranty of 
   13  # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU 
   14  # General Public License for more details. 
   15  # 
   16  # You should have received a copy of the GNU General Public License 
   17  # along with this program; if not, write to the Free Software 
   18  # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 
   19  # 02110-1301, USA. 
   20   
   21   
   22  """Module implementing the master-side code.""" 
   23   
   24  # pylint: disable=W0201,C0302 
   25   
   26  # W0201 since most LU attributes are defined in CheckPrereq or similar 
   27  # functions 
   28   
   29  # C0302: since we have waaaay to many lines in this module 
   30   
   31  import os 
   32  import os.path 
   33  import time 
   34  import re 
   35  import platform 
   36  import logging 
   37  import copy 
   38  import OpenSSL 
   39  import socket 
   40  import tempfile 
   41  import shutil 
   42  import itertools 
   43  import operator 
   44   
   45  from ganeti import ssh 
   46  from ganeti import utils 
   47  from ganeti import errors 
   48  from ganeti import hypervisor 
   49  from ganeti import locking 
   50  from ganeti import constants 
   51  from ganeti import objects 
   52  from ganeti import serializer 
   53  from ganeti import ssconf 
   54  from ganeti import uidpool 
   55  from ganeti import compat 
   56  from ganeti import masterd 
   57  from ganeti import netutils 
   58  from ganeti import query 
   59  from ganeti import qlang 
   60  from ganeti import opcodes 
   61  from ganeti import ht 
   62   
   63  import ganeti.masterd.instance # pylint: disable=W0611 
64 65 66 -class ResultWithJobs:
67 """Data container for LU results with jobs. 68 69 Instances of this class returned from L{LogicalUnit.Exec} will be recognized 70 by L{mcpu.Processor._ProcessResult}. The latter will then submit the jobs 71 contained in the C{jobs} attribute and include the job IDs in the opcode 72 result. 73 74 """
75 - def __init__(self, jobs, **kwargs):
76 """Initializes this class. 77 78 Additional return values can be specified as keyword arguments. 79 80 @type jobs: list of lists of L{opcode.OpCode} 81 @param jobs: A list of lists of opcode objects 82 83 """ 84 self.jobs = jobs 85 self.other = kwargs
86
87 88 -class LogicalUnit(object):
89 """Logical Unit base class. 90 91 Subclasses must follow these rules: 92 - implement ExpandNames 93 - implement CheckPrereq (except when tasklets are used) 94 - implement Exec (except when tasklets are used) 95 - implement BuildHooksEnv 96 - implement BuildHooksNodes 97 - redefine HPATH and HTYPE 98 - optionally redefine their run requirements: 99 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively 100 101 Note that all commands require root permissions. 102 103 @ivar dry_run_result: the value (if any) that will be returned to the caller 104 in dry-run mode (signalled by opcode dry_run parameter) 105 106 """ 107 HPATH = None 108 HTYPE = None 109 REQ_BGL = True 110
111 - def __init__(self, processor, op, context, rpc):
112 """Constructor for LogicalUnit. 113 114 This needs to be overridden in derived classes in order to check op 115 validity. 116 117 """ 118 self.proc = processor 119 self.op = op 120 self.cfg = context.cfg 121 self.glm = context.glm 122 # readability alias 123 self.owned_locks = context.glm.list_owned 124 self.context = context 125 self.rpc = rpc 126 # Dicts used to declare locking needs to mcpu 127 self.needed_locks = None 128 self.share_locks = dict.fromkeys(locking.LEVELS, 0) 129 self.add_locks = {} 130 self.remove_locks = {} 131 # Used to force good behavior when calling helper functions 132 self.recalculate_locks = {} 133 # logging 134 self.Log = processor.Log # pylint: disable=C0103 135 self.LogWarning = processor.LogWarning # pylint: disable=C0103 136 self.LogInfo = processor.LogInfo # pylint: disable=C0103 137 self.LogStep = processor.LogStep # pylint: disable=C0103 138 # support for dry-run 139 self.dry_run_result = None 140 # support for generic debug attribute 141 if (not hasattr(self.op, "debug_level") or 142 not isinstance(self.op.debug_level, int)): 143 self.op.debug_level = 0 144 145 # Tasklets 146 self.tasklets = None 147 148 # Validate opcode parameters and set defaults 149 self.op.Validate(True) 150 151 self.CheckArguments()
152
153 - def CheckArguments(self):
154 """Check syntactic validity for the opcode arguments. 155 156 This method is for doing a simple syntactic check and ensure 157 validity of opcode parameters, without any cluster-related 158 checks. While the same can be accomplished in ExpandNames and/or 159 CheckPrereq, doing these separate is better because: 160 161 - ExpandNames is left as as purely a lock-related function 162 - CheckPrereq is run after we have acquired locks (and possible 163 waited for them) 164 165 The function is allowed to change the self.op attribute so that 166 later methods can no longer worry about missing parameters. 167 168 """ 169 pass
170
171 - def ExpandNames(self):
172 """Expand names for this LU. 173 174 This method is called before starting to execute the opcode, and it should 175 update all the parameters of the opcode to their canonical form (e.g. a 176 short node name must be fully expanded after this method has successfully 177 completed). This way locking, hooks, logging, etc. can work correctly. 178 179 LUs which implement this method must also populate the self.needed_locks 180 member, as a dict with lock levels as keys, and a list of needed lock names 181 as values. Rules: 182 183 - use an empty dict if you don't need any lock 184 - if you don't need any lock at a particular level omit that level 185 - don't put anything for the BGL level 186 - if you want all locks at a level use locking.ALL_SET as a value 187 188 If you need to share locks (rather than acquire them exclusively) at one 189 level you can modify self.share_locks, setting a true value (usually 1) for 190 that level. By default locks are not shared. 191 192 This function can also define a list of tasklets, which then will be 193 executed in order instead of the usual LU-level CheckPrereq and Exec 194 functions, if those are not defined by the LU. 195 196 Examples:: 197 198 # Acquire all nodes and one instance 199 self.needed_locks = { 200 locking.LEVEL_NODE: locking.ALL_SET, 201 locking.LEVEL_INSTANCE: ['instance1.example.com'], 202 } 203 # Acquire just two nodes 204 self.needed_locks = { 205 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'], 206 } 207 # Acquire no locks 208 self.needed_locks = {} # No, you can't leave it to the default value None 209 210 """ 211 # The implementation of this method is mandatory only if the new LU is 212 # concurrent, so that old LUs don't need to be changed all at the same 213 # time. 214 if self.REQ_BGL: 215 self.needed_locks = {} # Exclusive LUs don't need locks. 216 else: 217 raise NotImplementedError
218
219 - def DeclareLocks(self, level):
220 """Declare LU locking needs for a level 221 222 While most LUs can just declare their locking needs at ExpandNames time, 223 sometimes there's the need to calculate some locks after having acquired 224 the ones before. This function is called just before acquiring locks at a 225 particular level, but after acquiring the ones at lower levels, and permits 226 such calculations. It can be used to modify self.needed_locks, and by 227 default it does nothing. 228 229 This function is only called if you have something already set in 230 self.needed_locks for the level. 231 232 @param level: Locking level which is going to be locked 233 @type level: member of ganeti.locking.LEVELS 234 235 """
236
237 - def CheckPrereq(self):
238 """Check prerequisites for this LU. 239 240 This method should check that the prerequisites for the execution 241 of this LU are fulfilled. It can do internode communication, but 242 it should be idempotent - no cluster or system changes are 243 allowed. 244 245 The method should raise errors.OpPrereqError in case something is 246 not fulfilled. Its return value is ignored. 247 248 This method should also update all the parameters of the opcode to 249 their canonical form if it hasn't been done by ExpandNames before. 250 251 """ 252 if self.tasklets is not None: 253 for (idx, tl) in enumerate(self.tasklets): 254 logging.debug("Checking prerequisites for tasklet %s/%s", 255 idx + 1, len(self.tasklets)) 256 tl.CheckPrereq() 257 else: 258 pass
259
260 - def Exec(self, feedback_fn):
261 """Execute the LU. 262 263 This method should implement the actual work. It should raise 264 errors.OpExecError for failures that are somewhat dealt with in 265 code, or expected. 266 267 """ 268 if self.tasklets is not None: 269 for (idx, tl) in enumerate(self.tasklets): 270 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets)) 271 tl.Exec(feedback_fn) 272 else: 273 raise NotImplementedError
274
275 - def BuildHooksEnv(self):
276 """Build hooks environment for this LU. 277 278 @rtype: dict 279 @return: Dictionary containing the environment that will be used for 280 running the hooks for this LU. The keys of the dict must not be prefixed 281 with "GANETI_"--that'll be added by the hooks runner. The hooks runner 282 will extend the environment with additional variables. If no environment 283 should be defined, an empty dictionary should be returned (not C{None}). 284 @note: If the C{HPATH} attribute of the LU class is C{None}, this function 285 will not be called. 286 287 """ 288 raise NotImplementedError
289
290 - def BuildHooksNodes(self):
291 """Build list of nodes to run LU's hooks. 292 293 @rtype: tuple; (list, list) 294 @return: Tuple containing a list of node names on which the hook 295 should run before the execution and a list of node names on which the 296 hook should run after the execution. No nodes should be returned as an 297 empty list (and not None). 298 @note: If the C{HPATH} attribute of the LU class is C{None}, this function 299 will not be called. 300 301 """ 302 raise NotImplementedError
303
304 - def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
305 """Notify the LU about the results of its hooks. 306 307 This method is called every time a hooks phase is executed, and notifies 308 the Logical Unit about the hooks' result. The LU can then use it to alter 309 its result based on the hooks. By default the method does nothing and the 310 previous result is passed back unchanged but any LU can define it if it 311 wants to use the local cluster hook-scripts somehow. 312 313 @param phase: one of L{constants.HOOKS_PHASE_POST} or 314 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase 315 @param hook_results: the results of the multi-node hooks rpc call 316 @param feedback_fn: function used send feedback back to the caller 317 @param lu_result: the previous Exec result this LU had, or None 318 in the PRE phase 319 @return: the new Exec result, based on the previous result 320 and hook results 321 322 """ 323 # API must be kept, thus we ignore the unused argument and could 324 # be a function warnings 325 # pylint: disable=W0613,R0201 326 return lu_result
327
328 - def _ExpandAndLockInstance(self):
329 """Helper function to expand and lock an instance. 330 331 Many LUs that work on an instance take its name in self.op.instance_name 332 and need to expand it and then declare the expanded name for locking. This 333 function does it, and then updates self.op.instance_name to the expanded 334 name. It also initializes needed_locks as a dict, if this hasn't been done 335 before. 336 337 """ 338 if self.needed_locks is None: 339 self.needed_locks = {} 340 else: 341 assert locking.LEVEL_INSTANCE not in self.needed_locks, \ 342 "_ExpandAndLockInstance called with instance-level locks set" 343 self.op.instance_name = _ExpandInstanceName(self.cfg, 344 self.op.instance_name) 345 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
346
347 - def _LockInstancesNodes(self, primary_only=False):
348 """Helper function to declare instances' nodes for locking. 349 350 This function should be called after locking one or more instances to lock 351 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE] 352 with all primary or secondary nodes for instances already locked and 353 present in self.needed_locks[locking.LEVEL_INSTANCE]. 354 355 It should be called from DeclareLocks, and for safety only works if 356 self.recalculate_locks[locking.LEVEL_NODE] is set. 357 358 In the future it may grow parameters to just lock some instance's nodes, or 359 to just lock primaries or secondary nodes, if needed. 360 361 If should be called in DeclareLocks in a way similar to:: 362 363 if level == locking.LEVEL_NODE: 364 self._LockInstancesNodes() 365 366 @type primary_only: boolean 367 @param primary_only: only lock primary nodes of locked instances 368 369 """ 370 assert locking.LEVEL_NODE in self.recalculate_locks, \ 371 "_LockInstancesNodes helper function called with no nodes to recalculate" 372 373 # TODO: check if we're really been called with the instance locks held 374 375 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the 376 # future we might want to have different behaviors depending on the value 377 # of self.recalculate_locks[locking.LEVEL_NODE] 378 wanted_nodes = [] 379 locked_i = self.owned_locks(locking.LEVEL_INSTANCE) 380 for _, instance in self.cfg.GetMultiInstanceInfo(locked_i): 381 wanted_nodes.append(instance.primary_node) 382 if not primary_only: 383 wanted_nodes.extend(instance.secondary_nodes) 384 385 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE: 386 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes 387 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND: 388 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes) 389 390 del self.recalculate_locks[locking.LEVEL_NODE]
391
392 393 -class NoHooksLU(LogicalUnit): # pylint: disable=W0223
394 """Simple LU which runs no hooks. 395 396 This LU is intended as a parent for other LogicalUnits which will 397 run no hooks, in order to reduce duplicate code. 398 399 """ 400 HPATH = None 401 HTYPE = None 402
403 - def BuildHooksEnv(self):
404 """Empty BuildHooksEnv for NoHooksLu. 405 406 This just raises an error. 407 408 """ 409 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
410
411 - def BuildHooksNodes(self):
412 """Empty BuildHooksNodes for NoHooksLU. 413 414 """ 415 raise AssertionError("BuildHooksNodes called for NoHooksLU")
416
417 418 -class Tasklet:
419 """Tasklet base class. 420 421 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or 422 they can mix legacy code with tasklets. Locking needs to be done in the LU, 423 tasklets know nothing about locks. 424 425 Subclasses must follow these rules: 426 - Implement CheckPrereq 427 - Implement Exec 428 429 """
430 - def __init__(self, lu):
431 self.lu = lu 432 433 # Shortcuts 434 self.cfg = lu.cfg 435 self.rpc = lu.rpc
436
437 - def CheckPrereq(self):
438 """Check prerequisites for this tasklets. 439 440 This method should check whether the prerequisites for the execution of 441 this tasklet are fulfilled. It can do internode communication, but it 442 should be idempotent - no cluster or system changes are allowed. 443 444 The method should raise errors.OpPrereqError in case something is not 445 fulfilled. Its return value is ignored. 446 447 This method should also update all parameters to their canonical form if it 448 hasn't been done before. 449 450 """ 451 pass
452
453 - def Exec(self, feedback_fn):
454 """Execute the tasklet. 455 456 This method should implement the actual work. It should raise 457 errors.OpExecError for failures that are somewhat dealt with in code, or 458 expected. 459 460 """ 461 raise NotImplementedError
462
463 464 -class _QueryBase:
465 """Base for query utility classes. 466 467 """ 468 #: Attribute holding field definitions 469 FIELDS = None 470
471 - def __init__(self, filter_, fields, use_locking):
472 """Initializes this class. 473 474 """ 475 self.use_locking = use_locking 476 477 self.query = query.Query(self.FIELDS, fields, filter_=filter_, 478 namefield="name") 479 self.requested_data = self.query.RequestedData() 480 self.names = self.query.RequestedNames() 481 482 # Sort only if no names were requested 483 self.sort_by_name = not self.names 484 485 self.do_locking = None 486 self.wanted = None
487
488 - def _GetNames(self, lu, all_names, lock_level):
489 """Helper function to determine names asked for in the query. 490 491 """ 492 if self.do_locking: 493 names = lu.owned_locks(lock_level) 494 else: 495 names = all_names 496 497 if self.wanted == locking.ALL_SET: 498 assert not self.names 499 # caller didn't specify names, so ordering is not important 500 return utils.NiceSort(names) 501 502 # caller specified names and we must keep the same order 503 assert self.names 504 assert not self.do_locking or lu.glm.is_owned(lock_level) 505 506 missing = set(self.wanted).difference(names) 507 if missing: 508 raise errors.OpExecError("Some items were removed before retrieving" 509 " their data: %s" % missing) 510 511 # Return expanded names 512 return self.wanted
513
514 - def ExpandNames(self, lu):
515 """Expand names for this query. 516 517 See L{LogicalUnit.ExpandNames}. 518 519 """ 520 raise NotImplementedError()
521
522 - def DeclareLocks(self, lu, level):
523 """Declare locks for this query. 524 525 See L{LogicalUnit.DeclareLocks}. 526 527 """ 528 raise NotImplementedError()
529
530 - def _GetQueryData(self, lu):
531 """Collects all data for this query. 532 533 @return: Query data object 534 535 """ 536 raise NotImplementedError()
537
538 - def NewStyleQuery(self, lu):
539 """Collect data and execute query. 540 541 """ 542 return query.GetQueryResponse(self.query, self._GetQueryData(lu), 543 sort_by_name=self.sort_by_name)
544
545 - def OldStyleQuery(self, lu):
546 """Collect data and execute query. 547 548 """ 549 return self.query.OldStyleQuery(self._GetQueryData(lu), 550 sort_by_name=self.sort_by_name)
551
552 553 -def _ShareAll():
554 """Returns a dict declaring all lock levels shared. 555 556 """ 557 return dict.fromkeys(locking.LEVELS, 1)
558
559 560 -def _CheckInstanceNodeGroups(cfg, instance_name, owned_groups):
561 """Checks if the owned node groups are still correct for an instance. 562 563 @type cfg: L{config.ConfigWriter} 564 @param cfg: The cluster configuration 565 @type instance_name: string 566 @param instance_name: Instance name 567 @type owned_groups: set or frozenset 568 @param owned_groups: List of currently owned node groups 569 570 """ 571 inst_groups = cfg.GetInstanceNodeGroups(instance_name) 572 573 if not owned_groups.issuperset(inst_groups): 574 raise errors.OpPrereqError("Instance %s's node groups changed since" 575 " locks were acquired, current groups are" 576 " are '%s', owning groups '%s'; retry the" 577 " operation" % 578 (instance_name, 579 utils.CommaJoin(inst_groups), 580 utils.CommaJoin(owned_groups)), 581 errors.ECODE_STATE) 582 583 return inst_groups
584
585 586 -def _CheckNodeGroupInstances(cfg, group_uuid, owned_instances):
587 """Checks if the instances in a node group are still correct. 588 589 @type cfg: L{config.ConfigWriter} 590 @param cfg: The cluster configuration 591 @type group_uuid: string 592 @param group_uuid: Node group UUID 593 @type owned_instances: set or frozenset 594 @param owned_instances: List of currently owned instances 595 596 """ 597 wanted_instances = cfg.GetNodeGroupInstances(group_uuid) 598 if owned_instances != wanted_instances: 599 raise errors.OpPrereqError("Instances in node group '%s' changed since" 600 " locks were acquired, wanted '%s', have '%s';" 601 " retry the operation" % 602 (group_uuid, 603 utils.CommaJoin(wanted_instances), 604 utils.CommaJoin(owned_instances)), 605 errors.ECODE_STATE) 606 607 return wanted_instances
608
609 610 -def _SupportsOob(cfg, node):
611 """Tells if node supports OOB. 612 613 @type cfg: L{config.ConfigWriter} 614 @param cfg: The cluster configuration 615 @type node: L{objects.Node} 616 @param node: The node 617 @return: The OOB script if supported or an empty string otherwise 618 619 """ 620 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
621
622 623 -def _GetWantedNodes(lu, nodes):
624 """Returns list of checked and expanded node names. 625 626 @type lu: L{LogicalUnit} 627 @param lu: the logical unit on whose behalf we execute 628 @type nodes: list 629 @param nodes: list of node names or None for all nodes 630 @rtype: list 631 @return: the list of nodes, sorted 632 @raise errors.ProgrammerError: if the nodes parameter is wrong type 633 634 """ 635 if nodes: 636 return [_ExpandNodeName(lu.cfg, name) for name in nodes] 637 638 return utils.NiceSort(lu.cfg.GetNodeList())
639
640 641 -def _GetWantedInstances(lu, instances):
642 """Returns list of checked and expanded instance names. 643 644 @type lu: L{LogicalUnit} 645 @param lu: the logical unit on whose behalf we execute 646 @type instances: list 647 @param instances: list of instance names or None for all instances 648 @rtype: list 649 @return: the list of instances, sorted 650 @raise errors.OpPrereqError: if the instances parameter is wrong type 651 @raise errors.OpPrereqError: if any of the passed instances is not found 652 653 """ 654 if instances: 655 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances] 656 else: 657 wanted = utils.NiceSort(lu.cfg.GetInstanceList()) 658 return wanted
659
660 661 -def _GetUpdatedParams(old_params, update_dict, 662 use_default=True, use_none=False):
663 """Return the new version of a parameter dictionary. 664 665 @type old_params: dict 666 @param old_params: old parameters 667 @type update_dict: dict 668 @param update_dict: dict containing new parameter values, or 669 constants.VALUE_DEFAULT to reset the parameter to its default 670 value 671 @param use_default: boolean 672 @type use_default: whether to recognise L{constants.VALUE_DEFAULT} 673 values as 'to be deleted' values 674 @param use_none: boolean 675 @type use_none: whether to recognise C{None} values as 'to be 676 deleted' values 677 @rtype: dict 678 @return: the new parameter dictionary 679 680 """ 681 params_copy = copy.deepcopy(old_params) 682 for key, val in update_dict.iteritems(): 683 if ((use_default and val == constants.VALUE_DEFAULT) or 684 (use_none and val is None)): 685 try: 686 del params_copy[key] 687 except KeyError: 688 pass 689 else: 690 params_copy[key] = val 691 return params_copy
692
693 694 -def _ReleaseLocks(lu, level, names=None, keep=None):
695 """Releases locks owned by an LU. 696 697 @type lu: L{LogicalUnit} 698 @param level: Lock level 699 @type names: list or None 700 @param names: Names of locks to release 701 @type keep: list or None 702 @param keep: Names of locks to retain 703 704 """ 705 assert not (keep is not None and names is not None), \ 706 "Only one of the 'names' and the 'keep' parameters can be given" 707 708 if names is not None: 709 should_release = names.__contains__ 710 elif keep: 711 should_release = lambda name: name not in keep 712 else: 713 should_release = None 714 715 if should_release: 716 retain = [] 717 release = [] 718 719 # Determine which locks to release 720 for name in lu.owned_locks(level): 721 if should_release(name): 722 release.append(name) 723 else: 724 retain.append(name) 725 726 assert len(lu.owned_locks(level)) == (len(retain) + len(release)) 727 728 # Release just some locks 729 lu.glm.release(level, names=release) 730 731 assert frozenset(lu.owned_locks(level)) == frozenset(retain) 732 else: 733 # Release everything 734 lu.glm.release(level) 735 736 assert not lu.glm.is_owned(level), "No locks should be owned"
737
738 739 -def _MapInstanceDisksToNodes(instances):
740 """Creates a map from (node, volume) to instance name. 741 742 @type instances: list of L{objects.Instance} 743 @rtype: dict; tuple of (node name, volume name) as key, instance name as value 744 745 """ 746 return dict(((node, vol), inst.name) 747 for inst in instances 748 for (node, vols) in inst.MapLVsByNode().items() 749 for vol in vols)
750
751 752 -def _RunPostHook(lu, node_name):
753 """Runs the post-hook for an opcode on a single node. 754 755 """ 756 hm = lu.proc.hmclass(lu.rpc.call_hooks_runner, lu) 757 try: 758 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name]) 759 except: 760 # pylint: disable=W0702 761 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
762
763 764 -def _CheckOutputFields(static, dynamic, selected):
765 """Checks whether all selected fields are valid. 766 767 @type static: L{utils.FieldSet} 768 @param static: static fields set 769 @type dynamic: L{utils.FieldSet} 770 @param dynamic: dynamic fields set 771 772 """ 773 f = utils.FieldSet() 774 f.Extend(static) 775 f.Extend(dynamic) 776 777 delta = f.NonMatching(selected) 778 if delta: 779 raise errors.OpPrereqError("Unknown output fields selected: %s" 780 % ",".join(delta), errors.ECODE_INVAL)
781
782 783 -def _CheckGlobalHvParams(params):
784 """Validates that given hypervisor params are not global ones. 785 786 This will ensure that instances don't get customised versions of 787 global params. 788 789 """ 790 used_globals = constants.HVC_GLOBALS.intersection(params) 791 if used_globals: 792 msg = ("The following hypervisor parameters are global and cannot" 793 " be customized at instance level, please modify them at" 794 " cluster level: %s" % utils.CommaJoin(used_globals)) 795 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
796
797 798 -def _CheckNodeOnline(lu, node, msg=None):
799 """Ensure that a given node is online. 800 801 @param lu: the LU on behalf of which we make the check 802 @param node: the node to check 803 @param msg: if passed, should be a message to replace the default one 804 @raise errors.OpPrereqError: if the node is offline 805 806 """ 807 if msg is None: 808 msg = "Can't use offline node" 809 if lu.cfg.GetNodeInfo(node).offline: 810 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
811
812 813 -def _CheckNodeNotDrained(lu, node):
814 """Ensure that a given node is not drained. 815 816 @param lu: the LU on behalf of which we make the check 817 @param node: the node to check 818 @raise errors.OpPrereqError: if the node is drained 819 820 """ 821 if lu.cfg.GetNodeInfo(node).drained: 822 raise errors.OpPrereqError("Can't use drained node %s" % node, 823 errors.ECODE_STATE)
824
825 826 -def _CheckNodeVmCapable(lu, node):
827 """Ensure that a given node is vm capable. 828 829 @param lu: the LU on behalf of which we make the check 830 @param node: the node to check 831 @raise errors.OpPrereqError: if the node is not vm capable 832 833 """ 834 if not lu.cfg.GetNodeInfo(node).vm_capable: 835 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node, 836 errors.ECODE_STATE)
837
838 839 -def _CheckNodeHasOS(lu, node, os_name, force_variant):
840 """Ensure that a node supports a given OS. 841 842 @param lu: the LU on behalf of which we make the check 843 @param node: the node to check 844 @param os_name: the OS to query about 845 @param force_variant: whether to ignore variant errors 846 @raise errors.OpPrereqError: if the node is not supporting the OS 847 848 """ 849 result = lu.rpc.call_os_get(node, os_name) 850 result.Raise("OS '%s' not in supported OS list for node %s" % 851 (os_name, node), 852 prereq=True, ecode=errors.ECODE_INVAL) 853 if not force_variant: 854 _CheckOSVariant(result.payload, os_name)
855
856 857 -def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
858 """Ensure that a node has the given secondary ip. 859 860 @type lu: L{LogicalUnit} 861 @param lu: the LU on behalf of which we make the check 862 @type node: string 863 @param node: the node to check 864 @type secondary_ip: string 865 @param secondary_ip: the ip to check 866 @type prereq: boolean 867 @param prereq: whether to throw a prerequisite or an execute error 868 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True 869 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False 870 871 """ 872 result = lu.rpc.call_node_has_ip_address(node, secondary_ip) 873 result.Raise("Failure checking secondary ip on node %s" % node, 874 prereq=prereq, ecode=errors.ECODE_ENVIRON) 875 if not result.payload: 876 msg = ("Node claims it doesn't have the secondary ip you gave (%s)," 877 " please fix and re-run this command" % secondary_ip) 878 if prereq: 879 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON) 880 else: 881 raise errors.OpExecError(msg)
882
883 884 -def _GetClusterDomainSecret():
885 """Reads the cluster domain secret. 886 887 """ 888 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE, 889 strict=True)
890
891 892 -def _CheckInstanceDown(lu, instance, reason):
893 """Ensure that an instance is not running.""" 894 if instance.admin_up: 895 raise errors.OpPrereqError("Instance %s is marked to be up, %s" % 896 (instance.name, reason), errors.ECODE_STATE) 897 898 pnode = instance.primary_node 899 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode] 900 ins_l.Raise("Can't contact node %s for instance information" % pnode, 901 prereq=True, ecode=errors.ECODE_ENVIRON) 902 903 if instance.name in ins_l.payload: 904 raise errors.OpPrereqError("Instance %s is running, %s" % 905 (instance.name, reason), errors.ECODE_STATE)
906
907 908 -def _ExpandItemName(fn, name, kind):
909 """Expand an item name. 910 911 @param fn: the function to use for expansion 912 @param name: requested item name 913 @param kind: text description ('Node' or 'Instance') 914 @return: the resolved (full) name 915 @raise errors.OpPrereqError: if the item is not found 916 917 """ 918 full_name = fn(name) 919 if full_name is None: 920 raise errors.OpPrereqError("%s '%s' not known" % (kind, name), 921 errors.ECODE_NOENT) 922 return full_name
923
924 925 -def _ExpandNodeName(cfg, name):
926 """Wrapper over L{_ExpandItemName} for nodes.""" 927 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
928
929 930 -def _ExpandInstanceName(cfg, name):
931 """Wrapper over L{_ExpandItemName} for instance.""" 932 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
933
934 935 -def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status, 936 memory, vcpus, nics, disk_template, disks, 937 bep, hvp, hypervisor_name, tags):
938 """Builds instance related env variables for hooks 939 940 This builds the hook environment from individual variables. 941 942 @type name: string 943 @param name: the name of the instance 944 @type primary_node: string 945 @param primary_node: the name of the instance's primary node 946 @type secondary_nodes: list 947 @param secondary_nodes: list of secondary nodes as strings 948 @type os_type: string 949 @param os_type: the name of the instance's OS 950 @type status: boolean 951 @param status: the should_run status of the instance 952 @type memory: string 953 @param memory: the memory size of the instance 954 @type vcpus: string 955 @param vcpus: the count of VCPUs the instance has 956 @type nics: list 957 @param nics: list of tuples (ip, mac, mode, link) representing 958 the NICs the instance has 959 @type disk_template: string 960 @param disk_template: the disk template of the instance 961 @type disks: list 962 @param disks: the list of (size, mode) pairs 963 @type bep: dict 964 @param bep: the backend parameters for the instance 965 @type hvp: dict 966 @param hvp: the hypervisor parameters for the instance 967 @type hypervisor_name: string 968 @param hypervisor_name: the hypervisor for the instance 969 @type tags: list 970 @param tags: list of instance tags as strings 971 @rtype: dict 972 @return: the hook environment for this instance 973 974 """ 975 if status: 976 str_status = "up" 977 else: 978 str_status = "down" 979 env = { 980 "OP_TARGET": name, 981 "INSTANCE_NAME": name, 982 "INSTANCE_PRIMARY": primary_node, 983 "INSTANCE_SECONDARIES": " ".join(secondary_nodes), 984 "INSTANCE_OS_TYPE": os_type, 985 "INSTANCE_STATUS": str_status, 986 "INSTANCE_MEMORY": memory, 987 "INSTANCE_VCPUS": vcpus, 988 "INSTANCE_DISK_TEMPLATE": disk_template, 989 "INSTANCE_HYPERVISOR": hypervisor_name, 990 } 991 992 if nics: 993 nic_count = len(nics) 994 for idx, (ip, mac, mode, link) in enumerate(nics): 995 if ip is None: 996 ip = "" 997 env["INSTANCE_NIC%d_IP" % idx] = ip 998 env["INSTANCE_NIC%d_MAC" % idx] = mac 999 env["INSTANCE_NIC%d_MODE" % idx] = mode 1000 env["INSTANCE_NIC%d_LINK" % idx] = link 1001 if mode == constants.NIC_MODE_BRIDGED: 1002 env["INSTANCE_NIC%d_BRIDGE" % idx] = link 1003 else: 1004 nic_count = 0 1005 1006 env["INSTANCE_NIC_COUNT"] = nic_count 1007 1008 if disks: 1009 disk_count = len(disks) 1010 for idx, (size, mode) in enumerate(disks): 1011 env["INSTANCE_DISK%d_SIZE" % idx] = size 1012 env["INSTANCE_DISK%d_MODE" % idx] = mode 1013 else: 1014 disk_count = 0 1015 1016 env["INSTANCE_DISK_COUNT"] = disk_count 1017 1018 if not tags: 1019 tags = [] 1020 1021 env["INSTANCE_TAGS"] = " ".join(tags) 1022 1023 for source, kind in [(bep, "BE"), (hvp, "HV")]: 1024 for key, value in source.items(): 1025 env["INSTANCE_%s_%s" % (kind, key)] = value 1026 1027 return env
1028
1029 1030 -def _NICListToTuple(lu, nics):
1031 """Build a list of nic information tuples. 1032 1033 This list is suitable to be passed to _BuildInstanceHookEnv or as a return 1034 value in LUInstanceQueryData. 1035 1036 @type lu: L{LogicalUnit} 1037 @param lu: the logical unit on whose behalf we execute 1038 @type nics: list of L{objects.NIC} 1039 @param nics: list of nics to convert to hooks tuples 1040 1041 """ 1042 hooks_nics = [] 1043 cluster = lu.cfg.GetClusterInfo() 1044 for nic in nics: 1045 ip = nic.ip 1046 mac = nic.mac 1047 filled_params = cluster.SimpleFillNIC(nic.nicparams) 1048 mode = filled_params[constants.NIC_MODE] 1049 link = filled_params[constants.NIC_LINK] 1050 hooks_nics.append((ip, mac, mode, link)) 1051 return hooks_nics
1052
1053 1054 -def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1055 """Builds instance related env variables for hooks from an object. 1056 1057 @type lu: L{LogicalUnit} 1058 @param lu: the logical unit on whose behalf we execute 1059 @type instance: L{objects.Instance} 1060 @param instance: the instance for which we should build the 1061 environment 1062 @type override: dict 1063 @param override: dictionary with key/values that will override 1064 our values 1065 @rtype: dict 1066 @return: the hook environment dictionary 1067 1068 """ 1069 cluster = lu.cfg.GetClusterInfo() 1070 bep = cluster.FillBE(instance) 1071 hvp = cluster.FillHV(instance) 1072 args = { 1073 "name": instance.name, 1074 "primary_node": instance.primary_node, 1075 "secondary_nodes": instance.secondary_nodes, 1076 "os_type": instance.os, 1077 "status": instance.admin_up, 1078 "memory": bep[constants.BE_MEMORY], 1079 "vcpus": bep[constants.BE_VCPUS], 1080 "nics": _NICListToTuple(lu, instance.nics), 1081 "disk_template": instance.disk_template, 1082 "disks": [(disk.size, disk.mode) for disk in instance.disks], 1083 "bep": bep, 1084 "hvp": hvp, 1085 "hypervisor_name": instance.hypervisor, 1086 "tags": instance.tags, 1087 } 1088 if override: 1089 args.update(override) 1090 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
1091
1092 1093 -def _AdjustCandidatePool(lu, exceptions):
1094 """Adjust the candidate pool after node operations. 1095 1096 """ 1097 mod_list = lu.cfg.MaintainCandidatePool(exceptions) 1098 if mod_list: 1099 lu.LogInfo("Promoted nodes to master candidate role: %s", 1100 utils.CommaJoin(node.name for node in mod_list)) 1101 for name in mod_list: 1102 lu.context.ReaddNode(name) 1103 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions) 1104 if mc_now > mc_max: 1105 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" % 1106 (mc_now, mc_max))
1107
1108 1109 -def _DecideSelfPromotion(lu, exceptions=None):
1110 """Decide whether I should promote myself as a master candidate. 1111 1112 """ 1113 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size 1114 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions) 1115 # the new node will increase mc_max with one, so: 1116 mc_should = min(mc_should + 1, cp_size) 1117 return mc_now < mc_should
1118
1119 1120 -def _CheckNicsBridgesExist(lu, target_nics, target_node):
1121 """Check that the brigdes needed by a list of nics exist. 1122 1123 """ 1124 cluster = lu.cfg.GetClusterInfo() 1125 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics] 1126 brlist = [params[constants.NIC_LINK] for params in paramslist 1127 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED] 1128 if brlist: 1129 result = lu.rpc.call_bridges_exist(target_node, brlist) 1130 result.Raise("Error checking bridges on destination node '%s'" % 1131 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1132
1133 1134 -def _CheckInstanceBridgesExist(lu, instance, node=None):
1135 """Check that the brigdes needed by an instance exist. 1136 1137 """ 1138 if node is None: 1139 node = instance.primary_node 1140 _CheckNicsBridgesExist(lu, instance.nics, node)
1141
1142 1143 -def _CheckOSVariant(os_obj, name):
1144 """Check whether an OS name conforms to the os variants specification. 1145 1146 @type os_obj: L{objects.OS} 1147 @param os_obj: OS object to check 1148 @type name: string 1149 @param name: OS name passed by the user, to check for validity 1150 1151 """ 1152 variant = objects.OS.GetVariant(name) 1153 if not os_obj.supported_variants: 1154 if variant: 1155 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'" 1156 " passed)" % (os_obj.name, variant), 1157 errors.ECODE_INVAL) 1158 return 1159 if not variant: 1160 raise errors.OpPrereqError("OS name must include a variant", 1161 errors.ECODE_INVAL) 1162 1163 if variant not in os_obj.supported_variants: 1164 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1165
1166 1167 -def _GetNodeInstancesInner(cfg, fn):
1168 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1169
1170 1171 -def _GetNodeInstances(cfg, node_name):
1172 """Returns a list of all primary and secondary instances on a node. 1173 1174 """ 1175 1176 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1177
1178 1179 -def _GetNodePrimaryInstances(cfg, node_name):
1180 """Returns primary instances on a node. 1181 1182 """ 1183 return _GetNodeInstancesInner(cfg, 1184 lambda inst: node_name == inst.primary_node)
1185
1186 1187 -def _GetNodeSecondaryInstances(cfg, node_name):
1188 """Returns secondary instances on a node. 1189 1190 """ 1191 return _GetNodeInstancesInner(cfg, 1192 lambda inst: node_name in inst.secondary_nodes)
1193
1194 1195 -def _GetStorageTypeArgs(cfg, storage_type):
1196 """Returns the arguments for a storage type. 1197 1198 """ 1199 # Special case for file storage 1200 if storage_type == constants.ST_FILE: 1201 # storage.FileStorage wants a list of storage directories 1202 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]] 1203 1204 return []
1205
1206 1207 -def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1208 faulty = [] 1209 1210 for dev in instance.disks: 1211 cfg.SetDiskID(dev, node_name) 1212 1213 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks) 1214 result.Raise("Failed to get disk status from node %s" % node_name, 1215 prereq=prereq, ecode=errors.ECODE_ENVIRON) 1216 1217 for idx, bdev_status in enumerate(result.payload): 1218 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY: 1219 faulty.append(idx) 1220 1221 return faulty
1222
1223 1224 -def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1225 """Check the sanity of iallocator and node arguments and use the 1226 cluster-wide iallocator if appropriate. 1227 1228 Check that at most one of (iallocator, node) is specified. If none is 1229 specified, then the LU's opcode's iallocator slot is filled with the 1230 cluster-wide default iallocator. 1231 1232 @type iallocator_slot: string 1233 @param iallocator_slot: the name of the opcode iallocator slot 1234 @type node_slot: string 1235 @param node_slot: the name of the opcode target node slot 1236 1237 """ 1238 node = getattr(lu.op, node_slot, None) 1239 iallocator = getattr(lu.op, iallocator_slot, None) 1240 1241 if node is not None and iallocator is not None: 1242 raise errors.OpPrereqError("Do not specify both, iallocator and node", 1243 errors.ECODE_INVAL) 1244 elif node is None and iallocator is None: 1245 default_iallocator = lu.cfg.GetDefaultIAllocator() 1246 if default_iallocator: 1247 setattr(lu.op, iallocator_slot, default_iallocator) 1248 else: 1249 raise errors.OpPrereqError("No iallocator or node given and no" 1250 " cluster-wide default iallocator found;" 1251 " please specify either an iallocator or a" 1252 " node, or set a cluster-wide default" 1253 " iallocator")
1254
1255 1256 -def _GetDefaultIAllocator(cfg, iallocator):
1257 """Decides on which iallocator to use. 1258 1259 @type cfg: L{config.ConfigWriter} 1260 @param cfg: Cluster configuration object 1261 @type iallocator: string or None 1262 @param iallocator: Iallocator specified in opcode 1263 @rtype: string 1264 @return: Iallocator name 1265 1266 """ 1267 if not iallocator: 1268 # Use default iallocator 1269 iallocator = cfg.GetDefaultIAllocator() 1270 1271 if not iallocator: 1272 raise errors.OpPrereqError("No iallocator was specified, neither in the" 1273 " opcode nor as a cluster-wide default", 1274 errors.ECODE_INVAL) 1275 1276 return iallocator
1277
1278 1279 -class LUClusterPostInit(LogicalUnit):
1280 """Logical unit for running hooks after cluster initialization. 1281 1282 """ 1283 HPATH = "cluster-init" 1284 HTYPE = constants.HTYPE_CLUSTER 1285
1286 - def BuildHooksEnv(self):
1287 """Build hooks env. 1288 1289 """ 1290 return { 1291 "OP_TARGET": self.cfg.GetClusterName(), 1292 }
1293
1294 - def BuildHooksNodes(self):
1295 """Build hooks nodes. 1296 1297 """ 1298 return ([], [self.cfg.GetMasterNode()])
1299
1300 - def Exec(self, feedback_fn):
1301 """Nothing to do. 1302 1303 """ 1304 return True
1305
1306 1307 -class LUClusterDestroy(LogicalUnit):
1308 """Logical unit for destroying the cluster. 1309 1310 """ 1311 HPATH = "cluster-destroy" 1312 HTYPE = constants.HTYPE_CLUSTER 1313
1314 - def BuildHooksEnv(self):
1315 """Build hooks env. 1316 1317 """ 1318 return { 1319 "OP_TARGET": self.cfg.GetClusterName(), 1320 }
1321
1322 - def BuildHooksNodes(self):
1323 """Build hooks nodes. 1324 1325 """ 1326 return ([], [])
1327
1328 - def CheckPrereq(self):
1329 """Check prerequisites. 1330 1331 This checks whether the cluster is empty. 1332 1333 Any errors are signaled by raising errors.OpPrereqError. 1334 1335 """ 1336 master = self.cfg.GetMasterNode() 1337 1338 nodelist = self.cfg.GetNodeList() 1339 if len(nodelist) != 1 or nodelist[0] != master: 1340 raise errors.OpPrereqError("There are still %d node(s) in" 1341 " this cluster." % (len(nodelist) - 1), 1342 errors.ECODE_INVAL) 1343 instancelist = self.cfg.GetInstanceList() 1344 if instancelist: 1345 raise errors.OpPrereqError("There are still %d instance(s) in" 1346 " this cluster." % len(instancelist), 1347 errors.ECODE_INVAL)
1348
1349 - def Exec(self, feedback_fn):
1350 """Destroys the cluster. 1351 1352 """ 1353 master = self.cfg.GetMasterNode() 1354 1355 # Run post hooks on master node before it's removed 1356 _RunPostHook(self, master) 1357 1358 result = self.rpc.call_node_stop_master(master, False) 1359 result.Raise("Could not disable the master role") 1360 1361 return master
1362
1363 1364 -def _VerifyCertificate(filename):
1365 """Verifies a certificate for L{LUClusterVerifyConfig}. 1366 1367 @type filename: string 1368 @param filename: Path to PEM file 1369 1370 """ 1371 try: 1372 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM, 1373 utils.ReadFile(filename)) 1374 except Exception, err: # pylint: disable=W0703 1375 return (LUClusterVerifyConfig.ETYPE_ERROR, 1376 "Failed to load X509 certificate %s: %s" % (filename, err)) 1377 1378 (errcode, msg) = \ 1379 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN, 1380 constants.SSL_CERT_EXPIRATION_ERROR) 1381 1382 if msg: 1383 fnamemsg = "While verifying %s: %s" % (filename, msg) 1384 else: 1385 fnamemsg = None 1386 1387 if errcode is None: 1388 return (None, fnamemsg) 1389 elif errcode == utils.CERT_WARNING: 1390 return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg) 1391 elif errcode == utils.CERT_ERROR: 1392 return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg) 1393 1394 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1395
1396 1397 -def _GetAllHypervisorParameters(cluster, instances):
1398 """Compute the set of all hypervisor parameters. 1399 1400 @type cluster: L{objects.Cluster} 1401 @param cluster: the cluster object 1402 @param instances: list of L{objects.Instance} 1403 @param instances: additional instances from which to obtain parameters 1404 @rtype: list of (origin, hypervisor, parameters) 1405 @return: a list with all parameters found, indicating the hypervisor they 1406 apply to, and the origin (can be "cluster", "os X", or "instance Y") 1407 1408 """ 1409 hvp_data = [] 1410 1411 for hv_name in cluster.enabled_hypervisors: 1412 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name))) 1413 1414 for os_name, os_hvp in cluster.os_hvp.items(): 1415 for hv_name, hv_params in os_hvp.items(): 1416 if hv_params: 1417 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name) 1418 hvp_data.append(("os %s" % os_name, hv_name, full_params)) 1419 1420 # TODO: collapse identical parameter values in a single one 1421 for instance in instances: 1422 if instance.hvparams: 1423 hvp_data.append(("instance %s" % instance.name, instance.hypervisor, 1424 cluster.FillHV(instance))) 1425 1426 return hvp_data
1427
1428 1429 -class _VerifyErrors(object):
1430 """Mix-in for cluster/group verify LUs. 1431 1432 It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects 1433 self.op and self._feedback_fn to be available.) 1434 1435 """ 1436 TCLUSTER = "cluster" 1437 TNODE = "node" 1438 TINSTANCE = "instance" 1439 1440 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG") 1441 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT") 1442 ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK") 1443 ECLUSTERDANGLINGNODES = (TNODE, "ECLUSTERDANGLINGNODES") 1444 ECLUSTERDANGLINGINST = (TNODE, "ECLUSTERDANGLINGINST") 1445 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE") 1446 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN") 1447 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT") 1448 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK") 1449 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK") 1450 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE") 1451 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS") 1452 ENODEDRBD = (TNODE, "ENODEDRBD") 1453 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER") 1454 ENODEFILECHECK = (TNODE, "ENODEFILECHECK") 1455 ENODEHOOKS = (TNODE, "ENODEHOOKS") 1456 ENODEHV = (TNODE, "ENODEHV") 1457 ENODELVM = (TNODE, "ENODELVM") 1458 ENODEN1 = (TNODE, "ENODEN1") 1459 ENODENET = (TNODE, "ENODENET") 1460 ENODEOS = (TNODE, "ENODEOS") 1461 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE") 1462 ENODEORPHANLV = (TNODE, "ENODEORPHANLV") 1463 ENODERPC = (TNODE, "ENODERPC") 1464 ENODESSH = (TNODE, "ENODESSH") 1465 ENODEVERSION = (TNODE, "ENODEVERSION") 1466 ENODESETUP = (TNODE, "ENODESETUP") 1467 ENODETIME = (TNODE, "ENODETIME") 1468 ENODEOOBPATH = (TNODE, "ENODEOOBPATH") 1469 1470 ETYPE_FIELD = "code" 1471 ETYPE_ERROR = "ERROR" 1472 ETYPE_WARNING = "WARNING" 1473
1474 - def _Error(self, ecode, item, msg, *args, **kwargs):
1475 """Format an error message. 1476 1477 Based on the opcode's error_codes parameter, either format a 1478 parseable error code, or a simpler error string. 1479 1480 This must be called only from Exec and functions called from Exec. 1481 1482 """ 1483 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) 1484 itype, etxt = ecode 1485 # first complete the msg 1486 if args: 1487 msg = msg % args 1488 # then format the whole message 1489 if self.op.error_codes: # This is a mix-in. pylint: disable=E1101 1490 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg) 1491 else: 1492 if item: 1493 item = " " + item 1494 else: 1495 item = "" 1496 msg = "%s: %s%s: %s" % (ltype, itype, item, msg) 1497 # and finally report it via the feedback_fn 1498 self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable=E1101
1499
1500 - def _ErrorIf(self, cond, *args, **kwargs):
1501 """Log an error message if the passed condition is True. 1502 1503 """ 1504 cond = (bool(cond) 1505 or self.op.debug_simulate_errors) # pylint: disable=E1101 1506 if cond: 1507 self._Error(*args, **kwargs) 1508 # do not mark the operation as failed for WARN cases only 1509 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR: 1510 self.bad = self.bad or cond
1511
1512 1513 -class LUClusterVerify(NoHooksLU):
1514 """Submits all jobs necessary to verify the cluster. 1515 1516 """ 1517 REQ_BGL = False 1518
1519 - def ExpandNames(self):
1520 self.needed_locks = {}
1521
1522 - def Exec(self, feedback_fn):
1523 jobs = [] 1524 1525 if self.op.group_name: 1526 groups = [self.op.group_name] 1527 depends_fn = lambda: None 1528 else: 1529 groups = self.cfg.GetNodeGroupList() 1530 1531 # Verify global configuration 1532 jobs.append([opcodes.OpClusterVerifyConfig()]) 1533 1534 # Always depend on global verification 1535 depends_fn = lambda: [(-len(jobs), [])] 1536 1537 jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group, 1538 depends=depends_fn())] 1539 for group in groups) 1540 1541 # Fix up all parameters 1542 for op in itertools.chain(*jobs): # pylint: disable=W0142 1543 op.debug_simulate_errors = self.op.debug_simulate_errors 1544 op.verbose = self.op.verbose 1545 op.error_codes = self.op.error_codes 1546 try: 1547 op.skip_checks = self.op.skip_checks 1548 except AttributeError: 1549 assert not isinstance(op, opcodes.OpClusterVerifyGroup) 1550 1551 return ResultWithJobs(jobs)
1552
1553 1554 -class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
1555 """Verifies the cluster config. 1556 1557 """ 1558 REQ_BGL = False 1559
1560 - def _VerifyHVP(self, hvp_data):
1561 """Verifies locally the syntax of the hypervisor parameters. 1562 1563 """ 1564 for item, hv_name, hv_params in hvp_data: 1565 msg = ("hypervisor %s parameters syntax check (source %s): %%s" % 1566 (item, hv_name)) 1567 try: 1568 hv_class = hypervisor.GetHypervisor(hv_name) 1569 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) 1570 hv_class.CheckParameterSyntax(hv_params) 1571 except errors.GenericError, err: 1572 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
1573
1574 - def ExpandNames(self):
1575 self.needed_locks = dict.fromkeys(locking.LEVELS, locking.ALL_SET) 1576 self.share_locks = _ShareAll()
1577
1578 - def CheckPrereq(self):
1579 """Check prerequisites. 1580 1581 """ 1582 # Retrieve all information 1583 self.all_group_info = self.cfg.GetAllNodeGroupsInfo() 1584 self.all_node_info = self.cfg.GetAllNodesInfo() 1585 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1586
1587 - def Exec(self, feedback_fn):
1588 """Verify integrity of cluster, performing various test on nodes. 1589 1590 """ 1591 self.bad = False 1592 self._feedback_fn = feedback_fn 1593 1594 feedback_fn("* Verifying cluster config") 1595 1596 for msg in self.cfg.VerifyConfig(): 1597 self._ErrorIf(True, self.ECLUSTERCFG, None, msg) 1598 1599 feedback_fn("* Verifying cluster certificate files") 1600 1601 for cert_filename in constants.ALL_CERT_FILES: 1602 (errcode, msg) = _VerifyCertificate(cert_filename) 1603 self._ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode) 1604 1605 feedback_fn("* Verifying hypervisor parameters") 1606 1607 self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(), 1608 self.all_inst_info.values())) 1609 1610 feedback_fn("* Verifying all nodes belong to an existing group") 1611 1612 # We do this verification here because, should this bogus circumstance 1613 # occur, it would never be caught by VerifyGroup, which only acts on 1614 # nodes/instances reachable from existing node groups. 1615 1616 dangling_nodes = set(node.name for node in self.all_node_info.values() 1617 if node.group not in self.all_group_info) 1618 1619 dangling_instances = {} 1620 no_node_instances = [] 1621 1622 for inst in self.all_inst_info.values(): 1623 if inst.primary_node in dangling_nodes: 1624 dangling_instances.setdefault(inst.primary_node, []).append(inst.name) 1625 elif inst.primary_node not in self.all_node_info: 1626 no_node_instances.append(inst.name) 1627 1628 pretty_dangling = [ 1629 "%s (%s)" % 1630 (node.name, 1631 utils.CommaJoin(dangling_instances.get(node.name, 1632 ["no instances"]))) 1633 for node in dangling_nodes] 1634 1635 self._ErrorIf(bool(dangling_nodes), self.ECLUSTERDANGLINGNODES, None, 1636 "the following nodes (and their instances) belong to a non" 1637 " existing group: %s", utils.CommaJoin(pretty_dangling)) 1638 1639 self._ErrorIf(bool(no_node_instances), self.ECLUSTERDANGLINGINST, None, 1640 "the following instances have a non-existing primary-node:" 1641 " %s", utils.CommaJoin(no_node_instances)) 1642 1643 return not self.bad
1644
1645 1646 -class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
1647 """Verifies the status of a node group. 1648 1649 """ 1650 HPATH = "cluster-verify" 1651 HTYPE = constants.HTYPE_CLUSTER 1652 REQ_BGL = False 1653 1654 _HOOKS_INDENT_RE = re.compile("^", re.M) 1655
1656 - class NodeImage(object):
1657 """A class representing the logical and physical status of a node. 1658 1659 @type name: string 1660 @ivar name: the node name to which this object refers 1661 @ivar volumes: a structure as returned from 1662 L{ganeti.backend.GetVolumeList} (runtime) 1663 @ivar instances: a list of running instances (runtime) 1664 @ivar pinst: list of configured primary instances (config) 1665 @ivar sinst: list of configured secondary instances (config) 1666 @ivar sbp: dictionary of {primary-node: list of instances} for all 1667 instances for which this node is secondary (config) 1668 @ivar mfree: free memory, as reported by hypervisor (runtime) 1669 @ivar dfree: free disk, as reported by the node (runtime) 1670 @ivar offline: the offline status (config) 1671 @type rpc_fail: boolean 1672 @ivar rpc_fail: whether the RPC verify call was successfull (overall, 1673 not whether the individual keys were correct) (runtime) 1674 @type lvm_fail: boolean 1675 @ivar lvm_fail: whether the RPC call didn't return valid LVM data 1676 @type hyp_fail: boolean 1677 @ivar hyp_fail: whether the RPC call didn't return the instance list 1678 @type ghost: boolean 1679 @ivar ghost: whether this is a known node or not (config) 1680 @type os_fail: boolean 1681 @ivar os_fail: whether the RPC call didn't return valid OS data 1682 @type oslist: list 1683 @ivar oslist: list of OSes as diagnosed by DiagnoseOS 1684 @type vm_capable: boolean 1685 @ivar vm_capable: whether the node can host instances 1686 1687 """
1688 - def __init__(self, offline=False, name=None, vm_capable=True):
1689 self.name = name 1690 self.volumes = {} 1691 self.instances = [] 1692 self.pinst = [] 1693 self.sinst = [] 1694 self.sbp = {} 1695 self.mfree = 0 1696 self.dfree = 0 1697 self.offline = offline 1698 self.vm_capable = vm_capable 1699 self.rpc_fail = False 1700 self.lvm_fail = False 1701 self.hyp_fail = False 1702 self.ghost = False 1703 self.os_fail = False 1704 self.oslist = {}
1705
1706 - def ExpandNames(self):
1707 # This raises errors.OpPrereqError on its own: 1708 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 1709 1710 # Get instances in node group; this is unsafe and needs verification later 1711 inst_names = \ 1712 self.cfg.GetNodeGroupInstances(self.group_uuid, primary_only=True) 1713 1714 self.needed_locks = { 1715 locking.LEVEL_INSTANCE: inst_names, 1716 locking.LEVEL_NODEGROUP: [self.group_uuid], 1717 locking.LEVEL_NODE: [], 1718 } 1719 1720 self.share_locks = _ShareAll()
1721
1722 - def DeclareLocks(self, level):
1723 if level == locking.LEVEL_NODE: 1724 # Get members of node group; this is unsafe and needs verification later 1725 nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members) 1726 1727 all_inst_info = self.cfg.GetAllInstancesInfo() 1728 1729 # In Exec(), we warn about mirrored instances that have primary and 1730 # secondary living in separate node groups. To fully verify that 1731 # volumes for these instances are healthy, we will need to do an 1732 # extra call to their secondaries. We ensure here those nodes will 1733 # be locked. 1734 for inst in self.owned_locks(locking.LEVEL_INSTANCE): 1735 # Important: access only the instances whose lock is owned 1736 if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR: 1737 nodes.update(all_inst_info[inst].secondary_nodes) 1738 1739 self.needed_locks[locking.LEVEL_NODE] = nodes
1740
1741 - def CheckPrereq(self):
1742 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP) 1743 self.group_info = self.cfg.GetNodeGroup(self.group_uuid) 1744 1745 group_nodes = set(self.group_info.members) 1746 group_instances = \ 1747 self.cfg.GetNodeGroupInstances(self.group_uuid, primary_only=True) 1748 1749 unlocked_nodes = \ 1750 group_nodes.difference(self.owned_locks(locking.LEVEL_NODE)) 1751 1752 unlocked_instances = \ 1753 group_instances.difference(self.owned_locks(locking.LEVEL_INSTANCE)) 1754 1755 if unlocked_nodes: 1756 raise errors.OpPrereqError("Missing lock for nodes: %s" % 1757 utils.CommaJoin(unlocked_nodes), 1758 errors.ECODE_STATE) 1759 1760 if unlocked_instances: 1761 raise errors.OpPrereqError("Missing lock for instances: %s" % 1762 utils.CommaJoin(unlocked_instances), 1763 errors.ECODE_STATE) 1764 1765 self.all_node_info = self.cfg.GetAllNodesInfo() 1766 self.all_inst_info = self.cfg.GetAllInstancesInfo() 1767 1768 self.my_node_names = utils.NiceSort(group_nodes) 1769 self.my_inst_names = utils.NiceSort(group_instances) 1770 1771 self.my_node_info = dict((name, self.all_node_info[name]) 1772 for name in self.my_node_names) 1773 1774 self.my_inst_info = dict((name, self.all_inst_info[name]) 1775 for name in self.my_inst_names) 1776 1777 # We detect here the nodes that will need the extra RPC calls for verifying 1778 # split LV volumes; they should be locked. 1779 extra_lv_nodes = set() 1780 1781 for inst in self.my_inst_info.values(): 1782 if inst.disk_template in constants.DTS_INT_MIRROR: 1783 for nname in inst.all_nodes: 1784 if self.all_node_info[nname].group != self.group_uuid: 1785 extra_lv_nodes.add(nname) 1786 1787 unlocked_lv_nodes = \ 1788 extra_lv_nodes.difference(self.owned_locks(locking.LEVEL_NODE)) 1789 1790 if unlocked_lv_nodes: 1791 raise errors.OpPrereqError("Missing node locks for LV check: %s" % 1792 utils.CommaJoin(unlocked_lv_nodes), 1793 errors.ECODE_STATE) 1794 self.extra_lv_nodes = list(extra_lv_nodes)
