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

Source Code for Module ganeti.backend

   1  # 
   2  # 
   3   
   4  # Copyright (C) 2006, 2007, 2008, 2009, 2010 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  """Functions used by the node daemon 
  23   
  24  @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in 
  25       the L{UploadFile} function 
  26  @var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted 
  27       in the L{_CleanDirectory} function 
  28   
  29  """ 
  30   
  31  # pylint: disable-msg=E1103 
  32   
  33  # E1103: %s %r has no %r member (but some types could not be 
  34  # inferred), because the _TryOSFromDisk returns either (True, os_obj) 
  35  # or (False, "string") which confuses pylint 
  36   
  37   
  38  import os 
  39  import os.path 
  40  import shutil 
  41  import time 
  42  import stat 
  43  import errno 
  44  import re 
  45  import random 
  46  import logging 
  47  import tempfile 
  48  import zlib 
  49  import base64 
  50  import signal 
  51   
  52  from ganeti import errors 
  53  from ganeti import utils 
  54  from ganeti import ssh 
  55  from ganeti import hypervisor 
  56  from ganeti import constants 
  57  from ganeti import bdev 
  58  from ganeti import objects 
  59  from ganeti import ssconf 
  60  from ganeti import serializer 
  61  from ganeti import netutils 
  62   
  63   
  64  _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id" 
  65  _ALLOWED_CLEAN_DIRS = frozenset([ 
  66    constants.DATA_DIR, 
  67    constants.JOB_QUEUE_ARCHIVE_DIR, 
  68    constants.QUEUE_DIR, 
  69    constants.CRYPTO_KEYS_DIR, 
  70    ]) 
  71  _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60 
  72  _X509_KEY_FILE = "key" 
  73  _X509_CERT_FILE = "cert" 
  74  _IES_STATUS_FILE = "status" 
  75  _IES_PID_FILE = "pid" 
  76  _IES_CA_FILE = "ca" 
77 78 79 -class RPCFail(Exception):
80 """Class denoting RPC failure. 81 82 Its argument is the error message. 83 84 """
85
86 87 -def _Fail(msg, *args, **kwargs):
88 """Log an error and the raise an RPCFail exception. 89 90 This exception is then handled specially in the ganeti daemon and 91 turned into a 'failed' return type. As such, this function is a 92 useful shortcut for logging the error and returning it to the master 93 daemon. 94 95 @type msg: string 96 @param msg: the text of the exception 97 @raise RPCFail 98 99 """ 100 if args: 101 msg = msg % args 102 if "log" not in kwargs or kwargs["log"]: # if we should log this error 103 if "exc" in kwargs and kwargs["exc"]: 104 logging.exception(msg) 105 else: 106 logging.error(msg) 107 raise RPCFail(msg)
108
109 110 -def _GetConfig():
111 """Simple wrapper to return a SimpleStore. 112 113 @rtype: L{ssconf.SimpleStore} 114 @return: a SimpleStore instance 115 116 """ 117 return ssconf.SimpleStore()
118
119 120 -def _GetSshRunner(cluster_name):
121 """Simple wrapper to return an SshRunner. 122 123 @type cluster_name: str 124 @param cluster_name: the cluster name, which is needed 125 by the SshRunner constructor 126 @rtype: L{ssh.SshRunner} 127 @return: an SshRunner instance 128 129 """ 130 return ssh.SshRunner(cluster_name)
131
132 133 -def _Decompress(data):
134 """Unpacks data compressed by the RPC client. 135 136 @type data: list or tuple 137 @param data: Data sent by RPC client 138 @rtype: str 139 @return: Decompressed data 140 141 """ 142 assert isinstance(data, (list, tuple)) 143 assert len(data) == 2 144 (encoding, content) = data 145 if encoding == constants.RPC_ENCODING_NONE: 146 return content 147 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64: 148 return zlib.decompress(base64.b64decode(content)) 149 else: 150 raise AssertionError("Unknown data encoding")
151
152 153 -def _CleanDirectory(path, exclude=None):
154 """Removes all regular files in a directory. 155 156 @type path: str 157 @param path: the directory to clean 158 @type exclude: list 159 @param exclude: list of files to be excluded, defaults 160 to the empty list 161 162 """ 163 if path not in _ALLOWED_CLEAN_DIRS: 164 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'", 165 path) 166 167 if not os.path.isdir(path): 168 return 169 if exclude is None: 170 exclude = [] 171 else: 172 # Normalize excluded paths 173 exclude = [os.path.normpath(i) for i in exclude] 174 175 for rel_name in utils.ListVisibleFiles(path): 176 full_name = utils.PathJoin(path, rel_name) 177 if full_name in exclude: 178 continue 179 if os.path.isfile(full_name) and not os.path.islink(full_name): 180 utils.RemoveFile(full_name)
181
182 183 -def _BuildUploadFileList():
184 """Build the list of allowed upload files. 185 186 This is abstracted so that it's built only once at module import time. 187 188 """ 189 allowed_files = set([ 190 constants.CLUSTER_CONF_FILE, 191 constants.ETC_HOSTS, 192 constants.SSH_KNOWN_HOSTS_FILE, 193 constants.VNC_PASSWORD_FILE, 194 constants.RAPI_CERT_FILE, 195 constants.RAPI_USERS_FILE, 196 constants.CONFD_HMAC_KEY, 197 constants.CLUSTER_DOMAIN_SECRET_FILE, 198 ]) 199 200 for hv_name in constants.HYPER_TYPES: 201 hv_class = hypervisor.GetHypervisorClass(hv_name) 202 allowed_files.update(hv_class.GetAncillaryFiles()) 203 204 return frozenset(allowed_files)
205 206 207 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
208 209 210 -def JobQueuePurge():
211 """Removes job queue files and archived jobs. 212 213 @rtype: tuple 214 @return: True, None 215 216 """ 217 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE]) 218 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
219
220 221 -def GetMasterInfo():
222 """Returns master information. 223 224 This is an utility function to compute master information, either 225 for consumption here or from the node daemon. 226 227 @rtype: tuple 228 @return: master_netdev, master_ip, master_name 229 @raise RPCFail: in case of errors 230 231 """ 232 try: 233 cfg = _GetConfig() 234 master_netdev = cfg.GetMasterNetdev() 235 master_ip = cfg.GetMasterIP() 236 master_node = cfg.GetMasterNode() 237 except errors.ConfigurationError, err: 238 _Fail("Cluster configuration incomplete: %s", err, exc=True) 239 return (master_netdev, master_ip, master_node)
240
241 242 -def StartMaster(start_daemons, no_voting):
243 """Activate local node as master node. 244 245 The function will either try activate the IP address of the master 246 (unless someone else has it) or also start the master daemons, based 247 on the start_daemons parameter. 248 249 @type start_daemons: boolean 250 @param start_daemons: whether to start the master daemons 251 (ganeti-masterd and ganeti-rapi), or (if false) activate the 252 master ip 253 @type no_voting: boolean 254 @param no_voting: whether to start ganeti-masterd without a node vote 255 (if start_daemons is True), but still non-interactively 256 @rtype: None 257 258 """ 259 # GetMasterInfo will raise an exception if not able to return data 260 master_netdev, master_ip, _ = GetMasterInfo() 261 262 err_msgs = [] 263 # either start the master and rapi daemons 264 if start_daemons: 265 if no_voting: 266 masterd_args = "--no-voting --yes-do-it" 267 else: 268 masterd_args = "" 269 270 env = { 271 "EXTRA_MASTERD_ARGS": masterd_args, 272 } 273 274 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env) 275 if result.failed: 276 msg = "Can't start Ganeti master: %s" % result.output 277 logging.error(msg) 278 err_msgs.append(msg) 279 # or activate the IP 280 else: 281 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT): 282 if netutils.OwnIpAddress(master_ip): 283 # we already have the ip: 284 logging.debug("Master IP already configured, doing nothing") 285 else: 286 msg = "Someone else has the master ip, not activating" 287 logging.error(msg) 288 err_msgs.append(msg) 289 else: 290 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip, 291 "dev", master_netdev, "label", 292 "%s:0" % master_netdev]) 293 if result.failed: 294 msg = "Can't activate master IP: %s" % result.output 295 logging.error(msg) 296 err_msgs.append(msg) 297 298 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev, 299 "-s", master_ip, master_ip]) 300 # we'll ignore the exit code of arping 301 302 if err_msgs: 303 _Fail("; ".join(err_msgs))
304
305 306 -def StopMaster(stop_daemons):
307 """Deactivate this node as master. 308 309 The function will always try to deactivate the IP address of the 310 master. It will also stop the master daemons depending on the 311 stop_daemons parameter. 312 313 @type stop_daemons: boolean 314 @param stop_daemons: whether to also stop the master daemons 315 (ganeti-masterd and ganeti-rapi) 316 @rtype: None 317 318 """ 319 # TODO: log and report back to the caller the error failures; we 320 # need to decide in which case we fail the RPC for this 321 322 # GetMasterInfo will raise an exception if not able to return data 323 master_netdev, master_ip, _ = GetMasterInfo() 324 325 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip, 326 "dev", master_netdev]) 327 if result.failed: 328 logging.error("Can't remove the master IP, error: %s", result.output) 329 # but otherwise ignore the failure 330 331 if stop_daemons: 332 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"]) 333 if result.failed: 334 logging.error("Could not stop Ganeti master, command %s had exitcode %s" 335 " and error %s", 336 result.cmd, result.exit_code, result.output)
337
338 339 -def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
340 """Joins this node to the cluster. 341 342 This does the following: 343 - updates the hostkeys of the machine (rsa and dsa) 344 - adds the ssh private key to the user 345 - adds the ssh public key to the users' authorized_keys file 346 347 @type dsa: str 348 @param dsa: the DSA private key to write 349 @type dsapub: str 350 @param dsapub: the DSA public key to write 351 @type rsa: str 352 @param rsa: the RSA private key to write 353 @type rsapub: str 354 @param rsapub: the RSA public key to write 355 @type sshkey: str 356 @param sshkey: the SSH private key to write 357 @type sshpub: str 358 @param sshpub: the SSH public key to write 359 @rtype: boolean 360 @return: the success of the operation 361 362 """ 363 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600), 364 (constants.SSH_HOST_RSA_PUB, rsapub, 0644), 365 (constants.SSH_HOST_DSA_PRIV, dsa, 0600), 366 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)] 367 for name, content, mode in sshd_keys: 368 utils.WriteFile(name, data=content, mode=mode) 369 370 try: 371 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS, 372 mkdir=True) 373 except errors.OpExecError, err: 374 _Fail("Error while processing user ssh files: %s", err, exc=True) 375 376 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]: 377 utils.WriteFile(name, data=content, mode=0600) 378 379 utils.AddAuthorizedKey(auth_keys, sshpub) 380 381 result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"]) 382 if result.failed: 383 _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)", 384 result.cmd, result.exit_code, result.output)
385
386 387 -def LeaveCluster(modify_ssh_setup):