1795
1796 - def _VerifyNode(self, ninfo, nresult):
1797 """Perform some basic validation on data returned from a node. 1798 1799 - check the result data structure is well formed and has all the 1800 mandatory fields 1801 - check ganeti version 1802 1803 @type ninfo: L{objects.Node} 1804 @param ninfo: the node to check 1805 @param nresult: the results from the node 1806 @rtype: boolean 1807 @return: whether overall this call was successful (and we can expect 1808 reasonable values in the respose) 1809 1810 """ 1811 node = ninfo.name 1812 _ErrorIf = self._ErrorIf # pylint: disable=C0103 1813 1814 # main result, nresult should be a non-empty dict 1815 test = not nresult or not isinstance(nresult, dict) 1816 _ErrorIf(test, self.ENODERPC, node, 1817 "unable to verify node: no data returned") 1818 if test: 1819 return False 1820 1821 # compares ganeti version 1822 local_version = constants.PROTOCOL_VERSION 1823 remote_version = nresult.get("version", None) 1824 test = not (remote_version and 1825 isinstance(remote_version, (list, tuple)) and 1826 len(remote_version) == 2) 1827 _ErrorIf(test, self.ENODERPC, node, 1828 "connection to node returned invalid data") 1829 if test: 1830 return False 1831 1832 test = local_version != remote_version[0] 1833 _ErrorIf(test, self.ENODEVERSION, node, 1834 "incompatible protocol versions: master %s," 1835 " node %s", local_version, remote_version[0]) 1836 if test: 1837 return False 1838 1839 # node seems compatible, we can actually try to look into its results 1840 1841 # full package version 1842 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1], 1843 self.ENODEVERSION, node, 1844 "software version mismatch: master %s, node %s", 1845 constants.RELEASE_VERSION, remote_version[1], 1846 code=self.ETYPE_WARNING) 1847 1848 hyp_result = nresult.get(constants.NV_HYPERVISOR, None) 1849 if ninfo.vm_capable and isinstance(hyp_result, dict): 1850 for hv_name, hv_result in hyp_result.iteritems(): 1851 test = hv_result is not None 1852 _ErrorIf(test, self.ENODEHV, node, 1853 "hypervisor %s verify failure: '%s'", hv_name, hv_result) 1854 1855 hvp_result = nresult.get(constants.NV_HVPARAMS, None) 1856 if ninfo.vm_capable and isinstance(hvp_result, list): 1857 for item, hv_name, hv_result in hvp_result: 1858 _ErrorIf(True, self.ENODEHV, node, 1859 "hypervisor %s parameter verify failure (source %s): %s", 1860 hv_name, item, hv_result) 1861 1862 test = nresult.get(constants.NV_NODESETUP, 1863 ["Missing NODESETUP results"]) 1864 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s", 1865 "; ".join(test)) 1866 1867 return True
1868
1869 - def _VerifyNodeTime(self, ninfo, nresult, 1870 nvinfo_starttime, nvinfo_endtime):
1871 """Check the node time. 1872 1873 @type ninfo: L{objects.Node} 1874 @param ninfo: the node to check 1875 @param nresult: the remote results for the node 1876 @param nvinfo_starttime: the start time of the RPC call 1877 @param nvinfo_endtime: the end time of the RPC call 1878 1879 """ 1880 node = ninfo.name 1881 _ErrorIf = self._ErrorIf # pylint: disable=C0103 1882 1883 ntime = nresult.get(constants.NV_TIME, None) 1884 try: 1885 ntime_merged = utils.MergeTime(ntime) 1886 except (ValueError, TypeError): 1887 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time") 1888 return 1889 1890 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW): 1891 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged) 1892 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW): 1893 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime) 1894 else: 1895 ntime_diff = None 1896 1897 _ErrorIf(ntime_diff is not None, self.ENODETIME, node, 1898 "Node time diverges by at least %s from master node time", 1899 ntime_diff)
1900
1901 - def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1902 """Check the node LVM results. 1903 1904 @type ninfo: L{objects.Node} 1905 @param ninfo: the node to check 1906 @param nresult: the remote results for the node 1907 @param vg_name: the configured VG name 1908 1909 """ 1910 if vg_name is None: 1911 return 1912 1913 node = ninfo.name 1914 _ErrorIf = self._ErrorIf # pylint: disable=C0103 1915 1916 # checks vg existence and size > 20G 1917 vglist = nresult.get(constants.NV_VGLIST, None) 1918 test = not vglist 1919 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups") 1920 if not test: 1921 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name, 1922 constants.MIN_VG_SIZE) 1923 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus) 1924 1925 # check pv names 1926 pvlist = nresult.get(constants.NV_PVLIST, None) 1927 test = pvlist is None 1928 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node") 1929 if not test: 1930 # check that ':' is not present in PV names, since it's a 1931 # special character for lvcreate (denotes the range of PEs to 1932 # use on the PV) 1933 for _, pvname, owner_vg in pvlist: 1934 test = ":" in pvname 1935 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV" 1936 " '%s' of VG '%s'", pvname, owner_vg)
1937
1938 - def _VerifyNodeBridges(self, ninfo, nresult, bridges):
1939 """Check the node bridges. 1940 1941 @type ninfo: L{objects.Node} 1942 @param ninfo: the node to check 1943 @param nresult: the remote results for the node 1944 @param bridges: the expected list of bridges 1945 1946 """ 1947 if not bridges: 1948 return 1949 1950 node = ninfo.name 1951 _ErrorIf = self._ErrorIf # pylint: disable=C0103 1952 1953 missing = nresult.get(constants.NV_BRIDGES, None) 1954 test = not isinstance(missing, list) 1955 _ErrorIf(test, self.ENODENET, node, 1956 "did not return valid bridge information") 1957 if not test: 1958 _ErrorIf(bool(missing), self.ENODENET, node, "missing bridges: %s" % 1959 utils.CommaJoin(sorted(missing)))
1960
1961 - def _VerifyNodeNetwork(self, ninfo, nresult):
1962 """Check the node network connectivity results. 1963 1964 @type ninfo: L{objects.Node} 1965 @param ninfo: the node to check 1966 @param nresult: the remote results for the node 1967 1968 """ 1969 node = ninfo.name 1970 _ErrorIf = self._ErrorIf # pylint: disable=C0103 1971 1972 test = constants.NV_NODELIST not in nresult 1973 _ErrorIf(test, self.ENODESSH, node, 1974 "node hasn't returned node ssh connectivity data") 1975 if not test: 1976 if nresult[constants.NV_NODELIST]: 1977 for a_node, a_msg in nresult[constants.NV_NODELIST].items(): 1978 _ErrorIf(True, self.ENODESSH, node, 1979 "ssh communication with node '%s': %s", a_node, a_msg) 1980 1981 test = constants.NV_NODENETTEST not in nresult 1982 _ErrorIf(test, self.ENODENET, node, 1983 "node hasn't returned node tcp connectivity data") 1984 if not test: 1985 if nresult[constants.NV_NODENETTEST]: 1986 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys()) 1987 for anode in nlist: 1988 _ErrorIf(True, self.ENODENET, node, 1989 "tcp communication with node '%s': %s", 1990 anode, nresult[constants.NV_NODENETTEST][anode]) 1991 1992 test = constants.NV_MASTERIP not in nresult 1993 _ErrorIf(test, self.ENODENET, node, 1994 "node hasn't returned node master IP reachability data") 1995 if not test: 1996 if not nresult[constants.NV_MASTERIP]: 1997 if node == self.master_node: 1998 msg = "the master node cannot reach the master IP (not configured?)" 1999 else: 2000 msg = "cannot reach the master IP" 2001 _ErrorIf(True, self.ENODENET, node, msg)
2002
2003 - def _VerifyInstance(self, instance, instanceconfig, node_image, 2004 diskstatus):
2005 """Verify an instance. 2006 2007 This function checks to see if the required block devices are 2008 available on the instance's node. 2009 2010 """ 2011 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2012 node_current = instanceconfig.primary_node 2013 2014 node_vol_should = {} 2015 instanceconfig.MapLVsByNode(node_vol_should) 2016 2017 for node in node_vol_should: 2018 n_img = node_image[node] 2019 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail: 2020 # ignore missing volumes on offline or broken nodes 2021 continue 2022 for volume in node_vol_should[node]: 2023 test = volume not in n_img.volumes 2024 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance, 2025 "volume %s missing on node %s", volume, node) 2026 2027 if instanceconfig.admin_up: 2028 pri_img = node_image[node_current] 2029 test = instance not in pri_img.instances and not pri_img.offline 2030 _ErrorIf(test, self.EINSTANCEDOWN, instance, 2031 "instance not running on its primary node %s", 2032 node_current) 2033 2034 diskdata = [(nname, success, status, idx) 2035 for (nname, disks) in diskstatus.items() 2036 for idx, (success, status) in enumerate(disks)] 2037 2038 for nname, success, bdev_status, idx in diskdata: 2039 # the 'ghost node' construction in Exec() ensures that we have a 2040 # node here 2041 snode = node_image[nname] 2042 bad_snode = snode.ghost or snode.offline 2043 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode, 2044 self.EINSTANCEFAULTYDISK, instance, 2045 "couldn't retrieve status for disk/%s on %s: %s", 2046 idx, nname, bdev_status) 2047 _ErrorIf((instanceconfig.admin_up and success and 2048 bdev_status.ldisk_status == constants.LDS_FAULTY), 2049 self.EINSTANCEFAULTYDISK, instance, 2050 "disk/%s on %s is faulty", idx, nname)
2051
2052 - def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
2053 """Verify if there are any unknown volumes in the cluster. 2054 2055 The .os, .swap and backup volumes are ignored. All other volumes are 2056 reported as unknown. 2057 2058 @type reserved: L{ganeti.utils.FieldSet} 2059 @param reserved: a FieldSet of reserved volume names 2060 2061 """ 2062 for node, n_img in node_image.items(): 2063 if (n_img.offline or n_img.rpc_fail or n_img.lvm_fail or 2064 self.all_node_info[node].group != self.group_uuid): 2065 # skip non-healthy nodes 2066 continue 2067 for volume in n_img.volumes: 2068 test = ((node not in node_vol_should or 2069 volume not in node_vol_should[node]) and 2070 not reserved.Matches(volume)) 2071 self._ErrorIf(test, self.ENODEORPHANLV, node, 2072 "volume %s is unknown", volume)
2073
2074 - def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
2075 """Verify N+1 Memory Resilience. 2076 2077 Check that if one single node dies we can still start all the 2078 instances it was primary for. 2079 2080 """ 2081 cluster_info = self.cfg.GetClusterInfo() 2082 for node, n_img in node_image.items(): 2083 # This code checks that every node which is now listed as 2084 # secondary has enough memory to host all instances it is 2085 # supposed to should a single other node in the cluster fail. 2086 # FIXME: not ready for failover to an arbitrary node 2087 # FIXME: does not support file-backed instances 2088 # WARNING: we currently take into account down instances as well 2089 # as up ones, considering that even if they're down someone 2090 # might want to start them even in the event of a node failure. 2091 if n_img.offline or self.all_node_info[node].group != self.group_uuid: 2092 # we're skipping nodes marked offline and nodes in other groups from 2093 # the N+1 warning, since most likely we don't have good memory 2094 # infromation from them; we already list instances living on such 2095 # nodes, and that's enough warning 2096 continue 2097 for prinode, instances in n_img.sbp.items(): 2098 needed_mem = 0 2099 for instance in instances: 2100 bep = cluster_info.FillBE(instance_cfg[instance]) 2101 if bep[constants.BE_AUTO_BALANCE]: 2102 needed_mem += bep[constants.BE_MEMORY] 2103 test = n_img.mfree < needed_mem 2104 self._ErrorIf(test, self.ENODEN1, node, 2105 "not enough memory to accomodate instance failovers" 2106 " should node %s fail (%dMiB needed, %dMiB available)", 2107 prinode, needed_mem, n_img.mfree)
2108 2109 @classmethod
2110 - def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo, 2111 (files_all, files_all_opt, files_mc, files_vm)):
2112 """Verifies file checksums collected from all nodes. 2113 2114 @param errorif: Callback for reporting errors 2115 @param nodeinfo: List of L{objects.Node} objects 2116 @param master_node: Name of master node 2117 @param all_nvinfo: RPC results 2118 2119 """ 2120 assert (len(files_all | files_all_opt | files_mc | files_vm) == 2121 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \ 2122 "Found file listed in more than one file list" 2123 2124 # Define functions determining which nodes to consider for a file 2125 files2nodefn = [ 2126 (files_all, None), 2127 (files_all_opt, None), 2128 (files_mc, lambda node: (node.master_candidate or 2129 node.name == master_node)), 2130 (files_vm, lambda node: node.vm_capable), 2131 ] 2132 2133 # Build mapping from filename to list of nodes which should have the file 2134 nodefiles = {} 2135 for (files, fn) in files2nodefn: 2136 if fn is None: 2137 filenodes = nodeinfo 2138 else: 2139 filenodes = filter(fn, nodeinfo) 2140 nodefiles.update((filename, 2141 frozenset(map(operator.attrgetter("name"), filenodes))) 2142 for filename in files) 2143 2144 assert set(nodefiles) == (files_all | files_all_opt | files_mc | files_vm) 2145 2146 fileinfo = dict((filename, {}) for filename in nodefiles) 2147 ignore_nodes = set() 2148 2149 for node in nodeinfo: 2150 if node.offline: 2151 ignore_nodes.add(node.name) 2152 continue 2153 2154 nresult = all_nvinfo[node.name] 2155 2156 if nresult.fail_msg or not nresult.payload: 2157 node_files = None 2158 else: 2159 node_files = nresult.payload.get(constants.NV_FILELIST, None) 2160 2161 test = not (node_files and isinstance(node_files, dict)) 2162 errorif(test, cls.ENODEFILECHECK, node.name, 2163 "Node did not return file checksum data") 2164 if test: 2165 ignore_nodes.add(node.name) 2166 continue 2167 2168 # Build per-checksum mapping from filename to nodes having it 2169 for (filename, checksum) in node_files.items(): 2170 assert filename in nodefiles 2171 fileinfo[filename].setdefault(checksum, set()).add(node.name) 2172 2173 for (filename, checksums) in fileinfo.items(): 2174 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum" 2175 2176 # Nodes having the file 2177 with_file = frozenset(node_name 2178 for nodes in fileinfo[filename].values() 2179 for node_name in nodes) - ignore_nodes 2180 2181 expected_nodes = nodefiles[filename] - ignore_nodes 2182 2183 # Nodes missing file 2184 missing_file = expected_nodes - with_file 2185 2186 if filename in files_all_opt: 2187 # All or no nodes 2188 errorif(missing_file and missing_file != expected_nodes, 2189 cls.ECLUSTERFILECHECK, None, 2190 "File %s is optional, but it must exist on all or no" 2191 " nodes (not found on %s)", 2192 filename, utils.CommaJoin(utils.NiceSort(missing_file))) 2193 else: 2194 # Non-optional files 2195 errorif(missing_file, cls.ECLUSTERFILECHECK, None, 2196 "File %s is missing from node(s) %s", filename, 2197 utils.CommaJoin(utils.NiceSort(missing_file))) 2198 2199 # Warn if a node has a file it shouldn't 2200 unexpected = with_file - expected_nodes 2201 errorif(unexpected, 2202 cls.ECLUSTERFILECHECK, None, 2203 "File %s should not exist on node(s) %s", 2204 filename, utils.CommaJoin(utils.NiceSort(unexpected))) 2205 2206 # See if there are multiple versions of the file 2207 test = len(checksums) > 1 2208 if test: 2209 variants = ["variant %s on %s" % 2210 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes))) 2211 for (idx, (checksum, nodes)) in 2212 enumerate(sorted(checksums.items()))] 2213 else: 2214 variants = [] 2215 2216 errorif(test, cls.ECLUSTERFILECHECK, None, 2217 "File %s found with %s different checksums (%s)", 2218 filename, len(checksums), "; ".join(variants))
2219
2220 - def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper, 2221 drbd_map):
2222 """Verifies and the node DRBD status. 2223 2224 @type ninfo: L{objects.Node} 2225 @param ninfo: the node to check 2226 @param nresult: the remote results for the node 2227 @param instanceinfo: the dict of instances 2228 @param drbd_helper: the configured DRBD usermode helper 2229 @param drbd_map: the DRBD map as returned by 2230 L{ganeti.config.ConfigWriter.ComputeDRBDMap} 2231 2232 """ 2233 node = ninfo.name 2234 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2235 2236 if drbd_helper: 2237 helper_result = nresult.get(constants.NV_DRBDHELPER, None) 2238 test = (helper_result == None) 2239 _ErrorIf(test, self.ENODEDRBDHELPER, node, 2240 "no drbd usermode helper returned") 2241 if helper_result: 2242 status, payload = helper_result 2243 test = not status 2244 _ErrorIf(test, self.ENODEDRBDHELPER, node, 2245 "drbd usermode helper check unsuccessful: %s", payload) 2246 test = status and (payload != drbd_helper) 2247 _ErrorIf(test, self.ENODEDRBDHELPER, node, 2248 "wrong drbd usermode helper: %s", payload) 2249 2250 # compute the DRBD minors 2251 node_drbd = {} 2252 for minor, instance in drbd_map[node].items(): 2253 test = instance not in instanceinfo 2254 _ErrorIf(test, self.ECLUSTERCFG, None, 2255 "ghost instance '%s' in temporary DRBD map", instance) 2256 # ghost instance should not be running, but otherwise we 2257 # don't give double warnings (both ghost instance and 2258 # unallocated minor in use) 2259 if test: 2260 node_drbd[minor] = (instance, False) 2261 else: 2262 instance = instanceinfo[instance] 2263 node_drbd[minor] = (instance.name, instance.admin_up) 2264 2265 # and now check them 2266 used_minors = nresult.get(constants.NV_DRBDLIST, []) 2267 test = not isinstance(used_minors, (tuple, list)) 2268 _ErrorIf(test, self.ENODEDRBD, node, 2269 "cannot parse drbd status file: %s", str(used_minors)) 2270 if test: 2271 # we cannot check drbd status 2272 return 2273 2274 for minor, (iname, must_exist) in node_drbd.items(): 2275 test = minor not in used_minors and must_exist 2276 _ErrorIf(test, self.ENODEDRBD, node, 2277 "drbd minor %d of instance %s is not active", minor, iname) 2278 for minor in used_minors: 2279 test = minor not in node_drbd 2280 _ErrorIf(test, self.ENODEDRBD, node, 2281 "unallocated drbd minor %d is in use", minor)
2282
2283 - def _UpdateNodeOS(self, ninfo, nresult, nimg):
2284 """Builds the node OS structures. 2285 2286 @type ninfo: L{objects.Node} 2287 @param ninfo: the node to check 2288 @param nresult: the remote results for the node 2289 @param nimg: the node image object 2290 2291 """ 2292 node = ninfo.name 2293 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2294 2295 remote_os = nresult.get(constants.NV_OSLIST, None) 2296 test = (not isinstance(remote_os, list) or 2297 not compat.all(isinstance(v, list) and len(v) == 7 2298 for v in remote_os)) 2299 2300 _ErrorIf(test, self.ENODEOS, node, 2301 "node hasn't returned valid OS data") 2302 2303 nimg.os_fail = test 2304 2305 if test: 2306 return 2307 2308 os_dict = {} 2309 2310 for (name, os_path, status, diagnose, 2311 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]: 2312 2313 if name not in os_dict: 2314 os_dict[name] = [] 2315 2316 # parameters is a list of lists instead of list of tuples due to 2317 # JSON lacking a real tuple type, fix it: 2318 parameters = [tuple(v) for v in parameters] 2319 os_dict[name].append((os_path, status, diagnose, 2320 set(variants), set(parameters), set(api_ver))) 2321 2322 nimg.oslist = os_dict
2323
2324 - def _VerifyNodeOS(self, ninfo, nimg, base):
2325 """Verifies the node OS list. 2326 2327 @type ninfo: L{objects.Node} 2328 @param ninfo: the node to check 2329 @param nimg: the node image object 2330 @param base: the 'template' node we match against (e.g. from the master) 2331 2332 """ 2333 node = ninfo.name 2334 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2335 2336 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?" 2337 2338 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l] 2339 for os_name, os_data in nimg.oslist.items(): 2340 assert os_data, "Empty OS status for OS %s?!" % os_name 2341 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0] 2342 _ErrorIf(not f_status, self.ENODEOS, node, 2343 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag) 2344 _ErrorIf(len(os_data) > 1, self.ENODEOS, node, 2345 "OS '%s' has multiple entries (first one shadows the rest): %s", 2346 os_name, utils.CommaJoin([v[0] for v in os_data])) 2347 # comparisons with the 'base' image 2348 test = os_name not in base.oslist 2349 _ErrorIf(test, self.ENODEOS, node, 2350 "Extra OS %s not present on reference node (%s)", 2351 os_name, base.name) 2352 if test: 2353 continue 2354 assert base.oslist[os_name], "Base node has empty OS status?" 2355 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0] 2356 if not b_status: 2357 # base OS is invalid, skipping 2358 continue 2359 for kind, a, b in [("API version", f_api, b_api), 2360 ("variants list", f_var, b_var), 2361 ("parameters", beautify_params(f_param), 2362 beautify_params(b_param))]: 2363 _ErrorIf(a != b, self.ENODEOS, node, 2364 "OS %s for %s differs from reference node %s: [%s] vs. [%s]", 2365 kind, os_name, base.name, 2366 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b))) 2367 2368 # check any missing OSes 2369 missing = set(base.oslist.keys()).difference(nimg.oslist.keys()) 2370 _ErrorIf(missing, self.ENODEOS, node, 2371 "OSes present on reference node %s but missing on this node: %s", 2372 base.name, utils.CommaJoin(missing))
2373
2374 - def _VerifyOob(self, ninfo, nresult):
2375 """Verifies out of band functionality of a node. 2376 2377 @type ninfo: L{objects.Node} 2378 @param ninfo: the node to check 2379 @param nresult: the remote results for the node 2380 2381 """ 2382 node = ninfo.name 2383 # We just have to verify the paths on master and/or master candidates 2384 # as the oob helper is invoked on the master 2385 if ((ninfo.master_candidate or ninfo.master_capable) and 2386 constants.NV_OOB_PATHS in nresult): 2387 for path_result in nresult[constants.NV_OOB_PATHS]: 2388 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
2389
2390 - def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
2391 """Verifies and updates the node volume data. 2392 2393 This function will update a L{NodeImage}'s internal structures 2394 with data from the remote call. 2395 2396 @type ninfo: L{objects.Node} 2397 @param ninfo: the node to check 2398 @param nresult: the remote results for the node 2399 @param nimg: the node image object 2400 @param vg_name: the configured VG name 2401 2402 """ 2403 node = ninfo.name 2404 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2405 2406 nimg.lvm_fail = True 2407 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data") 2408 if vg_name is None: 2409 pass 2410 elif isinstance(lvdata, basestring): 2411 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s", 2412 utils.SafeEncode(lvdata)) 2413 elif not isinstance(lvdata, dict): 2414 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)") 2415 else: 2416 nimg.volumes = lvdata 2417 nimg.lvm_fail = False
2418
2419 - def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2420 """Verifies and updates the node instance list. 2421 2422 If the listing was successful, then updates this node's instance 2423 list. Otherwise, it marks the RPC call as failed for the instance 2424 list key. 2425 2426 @type ninfo: L{objects.Node} 2427 @param ninfo: the node to check 2428 @param nresult: the remote results for the node 2429 @param nimg: the node image object 2430 2431 """ 2432 idata = nresult.get(constants.NV_INSTANCELIST, None) 2433 test = not isinstance(idata, list) 2434 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed" 2435 " (instancelist): %s", utils.SafeEncode(str(idata))) 2436 if test: 2437 nimg.hyp_fail = True 2438 else: 2439 nimg.instances = idata
2440
2441 - def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2442 """Verifies and computes a node information map 2443 2444 @type ninfo: L{objects.Node} 2445 @param ninfo: the node to check 2446 @param nresult: the remote results for the node 2447 @param nimg: the node image object 2448 @param vg_name: the configured VG name 2449 2450 """ 2451 node = ninfo.name 2452 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2453 2454 # try to read free memory (from the hypervisor) 2455 hv_info = nresult.get(constants.NV_HVINFO, None) 2456 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info 2457 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)") 2458 if not test: 2459 try: 2460 nimg.mfree = int(hv_info["memory_free"]) 2461 except (ValueError, TypeError): 2462 _ErrorIf(True, self.ENODERPC, node, 2463 "node returned invalid nodeinfo, check hypervisor") 2464 2465 # FIXME: devise a free space model for file based instances as well 2466 if vg_name is not None: 2467 test = (constants.NV_VGLIST not in nresult or 2468 vg_name not in nresult[constants.NV_VGLIST]) 2469 _ErrorIf(test, self.ENODELVM, node, 2470 "node didn't return data for the volume group '%s'" 2471 " - it is either missing or broken", vg_name) 2472 if not test: 2473 try: 2474 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name]) 2475 except (ValueError, TypeError): 2476 _ErrorIf(True, self.ENODERPC, node, 2477 "node returned invalid LVM info, check LVM status")
2478
2479 - def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2480 """Gets per-disk status information for all instances. 2481 2482 @type nodelist: list of strings 2483 @param nodelist: Node names 2484 @type node_image: dict of (name, L{objects.Node}) 2485 @param node_image: Node objects 2486 @type instanceinfo: dict of (name, L{objects.Instance}) 2487 @param instanceinfo: Instance objects 2488 @rtype: {instance: {node: [(succes, payload)]}} 2489 @return: a dictionary of per-instance dictionaries with nodes as 2490 keys and disk information as values; the disk information is a 2491 list of tuples (success, payload) 2492 2493 """ 2494 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2495 2496 node_disks = {} 2497 node_disks_devonly = {} 2498 diskless_instances = set() 2499 diskless = constants.DT_DISKLESS 2500 2501 for nname in nodelist: 2502 node_instances = list(itertools.chain(node_image[nname].pinst, 2503 node_image[nname].sinst)) 2504 diskless_instances.update(inst for inst in node_instances 2505 if instanceinfo[inst].disk_template == diskless) 2506 disks = [(inst, disk) 2507 for inst in node_instances 2508 for disk in instanceinfo[inst].disks] 2509 2510 if not disks: 2511 # No need to collect data 2512 continue 2513 2514 node_disks[nname] = disks 2515 2516 # Creating copies as SetDiskID below will modify the objects and that can 2517 # lead to incorrect data returned from nodes 2518 devonly = [dev.Copy() for (_, dev) in disks] 2519 2520 for dev in devonly: 2521 self.cfg.SetDiskID(dev, nname) 2522 2523 node_disks_devonly[nname] = devonly 2524 2525 assert len(node_disks) == len(node_disks_devonly) 2526 2527 # Collect data from all nodes with disks 2528 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(), 2529 node_disks_devonly) 2530 2531 assert len(result) == len(node_disks) 2532 2533 instdisk = {} 2534 2535 for (nname, nres) in result.items(): 2536 disks = node_disks[nname] 2537 2538 if nres.offline: 2539 # No data from this node 2540 data = len(disks) * [(False, "node offline")] 2541 else: 2542 msg = nres.fail_msg 2543 _ErrorIf(msg, self.ENODERPC, nname, 2544 "while getting disk information: %s", msg) 2545 if msg: 2546 # No data from this node 2547 data = len(disks) * [(False, msg)] 2548 else: 2549 data = [] 2550 for idx, i in enumerate(nres.payload): 2551 if isinstance(i, (tuple, list)) and len(i) == 2: 2552 data.append(i) 2553 else: 2554 logging.warning("Invalid result from node %s, entry %d: %s", 2555 nname, idx, i) 2556 data.append((False, "Invalid result from the remote node")) 2557 2558 for ((inst, _), status) in zip(disks, data): 2559 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status) 2560 2561 # Add empty entries for diskless instances. 2562 for inst in diskless_instances: 2563 assert inst not in instdisk 2564 instdisk[inst] = {} 2565 2566 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and 2567 len(nnames) <= len(instanceinfo[inst].all_nodes) and 2568 compat.all(isinstance(s, (tuple, list)) and 2569 len(s) == 2 for s in statuses) 2570 for inst, nnames in instdisk.items() 2571 for nname, statuses in nnames.items()) 2572 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure" 2573 2574 return instdisk
2575 2576 @staticmethod
2577 - def _SshNodeSelector(group_uuid, all_nodes):
2578 """Create endless iterators for all potential SSH check hosts. 2579 2580 """ 2581 nodes = [node for node in all_nodes 2582 if (node.group != group_uuid and 2583 not node.offline)] 2584 keyfunc = operator.attrgetter("group") 2585 2586 return map(itertools.cycle, 2587 [sorted(map(operator.attrgetter("name"), names)) 2588 for _, names in itertools.groupby(sorted(nodes, key=keyfunc), 2589 keyfunc)])
2590 2591 @classmethod
2592 - def _SelectSshCheckNodes(cls, group_nodes, group_uuid, all_nodes):
2593 """Choose which nodes should talk to which other nodes. 2594 2595 We will make nodes contact all nodes in their group, and one node from 2596 every other group. 2597 2598 @warning: This algorithm has a known issue if one node group is much 2599 smaller than others (e.g. just one node). In such a case all other 2600 nodes will talk to the single node. 2601 2602 """ 2603 online_nodes = sorted(node.name for node in group_nodes if not node.offline) 2604 sel = cls._SshNodeSelector(group_uuid, all_nodes) 2605 2606 return (online_nodes, 2607 dict((name, sorted([i.next() for i in sel])) 2608 for name in online_nodes))
2609
2610 - def BuildHooksEnv(self):
2611 """Build hooks env. 2612 2613 Cluster-Verify hooks just ran in the post phase and their failure makes 2614 the output be logged in the verify output and the verification to fail. 2615 2616 """ 2617 env = { 2618 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags()) 2619 } 2620 2621 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags())) 2622 for node in self.my_node_info.values()) 2623 2624 return env
2625
2626 - def BuildHooksNodes(self):
2627 """Build hooks nodes. 2628 2629 """ 2630 return ([], self.my_node_names)
2631
2632 - def Exec(self, feedback_fn):
2633 """Verify integrity of the node group, performing various test on nodes. 2634 2635 """ 2636 # This method has too many local variables. pylint: disable=R0914 2637 feedback_fn("* Verifying group '%s'" % self.group_info.name) 2638 2639 if not self.my_node_names: 2640 # empty node group 2641 feedback_fn("* Empty node group, skipping verification") 2642 return True 2643 2644 self.bad = False 2645 _ErrorIf = self._ErrorIf # pylint: disable=C0103 2646 verbose = self.op.verbose 2647 self._feedback_fn = feedback_fn 2648 2649 vg_name = self.cfg.GetVGName() 2650 drbd_helper = self.cfg.GetDRBDHelper() 2651 cluster = self.cfg.GetClusterInfo() 2652 groupinfo = self.cfg.GetAllNodeGroupsInfo() 2653 hypervisors = cluster.enabled_hypervisors 2654 node_data_list = [self.my_node_info[name] for name in self.my_node_names] 2655 2656 i_non_redundant = [] # Non redundant instances 2657 i_non_a_balanced = [] # Non auto-balanced instances 2658 n_offline = 0 # Count of offline nodes 2659 n_drained = 0 # Count of nodes being drained 2660 node_vol_should = {} 2661 2662 # FIXME: verify OS list 2663 2664 # File verification 2665 filemap = _ComputeAncillaryFiles(cluster, False) 2666 2667 # do local checksums 2668 master_node = self.master_node = self.cfg.GetMasterNode() 2669 master_ip = self.cfg.GetMasterIP() 2670 2671 feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names)) 2672 2673 node_verify_param = { 2674 constants.NV_FILELIST: 2675 utils.UniqueSequence(filename 2676 for files in filemap 2677 for filename in files), 2678 constants.NV_NODELIST: 2679 self._SelectSshCheckNodes(node_data_list, self.group_uuid, 2680 self.all_node_info.values()), 2681 constants.NV_HYPERVISOR: hypervisors, 2682 constants.NV_HVPARAMS: 2683 _GetAllHypervisorParameters(cluster, self.all_inst_info.values()), 2684 constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip) 2685 for node in node_data_list 2686 if not node.offline], 2687 constants.NV_INSTANCELIST: hypervisors, 2688 constants.NV_VERSION: None, 2689 constants.NV_HVINFO: self.cfg.GetHypervisorType(), 2690 constants.NV_NODESETUP: None, 2691 constants.NV_TIME: None, 2692 constants.NV_MASTERIP: (master_node, master_ip), 2693 constants.NV_OSLIST: None, 2694 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(), 2695 } 2696 2697 if vg_name is not None: 2698 node_verify_param[constants.NV_VGLIST] = None 2699 node_verify_param[constants.NV_LVLIST] = vg_name 2700 node_verify_param[constants.NV_PVLIST] = [vg_name] 2701 node_verify_param[constants.NV_DRBDLIST] = None 2702 2703 if drbd_helper: 2704 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper 2705 2706 # bridge checks 2707 # FIXME: this needs to be changed per node-group, not cluster-wide 2708 bridges = set() 2709 default_nicpp = cluster.nicparams[constants.PP_DEFAULT] 2710 if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED: 2711 bridges.add(default_nicpp[constants.NIC_LINK]) 2712 for instance in self.my_inst_info.values(): 2713 for nic in instance.nics: 2714 full_nic = cluster.SimpleFillNIC(nic.nicparams) 2715 if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED: 2716 bridges.add(full_nic[constants.NIC_LINK]) 2717 2718 if bridges: 2719 node_verify_param[constants.NV_BRIDGES] = list(bridges) 2720 2721 # Build our expected cluster state 2722 node_image = dict((node.name, self.NodeImage(offline=node.offline, 2723 name=node.name, 2724 vm_capable=node.vm_capable)) 2725 for node in node_data_list) 2726 2727 # Gather OOB paths 2728 oob_paths = [] 2729 for node in self.all_node_info.values(): 2730 path = _SupportsOob(self.cfg, node) 2731 if path and path not in oob_paths: 2732 oob_paths.append(path) 2733 2734 if oob_paths: 2735 node_verify_param[constants.NV_OOB_PATHS] = oob_paths 2736 2737 for instance in self.my_inst_names: 2738 inst_config = self.my_inst_info[instance] 2739 2740 for nname in inst_config.all_nodes: 2741 if nname not in node_image: 2742 gnode = self.NodeImage(name=nname) 2743 gnode.ghost = (nname not in self.all_node_info) 2744 node_image[nname] = gnode 2745 2746 inst_config.MapLVsByNode(node_vol_should) 2747 2748 pnode = inst_config.primary_node 2749 node_image[pnode].pinst.append(instance) 2750 2751 for snode in inst_config.secondary_nodes: 2752 nimg = node_image[snode] 2753 nimg.sinst.append(instance) 2754 if pnode not in nimg.sbp: 2755 nimg.sbp[pnode] = [] 2756 nimg.sbp[pnode].append(instance) 2757 2758 # At this point, we have the in-memory data structures complete, 2759 # except for the runtime information, which we'll gather next 2760 2761 # Due to the way our RPC system works, exact response times cannot be 2762 # guaranteed (e.g. a broken node could run into a timeout). By keeping the 2763 # time before and after executing the request, we can at least have a time 2764 # window. 2765 nvinfo_starttime = time.time() 2766 all_nvinfo = self.rpc.call_node_verify(self.my_node_names, 2767 node_verify_param, 2768 self.cfg.GetClusterName()) 2769 nvinfo_endtime = time.time() 2770 2771 if self.extra_lv_nodes and vg_name is not None: 2772 extra_lv_nvinfo = \ 2773 self.rpc.call_node_verify(self.extra_lv_nodes, 2774 {constants.NV_LVLIST: vg_name}, 2775 self.cfg.GetClusterName()) 2776 else: 2777 extra_lv_nvinfo = {} 2778 2779 all_drbd_map = self.cfg.ComputeDRBDMap() 2780 2781 feedback_fn("* Gathering disk information (%s nodes)" % 2782 len(self.my_node_names)) 2783 instdisk = self._CollectDiskInfo(self.my_node_names, node_image, 2784 self.my_inst_info) 2785 2786 feedback_fn("* Verifying configuration file consistency") 2787 2788 # If not all nodes are being checked, we need to make sure the master node 2789 # and a non-checked vm_capable node are in the list. 2790 absent_nodes = set(self.all_node_info).difference(self.my_node_info) 2791 if absent_nodes: 2792 vf_nvinfo = all_nvinfo.copy() 2793 vf_node_info = list(self.my_node_info.values()) 2794 additional_nodes = [] 2795 if master_node not in self.my_node_info: 2796 additional_nodes.append(master_node) 2797 vf_node_info.append(self.all_node_info[master_node]) 2798 # Add the first vm_capable node we find which is not included 2799 for node in absent_nodes: 2800 nodeinfo = self.all_node_info[node] 2801 if nodeinfo.vm_capable and not nodeinfo.offline: 2802 additional_nodes.append(node) 2803 vf_node_info.append(self.all_node_info[node]) 2804 break 2805 key = constants.NV_FILELIST 2806 vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes, 2807 {key: node_verify_param[key]}, 2808 self.cfg.GetClusterName())) 2809 else: 2810 vf_nvinfo = all_nvinfo 2811 vf_node_info = self.my_node_info.values() 2812 2813 self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap) 2814 2815 feedback_fn("* Verifying node status") 2816 2817 refos_img = None 2818 2819 for node_i in node_data_list: 2820 node = node_i.name 2821 nimg = node_image[node] 2822 2823 if node_i.offline: 2824 if verbose: 2825 feedback_fn("* Skipping offline node %s" % (node,)) 2826 n_offline += 1 2827 continue 2828 2829 if node == master_node: 2830 ntype = "master" 2831 elif node_i.master_candidate: 2832 ntype = "master candidate" 2833 elif node_i.drained: 2834 ntype = "drained" 2835 n_drained += 1 2836 else: 2837 ntype = "regular" 2838 if verbose: 2839 feedback_fn("* Verifying node %s (%s)" % (node, ntype)) 2840 2841 msg = all_nvinfo[node].fail_msg 2842 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg) 2843 if msg: 2844 nimg.rpc_fail = True 2845 continue 2846 2847 nresult = all_nvinfo[node].payload 2848 2849 nimg.call_ok = self._VerifyNode(node_i, nresult) 2850 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime) 2851 self._VerifyNodeNetwork(node_i, nresult) 2852 self._VerifyOob(node_i, nresult) 2853 2854 if nimg.vm_capable: 2855 self._VerifyNodeLVM(node_i, nresult, vg_name) 2856 self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper, 2857 all_drbd_map) 2858 2859 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name) 2860 self._UpdateNodeInstances(node_i, nresult, nimg) 2861 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name) 2862 self._UpdateNodeOS(node_i, nresult, nimg) 2863 2864 if not nimg.os_fail: 2865 if refos_img is None: 2866 refos_img = nimg 2867 self._VerifyNodeOS(node_i, nimg, refos_img) 2868 self._VerifyNodeBridges(node_i, nresult, bridges) 2869 2870 # Check whether all running instancies are primary for the node. (This 2871 # can no longer be done from _VerifyInstance below, since some of the 2872 # wrong instances could be from other node groups.) 2873 non_primary_inst = set(nimg.instances).difference(nimg.pinst) 2874 2875 for inst in non_primary_inst: 2876 test = inst in self.all_inst_info 2877 _ErrorIf(test, self.EINSTANCEWRONGNODE, inst, 2878 "instance should not run on node %s", node_i.name) 2879 _ErrorIf(not test, self.ENODEORPHANINSTANCE, node_i.name, 2880 "node is running unknown instance %s", inst) 2881 2882 for node, result in extra_lv_nvinfo.items(): 2883 self._UpdateNodeVolumes(self.all_node_info[node], result.payload, 2884 node_image[node], vg_name) 2885 2886 feedback_fn("* Verifying instance status") 2887 for instance in self.my_inst_names: 2888 if verbose: 2889 feedback_fn("* Verifying instance %s" % instance) 2890 inst_config = self.my_inst_info[instance] 2891 self._VerifyInstance(instance, inst_config, node_image, 2892 instdisk[instance]) 2893 inst_nodes_offline = [] 2894 2895 pnode = inst_config.primary_node 2896 pnode_img = node_image[pnode] 2897 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline, 2898 self.ENODERPC, pnode, "instance %s, connection to" 2899 " primary node failed", instance) 2900 2901 _ErrorIf(inst_config.admin_up and pnode_img.offline, 2902 self.EINSTANCEBADNODE, instance, 2903 "instance is marked as running and lives on offline node %s", 2904 inst_config.primary_node) 2905 2906 # If the instance is non-redundant we cannot survive losing its primary 2907 # node, so we are not N+1 compliant. On the other hand we have no disk 2908 # templates with more than one secondary so that situation is not well 2909 # supported either. 2910 # FIXME: does not support file-backed instances 2911 if not inst_config.secondary_nodes: 2912 i_non_redundant.append(instance) 2913 2914 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT, 2915 instance, "instance has multiple secondary nodes: %s", 2916 utils.CommaJoin(inst_config.secondary_nodes), 2917 code=self.ETYPE_WARNING) 2918 2919 if inst_config.disk_template in constants.DTS_INT_MIRROR: 2920 pnode = inst_config.primary_node 2921 instance_nodes = utils.NiceSort(inst_config.all_nodes) 2922 instance_groups = {} 2923 2924 for node in instance_nodes: 2925 instance_groups.setdefault(self.all_node_info[node].group, 2926 []).append(node) 2927 2928 pretty_list = [ 2929 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name) 2930 # Sort so that we always list the primary node first. 2931 for group, nodes in sorted(instance_groups.items(), 2932 key=lambda (_, nodes): pnode in nodes, 2933 reverse=True)] 2934 2935 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS, 2936 instance, "instance has primary and secondary nodes in" 2937 " different groups: %s", utils.CommaJoin(pretty_list), 2938 code=self.ETYPE_WARNING) 2939 2940 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]: 2941 i_non_a_balanced.append(instance) 2942 2943 for snode in inst_config.secondary_nodes: 2944 s_img = node_image[snode] 2945 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode, 2946 "instance %s, connection to secondary node failed", instance) 2947 2948 if s_img.offline: 2949 inst_nodes_offline.append(snode) 2950 2951 # warn that the instance lives on offline nodes 2952 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance, 2953 "instance has offline secondary node(s) %s", 2954 utils.CommaJoin(inst_nodes_offline)) 2955 # ... or ghost/non-vm_capable nodes 2956 for node in inst_config.all_nodes: 2957 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance, 2958 "instance lives on ghost node %s", node) 2959 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE, 2960 instance, "instance lives on non-vm_capable node %s", node) 2961 2962 feedback_fn("* Verifying orphan volumes") 2963 reserved = utils.FieldSet(*cluster.reserved_lvs) 2964 2965 # We will get spurious "unknown volume" warnings if any node of this group 2966 # is secondary for an instance whose primary is in another group. To avoid 2967 # them, we find these instances and add their volumes to node_vol_should. 2968 for inst in self.all_inst_info.values(): 2969 for secondary in inst.secondary_nodes: 2970 if (secondary in self.my_node_info 2971 and inst.name not in self.my_inst_info): 2972 inst.MapLVsByNode(node_vol_should) 2973 break 2974 2975 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved) 2976 2977 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks: 2978 feedback_fn("* Verifying N+1 Memory redundancy") 2979 self._VerifyNPlusOneMemory(node_image, self.my_inst_info) 2980 2981 feedback_fn("* Other Notes") 2982 if i_non_redundant: 2983 feedback_fn(" - NOTICE: %d non-redundant instance(s) found." 2984 % len(i_non_redundant)) 2985 2986 if i_non_a_balanced: 2987 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found." 2988 % len(i_non_a_balanced)) 2989 2990 if n_offline: 2991 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline) 2992 2993 if n_drained: 2994 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained) 2995 2996 return not self.bad
2997
2998 - def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2999 """Analyze the post-hooks' result 3000 3001 This method analyses the hook result, handles it, and sends some 3002 nicely-formatted feedback back to the user. 3003 3004 @param phase: one of L{constants.HOOKS_PHASE_POST} or 3005 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase 3006 @param hooks_results: the results of the multi-node hooks rpc call 3007 @param feedback_fn: function used send feedback back to the caller 3008 @param lu_result: previous Exec result 3009 @return: the new Exec result, based on the previous result 3010 and hook results 3011 3012 """ 3013 # We only really run POST phase hooks, only for non-empty groups, 3014 # and are only interested in their results 3015 if not self.my_node_names: 3016 # empty node group 3017 pass 3018 elif phase == constants.HOOKS_PHASE_POST: 3019 # Used to change hooks' output to proper indentation 3020 feedback_fn("* Hooks Results") 3021 assert hooks_results, "invalid result from hooks" 3022 3023 for node_name in hooks_results: 3024 res = hooks_results[node_name] 3025 msg = res.fail_msg 3026 test = msg and not res.offline 3027 self._ErrorIf(test, self.ENODEHOOKS, node_name, 3028 "Communication failure in hooks execution: %s", msg) 3029 if res.offline or msg: 3030 # No need to investigate payload if node is offline or gave 3031 # an error. 3032 continue 3033 for script, hkr, output in res.payload: 3034 test = hkr == constants.HKR_FAIL 3035 self._ErrorIf(test, self.ENODEHOOKS, node_name, 3036 "Script %s failed, output:", script) 3037 if test: 3038 output = self._HOOKS_INDENT_RE.sub(" ", output) 3039 feedback_fn("%s" % output) 3040 lu_result = False 3041 3042 return lu_result
3043
3044 3045 -class LUClusterVerifyDisks(NoHooksLU):
3046 """Verifies the cluster disks status. 3047 3048 """ 3049 REQ_BGL = False 3050
3051 - def ExpandNames(self):
3052 self.share_locks = _ShareAll() 3053 self.needed_locks = { 3054 locking.LEVEL_NODEGROUP: locking.ALL_SET, 3055 }
3056
3057 - def Exec(self, feedback_fn):
3058 group_names = self.owned_locks(locking.LEVEL_NODEGROUP) 3059 3060 # Submit one instance of L{opcodes.OpGroupVerifyDisks} per node group 3061 return ResultWithJobs([[opcodes.OpGroupVerifyDisks(group_name=group)] 3062 for group in group_names])
3063
3064 3065 -class LUGroupVerifyDisks(NoHooksLU):
3066 """Verifies the status of all disks in a node group. 3067 3068 """ 3069 REQ_BGL = False 3070
3071 - def ExpandNames(self):
3072 # Raises errors.OpPrereqError on its own if group can't be found 3073 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 3074 3075 self.share_locks = _ShareAll() 3076 self.needed_locks = { 3077 locking.LEVEL_INSTANCE: [], 3078 locking.LEVEL_NODEGROUP: [], 3079 locking.LEVEL_NODE: [], 3080 }
3081
3082 - def DeclareLocks(self, level):
3083 if level == locking.LEVEL_INSTANCE: 3084 assert not self.needed_locks[locking.LEVEL_INSTANCE] 3085 3086 # Lock instances optimistically, needs verification once node and group 3087 # locks have been acquired 3088 self.needed_locks[locking.LEVEL_INSTANCE] = \ 3089 self.cfg.GetNodeGroupInstances(self.group_uuid) 3090 3091 elif level == locking.LEVEL_NODEGROUP: 3092 assert not self.needed_locks[locking.LEVEL_NODEGROUP] 3093 3094 self.needed_locks[locking.LEVEL_NODEGROUP] = \ 3095 set([self.group_uuid] + 3096 # Lock all groups used by instances optimistically; this requires 3097 # going via the node before it's locked, requiring verification 3098 # later on 3099 [group_uuid 3100 for instance_name in self.owned_locks(locking.LEVEL_INSTANCE) 3101 for group_uuid in self.cfg.GetInstanceNodeGroups(instance_name)]) 3102 3103 elif level == locking.LEVEL_NODE: 3104 # This will only lock the nodes in the group to be verified which contain 3105 # actual instances 3106 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND 3107 self._LockInstancesNodes() 3108 3109 # Lock all nodes in group to be verified 3110 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP) 3111 member_nodes = self.cfg.GetNodeGroup(self.group_uuid).members 3112 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
3113
3114 - def CheckPrereq(self):
3115 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE)) 3116 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) 3117 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) 3118 3119 assert self.group_uuid in owned_groups 3120 3121 # Check if locked instances are still correct 3122 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances) 3123 3124 # Get instance information 3125 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances)) 3126 3127 # Check if node groups for locked instances are still correct 3128 for (instance_name, inst) in self.instances.items(): 3129 assert owned_nodes.issuperset(inst.all_nodes), \ 3130 "Instance %s's nodes changed while we kept the lock" % instance_name 3131 3132 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name, 3133 owned_groups) 3134 3135 assert self.group_uuid in inst_groups, \ 3136 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
3137
3138 - def Exec(self, feedback_fn):
3139 """Verify integrity of cluster disks. 3140 3141 @rtype: tuple of three items 3142 @return: a tuple of (dict of node-to-node_error, list of instances 3143 which need activate-disks, dict of instance: (node, volume) for 3144 missing volumes 3145 3146 """ 3147 res_nodes = {} 3148 res_instances = set() 3149 res_missing = {} 3150 3151 nv_dict = _MapInstanceDisksToNodes([inst 3152 for inst in self.instances.values() 3153 if inst.admin_up]) 3154 3155 if nv_dict: 3156 nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) & 3157 set(self.cfg.GetVmCapableNodeList())) 3158 3159 node_lvs = self.rpc.call_lv_list(nodes, []) 3160 3161 for (node, node_res) in node_lvs.items(): 3162 if node_res.offline: 3163 continue 3164 3165 msg = node_res.fail_msg 3166 if msg: 3167 logging.warning("Error enumerating LVs on node %s: %s", node, msg) 3168 res_nodes[node] = msg 3169 continue 3170 3171 for lv_name, (_, _, lv_online) in node_res.payload.items(): 3172 inst = nv_dict.pop((node, lv_name), None) 3173 if not (lv_online or inst is None): 3174 res_instances.add(inst) 3175 3176 # any leftover items in nv_dict are missing LVs, let's arrange the data 3177 # better 3178 for key, inst in nv_dict.iteritems(): 3179 res_missing.setdefault(inst, []).append(list(key)) 3180 3181 return (res_nodes, list(res_instances), res_missing)
3182
3183 3184 -class LUClusterRepairDiskSizes(NoHooksLU):
3185 """Verifies the cluster disks sizes. 3186 3187 """ 3188 REQ_BGL = False 3189
3190 - def ExpandNames(self):
3191 if self.op.instances: 3192 self.wanted_names = _GetWantedInstances(self, self.op.instances) 3193 self.needed_locks = { 3194 locking.LEVEL_NODE: [], 3195 locking.LEVEL_INSTANCE: self.wanted_names, 3196 } 3197 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE 3198 else: 3199 self.wanted_names = None 3200 self.needed_locks = { 3201 locking.LEVEL_NODE: locking.ALL_SET, 3202 locking.LEVEL_INSTANCE: locking.ALL_SET, 3203 } 3204 self.share_locks = { 3205 locking.LEVEL_NODE: 1, 3206 locking.LEVEL_INSTANCE: 0, 3207 }
3208
3209 - def DeclareLocks(self, level):
3210 if level == locking.LEVEL_NODE and self.wanted_names is not None: 3211 self._LockInstancesNodes(primary_only=True)
3212
3213 - def CheckPrereq(self):
3214 """Check prerequisites. 3215 3216 This only checks the optional instance list against the existing names. 3217 3218 """ 3219 if self.wanted_names is None: 3220 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE) 3221 3222 self.wanted_instances = \ 3223 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
3224
3225 - def _EnsureChildSizes(self, disk):
3226 """Ensure children of the disk have the needed disk size. 3227 3228 This is valid mainly for DRBD8 and fixes an issue where the 3229 children have smaller disk size. 3230 3231 @param disk: an L{ganeti.objects.Disk} object 3232 3233 """ 3234 if disk.dev_type == constants.LD_DRBD8: 3235 assert disk.children, "Empty children for DRBD8?" 3236 fchild = disk.children[0] 3237 mismatch = fchild.size < disk.size 3238 if mismatch: 3239 self.LogInfo("Child disk has size %d, parent %d, fixing", 3240 fchild.size, disk.size) 3241 fchild.size = disk.size 3242 3243 # and we recurse on this child only, not on the metadev 3244 return self._EnsureChildSizes(fchild) or mismatch 3245 else: 3246 return False
3247
3248 - def Exec(self, feedback_fn):
3249 """Verify the size of cluster disks. 3250 3251 """ 3252 # TODO: check child disks too 3253 # TODO: check differences in size between primary/secondary nodes 3254 per_node_disks = {} 3255 for instance in self.wanted_instances: 3256 pnode = instance.primary_node 3257 if pnode not in per_node_disks: 3258 per_node_disks[pnode] = [] 3259 for idx, disk in enumerate(instance.disks): 3260 per_node_disks[pnode].append((instance, idx, disk)) 3261 3262 changed = [] 3263 for node, dskl in per_node_disks.items(): 3264 newl = [v[2].Copy() for v in dskl] 3265 for dsk in newl: 3266 self.cfg.SetDiskID(dsk, node) 3267 result = self.rpc.call_blockdev_getsize(node, newl) 3268 if result.fail_msg: 3269 self.LogWarning("Failure in blockdev_getsize call to node" 3270 " %s, ignoring", node) 3271 continue 3272 if len(result.payload) != len(dskl): 3273 logging.warning("Invalid result from node %s: len(dksl)=%d," 3274 " result.payload=%s", node, len(dskl), result.payload) 3275 self.LogWarning("Invalid result from node %s, ignoring node results", 3276 node) 3277 continue 3278 for ((instance, idx, disk), size) in zip(dskl, result.payload): 3279 if size is None: 3280 self.LogWarning("Disk %d of instance %s did not return size" 3281 " information, ignoring", idx, instance.name) 3282 continue 3283 if not isinstance(size, (int, long)): 3284 self.LogWarning("Disk %d of instance %s did not return valid" 3285 " size information, ignoring", idx, instance.name) 3286 continue 3287 size = size >> 20 3288 if size != disk.size: 3289 self.LogInfo("Disk %d of instance %s has mismatched size," 3290 " correcting: recorded %d, actual %d", idx, 3291 instance.name, disk.size, size) 3292 disk.size = size 3293 self.cfg.Update(instance, feedback_fn) 3294 changed.append((instance.name, idx, size)) 3295 if self._EnsureChildSizes(disk): 3296 self.cfg.Update(instance, feedback_fn) 3297 changed.append((instance.name, idx, disk.size)) 3298 return changed
3299
3300 3301 -class LUClusterRename(LogicalUnit):
3302 """Rename the cluster. 3303 3304 """ 3305 HPATH = "cluster-rename" 3306 HTYPE = constants.HTYPE_CLUSTER 3307
3308 - def BuildHooksEnv(self):
3309 """Build hooks env. 3310 3311 """ 3312 return { 3313 "OP_TARGET": self.cfg.GetClusterName(), 3314 "NEW_NAME": self.op.name, 3315 }
3316
3317 - def BuildHooksNodes(self):
3318 """Build hooks nodes. 3319 3320 """ 3321 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
3322
3323 - def CheckPrereq(self):
3324 """Verify that the passed name is a valid one. 3325 3326 """ 3327 hostname = netutils.GetHostname(name=self.op.name, 3328 family=self.cfg.GetPrimaryIPFamily()) 3329 3330 new_name = hostname.name 3331 self.ip = new_ip = hostname.ip 3332 old_name = self.cfg.GetClusterName() 3333 old_ip = self.cfg.GetMasterIP() 3334 if new_name == old_name and new_ip == old_ip: 3335 raise errors.OpPrereqError("Neither the name nor the IP address of the" 3336 " cluster has changed", 3337 errors.ECODE_INVAL) 3338 if new_ip != old_ip: 3339 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT): 3340 raise errors.OpPrereqError("The given cluster IP address (%s) is" 3341 " reachable on the network" % 3342 new_ip, errors.ECODE_NOTUNIQUE) 3343 3344 self.op.name = new_name
3345
3346 - def Exec(self, feedback_fn):
3347 """Rename the cluster. 3348 3349 """ 3350 clustername = self.op.name 3351 ip = self.ip 3352 3353 # shutdown the master IP 3354 master = self.cfg.GetMasterNode() 3355 result = self.rpc.call_node_stop_master(master, False) 3356 result.Raise("Could not disable the master role") 3357 3358 try: 3359 cluster = self.cfg.GetClusterInfo() 3360 cluster.cluster_name = clustername 3361 cluster.master_ip = ip 3362 self.cfg.Update(cluster, feedback_fn) 3363 3364 # update the known hosts file 3365 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE) 3366 node_list = self.cfg.GetOnlineNodeList() 3367 try: 3368 node_list.remove(master) 3369 except ValueError: 3370 pass 3371 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE) 3372 finally: 3373 result = self.rpc.call_node_start_master(master, False, False) 3374 msg = result.fail_msg 3375 if msg: 3376 self.LogWarning("Could not re-enable the master role on" 3377 " the master, please restart manually: %s", msg) 3378 3379 return clustername
3380
3381 3382 -class LUClusterSetParams(LogicalUnit):
3383 """Change the parameters of the cluster. 3384 3385 """ 3386 HPATH = "cluster-modify" 3387 HTYPE = constants.HTYPE_CLUSTER 3388 REQ_BGL = False 3389
3390 - def CheckArguments(self):
3391 """Check parameters 3392 3393 """ 3394 if self.op.uid_pool: 3395 uidpool.CheckUidPool(self.op.uid_pool) 3396 3397 if self.op.add_uids: 3398 uidpool.CheckUidPool(self.op.add_uids) 3399 3400 if self.op.remove_uids: 3401 uidpool.CheckUidPool(self.op.remove_uids)
3402
3403 - def ExpandNames(self):
3404 # FIXME: in the future maybe other cluster params won't require checking on 3405 # all nodes to be modified. 3406 self.needed_locks = { 3407 locking.LEVEL_NODE: locking.ALL_SET, 3408 } 3409 self.share_locks[locking.LEVEL_NODE] = 1
3410
3411 - def BuildHooksEnv(self):
3412 """Build hooks env. 3413 3414 """ 3415 return { 3416 "OP_TARGET": self.cfg.GetClusterName(), 3417 "NEW_VG_NAME": self.op.vg_name, 3418 }
3419
3420 - def BuildHooksNodes(self):
3421 """Build hooks nodes. 3422 3423 """ 3424 mn = self.cfg.GetMasterNode() 3425 return ([mn], [mn])
3426
3427 - def CheckPrereq(self):