388 """Cleans up and remove the current node. 389 390 This function cleans up and prepares the current node to be removed 391 from the cluster. 392 393 If processing is successful, then it raises an 394 L{errors.QuitGanetiException} which is used as a special case to 395 shutdown the node daemon. 396 397 @param modify_ssh_setup: boolean 398 399 """ 400 _CleanDirectory(constants.DATA_DIR) 401 _CleanDirectory(constants.CRYPTO_KEYS_DIR) 402 JobQueuePurge() 403 404 if modify_ssh_setup: 405 try: 406 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) 407 408 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) 409 410 utils.RemoveFile(priv_key) 411 utils.RemoveFile(pub_key) 412 except errors.OpExecError: 413 logging.exception("Error while processing ssh files") 414 415 try: 416 utils.RemoveFile(constants.CONFD_HMAC_KEY) 417 utils.RemoveFile(constants.RAPI_CERT_FILE) 418 utils.RemoveFile(constants.NODED_CERT_FILE) 419 except: # pylint: disable-msg=W0702 420 logging.exception("Error while removing cluster secrets") 421 422 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD]) 423 if result.failed: 424 logging.error("Command %s failed with exitcode %s and error %s", 425 result.cmd, result.exit_code, result.output) 426 427 # Raise a custom exception (handled in ganeti-noded) 428 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
429
430 431 -def GetNodeInfo(vgname, hypervisor_type):
432 """Gives back a hash with different information about the node. 433 434 @type vgname: C{string} 435 @param vgname: the name of the volume group to ask for disk space information 436 @type hypervisor_type: C{str} 437 @param hypervisor_type: the name of the hypervisor to ask for 438 memory information 439 @rtype: C{dict} 440 @return: dictionary with the following keys: 441 - vg_size is the size of the configured volume group in MiB 442 - vg_free is the free size of the volume group in MiB 443 - memory_dom0 is the memory allocated for domain0 in MiB 444 - memory_free is the currently available (free) ram in MiB 445 - memory_total is the total number of ram in MiB 446 447 """ 448 outputarray = {} 449 vginfo = _GetVGInfo(vgname) 450 outputarray['vg_size'] = vginfo['vg_size'] 451 outputarray['vg_free'] = vginfo['vg_free'] 452 453 hyper = hypervisor.GetHypervisor(hypervisor_type) 454 hyp_info = hyper.GetNodeInfo() 455 if hyp_info is not None: 456 outputarray.update(hyp_info) 457 458 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n") 459 460 return outputarray
461
462 463 -def VerifyNode(what, cluster_name):
464 """Verify the status of the local node. 465 466 Based on the input L{what} parameter, various checks are done on the 467 local node. 468 469 If the I{filelist} key is present, this list of 470 files is checksummed and the file/checksum pairs are returned. 471 472 If the I{nodelist} key is present, we check that we have 473 connectivity via ssh with the target nodes (and check the hostname 474 report). 475 476 If the I{node-net-test} key is present, we check that we have 477 connectivity to the given nodes via both primary IP and, if 478 applicable, secondary IPs. 479 480 @type what: C{dict} 481 @param what: a dictionary of things to check: 482 - filelist: list of files for which to compute checksums 483 - nodelist: list of nodes we should check ssh communication with 484 - node-net-test: list of nodes we should check node daemon port 485 connectivity with 486 - hypervisor: list with hypervisors to run the verify for 487 @rtype: dict 488 @return: a dictionary with the same keys as the input dict, and 489 values representing the result of the checks 490 491 """ 492 result = {} 493 my_name = netutils.HostInfo().name 494 port = netutils.GetDaemonPort(constants.NODED) 495 496 if constants.NV_HYPERVISOR in what: 497 result[constants.NV_HYPERVISOR] = tmp = {} 498 for hv_name in what[constants.NV_HYPERVISOR]: 499 try: 500 val = hypervisor.GetHypervisor(hv_name).Verify() 501 except errors.HypervisorError, err: 502 val = "Error while checking hypervisor: %s" % str(err) 503 tmp[hv_name] = val 504 505 if constants.NV_FILELIST in what: 506 result[constants.NV_FILELIST] = utils.FingerprintFiles( 507 what[constants.NV_FILELIST]) 508 509 if constants.NV_NODELIST in what: 510 result[constants.NV_NODELIST] = tmp = {} 511 random.shuffle(what[constants.NV_NODELIST]) 512 for node in what[constants.NV_NODELIST]: 513 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node) 514 if not success: 515 tmp[node] = message 516 517 if constants.NV_NODENETTEST in what: 518 result[constants.NV_NODENETTEST] = tmp = {} 519 my_pip = my_sip = None 520 for name, pip, sip in what[constants.NV_NODENETTEST]: 521 if name == my_name: 522 my_pip = pip 523 my_sip = sip 524 break 525 if not my_pip: 526 tmp[my_name] = ("Can't find my own primary/secondary IP" 527 " in the node list") 528 else: 529 for name, pip, sip in what[constants.NV_NODENETTEST]: 530 fail = [] 531 if not netutils.TcpPing(pip, port, source=my_pip): 532 fail.append("primary") 533 if sip != pip: 534 if not netutils.TcpPing(sip, port, source=my_sip): 535 fail.append("secondary") 536 if fail: 537 tmp[name] = ("failure using the %s interface(s)" % 538 " and ".join(fail)) 539 540 if constants.NV_MASTERIP in what: 541 # FIXME: add checks on incoming data structures (here and in the 542 # rest of the function) 543 master_name, master_ip = what[constants.NV_MASTERIP] 544 if master_name == my_name: 545 source = constants.IP4_ADDRESS_LOCALHOST 546 else: 547 source = None 548 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port, 549 source=source) 550 551 if constants.NV_LVLIST in what: 552 try: 553 val = GetVolumeList(what[constants.NV_LVLIST]) 554 except RPCFail, err: 555 val = str(err) 556 result[constants.NV_LVLIST] = val 557 558 if constants.NV_INSTANCELIST in what: 559 # GetInstanceList can fail 560 try: 561 val = GetInstanceList(what[constants.NV_INSTANCELIST]) 562 except RPCFail, err: 563 val = str(err) 564 result[constants.NV_INSTANCELIST] = val 565 566 if constants.NV_VGLIST in what: 567 result[constants.NV_VGLIST] = utils.ListVolumeGroups() 568 569 if constants.NV_PVLIST in what: 570 result[constants.NV_PVLIST] = \ 571 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST], 572 filter_allocatable=False) 573 574 if constants.NV_VERSION in what: 575 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION, 576 constants.RELEASE_VERSION) 577 578 if constants.NV_HVINFO in what: 579 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO]) 580 result[constants.NV_HVINFO] = hyper.GetNodeInfo() 581 582 if constants.NV_DRBDLIST in what: 583 try: 584 used_minors = bdev.DRBD8.GetUsedDevs().keys() 585 except errors.BlockDeviceError, err: 586 logging.warning("Can't get used minors list", exc_info=True) 587 used_minors = str(err) 588 result[constants.NV_DRBDLIST] = used_minors 589 590 if constants.NV_DRBDHELPER in what: 591 status = True 592 try: 593 payload = bdev.BaseDRBD.GetUsermodeHelper() 594 except errors.BlockDeviceError, err: 595 logging.error("Can't get DRBD usermode helper: %s", str(err)) 596 status = False 597 payload = str(err) 598 result[constants.NV_DRBDHELPER] = (status, payload) 599 600 if constants.NV_NODESETUP in what: 601 result[constants.NV_NODESETUP] = tmpr = [] 602 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"): 603 tmpr.append("The sysfs filesytem doesn't seem to be mounted" 604 " under /sys, missing required directories /sys/block" 605 " and /sys/class/net") 606 if (not os.path.isdir("/proc/sys") or 607 not os.path.isfile("/proc/sysrq-trigger")): 608 tmpr.append("The procfs filesystem doesn't seem to be mounted" 609 " under /proc, missing required directory /proc/sys and" 610 " the file /proc/sysrq-trigger") 611 612 if constants.NV_TIME in what: 613 result[constants.NV_TIME] = utils.SplitTime(time.time()) 614 615 if constants.NV_OSLIST in what: 616 result[constants.NV_OSLIST] = DiagnoseOS() 617 618 return result
619
620 621 -def GetVolumeList(vg_name):
622 """Compute list of logical volumes and their size. 623 624 @type vg_name: str 625 @param vg_name: the volume group whose LVs we should list 626 @rtype: dict 627 @return: 628 dictionary of all partions (key) with value being a tuple of 629 their size (in MiB), inactive and online status:: 630 631 {'test1': ('20.06', True, True)} 632 633 in case of errors, a string is returned with the error 634 details. 635 636 """ 637 lvs = {} 638 sep = '|' 639 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix", 640 "--separator=%s" % sep, 641 "-olv_name,lv_size,lv_attr", vg_name]) 642 if result.failed: 643 _Fail("Failed to list logical volumes, lvs output: %s", result.output) 644 645 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$") 646 for line in result.stdout.splitlines(): 647 line = line.strip() 648 match = valid_line_re.match(line) 649 if not match: 650 logging.error("Invalid line returned from lvs output: '%s'", line) 651 continue 652 name, size, attr = match.groups() 653 inactive = attr[4] == '-' 654 online = attr[5] == 'o' 655 virtual = attr[0] == 'v' 656 if virtual: 657 # we don't want to report such volumes as existing, since they 658 # don't really hold data 659 continue 660 lvs[name] = (size, inactive, online) 661 662 return lvs
663
664 665 -def ListVolumeGroups():
666 """List the volume groups and their size. 667 668 @rtype: dict 669 @return: dictionary with keys volume name and values the 670 size of the volume 671 672 """ 673 return utils.ListVolumeGroups()
674
675 676 -def NodeVolumes():
677 """List all volumes on this node. 678 679 @rtype: list 680 @return: 681 A list of dictionaries, each having four keys: 682 - name: the logical volume name, 683 - size: the size of the logical volume 684 - dev: the physical device on which the LV lives 685 - vg: the volume group to which it belongs 686 687 In case of errors, we return an empty list and log the 688 error. 689 690 Note that since a logical volume can live on multiple physical 691 volumes, the resulting list might include a logical volume 692 multiple times. 693 694 """ 695 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix", 696 "--separator=|", 697 "--options=lv_name,lv_size,devices,vg_name"]) 698 if result.failed: 699 _Fail("Failed to list logical volumes, lvs output: %s", 700 result.output) 701 702 def parse_dev(dev): 703 return dev.split('(')[0]
704 705 def handle_dev(dev): 706 return [parse_dev(x) for x in dev.split(",")] 707 708 def map_line(line): 709 line = [v.strip() for v in line] 710 return [{'name': line[0], 'size': line[1], 711 'dev': dev, 'vg': line[3]} for dev in handle_dev(line[2])] 712 713 all_devs = [] 714 for line in result.stdout.splitlines(): 715 if line.count('|') >= 3: 716 all_devs.extend(map_line(line.split('|'))) 717 else: 718 logging.warning("Strange line in the output from lvs: '%s'", line) 719 return all_devs 720
721 722 -def BridgesExist(bridges_list):
723 """Check if a list of bridges exist on the current node. 724 725 @rtype: boolean 726 @return: C{True} if all of them exist, C{False} otherwise 727 728 """ 729 missing = [] 730 for bridge in bridges_list: 731 if not utils.BridgeExists(bridge): 732 missing.append(bridge) 733 734 if missing: 735 _Fail("Missing bridges %s", utils.CommaJoin(missing))
736
737 738 -def GetInstanceList(hypervisor_list):
739 """Provides a list of instances. 740 741 @type hypervisor_list: list 742 @param hypervisor_list: the list of hypervisors to query information 743 744 @rtype: list 745 @return: a list of all running instances on the current node 746 - instance1.example.com 747 - instance2.example.com 748 749 """ 750 results = [] 751 for hname in hypervisor_list: 752 try: 753 names = hypervisor.GetHypervisor(hname).ListInstances() 754 results.extend(names) 755 except errors.HypervisorError, err: 756 _Fail("Error enumerating instances (hypervisor %s): %s", 757 hname, err, exc=True) 758 759 return results
760
761 762 -def GetInstanceInfo(instance, hname):
763 """Gives back the information about an instance as a dictionary. 764 765 @type instance: string 766 @param instance: the instance name 767 @type hname: string 768 @param hname: the hypervisor type of the instance 769 770 @rtype: dict 771 @return: dictionary with the following keys: 772 - memory: memory size of instance (int) 773 - state: xen state of instance (string) 774 - time: cpu time of instance (float) 775 776 """ 777 output = {} 778 779 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance) 780 if iinfo is not None: 781 output['memory'] = iinfo[2] 782 output['state'] = iinfo[4] 783 output['time'] = iinfo[5] 784 785 return output
786
787 788 -def GetInstanceMigratable(instance):
789 """Gives whether an instance can be migrated. 790 791 @type instance: L{objects.Instance} 792 @param instance: object representing the instance to be checked. 793 794 @rtype: tuple 795 @return: tuple of (result, description) where: 796 - result: whether the instance can be migrated or not 797 - description: a description of the issue, if relevant 798 799 """ 800 hyper = hypervisor.GetHypervisor(instance.hypervisor) 801 iname = instance.name 802 if iname not in hyper.ListInstances(): 803 _Fail("Instance %s is not running", iname) 804 805 for idx in range(len(instance.disks)): 806 link_name = _GetBlockDevSymlinkPath(iname, idx) 807 if not os.path.islink(link_name): 808 logging.warning("Instance %s is missing symlink %s for disk %d", 809 iname, link_name, idx)
810
811 812 -def GetAllInstancesInfo(hypervisor_list):
813 """Gather data about all instances. 814 815 This is the equivalent of L{GetInstanceInfo}, except that it 816 computes data for all instances at once, thus being faster if one 817 needs data about more than one instance. 818 819 @type hypervisor_list: list 820 @param hypervisor_list: list of hypervisors to query for instance data 821 822 @rtype: dict 823 @return: dictionary of instance: data, with data having the following keys: 824 - memory: memory size of instance (int) 825 - state: xen state of instance (string) 826 - time: cpu time of instance (float) 827 - vcpus: the number of vcpus 828 829 """ 830 output = {} 831 832 for hname in hypervisor_list: 833 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo() 834 if iinfo: 835 for name, _, memory, vcpus, state, times in iinfo: 836 value = { 837 'memory': memory, 838 'vcpus': vcpus, 839 'state': state, 840 'time': times, 841 } 842 if name in output: 843 # we only check static parameters, like memory and vcpus, 844 # and not state and time which can change between the 845 # invocations of the different hypervisors 846 for key in 'memory', 'vcpus': 847 if value[key] != output[name][key]: 848 _Fail("Instance %s is running twice" 849 " with different parameters", name) 850 output[name] = value 851 852 return output
853
854 855 -def _InstanceLogName(kind, os_name, instance):
856 """Compute the OS log filename for a given instance and operation. 857 858 The instance name and os name are passed in as strings since not all 859 operations have these as part of an instance object. 860 861 @type kind: string 862 @param kind: the operation type (e.g. add, import, etc.) 863 @type os_name: string 864 @param os_name: the os name 865 @type instance: string 866 @param instance: the name of the instance being imported/added/etc. 867 868 """ 869 # TODO: Use tempfile.mkstemp to create unique filename 870 base = ("%s-%s-%s-%s.log" % 871 (kind, os_name, instance, utils.TimestampForFilename())) 872 return utils.PathJoin(constants.LOG_OS_DIR, base)
873
874 875 -def InstanceOsAdd(instance, reinstall, debug):
876 """Add an OS to an instance. 877 878 @type instance: L{objects.Instance} 879 @param instance: Instance whose OS is to be installed 880 @type reinstall: boolean 881 @param reinstall: whether this is an instance reinstall 882 @type debug: integer 883 @param debug: debug level, passed to the OS scripts 884 @rtype: None 885 886 """ 887 inst_os = OSFromDisk(instance.os) 888 889 create_env = OSEnvironment(instance, inst_os, debug) 890 if reinstall: 891 create_env['INSTANCE_REINSTALL'] = "1" 892 893 logfile = _InstanceLogName("add", instance.os, instance.name) 894 895 result = utils.RunCmd([inst_os.create_script], env=create_env, 896 cwd=inst_os.path, output=logfile,) 897 if result.failed: 898 logging.error("os create command '%s' returned error: %s, logfile: %s," 899 " output: %s", result.cmd, result.fail_reason, logfile, 900 result.output) 901 lines = [utils.SafeEncode(val) 902 for val in utils.TailFile(logfile, lines=20)] 903 _Fail("OS create script failed (%s), last lines in the" 904 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
905
906 907 -def RunRenameInstance(instance, old_name, debug):
908 """Run the OS rename script for an instance. 909 910 @type instance: L{objects.Instance} 911 @param instance: Instance whose OS is to be installed 912 @type old_name: string 913 @param old_name: previous instance name 914 @type debug: integer 915 @param debug: debug level, passed to the OS scripts 916 @rtype: boolean 917 @return: the success of the operation 918 919 """ 920 inst_os = OSFromDisk(instance.os) 921 922 rename_env = OSEnvironment(instance, inst_os, debug) 923 rename_env['OLD_INSTANCE_NAME'] = old_name 924 925 logfile = _InstanceLogName("rename", instance.os, 926 "%s-%s" % (old_name, instance.name)) 927 928 result = utils.RunCmd([inst_os.rename_script], env=rename_env, 929 cwd=inst_os.path, output=logfile) 930 931 if result.failed: 932 logging.error("os create command '%s' returned error: %s output: %s", 933 result.cmd, result.fail_reason, result.output) 934 lines = [utils.SafeEncode(val) 935 for val in utils.TailFile(logfile, lines=20)] 936 _Fail("OS rename script failed (%s), last lines in the" 937 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
938
939 940 -def _GetVGInfo(vg_name):
941 """Get information about the volume group. 942 943 @type vg_name: str 944 @param vg_name: the volume group which we query 945 @rtype: dict 946 @return: 947 A dictionary with the following keys: 948 - C{vg_size} is the total size of the volume group in MiB 949 - C{vg_free} is the free size of the volume group in MiB 950 - C{pv_count} are the number of physical disks in that VG 951 952 If an error occurs during gathering of data, we return the same dict 953 with keys all set to None. 954 955 """ 956 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"]) 957 958 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings", 959 "--nosuffix", "--units=m", "--separator=:", vg_name]) 960 961 if retval.failed: 962 logging.error("volume group %s not present", vg_name) 963 return retdic 964 valarr = retval.stdout.strip().rstrip(':').split(':') 965 if len(valarr) == 3: 966 try: 967 retdic = { 968 "vg_size": int(round(float(valarr[0]), 0)), 969 "vg_free": int(round(float(valarr[1]), 0)), 970 "pv_count": int(valarr[2]), 971 } 972 except (TypeError, ValueError), err: 973 logging.exception("Fail to parse vgs output: %s", err) 974 else: 975 logging.error("vgs output has the wrong number of fields (expected" 976 " three): %s", str(valarr)) 977 return retdic
978
979 980 -def _GetBlockDevSymlinkPath(instance_name, idx):
981 return utils.PathJoin(constants.DISK_LINKS_DIR, 982 "%s:%d" % (instance_name, idx))
983
984 985 -def _SymlinkBlockDev(instance_name, device_path, idx):
986 """Set up symlinks to a instance's block device. 987 988 This is an auxiliary function run when an instance is start (on the primary 989 node) or when an instance is migrated (on the target node). 990 991 992 @param instance_name: the name of the target instance 993 @param device_path: path of the physical block device, on the node 994 @param idx: the disk index 995 @return: absolute path to the disk's symlink 996 997 """ 998 link_name = _GetBlockDevSymlinkPath(instance_name, idx) 999 try: 1000 os.symlink(device_path, link_name) 1001 except OSError, err: 1002 if err.errno == errno.EEXIST: 1003 if (not os.path.islink(link_name) or 1004 os.readlink(link_name) != device_path): 1005 os.remove(link_name) 1006 os.symlink(device_path, link_name) 1007 else: 1008 raise 1009 1010 return link_name
1011 1024
1025 1026 -def _GatherAndLinkBlockDevs(instance):
1027 """Set up an instance's block device(s). 1028 1029 This is run on the primary node at instance startup. The block 1030 devices must be already assembled. 1031 1032 @type instance: L{objects.Instance} 1033 @param instance: the instance whose disks we shoul assemble 1034 @rtype: list 1035 @return: list of (disk_object, device_path) 1036 1037 """ 1038 block_devices = [] 1039 for idx, disk in enumerate(instance.disks): 1040 device = _RecursiveFindBD(disk) 1041 if device is None: 1042 raise errors.BlockDeviceError("Block device '%s' is not set up." % 1043 str(disk)) 1044 device.Open() 1045 try: 1046 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx) 1047 except OSError, e: 1048 raise errors.BlockDeviceError("Cannot create block device symlink: %s" % 1049 e.strerror) 1050 1051 block_devices.append((disk, link_name)) 1052 1053 return block_devices
1054
1055 1056 -def StartInstance(instance):
1057 """Start an instance. 1058 1059 @type instance: L{objects.Instance} 1060 @param instance: the instance object 1061 @rtype: None 1062 1063 """ 1064 running_instances = GetInstanceList([instance.hypervisor]) 1065 1066 if instance.name in running_instances: 1067 logging.info("Instance %s already running, not starting", instance.name) 1068 return 1069 1070 try: 1071 block_devices = _GatherAndLinkBlockDevs(instance) 1072 hyper = hypervisor.GetHypervisor(instance.hypervisor) 1073 hyper.StartInstance(instance, block_devices) 1074 except errors.BlockDeviceError, err: 1075 _Fail("Block device error: %s", err, exc=True) 1076 except errors.HypervisorError, err: 1077 _RemoveBlockDevLinks(instance.name, instance.disks) 1078 _Fail("Hypervisor error: %s", err, exc=True)
1079
1080 1081 -def InstanceShutdown(instance, timeout):
1082 """Shut an instance down. 1083 1084 @note: this functions uses polling with a hardcoded timeout. 1085 1086 @type instance: L{objects.Instance} 1087 @param instance: the instance object 1088 @type timeout: integer 1089 @param timeout: maximum timeout for soft shutdown 1090 @rtype: None 1091 1092 """ 1093 hv_name = instance.hypervisor 1094 hyper = hypervisor.GetHypervisor(hv_name) 1095 iname = instance.name 1096 1097 if instance.name not in hyper.ListInstances(): 1098 logging.info("Instance %s not running, doing nothing", iname) 1099 return 1100 1101 class _TryShutdown: 1102 def __init__(self): 1103 self.tried_once = False
1104 1105 def __call__(self): 1106 if iname not in hyper.ListInstances(): 1107 return 1108 1109 try: 1110 hyper.StopInstance(instance, retry=self.tried_once) 1111 except errors.HypervisorError, err: 1112 if iname not in hyper.ListInstances(): 1113 # if the instance is no longer existing, consider this a 1114 # success and go to cleanup 1115 return 1116 1117 _Fail("Failed to stop instance %s: %s", iname, err) 1118 1119 self.tried_once = True 1120 1121 raise utils.RetryAgain() 1122 1123 try: 1124 utils.Retry(_TryShutdown(), 5, timeout) 1125 except utils.RetryTimeout: 1126 # the shutdown did not succeed 1127 logging.error("Shutdown of '%s' unsuccessful, forcing", iname) 1128 1129 try: 1130 hyper.StopInstance(instance, force=True) 1131 except errors.HypervisorError, err: 1132 if iname in hyper.ListInstances(): 1133 # only raise an error if the instance still exists, otherwise 1134 # the error could simply be "instance ... unknown"! 1135 _Fail("Failed to force stop instance %s: %s", iname, err) 1136 1137 time.sleep(1) 1138 1139 if iname in hyper.ListInstances(): 1140 _Fail("Could not shutdown instance %s even by destroy", iname) 1141 1142 try: 1143 hyper.CleanupInstance(instance.name) 1144 except errors.HypervisorError, err: 1145 logging.warning("Failed to execute post-shutdown cleanup step: %s", err) 1146 1147 _RemoveBlockDevLinks(iname, instance.disks) 1148