3428 """Check prerequisites. 3429 3430 This checks whether the given params don't conflict and 3431 if the given volume group is valid. 3432 3433 """ 3434 if self.op.vg_name is not None and not self.op.vg_name: 3435 if self.cfg.HasAnyDiskOfType(constants.LD_LV): 3436 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based" 3437 " instances exist", errors.ECODE_INVAL) 3438 3439 if self.op.drbd_helper is not None and not self.op.drbd_helper: 3440 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8): 3441 raise errors.OpPrereqError("Cannot disable drbd helper while" 3442 " drbd-based instances exist", 3443 errors.ECODE_INVAL) 3444 3445 node_list = self.owned_locks(locking.LEVEL_NODE) 3446 3447 # if vg_name not None, checks given volume group on all nodes 3448 if self.op.vg_name: 3449 vglist = self.rpc.call_vg_list(node_list) 3450 for node in node_list: 3451 msg = vglist[node].fail_msg 3452 if msg: 3453 # ignoring down node 3454 self.LogWarning("Error while gathering data on node %s" 3455 " (ignoring node): %s", node, msg) 3456 continue 3457 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload, 3458 self.op.vg_name, 3459 constants.MIN_VG_SIZE) 3460 if vgstatus: 3461 raise errors.OpPrereqError("Error on node '%s': %s" % 3462 (node, vgstatus), errors.ECODE_ENVIRON) 3463 3464 if self.op.drbd_helper: 3465 # checks given drbd helper on all nodes 3466 helpers = self.rpc.call_drbd_helper(node_list) 3467 for (node, ninfo) in self.cfg.GetMultiNodeInfo(node_list): 3468 if ninfo.offline: 3469 self.LogInfo("Not checking drbd helper on offline node %s", node) 3470 continue 3471 msg = helpers[node].fail_msg 3472 if msg: 3473 raise errors.OpPrereqError("Error checking drbd helper on node" 3474 " '%s': %s" % (node, msg), 3475 errors.ECODE_ENVIRON) 3476 node_helper = helpers[node].payload 3477 if node_helper != self.op.drbd_helper: 3478 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" % 3479 (node, node_helper), errors.ECODE_ENVIRON) 3480 3481 self.cluster = cluster = self.cfg.GetClusterInfo() 3482 # validate params changes 3483 if self.op.beparams: 3484 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES) 3485 self.new_beparams = cluster.SimpleFillBE(self.op.beparams) 3486 3487 if self.op.ndparams: 3488 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES) 3489 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams) 3490 3491 # TODO: we need a more general way to handle resetting 3492 # cluster-level parameters to default values 3493 if self.new_ndparams["oob_program"] == "": 3494 self.new_ndparams["oob_program"] = \ 3495 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM] 3496 3497 if self.op.nicparams: 3498 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES) 3499 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams) 3500 objects.NIC.CheckParameterSyntax(self.new_nicparams) 3501 nic_errors = [] 3502 3503 # check all instances for consistency 3504 for instance in self.cfg.GetAllInstancesInfo().values(): 3505 for nic_idx, nic in enumerate(instance.nics): 3506 params_copy = copy.deepcopy(nic.nicparams) 3507 params_filled = objects.FillDict(self.new_nicparams, params_copy) 3508 3509 # check parameter syntax 3510 try: 3511 objects.NIC.CheckParameterSyntax(params_filled) 3512 except errors.ConfigurationError, err: 3513 nic_errors.append("Instance %s, nic/%d: %s" % 3514 (instance.name, nic_idx, err)) 3515 3516 # if we're moving instances to routed, check that they have an ip 3517 target_mode = params_filled[constants.NIC_MODE] 3518 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip: 3519 nic_errors.append("Instance %s, nic/%d: routed NIC with no ip" 3520 " address" % (instance.name, nic_idx)) 3521 if nic_errors: 3522 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" % 3523 "\n".join(nic_errors)) 3524 3525 # hypervisor list/parameters 3526 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {}) 3527 if self.op.hvparams: 3528 for hv_name, hv_dict in self.op.hvparams.items(): 3529 if hv_name not in self.new_hvparams: 3530 self.new_hvparams[hv_name] = hv_dict 3531 else: 3532 self.new_hvparams[hv_name].update(hv_dict) 3533 3534 # os hypervisor parameters 3535 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {}) 3536 if self.op.os_hvp: 3537 for os_name, hvs in self.op.os_hvp.items(): 3538 if os_name not in self.new_os_hvp: 3539 self.new_os_hvp[os_name] = hvs 3540 else: 3541 for hv_name, hv_dict in hvs.items(): 3542 if hv_name not in self.new_os_hvp[os_name]: 3543 self.new_os_hvp[os_name][hv_name] = hv_dict 3544 else: 3545 self.new_os_hvp[os_name][hv_name].update(hv_dict) 3546 3547 # os parameters 3548 self.new_osp = objects.FillDict(cluster.osparams, {}) 3549 if self.op.osparams: 3550 for os_name, osp in self.op.osparams.items(): 3551 if os_name not in self.new_osp: 3552 self.new_osp[os_name] = {} 3553 3554 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp, 3555 use_none=True) 3556 3557 if not self.new_osp[os_name]: 3558 # we removed all parameters 3559 del self.new_osp[os_name] 3560 else: 3561 # check the parameter validity (remote check) 3562 _CheckOSParams(self, False, [self.cfg.GetMasterNode()], 3563 os_name, self.new_osp[os_name]) 3564 3565 # changes to the hypervisor list 3566 if self.op.enabled_hypervisors is not None: 3567 self.hv_list = self.op.enabled_hypervisors 3568 for hv in self.hv_list: 3569 # if the hypervisor doesn't already exist in the cluster 3570 # hvparams, we initialize it to empty, and then (in both 3571 # cases) we make sure to fill the defaults, as we might not 3572 # have a complete defaults list if the hypervisor wasn't 3573 # enabled before 3574 if hv not in new_hvp: 3575 new_hvp[hv] = {} 3576 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv]) 3577 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES) 3578 else: 3579 self.hv_list = cluster.enabled_hypervisors 3580 3581 if self.op.hvparams or self.op.enabled_hypervisors is not None: 3582 # either the enabled list has changed, or the parameters have, validate 3583 for hv_name, hv_params in self.new_hvparams.items(): 3584 if ((self.op.hvparams and hv_name in self.op.hvparams) or 3585 (self.op.enabled_hypervisors and 3586 hv_name in self.op.enabled_hypervisors)): 3587 # either this is a new hypervisor, or its parameters have changed 3588 hv_class = hypervisor.GetHypervisor(hv_name) 3589 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) 3590 hv_class.CheckParameterSyntax(hv_params) 3591 _CheckHVParams(self, node_list, hv_name, hv_params) 3592 3593 if self.op.os_hvp: 3594 # no need to check any newly-enabled hypervisors, since the 3595 # defaults have already been checked in the above code-block 3596 for os_name, os_hvp in self.new_os_hvp.items(): 3597 for hv_name, hv_params in os_hvp.items(): 3598 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) 3599 # we need to fill in the new os_hvp on top of the actual hv_p 3600 cluster_defaults = self.new_hvparams.get(hv_name, {}) 3601 new_osp = objects.FillDict(cluster_defaults, hv_params) 3602 hv_class = hypervisor.GetHypervisor(hv_name) 3603 hv_class.CheckParameterSyntax(new_osp) 3604 _CheckHVParams(self, node_list, hv_name, new_osp) 3605 3606 if self.op.default_iallocator: 3607 alloc_script = utils.FindFile(self.op.default_iallocator, 3608 constants.IALLOCATOR_SEARCH_PATH, 3609 os.path.isfile) 3610 if alloc_script is None: 3611 raise errors.OpPrereqError("Invalid default iallocator script '%s'" 3612 " specified" % self.op.default_iallocator, 3613 errors.ECODE_INVAL)
3614
3615 - def Exec(self, feedback_fn):
3616 """Change the parameters of the cluster. 3617 3618 """ 3619 if self.op.vg_name is not None: 3620 new_volume = self.op.vg_name 3621 if not new_volume: 3622 new_volume = None 3623 if new_volume != self.cfg.GetVGName(): 3624 self.cfg.SetVGName(new_volume) 3625 else: 3626 feedback_fn("Cluster LVM configuration already in desired" 3627 " state, not changing") 3628 if self.op.drbd_helper is not None: 3629 new_helper = self.op.drbd_helper 3630 if not new_helper: 3631 new_helper = None 3632 if new_helper != self.cfg.GetDRBDHelper(): 3633 self.cfg.SetDRBDHelper(new_helper) 3634 else: 3635 feedback_fn("Cluster DRBD helper already in desired state," 3636 " not changing") 3637 if self.op.hvparams: 3638 self.cluster.hvparams = self.new_hvparams 3639 if self.op.os_hvp: 3640 self.cluster.os_hvp = self.new_os_hvp 3641 if self.op.enabled_hypervisors is not None: 3642 self.cluster.hvparams = self.new_hvparams 3643 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors 3644 if self.op.beparams: 3645 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams 3646 if self.op.nicparams: 3647 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams 3648 if self.op.osparams: 3649 self.cluster.osparams = self.new_osp 3650 if self.op.ndparams: 3651 self.cluster.ndparams = self.new_ndparams 3652 3653 if self.op.candidate_pool_size is not None: 3654 self.cluster.candidate_pool_size = self.op.candidate_pool_size 3655 # we need to update the pool size here, otherwise the save will fail 3656 _AdjustCandidatePool(self, []) 3657 3658 if self.op.maintain_node_health is not None: 3659 self.cluster.maintain_node_health = self.op.maintain_node_health 3660 3661 if self.op.prealloc_wipe_disks is not None: 3662 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks 3663 3664 if self.op.add_uids is not None: 3665 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids) 3666 3667 if self.op.remove_uids is not None: 3668 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids) 3669 3670 if self.op.uid_pool is not None: 3671 self.cluster.uid_pool = self.op.uid_pool 3672 3673 if self.op.default_iallocator is not None: 3674 self.cluster.default_iallocator = self.op.default_iallocator 3675 3676 if self.op.reserved_lvs is not None: 3677 self.cluster.reserved_lvs = self.op.reserved_lvs 3678 3679 def helper_os(aname, mods, desc): 3680 desc += " OS list" 3681 lst = getattr(self.cluster, aname) 3682 for key, val in mods: 3683 if key == constants.DDM_ADD: 3684 if val in lst: 3685 feedback_fn("OS %s already in %s, ignoring" % (val, desc)) 3686 else: 3687 lst.append(val) 3688 elif key == constants.DDM_REMOVE: 3689 if val in lst: 3690 lst.remove(val) 3691 else: 3692 feedback_fn("OS %s not found in %s, ignoring" % (val, desc)) 3693 else: 3694 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3695 3696 if self.op.hidden_os: 3697 helper_os("hidden_os", self.op.hidden_os, "hidden") 3698 3699 if self.op.blacklisted_os: 3700 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted") 3701 3702 if self.op.master_netdev: 3703 master = self.cfg.GetMasterNode() 3704 feedback_fn("Shutting down master ip on the current netdev (%s)" % 3705 self.cluster.master_netdev) 3706 result = self.rpc.call_node_stop_master(master, False) 3707 result.Raise("Could not disable the master ip") 3708 feedback_fn("Changing master_netdev from %s to %s" % 3709 (self.cluster.master_netdev, self.op.master_netdev)) 3710 self.cluster.master_netdev = self.op.master_netdev 3711 3712 self.cfg.Update(self.cluster, feedback_fn) 3713 3714 if self.op.master_netdev: 3715 feedback_fn("Starting the master ip on the new master netdev (%s)" % 3716 self.op.master_netdev) 3717 result = self.rpc.call_node_start_master(master, False, False) 3718 if result.fail_msg: 3719 self.LogWarning("Could not re-enable the master ip on" 3720 " the master, please restart manually: %s", 3721 result.fail_msg)
3722
3723 3724 -def _UploadHelper(lu, nodes, fname):
3725 """Helper for uploading a file and showing warnings. 3726 3727 """ 3728 if os.path.exists(fname): 3729 result = lu.rpc.call_upload_file(nodes, fname) 3730 for to_node, to_result in result.items(): 3731 msg = to_result.fail_msg 3732 if msg: 3733 msg = ("Copy of file %s to node %s failed: %s" % 3734 (fname, to_node, msg)) 3735 lu.proc.LogWarning(msg)
3736
3737 3738 -def _ComputeAncillaryFiles(cluster, redist):
3739 """Compute files external to Ganeti which need to be consistent. 3740 3741 @type redist: boolean 3742 @param redist: Whether to include files which need to be redistributed 3743 3744 """ 3745 # Compute files for all nodes 3746 files_all = set([ 3747 constants.SSH_KNOWN_HOSTS_FILE, 3748 constants.CONFD_HMAC_KEY, 3749 constants.CLUSTER_DOMAIN_SECRET_FILE, 3750 ]) 3751 3752 if not redist: 3753 files_all.update(constants.ALL_CERT_FILES) 3754 files_all.update(ssconf.SimpleStore().GetFileList()) 3755 else: 3756 # we need to ship at least the RAPI certificate 3757 files_all.add(constants.RAPI_CERT_FILE) 3758 3759 if cluster.modify_etc_hosts: 3760 files_all.add(constants.ETC_HOSTS) 3761 3762 # Files which must either exist on all nodes or on none 3763 files_all_opt = set([ 3764 constants.RAPI_USERS_FILE, 3765 ]) 3766 3767 # Files which should only be on master candidates 3768 files_mc = set() 3769 if not redist: 3770 files_mc.add(constants.CLUSTER_CONF_FILE) 3771 3772 # Files which should only be on VM-capable nodes 3773 files_vm = set(filename 3774 for hv_name in cluster.enabled_hypervisors 3775 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()) 3776 3777 # Filenames must be unique 3778 assert (len(files_all | files_all_opt | files_mc | files_vm) == 3779 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \ 3780 "Found file listed in more than one file list" 3781 3782 return (files_all, files_all_opt, files_mc, files_vm)
3783
3784 3785 -def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3786 """Distribute additional files which are part of the cluster configuration. 3787 3788 ConfigWriter takes care of distributing the config and ssconf files, but 3789 there are more files which should be distributed to all nodes. This function 3790 makes sure those are copied. 3791 3792 @param lu: calling logical unit 3793 @param additional_nodes: list of nodes not in the config to distribute to 3794 @type additional_vm: boolean 3795 @param additional_vm: whether the additional nodes are vm-capable or not 3796 3797 """ 3798 # Gather target nodes 3799 cluster = lu.cfg.GetClusterInfo() 3800 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode()) 3801 3802 online_nodes = lu.cfg.GetOnlineNodeList() 3803 vm_nodes = lu.cfg.GetVmCapableNodeList() 3804 3805 if additional_nodes is not None: 3806 online_nodes.extend(additional_nodes) 3807 if additional_vm: 3808 vm_nodes.extend(additional_nodes) 3809 3810 # Never distribute to master node 3811 for nodelist in [online_nodes, vm_nodes]: 3812 if master_info.name in nodelist: 3813 nodelist.remove(master_info.name) 3814 3815 # Gather file lists 3816 (files_all, files_all_opt, files_mc, files_vm) = \ 3817 _ComputeAncillaryFiles(cluster, True) 3818 3819 # Never re-distribute configuration file from here 3820 assert not (constants.CLUSTER_CONF_FILE in files_all or 3821 constants.CLUSTER_CONF_FILE in files_vm) 3822 assert not files_mc, "Master candidates not handled in this function" 3823 3824 filemap = [ 3825 (online_nodes, files_all), 3826 (online_nodes, files_all_opt), 3827 (vm_nodes, files_vm), 3828 ] 3829 3830 # Upload the files 3831 for (node_list, files) in filemap: 3832 for fname in files: 3833 _UploadHelper(lu, node_list, fname)
3834
3835 3836 -class LUClusterRedistConf(NoHooksLU):
3837 """Force the redistribution of cluster configuration. 3838 3839 This is a very simple LU. 3840 3841 """ 3842 REQ_BGL = False 3843
3844 - def ExpandNames(self):
3845 self.needed_locks = { 3846 locking.LEVEL_NODE: locking.ALL_SET, 3847 } 3848 self.share_locks[locking.LEVEL_NODE] = 1
3849
3850 - def Exec(self, feedback_fn):
3851 """Redistribute the configuration. 3852 3853 """ 3854 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn) 3855 _RedistributeAncillaryFiles(self)
3856
3857 3858 -def _WaitForSync(lu, instance, disks=None, oneshot=False):
3859 """Sleep and poll for an instance's disk to sync. 3860 3861 """ 3862 if not instance.disks or disks is not None and not disks: 3863 return True 3864 3865 disks = _ExpandCheckDisks(instance, disks) 3866 3867 if not oneshot: 3868 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name) 3869 3870 node = instance.primary_node 3871 3872 for dev in disks: 3873 lu.cfg.SetDiskID(dev, node) 3874 3875 # TODO: Convert to utils.Retry 3876 3877 retries = 0 3878 degr_retries = 10 # in seconds, as we sleep 1 second each time 3879 while True: 3880 max_time = 0 3881 done = True 3882 cumul_degraded = False 3883 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks) 3884 msg = rstats.fail_msg 3885 if msg: 3886 lu.LogWarning("Can't get any data from node %s: %s", node, msg) 3887 retries += 1 3888 if retries >= 10: 3889 raise errors.RemoteError("Can't contact node %s for mirror data," 3890 " aborting." % node) 3891 time.sleep(6) 3892 continue 3893 rstats = rstats.payload 3894 retries = 0 3895 for i, mstat in enumerate(rstats): 3896 if mstat is None: 3897 lu.LogWarning("Can't compute data for node %s/%s", 3898 node, disks[i].iv_name) 3899 continue 3900 3901 cumul_degraded = (cumul_degraded or 3902 (mstat.is_degraded and mstat.sync_percent is None)) 3903 if mstat.sync_percent is not None: 3904 done = False 3905 if mstat.estimated_time is not None: 3906 rem_time = ("%s remaining (estimated)" % 3907 utils.FormatSeconds(mstat.estimated_time)) 3908 max_time = mstat.estimated_time 3909 else: 3910 rem_time = "no time estimate" 3911 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" % 3912 (disks[i].iv_name, mstat.sync_percent, rem_time)) 3913 3914 # if we're done but degraded, let's do a few small retries, to 3915 # make sure we see a stable and not transient situation; therefore 3916 # we force restart of the loop 3917 if (done or oneshot) and cumul_degraded and degr_retries > 0: 3918 logging.info("Degraded disks found, %d retries left", degr_retries) 3919 degr_retries -= 1 3920 time.sleep(1) 3921 continue 3922 3923 if done or oneshot: 3924 break 3925 3926 time.sleep(min(60, max_time)) 3927 3928 if done: 3929 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name) 3930 return not cumul_degraded
3931
3932 3933 -def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3934 """Check that mirrors are not degraded. 3935 3936 The ldisk parameter, if True, will change the test from the 3937 is_degraded attribute (which represents overall non-ok status for 3938 the device(s)) to the ldisk (representing the local storage status). 3939 3940 """ 3941 lu.cfg.SetDiskID(dev, node) 3942 3943 result = True 3944 3945 if on_primary or dev.AssembleOnSecondary(): 3946 rstats = lu.rpc.call_blockdev_find(node, dev) 3947 msg = rstats.fail_msg 3948 if msg: 3949 lu.LogWarning("Can't find disk on node %s: %s", node, msg) 3950 result = False 3951 elif not rstats.payload: 3952 lu.LogWarning("Can't find disk on node %s", node) 3953 result = False 3954 else: 3955 if ldisk: 3956 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY 3957 else: 3958 result = result and not rstats.payload.is_degraded 3959 3960 if dev.children: 3961 for child in dev.children: 3962 result = result and _CheckDiskConsistency(lu, child, node, on_primary) 3963 3964 return result
3965
3966 3967 -class LUOobCommand(NoHooksLU):
3968 """Logical unit for OOB handling. 3969 3970 """ 3971 REQ_BGL = False 3972 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE) 3973
3974 - def ExpandNames(self):
3975 """Gather locks we need. 3976 3977 """ 3978 if self.op.node_names: 3979 self.op.node_names = _GetWantedNodes(self, self.op.node_names) 3980 lock_names = self.op.node_names 3981 else: 3982 lock_names = locking.ALL_SET 3983 3984 self.needed_locks = { 3985 locking.LEVEL_NODE: lock_names, 3986 }
3987
3988 - def CheckPrereq(self):
3989 """Check prerequisites. 3990 3991 This checks: 3992 - the node exists in the configuration 3993 - OOB is supported 3994 3995 Any errors are signaled by raising errors.OpPrereqError. 3996 3997 """ 3998 self.nodes = [] 3999 self.master_node = self.cfg.GetMasterNode() 4000 4001 assert self.op.power_delay >= 0.0 4002 4003 if self.op.node_names: 4004 if (self.op.command in self._SKIP_MASTER and 4005 self.master_node in self.op.node_names): 4006 master_node_obj = self.cfg.GetNodeInfo(self.master_node) 4007 master_oob_handler = _SupportsOob(self.cfg, master_node_obj) 4008 4009 if master_oob_handler: 4010 additional_text = ("run '%s %s %s' if you want to operate on the" 4011 " master regardless") % (master_oob_handler, 4012 self.op.command, 4013 self.master_node) 4014 else: 4015 additional_text = "it does not support out-of-band operations" 4016 4017 raise errors.OpPrereqError(("Operating on the master node %s is not" 4018 " allowed for %s; %s") % 4019 (self.master_node, self.op.command, 4020 additional_text), errors.ECODE_INVAL) 4021 else: 4022 self.op.node_names = self.cfg.GetNodeList() 4023 if self.op.command in self._SKIP_MASTER: 4024 self.op.node_names.remove(self.master_node) 4025 4026 if self.op.command in self._SKIP_MASTER: 4027 assert self.master_node not in self.op.node_names 4028 4029 for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names): 4030 if node is None: 4031 raise errors.OpPrereqError("Node %s not found" % node_name, 4032 errors.ECODE_NOENT) 4033 else: 4034 self.nodes.append(node) 4035 4036 if (not self.op.ignore_status and 4037 (self.op.command == constants.OOB_POWER_OFF and not node.offline)): 4038 raise errors.OpPrereqError(("Cannot power off node %s because it is" 4039 " not marked offline") % node_name, 4040 errors.ECODE_STATE)
4041
4042 - def Exec(self, feedback_fn):
4043 """Execute OOB and return result if we expect any. 4044 4045 """ 4046 master_node = self.master_node 4047 ret = [] 4048 4049 for idx, node in enumerate(utils.NiceSort(self.nodes, 4050 key=lambda node: node.name)): 4051 node_entry = [(constants.RS_NORMAL, node.name)] 4052 ret.append(node_entry) 4053 4054 oob_program = _SupportsOob(self.cfg, node) 4055 4056 if not oob_program: 4057 node_entry.append((constants.RS_UNAVAIL, None)) 4058 continue 4059 4060 logging.info("Executing out-of-band command '%s' using '%s' on %s", 4061 self.op.command, oob_program, node.name) 4062 result = self.rpc.call_run_oob(master_node, oob_program, 4063 self.op.command, node.name, 4064 self.op.timeout) 4065 4066 if result.fail_msg: 4067 self.LogWarning("Out-of-band RPC failed on node '%s': %s", 4068 node.name, result.fail_msg) 4069 node_entry.append((constants.RS_NODATA, None)) 4070 else: 4071 try: 4072 self._CheckPayload(result) 4073 except errors.OpExecError, err: 4074 self.LogWarning("Payload returned by node '%s' is not valid: %s", 4075 node.name, err) 4076 node_entry.append((constants.RS_NODATA, None)) 4077 else: 4078 if self.op.command == constants.OOB_HEALTH: 4079 # For health we should log important events 4080 for item, status in result.payload: 4081 if status in [constants.OOB_STATUS_WARNING, 4082 constants.OOB_STATUS_CRITICAL]: 4083 self.LogWarning("Item '%s' on node '%s' has status '%s'", 4084 item, node.name, status) 4085 4086 if self.op.command == constants.OOB_POWER_ON: 4087 node.powered = True 4088 elif self.op.command == constants.OOB_POWER_OFF: 4089 node.powered = False 4090 elif self.op.command == constants.OOB_POWER_STATUS: 4091 powered = result.payload[constants.OOB_POWER_STATUS_POWERED] 4092 if powered != node.powered: 4093 logging.warning(("Recorded power state (%s) of node '%s' does not" 4094 " match actual power state (%s)"), node.powered, 4095 node.name, powered) 4096 4097 # For configuration changing commands we should update the node 4098 if self.op.command in (constants.OOB_POWER_ON, 4099 constants.OOB_POWER_OFF): 4100 self.cfg.Update(node, feedback_fn) 4101 4102 node_entry.append((constants.RS_NORMAL, result.payload)) 4103 4104 if (self.op.command == constants.OOB_POWER_ON and 4105 idx < len(self.nodes) - 1): 4106 time.sleep(self.op.power_delay) 4107 4108 return ret
4109
4110 - def _CheckPayload(self, result):
4111 """Checks if the payload is valid. 4112 4113 @param result: RPC result 4114 @raises errors.OpExecError: If payload is not valid 4115 4116 """ 4117 errs = [] 4118 if self.op.command == constants.OOB_HEALTH: 4119 if not isinstance(result.payload, list): 4120 errs.append("command 'health' is expected to return a list but got %s" % 4121 type(result.payload)) 4122 else: 4123 for item, status in result.payload: 4124 if status not in constants.OOB_STATUSES: 4125 errs.append("health item '%s' has invalid status '%s'" % 4126 (item, status)) 4127 4128 if self.op.command == constants.OOB_POWER_STATUS: 4129 if not isinstance(result.payload, dict): 4130 errs.append("power-status is expected to return a dict but got %s" % 4131 type(result.payload)) 4132 4133 if self.op.command in [ 4134 constants.OOB_POWER_ON, 4135 constants.OOB_POWER_OFF, 4136 constants.OOB_POWER_CYCLE, 4137 ]: 4138 if result.payload is not None: 4139 errs.append("%s is expected to not return payload but got '%s'" % 4140 (self.op.command, result.payload)) 4141 4142 if errs: 4143 raise errors.OpExecError("Check of out-of-band payload failed due to %s" % 4144 utils.CommaJoin(errs))
4145
4146 4147 -class _OsQuery(_QueryBase):
4148 FIELDS = query.OS_FIELDS 4149
4150 - def ExpandNames(self, lu):
4151 # Lock all nodes in shared mode 4152 # Temporary removal of locks, should be reverted later 4153 # TODO: reintroduce locks when they are lighter-weight 4154 lu.needed_locks = {} 4155 #self.share_locks[locking.LEVEL_NODE] = 1 4156 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 4157 4158 # The following variables interact with _QueryBase._GetNames 4159 if self.names: 4160 self.wanted = self.names 4161 else: 4162 self.wanted = locking.ALL_SET 4163 4164 self.do_locking = self.use_locking
4165
4166 - def DeclareLocks(self, lu, level):
4167 pass
4168 4169 @staticmethod
4170 - def _DiagnoseByOS(rlist):
4171 """Remaps a per-node return list into an a per-os per-node dictionary 4172 4173 @param rlist: a map with node names as keys and OS objects as values 4174 4175 @rtype: dict 4176 @return: a dictionary with osnames as keys and as value another 4177 map, with nodes as keys and tuples of (path, status, diagnose, 4178 variants, parameters, api_versions) as values, eg:: 4179 4180 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []), 4181 (/srv/..., False, "invalid api")], 4182 "node2": [(/srv/..., True, "", [], [])]} 4183 } 4184 4185 """ 4186 all_os = {} 4187 # we build here the list of nodes that didn't fail the RPC (at RPC 4188 # level), so that nodes with a non-responding node daemon don't 4189 # make all OSes invalid 4190 good_nodes = [node_name for node_name in rlist 4191 if not rlist[node_name].fail_msg] 4192 for node_name, nr in rlist.items(): 4193 if nr.fail_msg or not nr.payload: 4194 continue 4195 for (name, path, status, diagnose, variants, 4196 params, api_versions) in nr.payload: 4197 if name not in all_os: 4198 # build a list of nodes for this os containing empty lists 4199 # for each node in node_list 4200 all_os[name] = {} 4201 for nname in good_nodes: 4202 all_os[name][nname] = [] 4203 # convert params from [name, help] to (name, help) 4204 params = [tuple(v) for v in params] 4205 all_os[name][node_name].append((path, status, diagnose, 4206 variants, params, api_versions)) 4207 return all_os
4208
4209 - def _GetQueryData(self, lu):
4210 """Computes the list of nodes and their attributes. 4211 4212 """ 4213 # Locking is not used 4214 assert not (compat.any(lu.glm.is_owned(level) 4215 for level in locking.LEVELS 4216 if level != locking.LEVEL_CLUSTER) or 4217 self.do_locking or self.use_locking) 4218 4219 valid_nodes = [node.name 4220 for node in lu.cfg.GetAllNodesInfo().values() 4221 if not node.offline and node.vm_capable] 4222 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes)) 4223 cluster = lu.cfg.GetClusterInfo() 4224 4225 data = {} 4226 4227 for (os_name, os_data) in pol.items(): 4228 info = query.OsInfo(name=os_name, valid=True, node_status=os_data, 4229 hidden=(os_name in cluster.hidden_os), 4230 blacklisted=(os_name in cluster.blacklisted_os)) 4231 4232 variants = set() 4233 parameters = set() 4234 api_versions = set() 4235 4236 for idx, osl in enumerate(os_data.values()): 4237 info.valid = bool(info.valid and osl and osl[0][1]) 4238 if not info.valid: 4239 break 4240 4241 (node_variants, node_params, node_api) = osl[0][3:6] 4242 if idx == 0: 4243 # First entry 4244 variants.update(node_variants) 4245 parameters.update(node_params) 4246 api_versions.update(node_api) 4247 else: 4248 # Filter out inconsistent values 4249 variants.intersection_update(node_variants) 4250 parameters.intersection_update(node_params) 4251 api_versions.intersection_update(node_api) 4252 4253 info.variants = list(variants) 4254 info.parameters = list(parameters) 4255 info.api_versions = list(api_versions) 4256 4257 data[os_name] = info 4258 4259 # Prepare data in requested order 4260 return [data[name] for name in self._GetNames(lu, pol.keys(), None) 4261 if name in data]
4262
4263 4264 -class LUOsDiagnose(NoHooksLU):
4265 """Logical unit for OS diagnose/query. 4266 4267 """ 4268 REQ_BGL = False 4269 4270 @staticmethod
4271 - def _BuildFilter(fields, names):
4272 """Builds a filter for querying OSes. 4273 4274 """ 4275 name_filter = qlang.MakeSimpleFilter("name", names) 4276 4277 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the 4278 # respective field is not requested 4279 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]] 4280 for fname in ["hidden", "blacklisted"] 4281 if fname not in fields] 4282 if "valid" not in fields: 4283 status_filter.append([qlang.OP_TRUE, "valid"]) 4284 4285 if status_filter: 4286 status_filter.insert(0, qlang.OP_AND) 4287 else: 4288 status_filter = None 4289 4290 if name_filter and status_filter: 4291 return [qlang.OP_AND, name_filter, status_filter] 4292 elif name_filter: 4293 return name_filter 4294 else: 4295 return status_filter
4296
4297 - def CheckArguments(self):
4298 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names), 4299 self.op.output_fields, False)
4300
4301 - def ExpandNames(self):
4302 self.oq.ExpandNames(self)
4303
4304 - def Exec(self, feedback_fn):
4305 return self.oq.OldStyleQuery(self)
4306
4307 4308 -class LUNodeRemove(LogicalUnit):
4309 """Logical unit for removing a node. 4310 4311 """ 4312 HPATH = "node-remove" 4313 HTYPE = constants.HTYPE_NODE 4314
4315 - def BuildHooksEnv(self):
4316 """Build hooks env. 4317 4318 This doesn't run on the target node in the pre phase as a failed 4319 node would then be impossible to remove. 4320 4321 """ 4322 return { 4323 "OP_TARGET": self.op.node_name, 4324 "NODE_NAME": self.op.node_name, 4325 }
4326
4327 - def BuildHooksNodes(self):
4328 """Build hooks nodes. 4329 4330 """ 4331 all_nodes = self.cfg.GetNodeList() 4332 try: 4333 all_nodes.remove(self.op.node_name) 4334 except ValueError: 4335 logging.warning("Node '%s', which is about to be removed, was not found" 4336 " in the list of all nodes", self.op.node_name) 4337 return (all_nodes, all_nodes)
4338
4339 - def CheckPrereq(self):
4340 """Check prerequisites. 4341 4342 This checks: 4343 - the node exists in the configuration 4344 - it does not have primary or secondary instances 4345 - it's not the master 4346 4347 Any errors are signaled by raising errors.OpPrereqError. 4348 4349 """ 4350 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 4351 node = self.cfg.GetNodeInfo(self.op.node_name) 4352 assert node is not None 4353 4354 masternode = self.cfg.GetMasterNode() 4355 if node.name == masternode: 4356 raise errors.OpPrereqError("Node is the master node, failover to another" 4357 " node is required", errors.ECODE_INVAL) 4358 4359 for instance_name, instance in self.cfg.GetAllInstancesInfo().items(): 4360 if node.name in instance.all_nodes: 4361 raise errors.OpPrereqError("Instance %s is still running on the node," 4362 " please remove first" % instance_name, 4363 errors.ECODE_INVAL) 4364 self.op.node_name = node.name 4365 self.node = node
4366
4367 - def Exec(self, feedback_fn):
4368 """Removes the node from the cluster. 4369 4370 """ 4371 node = self.node 4372 logging.info("Stopping the node daemon and removing configs from node %s", 4373 node.name) 4374 4375 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup 4376 4377 # Promote nodes to master candidate as needed 4378 _AdjustCandidatePool(self, exceptions=[node.name]) 4379 self.context.RemoveNode(node.name) 4380 4381 # Run post hooks on the node before it's removed 4382 _RunPostHook(self, node.name) 4383 4384 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup) 4385 msg = result.fail_msg 4386 if msg: 4387 self.LogWarning("Errors encountered on the remote node while leaving" 4388 " the cluster: %s", msg) 4389 4390 # Remove node from our /etc/hosts 4391 if self.cfg.GetClusterInfo().modify_etc_hosts: 4392 master_node = self.cfg.GetMasterNode() 4393 result = self.rpc.call_etc_hosts_modify(master_node, 4394 constants.ETC_HOSTS_REMOVE, 4395 node.name, None) 4396 result.Raise("Can't update hosts file with new host data") 4397 _RedistributeAncillaryFiles(self)
4398
4399 4400 -class _NodeQuery(_QueryBase):
4401 FIELDS = query.NODE_FIELDS 4402
4403 - def ExpandNames(self, lu):
4404 lu.needed_locks = {} 4405 lu.share_locks = _ShareAll() 4406 4407 if self.names: 4408 self.wanted = _GetWantedNodes(lu, self.names) 4409 else: 4410 self.wanted = locking.ALL_SET 4411 4412 self.do_locking = (self.use_locking and 4413 query.NQ_LIVE in self.requested_data) 4414 4415 if self.do_locking: 4416 # If any non-static field is requested we need to lock the nodes 4417 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
4418
4419 - def DeclareLocks(self, lu, level):
4420 pass
4421
4422 - def _GetQueryData(self, lu):
4423 """Computes the list of nodes and their attributes. 4424 4425 """ 4426 all_info = lu.cfg.GetAllNodesInfo() 4427 4428 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE) 4429 4430 # Gather data as requested 4431 if query.NQ_LIVE in self.requested_data: 4432 # filter out non-vm_capable nodes 4433 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable] 4434 4435 node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(), 4436 lu.cfg.GetHypervisorType()) 4437 live_data = dict((name, nresult.payload) 4438 for (name, nresult) in node_data.items() 4439 if not nresult.fail_msg and nresult.payload) 4440 else: 4441 live_data = None 4442 4443 if query.NQ_INST in self.requested_data: 4444 node_to_primary = dict([(name, set()) for name in nodenames]) 4445 node_to_secondary = dict([(name, set()) for name in nodenames]) 4446 4447 inst_data = lu.cfg.GetAllInstancesInfo() 4448 4449 for inst in inst_data.values(): 4450 if inst.primary_node in node_to_primary: 4451 node_to_primary[inst.primary_node].add(inst.name) 4452 for secnode in inst.secondary_nodes: 4453 if secnode in node_to_secondary: 4454 node_to_secondary[secnode].add(inst.name) 4455 else: 4456 node_to_primary = None 4457 node_to_secondary = None 4458 4459 if query.NQ_OOB in self.requested_data: 4460 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node))) 4461 for name, node in all_info.iteritems()) 4462 else: 4463 oob_support = None 4464 4465 if query.NQ_GROUP in self.requested_data: 4466 groups = lu.cfg.GetAllNodeGroupsInfo() 4467 else: 4468 groups = {} 4469 4470 return query.NodeQueryData([all_info[name] for name in nodenames], 4471 live_data, lu.cfg.GetMasterNode(), 4472 node_to_primary, node_to_secondary, groups, 4473 oob_support, lu.cfg.GetClusterInfo())
4474
4475 4476 -class LUNodeQuery(NoHooksLU):
4477 """Logical unit for querying nodes. 4478 4479 """ 4480 # pylint: disable=W0142 4481 REQ_BGL = False 4482
4483 - def CheckArguments(self):
4484 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names), 4485 self.op.output_fields, self.op.use_locking)
4486
4487 - def ExpandNames(self):
4488 self.nq.ExpandNames(self)
4489
4490 - def Exec(self, feedback_fn):
4491 return self.nq.OldStyleQuery(self)
4492
4493 4494 -class LUNodeQueryvols(NoHooksLU):
4495 """Logical unit for getting volumes on node(s). 4496 4497 """ 4498 REQ_BGL = False 4499 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance") 4500 _FIELDS_STATIC = utils.FieldSet("node") 4501
4502 - def CheckArguments(self):
4503 _CheckOutputFields(static=self._FIELDS_STATIC, 4504 dynamic=self._FIELDS_DYNAMIC, 4505 selected=self.op.output_fields)
4506
4507 - def ExpandNames(self):
4508 self.needed_locks = {} 4509 self.share_locks[locking.LEVEL_NODE] = 1 4510 if not self.op.nodes: 4511 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 4512 else: 4513 self.needed_locks[locking.LEVEL_NODE] = \ 4514 _GetWantedNodes(self, self.op.nodes)
4515
4516 - def Exec(self, feedback_fn):
4517 """Computes the list of nodes and their attributes. 4518 4519 """ 4520 nodenames = self.owned_locks(locking.LEVEL_NODE) 4521 volumes = self.rpc.call_node_volumes(nodenames) 4522 4523 ilist = self.cfg.GetAllInstancesInfo() 4524 vol2inst = _MapInstanceDisksToNodes(ilist.values()) 4525 4526 output = [] 4527 for node in nodenames: 4528 nresult = volumes[node] 4529 if nresult.offline: 4530 continue 4531 msg = nresult.fail_msg 4532 if msg: 4533 self.LogWarning("Can't compute volume data on node %s: %s", node, msg) 4534 continue 4535 4536 node_vols = sorted(nresult.payload, 4537 key=operator.itemgetter("dev")) 4538 4539 for vol in node_vols: 4540 node_output = [] 4541 for field in self.op.output_fields: 4542 if field == "node": 4543 val = node 4544 elif field == "phys": 4545 val = vol["dev"] 4546 elif field == "vg": 4547 val = vol["vg"] 4548 elif field == "name": 4549 val = vol["name"] 4550 elif field == "size": 4551 val = int(float(vol["size"])) 4552 elif field == "instance": 4553 val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-") 4554 else: 4555 raise errors.ParameterError(field) 4556 node_output.append(str(val)) 4557 4558 output.append(node_output) 4559 4560 return output
4561
4562 4563 -class LUNodeQueryStorage(NoHooksLU):
4564 """Logical unit for getting information on storage units on node(s). 4565 4566 """ 4567 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE) 4568 REQ_BGL = False 4569
4570 - def CheckArguments(self):
4571 _CheckOutputFields(static=self._FIELDS_STATIC, 4572 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS), 4573 selected=self.op.output_fields)
4574
4575 - def ExpandNames(self):
4576 self.needed_locks = {} 4577 self.share_locks[locking.LEVEL_NODE] = 1 4578 4579 if self.op.nodes: 4580 self.needed_locks[locking.LEVEL_NODE] = \ 4581 _GetWantedNodes(self, self.op.nodes) 4582 else: 4583 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4584
4585 - def Exec(self, feedback_fn):
4586 """Computes the list of nodes and their attributes. 4587 4588 """ 4589 self.nodes = self.owned_locks(locking.LEVEL_NODE) 4590 4591 # Always get name to sort by 4592 if constants.SF_NAME in self.op.output_fields: 4593 fields = self.op.output_fields[:] 4594 else: 4595 fields = [constants.SF_NAME] + self.op.output_fields 4596 4597 # Never ask for node or type as it's only known to the LU 4598 for extra in [constants.SF_NODE, constants.SF_TYPE]: 4599 while extra in fields: 4600 fields.remove(extra) 4601 4602 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)]) 4603 name_idx = field_idx[constants.SF_NAME] 4604 4605 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) 4606 data = self.rpc.call_storage_list(self.nodes, 4607 self.op.storage_type, st_args, 4608 self.op.name, fields) 4609 4610 result = [] 4611 4612 for node in utils.NiceSort(self.nodes): 4613 nresult = data[node] 4614 if nresult.offline: 4615 continue 4616 4617 msg = nresult.fail_msg 4618 if msg: 4619 self.LogWarning("Can't get storage data from node %s: %s", node, msg) 4620 continue 4621 4622 rows = dict([(row[name_idx], row) for row in nresult.payload]) 4623 4624 for name in utils.NiceSort(rows.keys()): 4625 row = rows[name] 4626 4627 out = [] 4628 4629 for field in self.op.output_fields: 4630 if field == constants.SF_NODE: 4631 val = node 4632 elif field == constants.SF_TYPE: 4633 val = self.op.storage_type 4634 elif field in field_idx: 4635 val = row[field_idx[field]] 4636 else: 4637 raise errors.ParameterError(field) 4638 4639 out.append(val) 4640 4641 result.append(out) 4642 4643 return result
4644
4645 4646 -class _InstanceQuery(_QueryBase):
4647 FIELDS = query.INSTANCE_FIELDS 4648
4649 - def ExpandNames(self, lu):
4650 lu.needed_locks = {} 4651 lu.share_locks = _ShareAll() 4652 4653 if self.names: 4654 self.wanted = _GetWantedInstances(lu, self.names) 4655 else: 4656 self.wanted = locking.ALL_SET 4657 4658 self.do_locking = (self.use_locking and 4659 query.IQ_LIVE in self.requested_data) 4660 if self.do_locking: 4661 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted 4662 lu.needed_locks[locking.LEVEL_NODEGROUP] = [] 4663 lu.needed_locks[locking.LEVEL_NODE] = [] 4664 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE 4665 4666 self.do_grouplocks = (self.do_locking and 4667 query.IQ_NODES in self.requested_data)
4668
4669 - def DeclareLocks(self, lu, level):
4670 if self.do_locking: 4671 if level == locking.LEVEL_NODEGROUP and self.do_grouplocks: 4672 assert not lu.needed_locks[locking.LEVEL_NODEGROUP] 4673 4674 # Lock all groups used by instances optimistically; this requires going 4675 # via the node before it's locked, requiring verification later on 4676 lu.needed_locks[locking.LEVEL_NODEGROUP] = \ 4677 set(group_uuid 4678 for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE) 4679 for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name)) 4680 elif level == locking.LEVEL_NODE: 4681 lu._LockInstancesNodes() # pylint: disable=W0212
4682 4683 @staticmethod
4684 - def _CheckGroupLocks(lu):
4685 owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE)) 4686 owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP)) 4687 4688 # Check if node groups for locked instances are still correct 4689 for instance_name in owned_instances: 4690 _CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups)
4691
4692 - def _GetQueryData(self, lu):
4693 """Computes the list of instances and their attributes. 4694 4695 """ 4696 if self.do_grouplocks: 4697 self._CheckGroupLocks(lu) 4698 4699 cluster = lu.cfg.GetClusterInfo() 4700 all_info = lu.cfg.GetAllInstancesInfo() 4701 4702 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE) 4703 4704 instance_list = [all_info[name] for name in instance_names] 4705 nodes = frozenset(itertools.chain(*(inst.all_nodes 4706 for inst in instance_list))) 4707 hv_list = list(set([inst.hypervisor for inst in instance_list])) 4708 bad_nodes = [] 4709 offline_nodes = [] 4710 wrongnode_inst = set() 4711 4712 # Gather data as requested 4713 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]): 4714 live_data = {} 4715 node_data = lu.rpc.call_all_instances_info(nodes, hv_list) 4716 for name in nodes: 4717 result = node_data[name] 4718 if result.offline: 4719 # offline nodes will be in both lists 4720 assert result.fail_msg 4721 offline_nodes.append(name) 4722 if result.fail_msg: 4723 bad_nodes.append(name) 4724 elif result.payload: 4725 for inst in result.payload: 4726 if inst in all_info: 4727 if all_info[inst].primary_node == name: 4728 live_data.update(result.payload) 4729 else: 4730 wrongnode_inst.add(inst) 4731 else: 4732 # orphan instance; we don't list it here as we don't 4733 # handle this case yet in the output of instance listing 4734 logging.warning("Orphan instance '%s' found on node %s", 4735 inst, name) 4736 # else no instance is alive 4737 else: 4738 live_data = {} 4739 4740 if query.IQ_DISKUSAGE in self.requested_data: 4741 disk_usage = dict((inst.name, 4742 _ComputeDiskSize(inst.disk_template, 4743 [{constants.IDISK_SIZE: disk.size} 4744 for disk in inst.disks])) 4745 for inst in instance_list) 4746 else: 4747 disk_usage = None 4748 4749 if query.IQ_CONSOLE in self.requested_data: 4750 consinfo = {} 4751 for inst in instance_list: 4752 if inst.name in live_data: 4753 # Instance is running 4754 consinfo[inst.name] = _GetInstanceConsole(cluster, inst) 4755 else: 4756 consinfo[inst.name] = None 4757 assert set(consinfo.keys()) == set(instance_names) 4758 else: 4759 consinfo = None 4760 4761 if query.IQ_NODES in self.requested_data: 4762 node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"), 4763 instance_list))) 4764 nodes = dict(lu.cfg.GetMultiNodeInfo(node_names)) 4765 groups = dict((uuid, lu.cfg.GetNodeGroup(uuid)) 4766 for uuid in set(map(operator.attrgetter("group"), 4767 nodes.values()))) 4768 else: 4769 nodes = None 4770 groups = None 4771 4772 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(), 4773 disk_usage, offline_nodes, bad_nodes, 4774 live_data, wrongnode_inst, consinfo, 4775 nodes, groups)
4776
4777 4778 -class LUQuery(NoHooksLU):
4779 """Query for resources/items of a certain kind. 4780 4781 """ 4782 # pylint: disable=W0142 4783 REQ_BGL = False 4784
4785 - def CheckArguments(self):
4786 qcls = _GetQueryImplementation(self.op.what) 4787 4788 self.impl = qcls(self.op.filter, self.op.fields, self.op.use_locking)
4789
4790 - def ExpandNames(self):
4791 self.impl.ExpandNames(self)
4792
4793 - def DeclareLocks(self, level):
4794 self.impl.DeclareLocks(self, level)
4795
4796 - def Exec(self, feedback_fn):
4797 return self.impl.NewStyleQuery(self)
4798
4799 4800 -class LUQueryFields(NoHooksLU):
4801 """Query for resources/items of a certain kind. 4802 4803 """ 4804 # pylint: disable=W0142 4805 REQ_BGL = False 4806
4807 - def CheckArguments(self):
4808 self.qcls = _GetQueryImplementation(self.op.what)
4809
4810 - def ExpandNames(self):
4811 self.needed_locks = {}
4812
4813 - def Exec(self, feedback_fn):
4814 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
4815
4816 4817 -class LUNodeModifyStorage(NoHooksLU):
4818 """Logical unit for modifying a storage volume on a node. 4819 4820 """ 4821 REQ_BGL = False 4822
4823 - def CheckArguments(self):
4824 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 4825 4826 storage_type = self.op.storage_type 4827 4828 try: 4829 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type] 4830 except KeyError: 4831 raise errors.OpPrereqError("Storage units of type '%s' can not be" 4832 " modified" % storage_type, 4833 errors.ECODE_INVAL) 4834 4835 diff = set(self.op.changes.keys()) - modifiable 4836 if diff: 4837 raise errors.OpPrereqError("The following fields can not be modified for" 4838 " storage units of type '%s': %r" % 4839 (storage_type, list(diff)), 4840 errors.ECODE_INVAL)
4841
4842 - def ExpandNames(self):
4843 self.needed_locks = { 4844 locking.LEVEL_NODE: self.op.node_name, 4845 }
4846
4847 - def Exec(self, feedback_fn):
4848 """Computes the list of nodes and their attributes. 4849 4850 """ 4851 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) 4852 result = self.rpc.call_storage_modify(self.op.node_name, 4853 self.op.storage_type, st_args, 4854 self.op.name, self.op.changes) 4855 result.Raise("Failed to modify storage unit '%s' on %s" % 4856 (self.op.name, self.op.node_name))
4857
4858 4859 -class LUNodeAdd(LogicalUnit):
4860 """Logical unit for adding node to the cluster. 4861 4862 """ 4863 HPATH = "node-add" 4864 HTYPE = constants.HTYPE_NODE 4865 _NFLAGS = ["master_capable", "vm_capable"] 4866
4867 - def CheckArguments(self):
4868 self.primary_ip_family = self.cfg.GetPrimaryIPFamily() 4869 # validate/normalize the node name 4870 self.hostname = netutils.GetHostname(name=self.op.node_name, 4871 family=self.primary_ip_family) 4872 self.op.node_name = self.hostname.name 4873 4874 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode(): 4875 raise errors.OpPrereqError("Cannot readd the master node", 4876 errors.ECODE_STATE) 4877 4878 if self.op.readd and self.op.group: 4879 raise errors.OpPrereqError("Cannot pass a node group when a node is" 4880 " being readded", errors.ECODE_INVAL)
4881
4882 - def BuildHooksEnv(self):
4883 """Build hooks env. 4884 4885 This will run on all nodes before, and on all nodes + the new node after. 4886 4887 """ 4888 return { 4889 "OP_TARGET": self.op.node_name, 4890 "NODE_NAME": self.op.node_name, 4891 "NODE_PIP": self.op.primary_ip, 4892 "NODE_SIP": self.op.secondary_ip, 4893 "MASTER_CAPABLE": str(self.op.master_capable), 4894 "VM_CAPABLE": str(self.op.vm_capable), 4895 }
4896
4897 - def BuildHooksNodes(self):
4898 """Build hooks nodes. 4899 4900 """ 4901 # Exclude added node 4902 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name])) 4903 post_nodes = pre_nodes + [self.op.node_name, ] 4904 4905 return (pre_nodes, post_nodes)
4906
4907 - def CheckPrereq(self):
4908 """Check prerequisites. 4909 4910 This checks: 4911 - the new node is not already in the config 4912 - it is resolvable 4913 - its parameters (single/dual homed) matches the cluster 4914 4915 Any errors are signaled by raising errors.OpPrereqError. 4916 4917 """ 4918 cfg = self.cfg 4919 hostname = self.hostname 4920 node = hostname.name 4921 primary_ip = self.op.primary_ip = hostname.ip 4922 if self.op.secondary_ip is None: 4923 if self.primary_ip_family == netutils.IP6Address.family: 4924 raise errors.OpPrereqError("When using a IPv6 primary address, a valid" 4925 " IPv4 address must be given as secondary", 4926 errors.ECODE_INVAL) 4927 self.op.secondary_ip = primary_ip 4928 4929 secondary_ip = self.op.secondary_ip 4930 if not netutils.IP4Address.IsValid(secondary_ip): 4931 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4" 4932 " address" % secondary_ip, errors.ECODE_INVAL) 4933 4934 node_list = cfg.GetNodeList() 4935 if not self.op.readd and node in node_list: 4936 raise errors.OpPrereqError("Node %s is already in the configuration" % 4937 node, errors.ECODE_EXISTS) 4938 elif self.op.readd and node not in node_list: 4939 raise errors.OpPrereqError("Node %s is not in the configuration" % node, 4940 errors.ECODE_NOENT) 4941 4942 self.changed_primary_ip = False 4943 4944 for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list): 4945 if self.op.readd and node == existing_node_name: 4946 if existing_node.secondary_ip != secondary_ip: 4947 raise errors.OpPrereqError("Readded node doesn't have the same IP" 4948 " address configuration as before", 4949 errors.ECODE_INVAL) 4950 if existing_node.primary_ip != primary_ip: 4951 self.changed_primary_ip = True 4952 4953 continue 4954 4955 if (existing_node.primary_ip == primary_ip or 4956 existing_node.secondary_ip == primary_ip or 4957 existing_node.primary_ip == secondary_ip or 4958 existing_node.secondary_ip == secondary_ip): 4959 raise errors.OpPrereqError("New node ip address(es) conflict with" 4960 " existing node %s" % existing_node.name, 4961 errors.ECODE_NOTUNIQUE) 4962 4963 # After this 'if' block, None is no longer a valid value for the 4964 # _capable op attributes 4965 if self.op.readd: 4966 old_node = self.cfg.GetNodeInfo(node) 4967 assert old_node is not None, "Can't retrieve locked node %s" % node 4968 for attr in self._NFLAGS: 4969 if getattr(self.op, attr) is None: 4970 setattr(self.op, attr, getattr(old_node, attr)) 4971 else: 4972 for attr in self._NFLAGS: 4973 if getattr(self.op, attr) is None: 4974 setattr(self.op, attr, True) 4975 4976 if self.op.readd and not self.op.vm_capable: 4977 pri, sec = cfg.GetNodeInstances(node) 4978 if pri or sec: 4979 raise errors.OpPrereqError("Node %s being re-added with vm_capable" 4980 " flag set to false, but it already holds" 4981 " instances" % node, 4982 errors.ECODE_STATE) 4983 4984 # check that the type of the node (single versus dual homed) is the 4985 # same as for the master 4986 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode()) 4987 master_singlehomed = myself.secondary_ip == myself.primary_ip 4988 newbie_singlehomed = secondary_ip == primary_ip 4989 if master_singlehomed != newbie_singlehomed: 4990 if master_singlehomed: 4991 raise errors.OpPrereqError("The master has no secondary ip but the" 4992 " new node has one", 4993 errors.ECODE_INVAL) 4994 else: 4995 raise errors.OpPrereqError("The master has a secondary ip but the" 4996 " new node doesn't have one", 4997 errors.ECODE_INVAL) 4998 4999 # checks reachability 5000 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT): 5001 raise errors.OpPrereqError("Node not reachable by ping", 5002 errors.ECODE_ENVIRON) 5003 5004 if not newbie_singlehomed: 5005 # check reachability from my secondary ip to newbie's secondary ip 5006 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, 5007 source=myself.secondary_ip): 5008 raise errors.OpPrereqError("Node secondary ip not reachable by TCP" 5009 " based ping to node daemon port", 5010 errors.ECODE_ENVIRON) 5011 5012 if self.op.readd: 5013 exceptions = [node] 5014 else: 5015 exceptions = [] 5016 5017 if self.op.master_capable: 5018 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions) 5019 else: 5020 self.master_candidate = False 5021 5022 if self.op.readd: 5023 self.new_node = old_node 5024 else: 5025 node_group = cfg.LookupNodeGroup(self.op.group) 5026 self.new_node = objects.Node(name=node, 5027 primary_ip=primary_ip, 5028 secondary_ip=secondary_ip, 5029 master_candidate=self.master_candidate, 5030 offline=False, drained=False, 5031 group=node_group) 5032 5033 if self.op.ndparams: 5034 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES) 5035 5036 # check connectivity 5037 result = self.rpc.call_version([self.new_node.name])[self.new_node.name] 5038 result.Raise("Can't get version information from node %s" % node) 5039 if constants.PROTOCOL_VERSION == result.payload: 5040 logging.info("Communication to node %s fine, sw version %s match", 5041 node, result.payload) 5042 else: 5043 raise errors.OpPrereqError("Version mismatch master version %s," 5044 " node version %s" % 5045 (constants.PROTOCOL_VERSION, result.payload), 5046 errors.ECODE_ENVIRON)
5047
5048 - def Exec(self, feedback_fn):
5049 """Adds the new node to the cluster. 5050 5051 """ 5052 new_node = self.new_node 5053 node = new_node.name 5054 5055 # We adding a new node so we assume it's powered 5056 new_node.powered = True 5057 5058 # for re-adds, reset the offline/drained/master-candidate flags; 5059 # we need to reset here, otherwise offline would prevent RPC calls 5060 # later in the procedure; this also means that if the re-add 5061 # fails, we are left with a non-offlined, broken node 5062 if self.op.readd: 5063 new_node.drained = new_node.offline = False # pylint: disable=W0201 5064 self.LogInfo("Readding a node, the offline/drained flags were reset") 5065 # if we demote the node, we do cleanup later in the procedure 5066 new_node.master_candidate = self.master_candidate 5067 if self.changed_primary_ip: 5068 new_node.primary_ip = self.op.primary_ip 5069 5070 # copy the master/vm_capable flags 5071 for attr in self._NFLAGS: 5072 setattr(new_node, attr, getattr(self.op, attr)) 5073 5074 # notify the user about any possible mc promotion 5075 if new_node.master_candidate: 5076 self.LogInfo("Node will be a master candidate") 5077 5078 if self.op.ndparams: 5079 new_node.ndparams = self.op.ndparams 5080 else: 5081 new_node.ndparams = {} 5082 5083 # Add node to our /etc/hosts, and add key to known_hosts 5084 if self.cfg.GetClusterInfo().modify_etc_hosts: 5085 master_node = self.cfg.GetMasterNode() 5086 result = self.rpc.call_etc_hosts_modify(master_node, 5087 constants.ETC_HOSTS_ADD, 5088 self.hostname.name, 5089 self.hostname.ip) 5090 result.Raise("Can't update hosts file with new host data") 5091 5092 if new_node.secondary_ip != new_node.primary_ip: 5093 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip, 5094 False) 5095 5096 node_verify_list = [self.cfg.GetMasterNode()] 5097 node_verify_param = { 5098 constants.NV_NODELIST: ([node], {}), 5099 # TODO: do a node-net-test as well? 5100 } 5101 5102 result = self.rpc.call_node_verify(node_verify_list, node_verify_param, 5103 self.cfg.GetClusterName()) 5104 for verifier in node_verify_list: 5105 result[verifier].Raise("Cannot communicate with node %s" % verifier) 5106 nl_payload = result[verifier].payload[constants.NV_NODELIST] 5107 if nl_payload: 5108 for failed in nl_payload: 5109 feedback_fn("ssh/hostname verification failed" 5110 " (checking from %s): %s" % 5111 (verifier, nl_payload[failed])) 5112 raise errors.OpExecError("ssh/hostname verification failed") 5113 5114 if self.op.readd: 5115 _RedistributeAncillaryFiles(self) 5116 self.context.ReaddNode(new_node) 5117 # make sure we redistribute the config 5118 self.cfg.Update(new_node, feedback_fn) 5119 # and make sure the new node will not have old files around 5120 if not new_node.master_candidate: 5121 result = self.rpc.call_node_demote_from_mc(new_node.name) 5122 msg = result.fail_msg 5123 if msg: 5124 self.LogWarning("Node failed to demote itself from master" 5125 " candidate status: %s" % msg) 5126 else: 5127 _RedistributeAncillaryFiles(self, additional_nodes=[node], 5128 additional_vm=self.op.vm_capable) 5129 self.context.AddNode(new_node, self.proc.GetECId())
5130
5131 5132 -class LUNodeSetParams(LogicalUnit):
5133 """Modifies the parameters of a node. 5134 5135 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline) 5136 to the node role (as _ROLE_*) 5137 @cvar _R2F: a dictionary from node role to tuples of flags 5138 @cvar _FLAGS: a list of attribute names corresponding to the flags 5139 5140 """ 5141 HPATH = "node-modify" 5142 HTYPE = constants.HTYPE_NODE 5143 REQ_BGL = False 5144 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4) 5145 _F2R = { 5146 (True, False, False): _ROLE_CANDIDATE, 5147 (False, True, False): _ROLE_DRAINED, 5148 (False, False, True): _ROLE_OFFLINE, 5149 (False, False, False): _ROLE_REGULAR, 5150 } 5151 _R2F = dict((v, k) for k, v in _F2R.items()) 5152 _FLAGS = ["master_candidate", "drained", "offline"] 5153
5154 - def CheckArguments(self):
5155 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 5156 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained, 5157 self.op.master_capable, self.op.vm_capable, 5158 self.op.secondary_ip, self.op.ndparams] 5159 if all_mods.count(None) == len(all_mods): 5160 raise errors.OpPrereqError("Please pass at least one modification", 5161 errors.ECODE_INVAL) 5162 if all_mods.count(True) > 1: 5163 raise errors.OpPrereqError("Can't set the node into more than one" 5164 " state at the same time", 5165 errors.ECODE_INVAL) 5166 5167 # Boolean value that tells us whether we might be demoting from MC 5168 self.might_demote = (self.op.master_candidate == False or 5169 self.op.offline == True or 5170 self.op.drained == True or 5171 self.op.master_capable == False) 5172 5173 if self.op.secondary_ip: 5174 if not netutils.IP4Address.IsValid(self.op.secondary_ip): 5175 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4" 5176 " address" % self.op.secondary_ip, 5177 errors.ECODE_INVAL) 5178 5179 self.lock_all = self.op.auto_promote and self.might_demote 5180 self.lock_instances = self.op.secondary_ip is not None
5181
5182 - def ExpandNames(self):
5183 if self.lock_all: 5184 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET} 5185 else: 5186 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name} 5187 5188 if self.lock_instances: 5189 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5190
5191 - def DeclareLocks(self, level):
5192 # If we have locked all instances, before waiting to lock nodes, release 5193 # all the ones living on nodes unrelated to the current operation. 5194 if level == locking.LEVEL_NODE and self.lock_instances: 5195 self.affected_instances = [] 5196 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET: 5197 instances_keep = [] 5198 5199 # Build list of instances to release 5200 locked_i = self.owned_locks(locking.LEVEL_INSTANCE) 5201 for instance_name, instance in self.cfg.GetMultiInstanceInfo(locked_i): 5202 if (instance.disk_template in constants.DTS_INT_MIRROR and 5203 self.op.node_name in instance.all_nodes): 5204 instances_keep.append(instance_name) 5205 self.affected_instances.append(instance) 5206 5207 _ReleaseLocks(self, locking.LEVEL_INSTANCE, keep=instances_keep) 5208 5209 assert (set(self.owned_locks(locking.LEVEL_INSTANCE)) == 5210 set(instances_keep))