1149 1150 -def InstanceReboot(instance, reboot_type, shutdown_timeout):
1151 """Reboot an instance. 1152 1153 @type instance: L{objects.Instance} 1154 @param instance: the instance object to reboot 1155 @type reboot_type: str 1156 @param reboot_type: the type of reboot, one the following 1157 constants: 1158 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the 1159 instance OS, do not recreate the VM 1160 - L{constants.INSTANCE_REBOOT_HARD}: tear down and 1161 restart the VM (at the hypervisor level) 1162 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is 1163 not accepted here, since that mode is handled differently, in 1164 cmdlib, and translates into full stop and start of the 1165 instance (instead of a call_instance_reboot RPC) 1166 @type shutdown_timeout: integer 1167 @param shutdown_timeout: maximum timeout for soft shutdown 1168 @rtype: None 1169 1170 """ 1171 running_instances = GetInstanceList([instance.hypervisor]) 1172 1173 if instance.name not in running_instances: 1174 _Fail("Cannot reboot instance %s that is not running", instance.name) 1175 1176 hyper = hypervisor.GetHypervisor(instance.hypervisor) 1177 if reboot_type == constants.INSTANCE_REBOOT_SOFT: 1178 try: 1179 hyper.RebootInstance(instance) 1180 except errors.HypervisorError, err: 1181 _Fail("Failed to soft reboot instance %s: %s", instance.name, err) 1182 elif reboot_type == constants.INSTANCE_REBOOT_HARD: 1183 try: 1184 InstanceShutdown(instance, shutdown_timeout) 1185 return StartInstance(instance) 1186 except errors.HypervisorError, err: 1187 _Fail("Failed to hard reboot instance %s: %s", instance.name, err) 1188 else: 1189 _Fail("Invalid reboot_type received: %s", reboot_type)
1190
1191 1192 -def MigrationInfo(instance):
1193 """Gather information about an instance to be migrated. 1194 1195 @type instance: L{objects.Instance} 1196 @param instance: the instance definition 1197 1198 """ 1199 hyper = hypervisor.GetHypervisor(instance.hypervisor) 1200 try: 1201 info = hyper.MigrationInfo(instance) 1202 except errors.HypervisorError, err: 1203 _Fail("Failed to fetch migration information: %s", err, exc=True) 1204 return info
1205
1206 1207 -def AcceptInstance(instance, info, target):
1208 """Prepare the node to accept an instance. 1209 1210 @type instance: L{objects.Instance} 1211 @param instance: the instance definition 1212 @type info: string/data (opaque) 1213 @param info: migration information, from the source node 1214 @type target: string 1215 @param target: target host (usually ip), on this node 1216 1217 """ 1218 hyper = hypervisor.GetHypervisor(instance.hypervisor) 1219 try: 1220 hyper.AcceptInstance(instance, info, target) 1221 except errors.HypervisorError, err: 1222 _Fail("Failed to accept instance: %s", err, exc=True)
1223
1224 1225 -def FinalizeMigration(instance, info, success):
1226 """Finalize any preparation to accept an instance. 1227 1228 @type instance: L{objects.Instance} 1229 @param instance: the instance definition 1230 @type info: string/data (opaque) 1231 @param info: migration information, from the source node 1232 @type success: boolean 1233 @param success: whether the migration was a success or a failure 1234 1235 """ 1236 hyper = hypervisor.GetHypervisor(instance.hypervisor) 1237 try: 1238 hyper.FinalizeMigration(instance, info, success) 1239 except errors.HypervisorError, err: 1240 _Fail("Failed to finalize migration: %s", err, exc=True)
1241
1242 1243 -def MigrateInstance(instance, target, live):
1244 """Migrates an instance to another node. 1245 1246 @type instance: L{objects.Instance} 1247 @param instance: the instance definition 1248 @type target: string 1249 @param target: the target node name 1250 @type live: boolean 1251 @param live: whether the migration should be done live or not (the 1252 interpretation of this parameter is left to the hypervisor) 1253 @rtype: tuple 1254 @return: a tuple of (success, msg) where: 1255 - succes is a boolean denoting the success/failure of the operation 1256 - msg is a string with details in case of failure 1257 1258 """ 1259 hyper = hypervisor.GetHypervisor(instance.hypervisor) 1260 1261 try: 1262 hyper.MigrateInstance(instance, target, live) 1263 except errors.HypervisorError, err: 1264 _Fail("Failed to migrate instance: %s", err, exc=True)
1265
1266 1267 -def BlockdevCreate(disk, size, owner, on_primary, info):
1268 """Creates a block device for an instance. 1269 1270 @type disk: L{objects.Disk} 1271 @param disk: the object describing the disk we should create 1272 @type size: int 1273 @param size: the size of the physical underlying device, in MiB 1274 @type owner: str 1275 @param owner: the name of the instance for which disk is created, 1276 used for device cache data 1277 @type on_primary: boolean 1278 @param on_primary: indicates if it is the primary node or not 1279 @type info: string 1280 @param info: string that will be sent to the physical device 1281 creation, used for example to set (LVM) tags on LVs 1282 1283 @return: the new unique_id of the device (this can sometime be 1284 computed only after creation), or None. On secondary nodes, 1285 it's not required to return anything. 1286 1287 """ 1288 # TODO: remove the obsolete 'size' argument 1289 # pylint: disable-msg=W0613 1290 clist = [] 1291 if disk.children: 1292 for child in disk.children: 1293 try: 1294 crdev = _RecursiveAssembleBD(child, owner, on_primary) 1295 except errors.BlockDeviceError, err: 1296 _Fail("Can't assemble device %s: %s", child, err) 1297 if on_primary or disk.AssembleOnSecondary(): 1298 # we need the children open in case the device itself has to 1299 # be assembled 1300 try: 1301 # pylint: disable-msg=E1103 1302 crdev.Open() 1303 except errors.BlockDeviceError, err: 1304 _Fail("Can't make child '%s' read-write: %s", child, err) 1305 clist.append(crdev) 1306 1307 try: 1308 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size) 1309 except errors.BlockDeviceError, err: 1310 _Fail("Can't create block device: %s", err) 1311 1312 if on_primary or disk.AssembleOnSecondary(): 1313 try: 1314 device.Assemble() 1315 except errors.BlockDeviceError, err: 1316 _Fail("Can't assemble device after creation, unusual event: %s", err) 1317 device.SetSyncSpeed(constants.SYNC_SPEED) 1318 if on_primary or disk.OpenOnSecondary(): 1319 try: 1320 device.Open(force=True) 1321 except errors.BlockDeviceError, err: 1322 _Fail("Can't make device r/w after creation, unusual event: %s", err) 1323 DevCacheManager.UpdateCache(device.dev_path, owner, 1324 on_primary, disk.iv_name) 1325 1326 device.SetInfo(info) 1327 1328 return device.unique_id
1329
1330 1331 -def BlockdevRemove(disk):
1332 """Remove a block device. 1333 1334 @note: This is intended to be called recursively. 1335 1336 @type disk: L{objects.Disk} 1337 @param disk: the disk object we should remove 1338 @rtype: boolean 1339 @return: the success of the operation 1340 1341 """ 1342 msgs = [] 1343 try: 1344 rdev = _RecursiveFindBD(disk) 1345 except errors.BlockDeviceError, err: 1346 # probably can't attach 1347 logging.info("Can't attach to device %s in remove", disk) 1348 rdev = None 1349 if rdev is not None: 1350 r_path = rdev.dev_path 1351 try: 1352 rdev.Remove() 1353 except errors.BlockDeviceError, err: 1354 msgs.append(str(err)) 1355 if not msgs: 1356 DevCacheManager.RemoveCache(r_path) 1357 1358 if disk.children: 1359 for child in disk.children: 1360 try: 1361 BlockdevRemove(child) 1362 except RPCFail, err: 1363 msgs.append(str(err)) 1364 1365 if msgs: 1366 _Fail("; ".join(msgs))
1367
1368 1369 -def _RecursiveAssembleBD(disk, owner, as_primary):
1370 """Activate a block device for an instance. 1371 1372 This is run on the primary and secondary nodes for an instance. 1373 1374 @note: this function is called recursively. 1375 1376 @type disk: L{objects.Disk} 1377 @param disk: the disk we try to assemble 1378 @type owner: str 1379 @param owner: the name of the instance which owns the disk 1380 @type as_primary: boolean 1381 @param as_primary: if we should make the block device 1382 read/write 1383 1384 @return: the assembled device or None (in case no device 1385 was assembled) 1386 @raise errors.BlockDeviceError: in case there is an error 1387 during the activation of the children or the device 1388 itself 1389 1390 """ 1391 children = [] 1392 if disk.children: 1393 mcn = disk.ChildrenNeeded() 1394 if mcn == -1: 1395 mcn = 0 # max number of Nones allowed 1396 else: 1397 mcn = len(disk.children) - mcn # max number of Nones 1398 for chld_disk in disk.children: 1399 try: 1400 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary) 1401 except errors.BlockDeviceError, err: 1402 if children.count(None) >= mcn: 1403 raise 1404 cdev = None 1405 logging.error("Error in child activation (but continuing): %s", 1406 str(err)) 1407 children.append(cdev) 1408 1409 if as_primary or disk.AssembleOnSecondary(): 1410 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size) 1411 r_dev.SetSyncSpeed(constants.SYNC_SPEED) 1412 result = r_dev 1413 if as_primary or disk.OpenOnSecondary(): 1414 r_dev.Open() 1415 DevCacheManager.UpdateCache(r_dev.dev_path, owner, 1416 as_primary, disk.iv_name) 1417 1418 else: 1419 result = True 1420 return result
1421
1422 1423 -def BlockdevAssemble(disk, owner, as_primary):
1424 """Activate a block device for an instance. 1425 1426 This is a wrapper over _RecursiveAssembleBD. 1427 1428 @rtype: str or boolean 1429 @return: a C{/dev/...} path for primary nodes, and 1430 C{True} for secondary nodes 1431 1432 """ 1433 try: 1434 result = _RecursiveAssembleBD(disk, owner, as_primary) 1435 if isinstance(result, bdev.BlockDev): 1436 # pylint: disable-msg=E1103 1437 result = result.dev_path 1438 except errors.BlockDeviceError, err: 1439 _Fail("Error while assembling disk: %s", err, exc=True) 1440 1441 return result
1442
1443 1444 -def BlockdevShutdown(disk):
1445 """Shut down a block device. 1446 1447 First, if the device is assembled (Attach() is successful), then 1448 the device is shutdown. Then the children of the device are 1449 shutdown. 1450 1451 This function is called recursively. Note that we don't cache the 1452 children or such, as oppossed to assemble, shutdown of different 1453 devices doesn't require that the upper device was active. 1454 1455 @type disk: L{objects.Disk} 1456 @param disk: the description of the disk we should 1457 shutdown 1458 @rtype: None 1459 1460 """ 1461 msgs = [] 1462 r_dev = _RecursiveFindBD(disk) 1463 if r_dev is not None: 1464 r_path = r_dev.dev_path 1465 try: 1466 r_dev.Shutdown() 1467 DevCacheManager.RemoveCache(r_path) 1468 except errors.BlockDeviceError, err: 1469 msgs.append(str(err)) 1470 1471 if disk.children: 1472 for child in disk.children: 1473 try: 1474 BlockdevShutdown(child) 1475 except RPCFail, err: 1476 msgs.append(str(err)) 1477 1478 if msgs: 1479 _Fail("; ".join(msgs))
1480
1481 1482 -def BlockdevAddchildren(parent_cdev, new_cdevs):
1483 """Extend a mirrored block device. 1484 1485 @type parent_cdev: L{objects.Disk} 1486 @param parent_cdev: the disk to which we should add children 1487 @type new_cdevs: list of L{objects.Disk} 1488 @param new_cdevs: the list of children which we should add 1489 @rtype: None 1490 1491 """ 1492 parent_bdev = _RecursiveFindBD(parent_cdev) 1493 if parent_bdev is None: 1494 _Fail("Can't find parent device '%s' in add children", parent_cdev) 1495 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs] 1496 if new_bdevs.count(None) > 0: 1497 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs) 1498 parent_bdev.AddChildren(new_bdevs)
1499
1500 1501 -def BlockdevRemovechildren(parent_cdev, new_cdevs):
1502 """Shrink a mirrored block device. 1503 1504 @type parent_cdev: L{objects.Disk} 1505 @param parent_cdev: the disk from which we should remove children 1506 @type new_cdevs: list of L{objects.Disk} 1507 @param new_cdevs: the list of children which we should remove 1508 @rtype: None 1509 1510 """ 1511 parent_bdev = _RecursiveFindBD(parent_cdev) 1512 if parent_bdev is None: 1513 _Fail("Can't find parent device '%s' in remove children", parent_cdev) 1514 devs = [] 1515 for disk in new_cdevs: 1516 rpath = disk.StaticDevPath() 1517 if rpath is None: 1518 bd = _RecursiveFindBD(disk) 1519 if bd is None: 1520 _Fail("Can't find device %s while removing children", disk) 1521 else: 1522 devs.append(bd.dev_path) 1523 else: 1524 if not utils.IsNormAbsPath(rpath): 1525 _Fail("Strange path returned from StaticDevPath: '%s'", rpath) 1526 devs.append(rpath) 1527 parent_bdev.RemoveChildren(devs)
1528
1529 1530 -def BlockdevGetmirrorstatus(disks):
1531 """Get the mirroring status of a list of devices. 1532 1533 @type disks: list of L{objects.Disk} 1534 @param disks: the list of disks which we should query 1535 @rtype: disk 1536 @return: 1537 a list of (mirror_done, estimated_time) tuples, which 1538 are the result of L{bdev.BlockDev.CombinedSyncStatus} 1539 @raise errors.BlockDeviceError: if any of the disks cannot be 1540 found 1541 1542 """ 1543 stats = [] 1544 for dsk in disks: 1545 rbd = _RecursiveFindBD(dsk) 1546 if rbd is None: 1547 _Fail("Can't find device %s", dsk) 1548 1549 stats.append(rbd.CombinedSyncStatus()) 1550 1551 return stats
1552
1553 1554 -def _RecursiveFindBD(disk):
1555 """Check if a device is activated. 1556 1557 If so, return information about the real device. 1558 1559 @type disk: L{objects.Disk} 1560 @param disk: the disk object we need to find 1561 1562 @return: None if the device can't be found, 1563 otherwise the device instance 1564 1565 """ 1566 children = [] 1567 if disk.children: 1568 for chdisk in disk.children: 1569 children.append(_RecursiveFindBD(chdisk)) 1570 1571 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1572
1573 1574 -def _OpenRealBD(disk):
1575 """Opens the underlying block device of a disk. 1576 1577 @type disk: L{objects.Disk} 1578 @param disk: the disk object we want to open 1579 1580 """ 1581 real_disk = _RecursiveFindBD(disk) 1582 if real_disk is None: 1583 _Fail("Block device '%s' is not set up", disk) 1584 1585 real_disk.Open() 1586 1587 return real_disk
1588
1589 1590 -def BlockdevFind(disk):
1591 """Check if a device is activated. 1592 1593 If it is, return information about the real device. 1594 1595 @type disk: L{objects.Disk} 1596 @param disk: the disk to find 1597 @rtype: None or objects.BlockDevStatus 1598 @return: None if the disk cannot be found, otherwise a the current 1599 information 1600 1601 """ 1602 try: 1603 rbd = _RecursiveFindBD(disk) 1604 except errors.BlockDeviceError, err: 1605 _Fail("Failed to find device: %s", err, exc=True) 1606 1607 if rbd is None: 1608 return None 1609 1610 return rbd.GetSyncStatus()
1611
1612 1613 -def BlockdevGetsize(disks):
1614 """Computes the size of the given disks. 1615 1616 If a disk is not found, returns None instead. 1617 1618 @type disks: list of L{objects.Disk} 1619 @param disks: the list of disk to compute the size for 1620 @rtype: list 1621 @return: list with elements None if the disk cannot be found, 1622 otherwise the size 1623 1624 """ 1625 result = [] 1626 for cf in disks: 1627 try: 1628 rbd = _RecursiveFindBD(cf) 1629 except errors.BlockDeviceError: 1630 result.append(None) 1631 continue 1632 if rbd is None: 1633 result.append(None) 1634 else: 1635 result.append(rbd.GetActualSize()) 1636 return result
1637
1638 1639 -def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1640 """Export a block device to a remote node. 1641 1642 @type disk: L{objects.Disk} 1643 @param disk: the description of the disk to export 1644 @type dest_node: str 1645 @param dest_node: the destination node to export to 1646 @type dest_path: str 1647 @param dest_path: the destination path on the target node 1648 @type cluster_name: str 1649 @param cluster_name: the cluster name, needed for SSH hostalias 1650 @rtype: None 1651 1652 """ 1653 real_disk = _OpenRealBD(disk) 1654 1655 # the block size on the read dd is 1MiB to match our units 1656 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; " 1657 "dd if=%s bs=1048576 count=%s", 1658 real_disk.dev_path, str(disk.size)) 1659 1660 # we set here a smaller block size as, due to ssh buffering, more 1661 # than 64-128k will mostly ignored; we use nocreat to fail if the 1662 # device is not already there or we pass a wrong path; we use 1663 # notrunc to no attempt truncate on an LV device; we use oflag=dsync 1664 # to not buffer too much memory; this means that at best, we flush 1665 # every 64k, which will not be very fast 1666 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536" 1667 " oflag=dsync", dest_path) 1668 1669 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node, 1670 constants.GANETI_RUNAS, 1671 destcmd) 1672 1673 # all commands have been checked, so we're safe to combine them 1674 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)]) 1675 1676 result = utils.RunCmd(["bash", "-c", command]) 1677 1678 if result.failed: 1679 _Fail("Disk copy command '%s' returned error: %s" 1680 " output: %s", command, result.fail_reason, result.output)
1681
1682 1683 -def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1684 """Write a file to the filesystem. 1685 1686 This allows the master to overwrite(!) a file. It will only perform 1687 the operation if the file belongs to a list of configuration files. 1688 1689 @type file_name: str 1690 @param file_name: the target file name 1691 @type data: str 1692 @param data: the new contents of the file 1693 @type mode: int 1694 @param mode: the mode to give the file (can be None) 1695 @type uid: int 1696 @param uid: the owner of the file (can be -1 for default) 1697 @type gid: int 1698 @param gid: the group of the file (can be -1 for default) 1699 @type atime: float 1700 @param atime: the atime to set on the file (can be None) 1701 @type mtime: float 1702 @param mtime: the mtime to set on the file (can be None) 1703 @rtype: None 1704 1705 """ 1706 if not os.path.isabs(file_name): 1707 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name) 1708 1709 if file_name not in _ALLOWED_UPLOAD_FILES: 1710 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'", 1711 file_name) 1712 1713 raw_data = _Decompress(data) 1714 1715 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid, 1716 atime=atime, mtime=mtime)
1717
1718 1719 -def WriteSsconfFiles(values):
1720 """Update all ssconf files. 1721 1722 Wrapper around the SimpleStore.WriteFiles. 1723 1724 """ 1725 ssconf.SimpleStore().WriteFiles(values)
1726
1727 1728 -def _ErrnoOrStr(err):
1729 """Format an EnvironmentError exception. 1730 1731 If the L{err} argument has an errno attribute, it will be looked up 1732 and converted into a textual C{E...} description. Otherwise the 1733 string representation of the error will be returned. 1734 1735 @type err: L{EnvironmentError} 1736 @param err: the exception to format 1737 1738 """ 1739 if hasattr(err, 'errno'): 1740 detail = errno.errorcode[err.errno] 1741 else: 1742 detail = str(err) 1743 return detail
1744
1745 1746 -def _OSOndiskAPIVersion(os_dir):
1747 """Compute and return the API version of a given OS. 1748 1749 This function will try to read the API version of the OS residing in 1750 the 'os_dir' directory. 1751 1752 @type os_dir: str 1753 @param os_dir: the directory in which we should look for the OS 1754 @rtype: tuple 1755 @return: tuple (status, data) with status denoting the validity and 1756 data holding either the vaid versions or an error message 1757 1758 """ 1759 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE) 1760 1761 try: 1762 st = os.stat(api_file) 1763 except EnvironmentError, err: 1764 return False, ("Required file '%s' not found under path %s: %s" % 1765 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err))) 1766 1767 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)): 1768 return False, ("File '%s' in %s is not a regular file" % 1769 (constants.OS_API_FILE, os_dir)) 1770 1771 try: 1772 api_versions = utils.ReadFile(api_file).splitlines() 1773 except EnvironmentError, err: 1774 return False, ("Error while reading the API version file at %s: %s" % 1775 (api_file, _ErrnoOrStr(err))) 1776 1777 try: 1778 api_versions = [int(version.strip()) for version in api_versions] 1779 except (TypeError, ValueError), err: 1780 return False, ("API version(s) can't be converted to integer: %s" % 1781 str(err)) 1782 1783 return True, api_versions
1784
1785 1786 -def DiagnoseOS(top_dirs=None):
1787 """Compute the validity for all OSes. 1788 1789 @type top_dirs: list 1790 @param top_dirs: the list of directories in which to 1791 search (if not given defaults to 1792 L{constants.OS_SEARCH_PATH}) 1793 @rtype: list of L{objects.OS} 1794 @return: a list of tuples (name, path, status, diagnose, variants, 1795 parameters, api_version) for all (potential) OSes under all 1796 search paths, where: 1797 - name is the (potential) OS name 1798 - path is the full path to the OS 1799 - status True/False is the validity of the OS 1800 - diagnose is the error message for an invalid OS, otherwise empty 1801 - variants is a list of supported OS variants, if any 1802 - parameters is a list of (name, help) parameters, if any 1803 - api_version is a list of support OS API versions 1804 1805 """ 1806 if top_dirs is None: 1807 top_dirs = constants.OS_SEARCH_PATH 1808 1809 result = [] 1810 for dir_name in top_dirs: 1811 if os.path.isdir(dir_name): 1812 try: 1813 f_names = utils.ListVisibleFiles(dir_name) 1814 except EnvironmentError, err: 1815 logging.exception("Can't list the OS directory %s: %s", dir_name, err) 1816 break 1817 for name in f_names: 1818 os_path = utils.PathJoin(dir_name, name) 1819 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name) 1820 if status: 1821 diagnose = "" 1822 variants = os_inst.supported_variants 1823 parameters = os_inst.supported_parameters 1824 api_versions = os_inst.api_versions 1825 else: 1826 diagnose = os_inst 1827 variants = parameters = api_versions = [] 1828 result.append((name, os_path, status, diagnose, variants, 1829 parameters, api_versions)) 1830 1831 return result
1832
1833 1834 -def _TryOSFromDisk(name, base_dir=None):