5211
5212 - def BuildHooksEnv(self):
5213 """Build hooks env. 5214 5215 This runs on the master node. 5216 5217 """ 5218 return { 5219 "OP_TARGET": self.op.node_name, 5220 "MASTER_CANDIDATE": str(self.op.master_candidate), 5221 "OFFLINE": str(self.op.offline), 5222 "DRAINED": str(self.op.drained), 5223 "MASTER_CAPABLE": str(self.op.master_capable), 5224 "VM_CAPABLE": str(self.op.vm_capable), 5225 }
5226
5227 - def BuildHooksNodes(self):
5228 """Build hooks nodes. 5229 5230 """ 5231 nl = [self.cfg.GetMasterNode(), self.op.node_name] 5232 return (nl, nl)
5233
5234 - def CheckPrereq(self):
5235 """Check prerequisites. 5236 5237 This only checks the instance list against the existing names. 5238 5239 """ 5240 node = self.node = self.cfg.GetNodeInfo(self.op.node_name) 5241 5242 if (self.op.master_candidate is not None or 5243 self.op.drained is not None or 5244 self.op.offline is not None): 5245 # we can't change the master's node flags 5246 if self.op.node_name == self.cfg.GetMasterNode(): 5247 raise errors.OpPrereqError("The master role can be changed" 5248 " only via master-failover", 5249 errors.ECODE_INVAL) 5250 5251 if self.op.master_candidate and not node.master_capable: 5252 raise errors.OpPrereqError("Node %s is not master capable, cannot make" 5253 " it a master candidate" % node.name, 5254 errors.ECODE_STATE) 5255 5256 if self.op.vm_capable == False: 5257 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name) 5258 if ipri or isec: 5259 raise errors.OpPrereqError("Node %s hosts instances, cannot unset" 5260 " the vm_capable flag" % node.name, 5261 errors.ECODE_STATE) 5262 5263 if node.master_candidate and self.might_demote and not self.lock_all: 5264 assert not self.op.auto_promote, "auto_promote set but lock_all not" 5265 # check if after removing the current node, we're missing master 5266 # candidates 5267 (mc_remaining, mc_should, _) = \ 5268 self.cfg.GetMasterCandidateStats(exceptions=[node.name]) 5269 if mc_remaining < mc_should: 5270 raise errors.OpPrereqError("Not enough master candidates, please" 5271 " pass auto promote option to allow" 5272 " promotion", errors.ECODE_STATE) 5273 5274 self.old_flags = old_flags = (node.master_candidate, 5275 node.drained, node.offline) 5276 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags) 5277 self.old_role = old_role = self._F2R[old_flags] 5278 5279 # Check for ineffective changes 5280 for attr in self._FLAGS: 5281 if (getattr(self.op, attr) == False and getattr(node, attr) == False): 5282 self.LogInfo("Ignoring request to unset flag %s, already unset", attr) 5283 setattr(self.op, attr, None) 5284 5285 # Past this point, any flag change to False means a transition 5286 # away from the respective state, as only real changes are kept 5287 5288 # TODO: We might query the real power state if it supports OOB 5289 if _SupportsOob(self.cfg, node): 5290 if self.op.offline is False and not (node.powered or 5291 self.op.powered == True): 5292 raise errors.OpPrereqError(("Node %s needs to be turned on before its" 5293 " offline status can be reset") % 5294 self.op.node_name) 5295 elif self.op.powered is not None: 5296 raise errors.OpPrereqError(("Unable to change powered state for node %s" 5297 " as it does not support out-of-band" 5298 " handling") % self.op.node_name) 5299 5300 # If we're being deofflined/drained, we'll MC ourself if needed 5301 if (self.op.drained == False or self.op.offline == False or 5302 (self.op.master_capable and not node.master_capable)): 5303 if _DecideSelfPromotion(self): 5304 self.op.master_candidate = True 5305 self.LogInfo("Auto-promoting node to master candidate") 5306 5307 # If we're no longer master capable, we'll demote ourselves from MC 5308 if self.op.master_capable == False and node.master_candidate: 5309 self.LogInfo("Demoting from master candidate") 5310 self.op.master_candidate = False 5311 5312 # Compute new role 5313 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1 5314 if self.op.master_candidate: 5315 new_role = self._ROLE_CANDIDATE 5316 elif self.op.drained: 5317 new_role = self._ROLE_DRAINED 5318 elif self.op.offline: 5319 new_role = self._ROLE_OFFLINE 5320 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]: 5321 # False is still in new flags, which means we're un-setting (the 5322 # only) True flag 5323 new_role = self._ROLE_REGULAR 5324 else: # no new flags, nothing, keep old role 5325 new_role = old_role 5326 5327 self.new_role = new_role 5328 5329 if old_role == self._ROLE_OFFLINE and new_role != old_role: 5330 # Trying to transition out of offline status 5331 result = self.rpc.call_version([node.name])[node.name] 5332 if result.fail_msg: 5333 raise errors.OpPrereqError("Node %s is being de-offlined but fails" 5334 " to report its version: %s" % 5335 (node.name, result.fail_msg), 5336 errors.ECODE_STATE) 5337 else: 5338 self.LogWarning("Transitioning node from offline to online state" 5339 " without using re-add. Please make sure the node" 5340 " is healthy!") 5341 5342 if self.op.secondary_ip: 5343 # Ok even without locking, because this can't be changed by any LU 5344 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode()) 5345 master_singlehomed = master.secondary_ip == master.primary_ip 5346 if master_singlehomed and self.op.secondary_ip: 5347 raise errors.OpPrereqError("Cannot change the secondary ip on a single" 5348 " homed cluster", errors.ECODE_INVAL) 5349 5350 if node.offline: 5351 if self.affected_instances: 5352 raise errors.OpPrereqError("Cannot change secondary ip: offline" 5353 " node has instances (%s) configured" 5354 " to use it" % self.affected_instances) 5355 else: 5356 # On online nodes, check that no instances are running, and that 5357 # the node has the new ip and we can reach it. 5358 for instance in self.affected_instances: 5359 _CheckInstanceDown(self, instance, "cannot change secondary ip") 5360 5361 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True) 5362 if master.name != node.name: 5363 # check reachability from master secondary ip to new secondary ip 5364 if not netutils.TcpPing(self.op.secondary_ip, 5365 constants.DEFAULT_NODED_PORT, 5366 source=master.secondary_ip): 5367 raise errors.OpPrereqError("Node secondary ip not reachable by TCP" 5368 " based ping to node daemon port", 5369 errors.ECODE_ENVIRON) 5370 5371 if self.op.ndparams: 5372 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams) 5373 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES) 5374 self.new_ndparams = new_ndparams
5375
5376 - def Exec(self, feedback_fn):
5377 """Modifies a node. 5378 5379 """ 5380 node = self.node 5381 old_role = self.old_role 5382 new_role = self.new_role 5383 5384 result = [] 5385 5386 if self.op.ndparams: 5387 node.ndparams = self.new_ndparams 5388 5389 if self.op.powered is not None: 5390 node.powered = self.op.powered 5391 5392 for attr in ["master_capable", "vm_capable"]: 5393 val = getattr(self.op, attr) 5394 if val is not None: 5395 setattr(node, attr, val) 5396 result.append((attr, str(val))) 5397 5398 if new_role != old_role: 5399 # Tell the node to demote itself, if no longer MC and not offline 5400 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE: 5401 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg 5402 if msg: 5403 self.LogWarning("Node failed to demote itself: %s", msg) 5404 5405 new_flags = self._R2F[new_role] 5406 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS): 5407 if of != nf: 5408 result.append((desc, str(nf))) 5409 (node.master_candidate, node.drained, node.offline) = new_flags 5410 5411 # we locked all nodes, we adjust the CP before updating this node 5412 if self.lock_all: 5413 _AdjustCandidatePool(self, [node.name]) 5414 5415 if self.op.secondary_ip: 5416 node.secondary_ip = self.op.secondary_ip 5417 result.append(("secondary_ip", self.op.secondary_ip)) 5418 5419 # this will trigger configuration file update, if needed 5420 self.cfg.Update(node, feedback_fn) 5421 5422 # this will trigger job queue propagation or cleanup if the mc 5423 # flag changed 5424 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1: 5425 self.context.ReaddNode(node) 5426 5427 return result
5428
5429 5430 -class LUNodePowercycle(NoHooksLU):
5431 """Powercycles a node. 5432 5433 """ 5434 REQ_BGL = False 5435
5436 - def CheckArguments(self):
5437 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 5438 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force: 5439 raise errors.OpPrereqError("The node is the master and the force" 5440 " parameter was not set", 5441 errors.ECODE_INVAL)
5442
5443 - def ExpandNames(self):
5444 """Locking for PowercycleNode. 5445 5446 This is a last-resort option and shouldn't block on other 5447 jobs. Therefore, we grab no locks. 5448 5449 """ 5450 self.needed_locks = {}
5451
5452 - def Exec(self, feedback_fn):
5453 """Reboots a node. 5454 5455 """ 5456 result = self.rpc.call_node_powercycle(self.op.node_name, 5457 self.cfg.GetHypervisorType()) 5458 result.Raise("Failed to schedule the reboot") 5459 return result.payload
5460
5461 5462 -class LUClusterQuery(NoHooksLU):
5463 """Query cluster configuration. 5464 5465 """ 5466 REQ_BGL = False 5467
5468 - def ExpandNames(self):
5469 self.needed_locks = {}
5470
5471 - def Exec(self, feedback_fn):
5472 """Return cluster config. 5473 5474 """ 5475 cluster = self.cfg.GetClusterInfo() 5476 os_hvp = {} 5477 5478 # Filter just for enabled hypervisors 5479 for os_name, hv_dict in cluster.os_hvp.items(): 5480 os_hvp[os_name] = {} 5481 for hv_name, hv_params in hv_dict.items(): 5482 if hv_name in cluster.enabled_hypervisors: 5483 os_hvp[os_name][hv_name] = hv_params 5484 5485 # Convert ip_family to ip_version 5486 primary_ip_version = constants.IP4_VERSION 5487 if cluster.primary_ip_family == netutils.IP6Address.family: 5488 primary_ip_version = constants.IP6_VERSION 5489 5490 result = { 5491 "software_version": constants.RELEASE_VERSION, 5492 "protocol_version": constants.PROTOCOL_VERSION, 5493 "config_version": constants.CONFIG_VERSION, 5494 "os_api_version": max(constants.OS_API_VERSIONS), 5495 "export_version": constants.EXPORT_VERSION, 5496 "architecture": (platform.architecture()[0], platform.machine()), 5497 "name": cluster.cluster_name, 5498 "master": cluster.master_node, 5499 "default_hypervisor": cluster.enabled_hypervisors[0], 5500 "enabled_hypervisors": cluster.enabled_hypervisors, 5501 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name]) 5502 for hypervisor_name in cluster.enabled_hypervisors]), 5503 "os_hvp": os_hvp, 5504 "beparams": cluster.beparams, 5505 "osparams": cluster.osparams, 5506 "nicparams": cluster.nicparams, 5507 "ndparams": cluster.ndparams, 5508 "candidate_pool_size": cluster.candidate_pool_size, 5509 "master_netdev": cluster.master_netdev, 5510 "volume_group_name": cluster.volume_group_name, 5511 "drbd_usermode_helper": cluster.drbd_usermode_helper, 5512 "file_storage_dir": cluster.file_storage_dir, 5513 "shared_file_storage_dir": cluster.shared_file_storage_dir, 5514 "maintain_node_health": cluster.maintain_node_health, 5515 "ctime": cluster.ctime, 5516 "mtime": cluster.mtime, 5517 "uuid": cluster.uuid, 5518 "tags": list(cluster.GetTags()), 5519 "uid_pool": cluster.uid_pool, 5520 "default_iallocator": cluster.default_iallocator, 5521 "reserved_lvs": cluster.reserved_lvs, 5522 "primary_ip_version": primary_ip_version, 5523 "prealloc_wipe_disks": cluster.prealloc_wipe_disks, 5524 "hidden_os": cluster.hidden_os, 5525 "blacklisted_os": cluster.blacklisted_os, 5526 } 5527 5528 return result
5529
5530 5531 -class LUClusterConfigQuery(NoHooksLU):
5532 """Return configuration values. 5533 5534 """ 5535 REQ_BGL = False 5536 _FIELDS_DYNAMIC = utils.FieldSet() 5537 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag", 5538 "watcher_pause", "volume_group_name") 5539
5540 - def CheckArguments(self):
5541 _CheckOutputFields(static=self._FIELDS_STATIC, 5542 dynamic=self._FIELDS_DYNAMIC, 5543 selected=self.op.output_fields)
5544
5545 - def ExpandNames(self):
5546 self.needed_locks = {}
5547
5548 - def Exec(self, feedback_fn):
5549 """Dump a representation of the cluster config to the standard output. 5550 5551 """ 5552 values = [] 5553 for field in self.op.output_fields: 5554 if field == "cluster_name": 5555 entry = self.cfg.GetClusterName() 5556 elif field == "master_node": 5557 entry = self.cfg.GetMasterNode() 5558 elif field == "drain_flag": 5559 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE) 5560 elif field == "watcher_pause": 5561 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE) 5562 elif field == "volume_group_name": 5563 entry = self.cfg.GetVGName() 5564 else: 5565 raise errors.ParameterError(field) 5566 values.append(entry) 5567 return values
5568
5569 5570 -class LUInstanceActivateDisks(NoHooksLU):
5571 """Bring up an instance's disks. 5572 5573 """ 5574 REQ_BGL = False 5575
5576 - def ExpandNames(self):
5577 self._ExpandAndLockInstance() 5578 self.needed_locks[locking.LEVEL_NODE] = [] 5579 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5580
5581 - def DeclareLocks(self, level):
5582 if level == locking.LEVEL_NODE: 5583 self._LockInstancesNodes()
5584
5585 - def CheckPrereq(self):
5586 """Check prerequisites. 5587 5588 This checks that the instance is in the cluster. 5589 5590 """ 5591 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 5592 assert self.instance is not None, \ 5593 "Cannot retrieve locked instance %s" % self.op.instance_name 5594 _CheckNodeOnline(self, self.instance.primary_node)
5595
5596 - def Exec(self, feedback_fn):
5597 """Activate the disks. 5598 5599 """ 5600 disks_ok, disks_info = \ 5601 _AssembleInstanceDisks(self, self.instance, 5602 ignore_size=self.op.ignore_size) 5603 if not disks_ok: 5604 raise errors.OpExecError("Cannot activate block devices") 5605 5606 return disks_info
5607
5608 5609 -def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False, 5610 ignore_size=False):
5611 """Prepare the block devices for an instance. 5612 5613 This sets up the block devices on all nodes. 5614 5615 @type lu: L{LogicalUnit} 5616 @param lu: the logical unit on whose behalf we execute 5617 @type instance: L{objects.Instance} 5618 @param instance: the instance for whose disks we assemble 5619 @type disks: list of L{objects.Disk} or None 5620 @param disks: which disks to assemble (or all, if None) 5621 @type ignore_secondaries: boolean 5622 @param ignore_secondaries: if true, errors on secondary nodes 5623 won't result in an error return from the function 5624 @type ignore_size: boolean 5625 @param ignore_size: if true, the current known size of the disk 5626 will not be used during the disk activation, useful for cases 5627 when the size is wrong 5628 @return: False if the operation failed, otherwise a list of 5629 (host, instance_visible_name, node_visible_name) 5630 with the mapping from node devices to instance devices 5631 5632 """ 5633 device_info = [] 5634 disks_ok = True 5635 iname = instance.name 5636 disks = _ExpandCheckDisks(instance, disks) 5637 5638 # With the two passes mechanism we try to reduce the window of 5639 # opportunity for the race condition of switching DRBD to primary 5640 # before handshaking occured, but we do not eliminate it 5641 5642 # The proper fix would be to wait (with some limits) until the 5643 # connection has been made and drbd transitions from WFConnection 5644 # into any other network-connected state (Connected, SyncTarget, 5645 # SyncSource, etc.) 5646 5647 # 1st pass, assemble on all nodes in secondary mode 5648 for idx, inst_disk in enumerate(disks): 5649 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): 5650 if ignore_size: 5651 node_disk = node_disk.Copy() 5652 node_disk.UnsetSize() 5653 lu.cfg.SetDiskID(node_disk, node) 5654 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx) 5655 msg = result.fail_msg 5656 if msg: 5657 lu.proc.LogWarning("Could not prepare block device %s on node %s" 5658 " (is_primary=False, pass=1): %s", 5659 inst_disk.iv_name, node, msg) 5660 if not ignore_secondaries: 5661 disks_ok = False 5662 5663 # FIXME: race condition on drbd migration to primary 5664 5665 # 2nd pass, do only the primary node 5666 for idx, inst_disk in enumerate(disks): 5667 dev_path = None 5668 5669 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): 5670 if node != instance.primary_node: 5671 continue 5672 if ignore_size: 5673 node_disk = node_disk.Copy() 5674 node_disk.UnsetSize() 5675 lu.cfg.SetDiskID(node_disk, node) 5676 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx) 5677 msg = result.fail_msg 5678 if msg: 5679 lu.proc.LogWarning("Could not prepare block device %s on node %s" 5680 " (is_primary=True, pass=2): %s", 5681 inst_disk.iv_name, node, msg) 5682 disks_ok = False 5683 else: 5684 dev_path = result.payload 5685 5686 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path)) 5687 5688 # leave the disks configured for the primary node 5689 # this is a workaround that would be fixed better by 5690 # improving the logical/physical id handling 5691 for disk in disks: 5692 lu.cfg.SetDiskID(disk, instance.primary_node) 5693 5694 return disks_ok, device_info
5695
5696 5697 -def _StartInstanceDisks(lu, instance, force):
5698 """Start the disks of an instance. 5699 5700 """ 5701 disks_ok, _ = _AssembleInstanceDisks(lu, instance, 5702 ignore_secondaries=force) 5703 if not disks_ok: 5704 _ShutdownInstanceDisks(lu, instance) 5705 if force is not None and not force: 5706 lu.proc.LogWarning("", hint="If the message above refers to a" 5707 " secondary node," 5708 " you can retry the operation using '--force'.") 5709 raise errors.OpExecError("Disk consistency error")
5710
5711 5712 -class LUInstanceDeactivateDisks(NoHooksLU):
5713 """Shutdown an instance's disks. 5714 5715 """ 5716 REQ_BGL = False 5717
5718 - def ExpandNames(self):
5719 self._ExpandAndLockInstance() 5720 self.needed_locks[locking.LEVEL_NODE] = [] 5721 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5722
5723 - def DeclareLocks(self, level):
5724 if level == locking.LEVEL_NODE: 5725 self._LockInstancesNodes()
5726
5727 - def CheckPrereq(self):
5728 """Check prerequisites. 5729 5730 This checks that the instance is in the cluster. 5731 5732 """ 5733 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 5734 assert self.instance is not None, \ 5735 "Cannot retrieve locked instance %s" % self.op.instance_name
5736
5737 - def Exec(self, feedback_fn):
5738 """Deactivate the disks 5739 5740 """ 5741 instance = self.instance 5742 if self.op.force: 5743 _ShutdownInstanceDisks(self, instance) 5744 else: 5745 _SafeShutdownInstanceDisks(self, instance)
5746
5747 5748 -def _SafeShutdownInstanceDisks(lu, instance, disks=None):
5749 """Shutdown block devices of an instance. 5750 5751 This function checks if an instance is running, before calling 5752 _ShutdownInstanceDisks. 5753 5754 """ 5755 _CheckInstanceDown(lu, instance, "cannot shutdown disks") 5756 _ShutdownInstanceDisks(lu, instance, disks=disks)
5757
5758 5759 -def _ExpandCheckDisks(instance, disks):
5760 """Return the instance disks selected by the disks list 5761 5762 @type disks: list of L{objects.Disk} or None 5763 @param disks: selected disks 5764 @rtype: list of L{objects.Disk} 5765 @return: selected instance disks to act on 5766 5767 """ 5768 if disks is None: 5769 return instance.disks 5770 else: 5771 if not set(disks).issubset(instance.disks): 5772 raise errors.ProgrammerError("Can only act on disks belonging to the" 5773 " target instance") 5774 return disks
5775
5776 5777 -def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
5778 """Shutdown block devices of an instance. 5779 5780 This does the shutdown on all nodes of the instance. 5781 5782 If the ignore_primary is false, errors on the primary node are 5783 ignored. 5784 5785 """ 5786 all_result = True 5787 disks = _ExpandCheckDisks(instance, disks) 5788 5789 for disk in disks: 5790 for node, top_disk in disk.ComputeNodeTree(instance.primary_node): 5791 lu.cfg.SetDiskID(top_disk, node) 5792 result = lu.rpc.call_blockdev_shutdown(node, top_disk) 5793 msg = result.fail_msg 5794 if msg: 5795 lu.LogWarning("Could not shutdown block device %s on node %s: %s", 5796 disk.iv_name, node, msg) 5797 if ((node == instance.primary_node and not ignore_primary) or 5798 (node != instance.primary_node and not result.offline)): 5799 all_result = False 5800 return all_result
5801
5802 5803 -def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
5804 """Checks if a node has enough free memory. 5805 5806 This function check if a given node has the needed amount of free 5807 memory. In case the node has less memory or we cannot get the 5808 information from the node, this function raise an OpPrereqError 5809 exception. 5810 5811 @type lu: C{LogicalUnit} 5812 @param lu: a logical unit from which we get configuration data 5813 @type node: C{str} 5814 @param node: the node to check 5815 @type reason: C{str} 5816 @param reason: string to use in the error message 5817 @type requested: C{int} 5818 @param requested: the amount of memory in MiB to check for 5819 @type hypervisor_name: C{str} 5820 @param hypervisor_name: the hypervisor to ask for memory stats 5821 @raise errors.OpPrereqError: if the node doesn't have enough memory, or 5822 we cannot check the node 5823 5824 """ 5825 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name) 5826 nodeinfo[node].Raise("Can't get data from node %s" % node, 5827 prereq=True, ecode=errors.ECODE_ENVIRON) 5828 free_mem = nodeinfo[node].payload.get("memory_free", None) 5829 if not isinstance(free_mem, int): 5830 raise errors.OpPrereqError("Can't compute free memory on node %s, result" 5831 " was '%s'" % (node, free_mem), 5832 errors.ECODE_ENVIRON) 5833 if requested > free_mem: 5834 raise errors.OpPrereqError("Not enough memory on node %s for %s:" 5835 " needed %s MiB, available %s MiB" % 5836 (node, reason, requested, free_mem), 5837 errors.ECODE_NORES)
5838
5839 5840 -def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5841 """Checks if nodes have enough free disk space in the all VGs. 5842 5843 This function check if all given nodes have the needed amount of 5844 free disk. In case any node has less disk or we cannot get the 5845 information from the node, this function raise an OpPrereqError 5846 exception. 5847 5848 @type lu: C{LogicalUnit} 5849 @param lu: a logical unit from which we get configuration data 5850 @type nodenames: C{list} 5851 @param nodenames: the list of node names to check 5852 @type req_sizes: C{dict} 5853 @param req_sizes: the hash of vg and corresponding amount of disk in 5854 MiB to check for 5855 @raise errors.OpPrereqError: if the node doesn't have enough disk, 5856 or we cannot check the node 5857 5858 """ 5859 for vg, req_size in req_sizes.items(): 5860 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5861
5862 5863 -def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5864 """Checks if nodes have enough free disk space in the specified VG. 5865 5866 This function check if all given nodes have the needed amount of 5867 free disk. In case any node has less disk or we cannot get the 5868 information from the node, this function raise an OpPrereqError 5869 exception. 5870 5871 @type lu: C{LogicalUnit} 5872 @param lu: a logical unit from which we get configuration data 5873 @type nodenames: C{list} 5874 @param nodenames: the list of node names to check 5875 @type vg: C{str} 5876 @param vg: the volume group to check 5877 @type requested: C{int} 5878 @param requested: the amount of disk in MiB to check for 5879 @raise errors.OpPrereqError: if the node doesn't have enough disk, 5880 or we cannot check the node 5881 5882 """ 5883 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None) 5884 for node in nodenames: 5885 info = nodeinfo[node] 5886 info.Raise("Cannot get current information from node %s" % node, 5887 prereq=True, ecode=errors.ECODE_ENVIRON) 5888 vg_free = info.payload.get("vg_free", None) 5889 if not isinstance(vg_free, int): 5890 raise errors.OpPrereqError("Can't compute free disk space on node" 5891 " %s for vg %s, result was '%s'" % 5892 (node, vg, vg_free), errors.ECODE_ENVIRON) 5893 if requested > vg_free: 5894 raise errors.OpPrereqError("Not enough disk space on target node %s" 5895 " vg %s: required %d MiB, available %d MiB" % 5896 (node, vg, requested, vg_free), 5897 errors.ECODE_NORES)
5898
5899 5900 -class LUInstanceStartup(LogicalUnit):
5901 """Starts an instance. 5902 5903 """ 5904 HPATH = "instance-start" 5905 HTYPE = constants.HTYPE_INSTANCE 5906 REQ_BGL = False 5907
5908 - def CheckArguments(self):
5909 # extra beparams 5910 if self.op.beparams: 5911 # fill the beparams dict 5912 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5913
5914 - def ExpandNames(self):
5916
5917 - def BuildHooksEnv(self):
5918 """Build hooks env. 5919 5920 This runs on master, primary and secondary nodes of the instance. 5921 5922 """ 5923 env = { 5924 "FORCE": self.op.force, 5925 } 5926 5927 env.update(_BuildInstanceHookEnvByObject(self, self.instance)) 5928 5929 return env
5930
5931 - def BuildHooksNodes(self):
5932 """Build hooks nodes. 5933 5934 """ 5935 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 5936 return (nl, nl)
5937
5938 - def CheckPrereq(self):
5939 """Check prerequisites. 5940 5941 This checks that the instance is in the cluster. 5942 5943 """ 5944 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) 5945 assert self.instance is not None, \ 5946 "Cannot retrieve locked instance %s" % self.op.instance_name 5947 5948 # extra hvparams 5949 if self.op.hvparams: 5950 # check hypervisor parameter syntax (locally) 5951 cluster = self.cfg.GetClusterInfo() 5952 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES) 5953 filled_hvp = cluster.FillHV(instance) 5954 filled_hvp.update(self.op.hvparams) 5955 hv_type = hypervisor.GetHypervisor(instance.hypervisor) 5956 hv_type.CheckParameterSyntax(filled_hvp) 5957 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp) 5958 5959 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline 5960 5961 if self.primary_offline and self.op.ignore_offline_nodes: 5962 self.proc.LogWarning("Ignoring offline primary node") 5963 5964 if self.op.hvparams or self.op.beparams: 5965 self.proc.LogWarning("Overridden parameters are ignored") 5966 else: 5967 _CheckNodeOnline(self, instance.primary_node) 5968 5969 bep = self.cfg.GetClusterInfo().FillBE(instance) 5970 5971 # check bridges existence 5972 _CheckInstanceBridgesExist(self, instance) 5973 5974 remote_info = self.rpc.call_instance_info(instance.primary_node, 5975 instance.name, 5976 instance.hypervisor) 5977 remote_info.Raise("Error checking node %s" % instance.primary_node, 5978 prereq=True, ecode=errors.ECODE_ENVIRON) 5979 if not remote_info.payload: # not running already 5980 _CheckNodeFreeMemory(self, instance.primary_node, 5981 "starting instance %s" % instance.name, 5982 bep[constants.BE_MEMORY], instance.hypervisor)
5983
5984 - def Exec(self, feedback_fn):
5985 """Start the instance. 5986 5987 """ 5988 instance = self.instance 5989 force = self.op.force 5990 5991 if not self.op.no_remember: 5992 self.cfg.MarkInstanceUp(instance.name) 5993 5994 if self.primary_offline: 5995 assert self.op.ignore_offline_nodes 5996 self.proc.LogInfo("Primary node offline, marked instance as started") 5997 else: 5998 node_current = instance.primary_node 5999 6000 _StartInstanceDisks(self, instance, force) 6001 6002 result = self.rpc.call_instance_start(node_current, instance, 6003 self.op.hvparams, self.op.beparams, 6004 self.op.startup_paused) 6005 msg = result.fail_msg 6006 if msg: 6007 _ShutdownInstanceDisks(self, instance) 6008 raise errors.OpExecError("Could not start instance: %s" % msg)
6009
6010 6011 -class LUInstanceReboot(LogicalUnit):
6012 """Reboot an instance. 6013 6014 """ 6015 HPATH = "instance-reboot" 6016 HTYPE = constants.HTYPE_INSTANCE 6017 REQ_BGL = False 6018
6019 - def ExpandNames(self):
6021
6022 - def BuildHooksEnv(self):
6023 """Build hooks env. 6024 6025 This runs on master, primary and secondary nodes of the instance. 6026 6027 """ 6028 env = { 6029 "IGNORE_SECONDARIES": self.op.ignore_secondaries, 6030 "REBOOT_TYPE": self.op.reboot_type, 6031 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, 6032 } 6033 6034 env.update(_BuildInstanceHookEnvByObject(self, self.instance)) 6035 6036 return env
6037
6038 - def BuildHooksNodes(self):
6039 """Build hooks nodes. 6040 6041 """ 6042 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 6043 return (nl, nl)
6044
6045 - def CheckPrereq(self):
6046 """Check prerequisites. 6047 6048 This checks that the instance is in the cluster. 6049 6050 """ 6051 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6052 assert self.instance is not None, \ 6053 "Cannot retrieve locked instance %s" % self.op.instance_name 6054 6055 _CheckNodeOnline(self, instance.primary_node) 6056 6057 # check bridges existence 6058 _CheckInstanceBridgesExist(self, instance)
6059
6060 - def Exec(self, feedback_fn):
6061 """Reboot the instance. 6062 6063 """ 6064 instance = self.instance 6065 ignore_secondaries = self.op.ignore_secondaries 6066 reboot_type = self.op.reboot_type 6067 6068 remote_info = self.rpc.call_instance_info(instance.primary_node, 6069 instance.name, 6070 instance.hypervisor) 6071 remote_info.Raise("Error checking node %s" % instance.primary_node) 6072 instance_running = bool(remote_info.payload) 6073 6074 node_current = instance.primary_node 6075 6076 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT, 6077 constants.INSTANCE_REBOOT_HARD]: 6078 for disk in instance.disks: 6079 self.cfg.SetDiskID(disk, node_current) 6080 result = self.rpc.call_instance_reboot(node_current, instance, 6081 reboot_type, 6082 self.op.shutdown_timeout) 6083 result.Raise("Could not reboot instance") 6084 else: 6085 if instance_running: 6086 result = self.rpc.call_instance_shutdown(node_current, instance, 6087 self.op.shutdown_timeout) 6088 result.Raise("Could not shutdown instance for full reboot") 6089 _ShutdownInstanceDisks(self, instance) 6090 else: 6091 self.LogInfo("Instance %s was already stopped, starting now", 6092 instance.name) 6093 _StartInstanceDisks(self, instance, ignore_secondaries) 6094 result = self.rpc.call_instance_start(node_current, instance, 6095 None, None, False) 6096 msg = result.fail_msg 6097 if msg: 6098 _ShutdownInstanceDisks(self, instance) 6099 raise errors.OpExecError("Could not start instance for" 6100 " full reboot: %s" % msg) 6101 6102 self.cfg.MarkInstanceUp(instance.name)
6103
6104 6105 -class LUInstanceShutdown(LogicalUnit):
6106 """Shutdown an instance. 6107 6108 """ 6109 HPATH = "instance-stop" 6110 HTYPE = constants.HTYPE_INSTANCE 6111 REQ_BGL = False 6112
6113 - def ExpandNames(self):
6115
6116 - def BuildHooksEnv(self):
6117 """Build hooks env. 6118 6119 This runs on master, primary and secondary nodes of the instance. 6120 6121 """ 6122 env = _BuildInstanceHookEnvByObject(self, self.instance) 6123 env["TIMEOUT"] = self.op.timeout 6124 return env
6125
6126 - def BuildHooksNodes(self):
6127 """Build hooks nodes. 6128 6129 """ 6130 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 6131 return (nl, nl)
6132
6133 - def CheckPrereq(self):
6134 """Check prerequisites. 6135 6136 This checks that the instance is in the cluster. 6137 6138 """ 6139 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6140 assert self.instance is not None, \ 6141 "Cannot retrieve locked instance %s" % self.op.instance_name 6142 6143 self.primary_offline = \ 6144 self.cfg.GetNodeInfo(self.instance.primary_node).offline 6145 6146 if self.primary_offline and self.op.ignore_offline_nodes: 6147 self.proc.LogWarning("Ignoring offline primary node") 6148 else: 6149 _CheckNodeOnline(self, self.instance.primary_node)
6150
6151 - def Exec(self, feedback_fn):
6152 """Shutdown the instance. 6153 6154 """ 6155 instance = self.instance 6156 node_current = instance.primary_node 6157 timeout = self.op.timeout 6158 6159 if not self.op.no_remember: 6160 self.cfg.MarkInstanceDown(instance.name) 6161 6162 if self.primary_offline: 6163 assert self.op.ignore_offline_nodes 6164 self.proc.LogInfo("Primary node offline, marked instance as stopped") 6165 else: 6166 result = self.rpc.call_instance_shutdown(node_current, instance, timeout) 6167 msg = result.fail_msg 6168 if msg: 6169 self.proc.LogWarning("Could not shutdown instance: %s" % msg) 6170 6171 _ShutdownInstanceDisks(self, instance)
6172
6173 6174 -class LUInstanceReinstall(LogicalUnit):
6175 """Reinstall an instance. 6176 6177 """ 6178 HPATH = "instance-reinstall" 6179 HTYPE = constants.HTYPE_INSTANCE 6180 REQ_BGL = False 6181
6182 - def ExpandNames(self):
6184
6185 - def BuildHooksEnv(self):
6186 """Build hooks env. 6187 6188 This runs on master, primary and secondary nodes of the instance. 6189 6190 """ 6191 return _BuildInstanceHookEnvByObject(self, self.instance)
6192
6193 - def BuildHooksNodes(self):
6194 """Build hooks nodes. 6195 6196 """ 6197 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 6198 return (nl, nl)
6199
6200 - def CheckPrereq(self):
6201 """Check prerequisites. 6202 6203 This checks that the instance is in the cluster and is not running. 6204 6205 """ 6206 instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6207 assert instance is not None, \ 6208 "Cannot retrieve locked instance %s" % self.op.instance_name 6209 _CheckNodeOnline(self, instance.primary_node, "Instance primary node" 6210 " offline, cannot reinstall") 6211 for node in instance.secondary_nodes: 6212 _CheckNodeOnline(self, node, "Instance secondary node offline," 6213 " cannot reinstall") 6214 6215 if instance.disk_template == constants.DT_DISKLESS: 6216 raise errors.OpPrereqError("Instance '%s' has no disks" % 6217 self.op.instance_name, 6218 errors.ECODE_INVAL) 6219 _CheckInstanceDown(self, instance, "cannot reinstall") 6220 6221 if self.op.os_type is not None: 6222 # OS verification 6223 pnode = _ExpandNodeName(self.cfg, instance.primary_node) 6224 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant) 6225 instance_os = self.op.os_type 6226 else: 6227 instance_os = instance.os 6228 6229 nodelist = list(instance.all_nodes) 6230 6231 if self.op.osparams: 6232 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams) 6233 _CheckOSParams(self, True, nodelist, instance_os, i_osdict) 6234 self.os_inst = i_osdict # the new dict (without defaults) 6235 else: 6236 self.os_inst = None 6237 6238 self.instance = instance
6239
6240 - def Exec(self, feedback_fn):
6241 """Reinstall the instance. 6242 6243 """ 6244 inst = self.instance 6245 6246 if self.op.os_type is not None: 6247 feedback_fn("Changing OS to '%s'..." % self.op.os_type) 6248 inst.os = self.op.os_type 6249 # Write to configuration 6250 self.cfg.Update(inst, feedback_fn) 6251 6252 _StartInstanceDisks(self, inst, None) 6253 try: 6254 feedback_fn("Running the instance OS create scripts...") 6255 # FIXME: pass debug option from opcode to backend 6256 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True, 6257 self.op.debug_level, 6258 osparams=self.os_inst) 6259 result.Raise("Could not install OS for instance %s on node %s" % 6260 (inst.name, inst.primary_node)) 6261 finally: 6262 _ShutdownInstanceDisks(self, inst)
6263
6264 6265 -class LUInstanceRecreateDisks(LogicalUnit):
6266 """Recreate an instance's missing disks. 6267 6268 """ 6269 HPATH = "instance-recreate-disks" 6270 HTYPE = constants.HTYPE_INSTANCE 6271 REQ_BGL = False 6272
6273 - def CheckArguments(self):
6274 # normalise the disk list 6275 self.op.disks = sorted(frozenset(self.op.disks))
6276
6277 - def ExpandNames(self):
6278 self._ExpandAndLockInstance() 6279 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND 6280 if self.op.nodes: 6281 self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes] 6282 self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes) 6283 else: 6284 self.needed_locks[locking.LEVEL_NODE] = []
6285
6286 - def DeclareLocks(self, level):
6287 if level == locking.LEVEL_NODE: 6288 # if we replace the nodes, we only need to lock the old primary, 6289 # otherwise we need to lock all nodes for disk re-creation 6290 primary_only = bool(self.op.nodes) 6291 self._LockInstancesNodes(primary_only=primary_only)
6292
6293 - def BuildHooksEnv(self):
6294 """Build hooks env. 6295 6296 This runs on master, primary and secondary nodes of the instance. 6297 6298 """ 6299 return _BuildInstanceHookEnvByObject(self, self.instance)
6300
6301 - def BuildHooksNodes(self):
6302 """Build hooks nodes. 6303 6304 """ 6305 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 6306 return (nl, nl)
6307
6308 - def CheckPrereq(self):
6309 """Check prerequisites. 6310 6311 This checks that the instance is in the cluster and is not running. 6312 6313 """ 6314 instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6315 assert instance is not None, \ 6316 "Cannot retrieve locked instance %s" % self.op.instance_name 6317 if self.op.nodes: 6318 if len(self.op.nodes) != len(instance.all_nodes): 6319 raise errors.OpPrereqError("Instance %s currently has %d nodes, but" 6320 " %d replacement nodes were specified" % 6321 (instance.name, len(instance.all_nodes), 6322 len(self.op.nodes)), 6323 errors.ECODE_INVAL) 6324 assert instance.disk_template != constants.DT_DRBD8 or \ 6325 len(self.op.nodes) == 2 6326 assert instance.disk_template != constants.DT_PLAIN or \ 6327 len(self.op.nodes) == 1 6328 primary_node = self.op.nodes[0] 6329 else: 6330 primary_node = instance.primary_node 6331 _CheckNodeOnline(self, primary_node) 6332 6333 if instance.disk_template == constants.DT_DISKLESS: 6334 raise errors.OpPrereqError("Instance '%s' has no disks" % 6335 self.op.instance_name, errors.ECODE_INVAL) 6336 # if we replace nodes *and* the old primary is offline, we don't 6337 # check 6338 assert instance.primary_node in self.needed_locks[locking.LEVEL_NODE] 6339 old_pnode = self.cfg.GetNodeInfo(instance.primary_node) 6340 if not (self.op.nodes and old_pnode.offline): 6341 _CheckInstanceDown(self, instance, "cannot recreate disks") 6342 6343 if not self.op.disks: 6344 self.op.disks = range(len(instance.disks)) 6345 else: 6346 for idx in self.op.disks: 6347 if idx >= len(instance.disks): 6348 raise errors.OpPrereqError("Invalid disk index '%s'" % idx, 6349 errors.ECODE_INVAL) 6350 if self.op.disks != range(len(instance.disks)) and self.op.nodes: 6351 raise errors.OpPrereqError("Can't recreate disks partially and" 6352 " change the nodes at the same time", 6353 errors.ECODE_INVAL) 6354 self.instance = instance
6355
6356 - def Exec(self, feedback_fn):
6357 """Recreate the disks. 6358 6359 """ 6360 instance = self.instance 6361 6362 to_skip = [] 6363 mods = [] # keeps track of needed logical_id changes 6364 6365 for idx, disk in enumerate(instance.disks): 6366 if idx not in self.op.disks: # disk idx has not been passed in 6367 to_skip.append(idx) 6368 continue 6369 # update secondaries for disks, if needed 6370 if self.op.nodes: 6371 if disk.dev_type == constants.LD_DRBD8: 6372 # need to update the nodes and minors 6373 assert len(self.op.nodes) == 2 6374 assert len(disk.logical_id) == 6 # otherwise disk internals 6375 # have changed 6376 (_, _, old_port, _, _, old_secret) = disk.logical_id 6377 new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name) 6378 new_id = (self.op.nodes[0], self.op.nodes[1], old_port, 6379 new_minors[0], new_minors[1], old_secret) 6380 assert len(disk.logical_id) == len(new_id) 6381 mods.append((idx, new_id)) 6382 6383 # now that we have passed all asserts above, we can apply the mods 6384 # in a single run (to avoid partial changes) 6385 for idx, new_id in mods: 6386 instance.disks[idx].logical_id = new_id 6387 6388 # change primary node, if needed 6389 if self.op.nodes: 6390 instance.primary_node = self.op.nodes[0] 6391 self.LogWarning("Changing the instance's nodes, you will have to" 6392 " remove any disks left on the older nodes manually") 6393 6394 if self.op.nodes: 6395 self.cfg.Update(instance, feedback_fn) 6396 6397 _CreateDisks(self, instance, to_skip=to_skip)
6398
6399 6400 -class LUInstanceRename(LogicalUnit):
6401 """Rename an instance. 6402 6403 """ 6404 HPATH = "instance-rename" 6405 HTYPE = constants.HTYPE_INSTANCE 6406
6407 - def CheckArguments(self):
6408 """Check arguments. 6409 6410 """ 6411 if self.op.ip_check and not self.op.name_check: 6412 # TODO: make the ip check more flexible and not depend on the name check 6413 raise errors.OpPrereqError("IP address check requires a name check", 6414 errors.ECODE_INVAL)
6415
6416 - def BuildHooksEnv(self):
6417 """Build hooks env. 6418 6419 This runs on master, primary and secondary nodes of the instance. 6420 6421 """ 6422 env = _BuildInstanceHookEnvByObject(self, self.instance) 6423 env["INSTANCE_NEW_NAME"] = self.op.new_name 6424 return env
6425
6426 - def BuildHooksNodes(self):
6427 """Build hooks nodes. 6428 6429 """ 6430 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 6431 return (nl, nl)
6432
6433 - def CheckPrereq(self):
6434 """Check prerequisites. 6435 6436 This checks that the instance is in the cluster and is not running. 6437 6438 """ 6439 self.op.instance_name = _ExpandInstanceName(self.cfg, 6440 self.op.instance_name) 6441 instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6442 assert instance is not None 6443 _CheckNodeOnline(self, instance.primary_node) 6444 _CheckInstanceDown(self, instance, "cannot rename") 6445 self.instance = instance 6446 6447 new_name = self.op.new_name 6448 if self.op.name_check: 6449 hostname = netutils.GetHostname(name=new_name) 6450 if hostname.name != new_name: 6451 self.LogInfo("Resolved given name '%s' to '%s'", new_name, 6452 hostname.name) 6453 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]): 6454 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the" 6455 " same as given hostname '%s'") % 6456 (hostname.name, self.op.new_name), 6457 errors.ECODE_INVAL) 6458 new_name = self.op.new_name = hostname.name 6459 if (self.op.ip_check and 6460 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)): 6461 raise errors.OpPrereqError("IP %s of instance %s already in use" % 6462 (hostname.ip, new_name), 6463 errors.ECODE_NOTUNIQUE) 6464 6465 instance_list = self.cfg.GetInstanceList() 6466 if new_name in instance_list and new_name != instance.name: 6467 raise errors.OpPrereqError("Instance '%s' is already in the cluster" % 6468 new_name, errors.ECODE_EXISTS)
6469
6470 - def Exec(self, feedback_fn):
6471 """Rename the instance. 6472 6473 """ 6474 inst = self.instance 6475 old_name = inst.name 6476 6477 rename_file_storage = False 6478 if (inst.disk_template in constants.DTS_FILEBASED and 6479 self.op.new_name != inst.name): 6480 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) 6481 rename_file_storage = True 6482 6483 self.cfg.RenameInstance(inst.name, self.op.new_name) 6484 # Change the instance lock. This is definitely safe while we hold the BGL. 6485 # Otherwise the new lock would have to be added in acquired mode. 6486 assert self.REQ_BGL 6487 self.glm.remove(locking.LEVEL_INSTANCE, old_name) 6488 self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name) 6489 6490 # re-read the instance from the configuration after rename 6491 inst = self.cfg.GetInstanceInfo(self.op.new_name) 6492 6493 if rename_file_storage: 6494 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) 6495 result = self.rpc.call_file_storage_dir_rename(inst.primary_node, 6496 old_file_storage_dir, 6497 new_file_storage_dir) 6498 result.Raise("Could not rename on node %s directory '%s' to '%s'" 6499 " (but the instance has been renamed in Ganeti)" % 6500 (inst.primary_node, old_file_storage_dir, 6501 new_file_storage_dir)) 6502 6503 _StartInstanceDisks(self, inst, None) 6504 try: 6505 result = self.rpc.call_instance_run_rename(inst.primary_node, inst, 6506 old_name, self.op.debug_level) 6507 msg = result.fail_msg 6508 if msg: 6509 msg = ("Could not run OS rename script for instance %s on node %s" 6510 " (but the instance has been renamed in Ganeti): %s" % 6511 (inst.name, inst.primary_node, msg)) 6512 self.proc.LogWarning(msg) 6513 finally: 6514 _ShutdownInstanceDisks(self, inst) 6515 6516 return inst.name
6517
6518 6519 -class LUInstanceRemove(LogicalUnit):
6520 """Remove an instance. 6521 6522 """ 6523 HPATH = "instance-remove" 6524 HTYPE = constants.HTYPE_INSTANCE 6525 REQ_BGL = False 6526
6527 - def ExpandNames(self):
6528 self._ExpandAndLockInstance() 6529 self.needed_locks[locking.LEVEL_NODE] = [] 6530 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6531
6532 - def DeclareLocks(self, level):
6533 if level == locking.LEVEL_NODE: 6534 self._LockInstancesNodes()
6535
6536 - def BuildHooksEnv(self):
6537 """Build hooks env. 6538 6539 This runs on master, primary and secondary nodes of the instance. 6540 6541 """ 6542 env = _BuildInstanceHookEnvByObject(self, self.instance) 6543 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout 6544 return env
6545
6546 - def BuildHooksNodes(self):
6547 """Build hooks nodes. 6548 6549 """ 6550 nl = [self.cfg.GetMasterNode()] 6551 nl_post = list(self.instance.all_nodes) + nl 6552 return (nl, nl_post)
6553
6554 - def CheckPrereq(self):
6555 """Check prerequisites. 6556 6557 This checks that the instance is in the cluster. 6558 6559 """ 6560 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6561 assert self.instance is not None, \ 6562 "Cannot retrieve locked instance %s" % self.op.instance_name
6563
6564 - def Exec(self, feedback_fn):
6565 """Remove the instance. 6566 6567 """ 6568 instance = self.instance 6569 logging.info("Shutting down instance %s on node %s", 6570 instance.name, instance.primary_node) 6571 6572 result = self.rpc.call_instance_shutdown(instance.primary_node, instance, 6573 self.op.shutdown_timeout) 6574 msg = result.fail_msg 6575 if msg: 6576 if self.op.ignore_failures: 6577 feedback_fn("Warning: can't shutdown instance: %s" % msg) 6578 else: 6579 raise errors.OpExecError("Could not shutdown instance %s on" 6580 " node %s: %s" % 6581 (instance.name, instance.primary_node, msg)) 6582 6583 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
6584
6585 6586 -def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
6587 """Utility function to remove an instance. 6588 6589 """ 6590 logging.info("Removing block devices for instance %s", instance.name) 6591 6592 if not _RemoveDisks(lu, instance, ignore_failures=ignore_failures): 6593 if not ignore_failures: 6594 raise errors.OpExecError("Can't remove instance's disks") 6595 feedback_fn("Warning: can't remove instance's disks") 6596 6597 logging.info("Removing instance %s out of cluster config", instance.name) 6598 6599 lu.cfg.RemoveInstance(instance.name) 6600 6601 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \ 6602 "Instance lock removal conflict" 6603 6604 # Remove lock for the instance 6605 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
6606
6607 6608 -class LUInstanceQuery(NoHooksLU):
6609 """Logical unit for querying instances. 6610 6611 """ 6612 # pylint: disable=W0142 6613 REQ_BGL = False 6614
6615 - def CheckArguments(self):
6616 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names), 6617 self.op.output_fields, self.op.use_locking)
6618
6619 - def ExpandNames(self):
6620 self.iq.ExpandNames(self)
6621
6622 - def DeclareLocks(self, level):
6623 self.iq.DeclareLocks(self, level)
6624
6625 - def Exec(self, feedback_fn):
6626 return self.iq.OldStyleQuery(self)
6627
6628 6629 -class LUInstanceFailover(LogicalUnit):
6630 """Failover an instance. 6631 6632 """ 6633 HPATH = "instance-failover" 6634 HTYPE = constants.HTYPE_INSTANCE 6635 REQ_BGL = False 6636
6637 - def CheckArguments(self):
6638 """Check the arguments. 6639 6640 """ 6641 self.iallocator = getattr(self.op, "iallocator", None) 6642 self.target_node = getattr(self.op, "target_node", None)
6643
6644 - def ExpandNames(self):
6645 self._ExpandAndLockInstance() 6646 6647 if self.op.target_node is not None: 6648 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node) 6649 6650 self.needed_locks[locking.LEVEL_NODE] = [] 6651 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE 6652 6653 ignore_consistency = self.op.ignore_consistency 6654 shutdown_timeout = self.op.shutdown_timeout 6655 self._migrater = TLMigrateInstance(self, self.op.instance_name, 6656 cleanup=False, 6657 failover=True, 6658 ignore_consistency=ignore_consistency, 6659 shutdown_timeout=shutdown_timeout) 6660 self.tasklets = [self._migrater]
6661
6662 - def DeclareLocks(self, level):
6663 if level == locking.LEVEL_NODE: 6664 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name) 6665 if instance.disk_template in constants.DTS_EXT_MIRROR: 6666 if self.op.target_node is None: 6667 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 6668 else: 6669 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node, 6670 self.op.target_node] 6671 del self.recalculate_locks[locking.LEVEL_NODE] 6672 else: 6673 self._LockInstancesNodes()
6674
6675 - def BuildHooksEnv(self):
6676 """Build hooks env. 6677 6678 This runs on master, primary and secondary nodes of the instance. 6679 6680 """ 6681 instance = self._migrater.instance 6682 source_node = instance.primary_node 6683 target_node = self.op.target_node 6684 env = { 6685 "IGNORE_CONSISTENCY": self.op.ignore_consistency, 6686 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, 6687 "OLD_PRIMARY": source_node, 6688 "NEW_PRIMARY": target_node, 6689 } 6690 6691 if instance.disk_template in constants.DTS_INT_MIRROR: 6692 env["OLD_SECONDARY"] = instance.secondary_nodes[0] 6693 env["NEW_SECONDARY"] = source_node 6694 else: 6695 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = "" 6696 6697 env.update(_BuildInstanceHookEnvByObject(self, instance)) 6698 6699 return env
6700
6701 - def BuildHooksNodes(self):
6702 """Build hooks nodes. 6703 6704 """ 6705 instance = self._migrater.instance 6706 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes) 6707 return (nl, nl + [instance.primary_node])
6708
6709 6710 -class LUInstanceMigrate(LogicalUnit):
6711 """Migrate an instance. 6712 6713 This is migration without shutting down, compared to the failover, 6714 which is done with shutdown. 6715 6716 """ 6717 HPATH = "instance-migrate" 6718 HTYPE = constants.HTYPE_INSTANCE 6719 REQ_BGL = False 6720
6721 - def ExpandNames(self):
6722 self._ExpandAndLockInstance() 6723 6724 if self.op.target_node is not None: 6725 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node) 6726 6727 self.needed_locks[locking.LEVEL_NODE] = [] 6728 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE 6729 6730 self._migrater = TLMigrateInstance(self, self.op.instance_name, 6731 cleanup=self.op.cleanup, 6732 failover=False, 6733 fallback=self.op.allow_failover) 6734 self.tasklets = [self._migrater]
6735
6736 - def DeclareLocks(self, level):
6737 if level == locking.LEVEL_NODE: 6738 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name) 6739 if instance.disk_template in constants.DTS_EXT_MIRROR: 6740 if self.op.target_node is None: 6741 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 6742 else: 6743 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node, 6744 self.op.target_node] 6745 del self.recalculate_locks[locking.LEVEL_NODE] 6746 else: 6747 self._LockInstancesNodes()
6748
6749 - def BuildHooksEnv(self):
6750 """Build hooks env. 6751 6752 This runs on master, primary and secondary nodes of the instance. 6753 6754 """ 6755 instance = self._migrater.instance 6756 source_node = instance.primary_node 6757 target_node = self.op.target_node 6758 env = _BuildInstanceHookEnvByObject(self, instance) 6759 env.update({ 6760 "MIGRATE_LIVE": self._migrater.live, 6761 "MIGRATE_CLEANUP": self.op.cleanup, 6762 "OLD_PRIMARY": source_node, 6763 "NEW_PRIMARY": target_node, 6764 }) 6765 6766 if instance.disk_template in constants.DTS_INT_MIRROR: 6767 env["OLD_SECONDARY"] = target_node 6768 env["NEW_SECONDARY"] = source_node 6769 else: 6770 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None 6771 6772 return env
6773
6774 - def BuildHooksNodes(self):
6775 """Build hooks nodes. 6776 6777 """ 6778 instance = self._migrater.instance 6779 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes) 6780 return (nl, nl + [instance.primary_node])
6781
6782 6783 -class LUInstanceMove(LogicalUnit):
6784 """Move an instance by data-copying. 6785 6786 """ 6787 HPATH = "instance-move" 6788 HTYPE = constants.HTYPE_INSTANCE 6789 REQ_BGL = False 6790
6791 - def ExpandNames(self):
6792 self._ExpandAndLockInstance() 6793 target_node = _ExpandNodeName(self.cfg, self.op.target_node) 6794 self.op.target_node = target_node 6795 self.needed_locks[locking.LEVEL_NODE] = [target_node] 6796 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6797
6798 - def DeclareLocks(self, level):
6799 if level == locking.LEVEL_NODE: 6800 self._LockInstancesNodes(primary_only=True)
6801
6802 - def BuildHooksEnv(self):
6803 """Build hooks env. 6804 6805 This runs on master, primary and secondary nodes of the instance. 6806 6807 """ 6808 env = { 6809 "TARGET_NODE": self.op.target_node, 6810 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, 6811 } 6812 env.update(_BuildInstanceHookEnvByObject(self, self.instance)) 6813 return env
6814
6815 - def BuildHooksNodes(self):
6816 """Build hooks nodes. 6817 6818 """ 6819 nl = [ 6820 self.cfg.GetMasterNode(), 6821 self.instance.primary_node, 6822 self.op.target_node, 6823 ] 6824 return (nl, nl)
6825
6826 - def CheckPrereq(self):
6827 """Check prerequisites. 6828 6829 This checks that the instance is in the cluster. 6830 6831 """ 6832 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) 6833 assert self.instance is not None, \ 6834 "Cannot retrieve locked instance %s" % self.op.instance_name 6835 6836 node = self.cfg.GetNodeInfo(self.op.target_node) 6837 assert node is not None, \ 6838 "Cannot retrieve locked node %s" % self.op.target_node 6839 6840 self.target_node = target_node = node.name 6841 6842 if target_node == instance.primary_node: 6843 raise errors.OpPrereqError("Instance %s is already on the node %s" % 6844 (instance.name, target_node), 6845 errors.ECODE_STATE) 6846 6847 bep = self.cfg.GetClusterInfo().FillBE(instance) 6848 6849 for idx, dsk in enumerate(instance.disks): 6850 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE): 6851 raise errors.OpPrereqError("Instance disk %d has a complex layout," 6852 " cannot copy" % idx, errors.ECODE_STATE) 6853 6854 _CheckNodeOnline(self, target_node) 6855 _CheckNodeNotDrained(self, target_node) 6856 _CheckNodeVmCapable(self, target_node) 6857 6858 if instance.admin_up: 6859 # check memory requirements on the secondary node 6860 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" % 6861 instance.name, bep[constants.BE_MEMORY], 6862 instance.hypervisor) 6863 else: 6864 self.LogInfo("Not checking memory on the secondary node as" 6865 " instance will not be started") 6866 6867 # check bridge existance 6868 _CheckInstanceBridgesExist(self, instance, node=target_node)
6869
6870 - def Exec(self, feedback_fn):
6871 """Move an instance. 6872 6873 The move is done by shutting it down on its present node, copying 6874 the data over (slow) and starting it on the new node. 6875 6876 """ 6877 instance = self.instance 6878 6879 source_node = instance.primary_node 6880 target_node = self.target_node 6881 6882 self.LogInfo("Shutting down instance %s on source node %s", 6883 instance.name, source_node) 6884 6885 result = self.rpc.call_instance_shutdown(source_node, instance, 6886 self.op.shutdown_timeout) 6887 msg = result.fail_msg 6888 if msg: 6889 if self.op.ignore_consistency: 6890 self.proc.LogWarning("Could not shutdown instance %s on node %s." 6891 " Proceeding anyway. Please make sure node" 6892 " %s is down. Error details: %s", 6893 instance.name, source_node, source_node, msg) 6894 else: 6895 raise errors.OpExecError("Could not shutdown instance %s on" 6896 " node %s: %s" % 6897 (instance.name, source_node, msg)) 6898 6899 # create the target disks 6900 try: 6901 _CreateDisks(self, instance, target_node=target_node) 6902 except errors.OpExecError: 6903 self.LogWarning("Device creation failed, reverting...") 6904 try: 6905 _RemoveDisks(self, instance, target_node=target_node) 6906 finally: 6907 self.cfg.ReleaseDRBDMinors(instance.name) 6908 raise 6909 6910 cluster_name = self.cfg.GetClusterInfo().cluster_name 6911 6912 errs = [] 6913 # activate, get path, copy the data over 6914 for idx, disk in enumerate(instance.disks): 6915 self.LogInfo("Copying data for disk %d", idx) 6916 result = self.rpc.call_blockdev_assemble(target_node, disk, 6917 instance.name, True, idx) 6918 if result.fail_msg: 6919 self.LogWarning("Can't assemble newly created disk %d: %s", 6920 idx, result.fail_msg) 6921 errs.append(result.fail_msg) 6922 break 6923 dev_path = result.payload 6924 result = self.rpc.call_blockdev_export(source_node, disk, 6925 target_node, dev_path, 6926 cluster_name) 6927 if result.fail_msg: 6928 self.LogWarning("Can't copy data over for disk %d: %s", 6929 idx, result.fail_msg) 6930 errs.append(result.fail_msg) 6931 break 6932 6933 if errs: 6934 self.LogWarning("Some disks failed to copy, aborting") 6935 try: 6936 _RemoveDisks(self, instance, target_node=target_node) 6937 finally: 6938 self.cfg.ReleaseDRBDMinors(instance.name) 6939 raise errors.OpExecError("Errors during disk copy: %s" % 6940 (",".join(errs),)) 6941 6942 instance.primary_node = target_node 6943 self.cfg.Update(instance, feedback_fn) 6944 6945 self.LogInfo("Removing the disks on the original node") 6946 _RemoveDisks(self, instance, target_node=source_node) 6947 6948 # Only start the instance if it's marked as up 6949 if instance.admin_up: 6950 self.LogInfo("Starting instance %s on node %s", 6951 instance.name, target_node) 6952 6953 disks_ok, _ = _AssembleInstanceDisks(self, instance, 6954 ignore_secondaries=True) 6955 if not disks_ok: 6956 _ShutdownInstanceDisks(self, instance) 6957 raise errors.OpExecError("Can't activate the instance's disks") 6958 6959 result = self.rpc.call_instance_start(target_node, instance, 6960 None, None, False) 6961 msg = result.fail_msg 6962 if msg: 6963 _ShutdownInstanceDisks(self, instance) 6964 raise errors.OpExecError("Could not start instance %s on node %s: %s" % 6965 (instance.name, target_node, msg))
6966
6967 6968 -class LUNodeMigrate(LogicalUnit):
6969 """Migrate all instances from a node. 6970 6971 """ 6972 HPATH = "node-migrate" 6973 HTYPE = constants.HTYPE_NODE 6974 REQ_BGL = False 6975
6976 - def CheckArguments(self):
6977 pass
6978
6979 - def ExpandNames(self):
6980 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 6981 6982 self.share_locks = _ShareAll() 6983 self.needed_locks = { 6984 locking.LEVEL_NODE: [self.op.node_name], 6985 }
6986
6987 - def BuildHooksEnv(self):
6988 """Build hooks env. 6989 6990 This runs on the master, the primary and all the secondaries. 6991 6992 """ 6993 return { 6994 "NODE_NAME": self.op.node_name, 6995 }
6996
6997 - def BuildHooksNodes(self):
6998 """Build hooks nodes. 6999 7000 """ 7001 nl = [self.cfg.GetMasterNode()] 7002 return (nl, nl)
7003
7004 - def CheckPrereq(self):
7005 pass
7006
7007 - def Exec(self, feedback_fn):
7008 # Prepare jobs for migration instances 7009 jobs = [ 7010 [opcodes.OpInstanceMigrate(instance_name=inst.name, 7011 mode=self.op.mode, 7012 live=self.op.live, 7013 iallocator=self.op.iallocator, 7014 target_node=self.op.target_node)] 7015 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name) 7016 ] 7017 7018 # TODO: Run iallocator in this opcode and pass correct placement options to 7019 # OpInstanceMigrate. Since other jobs can modify the cluster between 7020 # running the iallocator and the actual migration, a good consistency model 7021 # will have to be found. 7022 7023 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) == 7024 frozenset([self.op.node_name])) 7025 7026 return ResultWithJobs(jobs)
7027
7028 7029 -class TLMigrateInstance(Tasklet):
7030 """Tasklet class for instance migration. 7031 7032 @type live: boolean 7033 @ivar live: whether the migration will be done live or non-live; 7034 this variable is initalized only after CheckPrereq has run 7035 @type cleanup: boolean 7036 @ivar cleanup: Wheater we cleanup from a failed migration 7037 @type iallocator: string 7038 @ivar iallocator: The iallocator used to determine target_node 7039 @type target_node: string 7040 @ivar target_node: If given, the target_node to reallocate the instance to 7041 @type failover: boolean 7042 @ivar failover: Whether operation results in failover or migration 7043 @type fallback: boolean 7044 @ivar fallback: Whether fallback to failover is allowed if migration not 7045 possible 7046 @type ignore_consistency: boolean 7047 @ivar ignore_consistency: Wheter we should ignore consistency between source 7048 and target node 7049 @type shutdown_timeout: int 7050 @ivar shutdown_timeout: In case of failover timeout of the shutdown 7051 7052 """
7053 - def __init__(self, lu, instance_name, cleanup=False, 7054 failover=False, fallback=False, 7055 ignore_consistency=False, 7056 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT):
7057 """Initializes this class. 7058 7059 """ 7060 Tasklet.__init__(self, lu) 7061 7062 # Parameters 7063 self.instance_name = instance_name 7064 self.cleanup = cleanup 7065 self.live = False # will be overridden later 7066 self.failover = failover 7067 self.fallback = fallback 7068 self.ignore_consistency = ignore_consistency 7069 self.shutdown_timeout = shutdown_timeout
7070
7071 - def CheckPrereq(self):
7072 """Check prerequisites. 7073 7074 This checks that the instance is in the cluster. 7075 7076 """ 7077 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name) 7078 instance = self.cfg.GetInstanceInfo(instance_name) 7079 assert instance is not None 7080 self.instance = instance 7081 7082 if (not self.cleanup and not instance.admin_up and not self.failover and 7083 self.fallback): 7084 self.lu.LogInfo("Instance is marked down, fallback allowed, switching" 7085 " to failover") 7086 self.failover = True 7087 7088 if instance.disk_template not in constants.DTS_MIRRORED: 7089 if self.failover: 7090 text = "failovers" 7091 else: 7092 text = "migrations" 7093 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow" 7094 " %s" % (instance.disk_template, text), 7095 errors.ECODE_STATE) 7096 7097 if instance.disk_template in constants.DTS_EXT_MIRROR: 7098 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node") 7099 7100 if self.lu.op.iallocator: 7101 self._RunAllocator() 7102 else: 7103 # We set set self.target_node as it is required by 7104 # BuildHooksEnv 7105 self.target_node = self.lu.op.target_node 7106 7107 # self.target_node is already populated, either directly or by the 7108 # iallocator run 7109 target_node = self.target_node 7110 if self.target_node == instance.primary_node: 7111 raise errors.OpPrereqError("Cannot migrate instance %s" 7112 " to its primary (%s)" % 7113 (instance.name, instance.primary_node)) 7114 7115 if len(self.lu.tasklets) == 1: 7116 # It is safe to release locks only when we're the only tasklet 7117 # in the LU 7118 _ReleaseLocks(self.lu, locking.LEVEL_NODE, 7119 keep=[instance.primary_node, self.target_node]) 7120 7121 else: 7122 secondary_nodes = instance.secondary_nodes 7123 if not secondary_nodes: 7124 raise errors.ConfigurationError("No secondary node but using" 7125 " %s disk template" % 7126 instance.disk_template) 7127 target_node = secondary_nodes[0] 7128 if self.lu.op.iallocator or (self.lu.op.target_node and 7129 self.lu.op.target_node != target_node): 7130 if self.failover: 7131 text = "failed over" 7132 else: 7133 text = "migrated" 7134 raise errors.OpPrereqError("Instances with disk template %s cannot" 7135 " be %s to arbitrary nodes" 7136 " (neither an iallocator nor a target" 7137 " node can be passed)" % 7138 (instance.disk_template, text), 7139 errors.ECODE_INVAL) 7140 7141 i_be = self.cfg.GetClusterInfo().FillBE(instance) 7142 7143 # check memory requirements on the secondary node 7144 if not self.cleanup and (not self.failover or instance.admin_up): 7145 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" % 7146 instance.name, i_be[constants.BE_MEMORY], 7147 instance.hypervisor) 7148 else: 7149 self.lu.LogInfo("Not checking memory on the secondary node as" 7150 " instance will not be started") 7151 7152 # check bridge existance 7153 _CheckInstanceBridgesExist(self.lu, instance, node=target_node) 7154 7155 if not self.cleanup: 7156 _CheckNodeNotDrained(self.lu, target_node) 7157 if not self.failover: 7158 result = self.rpc.call_instance_migratable(instance.primary_node, 7159 instance) 7160 if result.fail_msg and self.fallback: 7161 self.lu.LogInfo("Can't migrate, instance offline, fallback to" 7162 " failover") 7163 self.failover = True 7164 else: 7165 result.Raise("Can't migrate, please use failover", 7166 prereq=True, ecode=errors.ECODE_STATE) 7167 7168 assert not (self.failover and self.cleanup) 7169 7170 if not self.failover: 7171 if self.lu.op.live is not None and self.lu.op.mode is not None: 7172 raise errors.OpPrereqError("Only one of the 'live' and 'mode'" 7173 " parameters are accepted", 7174 errors.ECODE_INVAL) 7175 if self.lu.op.live is not None: 7176 if self.lu.op.live: 7177 self.lu.op.mode = constants.HT_MIGRATION_LIVE 7178 else: 7179 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE 7180 # reset the 'live' parameter to None so that repeated 7181 # invocations of CheckPrereq do not raise an exception 7182 self.lu.op.live = None 7183 elif self.lu.op.mode is None: 7184 # read the default value from the hypervisor 7185 i_hv = self.cfg.GetClusterInfo().FillHV(self.instance, 7186 skip_globals=False) 7187 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE] 7188 7189 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE 7190 else: 7191 # Failover is never live 7192 self.live = False
7193
7194 - def _RunAllocator(self):
7195 """Run the allocator based on input opcode. 7196 7197 """ 7198 ial = IAllocator(self.cfg, self.rpc, 7199 mode=constants.IALLOCATOR_MODE_RELOC, 7200 name=self.instance_name, 7201 # TODO See why hail breaks with a single node below 7202 relocate_from=[self.instance.primary_node, 7203 self.instance.primary_node], 7204 ) 7205 7206 ial.Run(self.lu.op.iallocator) 7207 7208 if not ial.success: 7209 raise errors.OpPrereqError("Can't compute nodes using" 7210 " iallocator '%s': %s" % 7211 (self.lu.op.iallocator, ial.info), 7212 errors.ECODE_NORES) 7213 if len(ial.result) != ial.required_nodes: 7214 raise errors.OpPrereqError("iallocator '%s' returned invalid number" 7215 " of nodes (%s), required %s" % 7216 (self.lu.op.iallocator, len(ial.result), 7217 ial.required_nodes), errors.ECODE_FAULT) 7218 self.target_node = ial.result[0] 7219 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s", 7220 self.instance_name, self.lu.op.iallocator, 7221 utils.CommaJoin(ial.result))
7222
7223 - def _WaitUntilSync(self):
7224 """Poll with custom rpc for disk sync. 7225 7226 This uses our own step-based rpc call. 7227 7228 """ 7229 self.feedback_fn("* wait until resync is done") 7230 all_done = False 7231 while not all_done: 7232 all_done = True 7233 result = self.rpc.call_drbd_wait_sync(self.all_nodes, 7234 self.nodes_ip, 7235 self.instance.disks) 7236 min_percent = 100 7237 for node, nres in result.items(): 7238 nres.Raise("Cannot resync disks on node %s" % node) 7239 node_done, node_percent = nres.payload 7240 all_done = all_done and node_done 7241 if node_percent is not None: 7242 min_percent = min(min_percent, node_percent) 7243 if not all_done: 7244 if min_percent < 100: 7245 self.feedback_fn(" - progress: %.1f%%" % min_percent) 7246 time.sleep(2)
7247
7248 - def _EnsureSecondary(self, node):
7249 """Demote a node to secondary. 7250 7251 """ 7252 self.feedback_fn("* switching node %s to secondary mode" % node) 7253 7254 for dev in self.instance.disks: 7255 self.cfg.SetDiskID(dev, node) 7256 7257 result = self.rpc.call_blockdev_close(node, self.instance.name, 7258 self.instance.disks) 7259 result.Raise("Cannot change disk to secondary on node %s" % node)
7260
7261 - def _GoStandalone(self):
7262 """Disconnect from the network. 7263 7264 """ 7265 self.feedback_fn("* changing into standalone mode") 7266 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip, 7267 self.instance.disks) 7268 for node, nres in result.items(): 7269 nres.Raise("Cannot disconnect disks node %s" % node)
7270
7271 - def _GoReconnect(self, multimaster):
7272 """Reconnect to the network. 7273 7274 """ 7275 if multimaster: 7276 msg = "dual-master" 7277 else: 7278 msg = "single-master" 7279 self.feedback_fn("* changing disks into %s mode" % msg) 7280 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip, 7281 self.instance.disks, 7282 self.instance.name, multimaster) 7283 for node, nres in result.items(): 7284 nres.Raise("Cannot change disks config on node %s" % node)
7285
7286 - def _ExecCleanup(self):
7287 """Try to cleanup after a failed migration. 7288 7289 The cleanup is done by: 7290 - check that the instance is running only on one node 7291 (and update the config if needed) 7292 - change disks on its secondary node to secondary 7293 - wait until disks are fully synchronized 7294 - disconnect from the network 7295 - change disks into single-master mode 7296 - wait again until disks are fully synchronized 7297 7298 """ 7299 instance = self.instance 7300 target_node = self.target_node 7301 source_node = self.source_node 7302 7303 # check running on only one node 7304 self.feedback_fn("* checking where the instance actually runs" 7305 " (if this hangs, the hypervisor might be in" 7306 " a bad state)") 7307 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor]) 7308 for node, result in ins_l.items(): 7309 result.Raise("Can't contact node %s" % node) 7310 7311 runningon_source = instance.name in ins_l[source_node].payload 7312 runningon_target = instance.name in ins_l[target_node].payload 7313 7314 if runningon_source and runningon_target: 7315 raise errors.OpExecError("Instance seems to be running on two nodes," 7316 " or the hypervisor is confused; you will have" 7317 " to ensure manually that it runs only on one" 7318 " and restart this operation") 7319 7320 if not (runningon_source or runningon_target): 7321 raise errors.OpExecError("Instance does not seem to be running at all;" 7322 " in this case it's safer to repair by" 7323 " running 'gnt-instance stop' to ensure disk" 7324 " shutdown, and then restarting it") 7325 7326 if runningon_target: 7327 # the migration has actually succeeded, we need to update the config 7328 self.feedback_fn("* instance running on secondary node (%s)," 7329 " updating config" % target_node) 7330 instance.primary_node = target_node 7331 self.cfg.Update(instance, self.feedback_fn) 7332 demoted_node = source_node 7333 else: 7334 self.feedback_fn("* instance confirmed to be running on its" 7335 " primary node (%s)" % source_node) 7336 demoted_node = target_node 7337 7338 if instance.disk_template in constants.DTS_INT_MIRROR: 7339 self._EnsureSecondary(demoted_node) 7340 try: 7341 self._WaitUntilSync() 7342 except errors.OpExecError: 7343 # we ignore here errors, since if the device is standalone, it 7344 # won't be able to sync 7345 pass 7346 self._GoStandalone() 7347 self._GoReconnect(False) 7348 self._WaitUntilSync() 7349 7350 self.feedback_fn("* done")
7351
7352 - def _RevertDiskStatus(self):
7353 """Try to revert the disk status after a failed migration. 7354 7355 """ 7356 target_node = self.target_node 7357 if self.instance.disk_template in constants.DTS_EXT_MIRROR: 7358 return 7359 7360 try: 7361 self._EnsureSecondary(target_node) 7362 self._GoStandalone() 7363 self._GoReconnect(False) 7364 self._WaitUntilSync() 7365 except errors.OpExecError, err: 7366 self.lu.LogWarning("Migration failed and I can't reconnect the drives," 7367 " please try to recover the instance manually;" 7368 " error '%s'" % str(err))
7369
7370 - def _AbortMigration(self):
7371 """Call the hypervisor code to abort a started migration. 7372 7373 """ 7374 instance = self.instance 7375 target_node = self.target_node 7376 migration_info = self.migration_info 7377 7378 abort_result = self.rpc.call_finalize_migration(target_node, 7379 instance, 7380 migration_info, 7381 False) 7382 abort_msg = abort_result.fail_msg 7383 if abort_msg: 7384 logging.error("Aborting migration failed on target node %s: %s", 7385 target_node, abort_msg)
7386 # Don't raise an exception here, as we stil have to try to revert the 7387 # disk status, even if this step failed. 7388
7389 - def _ExecMigration(self):
7390 """Migrate an instance. 7391 7392 The migrate is done by: 7393 - change the disks into dual-master mode 7394 - wait until disks are fully synchronized again 7395 - migrate the instance 7396 - change disks on the new secondary node (the old primary) to secondary 7397 - wait until disks are fully synchronized 7398 - change disks into single-master mode 7399 7400 """ 7401 instance = self.instance 7402 target_node = self.target_node 7403 source_node = self.source_node 7404 7405 self.feedback_fn("* checking disk consistency between source and target") 7406 for dev in instance.disks: 7407 if not _CheckDiskConsistency(self.lu, dev, target_node, False): 7408 raise errors.OpExecError("Disk %s is degraded or not fully" 7409 " synchronized on target node," 7410 " aborting migration" % dev.iv_name) 7411 7412 # First get the migration information from the remote node 7413 result = self.rpc.call_migration_info(source_node, instance) 7414 msg = result.fail_msg 7415 if msg: 7416 log_err = ("Failed fetching source migration information from %s: %s" % 7417 (source_node, msg)) 7418 logging.error(log_err) 7419 raise errors.OpExecError(log_err) 7420 7421 self.migration_info = migration_info = result.payload 7422 7423 if self.instance.disk_template not in constants.DTS_EXT_MIRROR: 7424 # Then switch the disks to master/master mode 7425 self._EnsureSecondary(target_node) 7426 self._GoStandalone() 7427 self._GoReconnect(True) 7428 self._WaitUntilSync() 7429 7430 self.feedback_fn("* preparing %s to accept the instance" % target_node) 7431 result = self.rpc.call_accept_instance(target_node, 7432 instance, 7433 migration_info, 7434 self.nodes_ip[target_node]) 7435 7436 msg = result.fail_msg 7437 if msg: 7438 logging.error("Instance pre-migration failed, trying to revert" 7439 " disk status: %s", msg) 7440 self.feedback_fn("Pre-migration failed, aborting") 7441 self._AbortMigration() 7442 self._RevertDiskStatus() 7443 raise errors.OpExecError("Could not pre-migrate instance %s: %s" % 7444 (instance.name, msg)) 7445 7446 self.feedback_fn("* migrating instance to %s" % target_node) 7447 result = self.rpc.call_instance_migrate(source_node, instance, 7448 self.nodes_ip[target_node], 7449 self.live) 7450 msg = result.fail_msg 7451 if msg: 7452 logging.error("Instance migration failed, trying to revert" 7453 " disk status: %s", msg) 7454 self.feedback_fn("Migration failed, aborting") 7455 self._AbortMigration() 7456 self._RevertDiskStatus() 7457 raise errors.OpExecError("Could not migrate instance %s: %s" % 7458 (instance.name, msg)) 7459 7460 instance.primary_node = target_node 7461 # distribute new instance config to the other nodes 7462 self.cfg.Update(instance, self.feedback_fn) 7463 7464 result = self.rpc.call_finalize_migration(target_node, 7465 instance, 7466 migration_info, 7467 True) 7468 msg = result.fail_msg 7469 if msg: 7470 logging.error("Instance migration succeeded, but finalization failed:" 7471 " %s", msg) 7472 raise errors.OpExecError("Could not finalize instance migration: %s" % 7473 msg) 7474 7475 if self.instance.disk_template not in constants.DTS_EXT_MIRROR: 7476 self._EnsureSecondary(source_node) 7477 self._WaitUntilSync() 7478 self._GoStandalone() 7479 self._GoReconnect(False) 7480 self._WaitUntilSync() 7481 7482 self.feedback_fn("* done")
7483
7484 - def _ExecFailover(self):
7485 """Failover an instance. 7486 7487 The failover is done by shutting it down on its present node and 7488 starting it on the secondary. 7489 7490 """ 7491 instance = self.instance 7492 primary_node = self.cfg.GetNodeInfo(instance.primary_node) 7493 7494 source_node = instance.primary_node 7495 target_node = self.target_node 7496 7497 if instance.admin_up: 7498 self.feedback_fn("* checking disk consistency between source and target") 7499 for dev in instance.disks: 7500 # for drbd, these are drbd over lvm 7501 if not _CheckDiskConsistency(self.lu, dev, target_node, False): 7502 if primary_node.offline: 7503 self.feedback_fn("Node %s is offline, ignoring degraded disk %s on" 7504 " target node %s" % 7505 (primary_node.name, dev.iv_name, target_node)) 7506 elif not self.ignore_consistency: 7507 raise errors.OpExecError("Disk %s is degraded on target node," 7508 " aborting failover" % dev.iv_name) 7509 else: 7510 self.feedback_fn("* not checking disk consistency as instance is not" 7511 " running") 7512 7513 self.feedback_fn("* shutting down instance on source node") 7514 logging.info("Shutting down instance %s on node %s", 7515 instance.name, source_node) 7516 7517 result = self.rpc.call_instance_shutdown(source_node, instance, 7518 self.shutdown_timeout) 7519 msg = result.fail_msg 7520 if msg: 7521 if self.ignore_consistency or primary_node.offline: 7522 self.lu.LogWarning("Could not shutdown instance %s on node %s," 7523 " proceeding anyway; please make sure node" 7524 " %s is down; error details: %s", 7525 instance.name, source_node, source_node, msg) 7526 else: 7527 raise errors.OpExecError("Could not shutdown instance %s on" 7528 " node %s: %s" % 7529 (instance.name, source_node, msg)) 7530 7531 self.feedback_fn("* deactivating the instance's disks on source node") 7532 if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True): 7533 raise errors.OpExecError("Can't shut down the instance's disks") 7534 7535 instance.primary_node = target_node 7536 # distribute new instance config to the other nodes 7537 self.cfg.Update(instance, self.feedback_fn) 7538 7539 # Only start the instance if it's marked as up 7540 if instance.admin_up: 7541 self.feedback_fn("* activating the instance's disks on target node %s" % 7542 target_node) 7543 logging.info("Starting instance %s on node %s", 7544 instance.name, target_node) 7545 7546 disks_ok, _ = _AssembleInstanceDisks(self.lu, instance, 7547 ignore_secondaries=True) 7548 if not disks_ok: 7549 _ShutdownInstanceDisks(self.lu, instance) 7550 raise errors.OpExecError("Can't activate the instance's disks") 7551 7552 self.feedback_fn("* starting the instance on the target node %s" % 7553 target_node) 7554 result = self.rpc.call_instance_start(target_node, instance, None, None, 7555 False) 7556 msg = result.fail_msg 7557 if msg: 7558 _ShutdownInstanceDisks(self.lu, instance) 7559 raise errors.OpExecError("Could not start instance %s on node %s: %s" % 7560 (instance.name, target_node, msg))
7561
7562 - def Exec(self, feedback_fn):
7563 """Perform the migration. 7564 7565 """ 7566 self.feedback_fn = feedback_fn 7567 self.source_node = self.instance.primary_node 7568 7569 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing 7570 if self.instance.disk_template in constants.DTS_INT_MIRROR: 7571 self.target_node = self.instance.secondary_nodes[0] 7572 # Otherwise self.target_node has been populated either 7573 # directly, or through an iallocator. 7574 7575 self.all_nodes = [self.source_node, self.target_node] 7576 self.nodes_ip = dict((name, node.secondary_ip) for (name, node) 7577 in self.cfg.GetMultiNodeInfo(self.all_nodes)) 7578 7579 if self.failover: 7580 feedback_fn("Failover instance %s" % self.instance.name) 7581 self._ExecFailover() 7582 else: 7583 feedback_fn("Migrating instance %s" % self.instance.name) 7584 7585 if self.cleanup: 7586 return self._ExecCleanup() 7587 else: 7588 return self._ExecMigration()
7589
7590 7591 -def _CreateBlockDev(lu, node, instance, device, force_create, 7592 info, force_open):
7593 """Create a tree of block devices on a given node. 7594 7595 If this device type has to be created on secondaries, create it and 7596 all its children. 7597 7598 If not, just recurse to children keeping the same 'force' value. 7599 7600 @param lu: the lu on whose behalf we execute 7601 @param node: the node on which to create the device 7602 @type instance: L{objects.Instance} 7603 @param instance: the instance which owns the device 7604 @type device: L{objects.Disk} 7605 @param device: the device to create 7606 @type force_create: boolean 7607 @param force_create: whether to force creation of this device; this 7608 will be change to True whenever we find a device which has 7609 CreateOnSecondary() attribute 7610 @param info: the extra 'metadata' we should attach to the device 7611 (this will be represented as a LVM tag) 7612 @type force_open: boolean 7613 @param force_open: this parameter will be passes to the 7614 L{backend.BlockdevCreate} function where it specifies 7615 whether we run on primary or not, and it affects both 7616 the child assembly and the device own Open() execution 7617 7618 """ 7619 if device.CreateOnSecondary(): 7620 force_create = True 7621 7622 if device.children: 7623 for child in device.children: 7624 _CreateBlockDev(lu, node, instance, child, force_create, 7625 info, force_open) 7626 7627 if not force_create: 7628 return 7629 7630 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
7631
7632 7633 -def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
7634 """Create a single block device on a given node. 7635 7636 This will not recurse over children of the device, so they must be 7637 created in advance. 7638 7639 @param lu: the lu on whose behalf we execute 7640 @param node: the node on which to create the device 7641 @type instance: L{objects.Instance} 7642 @param instance: the instance which owns the device 7643 @type device: L{objects.Disk} 7644 @param device: the device to create 7645 @param info: the extra 'metadata' we should attach to the device 7646 (this will be represented as a LVM tag) 7647 @type force_open: boolean 7648 @param force_open: this parameter will be passes to the 7649 L{backend.BlockdevCreate} function where it specifies 7650 whether we run on primary or not, and it affects both 7651 the child assembly and the device own Open() execution 7652 7653 """ 7654 lu.cfg.SetDiskID(device, node) 7655 result = lu.rpc.call_blockdev_create(node, device, device.size, 7656 instance.name, force_open, info) 7657 result.Raise("Can't create block device %s on" 7658 " node %s for instance %s" % (device, node, instance.name)) 7659 if device.physical_id is None: 7660 device.physical_id = result.payload
7661
7662 7663 -def _GenerateUniqueNames(lu, exts):
7664 """Generate a suitable LV name. 7665 7666 This will generate a logical volume name for the given instance. 7667 7668 """ 7669 results = [] 7670 for val in exts: 7671 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId()) 7672 results.append("%s%s" % (new_id, val)) 7673 return results
7674
7675 7676 -def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names, 7677 iv_name, p_minor, s_minor):
7678 """Generate a drbd8 device complete with its children. 7679 7680 """ 7681 assert len(vgnames) == len(names) == 2 7682 port = lu.cfg.AllocatePort() 7683 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId()) 7684 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, 7685 logical_id=(vgnames[0], names[0])) 7686 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128, 7687 logical_id=(vgnames[1], names[1])) 7688 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size, 7689 logical_id=(primary, secondary, port, 7690 p_minor, s_minor, 7691 shared_secret), 7692 children=[dev_data, dev_meta], 7693 iv_name=iv_name) 7694 return drbd_dev
7695
7696 7697 -def _GenerateDiskTemplate(lu, template_name, 7698 instance_name, primary_node, 7699 secondary_nodes, disk_info, 7700 file_storage_dir, file_driver, 7701 base_index, feedback_fn):
7702 """Generate the entire disk layout for a given template type. 7703 7704 """ 7705 #TODO: compute space requirements 7706 7707 vgname = lu.cfg.GetVGName() 7708 disk_count = len(disk_info) 7709 disks = [] 7710 if template_name == constants.DT_DISKLESS: 7711 pass 7712 elif template_name == constants.DT_PLAIN: 7713 if len(secondary_nodes) != 0: 7714 raise errors.ProgrammerError("Wrong template configuration") 7715 7716 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i) 7717 for i in range(disk_count)]) 7718 for idx, disk in enumerate(disk_info): 7719 disk_index = idx + base_index 7720 vg = disk.get(constants.IDISK_VG, vgname) 7721 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx])) 7722 disk_dev = objects.Disk(dev_type=constants.LD_LV, 7723 size=disk[constants.IDISK_SIZE], 7724 logical_id=(vg, names[idx]), 7725 iv_name="disk/%d" % disk_index, 7726 mode=disk[constants.IDISK_MODE]) 7727 disks.append(disk_dev) 7728 elif template_name == constants.DT_DRBD8: 7729 if len(secondary_nodes) != 1: 7730 raise errors.ProgrammerError("Wrong template configuration") 7731 remote_node = secondary_nodes[0] 7732 minors = lu.cfg.AllocateDRBDMinor( 7733 [primary_node, remote_node] * len(disk_info), instance_name) 7734 7735 names = [] 7736 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i) 7737 for i in range(disk_count)]): 7738 names.append(lv_prefix + "_data") 7739 names.append(lv_prefix + "_meta") 7740 for idx, disk in enumerate(disk_info): 7741 disk_index = idx + base_index 7742 data_vg = disk.get(constants.IDISK_VG, vgname) 7743 meta_vg = disk.get(constants.IDISK_METAVG, data_vg) 7744 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node, 7745 disk[constants.IDISK_SIZE], 7746 [data_vg, meta_vg], 7747 names[idx * 2:idx * 2 + 2], 7748 "disk/%d" % disk_index, 7749 minors[idx * 2], minors[idx * 2 + 1]) 7750 disk_dev.mode = disk[constants.IDISK_MODE] 7751 disks.append(disk_dev) 7752 elif template_name == constants.DT_FILE: 7753 if len(secondary_nodes) != 0: 7754 raise errors.ProgrammerError("Wrong template configuration") 7755 7756 opcodes.RequireFileStorage() 7757 7758 for idx, disk in enumerate(disk_info): 7759 disk_index = idx + base_index 7760 disk_dev = objects.Disk(dev_type=constants.LD_FILE, 7761 size=disk[constants.IDISK_SIZE], 7762 iv_name="disk/%d" % disk_index, 7763 logical_id=(file_driver, 7764 "%s/disk%d" % (file_storage_dir, 7765 disk_index)), 7766 mode=disk[constants.IDISK_MODE]) 7767 disks.append(disk_dev) 7768 elif template_name == constants.DT_SHARED_FILE: 7769 if len(secondary_nodes) != 0: 7770 raise errors.ProgrammerError("Wrong template configuration") 7771 7772 opcodes.RequireSharedFileStorage() 7773 7774 for idx, disk in enumerate(disk_info): 7775 disk_index = idx + base_index 7776 disk_dev = objects.Disk(dev_type=constants.LD_FILE, 7777 size=disk[constants.IDISK_SIZE], 7778 iv_name="disk/%d" % disk_index, 7779 logical_id=(file_driver, 7780 "%s/disk%d" % (file_storage_dir, 7781 disk_index)), 7782 mode=disk[constants.IDISK_MODE]) 7783 disks.append(disk_dev) 7784 elif template_name == constants.DT_BLOCK: 7785 if len(secondary_nodes) != 0: 7786 raise errors.ProgrammerError("Wrong template configuration") 7787 7788 for idx, disk in enumerate(disk_info): 7789 disk_index = idx + base_index 7790 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV, 7791 size=disk[constants.IDISK_SIZE], 7792 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL, 7793 disk[constants.IDISK_ADOPT]), 7794 iv_name="disk/%d" % disk_index, 7795 mode=disk[constants.IDISK_MODE]) 7796 disks.append(disk_dev) 7797 7798 else: 7799 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name) 7800 return disks
7801
7802 7803 -def _GetInstanceInfoText(instance):
7804 """Compute that text that should be added to the disk's metadata. 7805 7806 """ 7807 return "originstname+%s" % instance.name
7808
7809 7810 -def _CalcEta(time_taken, written, total_size):
7811 """Calculates the ETA based on size written and total size. 7812 7813 @param time_taken: The time taken so far 7814 @param written: amount written so far 7815 @param total_size: The total size of data to be written 7816 @return: The remaining time in seconds 7817 7818 """ 7819 avg_time = time_taken / float(written) 7820 return (total_size - written) * avg_time
7821
7822 7823 -def _WipeDisks(lu, instance):
7824 """Wipes instance disks. 7825 7826 @type lu: L{LogicalUnit} 7827 @param lu: the logical unit on whose behalf we execute 7828 @type instance: L{objects.Instance} 7829 @param instance: the instance whose disks we should create 7830 @return: the success of the wipe 7831 7832 """ 7833 node = instance.primary_node 7834 7835 for device in instance.disks: 7836 lu.cfg.SetDiskID(device, node) 7837 7838 logging.info("Pause sync of instance %s disks", instance.name) 7839 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True) 7840 7841 for idx, success in enumerate(result.payload): 7842 if not success: 7843 logging.warn("pause-sync of instance %s for disks %d failed", 7844 instance.name, idx) 7845 7846 try: 7847 for idx, device in enumerate(instance.disks): 7848 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but 7849 # MAX_WIPE_CHUNK at max 7850 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 * 7851 constants.MIN_WIPE_CHUNK_PERCENT) 7852 # we _must_ make this an int, otherwise rounding errors will 7853 # occur 7854 wipe_chunk_size = int(wipe_chunk_size) 7855 7856 lu.LogInfo("* Wiping disk %d", idx) 7857 logging.info("Wiping disk %d for instance %s, node %s using" 7858 " chunk size %s", idx, instance.name, node, wipe_chunk_size) 7859 7860 offset = 0 7861 size = device.size 7862 last_output = 0 7863 start_time = time.time() 7864 7865 while offset < size: 7866 wipe_size = min(wipe_chunk_size, size - offset) 7867 logging.debug("Wiping disk %d, offset %s, chunk %s", 7868 idx, offset, wipe_size) 7869 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size) 7870 result.Raise("Could not wipe disk %d at offset %d for size %d" % 7871 (idx, offset, wipe_size)) 7872 now = time.time() 7873 offset += wipe_size 7874 if now - last_output >= 60: 7875 eta = _CalcEta(now - start_time, offset, size) 7876 lu.LogInfo(" - done: %.1f%% ETA: %s" % 7877 (offset / float(size) * 100, utils.FormatSeconds(eta))) 7878 last_output = now 7879 finally: 7880 logging.info("Resume sync of instance %s disks", instance.name) 7881 7882 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False) 7883 7884 for idx, success in enumerate(result.payload): 7885 if not success: 7886 lu.LogWarning("Resume sync of disk %d failed, please have a" 7887 " look at the status and troubleshoot the issue", idx) 7888 logging.warn("resume-sync of instance %s for disks %d failed", 7889 instance.name, idx)
7890
7891 7892 -def _CreateDisks(lu, instance, to_skip=None, target_node=None):
7893 """Create all disks for an instance. 7894 7895 This abstracts away some work from AddInstance. 7896 7897 @type lu: L{LogicalUnit} 7898 @param lu: the logical unit on whose behalf we execute 7899 @type instance: L{objects.Instance} 7900 @param instance: the instance whose disks we should create 7901 @type to_skip: list 7902 @param to_skip: list of indices to skip 7903 @type target_node: string 7904 @param target_node: if passed, overrides the target node for creation 7905 @rtype: boolean 7906 @return: the success of the creation 7907 7908 """ 7909 info = _GetInstanceInfoText(instance) 7910 if target_node is None: 7911 pnode = instance.primary_node 7912 all_nodes = instance.all_nodes 7913 else: 7914 pnode = target_node 7915 all_nodes = [pnode] 7916 7917 if instance.disk_template in constants.DTS_FILEBASED: 7918 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1]) 7919 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir) 7920 7921 result.Raise("Failed to create directory '%s' on" 7922 " node %s" % (file_storage_dir, pnode)) 7923 7924 # Note: this needs to be kept in sync with adding of disks in 7925 # LUInstanceSetParams 7926 for idx, device in enumerate(instance.disks): 7927 if to_skip and idx in to_skip: 7928 continue 7929 logging.info("Creating volume %s for instance %s", 7930 device.iv_name, instance.name) 7931 #HARDCODE 7932 for node in all_nodes: 7933 f_create = node == pnode 7934 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
7935
7936 7937 -def _RemoveDisks(lu, instance, target_node=None, ignore_failures=False):
7938 """Remove all disks for an instance. 7939 7940 This abstracts away some work from `AddInstance()` and 7941 `RemoveInstance()`. Note that in case some of the devices couldn't 7942 be removed, the removal will continue with the other ones (compare 7943 with `_CreateDisks()`). 7944 7945 @type lu: L{LogicalUnit} 7946 @param lu: the logical unit on whose behalf we execute 7947 @type instance: L{objects.Instance} 7948 @param instance: the instance whose disks we should remove 7949 @type target_node: string 7950 @param target_node: used to override the node on which to remove the disks 7951 @rtype: boolean 7952 @return: the success of the removal 7953 7954 """ 7955 logging.info("Removing block devices for instance %s", instance.name) 7956 7957 all_result = True 7958 ports_to_release = set() 7959 for device in instance.disks: 7960 if target_node: 7961 edata = [(target_node, device)] 7962 else: 7963 edata = device.ComputeNodeTree(instance.primary_node) 7964 for node, disk in edata: 7965 lu.cfg.SetDiskID(disk, node) 7966 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg 7967 if msg: 7968 lu.LogWarning("Could not remove block device %s on node %s," 7969 " continuing anyway: %s", device.iv_name, node, msg) 7970 all_result = False 7971 7972 # if this is a DRBD disk, return its port to the pool 7973 if device.dev_type in constants.LDS_DRBD: 7974 ports_to_release.add(device.logical_id[2]) 7975 7976 if all_result or ignore_failures: 7977 for port in ports_to_release: 7978 lu.cfg.AddTcpUdpPort(port) 7979 7980 if instance.disk_template == constants.DT_FILE: 7981 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1]) 7982 if target_node: 7983 tgt = target_node 7984 else: 7985 tgt = instance.primary_node 7986 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir) 7987 if result.fail_msg: 7988 lu.LogWarning("Could not remove directory '%s' on node %s: %s", 7989 file_storage_dir, instance.primary_node, result.fail_msg) 7990 all_result = False 7991 7992 return all_result
7993
7994 7995 -def _ComputeDiskSizePerVG(disk_template, disks):
7996 """Compute disk size requirements in the volume group 7997 7998 """ 7999 def _compute(disks, payload): 8000 """Universal algorithm. 8001 8002 """ 8003 vgs = {} 8004 for disk in disks: 8005 vgs[disk[constants.IDISK_VG]] = \ 8006 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload 8007 8008 return vgs
8009 8010 # Required free disk space as a function of disk and swap space 8011 req_size_dict = { 8012 constants.DT_DISKLESS: {}, 8013 constants.DT_PLAIN: _compute(disks, 0), 8014 # 128 MB are added for drbd metadata for each disk 8015 constants.DT_DRBD8: _compute(disks, 128), 8016 constants.DT_FILE: {}, 8017 constants.DT_SHARED_FILE: {}, 8018 } 8019 8020 if disk_template not in req_size_dict: 8021 raise errors.ProgrammerError("Disk template '%s' size requirement" 8022 " is unknown" % disk_template) 8023 8024 return req_size_dict[disk_template] 8025
8026 8027 -def _ComputeDiskSize(disk_template, disks):
8028 """Compute disk size requirements in the volume group 8029 8030 """ 8031 # Required free disk space as a function of disk and swap space 8032 req_size_dict = { 8033 constants.DT_DISKLESS: None, 8034 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks), 8035 # 128 MB are added for drbd metadata for each disk 8036 constants.DT_DRBD8: sum(d[constants.IDISK_SIZE] + 128 for d in disks), 8037 constants.DT_FILE: None, 8038 constants.DT_SHARED_FILE: 0, 8039 constants.DT_BLOCK: 0, 8040 } 8041 8042 if disk_template not in req_size_dict: 8043 raise errors.ProgrammerError("Disk template '%s' size requirement" 8044 " is unknown" % disk_template) 8045 8046 return req_size_dict[disk_template]
8047
8048 8049 -def _FilterVmNodes(lu, nodenames):
8050 """Filters out non-vm_capable nodes from a list. 8051 8052 @type lu: L{LogicalUnit} 8053 @param lu: the logical unit for which we check 8054 @type nodenames: list 8055 @param nodenames: the list of nodes on which we should check 8056 @rtype: list 8057 @return: the list of vm-capable nodes 8058 8059 """ 8060 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList()) 8061 return [name for name in nodenames if name not in vm_nodes]
8062
8063 8064 -def _CheckHVParams(lu, nodenames, hvname, hvparams):
8065 """Hypervisor parameter validation. 8066 8067 This function abstract the hypervisor parameter validation to be 8068 used in both instance create and instance modify. 8069 8070 @type lu: L{LogicalUnit} 8071 @param lu: the logical unit for which we check 8072 @type nodenames: list 8073 @param nodenames: the list of nodes on which we should check 8074 @type hvname: string 8075 @param hvname: the name of the hypervisor we should use 8076 @type hvparams: dict 8077 @param hvparams: the parameters which we need to check 8078 @raise errors.OpPrereqError: if the parameters are not valid 8079 8080 """ 8081 nodenames = _FilterVmNodes(lu, nodenames) 8082 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames, 8083 hvname, 8084 hvparams) 8085 for node in nodenames: 8086 info = hvinfo[node] 8087 if info.offline: 8088 continue 8089 info.Raise("Hypervisor parameter validation failed on node %s" % node)
8090
8091 8092 -def _CheckOSParams(lu, required, nodenames, osname, osparams):
8093 """OS parameters validation. 8094 8095 @type lu: L{LogicalUnit} 8096 @param lu: the logical unit for which we check 8097 @type required: boolean 8098 @param required: whether the validation should fail if the OS is not 8099 found 8100 @type nodenames: list 8101 @param nodenames: the list of nodes on which we should check 8102 @type osname: string 8103 @param osname: the name of the hypervisor we should use 8104 @type osparams: dict 8105 @param osparams: the parameters which we need to check 8106 @raise errors.OpPrereqError: if the parameters are not valid 8107 8108 """ 8109 nodenames = _FilterVmNodes(lu, nodenames) 8110 result = lu.rpc.call_os_validate(required, nodenames, osname, 8111 [constants.OS_VALIDATE_PARAMETERS], 8112 osparams) 8113 for node, nres in result.items(): 8114 # we don't check for offline cases since this should be run only 8115 # against the master node and/or an instance's nodes 8116 nres.Raise("OS Parameters validation failed on node %s" % node) 8117 if not nres.payload: 8118 lu.LogInfo("OS %s not found on node %s, validation skipped", 8119 osname, node)
8120
8121 8122 -class LUInstanceCreate(LogicalUnit):
8123 """Create an instance. 8124 8125 """ 8126 HPATH = "instance-add" 8127 HTYPE = constants.HTYPE_INSTANCE 8128 REQ_BGL = False 8129
8130 - def CheckArguments(self):
8131 """Check arguments. 8132 8133 """ 8134 # do not require name_check to ease forward/backward compatibility 8135 # for tools 8136 if self.op.no_install and self.op.start: 8137 self.LogInfo("No-installation mode selected, disabling startup") 8138 self.op.start = False 8139 # validate/normalize the instance name 8140 self.op.instance_name = \ 8141 netutils.Hostname.GetNormalizedName(self.op.instance_name) 8142 8143 if self.op.ip_check and not self.op.name_check: 8144 # TODO: make the ip check more flexible and not depend on the name check 8145 raise errors.OpPrereqError("Cannot do IP address check without a name" 8146 " check", errors.ECODE_INVAL) 8147 8148 # check nics' parameter names 8149 for nic in self.op.nics: 8150 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES) 8151 8152 # check disks. parameter names and consistent adopt/no-adopt strategy 8153 has_adopt = has_no_adopt = False 8154 for disk in self.op.disks: 8155 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES) 8156 if constants.IDISK_ADOPT in disk: 8157 has_adopt = True 8158 else: 8159 has_no_adopt = True 8160 if has_adopt and has_no_adopt: 8161 raise errors.OpPrereqError("Either all disks are adopted or none is", 8162 errors.ECODE_INVAL) 8163 if has_adopt: 8164 if self.op.disk_template not in constants.DTS_MAY_ADOPT: 8165 raise errors.OpPrereqError("Disk adoption is not supported for the" 8166 " '%s' disk template" % 8167 self.op.disk_template, 8168 errors.ECODE_INVAL) 8169 if self.op.iallocator is not None: 8170 raise errors.OpPrereqError("Disk adoption not allowed with an" 8171 " iallocator script", errors.ECODE_INVAL) 8172 if self.op.mode == constants.INSTANCE_IMPORT: 8173 raise errors.OpPrereqError("Disk adoption not allowed for" 8174 " instance import", errors.ECODE_INVAL) 8175 else: 8176 if self.op.disk_template in constants.DTS_MUST_ADOPT: 8177 raise errors.OpPrereqError("Disk template %s requires disk adoption," 8178 " but no 'adopt' parameter given" % 8179 self.op.disk_template, 8180 errors.ECODE_INVAL) 8181 8182 self.adopt_disks = has_adopt 8183 8184 # instance name verification 8185 if self.op.name_check: 8186 self.hostname1 = netutils.GetHostname(name=self.op.instance_name) 8187 self.op.instance_name = self.hostname1.name 8188 # used in CheckPrereq for ip ping check 8189 self.check_ip = self.hostname1.ip 8190 else: 8191 self.check_ip = None 8192 8193 # file storage checks 8194 if (self.op.file_driver and 8195 not self.op.file_driver in constants.FILE_DRIVER): 8196 raise errors.OpPrereqError("Invalid file driver name '%s'" % 8197 self.op.file_driver, errors.ECODE_INVAL) 8198 8199 if self.op.disk_template == constants.DT_FILE: 8200 opcodes.RequireFileStorage() 8201 elif self.op.disk_template == constants.DT_SHARED_FILE: 8202 opcodes.RequireSharedFileStorage() 8203 8204 ### Node/iallocator related checks 8205 _CheckIAllocatorOrNode(self, "iallocator", "pnode") 8206 8207 if self.op.pnode is not None: 8208 if self.op.disk_template in constants.DTS_INT_MIRROR: 8209 if self.op.snode is None: 8210 raise errors.OpPrereqError("The networked disk templates need" 8211 " a mirror node", errors.ECODE_INVAL) 8212 elif self.op.snode: 8213 self.LogWarning("Secondary node will be ignored on non-mirrored disk" 8214 " template") 8215 self.op.snode = None 8216 8217 self._cds = _GetClusterDomainSecret() 8218 8219 if self.op.mode == constants.INSTANCE_IMPORT: 8220 # On import force_variant must be True, because if we forced it at 8221 # initial install, our only chance when importing it back is that it 8222 # works again! 8223 self.op.force_variant = True 8224 8225 if self.op.no_install: 8226 self.LogInfo("No-installation mode has no effect during import") 8227 8228 elif self.op.mode == constants.INSTANCE_CREATE: 8229 if self.op.os_type is None: 8230 raise errors.OpPrereqError("No guest OS specified", 8231 errors.ECODE_INVAL) 8232 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os: 8233 raise errors.OpPrereqError("Guest OS '%s' is not allowed for" 8234 " installation" % self.op.os_type, 8235 errors.ECODE_STATE) 8236 if self.op.disk_template is None: 8237 raise errors.OpPrereqError("No disk template specified", 8238 errors.ECODE_INVAL) 8239 8240 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: 8241 # Check handshake to ensure both clusters have the same domain secret 8242 src_handshake = self.op.source_handshake 8243 if not src_handshake: 8244 raise errors.OpPrereqError("Missing source handshake", 8245 errors.ECODE_INVAL) 8246 8247 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds, 8248 src_handshake) 8249 if errmsg: 8250 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg, 8251 errors.ECODE_INVAL) 8252 8253 # Load and check source CA 8254 self.source_x509_ca_pem = self.op.source_x509_ca 8255 if not self.source_x509_ca_pem: 8256 raise errors.OpPrereqError("Missing source X509 CA", 8257 errors.ECODE_INVAL) 8258 8259 try: 8260 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem, 8261 self._cds) 8262 except OpenSSL.crypto.Error, err: 8263 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" % 8264 (err, ), errors.ECODE_INVAL) 8265 8266 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None) 8267 if errcode is not None: 8268 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ), 8269 errors.ECODE_INVAL) 8270 8271 self.source_x509_ca = cert 8272 8273 src_instance_name = self.op.source_instance_name 8274 if not src_instance_name: 8275 raise errors.OpPrereqError("Missing source instance name", 8276 errors.ECODE_INVAL) 8277 8278 self.source_instance_name = \ 8279 netutils.GetHostname(name=src_instance_name).name 8280 8281 else: 8282 raise errors.OpPrereqError("Invalid instance creation mode %r" % 8283 self.op.mode, errors.ECODE_INVAL)
8284
8285 - def ExpandNames(self):
8286 """ExpandNames for CreateInstance. 8287 8288 Figure out the right locks for instance creation. 8289 8290 """ 8291 self.needed_locks = {} 8292 8293 instance_name = self.op.instance_name 8294 # this is just a preventive check, but someone might still add this 8295 # instance in the meantime, and creation will fail at lock-add time 8296 if instance_name in self.cfg.GetInstanceList(): 8297 raise errors.OpPrereqError("Instance '%s' is already in the cluster" % 8298 instance_name, errors.ECODE_EXISTS) 8299 8300 self.add_locks[locking.LEVEL_INSTANCE] = instance_name 8301 8302 if self.op.iallocator: 8303 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 8304 else: 8305 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode) 8306 nodelist = [self.op.pnode] 8307 if self.op.snode is not None: 8308 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode) 8309 nodelist.append(self.op.snode) 8310 self.needed_locks[locking.LEVEL_NODE] = nodelist 8311 8312 # in case of import lock the source node too 8313 if self.op.mode == constants.INSTANCE_IMPORT: 8314 src_node = self.op.src_node 8315 src_path = self.op.src_path 8316 8317 if src_path is None: 8318 self.op.src_path = src_path = self.op.instance_name 8319 8320 if src_node is None: 8321 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 8322 self.op.src_node = None 8323 if os.path.isabs(src_path): 8324 raise errors.OpPrereqError("Importing an instance from a path" 8325 " requires a source node option", 8326 errors.ECODE_INVAL) 8327 else: 8328 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node) 8329 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET: 8330 self.needed_locks[locking.LEVEL_NODE].append(src_node) 8331 if not os.path.isabs(src_path): 8332 self.op.src_path = src_path = \ 8333 utils.PathJoin(constants.EXPORT_DIR, src_path)
8334
8335 - def _RunAllocator(self):
8336 """Run the allocator based on input opcode. 8337 8338 """ 8339 nics = [n.ToDict() for n in self.nics] 8340 ial = IAllocator(self.cfg, self.rpc, 8341 mode=constants.IALLOCATOR_MODE_ALLOC, 8342 name=self.op.instance_name, 8343 disk_template=self.op.disk_template, 8344 tags=self.op.tags, 8345 os=self.op.os_type, 8346 vcpus=self.be_full[constants.BE_VCPUS], 8347 memory=self.be_full[constants.BE_MEMORY], 8348 disks=self.disks, 8349 nics=nics, 8350 hypervisor=self.op.hypervisor, 8351 ) 8352 8353 ial.Run(self.op.iallocator) 8354 8355 if not ial.success: 8356 raise errors.OpPrereqError("Can't compute nodes using" 8357 " iallocator '%s': %s" % 8358 (self.op.iallocator, ial.info), 8359 errors.ECODE_NORES) 8360 if len(ial.result) != ial.required_nodes: 8361 raise errors.OpPrereqError("iallocator '%s' returned invalid number" 8362 " of nodes (%s), required %s" % 8363 (self.op.iallocator, len(ial.result), 8364 ial.required_nodes), errors.ECODE_FAULT) 8365 self.op.pnode = ial.result[0] 8366 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s", 8367 self.op.instance_name, self.op.iallocator, 8368 utils.CommaJoin(ial.result)) 8369 if ial.required_nodes == 2: 8370 self.op.snode = ial.result[1]
8371
8372 - def BuildHooksEnv(self):
8373 """Build hooks env. 8374 8375 This runs on master, primary and secondary nodes of the instance. 8376 8377 """ 8378 env = { 8379 "ADD_MODE": self.op.mode, 8380 } 8381 if self.op.mode == constants.INSTANCE_IMPORT: 8382 env["SRC_NODE"] = self.op.src_node 8383 env["SRC_PATH"] = self.op.src_path 8384 env["SRC_IMAGES"] = self.src_images 8385 8386 env.update(_BuildInstanceHookEnv( 8387 name=self.op.instance_name, 8388 primary_node=self.op.pnode, 8389 secondary_nodes=self.secondaries, 8390 status=self.op.start, 8391 os_type=self.op.os_type, 8392 memory=self.be_full[constants.BE_MEMORY], 8393 vcpus=self.be_full[constants.BE_VCPUS], 8394 nics=_NICListToTuple(self, self.nics), 8395 disk_template=self.op.disk_template, 8396 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE]) 8397 for d in self.disks], 8398 bep=self.be_full, 8399 hvp=self.hv_full, 8400 hypervisor_name=self.op.hypervisor, 8401 tags=self.op.tags, 8402 )) 8403 8404 return env
8405
8406 - def BuildHooksNodes(self):
8407 """Build hooks nodes. 8408 8409 """ 8410 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries 8411 return nl, nl
8412
8413 - def _ReadExportInfo(self):
8414 """Reads the export information from disk. 8415 8416 It will override the opcode source node and path with the actual 8417 information, if these two were not specified before. 8418 8419 @return: the export information 8420 8421 """ 8422 assert self.op.mode == constants.INSTANCE_IMPORT 8423 8424 src_node = self.op.src_node 8425 src_path = self.op.src_path 8426 8427 if src_node is None: 8428 locked_nodes = self.owned_locks(locking.LEVEL_NODE) 8429 exp_list = self.rpc.call_export_list(locked_nodes) 8430 found = False 8431 for node in exp_list: 8432 if exp_list[node].fail_msg: 8433 continue 8434 if src_path in exp_list[node].payload: 8435 found = True 8436 self.op.src_node = src_node = node 8437 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR, 8438 src_path) 8439 break 8440 if not found: 8441 raise errors.OpPrereqError("No export found for relative path %s" % 8442 src_path, errors.ECODE_INVAL) 8443 8444 _CheckNodeOnline(self, src_node) 8445 result = self.rpc.call_export_info(src_node, src_path) 8446 result.Raise("No export or invalid export found in dir %s" % src_path) 8447 8448 export_info = objects.SerializableConfigParser.Loads(str(result.payload)) 8449 if not export_info.has_section(constants.INISECT_EXP): 8450 raise errors.ProgrammerError("Corrupted export config", 8451 errors.ECODE_ENVIRON) 8452 8453 ei_version = export_info.get(constants.INISECT_EXP, "version") 8454 if (int(ei_version) != constants.EXPORT_VERSION): 8455 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" % 8456 (ei_version, constants.EXPORT_VERSION), 8457 errors.ECODE_ENVIRON) 8458 return export_info
8459
8460 - def _ReadExportParams(self, einfo):
8461 """Use export parameters as defaults. 8462 8463 In case the opcode doesn't specify (as in override) some instance 8464 parameters, then try to use them from the export information, if 8465 that declares them. 8466 8467 """ 8468 self.op.os_type = einfo.get(constants.INISECT_EXP, "os") 8469 8470 if self.op.disk_template is None: 8471 if einfo.has_option(constants.INISECT_INS, "disk_template"): 8472 self.op.disk_template = einfo.get(constants.INISECT_INS, 8473 "disk_template") 8474 else: 8475 raise errors.OpPrereqError("No disk template specified and the export" 8476 " is missing the disk_template information", 8477 errors.ECODE_INVAL) 8478 8479 if not self.op.disks: 8480 if einfo.has_option(constants.INISECT_INS, "disk_count"): 8481 disks = [] 8482 # TODO: import the disk iv_name too 8483 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")): 8484 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx) 8485 disks.append({constants.IDISK_SIZE: disk_sz}) 8486 self.op.disks = disks 8487 else: 8488 raise errors.OpPrereqError("No disk info specified and the export" 8489 " is missing the disk information", 8490 errors.ECODE_INVAL) 8491 8492 if (not self.op.nics and 8493 einfo.has_option(constants.INISECT_INS, "nic_count")): 8494 nics = [] 8495 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")): 8496 ndict = {} 8497 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]: 8498 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name)) 8499 ndict[name] = v 8500 nics.append(ndict) 8501 self.op.nics = nics 8502 8503 if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"): 8504 self.op.tags = einfo.get(constants.INISECT_INS, "tags").split() 8505 8506 if (self.op.hypervisor is None and 8507 einfo.has_option(constants.INISECT_INS, "hypervisor")): 8508 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor") 8509 8510 if einfo.has_section(constants.INISECT_HYP): 8511 # use the export parameters but do not override the ones 8512 # specified by the user 8513 for name, value in einfo.items(constants.INISECT_HYP): 8514 if name not in self.op.hvparams: 8515 self.op.hvparams[name] = value 8516 8517 if einfo.has_section(constants.INISECT_BEP): 8518 # use the parameters, without overriding 8519 for name, value in einfo.items(constants.INISECT_BEP): 8520 if name not in self.op.beparams: 8521 self.op.beparams[name] = value 8522 else: 8523 # try to read the parameters old style, from the main section 8524 for name in constants.BES_PARAMETERS: 8525 if (name not in self.op.beparams and 8526 einfo.has_option(constants.INISECT_INS, name)): 8527 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name) 8528 8529 if einfo.has_section(constants.INISECT_OSP): 8530 # use the parameters, without overriding 8531 for name, value in einfo.items(constants.INISECT_OSP): 8532 if name not in self.op.osparams: 8533 self.op.osparams[name] = value
8534
8535 - def _RevertToDefaults(self, cluster):
8536 """Revert the instance parameters to the default values. 8537 8538 """ 8539 # hvparams 8540 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {}) 8541 for name in self.op.hvparams.keys(): 8542 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]: 8543 del self.op.hvparams[name] 8544 # beparams 8545 be_defs = cluster.SimpleFillBE({}) 8546 for name in self.op.beparams.keys(): 8547 if name in be_defs and be_defs[name] == self.op.beparams[name]: 8548 del self.op.beparams[name] 8549 # nic params 8550 nic_defs = cluster.SimpleFillNIC({}) 8551 for nic in self.op.nics: 8552 for name in constants.NICS_PARAMETERS: 8553 if name in nic and name in nic_defs and nic[name] == nic_defs[name]: 8554 del nic[name] 8555 # osparams 8556 os_defs = cluster.SimpleFillOS(self.op.os_type, {}) 8557 for name in self.op.osparams.keys(): 8558 if name in os_defs and os_defs[name] == self.op.osparams[name]: 8559 del self.op.osparams[name]
8560
8561 - def _CalculateFileStorageDir(self):
8562 """Calculate final instance file storage dir. 8563 8564 """ 8565 # file storage dir calculation/check 8566 self.instance_file_storage_dir = None 8567 if self.op.disk_template in constants.DTS_FILEBASED: 8568 # build the full file storage dir path 8569 joinargs = [] 8570 8571 if self.op.disk_template == constants.DT_SHARED_FILE: 8572 get_fsd_fn = self.cfg.GetSharedFileStorageDir 8573 else: 8574 get_fsd_fn = self.cfg.GetFileStorageDir 8575 8576 cfg_storagedir = get_fsd_fn() 8577 if not cfg_storagedir: 8578 raise errors.OpPrereqError("Cluster file storage dir not defined") 8579 joinargs.append(cfg_storagedir) 8580 8581 if self.op.file_storage_dir is not None: 8582 joinargs.append(self.op.file_storage_dir) 8583 8584 joinargs.append(self.op.instance_name) 8585 8586 # pylint: disable=W0142 8587 self.instance_file_storage_dir = utils.PathJoin(*joinargs)
8588
8589 - def CheckPrereq(self):