1835 """Create an OS instance from disk. 1836 1837 This function will return an OS instance if the given name is a 1838 valid OS name. 1839 1840 @type base_dir: string 1841 @keyword base_dir: Base directory containing OS installations. 1842 Defaults to a search in all the OS_SEARCH_PATH dirs. 1843 @rtype: tuple 1844 @return: success and either the OS instance if we find a valid one, 1845 or error message 1846 1847 """ 1848 if base_dir is None: 1849 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir) 1850 else: 1851 os_dir = utils.FindFile(name, [base_dir], os.path.isdir) 1852 1853 if os_dir is None: 1854 return False, "Directory for OS %s not found in search path" % name 1855 1856 status, api_versions = _OSOndiskAPIVersion(os_dir) 1857 if not status: 1858 # push the error up 1859 return status, api_versions 1860 1861 if not constants.OS_API_VERSIONS.intersection(api_versions): 1862 return False, ("API version mismatch for path '%s': found %s, want %s." % 1863 (os_dir, api_versions, constants.OS_API_VERSIONS)) 1864 1865 # OS Files dictionary, we will populate it with the absolute path names 1866 os_files = dict.fromkeys(constants.OS_SCRIPTS) 1867 1868 if max(api_versions) >= constants.OS_API_V15: 1869 os_files[constants.OS_VARIANTS_FILE] = '' 1870 1871 if max(api_versions) >= constants.OS_API_V20: 1872 os_files[constants.OS_PARAMETERS_FILE] = '' 1873 else: 1874 del os_files[constants.OS_SCRIPT_VERIFY] 1875 1876 for filename in os_files: 1877 os_files[filename] = utils.PathJoin(os_dir, filename) 1878 1879 try: 1880 st = os.stat(os_files[filename]) 1881 except EnvironmentError, err: 1882 return False, ("File '%s' under path '%s' is missing (%s)" % 1883 (filename, os_dir, _ErrnoOrStr(err))) 1884 1885 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)): 1886 return False, ("File '%s' under path '%s' is not a regular file" % 1887 (filename, os_dir)) 1888 1889 if filename in constants.OS_SCRIPTS: 1890 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR: 1891 return False, ("File '%s' under path '%s' is not executable" % 1892 (filename, os_dir)) 1893 1894 variants = [] 1895 if constants.OS_VARIANTS_FILE in os_files: 1896 variants_file = os_files[constants.OS_VARIANTS_FILE] 1897 try: 1898 variants = utils.ReadFile(variants_file).splitlines() 1899 except EnvironmentError, err: 1900 return False, ("Error while reading the OS variants file at %s: %s" % 1901 (variants_file, _ErrnoOrStr(err))) 1902 if not variants: 1903 return False, ("No supported os variant found") 1904 1905 parameters = [] 1906 if constants.OS_PARAMETERS_FILE in os_files: 1907 parameters_file = os_files[constants.OS_PARAMETERS_FILE] 1908 try: 1909 parameters = utils.ReadFile(parameters_file).splitlines() 1910 except EnvironmentError, err: 1911 return False, ("Error while reading the OS parameters file at %s: %s" % 1912 (parameters_file, _ErrnoOrStr(err))) 1913 parameters = [v.split(None, 1) for v in parameters] 1914 1915 os_obj = objects.OS(name=name, path=os_dir, 1916 create_script=os_files[constants.OS_SCRIPT_CREATE], 1917 export_script=os_files[constants.OS_SCRIPT_EXPORT], 1918 import_script=os_files[constants.OS_SCRIPT_IMPORT], 1919 rename_script=os_files[constants.OS_SCRIPT_RENAME], 1920 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY, 1921 None), 1922 supported_variants=variants, 1923 supported_parameters=parameters, 1924 api_versions=api_versions) 1925 return True, os_obj
1926
1927 1928 -def OSFromDisk(name, base_dir=None):
1929 """Create an OS instance from disk. 1930 1931 This function will return an OS instance if the given name is a 1932 valid OS name. Otherwise, it will raise an appropriate 1933 L{RPCFail} exception, detailing why this is not a valid OS. 1934 1935 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise 1936 an exception but returns true/false status data. 1937 1938 @type base_dir: string 1939 @keyword base_dir: Base directory containing OS installations. 1940 Defaults to a search in all the OS_SEARCH_PATH dirs. 1941 @rtype: L{objects.OS} 1942 @return: the OS instance if we find a valid one 1943 @raise RPCFail: if we don't find a valid OS 1944 1945 """ 1946 name_only = objects.OS.GetName(name) 1947 status, payload = _TryOSFromDisk(name_only, base_dir) 1948 1949 if not status: 1950 _Fail(payload) 1951 1952 return payload
1953
1954 1955 -def OSCoreEnv(os_name, inst_os, os_params, debug=0):
1956 """Calculate the basic environment for an os script. 1957 1958 @type os_name: str 1959 @param os_name: full operating system name (including variant) 1960 @type inst_os: L{objects.OS} 1961 @param inst_os: operating system for which the environment is being built 1962 @type os_params: dict 1963 @param os_params: the OS parameters 1964 @type debug: integer 1965 @param debug: debug level (0 or 1, for OS Api 10) 1966 @rtype: dict 1967 @return: dict of environment variables 1968 @raise errors.BlockDeviceError: if the block device 1969 cannot be found 1970 1971 """ 1972 result = {} 1973 api_version = \ 1974 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions)) 1975 result['OS_API_VERSION'] = '%d' % api_version 1976 result['OS_NAME'] = inst_os.name 1977 result['DEBUG_LEVEL'] = '%d' % debug 1978 1979 # OS variants 1980 if api_version >= constants.OS_API_V15: 1981 variant = objects.OS.GetVariant(os_name) 1982 if not variant: 1983 variant = inst_os.supported_variants[0] 1984 result['OS_VARIANT'] = variant 1985 1986 # OS params 1987 for pname, pvalue in os_params.items(): 1988 result['OSP_%s' % pname.upper()] = pvalue 1989 1990 return result
1991
1992 1993 -def OSEnvironment(instance, inst_os, debug=0):
1994 """Calculate the environment for an os script. 1995 1996 @type instance: L{objects.Instance} 1997 @param instance: target instance for the os script run 1998 @type inst_os: L{objects.OS} 1999 @param inst_os: operating system for which the environment is being built 2000 @type debug: integer 2001 @param debug: debug level (0 or 1, for OS Api 10) 2002 @rtype: dict 2003 @return: dict of environment variables 2004 @raise errors.BlockDeviceError: if the block device 2005 cannot be found 2006 2007 """ 2008 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug) 2009 2010 result['INSTANCE_NAME'] = instance.name 2011 result['INSTANCE_OS'] = instance.os 2012 result['HYPERVISOR'] = instance.hypervisor 2013 result['DISK_COUNT'] = '%d' % len(instance.disks) 2014 result['NIC_COUNT'] = '%d' % len(instance.nics) 2015 2016 # Disks 2017 for idx, disk in enumerate(instance.disks): 2018 real_disk = _OpenRealBD(disk) 2019 result['DISK_%d_PATH' % idx] = real_disk.dev_path 2020 result['DISK_%d_ACCESS' % idx] = disk.mode 2021 if constants.HV_DISK_TYPE in instance.hvparams: 2022 result['DISK_%d_FRONTEND_TYPE' % idx] = \ 2023 instance.hvparams[constants.HV_DISK_TYPE] 2024 if disk.dev_type in constants.LDS_BLOCK: 2025 result['DISK_%d_BACKEND_TYPE' % idx] = 'block' 2026 elif disk.dev_type == constants.LD_FILE: 2027 result['DISK_%d_BACKEND_TYPE' % idx] = \ 2028 'file:%s' % disk.physical_id[0] 2029 2030 # NICs 2031 for idx, nic in enumerate(instance.nics): 2032 result['NIC_%d_MAC' % idx] = nic.mac 2033 if nic.ip: 2034 result['NIC_%d_IP' % idx] = nic.ip 2035 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE] 2036 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED: 2037 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK] 2038 if nic.nicparams[constants.NIC_LINK]: 2039 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK] 2040 if constants.HV_NIC_TYPE in instance.hvparams: 2041 result['NIC_%d_FRONTEND_TYPE' % idx] = \ 2042 instance.hvparams[constants.HV_NIC_TYPE] 2043 2044 # HV/BE params 2045 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]: 2046 for key, value in source.items(): 2047 result["INSTANCE_%s_%s" % (kind, key)] = str(value) 2048 2049 return result
2050
2051 2052 -def BlockdevGrow(disk, amount):
2053 """Grow a stack of block devices. 2054 2055 This function is called recursively, with the childrens being the 2056 first ones to resize. 2057 2058 @type disk: L{objects.Disk} 2059 @param disk: the disk to be grown 2060 @rtype: (status, result) 2061 @return: a tuple with the status of the operation 2062 (True/False), and the errors message if status 2063 is False 2064 2065 """ 2066 r_dev = _RecursiveFindBD(disk) 2067 if r_dev is None: 2068 _Fail("Cannot find block device %s", disk) 2069 2070 try: 2071 r_dev.Grow(amount) 2072 except errors.BlockDeviceError, err: 2073 _Fail("Failed to grow block device: %s", err, exc=True)
2074
2075 2076 -def BlockdevSnapshot(disk):
2077 """Create a snapshot copy of a block device. 2078 2079 This function is called recursively, and the snapshot is actually created 2080 just for the leaf lvm backend device. 2081 2082 @type disk: L{objects.Disk} 2083 @param disk: the disk to be snapshotted 2084 @rtype: string 2085 @return: snapshot disk path 2086 2087 """ 2088 if disk.dev_type == constants.LD_DRBD8: 2089 if not disk.children: 2090 _Fail("DRBD device '%s' without backing storage cannot be snapshotted", 2091 disk.unique_id) 2092 return BlockdevSnapshot(disk.children[0]) 2093 elif disk.dev_type == constants.LD_LV: 2094 r_dev = _RecursiveFindBD(disk) 2095 if r_dev is not None: 2096 # FIXME: choose a saner value for the snapshot size 2097 # let's stay on the safe side and ask for the full size, for now 2098 return r_dev.Snapshot(disk.size) 2099 else: 2100 _Fail("Cannot find block device %s", disk) 2101 else: 2102 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'", 2103 disk.unique_id, disk.dev_type)
2104
2105 2106 -def FinalizeExport(instance, snap_disks):
2107 """Write out the export configuration information. 2108 2109 @type instance: L{objects.Instance} 2110 @param instance: the instance which we export, used for 2111 saving configuration 2112 @type snap_disks: list of L{objects.Disk} 2113 @param snap_disks: list of snapshot block devices, which 2114 will be used to get the actual name of the dump file 2115 2116 @rtype: None 2117 2118 """ 2119 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new") 2120 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name) 2121 2122 config = objects.SerializableConfigParser() 2123 2124 config.add_section(constants.INISECT_EXP) 2125 config.set(constants.INISECT_EXP, 'version', '0') 2126 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time())) 2127 config.set(constants.INISECT_EXP, 'source', instance.primary_node) 2128 config.set(constants.INISECT_EXP, 'os', instance.os) 2129 config.set(constants.INISECT_EXP, 'compression', 'gzip') 2130 2131 config.add_section(constants.INISECT_INS) 2132 config.set(constants.INISECT_INS, 'name', instance.name) 2133 config.set(constants.INISECT_INS, 'memory', '%d' % 2134 instance.beparams[constants.BE_MEMORY]) 2135 config.set(constants.INISECT_INS, 'vcpus', '%d' % 2136 instance.beparams[constants.BE_VCPUS]) 2137 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template) 2138 config.set(constants.INISECT_INS, 'hypervisor', instance.hypervisor) 2139 2140 nic_total = 0 2141 for nic_count, nic in enumerate(instance.nics): 2142 nic_total += 1 2143 config.set(constants.INISECT_INS, 'nic%d_mac' % 2144 nic_count, '%s' % nic.mac) 2145 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip) 2146 for param in constants.NICS_PARAMETER_TYPES: 2147 config.set(constants.INISECT_INS, 'nic%d_%s' % (nic_count, param), 2148 '%s' % nic.nicparams.get(param, None)) 2149 # TODO: redundant: on load can read nics until it doesn't exist 2150 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total) 2151 2152 disk_total = 0 2153 for disk_count, disk in enumerate(snap_disks): 2154 if disk: 2155 disk_total += 1 2156 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count, 2157 ('%s' % disk.iv_name)) 2158 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count, 2159 ('%s' % disk.physical_id[1])) 2160 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count, 2161 ('%d' % disk.size)) 2162 2163 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total) 2164 2165 # New-style hypervisor/backend parameters 2166 2167 config.add_section(constants.INISECT_HYP) 2168 for name, value in instance.hvparams.items(): 2169 if name not in constants.HVC_GLOBALS: 2170 config.set(constants.INISECT_HYP, name, str(value)) 2171 2172 config.add_section(constants.INISECT_BEP) 2173 for name, value in instance.beparams.items(): 2174 config.set(constants.INISECT_BEP, name, str(value)) 2175 2176 config.add_section(constants.INISECT_OSP) 2177 for name, value in instance.osparams.items(): 2178 config.set(constants.INISECT_OSP, name, str(value)) 2179 2180 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE), 2181 data=config.Dumps()) 2182 shutil.rmtree(finaldestdir, ignore_errors=True) 2183 shutil.move(destdir, finaldestdir)
2184
2185 2186 -def ExportInfo(dest):