8590 """Check prerequisites. 8591 8592 """ 8593 self._CalculateFileStorageDir() 8594 8595 if self.op.mode == constants.INSTANCE_IMPORT: 8596 export_info = self._ReadExportInfo() 8597 self._ReadExportParams(export_info) 8598 8599 if (not self.cfg.GetVGName() and 8600 self.op.disk_template not in constants.DTS_NOT_LVM): 8601 raise errors.OpPrereqError("Cluster does not support lvm-based" 8602 " instances", errors.ECODE_STATE) 8603 8604 if self.op.hypervisor is None: 8605 self.op.hypervisor = self.cfg.GetHypervisorType() 8606 8607 cluster = self.cfg.GetClusterInfo() 8608 enabled_hvs = cluster.enabled_hypervisors 8609 if self.op.hypervisor not in enabled_hvs: 8610 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the" 8611 " cluster (%s)" % (self.op.hypervisor, 8612 ",".join(enabled_hvs)), 8613 errors.ECODE_STATE) 8614 8615 # Check tag validity 8616 for tag in self.op.tags: 8617 objects.TaggableObject.ValidateTag(tag) 8618 8619 # check hypervisor parameter syntax (locally) 8620 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES) 8621 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, 8622 self.op.hvparams) 8623 hv_type = hypervisor.GetHypervisor(self.op.hypervisor) 8624 hv_type.CheckParameterSyntax(filled_hvp) 8625 self.hv_full = filled_hvp 8626 # check that we don't specify global parameters on an instance 8627 _CheckGlobalHvParams(self.op.hvparams) 8628 8629 # fill and remember the beparams dict 8630 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES) 8631 self.be_full = cluster.SimpleFillBE(self.op.beparams) 8632 8633 # build os parameters 8634 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams) 8635 8636 # now that hvp/bep are in final format, let's reset to defaults, 8637 # if told to do so 8638 if self.op.identify_defaults: 8639 self._RevertToDefaults(cluster) 8640 8641 # NIC buildup 8642 self.nics = [] 8643 for idx, nic in enumerate(self.op.nics): 8644 nic_mode_req = nic.get(constants.INIC_MODE, None) 8645 nic_mode = nic_mode_req 8646 if nic_mode is None: 8647 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE] 8648 8649 # in routed mode, for the first nic, the default ip is 'auto' 8650 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0: 8651 default_ip_mode = constants.VALUE_AUTO 8652 else: 8653 default_ip_mode = constants.VALUE_NONE 8654 8655 # ip validity checks 8656 ip = nic.get(constants.INIC_IP, default_ip_mode) 8657 if ip is None or ip.lower() == constants.VALUE_NONE: 8658 nic_ip = None 8659 elif ip.lower() == constants.VALUE_AUTO: 8660 if not self.op.name_check: 8661 raise errors.OpPrereqError("IP address set to auto but name checks" 8662 " have been skipped", 8663 errors.ECODE_INVAL) 8664 nic_ip = self.hostname1.ip 8665 else: 8666 if not netutils.IPAddress.IsValid(ip): 8667 raise errors.OpPrereqError("Invalid IP address '%s'" % ip, 8668 errors.ECODE_INVAL) 8669 nic_ip = ip 8670 8671 # TODO: check the ip address for uniqueness 8672 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip: 8673 raise errors.OpPrereqError("Routed nic mode requires an ip address", 8674 errors.ECODE_INVAL) 8675 8676 # MAC address verification 8677 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO) 8678 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE): 8679 mac = utils.NormalizeAndValidateMac(mac) 8680 8681 try: 8682 self.cfg.ReserveMAC(mac, self.proc.GetECId()) 8683 except errors.ReservationError: 8684 raise errors.OpPrereqError("MAC address %s already in use" 8685 " in cluster" % mac, 8686 errors.ECODE_NOTUNIQUE) 8687 8688 # Build nic parameters 8689 link = nic.get(constants.INIC_LINK, None) 8690 nicparams = {} 8691 if nic_mode_req: 8692 nicparams[constants.NIC_MODE] = nic_mode_req 8693 if link: 8694 nicparams[constants.NIC_LINK] = link 8695 8696 check_params = cluster.SimpleFillNIC(nicparams) 8697 objects.NIC.CheckParameterSyntax(check_params) 8698 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams)) 8699 8700 # disk checks/pre-build 8701 default_vg = self.cfg.GetVGName() 8702 self.disks = [] 8703 for disk in self.op.disks: 8704 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR) 8705 if mode not in constants.DISK_ACCESS_SET: 8706 raise errors.OpPrereqError("Invalid disk access mode '%s'" % 8707 mode, errors.ECODE_INVAL) 8708 size = disk.get(constants.IDISK_SIZE, None) 8709 if size is None: 8710 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL) 8711 try: 8712 size = int(size) 8713 except (TypeError, ValueError): 8714 raise errors.OpPrereqError("Invalid disk size '%s'" % size, 8715 errors.ECODE_INVAL) 8716 8717 data_vg = disk.get(constants.IDISK_VG, default_vg) 8718 new_disk = { 8719 constants.IDISK_SIZE: size, 8720 constants.IDISK_MODE: mode, 8721 constants.IDISK_VG: data_vg, 8722 constants.IDISK_METAVG: disk.get(constants.IDISK_METAVG, data_vg), 8723 } 8724 if constants.IDISK_ADOPT in disk: 8725 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT] 8726 self.disks.append(new_disk) 8727 8728 if self.op.mode == constants.INSTANCE_IMPORT: 8729 8730 # Check that the new instance doesn't have less disks than the export 8731 instance_disks = len(self.disks) 8732 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count') 8733 if instance_disks < export_disks: 8734 raise errors.OpPrereqError("Not enough disks to import." 8735 " (instance: %d, export: %d)" % 8736 (instance_disks, export_disks), 8737 errors.ECODE_INVAL) 8738 8739 disk_images = [] 8740 for idx in range(export_disks): 8741 option = "disk%d_dump" % idx 8742 if export_info.has_option(constants.INISECT_INS, option): 8743 # FIXME: are the old os-es, disk sizes, etc. useful? 8744 export_name = export_info.get(constants.INISECT_INS, option) 8745 image = utils.PathJoin(self.op.src_path, export_name) 8746 disk_images.append(image) 8747 else: 8748 disk_images.append(False) 8749 8750 self.src_images = disk_images 8751 8752 old_name = export_info.get(constants.INISECT_INS, "name") 8753 try: 8754 exp_nic_count = export_info.getint(constants.INISECT_INS, "nic_count") 8755 except (TypeError, ValueError), err: 8756 raise errors.OpPrereqError("Invalid export file, nic_count is not" 8757 " an integer: %s" % str(err), 8758 errors.ECODE_STATE) 8759 if self.op.instance_name == old_name: 8760 for idx, nic in enumerate(self.nics): 8761 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx: 8762 nic_mac_ini = "nic%d_mac" % idx 8763 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini) 8764 8765 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT 8766 8767 # ip ping checks (we use the same ip that was resolved in ExpandNames) 8768 if self.op.ip_check: 8769 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT): 8770 raise errors.OpPrereqError("IP %s of instance %s already in use" % 8771 (self.check_ip, self.op.instance_name), 8772 errors.ECODE_NOTUNIQUE) 8773 8774 #### mac address generation 8775 # By generating here the mac address both the allocator and the hooks get 8776 # the real final mac address rather than the 'auto' or 'generate' value. 8777 # There is a race condition between the generation and the instance object 8778 # creation, which means that we know the mac is valid now, but we're not 8779 # sure it will be when we actually add the instance. If things go bad 8780 # adding the instance will abort because of a duplicate mac, and the 8781 # creation job will fail. 8782 for nic in self.nics: 8783 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE): 8784 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId()) 8785 8786 #### allocator run 8787 8788 if self.op.iallocator is not None: 8789 self._RunAllocator() 8790 8791 #### node related checks 8792 8793 # check primary node 8794 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode) 8795 assert self.pnode is not None, \ 8796 "Cannot retrieve locked node %s" % self.op.pnode 8797 if pnode.offline: 8798 raise errors.OpPrereqError("Cannot use offline primary node '%s'" % 8799 pnode.name, errors.ECODE_STATE) 8800 if pnode.drained: 8801 raise errors.OpPrereqError("Cannot use drained primary node '%s'" % 8802 pnode.name, errors.ECODE_STATE) 8803 if not pnode.vm_capable: 8804 raise errors.OpPrereqError("Cannot use non-vm_capable primary node" 8805 " '%s'" % pnode.name, errors.ECODE_STATE) 8806 8807 self.secondaries = [] 8808 8809 # mirror node verification 8810 if self.op.disk_template in constants.DTS_INT_MIRROR: 8811 if self.op.snode == pnode.name: 8812 raise errors.OpPrereqError("The secondary node cannot be the" 8813 " primary node", errors.ECODE_INVAL) 8814 _CheckNodeOnline(self, self.op.snode) 8815 _CheckNodeNotDrained(self, self.op.snode) 8816 _CheckNodeVmCapable(self, self.op.snode) 8817 self.secondaries.append(self.op.snode) 8818 8819 nodenames = [pnode.name] + self.secondaries 8820 8821 if not self.adopt_disks: 8822 # Check lv size requirements, if not adopting 8823 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks) 8824 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes) 8825 8826 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data 8827 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG], 8828 disk[constants.IDISK_ADOPT]) 8829 for disk in self.disks]) 8830 if len(all_lvs) != len(self.disks): 8831 raise errors.OpPrereqError("Duplicate volume names given for adoption", 8832 errors.ECODE_INVAL) 8833 for lv_name in all_lvs: 8834 try: 8835 # FIXME: lv_name here is "vg/lv" need to ensure that other calls 8836 # to ReserveLV uses the same syntax 8837 self.cfg.ReserveLV(lv_name, self.proc.GetECId()) 8838 except errors.ReservationError: 8839 raise errors.OpPrereqError("LV named %s used by another instance" % 8840 lv_name, errors.ECODE_NOTUNIQUE) 8841 8842 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name] 8843 vg_names.Raise("Cannot get VG information from node %s" % pnode.name) 8844 8845 node_lvs = self.rpc.call_lv_list([pnode.name], 8846 vg_names.payload.keys())[pnode.name] 8847 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name) 8848 node_lvs = node_lvs.payload 8849 8850 delta = all_lvs.difference(node_lvs.keys()) 8851 if delta: 8852 raise errors.OpPrereqError("Missing logical volume(s): %s" % 8853 utils.CommaJoin(delta), 8854 errors.ECODE_INVAL) 8855 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]] 8856 if online_lvs: 8857 raise errors.OpPrereqError("Online logical volumes found, cannot" 8858 " adopt: %s" % utils.CommaJoin(online_lvs), 8859 errors.ECODE_STATE) 8860 # update the size of disk based on what is found 8861 for dsk in self.disks: 8862 dsk[constants.IDISK_SIZE] = \ 8863 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG], 8864 dsk[constants.IDISK_ADOPT])][0])) 8865 8866 elif self.op.disk_template == constants.DT_BLOCK: 8867 # Normalize and de-duplicate device paths 8868 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT]) 8869 for disk in self.disks]) 8870 if len(all_disks) != len(self.disks): 8871 raise errors.OpPrereqError("Duplicate disk names given for adoption", 8872 errors.ECODE_INVAL) 8873 baddisks = [d for d in all_disks 8874 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)] 8875 if baddisks: 8876 raise errors.OpPrereqError("Device node(s) %s lie outside %s and" 8877 " cannot be adopted" % 8878 (", ".join(baddisks), 8879 constants.ADOPTABLE_BLOCKDEV_ROOT), 8880 errors.ECODE_INVAL) 8881 8882 node_disks = self.rpc.call_bdev_sizes([pnode.name], 8883 list(all_disks))[pnode.name] 8884 node_disks.Raise("Cannot get block device information from node %s" % 8885 pnode.name) 8886 node_disks = node_disks.payload 8887 delta = all_disks.difference(node_disks.keys()) 8888 if delta: 8889 raise errors.OpPrereqError("Missing block device(s): %s" % 8890 utils.CommaJoin(delta), 8891 errors.ECODE_INVAL) 8892 for dsk in self.disks: 8893 dsk[constants.IDISK_SIZE] = \ 8894 int(float(node_disks[dsk[constants.IDISK_ADOPT]])) 8895 8896 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams) 8897 8898 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant) 8899 # check OS parameters (remotely) 8900 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full) 8901 8902 _CheckNicsBridgesExist(self, self.nics, self.pnode.name) 8903 8904 # memory check on primary node 8905 if self.op.start: 8906 _CheckNodeFreeMemory(self, self.pnode.name, 8907 "creating instance %s" % self.op.instance_name, 8908 self.be_full[constants.BE_MEMORY], 8909 self.op.hypervisor) 8910 8911 self.dry_run_result = list(nodenames)
8912
8913 - def Exec(self, feedback_fn):
8914 """Create and add the instance to the cluster. 8915 8916 """ 8917 instance = self.op.instance_name 8918 pnode_name = self.pnode.name 8919 8920 ht_kind = self.op.hypervisor 8921 if ht_kind in constants.HTS_REQ_PORT: 8922 network_port = self.cfg.AllocatePort() 8923 else: 8924 network_port = None 8925 8926 disks = _GenerateDiskTemplate(self, 8927 self.op.disk_template, 8928 instance, pnode_name, 8929 self.secondaries, 8930 self.disks, 8931 self.instance_file_storage_dir, 8932 self.op.file_driver, 8933 0, 8934 feedback_fn) 8935 8936 iobj = objects.Instance(name=instance, os=self.op.os_type, 8937 primary_node=pnode_name, 8938 nics=self.nics, disks=disks, 8939 disk_template=self.op.disk_template, 8940 admin_up=False, 8941 network_port=network_port, 8942 beparams=self.op.beparams, 8943 hvparams=self.op.hvparams, 8944 hypervisor=self.op.hypervisor, 8945 osparams=self.op.osparams, 8946 ) 8947 8948 if self.op.tags: 8949 for tag in self.op.tags: 8950 iobj.AddTag(tag) 8951 8952 if self.adopt_disks: 8953 if self.op.disk_template == constants.DT_PLAIN: 8954 # rename LVs to the newly-generated names; we need to construct 8955 # 'fake' LV disks with the old data, plus the new unique_id 8956 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks] 8957 rename_to = [] 8958 for t_dsk, a_dsk in zip(tmp_disks, self.disks): 8959 rename_to.append(t_dsk.logical_id) 8960 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT]) 8961 self.cfg.SetDiskID(t_dsk, pnode_name) 8962 result = self.rpc.call_blockdev_rename(pnode_name, 8963 zip(tmp_disks, rename_to)) 8964 result.Raise("Failed to rename adoped LVs") 8965 else: 8966 feedback_fn("* creating instance disks...") 8967 try: 8968 _CreateDisks(self, iobj) 8969 except errors.OpExecError: 8970 self.LogWarning("Device creation failed, reverting...") 8971 try: 8972 _RemoveDisks(self, iobj) 8973 finally: 8974 self.cfg.ReleaseDRBDMinors(instance) 8975 raise 8976 8977 feedback_fn("adding instance %s to cluster config" % instance) 8978 8979 self.cfg.AddInstance(iobj, self.proc.GetECId()) 8980 8981 # Declare that we don't want to remove the instance lock anymore, as we've 8982 # added the instance to the config 8983 del self.remove_locks[locking.LEVEL_INSTANCE] 8984 8985 if self.op.mode == constants.INSTANCE_IMPORT: 8986 # Release unused nodes 8987 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node]) 8988 else: 8989 # Release all nodes 8990 _ReleaseLocks(self, locking.LEVEL_NODE) 8991 8992 disk_abort = False 8993 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks: 8994 feedback_fn("* wiping instance disks...") 8995 try: 8996 _WipeDisks(self, iobj) 8997 except errors.OpExecError, err: 8998 logging.exception("Wiping disks failed") 8999 self.LogWarning("Wiping instance disks failed (%s)", err) 9000 disk_abort = True 9001 9002 if disk_abort: 9003 # Something is already wrong with the disks, don't do anything else 9004 pass 9005 elif self.op.wait_for_sync: 9006 disk_abort = not _WaitForSync(self, iobj) 9007 elif iobj.disk_template in constants.DTS_INT_MIRROR: 9008 # make sure the disks are not degraded (still sync-ing is ok) 9009 feedback_fn("* checking mirrors status") 9010 disk_abort = not _WaitForSync(self, iobj, oneshot=True) 9011 else: 9012 disk_abort = False 9013 9014 if disk_abort: 9015 _RemoveDisks(self, iobj) 9016 self.cfg.RemoveInstance(iobj.name) 9017 # Make sure the instance lock gets removed 9018 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name 9019 raise errors.OpExecError("There are some degraded disks for" 9020 " this instance") 9021 9022 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks: 9023 if self.op.mode == constants.INSTANCE_CREATE: 9024 if not self.op.no_install: 9025 pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and 9026 not self.op.wait_for_sync) 9027 if pause_sync: 9028 feedback_fn("* pausing disk sync to install instance OS") 9029 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name, 9030 iobj.disks, True) 9031 for idx, success in enumerate(result.payload): 9032 if not success: 9033 logging.warn("pause-sync of instance %s for disk %d failed", 9034 instance, idx) 9035 9036 feedback_fn("* running the instance OS create scripts...") 9037 # FIXME: pass debug option from opcode to backend 9038 result = self.rpc.call_instance_os_add(pnode_name, iobj, False, 9039 self.op.debug_level) 9040 if pause_sync: 9041 feedback_fn("* resuming disk sync") 9042 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name, 9043 iobj.disks, False) 9044 for idx, success in enumerate(result.payload): 9045 if not success: 9046 logging.warn("resume-sync of instance %s for disk %d failed", 9047 instance, idx) 9048 9049 result.Raise("Could not add os for instance %s" 9050 " on node %s" % (instance, pnode_name)) 9051 9052 elif self.op.mode == constants.INSTANCE_IMPORT: 9053 feedback_fn("* running the instance OS import scripts...") 9054 9055 transfers = [] 9056 9057 for idx, image in enumerate(self.src_images): 9058 if not image: 9059 continue 9060 9061 # FIXME: pass debug option from opcode to backend 9062 dt = masterd.instance.DiskTransfer("disk/%s" % idx, 9063 constants.IEIO_FILE, (image, ), 9064 constants.IEIO_SCRIPT, 9065 (iobj.disks[idx], idx), 9066 None) 9067 transfers.append(dt) 9068 9069 import_result = \ 9070 masterd.instance.TransferInstanceData(self, feedback_fn, 9071 self.op.src_node, pnode_name, 9072 self.pnode.secondary_ip, 9073 iobj, transfers) 9074 if not compat.all(import_result): 9075 self.LogWarning("Some disks for instance %s on node %s were not" 9076 " imported successfully" % (instance, pnode_name)) 9077 9078 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT: 9079 feedback_fn("* preparing remote import...") 9080 # The source cluster will stop the instance before attempting to make a 9081 # connection. In some cases stopping an instance can take a long time, 9082 # hence the shutdown timeout is added to the connection timeout. 9083 connect_timeout = (constants.RIE_CONNECT_TIMEOUT + 9084 self.op.source_shutdown_timeout) 9085 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout) 9086 9087 assert iobj.primary_node == self.pnode.name 9088 disk_results = \ 9089 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode, 9090 self.source_x509_ca, 9091 self._cds, timeouts) 9092 if not compat.all(disk_results): 9093 # TODO: Should the instance still be started, even if some disks 9094 # failed to import (valid for local imports, too)? 9095 self.LogWarning("Some disks for instance %s on node %s were not" 9096 " imported successfully" % (instance, pnode_name)) 9097 9098 # Run rename script on newly imported instance 9099 assert iobj.name == instance 9100 feedback_fn("Running rename script for %s" % instance) 9101 result = self.rpc.call_instance_run_rename(pnode_name, iobj, 9102 self.source_instance_name, 9103 self.op.debug_level) 9104 if result.fail_msg: 9105 self.LogWarning("Failed to run rename script for %s on node" 9106 " %s: %s" % (instance, pnode_name, result.fail_msg)) 9107 9108 else: 9109 # also checked in the prereq part 9110 raise errors.ProgrammerError("Unknown OS initialization mode '%s'" 9111 % self.op.mode) 9112 9113 if self.op.start: 9114 iobj.admin_up = True 9115 self.cfg.Update(iobj, feedback_fn) 9116 logging.info("Starting instance %s on node %s", instance, pnode_name) 9117 feedback_fn("* starting instance...") 9118 result = self.rpc.call_instance_start(pnode_name, iobj, 9119 None, None, False) 9120 result.Raise("Could not start instance") 9121 9122 return list(iobj.all_nodes)
9123
9124 9125 -class LUInstanceConsole(NoHooksLU):
9126 """Connect to an instance's console. 9127 9128 This is somewhat special in that it returns the command line that 9129 you need to run on the master node in order to connect to the 9130 console. 9131 9132 """ 9133 REQ_BGL = False 9134
9135 - def ExpandNames(self):
9137
9138 - def CheckPrereq(self):
9139 """Check prerequisites. 9140 9141 This checks that the instance is in the cluster. 9142 9143 """ 9144 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 9145 assert self.instance is not None, \ 9146 "Cannot retrieve locked instance %s" % self.op.instance_name 9147 _CheckNodeOnline(self, self.instance.primary_node)
9148
9149 - def Exec(self, feedback_fn):
9150 """Connect to the console of an instance 9151 9152 """ 9153 instance = self.instance 9154 node = instance.primary_node 9155 9156 node_insts = self.rpc.call_instance_list([node], 9157 [instance.hypervisor])[node] 9158 node_insts.Raise("Can't get node information from %s" % node) 9159 9160 if instance.name not in node_insts.payload: 9161 if instance.admin_up: 9162 state = constants.INSTST_ERRORDOWN 9163 else: 9164 state = constants.INSTST_ADMINDOWN 9165 raise errors.OpExecError("Instance %s is not running (state %s)" % 9166 (instance.name, state)) 9167 9168 logging.debug("Connecting to console of %s on %s", instance.name, node) 9169 9170 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
9171
9172 9173 -def _GetInstanceConsole(cluster, instance):
9174 """Returns console information for an instance. 9175 9176 @type cluster: L{objects.Cluster} 9177 @type instance: L{objects.Instance} 9178 @rtype: dict 9179 9180 """ 9181 hyper = hypervisor.GetHypervisor(instance.hypervisor) 9182 # beparams and hvparams are passed separately, to avoid editing the 9183 # instance and then saving the defaults in the instance itself. 9184 hvparams = cluster.FillHV(instance) 9185 beparams = cluster.FillBE(instance) 9186 console = hyper.GetInstanceConsole(instance, hvparams, beparams) 9187 9188 assert console.instance == instance.name 9189 assert console.Validate() 9190 9191 return console.ToDict()
9192
9193 9194 -class LUInstanceReplaceDisks(LogicalUnit):
9195 """Replace the disks of an instance. 9196 9197 """ 9198 HPATH = "mirrors-replace" 9199 HTYPE = constants.HTYPE_INSTANCE 9200 REQ_BGL = False 9201
9202 - def CheckArguments(self):
9203 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node, 9204 self.op.iallocator)
9205
9206 - def ExpandNames(self):
9207 self._ExpandAndLockInstance() 9208 9209 assert locking.LEVEL_NODE not in self.needed_locks 9210 assert locking.LEVEL_NODEGROUP not in self.needed_locks 9211 9212 assert self.op.iallocator is None or self.op.remote_node is None, \ 9213 "Conflicting options" 9214 9215 if self.op.remote_node is not None: 9216 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) 9217 9218 # Warning: do not remove the locking of the new secondary here 9219 # unless DRBD8.AddChildren is changed to work in parallel; 9220 # currently it doesn't since parallel invocations of 9221 # FindUnusedMinor will conflict 9222 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node] 9223 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND 9224 else: 9225 self.needed_locks[locking.LEVEL_NODE] = [] 9226 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE 9227 9228 if self.op.iallocator is not None: 9229 # iallocator will select a new node in the same group 9230 self.needed_locks[locking.LEVEL_NODEGROUP] = [] 9231 9232 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode, 9233 self.op.iallocator, self.op.remote_node, 9234 self.op.disks, False, self.op.early_release) 9235 9236 self.tasklets = [self.replacer]
9237
9238 - def DeclareLocks(self, level):
9239 if level == locking.LEVEL_NODEGROUP: 9240 assert self.op.remote_node is None 9241 assert self.op.iallocator is not None 9242 assert not self.needed_locks[locking.LEVEL_NODEGROUP] 9243 9244 self.share_locks[locking.LEVEL_NODEGROUP] = 1 9245 self.needed_locks[locking.LEVEL_NODEGROUP] = \ 9246 self.cfg.GetInstanceNodeGroups(self.op.instance_name) 9247 9248 elif level == locking.LEVEL_NODE: 9249 if self.op.iallocator is not None: 9250 assert self.op.remote_node is None 9251 assert not self.needed_locks[locking.LEVEL_NODE] 9252 9253 # Lock member nodes of all locked groups 9254 self.needed_locks[locking.LEVEL_NODE] = [node_name 9255 for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP) 9256 for node_name in self.cfg.GetNodeGroup(group_uuid).members] 9257 else: 9258 self._LockInstancesNodes()
9259
9260 - def BuildHooksEnv(self):
9261 """Build hooks env. 9262 9263 This runs on the master, the primary and all the secondaries. 9264 9265 """ 9266 instance = self.replacer.instance 9267 env = { 9268 "MODE": self.op.mode, 9269 "NEW_SECONDARY": self.op.remote_node, 9270 "OLD_SECONDARY": instance.secondary_nodes[0], 9271 } 9272 env.update(_BuildInstanceHookEnvByObject(self, instance)) 9273 return env
9274
9275 - def BuildHooksNodes(self):
9276 """Build hooks nodes. 9277 9278 """ 9279 instance = self.replacer.instance 9280 nl = [ 9281 self.cfg.GetMasterNode(), 9282 instance.primary_node, 9283 ] 9284 if self.op.remote_node is not None: 9285 nl.append(self.op.remote_node) 9286 return nl, nl
9287
9288 - def CheckPrereq(self):
9289 """Check prerequisites. 9290 9291 """ 9292 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or 9293 self.op.iallocator is None) 9294 9295 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP) 9296 if owned_groups: 9297 _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups) 9298 9299 return LogicalUnit.CheckPrereq(self)
9300
9301 9302 -class TLReplaceDisks(Tasklet):
9303 """Replaces disks for an instance. 9304 9305 Note: Locking is not within the scope of this class. 9306 9307 """
9308 - def __init__(self, lu, instance_name, mode, iallocator_name, remote_node, 9309 disks, delay_iallocator, early_release):
9310 """Initializes this class. 9311 9312 """ 9313 Tasklet.__init__(self, lu) 9314 9315 # Parameters 9316 self.instance_name = instance_name 9317 self.mode = mode 9318 self.iallocator_name = iallocator_name 9319 self.remote_node = remote_node 9320 self.disks = disks 9321 self.delay_iallocator = delay_iallocator 9322 self.early_release = early_release 9323 9324 # Runtime data 9325 self.instance = None 9326 self.new_node = None 9327 self.target_node = None 9328 self.other_node = None 9329 self.remote_node_info = None 9330 self.node_secondary_ip = None
9331 9332 @staticmethod
9333 - def CheckArguments(mode, remote_node, iallocator):
9334 """Helper function for users of this class. 9335 9336 """ 9337 # check for valid parameter combination 9338 if mode == constants.REPLACE_DISK_CHG: 9339 if remote_node is None and iallocator is None: 9340 raise errors.OpPrereqError("When changing the secondary either an" 9341 " iallocator script must be used or the" 9342 " new node given", errors.ECODE_INVAL) 9343 9344 if remote_node is not None and iallocator is not None: 9345 raise errors.OpPrereqError("Give either the iallocator or the new" 9346 " secondary, not both", errors.ECODE_INVAL) 9347 9348 elif remote_node is not None or iallocator is not None: 9349 # Not replacing the secondary 9350 raise errors.OpPrereqError("The iallocator and new node options can" 9351 " only be used when changing the" 9352 " secondary node", errors.ECODE_INVAL)
9353 9354 @staticmethod
9355 - def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
9356 """Compute a new secondary node using an IAllocator. 9357 9358 """ 9359 ial = IAllocator(lu.cfg, lu.rpc, 9360 mode=constants.IALLOCATOR_MODE_RELOC, 9361 name=instance_name, 9362 relocate_from=list(relocate_from)) 9363 9364 ial.Run(iallocator_name) 9365 9366 if not ial.success: 9367 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':" 9368 " %s" % (iallocator_name, ial.info), 9369 errors.ECODE_NORES) 9370 9371 if len(ial.result) != ial.required_nodes: 9372 raise errors.OpPrereqError("iallocator '%s' returned invalid number" 9373 " of nodes (%s), required %s" % 9374 (iallocator_name, 9375 len(ial.result), ial.required_nodes), 9376 errors.ECODE_FAULT) 9377 9378 remote_node_name = ial.result[0] 9379 9380 lu.LogInfo("Selected new secondary for instance '%s': %s", 9381 instance_name, remote_node_name) 9382 9383 return remote_node_name
9384
9385 - def _FindFaultyDisks(self, node_name):
9386 """Wrapper for L{_FindFaultyInstanceDisks}. 9387 9388 """ 9389 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance, 9390 node_name, True)
9391
9392 - def _CheckDisksActivated(self, instance):
9393 """Checks if the instance disks are activated. 9394 9395 @param instance: The instance to check disks 9396 @return: True if they are activated, False otherwise 9397 9398 """ 9399 nodes = instance.all_nodes 9400 9401 for idx, dev in enumerate(instance.disks): 9402 for node in nodes: 9403 self.lu.LogInfo("Checking disk/%d on %s", idx, node) 9404 self.cfg.SetDiskID(dev, node) 9405 9406 result = self.rpc.call_blockdev_find(node, dev) 9407 9408 if result.offline: 9409 continue 9410 elif result.fail_msg or not result.payload: 9411 return False 9412 9413 return True
9414
9415 - def CheckPrereq(self):
9416 """Check prerequisites. 9417 9418 This checks that the instance is in the cluster. 9419 9420 """ 9421 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name) 9422 assert instance is not None, \ 9423 "Cannot retrieve locked instance %s" % self.instance_name 9424 9425 if instance.disk_template != constants.DT_DRBD8: 9426 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based" 9427 " instances", errors.ECODE_INVAL) 9428 9429 if len(instance.secondary_nodes) != 1: 9430 raise errors.OpPrereqError("The instance has a strange layout," 9431 " expected one secondary but found %d" % 9432 len(instance.secondary_nodes), 9433 errors.ECODE_FAULT) 9434 9435 if not self.delay_iallocator: 9436 self._CheckPrereq2()
9437
9438 - def _CheckPrereq2(self):
9439 """Check prerequisites, second part. 9440 9441 This function should always be part of CheckPrereq. It was separated and is 9442 now called from Exec because during node evacuation iallocator was only 9443 called with an unmodified cluster model, not taking planned changes into 9444 account. 9445 9446 """ 9447 instance = self.instance 9448 secondary_node = instance.secondary_nodes[0] 9449 9450 if self.iallocator_name is None: 9451 remote_node = self.remote_node 9452 else: 9453 remote_node = self._RunAllocator(self.lu, self.iallocator_name, 9454 instance.name, instance.secondary_nodes) 9455 9456 if remote_node is None: 9457 self.remote_node_info = None 9458 else: 9459 assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \ 9460 "Remote node '%s' is not locked" % remote_node 9461 9462 self.remote_node_info = self.cfg.GetNodeInfo(remote_node) 9463 assert self.remote_node_info is not None, \ 9464 "Cannot retrieve locked node %s" % remote_node 9465 9466 if remote_node == self.instance.primary_node: 9467 raise errors.OpPrereqError("The specified node is the primary node of" 9468 " the instance", errors.ECODE_INVAL) 9469 9470 if remote_node == secondary_node: 9471 raise errors.OpPrereqError("The specified node is already the" 9472 " secondary node of the instance", 9473 errors.ECODE_INVAL) 9474 9475 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO, 9476 constants.REPLACE_DISK_CHG): 9477 raise errors.OpPrereqError("Cannot specify disks to be replaced", 9478 errors.ECODE_INVAL) 9479 9480 if self.mode == constants.REPLACE_DISK_AUTO: 9481 if not self._CheckDisksActivated(instance): 9482 raise errors.OpPrereqError("Please run activate-disks on instance %s" 9483 " first" % self.instance_name, 9484 errors.ECODE_STATE) 9485 faulty_primary = self._FindFaultyDisks(instance.primary_node) 9486 faulty_secondary = self._FindFaultyDisks(secondary_node) 9487 9488 if faulty_primary and faulty_secondary: 9489 raise errors.OpPrereqError("Instance %s has faulty disks on more than" 9490 " one node and can not be repaired" 9491 " automatically" % self.instance_name, 9492 errors.ECODE_STATE) 9493 9494 if faulty_primary: 9495 self.disks = faulty_primary 9496 self.target_node = instance.primary_node 9497 self.other_node = secondary_node 9498 check_nodes = [self.target_node, self.other_node] 9499 elif faulty_secondary: 9500 self.disks = faulty_secondary 9501 self.target_node = secondary_node 9502 self.other_node = instance.primary_node 9503 check_nodes = [self.target_node, self.other_node] 9504 else: 9505 self.disks = [] 9506 check_nodes = [] 9507 9508 else: 9509 # Non-automatic modes 9510 if self.mode == constants.REPLACE_DISK_PRI: 9511 self.target_node = instance.primary_node 9512 self.other_node = secondary_node 9513 check_nodes = [self.target_node, self.other_node] 9514 9515 elif self.mode == constants.REPLACE_DISK_SEC: 9516 self.target_node = secondary_node 9517 self.other_node = instance.primary_node 9518 check_nodes = [self.target_node, self.other_node] 9519 9520 elif self.mode == constants.REPLACE_DISK_CHG: 9521 self.new_node = remote_node 9522 self.other_node = instance.primary_node 9523 self.target_node = secondary_node 9524 check_nodes = [self.new_node, self.other_node] 9525 9526 _CheckNodeNotDrained(self.lu, remote_node) 9527 _CheckNodeVmCapable(self.lu, remote_node) 9528 9529 old_node_info = self.cfg.GetNodeInfo(secondary_node) 9530 assert old_node_info is not None 9531 if old_node_info.offline and not self.early_release: 9532 # doesn't make sense to delay the release 9533 self.early_release = True 9534 self.lu.LogInfo("Old secondary %s is offline, automatically enabling" 9535 " early-release mode", secondary_node) 9536 9537 else: 9538 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" % 9539 self.mode) 9540 9541 # If not specified all disks should be replaced 9542 if not self.disks: 9543 self.disks = range(len(self.instance.disks)) 9544 9545 for node in check_nodes: 9546 _CheckNodeOnline(self.lu, node) 9547 9548 touched_nodes = frozenset(node_name for node_name in [self.new_node, 9549 self.other_node, 9550 self.target_node] 9551 if node_name is not None) 9552 9553 # Release unneeded node locks 9554 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes) 9555 9556 # Release any owned node group 9557 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP): 9558 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP) 9559 9560 # Check whether disks are valid 9561 for disk_idx in self.disks: 9562 instance.FindDisk(disk_idx) 9563 9564 # Get secondary node IP addresses 9565 self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node) 9566 in self.cfg.GetMultiNodeInfo(touched_nodes))
9567
9568 - def Exec(self, feedback_fn):
9569 """Execute disk replacement. 9570 9571 This dispatches the disk replacement to the appropriate handler. 9572 9573 """ 9574 if self.delay_iallocator: 9575 self._CheckPrereq2() 9576 9577 if __debug__: 9578 # Verify owned locks before starting operation 9579 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE) 9580 assert set(owned_nodes) == set(self.node_secondary_ip), \ 9581 ("Incorrect node locks, owning %s, expected %s" % 9582 (owned_nodes, self.node_secondary_ip.keys())) 9583 9584 owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE) 9585 assert list(owned_instances) == [self.instance_name], \ 9586 "Instance '%s' not locked" % self.instance_name 9587 9588 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \ 9589 "Should not own any node group lock at this point" 9590 9591 if not self.disks: 9592 feedback_fn("No disks need replacement") 9593 return 9594 9595 feedback_fn("Replacing disk(s) %s for %s" % 9596 (utils.CommaJoin(self.disks), self.instance.name)) 9597 9598 activate_disks = (not self.instance.admin_up) 9599 9600 # Activate the instance disks if we're replacing them on a down instance 9601 if activate_disks: 9602 _StartInstanceDisks(self.lu, self.instance, True) 9603 9604 try: 9605 # Should we replace the secondary node? 9606 if self.new_node is not None: 9607 fn = self._ExecDrbd8Secondary 9608 else: 9609 fn = self._ExecDrbd8DiskOnly 9610 9611 result = fn(feedback_fn) 9612 finally: 9613 # Deactivate the instance disks if we're replacing them on a 9614 # down instance 9615 if activate_disks: 9616 _SafeShutdownInstanceDisks(self.lu, self.instance) 9617 9618 if __debug__: 9619 # Verify owned locks 9620 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE) 9621 nodes = frozenset(self.node_secondary_ip) 9622 assert ((self.early_release and not owned_nodes) or 9623 (not self.early_release and not (set(owned_nodes) - nodes))), \ 9624 ("Not owning the correct locks, early_release=%s, owned=%r," 9625 " nodes=%r" % (self.early_release, owned_nodes, nodes)) 9626 9627 return result
9628
9629 - def _CheckVolumeGroup(self, nodes):
9630 self.lu.LogInfo("Checking volume groups") 9631 9632 vgname = self.cfg.GetVGName() 9633 9634 # Make sure volume group exists on all involved nodes 9635 results = self.rpc.call_vg_list(nodes) 9636 if not results: 9637 raise errors.OpExecError("Can't list volume groups on the nodes") 9638 9639 for node in nodes: 9640 res = results[node] 9641 res.Raise("Error checking node %s" % node) 9642 if vgname not in res.payload: 9643 raise errors.OpExecError("Volume group '%s' not found on node %s" % 9644 (vgname, node))
9645
9646 - def _CheckDisksExistence(self, nodes):
9647 # Check disk existence 9648 for idx, dev in enumerate(self.instance.disks): 9649 if idx not in self.disks: 9650 continue 9651 9652 for node in nodes: 9653 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node)) 9654 self.cfg.SetDiskID(dev, node) 9655 9656 result = self.rpc.call_blockdev_find(node, dev) 9657 9658 msg = result.fail_msg 9659 if msg or not result.payload: 9660 if not msg: 9661 msg = "disk not found" 9662 raise errors.OpExecError("Can't find disk/%d on node %s: %s" % 9663 (idx, node, msg))
9664
9665 - def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
9666 for idx, dev in enumerate(self.instance.disks): 9667 if idx not in self.disks: 9668 continue 9669 9670 self.lu.LogInfo("Checking disk/%d consistency on node %s" % 9671 (idx, node_name)) 9672 9673 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary, 9674 ldisk=ldisk): 9675 raise errors.OpExecError("Node %s has degraded storage, unsafe to" 9676 " replace disks for instance %s" % 9677 (node_name, self.instance.name))
9678
9679 - def _CreateNewStorage(self, node_name):
9680 """Create new storage on the primary or secondary node. 9681 9682 This is only used for same-node replaces, not for changing the 9683 secondary node, hence we don't want to modify the existing disk. 9684 9685 """ 9686 iv_names = {} 9687 9688 for idx, dev in enumerate(self.instance.disks): 9689 if idx not in self.disks: 9690 continue 9691 9692 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx)) 9693 9694 self.cfg.SetDiskID(dev, node_name) 9695 9696 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]] 9697 names = _GenerateUniqueNames(self.lu, lv_names) 9698 9699 vg_data = dev.children[0].logical_id[0] 9700 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size, 9701 logical_id=(vg_data, names[0])) 9702 vg_meta = dev.children[1].logical_id[0] 9703 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128, 9704 logical_id=(vg_meta, names[1])) 9705 9706 new_lvs = [lv_data, lv_meta] 9707 old_lvs = [child.Copy() for child in dev.children] 9708 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs) 9709 9710 # we pass force_create=True to force the LVM creation 9711 for new_lv in new_lvs: 9712 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True, 9713 _GetInstanceInfoText(self.instance), False) 9714 9715 return iv_names
9716
9717 - def _CheckDevices(self, node_name, iv_names):
9718 for name, (dev, _, _) in iv_names.iteritems(): 9719 self.cfg.SetDiskID(dev, node_name) 9720 9721 result = self.rpc.call_blockdev_find(node_name, dev) 9722 9723 msg = result.fail_msg 9724 if msg or not result.payload: 9725 if not msg: 9726 msg = "disk not found" 9727 raise errors.OpExecError("Can't find DRBD device %s: %s" % 9728 (name, msg)) 9729 9730 if result.payload.is_degraded: 9731 raise errors.OpExecError("DRBD device %s is degraded!" % name)
9732
9733 - def _RemoveOldStorage(self, node_name, iv_names):
9734 for name, (_, old_lvs, _) in iv_names.iteritems(): 9735 self.lu.LogInfo("Remove logical volumes for %s" % name) 9736 9737 for lv in old_lvs: 9738 self.cfg.SetDiskID(lv, node_name) 9739 9740 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg 9741 if msg: 9742 self.lu.LogWarning("Can't remove old LV: %s" % msg, 9743 hint="remove unused LVs manually")
9744
9745 - def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
9746 """Replace a disk on the primary or secondary for DRBD 8. 9747 9748 The algorithm for replace is quite complicated: 9749 9750 1. for each disk to be replaced: 9751 9752 1. create new LVs on the target node with unique names 9753 1. detach old LVs from the drbd device 9754 1. rename old LVs to name_replaced.<time_t> 9755 1. rename new LVs to old LVs 9756 1. attach the new LVs (with the old names now) to the drbd device 9757 9758 1. wait for sync across all devices 9759 9760 1. for each modified disk: 9761 9762 1. remove old LVs (which have the name name_replaces.<time_t>) 9763 9764 Failures are not very well handled. 9765 9766 """ 9767 steps_total = 6 9768 9769 # Step: check device activation 9770 self.lu.LogStep(1, steps_total, "Check device existence") 9771 self._CheckDisksExistence([self.other_node, self.target_node]) 9772 self._CheckVolumeGroup([self.target_node, self.other_node]) 9773 9774 # Step: check other node consistency 9775 self.lu.LogStep(2, steps_total, "Check peer consistency") 9776 self._CheckDisksConsistency(self.other_node, 9777 self.other_node == self.instance.primary_node, 9778 False) 9779 9780 # Step: create new storage 9781 self.lu.LogStep(3, steps_total, "Allocate new storage") 9782 iv_names = self._CreateNewStorage(self.target_node) 9783 9784 # Step: for each lv, detach+rename*2+attach 9785 self.lu.LogStep(4, steps_total, "Changing drbd configuration") 9786 for dev, old_lvs, new_lvs in iv_names.itervalues(): 9787 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name) 9788 9789 result = self.rpc.call_blockdev_removechildren(self.target_node, dev, 9790 old_lvs) 9791 result.Raise("Can't detach drbd from local storage on node" 9792 " %s for device %s" % (self.target_node, dev.iv_name)) 9793 #dev.children = [] 9794 #cfg.Update(instance) 9795 9796 # ok, we created the new LVs, so now we know we have the needed 9797 # storage; as such, we proceed on the target node to rename 9798 # old_lv to _old, and new_lv to old_lv; note that we rename LVs 9799 # using the assumption that logical_id == physical_id (which in 9800 # turn is the unique_id on that node) 9801 9802 # FIXME(iustin): use a better name for the replaced LVs 9803 temp_suffix = int(time.time()) 9804 ren_fn = lambda d, suff: (d.physical_id[0], 9805 d.physical_id[1] + "_replaced-%s" % suff) 9806 9807 # Build the rename list based on what LVs exist on the node 9808 rename_old_to_new = [] 9809 for to_ren in old_lvs: 9810 result = self.rpc.call_blockdev_find(self.target_node, to_ren) 9811 if not result.fail_msg and result.payload: 9812 # device exists 9813 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix))) 9814 9815 self.lu.LogInfo("Renaming the old LVs on the target node") 9816 result = self.rpc.call_blockdev_rename(self.target_node, 9817 rename_old_to_new) 9818 result.Raise("Can't rename old LVs on node %s" % self.target_node) 9819 9820 # Now we rename the new LVs to the old LVs 9821 self.lu.LogInfo("Renaming the new LVs on the target node") 9822 rename_new_to_old = [(new, old.physical_id) 9823 for old, new in zip(old_lvs, new_lvs)] 9824 result = self.rpc.call_blockdev_rename(self.target_node, 9825 rename_new_to_old) 9826 result.Raise("Can't rename new LVs on node %s" % self.target_node) 9827 9828 # Intermediate steps of in memory modifications 9829 for old, new in zip(old_lvs, new_lvs): 9830 new.logical_id = old.logical_id 9831 self.cfg.SetDiskID(new, self.target_node) 9832 9833 # We need to modify old_lvs so that removal later removes the 9834 # right LVs, not the newly added ones; note that old_lvs is a 9835 # copy here 9836 for disk in old_lvs: 9837 disk.logical_id = ren_fn(disk, temp_suffix) 9838 self.cfg.SetDiskID(disk, self.target_node) 9839 9840 # Now that the new lvs have the old name, we can add them to the device 9841 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node) 9842 result = self.rpc.call_blockdev_addchildren(self.target_node, dev, 9843 new_lvs) 9844 msg = result.fail_msg 9845 if msg: 9846 for new_lv in new_lvs: 9847 msg2 = self.rpc.call_blockdev_remove(self.target_node, 9848 new_lv).fail_msg 9849 if msg2: 9850 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2, 9851 hint=("cleanup manually the unused logical" 9852 "volumes")) 9853 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg) 9854 9855 cstep = 5 9856 if self.early_release: 9857 self.lu.LogStep(cstep, steps_total, "Removing old storage") 9858 cstep += 1 9859 self._RemoveOldStorage(self.target_node, iv_names) 9860 # WARNING: we release both node locks here, do not do other RPCs 9861 # than WaitForSync to the primary node 9862 _ReleaseLocks(self.lu, locking.LEVEL_NODE, 9863 names=[self.target_node, self.other_node]) 9864 9865 # Wait for sync 9866 # This can fail as the old devices are degraded and _WaitForSync 9867 # does a combined result over all disks, so we don't check its return value 9868 self.lu.LogStep(cstep, steps_total, "Sync devices") 9869 cstep += 1 9870 _WaitForSync(self.lu, self.instance) 9871 9872 # Check all devices manually 9873 self._CheckDevices(self.instance.primary_node, iv_names) 9874 9875 # Step: remove old storage 9876 if not self.early_release: 9877 self.lu.LogStep(cstep, steps_total, "Removing old storage") 9878 cstep += 1 9879 self._RemoveOldStorage(self.target_node, iv_names)
9880
9881 - def _ExecDrbd8Secondary(self, feedback_fn):
9882 """Replace the secondary node for DRBD 8. 9883 9884 The algorithm for replace is quite complicated: 9885 - for all disks of the instance: 9886 - create new LVs on the new node with same names 9887 - shutdown the drbd device on the old secondary 9888 - disconnect the drbd network on the primary 9889 - create the drbd device on the new secondary 9890 - network attach the drbd on the primary, using an artifice: 9891 the drbd code for Attach() will connect to the network if it 9892 finds a device which is connected to the good local disks but 9893 not network enabled 9894 - wait for sync across all devices 9895 - remove all disks from the old secondary 9896 9897 Failures are not very well handled. 9898 9899 """ 9900 steps_total = 6 9901 9902 pnode = self.instance.primary_node 9903 9904 # Step: check device activation 9905 self.lu.LogStep(1, steps_total, "Check device existence") 9906 self._CheckDisksExistence([self.instance.primary_node]) 9907 self._CheckVolumeGroup([self.instance.primary_node]) 9908 9909 # Step: check other node consistency 9910 self.lu.LogStep(2, steps_total, "Check peer consistency") 9911 self._CheckDisksConsistency(self.instance.primary_node, True, True) 9912 9913 # Step: create new storage 9914 self.lu.LogStep(3, steps_total, "Allocate new storage") 9915 for idx, dev in enumerate(self.instance.disks): 9916 self.lu.LogInfo("Adding new local storage on %s for disk/%d" % 9917 (self.new_node, idx)) 9918 # we pass force_create=True to force LVM creation 9919 for new_lv in dev.children: 9920 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True, 9921 _GetInstanceInfoText(self.instance), False) 9922 9923 # Step 4: dbrd minors and drbd setups changes 9924 # after this, we must manually remove the drbd minors on both the 9925 # error and the success paths 9926 self.lu.LogStep(4, steps_total, "Changing drbd configuration") 9927 minors = self.cfg.AllocateDRBDMinor([self.new_node 9928 for dev in self.instance.disks], 9929 self.instance.name) 9930 logging.debug("Allocated minors %r", minors) 9931 9932 iv_names = {} 9933 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)): 9934 self.lu.LogInfo("activating a new drbd on %s for disk/%d" % 9935 (self.new_node, idx)) 9936 # create new devices on new_node; note that we create two IDs: 9937 # one without port, so the drbd will be activated without 9938 # networking information on the new node at this stage, and one 9939 # with network, for the latter activation in step 4 9940 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id 9941 if self.instance.primary_node == o_node1: 9942 p_minor = o_minor1 9943 else: 9944 assert self.instance.primary_node == o_node2, "Three-node instance?" 9945 p_minor = o_minor2 9946 9947 new_alone_id = (self.instance.primary_node, self.new_node, None, 9948 p_minor, new_minor, o_secret) 9949 new_net_id = (self.instance.primary_node, self.new_node, o_port, 9950 p_minor, new_minor, o_secret) 9951 9952 iv_names[idx] = (dev, dev.children, new_net_id) 9953 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor, 9954 new_net_id) 9955 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8, 9956 logical_id=new_alone_id, 9957 children=dev.children, 9958 size=dev.size) 9959 try: 9960 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd, 9961 _GetInstanceInfoText(self.instance), False) 9962 except errors.GenericError: 9963 self.cfg.ReleaseDRBDMinors(self.instance.name) 9964 raise 9965 9966 # We have new devices, shutdown the drbd on the old secondary 9967 for idx, dev in enumerate(self.instance.disks): 9968 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx) 9969 self.cfg.SetDiskID(dev, self.target_node) 9970 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg 9971 if msg: 9972 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old" 9973 "node: %s" % (idx, msg), 9974 hint=("Please cleanup this device manually as" 9975 " soon as possible")) 9976 9977 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)") 9978 result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip, 9979 self.instance.disks)[pnode] 9980 9981 msg = result.fail_msg 9982 if msg: 9983 # detaches didn't succeed (unlikely) 9984 self.cfg.ReleaseDRBDMinors(self.instance.name) 9985 raise errors.OpExecError("Can't detach the disks from the network on" 9986 " old node: %s" % (msg,)) 9987 9988 # if we managed to detach at least one, we update all the disks of 9989 # the instance to point to the new secondary 9990 self.lu.LogInfo("Updating instance configuration") 9991 for dev, _, new_logical_id in iv_names.itervalues(): 9992 dev.logical_id = new_logical_id 9993 self.cfg.SetDiskID(dev, self.instance.primary_node) 9994 9995 self.cfg.Update(self.instance, feedback_fn) 9996 9997 # and now perform the drbd attach 9998 self.lu.LogInfo("Attaching primary drbds to new secondary" 9999 " (standalone => connected)") 10000 result = self.rpc.call_drbd_attach_net([self.instance.primary_node, 10001 self.new_node], 10002 self.node_secondary_ip, 10003 self.instance.disks, 10004 self.instance.name, 10005 False) 10006 for to_node, to_result in result.items(): 10007 msg = to_result.fail_msg 10008 if msg: 10009 self.lu.LogWarning("Can't attach drbd disks on node %s: %s", 10010 to_node, msg, 10011 hint=("please do a gnt-instance info to see the" 10012 " status of disks")) 10013 cstep = 5 10014 if self.early_release: 10015 self.lu.LogStep(cstep, steps_total, "Removing old storage") 10016 cstep += 1 10017 self._RemoveOldStorage(self.target_node, iv_names) 10018 # WARNING: we release all node locks here, do not do other RPCs 10019 # than WaitForSync to the primary node 10020 _ReleaseLocks(self.lu, locking.LEVEL_NODE, 10021 names=[self.instance.primary_node, 10022 self.target_node, 10023 self.new_node]) 10024 10025 # Wait for sync 10026 # This can fail as the old devices are degraded and _WaitForSync 10027 # does a combined result over all disks, so we don't check its return value 10028 self.lu.LogStep(cstep, steps_total, "Sync devices") 10029 cstep += 1 10030 _WaitForSync(self.lu, self.instance) 10031 10032 # Check all devices manually 10033 self._CheckDevices(self.instance.primary_node, iv_names) 10034 10035 # Step: remove old storage 10036 if not self.early_release: 10037 self.lu.LogStep(cstep, steps_total, "Removing old storage") 10038 self._RemoveOldStorage(self.target_node, iv_names)
10039
10040 10041 -class LURepairNodeStorage(NoHooksLU):
10042 """Repairs the volume group on a node. 10043 10044 """ 10045 REQ_BGL = False 10046
10047 - def CheckArguments(self):
10048 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 10049 10050 storage_type = self.op.storage_type 10051 10052 if (constants.SO_FIX_CONSISTENCY not in 10053 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])): 10054 raise errors.OpPrereqError("Storage units of type '%s' can not be" 10055 " repaired" % storage_type, 10056 errors.ECODE_INVAL)
10057
10058 - def ExpandNames(self):
10059 self.needed_locks = { 10060 locking.LEVEL_NODE: [self.op.node_name], 10061 }
10062
10063 - def _CheckFaultyDisks(self, instance, node_name):
10064 """Ensure faulty disks abort the opcode or at least warn.""" 10065 try: 10066 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance, 10067 node_name, True): 10068 raise errors.OpPrereqError("Instance '%s' has faulty disks on" 10069 " node '%s'" % (instance.name, node_name), 10070 errors.ECODE_STATE) 10071 except errors.OpPrereqError, err: 10072 if self.op.ignore_consistency: 10073 self.proc.LogWarning(str(err.args[0])) 10074 else: 10075 raise
10076
10077 - def CheckPrereq(self):
10078 """Check prerequisites. 10079 10080 """ 10081 # Check whether any instance on this node has faulty disks 10082 for inst in _GetNodeInstances(self.cfg, self.op.node_name): 10083 if not inst.admin_up: 10084 continue 10085 check_nodes = set(inst.all_nodes) 10086 check_nodes.discard(self.op.node_name) 10087 for inst_node_name in check_nodes: 10088 self._CheckFaultyDisks(inst, inst_node_name)
10089
10090 - def Exec(self, feedback_fn):
10091 feedback_fn("Repairing storage unit '%s' on %s ..." % 10092 (self.op.name, self.op.node_name)) 10093 10094 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) 10095 result = self.rpc.call_storage_execute(self.op.node_name, 10096 self.op.storage_type, st_args, 10097 self.op.name, 10098 constants.SO_FIX_CONSISTENCY) 10099 result.Raise("Failed to repair storage unit '%s' on %s" % 10100 (self.op.name, self.op.node_name))
10101
10102 10103 -class LUNodeEvacuate(NoHooksLU):
10104 """Evacuates instances off a list of nodes. 10105 10106 """ 10107 REQ_BGL = False 10108
10109 - def CheckArguments(self):
10110 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
10111
10112 - def ExpandNames(self):
10113 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name) 10114 10115 if self.op.remote_node is not None: 10116 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) 10117 assert self.op.remote_node 10118 10119 if self.op.remote_node == self.op.node_name: 10120 raise errors.OpPrereqError("Can not use evacuated node as a new" 10121 " secondary node", errors.ECODE_INVAL) 10122 10123 if self.op.mode != constants.IALLOCATOR_NEVAC_SEC: 10124 raise errors.OpPrereqError("Without the use of an iallocator only" 10125 " secondary instances can be evacuated", 10126 errors.ECODE_INVAL) 10127 10128 # Declare locks 10129 self.share_locks = _ShareAll() 10130 self.needed_locks = { 10131 locking.LEVEL_INSTANCE: [], 10132 locking.LEVEL_NODEGROUP: [], 10133 locking.LEVEL_NODE: [], 10134 } 10135 10136 # Determine nodes (via group) optimistically, needs verification once locks 10137 # have been acquired 10138 self.lock_nodes = self._DetermineNodes()
10139
10140 - def _DetermineNodes(self):
10141 """Gets the list of nodes to operate on. 10142 10143 """ 10144 if self.op.remote_node is None: 10145 # Iallocator will choose any node(s) in the same group 10146 group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name]) 10147 else: 10148 group_nodes = frozenset([self.op.remote_node]) 10149 10150 # Determine nodes to be locked 10151 return set([self.op.node_name]) | group_nodes
10152
10153 - def _DetermineInstances(self):
10154 """Builds list of instances to operate on. 10155 10156 """ 10157 assert self.op.mode in constants.IALLOCATOR_NEVAC_MODES 10158 10159 if self.op.mode == constants.IALLOCATOR_NEVAC_PRI: 10160 # Primary instances only 10161 inst_fn = _GetNodePrimaryInstances 10162 assert self.op.remote_node is None, \ 10163 "Evacuating primary instances requires iallocator" 10164 elif self.op.mode == constants.IALLOCATOR_NEVAC_SEC: 10165 # Secondary instances only 10166 inst_fn = _GetNodeSecondaryInstances 10167 else: 10168 # All instances 10169 assert self.op.mode == constants.IALLOCATOR_NEVAC_ALL 10170 inst_fn = _GetNodeInstances 10171 # TODO: In 2.6, change the iallocator interface to take an evacuation mode 10172 # per instance 10173 raise errors.OpPrereqError("Due to an issue with the iallocator" 10174 " interface it is not possible to evacuate" 10175 " all instances at once; specify explicitly" 10176 " whether to evacuate primary or secondary" 10177 " instances", 10178 errors.ECODE_INVAL) 10179 10180 return inst_fn(self.cfg, self.op.node_name)
10181
10182 - def DeclareLocks(self, level):
10183 if level == locking.LEVEL_INSTANCE: 10184 # Lock instances optimistically, needs verification once node and group 10185 # locks have been acquired 10186 self.needed_locks[locking.LEVEL_INSTANCE] = \ 10187 set(i.name for i in self._DetermineInstances()) 10188 10189 elif level == locking.LEVEL_NODEGROUP: 10190 # Lock node groups for all potential target nodes optimistically, needs 10191 # verification once nodes have been acquired 10192 self.needed_locks[locking.LEVEL_NODEGROUP] = \ 10193 self.cfg.GetNodeGroupsFromNodes(self.lock_nodes) 10194 10195 elif level == locking.LEVEL_NODE: 10196 self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes
10197
10198 - def CheckPrereq(self):
10199 # Verify locks 10200 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE) 10201 owned_nodes = self.owned_locks(locking.LEVEL_NODE) 10202 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP) 10203 10204 need_nodes = self._DetermineNodes() 10205 10206 if not owned_nodes.issuperset(need_nodes): 10207 raise errors.OpPrereqError("Nodes in same group as '%s' changed since" 10208 " locks were acquired, current nodes are" 10209 " are '%s', used to be '%s'; retry the" 10210 " operation" % 10211 (self.op.node_name, 10212 utils.CommaJoin(need_nodes), 10213 utils.CommaJoin(owned_nodes)), 10214 errors.ECODE_STATE) 10215 10216 wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes) 10217 if owned_groups != wanted_groups: 10218 raise errors.OpExecError("Node groups changed since locks were acquired," 10219 " current groups are '%s', used to be '%s';" 10220 " retry the operation" % 10221 (utils.CommaJoin(wanted_groups), 10222 utils.CommaJoin(owned_groups))) 10223 10224 # Determine affected instances 10225 self.instances = self._DetermineInstances() 10226 self.instance_names = [i.name for i in self.instances] 10227 10228 if set(self.instance_names) != owned_instances: 10229 raise errors.OpExecError("Instances on node '%s' changed since locks" 10230 " were acquired, current instances are '%s'," 10231 " used to be '%s'; retry the operation" % 10232 (self.op.node_name, 10233 utils.CommaJoin(self.instance_names), 10234 utils.CommaJoin(owned_instances))) 10235 10236 if self.instance_names: 10237 self.LogInfo("Evacuating instances from node '%s': %s", 10238 self.op.node_name, 10239 utils.CommaJoin(utils.NiceSort(self.instance_names))) 10240 else: 10241 self.LogInfo("No instances to evacuate from node '%s'", 10242 self.op.node_name) 10243 10244 if self.op.remote_node is not None: 10245 for i in self.instances: 10246 if i.primary_node == self.op.remote_node: 10247 raise errors.OpPrereqError("Node %s is the primary node of" 10248 " instance %s, cannot use it as" 10249 " secondary" % 10250 (self.op.remote_node, i.name), 10251 errors.ECODE_INVAL)
10252
10253 - def Exec(self, feedback_fn):
10254 assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None) 10255 10256 if not self.instance_names: 10257 # No instances to evacuate 10258 jobs = [] 10259 10260 elif self.op.iallocator is not None: 10261 # TODO: Implement relocation to other group 10262 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_NODE_EVAC, 10263 evac_mode=self.op.mode, 10264 instances=list(self.instance_names)) 10265 10266 ial.Run(self.op.iallocator) 10267 10268 if not ial.success: 10269 raise errors.OpPrereqError("Can't compute node evacuation using" 10270 " iallocator '%s': %s" % 10271 (self.op.iallocator, ial.info), 10272 errors.ECODE_NORES) 10273 10274 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True) 10275 10276 elif self.op.remote_node is not None: 10277 assert self.op.mode == constants.IALLOCATOR_NEVAC_SEC 10278 jobs = [ 10279 [opcodes.OpInstanceReplaceDisks(instance_name=instance_name, 10280 remote_node=self.op.remote_node, 10281 disks=[], 10282 mode=constants.REPLACE_DISK_CHG, 10283 early_release=self.op.early_release)] 10284 for instance_name in self.instance_names 10285 ] 10286 10287 else: 10288 raise errors.ProgrammerError("No iallocator or remote node") 10289 10290 return ResultWithJobs(jobs)
10291
10292 10293 -def _SetOpEarlyRelease(early_release, op):
10294 """Sets C{early_release} flag on opcodes if available. 10295 10296 """ 10297 try: 10298 op.early_release = early_release 10299 except AttributeError: 10300 assert not isinstance(op, opcodes.OpInstanceReplaceDisks) 10301 10302 return op
10303
10304 10305 -def _NodeEvacDest(use_nodes, group, nodes):
10306 """Returns group or nodes depending on caller's choice. 10307 10308 """ 10309 if use_nodes: 10310 return utils.CommaJoin(nodes) 10311 else: 10312 return group
10313
10314 10315 -def _LoadNodeEvacResult(lu, alloc_result, early_release, use_nodes):
10316 """Unpacks the result of change-group and node-evacuate iallocator requests. 10317 10318 Iallocator modes L{constants.IALLOCATOR_MODE_NODE_EVAC} and 10319 L{constants.IALLOCATOR_MODE_CHG_GROUP}. 10320 10321 @type lu: L{LogicalUnit} 10322 @param lu: Logical unit instance 10323 @type alloc_result: tuple/list 10324 @param alloc_result: Result from iallocator 10325 @type early_release: bool 10326 @param early_release: Whether to release locks early if possible 10327 @type use_nodes: bool 10328 @param use_nodes: Whether to display node names instead of groups 10329 10330 """ 10331 (moved, failed, jobs) = alloc_result 10332 10333 if failed: 10334 failreason = utils.CommaJoin("%s (%s)" % (name, reason) 10335 for (name, reason) in failed) 10336 lu.LogWarning("Unable to evacuate instances %s", failreason) 10337 raise errors.OpExecError("Unable to evacuate instances %s" % failreason) 10338 10339 if moved: 10340 lu.LogInfo("Instances to be moved: %s", 10341 utils.CommaJoin("%s (to %s)" % 10342 (name, _NodeEvacDest(use_nodes, group, nodes)) 10343 for (name, group, nodes) in moved)) 10344 10345 return [map(compat.partial(_SetOpEarlyRelease, early_release), 10346 map(opcodes.OpCode.LoadOpCode, ops)) 10347 for ops in jobs]
10348
10349 10350 -class LUInstanceGrowDisk(LogicalUnit):
10351 """Grow a disk of an instance. 10352 10353 """ 10354 HPATH = "disk-grow" 10355 HTYPE = constants.HTYPE_INSTANCE 10356 REQ_BGL = False 10357
10358 - def ExpandNames(self):
10359 self._ExpandAndLockInstance() 10360 self.needed_locks[locking.LEVEL_NODE] = [] 10361 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10362
10363 - def DeclareLocks(self, level):
10364 if level == locking.LEVEL_NODE: 10365 self._LockInstancesNodes()
10366
10367 - def BuildHooksEnv(self):
10368 """Build hooks env. 10369 10370 This runs on the master, the primary and all the secondaries. 10371 10372 """ 10373 env = { 10374 "DISK": self.op.disk, 10375 "AMOUNT": self.op.amount, 10376 } 10377 env.update(_BuildInstanceHookEnvByObject(self, self.instance)) 10378 return env
10379
10380 - def BuildHooksNodes(self):
10381 """Build hooks nodes. 10382 10383 """ 10384 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 10385 return (nl, nl)
10386
10387 - def CheckPrereq(self):
10388 """Check prerequisites. 10389 10390 This checks that the instance is in the cluster. 10391 10392 """ 10393 instance = self.cfg.GetInstanceInfo(self.op.instance_name) 10394 assert instance is not None, \ 10395 "Cannot retrieve locked instance %s" % self.op.instance_name 10396 nodenames = list(instance.all_nodes) 10397 for node in nodenames: 10398 _CheckNodeOnline(self, node) 10399 10400 self.instance = instance 10401 10402 if instance.disk_template not in constants.DTS_GROWABLE: 10403 raise errors.OpPrereqError("Instance's disk layout does not support" 10404 " growing", errors.ECODE_INVAL) 10405 10406 self.disk = instance.FindDisk(self.op.disk) 10407 10408 if instance.disk_template not in (constants.DT_FILE, 10409 constants.DT_SHARED_FILE): 10410 # TODO: check the free disk space for file, when that feature will be 10411 # supported 10412 _CheckNodesFreeDiskPerVG(self, nodenames, 10413 self.disk.ComputeGrowth(self.op.amount))
10414
10415 - def Exec(self, feedback_fn):
10416 """Execute disk grow. 10417 10418 """ 10419 instance = self.instance 10420 disk = self.disk 10421 10422 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk]) 10423 if not disks_ok: 10424 raise errors.OpExecError("Cannot activate block device to grow") 10425 10426 # First run all grow ops in dry-run mode 10427 for node in instance.all_nodes: 10428 self.cfg.SetDiskID(disk, node) 10429 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, True) 10430 result.Raise("Grow request failed to node %s" % node) 10431 10432 # We know that (as far as we can test) operations across different 10433 # nodes will succeed, time to run it for real 10434 for node in instance.all_nodes: 10435 self.cfg.SetDiskID(disk, node) 10436 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, False) 10437 result.Raise("Grow request failed to node %s" % node) 10438 10439 # TODO: Rewrite code to work properly 10440 # DRBD goes into sync mode for a short amount of time after executing the 10441 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby 10442 # calling "resize" in sync mode fails. Sleeping for a short amount of 10443 # time is a work-around. 10444 time.sleep(5) 10445 10446 disk.RecordGrow(self.op.amount) 10447 self.cfg.Update(instance, feedback_fn) 10448 if self.op.wait_for_sync: 10449 disk_abort = not _WaitForSync(self, instance, disks=[disk]) 10450 if disk_abort: 10451 self.proc.LogWarning("Disk sync-ing has not returned a good" 10452 " status; please check the instance") 10453 if not instance.admin_up: 10454 _SafeShutdownInstanceDisks(self, instance, disks=[disk]) 10455 elif not instance.admin_up: 10456 self.proc.LogWarning("Not shutting down the disk even if the instance is" 10457 " not supposed to be running because no wait for" 10458 " sync mode was requested")
10459
10460 10461 -class LUInstanceQueryData(NoHooksLU):
10462 """Query runtime instance data. 10463 10464 """ 10465 REQ_BGL = False 10466
10467 - def ExpandNames(self):
10468 self.needed_locks = {} 10469 10470 # Use locking if requested or when non-static information is wanted 10471 if not (self.op.static or self.op.use_locking): 10472 self.LogWarning("Non-static data requested, locks need to be acquired") 10473 self.op.use_locking = True 10474 10475 if self.op.instances or not self.op.use_locking: 10476 # Expand instance names right here 10477 self.wanted_names = _GetWantedInstances(self, self.op.instances) 10478 else: 10479 # Will use acquired locks 10480 self.wanted_names = None 10481 10482 if self.op.use_locking: 10483 self.share_locks = _ShareAll() 10484 10485 if self.wanted_names is None: 10486 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET 10487 else: 10488 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names 10489 10490 self.needed_locks[locking.LEVEL_NODE] = [] 10491 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10492
10493 - def DeclareLocks(self, level):
10494 if self.op.use_locking and level == locking.LEVEL_NODE: 10495 self._LockInstancesNodes()
10496
10497 - def CheckPrereq(self):
10498 """Check prerequisites. 10499 10500 This only checks the optional instance list against the existing names. 10501 10502 """ 10503 if self.wanted_names is None: 10504 assert self.op.use_locking, "Locking was not used" 10505 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE) 10506 10507 self.wanted_instances = \ 10508 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
10509
10510 - def _ComputeBlockdevStatus(self, node, instance_name, dev):
10511 """Returns the status of a block device 10512 10513 """ 10514 if self.op.static or not node: 10515 return None 10516 10517 self.cfg.SetDiskID(dev, node) 10518 10519 result = self.rpc.call_blockdev_find(node, dev) 10520 if result.offline: 10521 return None 10522 10523 result.Raise("Can't compute disk status for %s" % instance_name) 10524 10525 status = result.payload 10526 if status is None: 10527 return None 10528 10529 return (status.dev_path, status.major, status.minor, 10530 status.sync_percent, status.estimated_time, 10531 status.is_degraded, status.ldisk_status)
10532
10533 - def _ComputeDiskStatus(self, instance, snode, dev):
10534 """Compute block device status. 10535 10536 """ 10537 if dev.dev_type in constants.LDS_DRBD: 10538 # we change the snode then (otherwise we use the one passed in) 10539 if dev.logical_id[0] == instance.primary_node: 10540 snode = dev.logical_id[1] 10541 else: 10542 snode = dev.logical_id[0] 10543 10544 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node, 10545 instance.name, dev) 10546 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev) 10547 10548 if dev.children: 10549 dev_children = map(compat.partial(self._ComputeDiskStatus, 10550 instance, snode), 10551 dev.children) 10552 else: 10553 dev_children = [] 10554 10555 return { 10556 "iv_name": dev.iv_name, 10557 "dev_type": dev.dev_type, 10558 "logical_id": dev.logical_id, 10559 "physical_id": dev.physical_id, 10560 "pstatus": dev_pstatus, 10561 "sstatus": dev_sstatus, 10562 "children": dev_children, 10563 "mode": dev.mode, 10564 "size": dev.size, 10565 }
10566
10567 - def Exec(self, feedback_fn):
10568 """Gather and return data""" 10569 result = {} 10570 10571 cluster = self.cfg.GetClusterInfo() 10572 10573 pri_nodes = self.cfg.GetMultiNodeInfo(i.primary_node 10574 for i in self.wanted_instances) 10575 for instance, (_, pnode) in zip(self.wanted_instances, pri_nodes): 10576 if self.op.static or pnode.offline: 10577 remote_state = None 10578 if pnode.offline: 10579 self.LogWarning("Primary node %s is marked offline, returning static" 10580 " information only for instance %s" % 10581 (pnode.name, instance.name)) 10582 else: 10583 remote_info = self.rpc.call_instance_info(instance.primary_node, 10584 instance.name, 10585 instance.hypervisor) 10586 remote_info.Raise("Error checking node %s" % instance.primary_node) 10587 remote_info = remote_info.payload 10588 if remote_info and "state" in remote_info: 10589 remote_state = "up" 10590 else: 10591 remote_state = "down" 10592 10593 if instance.admin_up: 10594 config_state = "up" 10595 else: 10596 config_state = "down" 10597 10598 disks = map(compat.partial(self._ComputeDiskStatus, instance, None), 10599 instance.disks) 10600 10601 result[instance.name] = { 10602 "name": instance.name, 10603 "config_state": config_state, 10604 "run_state": remote_state, 10605 "pnode": instance.primary_node, 10606 "snodes": instance.secondary_nodes, 10607 "os": instance.os, 10608 # this happens to be the same format used for hooks 10609 "nics": _NICListToTuple(self, instance.nics), 10610 "disk_template": instance.disk_template, 10611 "disks": disks, 10612 "hypervisor": instance.hypervisor, 10613 "network_port": instance.network_port, 10614 "hv_instance": instance.hvparams, 10615 "hv_actual": cluster.FillHV(instance, skip_globals=True), 10616 "be_instance": instance.beparams, 10617 "be_actual": cluster.FillBE(instance), 10618 "os_instance": instance.osparams, 10619 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams), 10620 "serial_no": instance.serial_no, 10621 "mtime": instance.mtime, 10622 "ctime": instance.ctime, 10623 "uuid": instance.uuid, 10624 } 10625 10626 return result
10627
10628 10629 -class LUInstanceSetParams(LogicalUnit):
10630 """Modifies an instances's parameters. 10631 10632 """ 10633 HPATH = "instance-modify" 10634 HTYPE = constants.HTYPE_INSTANCE 10635 REQ_BGL = False 10636
10637 - def CheckArguments(self):
10638 if not (self.op.nics or self.op.disks or self.op.disk_template or 10639 self.op.hvparams or self.op.beparams or self.op.os_name): 10640 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL) 10641 10642 if self.op.hvparams: 10643 _CheckGlobalHvParams(self.op.hvparams) 10644 10645 # Disk validation 10646 disk_addremove = 0 10647 for disk_op, disk_dict in self.op.disks: 10648 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES) 10649 if disk_op == constants.DDM_REMOVE: 10650 disk_addremove += 1 10651 continue 10652 elif disk_op == constants.DDM_ADD: 10653 disk_addremove += 1 10654 else: 10655 if not isinstance(disk_op, int): 10656 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL) 10657 if not isinstance(disk_dict, dict): 10658 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict 10659 raise errors.OpPrereqError(msg, errors.ECODE_INVAL) 10660 10661 if disk_op == constants.DDM_ADD: 10662 mode = disk_dict.setdefault(constants.IDISK_MODE, constants.DISK_RDWR) 10663 if mode not in constants.DISK_ACCESS_SET: 10664 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode, 10665 errors.ECODE_INVAL) 10666 size = disk_dict.get(constants.IDISK_SIZE, None) 10667 if size is None: 10668 raise errors.OpPrereqError("Required disk parameter size missing", 10669 errors.ECODE_INVAL) 10670 try: 10671 size = int(size) 10672 except (TypeError, ValueError), err: 10673 raise errors.OpPrereqError("Invalid disk size parameter: %s" % 10674 str(err), errors.ECODE_INVAL) 10675 disk_dict[constants.IDISK_SIZE] = size 10676 else: 10677 # modification of disk 10678 if constants.IDISK_SIZE in disk_dict: 10679 raise errors.OpPrereqError("Disk size change not possible, use" 10680 " grow-disk", errors.ECODE_INVAL) 10681 10682 if disk_addremove > 1: 10683 raise errors.OpPrereqError("Only one disk add or remove operation" 10684 " supported at a time", errors.ECODE_INVAL) 10685 10686 if self.op.disks and self.op.disk_template is not None: 10687 raise errors.OpPrereqError("Disk template conversion and other disk" 10688 " changes not supported at the same time", 10689 errors.ECODE_INVAL) 10690 10691 if (self.op.disk_template and 10692 self.op.disk_template in constants.DTS_INT_MIRROR and 10693 self.op.remote_node is None): 10694 raise errors.OpPrereqError("Changing the disk template to a mirrored" 10695 " one requires specifying a secondary node", 10696 errors.ECODE_INVAL) 10697 10698 # NIC validation 10699 nic_addremove = 0 10700 for nic_op, nic_dict in self.op.nics: 10701 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES) 10702 if nic_op == constants.DDM_REMOVE: 10703 nic_addremove += 1 10704 continue 10705 elif nic_op == constants.DDM_ADD: 10706 nic_addremove += 1 10707 else: 10708 if not isinstance(nic_op, int): 10709 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL) 10710 if not isinstance(nic_dict, dict): 10711 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict 10712 raise errors.OpPrereqError(msg, errors.ECODE_INVAL) 10713 10714 # nic_dict should be a dict 10715 nic_ip = nic_dict.get(constants.INIC_IP, None) 10716 if nic_ip is not None: 10717 if nic_ip.lower() == constants.VALUE_NONE: 10718 nic_dict[constants.INIC_IP] = None 10719 else: 10720 if not netutils.IPAddress.IsValid(nic_ip): 10721 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip, 10722 errors.ECODE_INVAL) 10723 10724 nic_bridge = nic_dict.get("bridge", None) 10725 nic_link = nic_dict.get(constants.INIC_LINK, None) 10726 if nic_bridge and nic_link: 10727 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'" 10728 " at the same time", errors.ECODE_INVAL) 10729 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE: 10730 nic_dict["bridge"] = None 10731 elif nic_link and nic_link.lower() == constants.VALUE_NONE: 10732 nic_dict[constants.INIC_LINK] = None 10733 10734 if nic_op == constants.DDM_ADD: 10735 nic_mac = nic_dict.get(constants.INIC_MAC, None) 10736 if nic_mac is None: 10737 nic_dict[constants.INIC_MAC] = constants.VALUE_AUTO 10738 10739 if constants.INIC_MAC in nic_dict: 10740 nic_mac = nic_dict[constants.INIC_MAC] 10741 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE): 10742 nic_mac = utils.NormalizeAndValidateMac(nic_mac) 10743 10744 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO: 10745 raise errors.OpPrereqError("'auto' is not a valid MAC address when" 10746 " modifying an existing nic", 10747 errors.ECODE_INVAL) 10748 10749 if nic_addremove > 1: 10750 raise errors.OpPrereqError("Only one NIC add or remove operation" 10751 " supported at a time", errors.ECODE_INVAL)
10752
10753 - def ExpandNames(self):
10754 self._ExpandAndLockInstance() 10755 self.needed_locks[locking.LEVEL_NODE] = [] 10756 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10757
10758 - def DeclareLocks(self, level):
10759 if level == locking.LEVEL_NODE: 10760 self._LockInstancesNodes() 10761 if self.op.disk_template and self.op.remote_node: 10762 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node) 10763 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
10764
10765 - def BuildHooksEnv(self):
10766 """Build hooks env. 10767 10768 This runs on the master, primary and secondaries. 10769 10770 """ 10771 args = dict() 10772 if constants.BE_MEMORY in self.be_new: 10773 args["memory"] = self.be_new[constants.BE_MEMORY] 10774 if constants.BE_VCPUS in self.be_new: 10775 args["vcpus"] = self.be_new[constants.BE_VCPUS] 10776 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk 10777 # information at all. 10778 if self.op.nics: 10779 args["nics"] = [] 10780 nic_override = dict(self.op.nics) 10781 for idx, nic in enumerate(self.instance.nics): 10782 if idx in nic_override: 10783 this_nic_override = nic_override[idx] 10784 else: 10785 this_nic_override = {} 10786 if constants.INIC_IP in this_nic_override: 10787 ip = this_nic_override[constants.INIC_IP] 10788 else: 10789 ip = nic.ip 10790 if constants.INIC_MAC in this_nic_override: 10791 mac = this_nic_override[constants.INIC_MAC] 10792 else: 10793 mac = nic.mac 10794 if idx in self.nic_pnew: 10795 nicparams = self.nic_pnew[idx] 10796 else: 10797 nicparams = self.cluster.SimpleFillNIC(nic.nicparams) 10798 mode = nicparams[constants.NIC_MODE] 10799 link = nicparams[constants.NIC_LINK] 10800 args["nics"].append((ip, mac, mode, link)) 10801 if constants.DDM_ADD in nic_override: 10802 ip = nic_override[constants.DDM_ADD].get(constants.INIC_IP, None) 10803 mac = nic_override[constants.DDM_ADD][constants.INIC_MAC] 10804 nicparams = self.nic_pnew[constants.DDM_ADD] 10805 mode = nicparams[constants.NIC_MODE] 10806 link = nicparams[constants.NIC_LINK] 10807 args["nics"].append((ip, mac, mode, link)) 10808 elif constants.DDM_REMOVE in nic_override: 10809 del args["nics"][-1] 10810 10811 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args) 10812 if self.op.disk_template: 10813 env["NEW_DISK_TEMPLATE"] = self.op.disk_template 10814 10815 return env
10816
10817 - def BuildHooksNodes(self):
10818 """Build hooks nodes. 10819 10820 """ 10821 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes) 10822 return (nl, nl)
10823
10824 - def CheckPrereq(self):
10825 """Check prerequisites. 10826 10827 This only checks the instance list against the existing names. 10828 10829 """ 10830 # checking the new params on the primary/secondary nodes 10831 10832 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 10833 cluster = self.cluster = self.cfg.GetClusterInfo() 10834 assert self.instance is not None, \ 10835 "Cannot retrieve locked instance %s" % self.op.instance_name 10836 pnode = instance.primary_node 10837 nodelist = list(instance.all_nodes) 10838 10839 # OS change 10840 if self.op.os_name and not self.op.force: 10841 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name, 10842 self.op.force_variant) 10843 instance_os = self.op.os_name 10844 else: 10845 instance_os = instance.os 10846 10847 if self.op.disk_template: 10848 if instance.disk_template == self.op.disk_template: 10849 raise errors.OpPrereqError("Instance already has disk template %s" % 10850 instance.disk_template, errors.ECODE_INVAL) 10851 10852 if (instance.disk_template, 10853 self.op.disk_template) not in self._DISK_CONVERSIONS: 10854 raise errors.OpPrereqError("Unsupported disk template conversion from" 10855 " %s to %s" % (instance.disk_template, 10856 self.op.disk_template), 10857 errors.ECODE_INVAL) 10858 _CheckInstanceDown(self, instance, "cannot change disk template") 10859 if self.op.disk_template in constants.DTS_INT_MIRROR: 10860 if self.op.remote_node == pnode: 10861 raise errors.OpPrereqError("Given new secondary node %s is the same" 10862 " as the primary node of the instance" % 10863 self.op.remote_node, errors.ECODE_STATE) 10864 _CheckNodeOnline(self, self.op.remote_node) 10865 _CheckNodeNotDrained(self, self.op.remote_node) 10866 # FIXME: here we assume that the old instance type is DT_PLAIN 10867 assert instance.disk_template == constants.DT_PLAIN 10868 disks = [{constants.IDISK_SIZE: d.size, 10869 constants.IDISK_VG: d.logical_id[0]} 10870 for d in instance.disks] 10871 required = _ComputeDiskSizePerVG(self.op.disk_template, disks) 10872 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required) 10873 10874 # hvparams processing 10875 if self.op.hvparams: 10876 hv_type = instance.hypervisor 10877 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams) 10878 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES) 10879 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict) 10880 10881 # local check 10882 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new) 10883 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new) 10884 self.hv_new = hv_new # the new actual values 10885 self.hv_inst = i_hvdict # the new dict (without defaults) 10886 else: 10887 self.hv_new = self.hv_inst = {} 10888 10889 # beparams processing 10890 if self.op.beparams: 10891 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams, 10892 use_none=True) 10893 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES) 10894 be_new = cluster.SimpleFillBE(i_bedict) 10895 self.be_new = be_new # the new actual values 10896 self.be_inst = i_bedict # the new dict (without defaults) 10897 else: 10898 self.be_new = self.be_inst = {} 10899 be_old = cluster.FillBE(instance) 10900 10901 # osparams processing 10902 if self.op.osparams: 10903 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams) 10904 _CheckOSParams(self, True, nodelist, instance_os, i_osdict) 10905 self.os_inst = i_osdict # the new dict (without defaults) 10906 else: 10907 self.os_inst = {} 10908 10909 self.warn = [] 10910 10911 if (constants.BE_MEMORY in self.op.beparams and not self.op.force and 10912 be_new[constants.BE_MEMORY] > be_old[constants.BE_MEMORY]): 10913 mem_check_list = [pnode] 10914 if be_new[constants.BE_AUTO_BALANCE]: 10915 # either we changed auto_balance to yes or it was from before 10916 mem_check_list.extend(instance.secondary_nodes) 10917 instance_info = self.rpc.call_instance_info(pnode, instance.name, 10918 instance.hypervisor) 10919 nodeinfo = self.rpc.call_node_info(mem_check_list, None, 10920 instance.hypervisor) 10921 pninfo = nodeinfo[pnode] 10922 msg = pninfo.fail_msg 10923 if msg: 10924 # Assume the primary node is unreachable and go ahead 10925 self.warn.append("Can't get info from primary node %s: %s" % 10926 (pnode, msg)) 10927 elif not isinstance(pninfo.payload.get("memory_free", None), int): 10928 self.warn.append("Node data from primary node %s doesn't contain" 10929 " free memory information" % pnode) 10930 elif instance_info.fail_msg: 10931 self.warn.append("Can't get instance runtime information: %s" % 10932 instance_info.fail_msg) 10933 else: 10934 if instance_info.payload: 10935 current_mem = int(instance_info.payload["memory"]) 10936 else: 10937 # Assume instance not running 10938 # (there is a slight race condition here, but it's not very probable, 10939 # and we have no other way to check) 10940 current_mem = 0 10941 miss_mem = (be_new[constants.BE_MEMORY] - current_mem - 10942 pninfo.payload["memory_free"]) 10943 if miss_mem > 0: 10944 raise errors.OpPrereqError("This change will prevent the instance" 10945 " from starting, due to %d MB of memory" 10946 " missing on its primary node" % miss_mem, 10947 errors.ECODE_NORES) 10948 10949 if be_new[constants.BE_AUTO_BALANCE]: 10950 for node, nres in nodeinfo.items(): 10951 if node not in instance.secondary_nodes: 10952 continue 10953 nres.Raise("Can't get info from secondary node %s" % node, 10954 prereq=True, ecode=errors.ECODE_STATE) 10955 if not isinstance(nres.payload.get("memory_free", None), int): 10956 raise errors.OpPrereqError("Secondary node %s didn't return free" 10957 " memory information" % node, 10958 errors.ECODE_STATE) 10959 elif be_new[constants.BE_MEMORY] > nres.payload["memory_free"]: 10960 raise errors.OpPrereqError("This change will prevent the instance" 10961 " from failover to its secondary node" 10962 " %s, due to not enough memory" % node, 10963 errors.ECODE_STATE) 10964 10965 # NIC processing 10966 self.nic_pnew = {} 10967 self.nic_pinst = {} 10968 for nic_op, nic_dict in self.op.nics: 10969 if nic_op == constants.DDM_REMOVE: 10970 if not instance.nics: 10971 raise errors.OpPrereqError("Instance has no NICs, cannot remove", 10972 errors.ECODE_INVAL) 10973 continue 10974 if nic_op != constants.DDM_ADD: 10975 # an existing nic 10976 if not instance.nics: 10977 raise errors.OpPrereqError("Invalid NIC index %s, instance has" 10978 " no NICs" % nic_op, 10979 errors.ECODE_INVAL) 10980 if nic_op < 0 or nic_op >= len(instance.nics): 10981 raise errors.OpPrereqError("Invalid NIC index %s, valid values" 10982 " are 0 to %d" % 10983 (nic_op, len(instance.nics) - 1), 10984 errors.ECODE_INVAL) 10985 old_nic_params = instance.nics[nic_op].nicparams 10986 old_nic_ip = instance.nics[nic_op].ip 10987 else: 10988 old_nic_params = {} 10989 old_nic_ip = None 10990 10991 update_params_dict = dict([(key, nic_dict[key]) 10992 for key in constants.NICS_PARAMETERS 10993 if key in nic_dict]) 10994 10995 if "bridge" in nic_dict: 10996 update_params_dict[constants.NIC_LINK] = nic_dict["bridge"] 10997 10998 new_nic_params = _GetUpdatedParams(old_nic_params, 10999 update_params_dict) 11000 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES) 11001 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params) 11002 objects.NIC.CheckParameterSyntax(new_filled_nic_params) 11003 self.nic_pinst[nic_op] = new_nic_params 11004 self.nic_pnew[nic_op] = new_filled_nic_params 11005 new_nic_mode = new_filled_nic_params[constants.NIC_MODE] 11006 11007 if new_nic_mode == constants.NIC_MODE_BRIDGED: 11008 nic_bridge = new_filled_nic_params[constants.NIC_LINK] 11009 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg 11010 if msg: 11011 msg = "Error checking bridges on node %s: %s" % (pnode, msg) 11012 if self.op.force: 11013 self.warn.append(msg) 11014 else: 11015 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON) 11016 if new_nic_mode == constants.NIC_MODE_ROUTED: 11017 if constants.INIC_IP in nic_dict: 11018 nic_ip = nic_dict[constants.INIC_IP] 11019 else: 11020 nic_ip = old_nic_ip 11021 if nic_ip is None: 11022 raise errors.OpPrereqError("Cannot set the nic ip to None" 11023 " on a routed nic", errors.ECODE_INVAL) 11024 if constants.INIC_MAC in nic_dict: 11025 nic_mac = nic_dict[constants.INIC_MAC] 11026 if nic_mac is None: 11027 raise errors.OpPrereqError("Cannot set the nic mac to None", 11028 errors.ECODE_INVAL) 11029 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE): 11030 # otherwise generate the mac 11031 nic_dict[constants.INIC_MAC] = \ 11032 self.cfg.GenerateMAC(self.proc.GetECId()) 11033 else: 11034 # or validate/reserve the current one 11035 try: 11036 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId()) 11037 except errors.ReservationError: 11038 raise errors.OpPrereqError("MAC address %s already in use" 11039 " in cluster" % nic_mac, 11040 errors.ECODE_NOTUNIQUE) 11041 11042 # DISK processing 11043 if self.op.disks and instance.disk_template == constants.DT_DISKLESS: 11044 raise errors.OpPrereqError("Disk operations not supported for" 11045 " diskless instances", 11046 errors.ECODE_INVAL) 11047 for disk_op, _ in self.op.disks: 11048 if disk_op == constants.DDM_REMOVE: 11049 if len(instance.disks) == 1: 11050 raise errors.OpPrereqError("Cannot remove the last disk of" 11051 " an instance", errors.ECODE_INVAL) 11052 _CheckInstanceDown(self, instance, "cannot remove disks") 11053 11054 if (disk_op == constants.DDM_ADD and 11055 len(instance.disks) >= constants.MAX_DISKS): 11056 raise errors.OpPrereqError("Instance has too many disks (%d), cannot" 11057 " add more" % constants.MAX_DISKS, 11058 errors.ECODE_STATE) 11059 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE): 11060 # an existing disk 11061 if disk_op < 0 or disk_op >= len(instance.disks): 11062 raise errors.OpPrereqError("Invalid disk index %s, valid values" 11063 " are 0 to %d" % 11064 (disk_op, len(instance.disks)), 11065 errors.ECODE_INVAL) 11066 11067 return
11068
11069 - def _ConvertPlainToDrbd(self, feedback_fn):
11070 """Converts an instance from plain to drbd. 11071 11072 """ 11073 feedback_fn("Converting template to drbd") 11074 instance = self.instance 11075 pnode = instance.primary_node 11076 snode = self.op.remote_node 11077 11078 # create a fake disk info for _GenerateDiskTemplate 11079 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode, 11080 constants.IDISK_VG: d.logical_id[0]} 11081 for d in instance.disks] 11082 new_disks = _GenerateDiskTemplate(self, self.op.disk_template, 11083 instance.name, pnode, [snode], 11084 disk_info, None, None, 0, feedback_fn) 11085 info = _GetInstanceInfoText(instance) 11086 feedback_fn("Creating aditional volumes...") 11087 # first, create the missing data and meta devices 11088 for disk in new_disks: 11089 # unfortunately this is... not too nice 11090 _CreateSingleBlockDev(self, pnode, instance, disk.children[1], 11091 info, True) 11092 for child in disk.children: 11093 _CreateSingleBlockDev(self, snode, instance, child, info, True) 11094 # at this stage, all new LVs have been created, we can rename the 11095 # old ones 11096 feedback_fn("Renaming original volumes...") 11097 rename_list = [(o, n.children[0].logical_id) 11098 for (o, n) in zip(instance.disks, new_disks)] 11099 result = self.rpc.call_blockdev_rename(pnode, rename_list) 11100 result.Raise("Failed to rename original LVs") 11101 11102 feedback_fn("Initializing DRBD devices...") 11103 # all child devices are in place, we can now create the DRBD devices 11104 for disk in new_disks: 11105 for node in [pnode, snode]: 11106 f_create = node == pnode 11107 _CreateSingleBlockDev(self, node, instance, disk, info, f_create) 11108 11109 # at this point, the instance has been modified 11110 instance.disk_template = constants.DT_DRBD8 11111 instance.disks = new_disks 11112 self.cfg.Update(instance, feedback_fn) 11113 11114 # disks are created, waiting for sync 11115 disk_abort = not _WaitForSync(self, instance, 11116 oneshot=not self.op.wait_for_sync) 11117 if disk_abort: 11118 raise errors.OpExecError("There are some degraded disks for" 11119 " this instance, please cleanup manually")
11120
11121 - def _ConvertDrbdToPlain(self, feedback_fn):
11122 """Converts an instance from drbd to plain. 11123 11124 """ 11125 instance = self.instance 11126 assert len(instance.secondary_nodes) == 1 11127 pnode = instance.primary_node 11128 snode = instance.secondary_nodes[0] 11129 feedback_fn("Converting template to plain") 11130 11131 old_disks = instance.disks 11132 new_disks = [d.children[0] for d in old_disks] 11133 11134 # copy over size and mode 11135 for parent, child in zip(old_disks, new_disks): 11136 child.size = parent.size 11137 child.mode = parent.mode 11138 11139 # this is a DRBD disk, return its port to the pool 11140 # NOTE: this must be done right before the call to cfg.Update! 11141 for disk in old_disks: 11142 tcp_port = disk.logical_id[2] 11143 self.cfg.AddTcpUdpPort(tcp_port) 11144 11145 # update instance structure 11146 instance.disks = new_disks 11147 instance.disk_template = constants.DT_PLAIN 11148 self.cfg.Update(instance, feedback_fn) 11149 11150 feedback_fn("Removing volumes on the secondary node...") 11151 for disk in old_disks: 11152 self.cfg.SetDiskID(disk, snode) 11153 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg 11154 if msg: 11155 self.LogWarning("Could not remove block device %s on node %s," 11156 " continuing anyway: %s", disk.iv_name, snode, msg) 11157 11158 feedback_fn("Removing unneeded volumes on the primary node...") 11159 for idx, disk in enumerate(old_disks): 11160 meta = disk.children[1] 11161 self.cfg.SetDiskID(meta, pnode) 11162 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg 11163 if msg: 11164 self.LogWarning("Could not remove metadata for disk %d on node %s," 11165 " continuing anyway: %s", idx, pnode, msg)
11166
11167 - def Exec(self, feedback_fn):
11168 """Modifies an instance. 11169 11170 All parameters take effect only at the next restart of the instance. 11171 11172 """ 11173 # Process here the warnings from CheckPrereq, as we don't have a 11174 # feedback_fn there. 11175 for warn in self.warn: 11176 feedback_fn("WARNING: %s" % warn) 11177 11178 result = [] 11179 instance = self.instance 11180 # disk changes 11181 for disk_op, disk_dict in self.op.disks: 11182 if disk_op == constants.DDM_REMOVE: 11183 # remove the last disk 11184 device = instance.disks.pop() 11185 device_idx = len(instance.disks) 11186 for node, disk in device.ComputeNodeTree(instance.primary_node): 11187 self.cfg.SetDiskID(disk, node) 11188 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg 11189 if msg: 11190 self.LogWarning("Could not remove disk/%d on node %s: %s," 11191 " continuing anyway", device_idx, node, msg) 11192 result.append(("disk/%d" % device_idx, "remove")) 11193 11194 # if this is a DRBD disk, return its port to the pool 11195 if device.dev_type in constants.LDS_DRBD: 11196 tcp_port = device.logical_id[2] 11197 self.cfg.AddTcpUdpPort(tcp_port) 11198 elif disk_op == constants.DDM_ADD: 11199 # add a new disk 11200 if instance.disk_template in (constants.DT_FILE, 11201 constants.DT_SHARED_FILE): 11202 file_driver, file_path = instance.disks[0].logical_id 11203 file_path = os.path.dirname(file_path) 11204 else: 11205 file_driver = file_path = None 11206 disk_idx_base = len(instance.disks) 11207 new_disk = _GenerateDiskTemplate(self, 11208 instance.disk_template, 11209 instance.name, instance.primary_node, 11210 instance.secondary_nodes, 11211 [disk_dict], 11212 file_path, 11213 file_driver, 11214 disk_idx_base, feedback_fn)[0] 11215 instance.disks.append(new_disk) 11216 info = _GetInstanceInfoText(instance) 11217 11218 logging.info("Creating volume %s for instance %s", 11219 new_disk.iv_name, instance.name) 11220 # Note: this needs to be kept in sync with _CreateDisks 11221 #HARDCODE 11222 for node in instance.all_nodes: 11223 f_create = node == instance.primary_node 11224 try: 11225 _CreateBlockDev(self, node, instance, new_disk, 11226 f_create, info, f_create) 11227 except errors.OpExecError, err: 11228 self.LogWarning("Failed to create volume %s (%s) on" 11229 " node %s: %s", 11230 new_disk.iv_name, new_disk, node, err) 11231 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" % 11232 (new_disk.size, new_disk.mode))) 11233 else: 11234 # change a given disk 11235 instance.disks[disk_op].mode = disk_dict[constants.IDISK_MODE] 11236 result.append(("disk.mode/%d" % disk_op, 11237 disk_dict[constants.IDISK_MODE])) 11238 11239 if self.op.disk_template: 11240 r_shut = _ShutdownInstanceDisks(self, instance) 11241 if not r_shut: 11242 raise errors.OpExecError("Cannot shutdown instance disks, unable to" 11243 " proceed with disk template conversion") 11244 mode = (instance.disk_template, self.op.disk_template) 11245 try: 11246 self._DISK_CONVERSIONS[mode](self, feedback_fn) 11247 except: 11248 self.cfg.ReleaseDRBDMinors(instance.name) 11249 raise 11250 result.append(("disk_template", self.op.disk_template)) 11251 11252 # NIC changes 11253 for nic_op, nic_dict in self.op.nics: 11254 if nic_op == constants.DDM_REMOVE: 11255 # remove the last nic 11256 del instance.nics[-1] 11257 result.append(("nic.%d" % len(instance.nics), "remove")) 11258 elif nic_op == constants.DDM_ADD: 11259 # mac and bridge should be set, by now 11260 mac = nic_dict[constants.INIC_MAC] 11261 ip = nic_dict.get(constants.INIC_IP, None) 11262 nicparams = self.nic_pinst[constants.DDM_ADD] 11263 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams) 11264 instance.nics.append(new_nic) 11265 result.append(("nic.%d" % (len(instance.nics) - 1), 11266 "add:mac=%s,ip=%s,mode=%s,link=%s" % 11267 (new_nic.mac, new_nic.ip, 11268 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE], 11269 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK] 11270 ))) 11271 else: 11272 for key in (constants.INIC_MAC, constants.INIC_IP): 11273 if key in nic_dict: 11274 setattr(instance.nics[nic_op], key, nic_dict[key]) 11275 if nic_op in self.nic_pinst: 11276 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op] 11277 for key, val in nic_dict.iteritems(): 11278 result.append(("nic.%s/%d" % (key, nic_op), val)) 11279 11280 # hvparams changes 11281 if self.op.hvparams: 11282 instance.hvparams = self.hv_inst 11283 for key, val in self.op.hvparams.iteritems(): 11284 result.append(("hv/%s" % key, val)) 11285 11286 # beparams changes 11287 if self.op.beparams: 11288 instance.beparams = self.be_inst 11289 for key, val in self.op.beparams.iteritems(): 11290 result.append(("be/%s" % key, val)) 11291 11292 # OS change 11293 if self.op.os_name: 11294 instance.os = self.op.os_name 11295 11296 # osparams changes 11297 if self.op.osparams: 11298 instance.osparams = self.os_inst 11299 for key, val in self.op.osparams.iteritems(): 11300 result.append(("os/%s" % key, val)) 11301 11302 self.cfg.Update(instance, feedback_fn) 11303 11304 return result
11305 11306 _DISK_CONVERSIONS = { 11307 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd, 11308 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain, 11309 }
11310
11311 11312 -class LUInstanceChangeGroup(LogicalUnit):
11313 HPATH = "instance-change-group" 11314 HTYPE = constants.HTYPE_INSTANCE 11315 REQ_BGL = False 11316
11317 - def ExpandNames(self):
11318 self.share_locks = _ShareAll() 11319 self.needed_locks = { 11320 locking.LEVEL_NODEGROUP: [], 11321 locking.LEVEL_NODE: [], 11322 } 11323 11324 self._ExpandAndLockInstance() 11325 11326 if self.op.target_groups: 11327 self.req_target_uuids = map(self.cfg.LookupNodeGroup, 11328 self.op.target_groups) 11329 else: 11330 self.req_target_uuids = None 11331 11332 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
11333
11334 - def DeclareLocks(self, level):
11335 if level == locking.LEVEL_NODEGROUP: 11336 assert not self.needed_locks[locking.LEVEL_NODEGROUP] 11337 11338 if self.req_target_uuids: 11339 lock_groups = set(self.req_target_uuids) 11340 11341 # Lock all groups used by instance optimistically; this requires going 11342 # via the node before it's locked, requiring verification later on 11343 instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name) 11344 lock_groups.update(instance_groups) 11345 else: 11346 # No target groups, need to lock all of them 11347 lock_groups = locking.ALL_SET 11348 11349 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups 11350 11351 elif level == locking.LEVEL_NODE: 11352 if self.req_target_uuids: 11353 # Lock all nodes used by instances 11354 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND 11355 self._LockInstancesNodes() 11356 11357 # Lock all nodes in all potential target groups 11358 lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) - 11359 self.cfg.GetInstanceNodeGroups(self.op.instance_name)) 11360 member_nodes = [node_name 11361 for group in lock_groups 11362 for node_name in self.cfg.GetNodeGroup(group).members] 11363 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes) 11364 else: 11365 # Lock all nodes as all groups are potential targets 11366 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11367
11368 - def CheckPrereq(self):
11369 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE)) 11370 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) 11371 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) 11372 11373 assert (self.req_target_uuids is None or 11374 owned_groups.issuperset(self.req_target_uuids)) 11375 assert owned_instances == set([self.op.instance_name]) 11376 11377 # Get instance information 11378 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) 11379 11380 # Check if node groups for locked instance are still correct 11381 assert owned_nodes.issuperset(self.instance.all_nodes), \ 11382 ("Instance %s's nodes changed while we kept the lock" % 11383 self.op.instance_name) 11384 11385 inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, 11386 owned_groups) 11387 11388 if self.req_target_uuids: 11389 # User requested specific target groups 11390 self.target_uuids = frozenset(self.req_target_uuids) 11391 else: 11392 # All groups except those used by the instance are potential targets 11393 self.target_uuids = owned_groups - inst_groups 11394 11395 conflicting_groups = self.target_uuids & inst_groups 11396 if conflicting_groups: 11397 raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are" 11398 " used by the instance '%s'" % 11399 (utils.CommaJoin(conflicting_groups), 11400 self.op.instance_name), 11401 errors.ECODE_INVAL) 11402 11403 if not self.target_uuids: 11404 raise errors.OpPrereqError("There are no possible target groups", 11405 errors.ECODE_INVAL)
11406
11407 - def BuildHooksEnv(self):
11408 """Build hooks env. 11409 11410 """ 11411 assert self.target_uuids 11412 11413 env = { 11414 "TARGET_GROUPS": " ".join(self.target_uuids), 11415 } 11416 11417 env.update(_BuildInstanceHookEnvByObject(self, self.instance)) 11418 11419 return env
11420
11421 - def BuildHooksNodes(self):
11422 """Build hooks nodes. 11423 11424 """ 11425 mn = self.cfg.GetMasterNode() 11426 return ([mn], [mn])
11427
11428 - def Exec(self, feedback_fn):
11429 instances = list(self.owned_locks(locking.LEVEL_INSTANCE)) 11430 11431 assert instances == [self.op.instance_name], "Instance not locked" 11432 11433 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP, 11434 instances=instances, target_groups=list(self.target_uuids)) 11435 11436 ial.Run(self.op.iallocator) 11437 11438 if not ial.success: 11439 raise errors.OpPrereqError("Can't compute solution for changing group of" 11440 " instance '%s' using iallocator '%s': %s" % 11441 (self.op.instance_name, self.op.iallocator, 11442 ial.info), 11443 errors.ECODE_NORES) 11444 11445 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False) 11446 11447 self.LogInfo("Iallocator returned %s job(s) for changing group of" 11448 " instance '%s'", len(jobs), self.op.instance_name) 11449 11450 return ResultWithJobs(jobs)
11451
11452 11453 -class LUBackupQuery(NoHooksLU):
11454 """Query the exports list 11455 11456 """ 11457 REQ_BGL = False 11458
11459 - def ExpandNames(self):
11460 self.needed_locks = {} 11461 self.share_locks[locking.LEVEL_NODE] = 1 11462 if not self.op.nodes: 11463 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET 11464 else: 11465 self.needed_locks[locking.LEVEL_NODE] = \ 11466 _GetWantedNodes(self, self.op.nodes)
11467
11468 - def Exec(self, feedback_fn):
11469 """Compute the list of all the exported system images. 11470 11471 @rtype: dict 11472 @return: a dictionary with the structure node->(export-list) 11473 where export-list is a list of the instances exported on 11474 that node. 11475 11476 """ 11477 self.nodes = self.owned_locks(locking.LEVEL_NODE) 11478 rpcresult = self.rpc.call_export_list(self.nodes) 11479 result = {} 11480 for node in rpcresult: 11481 if rpcresult[node].fail_msg: 11482 result[node] = False 11483 else: 11484 result[node] = rpcresult[node].payload 11485 11486 return result
11487
11488 11489 -class LUBackupPrepare(NoHooksLU):
11490 """Prepares an instance for an export and returns useful information. 11491 11492 """ 11493 REQ_BGL = False 11494
11495 - def ExpandNames(self):
11497
11498 - def CheckPrereq(self):
11499 """Check prerequisites. 11500 11501 """ 11502 instance_name = self.op.instance_name 11503 11504 self.instance = self.cfg.GetInstanceInfo(instance_name) 11505 assert self.instance is not None, \ 11506 "Cannot retrieve locked instance %s" % self.op.instance_name 11507 _CheckNodeOnline(self, self.instance.primary_node) 11508 11509 self._cds = _GetClusterDomainSecret()
11510
11511 - def Exec(self, feedback_fn):
11512 """Prepares an instance for an export. 11513 11514 """ 11515 instance = self.instance 11516 11517 if self.op.mode == constants.EXPORT_MODE_REMOTE: 11518 salt = utils.GenerateSecret(8) 11519 11520 feedback_fn("Generating X509 certificate on %s" % instance.primary_node) 11521 result = self.rpc.call_x509_cert_create(instance.primary_node, 11522 constants.RIE_CERT_VALIDITY) 11523 result.Raise("Can't create X509 key and certificate on %s" % result.node) 11524 11525 (name, cert_pem) = result.payload 11526 11527 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM, 11528 cert_pem) 11529 11530 return { 11531 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds), 11532 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt), 11533 salt), 11534 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt), 11535 } 11536 11537 return None
11538
11539 11540 -class LUBackupExport(LogicalUnit):
11541 """Export an instance to an image in the cluster. 11542 11543 """ 11544 HPATH = "instance-export" 11545 HTYPE = constants.HTYPE_INSTANCE 11546 REQ_BGL = False 11547
11548 - def CheckArguments(self):
11549 """Check the arguments. 11550 11551 """ 11552 self.x509_key_name = self.op.x509_key_name 11553 self.dest_x509_ca_pem = self.op.destination_x509_ca 11554 11555 if self.op.mode == constants.EXPORT_MODE_REMOTE: 11556 if not self.x509_key_name: 11557 raise errors.OpPrereqError("Missing X509 key name for encryption", 11558 errors.ECODE_INVAL) 11559 11560 if not self.dest_x509_ca_pem: 11561 raise errors.OpPrereqError("Missing destination X509 CA", 11562 errors.ECODE_INVAL)
11563
11564 - def ExpandNames(self):
11565 self._ExpandAndLockInstance() 11566 11567 # Lock all nodes for local exports 11568 if self.op.mode == constants.EXPORT_MODE_LOCAL: 11569 # FIXME: lock only instance primary and destination node 11570 # 11571 # Sad but true, for now we have do lock all nodes, as we don't know where 11572 # the previous export might be, and in this LU we search for it and 11573 # remove it from its current node. In the future we could fix this by: 11574 # - making a tasklet to search (share-lock all), then create the 11575 # new one, then one to remove, after 11576 # - removing the removal operation altogether 11577 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11578
11579 - def DeclareLocks(self, level):
11580 """Last minute lock declaration."""
11581 # All nodes are locked anyway, so nothing to do here. 11582
11583 - def BuildHooksEnv(self):
11584 """Build hooks env. 11585 11586 This will run on the master, primary node and target node. 11587 11588 """ 11589 env = { 11590 "EXPORT_MODE": self.op.mode, 11591 "EXPORT_NODE": self.op.target_node, 11592 "EXPORT_DO_SHUTDOWN": self.op.shutdown, 11593 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout, 11594 # TODO: Generic function for boolean env variables 11595 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)), 11596 } 11597 11598 env.update(_BuildInstanceHookEnvByObject(self, self.instance)) 11599 11600 return env
11601
11602 - def BuildHooksNodes(self):
11603 """Build hooks nodes. 11604 11605 """ 11606 nl = [self.cfg.GetMasterNode(), self.instance.primary_node] 11607 11608 if self.op.mode == constants.EXPORT_MODE_LOCAL: 11609 nl.append(self.op.target_node) 11610 11611 return (nl, nl)
11612
11613 - def CheckPrereq(self):
11614 """Check prerequisites. 11615 11616 This checks that the instance and node names are valid. 11617 11618 """ 11619 instance_name = self.op.instance_name 11620 11621 self.instance = self.cfg.GetInstanceInfo(instance_name) 11622 assert self.instance is not None, \ 11623 "Cannot retrieve locked instance %s" % self.op.instance_name 11624 _CheckNodeOnline(self, self.instance.primary_node) 11625 11626 if (self.op.remove_instance and self.instance.admin_up and 11627 not self.op.shutdown): 11628 raise errors.OpPrereqError("Can not remove instance without shutting it" 11629 " down before") 11630 11631 if self.op.mode == constants.EXPORT_MODE_LOCAL: 11632 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node) 11633 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node) 11634 assert self.dst_node is not None 11635 11636 _CheckNodeOnline(self, self.dst_node.name) 11637 _CheckNodeNotDrained(self, self.dst_node.name) 11638 11639 self._cds = None 11640 self.dest_disk_info = None 11641 self.dest_x509_ca = None 11642 11643 elif self.op.mode == constants.EXPORT_MODE_REMOTE: 11644 self.dst_node = None 11645 11646 if len(self.op.target_node) != len(self.instance.disks): 11647 raise errors.OpPrereqError(("Received destination information for %s" 11648 " disks, but instance %s has %s disks") % 11649 (len(self.op.target_node), instance_name, 11650 len(self.instance.disks)), 11651 errors.ECODE_INVAL) 11652 11653 cds = _GetClusterDomainSecret() 11654 11655 # Check X509 key name 11656 try: 11657 (key_name, hmac_digest, hmac_salt) = self.x509_key_name 11658 except (TypeError, ValueError), err: 11659 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err) 11660 11661 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt): 11662 raise errors.OpPrereqError("HMAC for X509 key name is wrong", 11663 errors.ECODE_INVAL) 11664 11665 # Load and verify CA 11666 try: 11667 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds) 11668 except OpenSSL.crypto.Error, err: 11669 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" % 11670 (err, ), errors.ECODE_INVAL) 11671 11672 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None) 11673 if errcode is not None: 11674 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" % 11675 (msg, ), errors.ECODE_INVAL) 11676 11677 self.dest_x509_ca = cert 11678 11679 # Verify target information 11680 disk_info = [] 11681 for idx, disk_data in enumerate(self.op.target_node): 11682 try: 11683 (host, port, magic) = \ 11684 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data) 11685 except errors.GenericError, err: 11686 raise errors.OpPrereqError("Target info for disk %s: %s" % 11687 (idx, err), errors.ECODE_INVAL) 11688 11689 disk_info.append((host, port, magic)) 11690 11691 assert len(disk_info) == len(self.op.target_node) 11692 self.dest_disk_info = disk_info 11693 11694 else: 11695 raise errors.ProgrammerError("Unhandled export mode %r" % 11696 self.op.mode) 11697 11698 # instance disk type verification 11699 # TODO: Implement export support for file-based disks 11700 for disk in self.instance.disks: 11701 if disk.dev_type == constants.LD_FILE: 11702 raise errors.OpPrereqError("Export not supported for instances with" 11703 " file-based disks", errors.ECODE_INVAL)
11704
11705 - def _CleanupExports(self, feedback_fn):
11706 """Removes exports of current instance from all other nodes. 11707 11708 If an instance in a cluster with nodes A..D was exported to node C, its 11709 exports will be removed from the nodes A, B and D. 11710 11711 """ 11712 assert self.op.mode != constants.EXPORT_MODE_REMOTE 11713 11714 nodelist = self.cfg.GetNodeList() 11715 nodelist.remove(self.dst_node.name) 11716 11717 # on one-node clusters nodelist will be empty after the removal 11718 # if we proceed the backup would be removed because OpBackupQuery 11719 # substitutes an empty list with the full cluster node list. 11720 iname = self.instance.name 11721 if nodelist: 11722 feedback_fn("Removing old exports for instance %s" % iname) 11723 exportlist = self.rpc.call_export_list(nodelist) 11724 for node in exportlist: 11725 if exportlist[node].fail_msg: 11726 continue 11727 if iname in exportlist[node].payload: 11728 msg = self.rpc.call_export_remove(node, iname).fail_msg 11729 if msg: 11730 self.LogWarning("Could not remove older export for instance %s" 11731 " on node %s: %s", iname, node, msg)
11732
11733 - def Exec(self, feedback_fn):
11734 """Export an instance to an image in the cluster. 11735 11736 """ 11737 assert self.op.mode in constants.EXPORT_MODES 11738 11739 instance = self.instance 11740 src_node = instance.primary_node 11741 11742 if self.op.shutdown: 11743 # shutdown the instance, but not the disks 11744 feedback_fn("Shutting down instance %s" % instance.name) 11745 result = self.rpc.call_instance_shutdown(src_node, instance, 11746 self.op.shutdown_timeout) 11747 # TODO: Maybe ignore failures if ignore_remove_failures is set 11748 result.Raise("Could not shutdown instance %s on" 11749 " node %s" % (instance.name, src_node)) 11750 11751 # set the disks ID correctly since call_instance_start needs the 11752 # correct drbd minor to create the symlinks 11753 for disk in instance.disks: 11754 self.cfg.SetDiskID(disk, src_node) 11755 11756 activate_disks = (not instance.admin_up) 11757 11758 if activate_disks: 11759 # Activate the instance disks if we'exporting a stopped instance 11760 feedback_fn("Activating disks for %s" % instance.name) 11761 _StartInstanceDisks(self, instance, None) 11762 11763 try: 11764 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn, 11765 instance) 11766 11767 helper.CreateSnapshots() 11768 try: 11769 if (self.op.shutdown and instance.admin_up and 11770 not self.op.remove_instance): 11771 assert not activate_disks 11772 feedback_fn("Starting instance %s" % instance.name) 11773 result = self.rpc.call_instance_start(src_node, instance, 11774 None, None, False) 11775 msg = result.fail_msg 11776 if msg: 11777 feedback_fn("Failed to start instance: %s" % msg) 11778 _ShutdownInstanceDisks(self, instance) 11779 raise errors.OpExecError("Could not start instance: %s" % msg) 11780 11781 if self.op.mode == constants.EXPORT_MODE_LOCAL: 11782 (fin_resu, dresults) = helper.LocalExport(self.dst_node) 11783 elif self.op.mode == constants.EXPORT_MODE_REMOTE: 11784 connect_timeout = constants.RIE_CONNECT_TIMEOUT 11785 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout) 11786 11787 (key_name, _, _) = self.x509_key_name 11788 11789 dest_ca_pem = \ 11790 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM, 11791 self.dest_x509_ca) 11792 11793 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info, 11794 key_name, dest_ca_pem, 11795 timeouts) 11796 finally: 11797 helper.Cleanup() 11798 11799 # Check for backwards compatibility 11800 assert len(dresults) == len(instance.disks) 11801 assert compat.all(isinstance(i, bool) for i in dresults), \ 11802 "Not all results are boolean: %r" % dresults 11803 11804 finally: 11805 if activate_disks: 11806 feedback_fn("Deactivating disks for %s" % instance.name) 11807 _ShutdownInstanceDisks(self, instance) 11808 11809 if not (compat.all(dresults) and fin_resu): 11810 failures = [] 11811 if not fin_resu: 11812 failures.append("export finalization") 11813 if not compat.all(dresults): 11814 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults) 11815 if not dsk) 11816 failures.append("disk export: disk(s) %s" % fdsk) 11817 11818 raise errors.OpExecError("Export failed, errors in %s" % 11819 utils.CommaJoin(failures)) 11820 11821 # At this point, the export was successful, we can cleanup/finish 11822 11823 # Remove instance if requested 11824 if self.op.remove_instance: 11825 feedback_fn("Removing instance %s" % instance.name) 11826 _RemoveInstance(self, feedback_fn, instance, 11827 self.op.ignore_remove_failures) 11828 11829 if self.op.mode == constants.EXPORT_MODE_LOCAL: 11830 self._CleanupExports(feedback_fn) 11831 11832 return fin_resu, dresults
11833
11834 11835 -class LUBackupRemove(NoHooksLU):
11836 """Remove exports related to the named instance. 11837 11838 """ 11839 REQ_BGL = False 11840
11841 - def ExpandNames(self):
11842 self.needed_locks = {} 11843 # We need all nodes to be locked in order for RemoveExport to work, but we 11844 # don't need to lock the instance itself, as nothing will happen to it (and 11845 # we can remove exports also for a removed instance) 11846 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11847
11848 - def Exec(self, feedback_fn):
11849 """Remove any export. 11850 11851 """ 11852 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name) 11853 # If the instance was not found we'll try with the name that was passed in. 11854 # This will only work if it was an FQDN, though. 11855 fqdn_warn = False 11856 if not instance_name: 11857 fqdn_warn = True 11858 instance_name = self.op.instance_name 11859 11860 locked_nodes = self.owned_locks(locking.LEVEL_NODE) 11861 exportlist = self.rpc.call_export_list(locked_nodes) 11862 found = False 11863 for node in exportlist: 11864 msg = exportlist[node].fail_msg 11865 if msg: 11866 self.LogWarning("Failed to query node %s (continuing): %s", node, msg) 11867 continue 11868 if instance_name in exportlist[node].payload: 11869 found = True 11870 result = self.rpc.call_export_remove(node, instance_name) 11871 msg = result.fail_msg 11872 if msg: 11873 logging.error("Could not remove export for instance %s" 11874 " on node %s: %s", instance_name, node, msg) 11875 11876 if fqdn_warn and not found: 11877 feedback_fn("Export not found. If trying to remove an export belonging" 11878 " to a deleted instance please use its Fully Qualified" 11879 " Domain Name.")