2187 """Get export configuration information. 2188 2189 @type dest: str 2190 @param dest: directory containing the export 2191 2192 @rtype: L{objects.SerializableConfigParser} 2193 @return: a serializable config file containing the 2194 export info 2195 2196 """ 2197 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE) 2198 2199 config = objects.SerializableConfigParser() 2200 config.read(cff) 2201 2202 if (not config.has_section(constants.INISECT_EXP) or 2203 not config.has_section(constants.INISECT_INS)): 2204 _Fail("Export info file doesn't have the required fields") 2205 2206 return config.Dumps()
2207
2208 2209 -def ListExports():
2210 """Return a list of exports currently available on this machine. 2211 2212 @rtype: list 2213 @return: list of the exports 2214 2215 """ 2216 if os.path.isdir(constants.EXPORT_DIR): 2217 return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR)) 2218 else: 2219 _Fail("No exports directory")
2220
2221 2222 -def RemoveExport(export):
2223 """Remove an existing export from the node. 2224 2225 @type export: str 2226 @param export: the name of the export to remove 2227 @rtype: None 2228 2229 """ 2230 target = utils.PathJoin(constants.EXPORT_DIR, export) 2231 2232 try: 2233 shutil.rmtree(target) 2234 except EnvironmentError, err: 2235 _Fail("Error while removing the export: %s", err, exc=True)
2236
2237 2238 -def BlockdevRename(devlist):
2239 """Rename a list of block devices. 2240 2241 @type devlist: list of tuples 2242 @param devlist: list of tuples of the form (disk, 2243 new_logical_id, new_physical_id); disk is an 2244 L{objects.Disk} object describing the current disk, 2245 and new logical_id/physical_id is the name we 2246 rename it to 2247 @rtype: boolean 2248 @return: True if all renames succeeded, False otherwise 2249 2250 """ 2251 msgs = [] 2252 result = True 2253 for disk, unique_id in devlist: 2254 dev = _RecursiveFindBD(disk) 2255 if dev is None: 2256 msgs.append("Can't find device %s in rename" % str(disk)) 2257 result = False 2258 continue 2259 try: 2260 old_rpath = dev.dev_path 2261 dev.Rename(unique_id) 2262 new_rpath = dev.dev_path 2263 if old_rpath != new_rpath: 2264 DevCacheManager.RemoveCache(old_rpath) 2265 # FIXME: we should add the new cache information here, like: 2266 # DevCacheManager.UpdateCache(new_rpath, owner, ...) 2267 # but we don't have the owner here - maybe parse from existing 2268 # cache? for now, we only lose lvm data when we rename, which 2269 # is less critical than DRBD or MD 2270 except errors.BlockDeviceError, err: 2271 msgs.append("Can't rename device '%s' to '%s': %s" % 2272 (dev, unique_id, err)) 2273 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id) 2274 result = False 2275 if not result: 2276 _Fail("; ".join(msgs))
2277
2278 2279 -def _TransformFileStorageDir(file_storage_dir):
2280 """Checks whether given file_storage_dir is valid. 2281 2282 Checks wheter the given file_storage_dir is within the cluster-wide 2283 default file_storage_dir stored in SimpleStore. Only paths under that 2284 directory are allowed. 2285 2286 @type file_storage_dir: str 2287 @param file_storage_dir: the path to check 2288 2289 @return: the normalized path if valid, None otherwise 2290 2291 """ 2292 if not constants.ENABLE_FILE_STORAGE: 2293 _Fail("File storage disabled at configure time") 2294 cfg = _GetConfig() 2295 file_storage_dir = os.path.normpath(file_storage_dir) 2296 base_file_storage_dir = cfg.GetFileStorageDir() 2297 if (os.path.commonprefix([file_storage_dir, base_file_storage_dir]) != 2298 base_file_storage_dir): 2299 _Fail("File storage directory '%s' is not under base file" 2300 " storage directory '%s'", file_storage_dir, base_file_storage_dir) 2301 return file_storage_dir
2302
2303 2304 -def CreateFileStorageDir(file_storage_dir):
2305 """Create file storage directory. 2306 2307 @type file_storage_dir: str 2308 @param file_storage_dir: directory to create 2309 2310 @rtype: tuple 2311 @return: tuple with first element a boolean indicating wheter dir 2312 creation was successful or not 2313 2314 """ 2315 file_storage_dir = _TransformFileStorageDir(file_storage_dir) 2316 if os.path.exists(file_storage_dir): 2317 if not os.path.isdir(file_storage_dir): 2318 _Fail("Specified storage dir '%s' is not a directory", 2319 file_storage_dir) 2320 else: 2321 try: 2322 os.makedirs(file_storage_dir, 0750) 2323 except OSError, err: 2324 _Fail("Cannot create file storage directory '%s': %s", 2325 file_storage_dir, err, exc=True)
2326
2327 2328 -def RemoveFileStorageDir(file_storage_dir):
2329 """Remove file storage directory. 2330 2331 Remove it only if it's empty. If not log an error and return. 2332 2333 @type file_storage_dir: str 2334 @param file_storage_dir: the directory we should cleanup 2335 @rtype: tuple (success,) 2336 @return: tuple of one element, C{success}, denoting 2337 whether the operation was successful 2338 2339 """ 2340 file_storage_dir = _TransformFileStorageDir(file_storage_dir) 2341 if os.path.exists(file_storage_dir): 2342 if not os.path.isdir(file_storage_dir): 2343 _Fail("Specified Storage directory '%s' is not a directory", 2344 file_storage_dir) 2345 # deletes dir only if empty, otherwise we want to fail the rpc call 2346 try: 2347 os.rmdir(file_storage_dir) 2348 except OSError, err: 2349 _Fail("Cannot remove file storage directory '%s': %s", 2350 file_storage_dir, err)
2351
2352 2353 -def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2354 """Rename the file storage directory. 2355 2356 @type old_file_storage_dir: str 2357 @param old_file_storage_dir: the current path 2358 @type new_file_storage_dir: str 2359 @param new_file_storage_dir: the name we should rename to 2360 @rtype: tuple (success,) 2361 @return: tuple of one element, C{success}, denoting 2362 whether the operation was successful 2363 2364 """ 2365 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir) 2366 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir) 2367 if not os.path.exists(new_file_storage_dir): 2368 if os.path.isdir(old_file_storage_dir): 2369 try: 2370 os.rename(old_file_storage_dir, new_file_storage_dir) 2371 except OSError, err: 2372 _Fail("Cannot rename '%s' to '%s': %s", 2373 old_file_storage_dir, new_file_storage_dir, err) 2374 else: 2375 _Fail("Specified storage dir '%s' is not a directory", 2376 old_file_storage_dir) 2377 else: 2378 if os.path.exists(old_file_storage_dir): 2379 _Fail("Cannot rename '%s' to '%s': both locations exist", 2380 old_file_storage_dir, new_file_storage_dir)
2381
2382 2383 -def _EnsureJobQueueFile(file_name):
2384 """Checks whether the given filename is in the queue directory. 2385 2386 @type file_name: str 2387 @param file_name: the file name we should check 2388 @rtype: None 2389 @raises RPCFail: if the file is not valid 2390 2391 """ 2392 queue_dir = os.path.normpath(constants.QUEUE_DIR) 2393 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir) 2394 2395 if not result: 2396 _Fail("Passed job queue file '%s' does not belong to" 2397 " the queue directory '%s'", file_name, queue_dir)
2398
2399 2400 -def JobQueueUpdate(file_name, content):
2401 """Updates a file in the queue directory. 2402 2403 This is just a wrapper over L{utils.WriteFile}, with proper 2404 checking. 2405 2406 @type file_name: str 2407 @param file_name: the job file name 2408 @type content: str 2409 @param content: the new job contents 2410 @rtype: boolean 2411 @return: the success of the operation 2412 2413 """ 2414 _EnsureJobQueueFile(file_name) 2415 2416 # Write and replace the file atomically 2417 utils.WriteFile(file_name, data=_Decompress(content))
2418
2419 2420 -def JobQueueRename(old, new):
2421 """Renames a job queue file. 2422 2423 This is just a wrapper over os.rename with proper checking. 2424 2425 @type old: str 2426 @param old: the old (actual) file name 2427 @type new: str 2428 @param new: the desired file name 2429 @rtype: tuple 2430 @return: the success of the operation and payload 2431 2432 """ 2433 _EnsureJobQueueFile(old) 2434 _EnsureJobQueueFile(new) 2435 2436 utils.RenameFile(old, new, mkdir=True)
2437
2438 2439 -def BlockdevClose(instance_name, disks):
2440 """Closes the given block devices. 2441 2442 This means they will be switched to secondary mode (in case of 2443 DRBD). 2444 2445 @param instance_name: if the argument is not empty, the symlinks 2446 of this instance will be removed 2447 @type disks: list of L{objects.Disk} 2448 @param disks: the list of disks to be closed 2449 @rtype: tuple (success, message) 2450 @return: a tuple of success and message, where success 2451 indicates the succes of the operation, and message 2452 which will contain the error details in case we 2453 failed 2454 2455 """ 2456 bdevs = [] 2457 for cf in disks: 2458 rd = _RecursiveFindBD(cf) 2459 if rd is None: 2460 _Fail("Can't find device %s", cf) 2461 bdevs.append(rd) 2462 2463 msg = [] 2464 for rd in bdevs: 2465 try: 2466 rd.Close() 2467 except errors.BlockDeviceError, err: 2468 msg.append(str(err)) 2469 if msg: 2470 _Fail("Can't make devices secondary: %s", ",".join(msg)) 2471 else: 2472 if instance_name: 2473 _RemoveBlockDevLinks(instance_name, disks)
2474
2475 2476 -def ValidateHVParams(hvname, hvparams):
2477 """Validates the given hypervisor parameters. 2478 2479 @type hvname: string 2480 @param hvname: the hypervisor name 2481 @type hvparams: dict 2482 @param hvparams: the hypervisor parameters to be validated 2483 @rtype: None 2484 2485 """ 2486 try: 2487 hv_type = hypervisor.GetHypervisor(hvname) 2488 hv_type.ValidateParameters(hvparams) 2489 except errors.HypervisorError, err: 2490 _Fail(str(err), log=False)
2491
2492 2493 -def _CheckOSPList(os_obj, parameters):
2494 """Check whether a list of parameters is supported by the OS. 2495 2496 @type os_obj: L{objects.OS} 2497 @param os_obj: OS object to check 2498 @type parameters: list 2499 @param parameters: the list of parameters to check 2500 2501 """ 2502 supported = [v[0] for v in os_obj.supported_parameters] 2503 delta = frozenset(parameters).difference(supported) 2504 if delta: 2505 _Fail("The following parameters are not supported" 2506 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2507
2508 2509 -def ValidateOS(required, osname, checks, osparams):
2510 """Validate the given OS' parameters. 2511 2512 @type required: boolean 2513 @param required: whether absence of the OS should translate into 2514 failure or not 2515 @type osname: string 2516 @param osname: the OS to be validated 2517 @type checks: list 2518 @param checks: list of the checks to run (currently only 'parameters') 2519 @type osparams: dict 2520 @param osparams: dictionary with OS parameters 2521 @rtype: boolean 2522 @return: True if the validation passed, or False if the OS was not 2523 found and L{required} was false 2524 2525 """ 2526 if not constants.OS_VALIDATE_CALLS.issuperset(checks): 2527 _Fail("Unknown checks required for OS %s: %s", osname, 2528 set(checks).difference(constants.OS_VALIDATE_CALLS)) 2529 2530 name_only = objects.OS.GetName(osname) 2531 status, tbv = _TryOSFromDisk(name_only, None) 2532 2533 if not status: 2534 if required: 2535 _Fail(tbv) 2536 else: 2537 return False 2538 2539 if max(tbv.api_versions) < constants.OS_API_V20: 2540 return True 2541 2542 if constants.OS_VALIDATE_PARAMETERS in checks: 2543 _CheckOSPList(tbv, osparams.keys()) 2544 2545 validate_env = OSCoreEnv(osname, tbv, osparams) 2546 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env, 2547 cwd=tbv.path) 2548 if result.failed: 2549 logging.error("os validate command '%s' returned error: %s output: %s", 2550 result.cmd, result.fail_reason, result.output) 2551 _Fail("OS validation script failed (%s), output: %s", 2552 result.fail_reason, result.output, log=False) 2553 2554 return True