11880
11881 11882 -class LUGroupAdd(LogicalUnit):
11883 """Logical unit for creating node groups. 11884 11885 """ 11886 HPATH = "group-add" 11887 HTYPE = constants.HTYPE_GROUP 11888 REQ_BGL = False 11889
11890 - def ExpandNames(self):
11891 # We need the new group's UUID here so that we can create and acquire the 11892 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup 11893 # that it should not check whether the UUID exists in the configuration. 11894 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId()) 11895 self.needed_locks = {} 11896 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
11897
11898 - def CheckPrereq(self):
11899 """Check prerequisites. 11900 11901 This checks that the given group name is not an existing node group 11902 already. 11903 11904 """ 11905 try: 11906 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 11907 except errors.OpPrereqError: 11908 pass 11909 else: 11910 raise errors.OpPrereqError("Desired group name '%s' already exists as a" 11911 " node group (UUID: %s)" % 11912 (self.op.group_name, existing_uuid), 11913 errors.ECODE_EXISTS) 11914 11915 if self.op.ndparams: 11916 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
11917
11918 - def BuildHooksEnv(self):
11919 """Build hooks env. 11920 11921 """ 11922 return { 11923 "GROUP_NAME": self.op.group_name, 11924 }
11925
11926 - def BuildHooksNodes(self):
11927 """Build hooks nodes. 11928 11929 """ 11930 mn = self.cfg.GetMasterNode() 11931 return ([mn], [mn])
11932
11933 - def Exec(self, feedback_fn):
11934 """Add the node group to the cluster. 11935 11936 """ 11937 group_obj = objects.NodeGroup(name=self.op.group_name, members=[], 11938 uuid=self.group_uuid, 11939 alloc_policy=self.op.alloc_policy, 11940 ndparams=self.op.ndparams) 11941 11942 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False) 11943 del self.remove_locks[locking.LEVEL_NODEGROUP]
11944
11945 11946 -class LUGroupAssignNodes(NoHooksLU):
11947 """Logical unit for assigning nodes to groups. 11948 11949 """ 11950 REQ_BGL = False 11951
11952 - def ExpandNames(self):
11953 # These raise errors.OpPrereqError on their own: 11954 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 11955 self.op.nodes = _GetWantedNodes(self, self.op.nodes) 11956 11957 # We want to lock all the affected nodes and groups. We have readily 11958 # available the list of nodes, and the *destination* group. To gather the 11959 # list of "source" groups, we need to fetch node information later on. 11960 self.needed_locks = { 11961 locking.LEVEL_NODEGROUP: set([self.group_uuid]), 11962 locking.LEVEL_NODE: self.op.nodes, 11963 }
11964
11965 - def DeclareLocks(self, level):
11966 if level == locking.LEVEL_NODEGROUP: 11967 assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1 11968 11969 # Try to get all affected nodes' groups without having the group or node 11970 # lock yet. Needs verification later in the code flow. 11971 groups = self.cfg.GetNodeGroupsFromNodes(self.op.nodes) 11972 11973 self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)
11974
11975 - def CheckPrereq(self):
11976 """Check prerequisites. 11977 11978 """ 11979 assert self.needed_locks[locking.LEVEL_NODEGROUP] 11980 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) == 11981 frozenset(self.op.nodes)) 11982 11983 expected_locks = (set([self.group_uuid]) | 11984 self.cfg.GetNodeGroupsFromNodes(self.op.nodes)) 11985 actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP) 11986 if actual_locks != expected_locks: 11987 raise errors.OpExecError("Nodes changed groups since locks were acquired," 11988 " current groups are '%s', used to be '%s'" % 11989 (utils.CommaJoin(expected_locks), 11990 utils.CommaJoin(actual_locks))) 11991 11992 self.node_data = self.cfg.GetAllNodesInfo() 11993 self.group = self.cfg.GetNodeGroup(self.group_uuid) 11994 instance_data = self.cfg.GetAllInstancesInfo() 11995 11996 if self.group is None: 11997 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" % 11998 (self.op.group_name, self.group_uuid)) 11999 12000 (new_splits, previous_splits) = \ 12001 self.CheckAssignmentForSplitInstances([(node, self.group_uuid) 12002 for node in self.op.nodes], 12003 self.node_data, instance_data) 12004 12005 if new_splits: 12006 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits)) 12007 12008 if not self.op.force: 12009 raise errors.OpExecError("The following instances get split by this" 12010 " change and --force was not given: %s" % 12011 fmt_new_splits) 12012 else: 12013 self.LogWarning("This operation will split the following instances: %s", 12014 fmt_new_splits) 12015 12016 if previous_splits: 12017 self.LogWarning("In addition, these already-split instances continue" 12018 " to be split across groups: %s", 12019 utils.CommaJoin(utils.NiceSort(previous_splits)))
12020
12021 - def Exec(self, feedback_fn):
12022 """Assign nodes to a new group. 12023 12024 """ 12025 mods = [(node_name, self.group_uuid) for node_name in self.op.nodes] 12026 12027 self.cfg.AssignGroupNodes(mods)
12028 12029 @staticmethod
12030 - def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
12031 """Check for split instances after a node assignment. 12032 12033 This method considers a series of node assignments as an atomic operation, 12034 and returns information about split instances after applying the set of 12035 changes. 12036 12037 In particular, it returns information about newly split instances, and 12038 instances that were already split, and remain so after the change. 12039 12040 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are 12041 considered. 12042 12043 @type changes: list of (node_name, new_group_uuid) pairs. 12044 @param changes: list of node assignments to consider. 12045 @param node_data: a dict with data for all nodes 12046 @param instance_data: a dict with all instances to consider 12047 @rtype: a two-tuple 12048 @return: a list of instances that were previously okay and result split as a 12049 consequence of this change, and a list of instances that were previously 12050 split and this change does not fix. 12051 12052 """ 12053 changed_nodes = dict((node, group) for node, group in changes 12054 if node_data[node].group != group) 12055 12056 all_split_instances = set() 12057 previously_split_instances = set() 12058 12059 def InstanceNodes(instance): 12060 return [instance.primary_node] + list(instance.secondary_nodes)
12061 12062 for inst in instance_data.values(): 12063 if inst.disk_template not in constants.DTS_INT_MIRROR: 12064 continue 12065 12066 instance_nodes = InstanceNodes(inst) 12067 12068 if len(set(node_data[node].group for node in instance_nodes)) > 1: 12069 previously_split_instances.add(inst.name) 12070 12071 if len(set(changed_nodes.get(node, node_data[node].group) 12072 for node in instance_nodes)) > 1: 12073 all_split_instances.add(inst.name) 12074 12075 return (list(all_split_instances - previously_split_instances), 12076 list(previously_split_instances & all_split_instances))
12077
12078 12079 -class _GroupQuery(_QueryBase):
12080 FIELDS = query.GROUP_FIELDS 12081
12082 - def ExpandNames(self, lu):
12083 lu.needed_locks = {} 12084 12085 self._all_groups = lu.cfg.GetAllNodeGroupsInfo() 12086 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values()) 12087 12088 if not self.names: 12089 self.wanted = [name_to_uuid[name] 12090 for name in utils.NiceSort(name_to_uuid.keys())] 12091 else: 12092 # Accept names to be either names or UUIDs. 12093 missing = [] 12094 self.wanted = [] 12095 all_uuid = frozenset(self._all_groups.keys()) 12096 12097 for name in self.names: 12098 if name in all_uuid: 12099 self.wanted.append(name) 12100 elif name in name_to_uuid: 12101 self.wanted.append(name_to_uuid[name]) 12102 else: 12103 missing.append(name) 12104 12105 if missing: 12106 raise errors.OpPrereqError("Some groups do not exist: %s" % 12107 utils.CommaJoin(missing), 12108 errors.ECODE_NOENT)
12109
12110 - def DeclareLocks(self, lu, level):
12111 pass
12112
12113 - def _GetQueryData(self, lu):
12114 """Computes the list of node groups and their attributes. 12115 12116 """ 12117 do_nodes = query.GQ_NODE in self.requested_data 12118 do_instances = query.GQ_INST in self.requested_data 12119 12120 group_to_nodes = None 12121 group_to_instances = None 12122 12123 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for 12124 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the 12125 # latter GetAllInstancesInfo() is not enough, for we have to go through 12126 # instance->node. Hence, we will need to process nodes even if we only need 12127 # instance information. 12128 if do_nodes or do_instances: 12129 all_nodes = lu.cfg.GetAllNodesInfo() 12130 group_to_nodes = dict((uuid, []) for uuid in self.wanted) 12131 node_to_group = {} 12132 12133 for node in all_nodes.values(): 12134 if node.group in group_to_nodes: 12135 group_to_nodes[node.group].append(node.name) 12136 node_to_group[node.name] = node.group 12137 12138 if do_instances: 12139 all_instances = lu.cfg.GetAllInstancesInfo() 12140 group_to_instances = dict((uuid, []) for uuid in self.wanted) 12141 12142 for instance in all_instances.values(): 12143 node = instance.primary_node 12144 if node in node_to_group: 12145 group_to_instances[node_to_group[node]].append(instance.name) 12146 12147 if not do_nodes: 12148 # Do not pass on node information if it was not requested. 12149 group_to_nodes = None 12150 12151 return query.GroupQueryData([self._all_groups[uuid] 12152 for uuid in self.wanted], 12153 group_to_nodes, group_to_instances)
12154
12155 12156 -class LUGroupQuery(NoHooksLU):
12157 """Logical unit for querying node groups. 12158 12159 """ 12160 REQ_BGL = False 12161
12162 - def CheckArguments(self):
12163 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names), 12164 self.op.output_fields, False)
12165
12166 - def ExpandNames(self):
12167 self.gq.ExpandNames(self)
12168
12169 - def DeclareLocks(self, level):
12170 self.gq.DeclareLocks(self, level)
12171
12172 - def Exec(self, feedback_fn):
12173 return self.gq.OldStyleQuery(self)
12174
12175 12176 -class LUGroupSetParams(LogicalUnit):
12177 """Modifies the parameters of a node group. 12178 12179 """ 12180 HPATH = "group-modify" 12181 HTYPE = constants.HTYPE_GROUP 12182 REQ_BGL = False 12183
12184 - def CheckArguments(self):
12185 all_changes = [ 12186 self.op.ndparams, 12187 self.op.alloc_policy, 12188 ] 12189 12190 if all_changes.count(None) == len(all_changes): 12191 raise errors.OpPrereqError("Please pass at least one modification", 12192 errors.ECODE_INVAL)
12193
12194 - def ExpandNames(self):
12195 # This raises errors.OpPrereqError on its own: 12196 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 12197 12198 self.needed_locks = { 12199 locking.LEVEL_NODEGROUP: [self.group_uuid], 12200 }
12201
12202 - def CheckPrereq(self):
12203 """Check prerequisites. 12204 12205 """ 12206 self.group = self.cfg.GetNodeGroup(self.group_uuid) 12207 12208 if self.group is None: 12209 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" % 12210 (self.op.group_name, self.group_uuid)) 12211 12212 if self.op.ndparams: 12213 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams) 12214 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES) 12215 self.new_ndparams = new_ndparams
12216
12217 - def BuildHooksEnv(self):
12218 """Build hooks env. 12219 12220 """ 12221 return { 12222 "GROUP_NAME": self.op.group_name, 12223 "NEW_ALLOC_POLICY": self.op.alloc_policy, 12224 }
12225
12226 - def BuildHooksNodes(self):
12227 """Build hooks nodes. 12228 12229 """ 12230 mn = self.cfg.GetMasterNode() 12231 return ([mn], [mn])
12232
12233 - def Exec(self, feedback_fn):
12234 """Modifies the node group. 12235 12236 """ 12237 result = [] 12238 12239 if self.op.ndparams: 12240 self.group.ndparams = self.new_ndparams 12241 result.append(("ndparams", str(self.group.ndparams))) 12242 12243 if self.op.alloc_policy: 12244 self.group.alloc_policy = self.op.alloc_policy 12245 12246 self.cfg.Update(self.group, feedback_fn) 12247 return result
12248
12249 12250 -class LUGroupRemove(LogicalUnit):
12251 HPATH = "group-remove" 12252 HTYPE = constants.HTYPE_GROUP 12253 REQ_BGL = False 12254
12255 - def ExpandNames(self):
12256 # This will raises errors.OpPrereqError on its own: 12257 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 12258 self.needed_locks = { 12259 locking.LEVEL_NODEGROUP: [self.group_uuid], 12260 }
12261
12262 - def CheckPrereq(self):
12263 """Check prerequisites. 12264 12265 This checks that the given group name exists as a node group, that is 12266 empty (i.e., contains no nodes), and that is not the last group of the 12267 cluster. 12268 12269 """ 12270 # Verify that the group is empty. 12271 group_nodes = [node.name 12272 for node in self.cfg.GetAllNodesInfo().values() 12273 if node.group == self.group_uuid] 12274 12275 if group_nodes: 12276 raise errors.OpPrereqError("Group '%s' not empty, has the following" 12277 " nodes: %s" % 12278 (self.op.group_name, 12279 utils.CommaJoin(utils.NiceSort(group_nodes))), 12280 errors.ECODE_STATE) 12281 12282 # Verify the cluster would not be left group-less. 12283 if len(self.cfg.GetNodeGroupList()) == 1: 12284 raise errors.OpPrereqError("Group '%s' is the only group," 12285 " cannot be removed" % 12286 self.op.group_name, 12287 errors.ECODE_STATE)
12288
12289 - def BuildHooksEnv(self):
12290 """Build hooks env. 12291 12292 """ 12293 return { 12294 "GROUP_NAME": self.op.group_name, 12295 }
12296
12297 - def BuildHooksNodes(self):
12298 """Build hooks nodes. 12299 12300 """ 12301 mn = self.cfg.GetMasterNode() 12302 return ([mn], [mn])
12303
12304 - def Exec(self, feedback_fn):
12305 """Remove the node group. 12306 12307 """ 12308 try: 12309 self.cfg.RemoveNodeGroup(self.group_uuid) 12310 except errors.ConfigurationError: 12311 raise errors.OpExecError("Group '%s' with UUID %s disappeared" % 12312 (self.op.group_name, self.group_uuid)) 12313 12314 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
12315
12316 12317 -class LUGroupRename(LogicalUnit):
12318 HPATH = "group-rename" 12319 HTYPE = constants.HTYPE_GROUP 12320 REQ_BGL = False 12321
12322 - def ExpandNames(self):
12323 # This raises errors.OpPrereqError on its own: 12324 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 12325 12326 self.needed_locks = { 12327 locking.LEVEL_NODEGROUP: [self.group_uuid], 12328 }
12329
12330 - def CheckPrereq(self):
12331 """Check prerequisites. 12332 12333 Ensures requested new name is not yet used. 12334 12335 """ 12336 try: 12337 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name) 12338 except errors.OpPrereqError: 12339 pass 12340 else: 12341 raise errors.OpPrereqError("Desired new name '%s' clashes with existing" 12342 " node group (UUID: %s)" % 12343 (self.op.new_name, new_name_uuid), 12344 errors.ECODE_EXISTS)
12345
12346 - def BuildHooksEnv(self):
12347 """Build hooks env. 12348 12349 """ 12350 return { 12351 "OLD_NAME": self.op.group_name, 12352 "NEW_NAME": self.op.new_name, 12353 }
12354
12355 - def BuildHooksNodes(self):
12356 """Build hooks nodes. 12357 12358 """ 12359 mn = self.cfg.GetMasterNode() 12360 12361 all_nodes = self.cfg.GetAllNodesInfo() 12362 all_nodes.pop(mn, None) 12363 12364 run_nodes = [mn] 12365 run_nodes.extend(node.name for node in all_nodes.values() 12366 if node.group == self.group_uuid) 12367 12368 return (run_nodes, run_nodes)
12369
12370 - def Exec(self, feedback_fn):
12371 """Rename the node group. 12372 12373 """ 12374 group = self.cfg.GetNodeGroup(self.group_uuid) 12375 12376 if group is None: 12377 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" % 12378 (self.op.group_name, self.group_uuid)) 12379 12380 group.name = self.op.new_name 12381 self.cfg.Update(group, feedback_fn) 12382 12383 return self.op.new_name
12384
12385 12386 -class LUGroupEvacuate(LogicalUnit):
12387 HPATH = "group-evacuate" 12388 HTYPE = constants.HTYPE_GROUP 12389 REQ_BGL = False 12390
12391 - def ExpandNames(self):
12392 # This raises errors.OpPrereqError on its own: 12393 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name) 12394 12395 if self.op.target_groups: 12396 self.req_target_uuids = map(self.cfg.LookupNodeGroup, 12397 self.op.target_groups) 12398 else: 12399 self.req_target_uuids = [] 12400 12401 if self.group_uuid in self.req_target_uuids: 12402 raise errors.OpPrereqError("Group to be evacuated (%s) can not be used" 12403 " as a target group (targets are %s)" % 12404 (self.group_uuid, 12405 utils.CommaJoin(self.req_target_uuids)), 12406 errors.ECODE_INVAL) 12407 12408 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator) 12409 12410 self.share_locks = _ShareAll() 12411 self.needed_locks = { 12412 locking.LEVEL_INSTANCE: [], 12413 locking.LEVEL_NODEGROUP: [], 12414 locking.LEVEL_NODE: [], 12415 }
12416
12417 - def DeclareLocks(self, level):
12418 if level == locking.LEVEL_INSTANCE: 12419 assert not self.needed_locks[locking.LEVEL_INSTANCE] 12420 12421 # Lock instances optimistically, needs verification once node and group 12422 # locks have been acquired 12423 self.needed_locks[locking.LEVEL_INSTANCE] = \ 12424 self.cfg.GetNodeGroupInstances(self.group_uuid) 12425 12426 elif level == locking.LEVEL_NODEGROUP: 12427 assert not self.needed_locks[locking.LEVEL_NODEGROUP] 12428 12429 if self.req_target_uuids: 12430 lock_groups = set([self.group_uuid] + self.req_target_uuids) 12431 12432 # Lock all groups used by instances optimistically; this requires going 12433 # via the node before it's locked, requiring verification later on 12434 lock_groups.update(group_uuid 12435 for instance_name in 12436 self.owned_locks(locking.LEVEL_INSTANCE) 12437 for group_uuid in 12438 self.cfg.GetInstanceNodeGroups(instance_name)) 12439 else: 12440 # No target groups, need to lock all of them 12441 lock_groups = locking.ALL_SET 12442 12443 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups 12444 12445 elif level == locking.LEVEL_NODE: 12446 # This will only lock the nodes in the group to be evacuated which 12447 # contain actual instances 12448 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND 12449 self._LockInstancesNodes() 12450 12451 # Lock all nodes in group to be evacuated and target groups 12452 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) 12453 assert self.group_uuid in owned_groups 12454 member_nodes = [node_name 12455 for group in owned_groups 12456 for node_name in self.cfg.GetNodeGroup(group).members] 12457 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
12458
12459 - def CheckPrereq(self):
12460 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE)) 12461 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) 12462 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) 12463 12464 assert owned_groups.issuperset(self.req_target_uuids) 12465 assert self.group_uuid in owned_groups 12466 12467 # Check if locked instances are still correct 12468 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances) 12469 12470 # Get instance information 12471 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances)) 12472 12473 # Check if node groups for locked instances are still correct 12474 for instance_name in owned_instances: 12475 inst = self.instances[instance_name] 12476 assert owned_nodes.issuperset(inst.all_nodes), \ 12477 "Instance %s's nodes changed while we kept the lock" % instance_name 12478 12479 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name, 12480 owned_groups) 12481 12482 assert self.group_uuid in inst_groups, \ 12483 "Instance %s has no node in group %s" % (instance_name, self.group_uuid) 12484 12485 if self.req_target_uuids: 12486 # User requested specific target groups 12487 self.target_uuids = self.req_target_uuids 12488 else: 12489 # All groups except the one to be evacuated are potential targets 12490 self.target_uuids = [group_uuid for group_uuid in owned_groups 12491 if group_uuid != self.group_uuid] 12492 12493 if not self.target_uuids: 12494 raise errors.OpPrereqError("There are no possible target groups", 12495 errors.ECODE_INVAL)
12496
12497 - def BuildHooksEnv(self):
12498 """Build hooks env. 12499 12500 """ 12501 return { 12502 "GROUP_NAME": self.op.group_name, 12503 "TARGET_GROUPS": " ".join(self.target_uuids), 12504 }
12505
12506 - def BuildHooksNodes(self):
12507 """Build hooks nodes. 12508 12509 """ 12510 mn = self.cfg.GetMasterNode() 12511 12512 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP) 12513 12514 run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members 12515 12516 return (run_nodes, run_nodes)
12517
12518 - def Exec(self, feedback_fn):
12519 instances = list(self.owned_locks(locking.LEVEL_INSTANCE)) 12520 12521 assert self.group_uuid not in self.target_uuids 12522 12523 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP, 12524 instances=instances, target_groups=self.target_uuids) 12525 12526 ial.Run(self.op.iallocator) 12527 12528 if not ial.success: 12529 raise errors.OpPrereqError("Can't compute group evacuation using" 12530 " iallocator '%s': %s" % 12531 (self.op.iallocator, ial.info), 12532 errors.ECODE_NORES) 12533 12534 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False) 12535 12536 self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s", 12537 len(jobs), self.op.group_name) 12538 12539 return ResultWithJobs(jobs)
12540
12541 12542 -class TagsLU(NoHooksLU): # pylint: disable=W0223
12543 """Generic tags LU. 12544 12545 This is an abstract class which is the parent of all the other tags LUs. 12546 12547 """
12548 - def ExpandNames(self):
12549 self.group_uuid = None 12550 self.needed_locks = {} 12551 if self.op.kind == constants.TAG_NODE: 12552 self.op.name = _ExpandNodeName(self.cfg, self.op.name) 12553 self.needed_locks[locking.LEVEL_NODE] = self.op.name 12554 elif self.op.kind == constants.TAG_INSTANCE: 12555 self.op.name = _ExpandInstanceName(self.cfg, self.op.name) 12556 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name 12557 elif self.op.kind == constants.TAG_NODEGROUP: 12558 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
12559 12560 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's 12561 # not possible to acquire the BGL based on opcode parameters) 12562
12563 - def CheckPrereq(self):
12564 """Check prerequisites. 12565 12566 """ 12567 if self.op.kind == constants.TAG_CLUSTER: 12568 self.target = self.cfg.GetClusterInfo() 12569 elif self.op.kind == constants.TAG_NODE: 12570 self.target = self.cfg.GetNodeInfo(self.op.name) 12571 elif self.op.kind == constants.TAG_INSTANCE: 12572 self.target = self.cfg.GetInstanceInfo(self.op.name) 12573 elif self.op.kind == constants.TAG_NODEGROUP: 12574 self.target = self.cfg.GetNodeGroup(self.group_uuid) 12575 else: 12576 raise errors.OpPrereqError("Wrong tag type requested (%s)" % 12577 str(self.op.kind), errors.ECODE_INVAL)
12578
12579 12580 -class LUTagsGet(TagsLU):
12581 """Returns the tags of a given object. 12582 12583 """ 12584 REQ_BGL = False 12585
12586 - def ExpandNames(self):
12587 TagsLU.ExpandNames(self) 12588 12589 # Share locks as this is only a read operation 12590 self.share_locks = _ShareAll()
12591
12592 - def Exec(self, feedback_fn):
12593 """Returns the tag list. 12594 12595 """ 12596 return list(self.target.GetTags())
12597
12598 12599 -class LUTagsSearch(NoHooksLU):
12600 """Searches the tags for a given pattern. 12601 12602 """ 12603 REQ_BGL = False 12604
12605 - def ExpandNames(self):
12606 self.needed_locks = {}
12607
12608 - def CheckPrereq(self):
12609 """Check prerequisites. 12610 12611 This checks the pattern passed for validity by compiling it. 12612 12613 """ 12614 try: 12615 self.re = re.compile(self.op.pattern) 12616 except re.error, err: 12617 raise errors.OpPrereqError("Invalid search pattern '%s': %s" % 12618 (self.op.pattern, err), errors.ECODE_INVAL)
12619
12620 - def Exec(self, feedback_fn):
12621 """Returns the tag list. 12622 12623 """ 12624 cfg = self.cfg 12625 tgts = [("/cluster", cfg.GetClusterInfo())] 12626 ilist = cfg.GetAllInstancesInfo().values() 12627 tgts.extend([("/instances/%s" % i.name, i) for i in ilist]) 12628 nlist = cfg.GetAllNodesInfo().values() 12629 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist]) 12630 tgts.extend(("/nodegroup/%s" % n.name, n) 12631 for n in cfg.GetAllNodeGroupsInfo().values()) 12632 results = [] 12633 for path, target in tgts: 12634 for tag in target.GetTags(): 12635 if self.re.search(tag): 12636 results.append((path, tag)) 12637 return results
12638
12639 12640 -class LUTagsSet(TagsLU):
12641 """Sets a tag on a given object. 12642 12643 """ 12644 REQ_BGL = False 12645
12646 - def CheckPrereq(self):
12647 """Check prerequisites. 12648 12649 This checks the type and length of the tag name and value. 12650 12651 """ 12652 TagsLU.CheckPrereq(self) 12653 for tag in self.op.tags: 12654 objects.TaggableObject.ValidateTag(tag)
12655
12656 - def Exec(self, feedback_fn):
12657 """Sets the tag. 12658 12659 """ 12660 try: 12661 for tag in self.op.tags: 12662 self.target.AddTag(tag) 12663 except errors.TagError, err: 12664 raise errors.OpExecError("Error while setting tag: %s" % str(err)) 12665 self.cfg.Update(self.target, feedback_fn)
12666
12667 12668 -class LUTagsDel(TagsLU):
12669 """Delete a list of tags from a given object. 12670 12671 """ 12672 REQ_BGL = False 12673
12674 - def CheckPrereq(self):
12675 """Check prerequisites. 12676 12677 This checks that we have the given tag. 12678 12679 """ 12680 TagsLU.CheckPrereq(self) 12681 for tag in self.op.tags: 12682 objects.TaggableObject.ValidateTag(tag) 12683 del_tags = frozenset(self.op.tags) 12684 cur_tags = self.target.GetTags() 12685 12686 diff_tags = del_tags - cur_tags 12687 if diff_tags: 12688 diff_names = ("'%s'" % i for i in sorted(diff_tags)) 12689 raise errors.OpPrereqError("Tag(s) %s not found" % 12690 (utils.CommaJoin(diff_names), ), 12691 errors.ECODE_NOENT)
12692
12693 - def Exec(self, feedback_fn):
12694 """Remove the tag from the object. 12695 12696 """ 12697 for tag in self.op.tags: 12698 self.target.RemoveTag(tag) 12699 self.cfg.Update(self.target, feedback_fn)
12700
12701 12702 -class LUTestDelay(NoHooksLU):
12703 """Sleep for a specified amount of time. 12704 12705 This LU sleeps on the master and/or nodes for a specified amount of 12706 time. 12707 12708 """ 12709 REQ_BGL = False 12710
12711 - def ExpandNames(self):
12712 """Expand names and set required locks. 12713 12714 This expands the node list, if any. 12715 12716 """ 12717 self.needed_locks = {} 12718 if self.op.on_nodes: 12719 # _GetWantedNodes can be used here, but is not always appropriate to use 12720 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for 12721 # more information. 12722 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes) 12723 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
12724
12725 - def _TestDelay(self):
12726 """Do the actual sleep. 12727 12728 """ 12729 if self.op.on_master: 12730 if not utils.TestDelay(self.op.duration): 12731 raise errors.OpExecError("Error during master delay test") 12732 if self.op.on_nodes: 12733 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration) 12734 for node, node_result in result.items(): 12735 node_result.Raise("Failure during rpc call to node %s" % node)
12736
12737 - def Exec(self, feedback_fn):
12738 """Execute the test delay opcode, with the wanted repetitions. 12739 12740 """ 12741 if self.op.repeat == 0: 12742 self._TestDelay() 12743 else: 12744 top_value = self.op.repeat - 1 12745 for i in range(self.op.repeat): 12746 self.LogInfo("Test delay iteration %d/%d" % (i, top_value)) 12747 self._TestDelay()
12748
12749 12750 -class LUTestJqueue(NoHooksLU):
12751 """Utility LU to test some aspects of the job queue. 12752 12753 """ 12754 REQ_BGL = False 12755 12756 # Must be lower than default timeout for WaitForJobChange to see whether it 12757 # notices changed jobs 12758 _CLIENT_CONNECT_TIMEOUT = 20.0 12759 _CLIENT_CONFIRM_TIMEOUT = 60.0 12760 12761 @classmethod
12762 - def _NotifyUsingSocket(cls, cb, errcls):
12763 """Opens a Unix socket and waits for another program to connect. 12764 12765 @type cb: callable 12766 @param cb: Callback to send socket name to client 12767 @type errcls: class 12768 @param errcls: Exception class to use for errors 12769 12770 """ 12771 # Using a temporary directory as there's no easy way to create temporary 12772 # sockets without writing a custom loop around tempfile.mktemp and 12773 # socket.bind 12774 tmpdir = tempfile.mkdtemp() 12775 try: 12776 tmpsock = utils.PathJoin(tmpdir, "sock") 12777 12778 logging.debug("Creating temporary socket at %s", tmpsock) 12779 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) 12780 try: 12781 sock.bind(tmpsock) 12782 sock.listen(1) 12783 12784 # Send details to client 12785 cb(tmpsock) 12786 12787 # Wait for client to connect before continuing 12788 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT) 12789 try: 12790 (conn, _) = sock.accept() 12791 except socket.error, err: 12792 raise errcls("Client didn't connect in time (%s)" % err) 12793 finally: 12794 sock.close() 12795 finally: 12796 # Remove as soon as client is connected 12797 shutil.rmtree(tmpdir) 12798 12799 # Wait for client to close 12800 try: 12801 try: 12802 # pylint: disable=E1101 12803 # Instance of '_socketobject' has no ... member 12804 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT) 12805 conn.recv(1) 12806 except socket.error, err: 12807 raise errcls("Client failed to confirm notification (%s)" % err) 12808 finally: 12809 conn.close()
12810
12811 - def _SendNotification(self, test, arg, sockname):
12812 """Sends a notification to the client. 12813 12814 @type test: string 12815 @param test: Test name 12816 @param arg: Test argument (depends on test) 12817 @type sockname: string 12818 @param sockname: Socket path 12819 12820 """ 12821 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
12822
12823 - def _Notify(self, prereq, test, arg):
12824 """Notifies the client of a test. 12825 12826 @type prereq: bool 12827 @param prereq: Whether this is a prereq-phase test 12828 @type test: string 12829 @param test: Test name 12830 @param arg: Test argument (depends on test) 12831 12832 """ 12833 if prereq: 12834 errcls = errors.OpPrereqError 12835 else: 12836 errcls = errors.OpExecError 12837 12838 return self._NotifyUsingSocket(compat.partial(self._SendNotification, 12839 test, arg), 12840 errcls)
12841
12842 - def CheckArguments(self):
12843 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1 12844 self.expandnames_calls = 0
12845
12846 - def ExpandNames(self):
12847 checkargs_calls = getattr(self, "checkargs_calls", 0) 12848 if checkargs_calls < 1: 12849 raise errors.ProgrammerError("CheckArguments was not called") 12850 12851 self.expandnames_calls += 1 12852 12853 if self.op.notify_waitlock: 12854 self._Notify(True, constants.JQT_EXPANDNAMES, None) 12855 12856 self.LogInfo("Expanding names") 12857 12858 # Get lock on master node (just to get a lock, not for a particular reason) 12859 self.needed_locks = { 12860 locking.LEVEL_NODE: self.cfg.GetMasterNode(), 12861 }
12862
12863 - def Exec(self, feedback_fn):
12864 if self.expandnames_calls < 1: 12865 raise errors.ProgrammerError("ExpandNames was not called") 12866 12867 if self.op.notify_exec: 12868 self._Notify(False, constants.JQT_EXEC, None) 12869 12870 self.LogInfo("Executing") 12871 12872 if self.op.log_messages: 12873 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages)) 12874 for idx, msg in enumerate(self.op.log_messages): 12875 self.LogInfo("Sending log message %s", idx + 1) 12876 feedback_fn(constants.JQT_MSGPREFIX + msg) 12877 # Report how many test messages have been sent 12878 self._Notify(False, constants.JQT_LOGMSG, idx + 1) 12879 12880 if self.op.fail: 12881 raise errors.OpExecError("Opcode failure was requested") 12882 12883 return True
12884
12885 12886 -class IAllocator(object):
12887 """IAllocator framework. 12888 12889 An IAllocator instance has three sets of attributes: 12890 - cfg that is needed to query the cluster 12891 - input data (all members of the _KEYS class attribute are required) 12892 - four buffer attributes (in|out_data|text), that represent the 12893 input (to the external script) in text and data structure format, 12894 and the output from it, again in two formats 12895 - the result variables from the script (success, info, nodes) for 12896 easy usage 12897 12898 """ 12899 # pylint: disable=R0902 12900 # lots of instance attributes 12901
12902 - def __init__(self, cfg, rpc, mode, **kwargs):
12903 self.cfg = cfg 12904 self.rpc = rpc 12905 # init buffer variables 12906 self.in_text = self.out_text = self.in_data = self.out_data = None 12907 # init all input fields so that pylint is happy 12908 self.mode = mode 12909 self.memory = self.disks = self.disk_template = None 12910 self.os = self.tags = self.nics = self.vcpus = None 12911 self.hypervisor = None 12912 self.relocate_from = None 12913 self.name = None 12914 self.instances = None 12915 self.evac_mode = None 12916 self.target_groups = [] 12917 # computed fields 12918 self.required_nodes = None 12919 # init result fields 12920 self.success = self.info = self.result = None 12921 12922 try: 12923 (fn, keydata, self._result_check) = self._MODE_DATA[self.mode] 12924 except KeyError: 12925 raise errors.ProgrammerError("Unknown mode '%s' passed to the" 12926 " IAllocator" % self.mode) 12927 12928 keyset = [n for (n, _) in keydata] 12929 12930 for key in kwargs: 12931 if key not in keyset: 12932 raise errors.ProgrammerError("Invalid input parameter '%s' to" 12933 " IAllocator" % key) 12934 setattr(self, key, kwargs[key]) 12935 12936 for key in keyset: 12937 if key not in kwargs: 12938 raise errors.ProgrammerError("Missing input parameter '%s' to" 12939 " IAllocator" % key) 12940 self._BuildInputData(compat.partial(fn, self), keydata)
12941
12942 - def _ComputeClusterData(self):
12943 """Compute the generic allocator input data. 12944 12945 This is the data that is independent of the actual operation. 12946 12947 """ 12948 cfg = self.cfg 12949 cluster_info = cfg.GetClusterInfo() 12950 # cluster data 12951 data = { 12952 "version": constants.IALLOCATOR_VERSION, 12953 "cluster_name": cfg.GetClusterName(), 12954 "cluster_tags": list(cluster_info.GetTags()), 12955 "enabled_hypervisors": list(cluster_info.enabled_hypervisors), 12956 # we don't have job IDs 12957 } 12958 ninfo = cfg.GetAllNodesInfo() 12959 iinfo = cfg.GetAllInstancesInfo().values() 12960 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo] 12961 12962 # node data 12963 node_list = [n.name for n in ninfo.values() if n.vm_capable] 12964 12965 if self.mode == constants.IALLOCATOR_MODE_ALLOC: 12966 hypervisor_name = self.hypervisor 12967 elif self.mode == constants.IALLOCATOR_MODE_RELOC: 12968 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor 12969 else: 12970 hypervisor_name = cluster_info.enabled_hypervisors[0] 12971 12972 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(), 12973 hypervisor_name) 12974 node_iinfo = \ 12975 self.rpc.call_all_instances_info(node_list, 12976 cluster_info.enabled_hypervisors) 12977 12978 data["nodegroups"] = self._ComputeNodeGroupData(cfg) 12979 12980 config_ndata = self._ComputeBasicNodeData(ninfo) 12981 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo, 12982 i_list, config_ndata) 12983 assert len(data["nodes"]) == len(ninfo), \ 12984 "Incomplete node data computed" 12985 12986 data["instances"] = self._ComputeInstanceData(cluster_info, i_list) 12987 12988 self.in_data = data
12989 12990 @staticmethod
12991 - def _ComputeNodeGroupData(cfg):
12992 """Compute node groups data. 12993 12994 """ 12995 ng = dict((guuid, { 12996 "name": gdata.name, 12997 "alloc_policy": gdata.alloc_policy, 12998 }) 12999 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items()) 13000 13001 return ng
13002 13003 @staticmethod
13004 - def _ComputeBasicNodeData(node_cfg):
13005 """Compute global node data. 13006 13007 @rtype: dict 13008 @returns: a dict of name: (node dict, node config) 13009 13010 """ 13011 # fill in static (config-based) values 13012 node_results = dict((ninfo.name, { 13013 "tags": list(ninfo.GetTags()), 13014 "primary_ip": ninfo.primary_ip, 13015 "secondary_ip": ninfo.secondary_ip, 13016 "offline": ninfo.offline, 13017 "drained": ninfo.drained, 13018 "master_candidate": ninfo.master_candidate, 13019 "group": ninfo.group, 13020 "master_capable": ninfo.master_capable, 13021 "vm_capable": ninfo.vm_capable, 13022 }) 13023 for ninfo in node_cfg.values()) 13024 13025 return node_results
13026 13027 @staticmethod
13028 - def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list, 13029 node_results):
13030 """Compute global node data. 13031 13032 @param node_results: the basic node structures as filled from the config 13033 13034 """ 13035 # make a copy of the current dict 13036 node_results = dict(node_results) 13037 for nname, nresult in node_data.items(): 13038 assert nname in node_results, "Missing basic data for node %s" % nname 13039 ninfo = node_cfg[nname] 13040 13041 if not (ninfo.offline or ninfo.drained): 13042 nresult.Raise("Can't get data for node %s" % nname) 13043 node_iinfo[nname].Raise("Can't get node instance info from node %s" % 13044 nname) 13045 remote_info = nresult.payload 13046 13047 for attr in ["memory_total", "memory_free", "memory_dom0", 13048 "vg_size", "vg_free", "cpu_total"]: 13049 if attr not in remote_info: 13050 raise errors.OpExecError("Node '%s' didn't return attribute" 13051 " '%s'" % (nname, attr)) 13052 if not isinstance(remote_info[attr], int): 13053 raise errors.OpExecError("Node '%s' returned invalid value" 13054 " for '%s': %s" % 13055 (nname, attr, remote_info[attr])) 13056 # compute memory used by primary instances 13057 i_p_mem = i_p_up_mem = 0 13058 for iinfo, beinfo in i_list: 13059 if iinfo.primary_node == nname: 13060 i_p_mem += beinfo[constants.BE_MEMORY] 13061 if iinfo.name not in node_iinfo[nname].payload: 13062 i_used_mem = 0 13063 else: 13064 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]["memory"]) 13065 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem 13066 remote_info["memory_free"] -= max(0, i_mem_diff) 13067 13068 if iinfo.admin_up: 13069 i_p_up_mem += beinfo[constants.BE_MEMORY] 13070 13071 # compute memory used by instances 13072 pnr_dyn = { 13073 "total_memory": remote_info["memory_total"], 13074 "reserved_memory": remote_info["memory_dom0"], 13075 "free_memory": remote_info["memory_free"], 13076 "total_disk": remote_info["vg_size"], 13077 "free_disk": remote_info["vg_free"], 13078 "total_cpus": remote_info["cpu_total"], 13079 "i_pri_memory": i_p_mem, 13080 "i_pri_up_memory": i_p_up_mem, 13081 } 13082 pnr_dyn.update(node_results[nname]) 13083 node_results[nname] = pnr_dyn 13084 13085 return node_results
13086 13087 @staticmethod
13088 - def _ComputeInstanceData(cluster_info, i_list):
13089 """Compute global instance data. 13090 13091 """ 13092 instance_data = {} 13093 for iinfo, beinfo in i_list: 13094 nic_data = [] 13095 for nic in iinfo.nics: 13096 filled_params = cluster_info.SimpleFillNIC(nic.nicparams) 13097 nic_dict = { 13098 "mac": nic.mac, 13099 "ip": nic.ip, 13100 "mode": filled_params[constants.NIC_MODE], 13101 "link": filled_params[constants.NIC_LINK], 13102 } 13103 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED: 13104 nic_dict["bridge"] = filled_params[constants.NIC_LINK] 13105 nic_data.append(nic_dict) 13106 pir = { 13107 "tags": list(iinfo.GetTags()), 13108 "admin_up": iinfo.admin_up, 13109 "vcpus": beinfo[constants.BE_VCPUS], 13110 "memory": beinfo[constants.BE_MEMORY], 13111 "os": iinfo.os, 13112 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes), 13113 "nics": nic_data, 13114 "disks": [{constants.IDISK_SIZE: dsk.size, 13115 constants.IDISK_MODE: dsk.mode} 13116 for dsk in iinfo.disks], 13117 "disk_template": iinfo.disk_template, 13118 "hypervisor": iinfo.hypervisor, 13119 } 13120 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template, 13121 pir["disks"]) 13122 instance_data[iinfo.name] = pir 13123 13124 return instance_data
13125
13126 - def _AddNewInstance(self):
13127 """Add new instance data to allocator structure. 13128 13129 This in combination with _AllocatorGetClusterData will create the 13130 correct structure needed as input for the allocator. 13131 13132 The checks for the completeness of the opcode must have already been 13133 done. 13134 13135 """ 13136 disk_space = _ComputeDiskSize(self.disk_template, self.disks) 13137 13138 if self.disk_template in constants.DTS_INT_MIRROR: 13139 self.required_nodes = 2 13140 else: 13141 self.required_nodes = 1 13142 13143 request = { 13144 "name": self.name, 13145 "disk_template": self.disk_template, 13146 "tags": self.tags, 13147 "os": self.os, 13148 "vcpus": self.vcpus, 13149 "memory": self.memory, 13150 "disks": self.disks, 13151 "disk_space_total": disk_space, 13152 "nics": self.nics, 13153 "required_nodes": self.required_nodes, 13154 "hypervisor": self.hypervisor, 13155 } 13156 13157 return request
13158
13159 - def _AddRelocateInstance(self):
13160 """Add relocate instance data to allocator structure. 13161 13162 This in combination with _IAllocatorGetClusterData will create the 13163 correct structure needed as input for the allocator. 13164 13165 The checks for the completeness of the opcode must have already been 13166 done. 13167 13168 """ 13169 instance = self.cfg.GetInstanceInfo(self.name) 13170 if instance is None: 13171 raise errors.ProgrammerError("Unknown instance '%s' passed to" 13172 " IAllocator" % self.name) 13173 13174 if instance.disk_template not in constants.DTS_MIRRORED: 13175 raise errors.OpPrereqError("Can't relocate non-mirrored instances", 13176 errors.ECODE_INVAL) 13177 13178 if instance.disk_template in constants.DTS_INT_MIRROR and \ 13179 len(instance.secondary_nodes) != 1: 13180 raise errors.OpPrereqError("Instance has not exactly one secondary node", 13181 errors.ECODE_STATE) 13182 13183 self.required_nodes = 1 13184 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks] 13185 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes) 13186 13187 request = { 13188 "name": self.name, 13189 "disk_space_total": disk_space, 13190 "required_nodes": self.required_nodes, 13191 "relocate_from": self.relocate_from, 13192 } 13193 return request
13194
13195 - def _AddNodeEvacuate(self):
13196 """Get data for node-evacuate requests. 13197 13198 """ 13199 return { 13200 "instances": self.instances, 13201 "evac_mode": self.evac_mode, 13202 }
13203
13204 - def _AddChangeGroup(self):
13205 """Get data for node-evacuate requests. 13206 13207 """ 13208 return { 13209 "instances": self.instances, 13210 "target_groups": self.target_groups, 13211 }
13212
13213 - def _BuildInputData(self, fn, keydata):
13214 """Build input data structures. 13215 13216 """ 13217 self._ComputeClusterData() 13218 13219 request = fn() 13220 request["type"] = self.mode 13221 for keyname, keytype in keydata: 13222 if keyname not in request: 13223 raise errors.ProgrammerError("Request parameter %s is missing" % 13224 keyname) 13225 val = request[keyname] 13226 if not keytype(val): 13227 raise errors.ProgrammerError("Request parameter %s doesn't pass" 13228 " validation, value %s, expected" 13229 " type %s" % (keyname, val, keytype)) 13230 self.in_data["request"] = request 13231 13232 self.in_text = serializer.Dump(self.in_data)
13233 13234 _STRING_LIST = ht.TListOf(ht.TString) 13235 _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, { 13236 # pylint: disable=E1101 13237 # Class '...' has no 'OP_ID' member 13238 "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID, 13239 opcodes.OpInstanceMigrate.OP_ID, 13240 opcodes.OpInstanceReplaceDisks.OP_ID]) 13241 }))) 13242 13243 _NEVAC_MOVED = \ 13244 ht.TListOf(ht.TAnd(ht.TIsLength(3), 13245 ht.TItems([ht.TNonEmptyString, 13246 ht.TNonEmptyString, 13247 ht.TListOf(ht.TNonEmptyString), 13248 ]))) 13249 _NEVAC_FAILED = \ 13250 ht.TListOf(ht.TAnd(ht.TIsLength(2), 13251 ht.TItems([ht.TNonEmptyString, 13252 ht.TMaybeString, 13253 ]))) 13254 _NEVAC_RESULT = ht.TAnd(ht.TIsLength(3), 13255 ht.TItems([_NEVAC_MOVED, _NEVAC_FAILED, _JOB_LIST])) 13256 13257 _MODE_DATA = { 13258 constants.IALLOCATOR_MODE_ALLOC: 13259 (_AddNewInstance, 13260 [ 13261 ("name", ht.TString), 13262 ("memory", ht.TInt), 13263 ("disks", ht.TListOf(ht.TDict)), 13264 ("disk_template", ht.TString), 13265 ("os", ht.TString), 13266 ("tags", _STRING_LIST), 13267 ("nics", ht.TListOf(ht.TDict)), 13268 ("vcpus", ht.TInt), 13269 ("hypervisor", ht.TString), 13270 ], ht.TList), 13271 constants.IALLOCATOR_MODE_RELOC: 13272 (_AddRelocateInstance, 13273 [("name", ht.TString), ("relocate_from", _STRING_LIST)], 13274 ht.TList), 13275 constants.IALLOCATOR_MODE_NODE_EVAC: 13276 (_AddNodeEvacuate, [ 13277 ("instances", _STRING_LIST), 13278 ("evac_mode", ht.TElemOf(constants.IALLOCATOR_NEVAC_MODES)), 13279 ], _NEVAC_RESULT), 13280 constants.IALLOCATOR_MODE_CHG_GROUP: 13281 (_AddChangeGroup, [ 13282 ("instances", _STRING_LIST), 13283 ("target_groups", _STRING_LIST), 13284 ], _NEVAC_RESULT), 13285 } 13286
13287 - def Run(self, name, validate=True, call_fn=None):
13288 """Run an instance allocator and return the results. 13289 13290 """ 13291 if call_fn is None: 13292 call_fn = self.rpc.call_iallocator_runner 13293 13294 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text) 13295 result.Raise("Failure while running the iallocator script") 13296 13297 self.out_text = result.payload 13298 if validate: 13299 self._ValidateResult()
13300
13301 - def _ValidateResult(self):
13302 """Process the allocator results. 13303 13304 This will process and if successful save the result in 13305 self.out_data and the other parameters. 13306 13307 """ 13308 try: 13309 rdict = serializer.Load(self.out_text) 13310 except Exception, err: 13311 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err)) 13312 13313 if not isinstance(rdict, dict): 13314 raise errors.OpExecError("Can't parse iallocator results: not a dict") 13315 13316 # TODO: remove backwards compatiblity in later versions 13317 if "nodes" in rdict and "result" not in rdict: 13318 rdict["result"] = rdict["nodes"] 13319 del rdict["nodes"] 13320 13321 for key in "success", "info", "result": 13322 if key not in rdict: 13323 raise errors.OpExecError("Can't parse iallocator results:" 13324 " missing key '%s'" % key) 13325 setattr(self, key, rdict[key]) 13326 13327 if not self._result_check(self.result): 13328 raise errors.OpExecError("Iallocator returned invalid result," 13329 " expected %s, got %s" % 13330 (self._result_check, self.result), 13331 errors.ECODE_INVAL) 13332 13333 if self.mode == constants.IALLOCATOR_MODE_RELOC: 13334 assert self.relocate_from is not None 13335 assert self.required_nodes == 1 13336 13337 node2group = dict((name, ndata["group"]) 13338 for (name, ndata) in self.in_data["nodes"].items()) 13339 13340 fn = compat.partial(self._NodesToGroups, node2group, 13341 self.in_data["nodegroups"]) 13342 13343 instance = self.cfg.GetInstanceInfo(self.name) 13344 request_groups = fn(self.relocate_from + [instance.primary_node]) 13345 result_groups = fn(rdict["result"] + [instance.primary_node]) 13346 13347 if self.success and not set(result_groups).issubset(request_groups): 13348 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)" 13349 " differ from original groups (%s)" % 13350 (utils.CommaJoin(result_groups), 13351 utils.CommaJoin(request_groups))) 13352 13353 elif self.mode == constants.IALLOCATOR_MODE_NODE_EVAC: 13354 assert self.evac_mode in constants.IALLOCATOR_NEVAC_MODES 13355 13356 self.out_data = rdict
13357 13358 @staticmethod
13359 - def _NodesToGroups(node2group, groups, nodes):
13360 """Returns a list of unique group names for a list of nodes. 13361 13362 @type node2group: dict 13363 @param node2group: Map from node name to group UUID 13364 @type groups: dict 13365 @param groups: Group information 13366 @type nodes: list 13367 @param nodes: Node names 13368 13369 """ 13370 result = set() 13371 13372 for node in nodes: 13373 try: 13374 group_uuid = node2group[node] 13375 except KeyError: 13376 # Ignore unknown node 13377 pass 13378 else: 13379 try: 13380 group = groups[group_uuid] 13381 except KeyError: 13382 # Can't find group, let's use UUID 13383 group_name = group_uuid 13384 else: 13385 group_name = group["name"] 13386 13387 result.add(group_name) 13388 13389 return sorted(result)
13390
13391 13392 -class LUTestAllocator(NoHooksLU):
13393 """Run allocator tests. 13394 13395 This LU runs the allocator tests 13396 13397 """
13398 - def CheckPrereq(self):
13399 """Check prerequisites. 13400 13401 This checks the opcode parameters depending on the director and mode test. 13402 13403 """ 13404 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC: 13405 for attr in ["memory", "disks", "disk_template", 13406 "os", "tags", "nics", "vcpus"]: 13407 if not hasattr(self.op, attr): 13408 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" % 13409 attr, errors.ECODE_INVAL) 13410 iname = self.cfg.ExpandInstanceName(self.op.name) 13411 if iname is not None: 13412 raise errors.OpPrereqError("Instance '%s' already in the cluster" % 13413 iname, errors.ECODE_EXISTS) 13414 if not isinstance(self.op.nics, list): 13415 raise errors.OpPrereqError("Invalid parameter 'nics'", 13416 errors.ECODE_INVAL) 13417 if not isinstance(self.op.disks, list): 13418 raise errors.OpPrereqError("Invalid parameter 'disks'", 13419 errors.ECODE_INVAL) 13420 for row in self.op.disks: 13421 if (not isinstance(row, dict) or 13422 constants.IDISK_SIZE not in row or 13423 not isinstance(row[constants.IDISK_SIZE], int) or 13424 constants.IDISK_MODE not in row or 13425 row[constants.IDISK_MODE] not in constants.DISK_ACCESS_SET): 13426 raise errors.OpPrereqError("Invalid contents of the 'disks'" 13427 " parameter", errors.ECODE_INVAL) 13428 if self.op.hypervisor is None: 13429 self.op.hypervisor = self.cfg.GetHypervisorType() 13430 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC: 13431 fname = _ExpandInstanceName(self.cfg, self.op.name) 13432 self.op.name = fname 13433 self.relocate_from = \ 13434 list(self.cfg.GetInstanceInfo(fname).secondary_nodes) 13435 elif self.op.mode in (constants.IALLOCATOR_MODE_CHG_GROUP, 13436 constants.IALLOCATOR_MODE_NODE_EVAC): 13437 if not self.op.instances: 13438 raise errors.OpPrereqError("Missing instances", errors.ECODE_INVAL) 13439 self.op.instances = _GetWantedInstances(self, self.op.instances) 13440 else: 13441 raise errors.OpPrereqError("Invalid test allocator mode '%s'" % 13442 self.op.mode, errors.ECODE_INVAL) 13443 13444 if self.op.direction == constants.IALLOCATOR_DIR_OUT: 13445 if self.op.allocator is None: 13446 raise errors.OpPrereqError("Missing allocator name", 13447 errors.ECODE_INVAL) 13448 elif self.op.direction != constants.IALLOCATOR_DIR_IN: 13449 raise errors.OpPrereqError("Wrong allocator test '%s'" % 13450 self.op.direction, errors.ECODE_INVAL)
13451
13452 - def Exec(self, feedback_fn):
13453 """Run the allocator test. 13454 13455 """ 13456 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC: 13457 ial = IAllocator(self.cfg, self.rpc, 13458 mode=self.op.mode, 13459 name=self.op.name, 13460 memory=self.op.memory, 13461 disks=self.op.disks, 13462 disk_template=self.op.disk_template, 13463 os=self.op.os, 13464 tags=self.op.tags, 13465 nics=self.op.nics, 13466 vcpus=self.op.vcpus, 13467 hypervisor=self.op.hypervisor, 13468 ) 13469 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC: 13470 ial = IAllocator(self.cfg, self.rpc, 13471 mode=self.op.mode, 13472 name=self.op.name, 13473 relocate_from=list(self.relocate_from), 13474 ) 13475 elif self.op.mode == constants.IALLOCATOR_MODE_CHG_GROUP: 13476 ial = IAllocator(self.cfg, self.rpc, 13477 mode=self.op.mode, 13478 instances=self.op.instances, 13479 target_groups=self.op.target_groups) 13480 elif self.op.mode == constants.IALLOCATOR_MODE_NODE_EVAC: 13481 ial = IAllocator(self.cfg, self.rpc, 13482 mode=self.op.mode, 13483 instances=self.op.instances, 13484 evac_mode=self.op.evac_mode) 13485 else: 13486 raise errors.ProgrammerError("Uncatched mode %s in" 13487 " LUTestAllocator.Exec", self.op.mode) 13488 13489 if self.op.direction == constants.IALLOCATOR_DIR_IN: 13490 result = ial.in_text 13491 else: 13492 ial.Run(self.op.allocator, validate=False) 13493 result = ial.out_text 13494 return result
13495 13496 13497 #: Query type implementations 13498 _QUERY_IMPL = { 13499 constants.QR_INSTANCE: _InstanceQuery, 13500 constants.QR_NODE: _NodeQuery, 13501 constants.QR_GROUP: _GroupQuery, 13502 constants.QR_OS: _OsQuery, 13503 } 13504 13505 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
13506 13507 13508 -def _GetQueryImplementation(name):
13509 """Returns the implemtnation for a query type. 13510 13511 @param name: Query type, must be one of L{constants.QR_VIA_OP} 13512 13513 """ 13514 try: 13515 return _QUERY_IMPL[name] 13516 except KeyError: 13517 raise errors.OpPrereqError("Unknown query resource '%s'" % name, 13518 errors.ECODE_INVAL)
13519