2555
2556 2557 -def DemoteFromMC():
2558 """Demotes the current node from master candidate role. 2559 2560 """ 2561 # try to ensure we're not the master by mistake 2562 master, myself = ssconf.GetMasterAndMyself() 2563 if master == myself: 2564 _Fail("ssconf status shows I'm the master node, will not demote") 2565 2566 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD]) 2567 if not result.failed: 2568 _Fail("The master daemon is running, will not demote") 2569 2570 try: 2571 if os.path.isfile(constants.CLUSTER_CONF_FILE): 2572 utils.CreateBackup(constants.CLUSTER_CONF_FILE) 2573 except EnvironmentError, err: 2574 if err.errno != errno.ENOENT: 2575 _Fail("Error while backing up cluster file: %s", err, exc=True) 2576 2577 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2578
2579 2580 -def _GetX509Filenames(cryptodir, name):
2581 """Returns the full paths for the private key and certificate. 2582 2583 """ 2584 return (utils.PathJoin(cryptodir, name), 2585 utils.PathJoin(cryptodir, name, _X509_KEY_FILE), 2586 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2587
2588 2589 -def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2590 """Creates a new X509 certificate for SSL/TLS. 2591 2592 @type validity: int 2593 @param validity: Validity in seconds 2594 @rtype: tuple; (string, string) 2595 @return: Certificate name and public part 2596 2597 """ 2598 (key_pem, cert_pem) = \ 2599 utils.GenerateSelfSignedX509Cert(netutils.HostInfo.SysName(), 2600 min(validity, _MAX_SSL_CERT_VALIDITY)) 2601 2602 cert_dir = tempfile.mkdtemp(dir=cryptodir, 2603 prefix="x509-%s-" % utils.TimestampForFilename()) 2604 try: 2605 name = os.path.basename(cert_dir) 2606 assert len(name) > 5 2607 2608 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name) 2609 2610 utils.WriteFile(key_file, mode=0400, data=key_pem) 2611 utils.WriteFile(cert_file, mode=0400, data=cert_pem) 2612 2613 # Never return private key as it shouldn't leave the node 2614 return (name, cert_pem) 2615 except Exception: 2616 shutil.rmtree(cert_dir, ignore_errors=True) 2617 raise
2618
2619 2620 -def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2621 """Removes a X509 certificate. 2622 2623 @type name: string 2624 @param name: Certificate name 2625 2626 """ 2627 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name) 2628 2629 utils.RemoveFile(key_file) 2630 utils.RemoveFile(cert_file) 2631 2632 try: 2633 os.rmdir(cert_dir) 2634 except EnvironmentError, err: 2635 _Fail("Cannot remove certificate directory '%s': %s", 2636 cert_dir, err)
2637
2638 2639 -def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2640 """Returns the command for the requested input/output. 2641 2642 @type instance: L{objects.Instance} 2643 @param instance: The instance object 2644 @param mode: Import/export mode 2645 @param ieio: Input/output type 2646 @param ieargs: Input/output arguments 2647 2648 """ 2649 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT) 2650 2651 env = None 2652 prefix = None 2653 suffix = None 2654 exp_size = None 2655 2656 if ieio == constants.IEIO_FILE: 2657 (filename, ) = ieargs 2658 2659 if not utils.IsNormAbsPath(filename): 2660 _Fail("Path '%s' is not normalized or absolute", filename) 2661 2662 directory = os.path.normpath(os.path.dirname(filename)) 2663 2664 if (os.path.commonprefix([constants.EXPORT_DIR, directory]) != 2665 constants.EXPORT_DIR): 2666 _Fail("File '%s' is not under exports directory '%s'", 2667 filename, constants.EXPORT_DIR) 2668 2669 # Create directory 2670 utils.Makedirs(directory, mode=0750) 2671 2672 quoted_filename = utils.ShellQuote(filename) 2673 2674 if mode == constants.IEM_IMPORT: 2675 suffix = "> %s" % quoted_filename 2676 elif mode == constants.IEM_EXPORT: 2677 suffix = "< %s" % quoted_filename 2678 2679 # Retrieve file size 2680 try: 2681 st = os.stat(filename) 2682 except EnvironmentError, err: 2683 logging.error("Can't stat(2) %s: %s", filename, err) 2684 else: 2685 exp_size = utils.BytesToMebibyte(st.st_size) 2686 2687 elif ieio == constants.IEIO_RAW_DISK: 2688 (disk, ) = ieargs 2689 2690 real_disk = _OpenRealBD(disk) 2691 2692 if mode == constants.IEM_IMPORT: 2693 # we set here a smaller block size as, due to transport buffering, more 2694 # than 64-128k will mostly ignored; we use nocreat to fail if the device 2695 # is not already there or we pass a wrong path; we use notrunc to no 2696 # attempt truncate on an LV device; we use oflag=dsync to not buffer too 2697 # much memory; this means that at best, we flush every 64k, which will 2698 # not be very fast 2699 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc" 2700 " bs=%s oflag=dsync"), 2701 real_disk.dev_path, 2702 str(64 * 1024)) 2703 2704 elif mode == constants.IEM_EXPORT: 2705 # the block size on the read dd is 1MiB to match our units 2706 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |", 2707 real_disk.dev_path, 2708 str(1024 * 1024), # 1 MB 2709 str(disk.size)) 2710 exp_size = disk.size 2711 2712 elif ieio == constants.IEIO_SCRIPT: 2713 (disk, disk_index, ) = ieargs 2714 2715 assert isinstance(disk_index, (int, long)) 2716 2717 real_disk = _OpenRealBD(disk) 2718 2719 inst_os = OSFromDisk(instance.os) 2720 env = OSEnvironment(instance, inst_os) 2721 2722 if mode == constants.IEM_IMPORT: 2723 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index] 2724 env["IMPORT_INDEX"] = str(disk_index) 2725 script = inst_os.import_script 2726 2727 elif mode == constants.IEM_EXPORT: 2728 env["EXPORT_DEVICE"] = real_disk.dev_path 2729 env["EXPORT_INDEX"] = str(disk_index) 2730 script = inst_os.export_script 2731 2732 # TODO: Pass special environment only to script 2733 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script) 2734 2735 if mode == constants.IEM_IMPORT: 2736 suffix = "| %s" % script_cmd 2737 2738 elif mode == constants.IEM_EXPORT: 2739 prefix = "%s |" % script_cmd 2740 2741 # Let script predict size 2742 exp_size = constants.IE_CUSTOM_SIZE 2743 2744 else: 2745 _Fail("Invalid %s I/O mode %r", mode, ieio) 2746 2747 return (env, prefix, suffix, exp_size)
2748
2749 2750 -def _CreateImportExportStatusDir(prefix):
2751 """Creates status directory for import/export. 2752 2753 """ 2754 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR, 2755 prefix=("%s-%s-" % 2756 (prefix, utils.TimestampForFilename())))
2757
2758 2759 -def StartImportExportDaemon(mode, opts, host, port, instance, ieio, ieioargs):
2760 """Starts an import or export daemon. 2761 2762 @param mode: Import/output mode 2763 @type opts: L{objects.ImportExportOptions} 2764 @param opts: Daemon options 2765 @type host: string 2766 @param host: Remote host for export (None for import) 2767 @type port: int 2768 @param port: Remote port for export (None for import) 2769 @type instance: L{objects.Instance} 2770 @param instance: Instance object 2771 @param ieio: Input/output type 2772 @param ieioargs: Input/output arguments 2773 2774 """ 2775 if mode == constants.IEM_IMPORT: 2776 prefix = "import" 2777 2778 if not (host is None and port is None): 2779 _Fail("Can not specify host or port on import") 2780 2781 elif mode == constants.IEM_EXPORT: 2782 prefix = "export" 2783 2784 if host is None or port is None: 2785 _Fail("Host and port must be specified for an export") 2786 2787 else: 2788 _Fail("Invalid mode %r", mode) 2789 2790 if (opts.key_name is None) ^ (opts.ca_pem is None): 2791 _Fail("Cluster certificate can only be used for both key and CA") 2792 2793 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \ 2794 _GetImportExportIoCommand(instance, mode, ieio, ieioargs) 2795 2796 if opts.key_name is None: 2797 # Use server.pem 2798 key_path = constants.NODED_CERT_FILE 2799 cert_path = constants.NODED_CERT_FILE 2800 assert opts.ca_pem is None 2801 else: 2802 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR, 2803 opts.key_name) 2804 assert opts.ca_pem is not None 2805 2806 for i in [key_path, cert_path]: 2807 if not os.path.exists(i): 2808 _Fail("File '%s' does not exist" % i) 2809 2810 status_dir = _CreateImportExportStatusDir(prefix) 2811 try: 2812 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE) 2813 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE) 2814 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE) 2815 2816 if opts.ca_pem is None: 2817 # Use server.pem 2818 ca = utils.ReadFile(constants.NODED_CERT_FILE) 2819 else: 2820 ca = opts.ca_pem 2821 2822 # Write CA file 2823 utils.WriteFile(ca_file, data=ca, mode=0400) 2824 2825 cmd = [ 2826 constants.IMPORT_EXPORT_DAEMON, 2827 status_file, mode, 2828 "--key=%s" % key_path, 2829 "--cert=%s" % cert_path, 2830 "--ca=%s" % ca_file, 2831 ] 2832 2833 if host: 2834 cmd.append("--host=%s" % host) 2835 2836 if port: 2837 cmd.append("--port=%s" % port) 2838 2839 if opts.compress: 2840 cmd.append("--compress=%s" % opts.compress) 2841 2842 if opts.magic: 2843 cmd.append("--magic=%s" % opts.magic) 2844 2845 if exp_size is not None: 2846 cmd.append("--expected-size=%s" % exp_size) 2847 2848 if cmd_prefix: 2849 cmd.append("--cmd-prefix=%s" % cmd_prefix) 2850 2851 if cmd_suffix: 2852 cmd.append("--cmd-suffix=%s" % cmd_suffix) 2853 2854 logfile = _InstanceLogName(prefix, instance.os, instance.name) 2855 2856 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has 2857 # support for receiving a file descriptor for output 2858 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file, 2859 output=logfile) 2860 2861 # The import/export name is simply the status directory name 2862 return os.path.basename(status_dir) 2863 2864 except Exception: 2865 shutil.rmtree(status_dir, ignore_errors=True) 2866 raise
2867
2868 2869 -def GetImportExportStatus(names):
2870 """Returns import/export daemon status. 2871 2872 @type names: sequence 2873 @param names: List of names 2874 @rtype: List of dicts 2875 @return: Returns a list of the state of each named import/export or None if a 2876 status couldn't be read 2877 2878 """ 2879 result = [] 2880 2881 for name in names: 2882 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name, 2883 _IES_STATUS_FILE) 2884 2885 try: 2886 data = utils.ReadFile(status_file) 2887 except EnvironmentError, err: 2888 if err.errno != errno.ENOENT: 2889 raise 2890 data = None 2891 2892 if not data: 2893 result.append(None) 2894 continue 2895 2896 result.append(serializer.LoadJson(data)) 2897 2898 return result
2899
2900 2901 -def AbortImportExport(name):
2902 """Sends SIGTERM to a running import/export daemon. 2903 2904 """ 2905 logging.info("Abort import/export %s", name) 2906 2907 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name) 2908 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE)) 2909 2910 if pid: 2911 logging.info("Import/export %s is running with PID %s, sending SIGTERM", 2912 name, pid) 2913 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
2914
2915 2916 -def CleanupImportExport(name):
2917 """Cleanup after an import or export. 2918 2919 If the import/export daemon is still running it's killed. Afterwards the 2920 whole status directory is removed. 2921 2922 """ 2923 logging.info("Finalizing import/export %s", name) 2924 2925 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name) 2926 2927 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE)) 2928 2929 if pid: 2930 logging.info("Import/export %s is still running with PID %s", 2931 name, pid) 2932 utils.KillProcess(pid, waitpid=False) 2933 2934 shutil.rmtree(status_dir, ignore_errors=True)
2935
2936 2937 -def _FindDisks(nodes_ip, disks):
2938 """Sets the physical ID on disks and returns the block devices. 2939 2940 """ 2941 # set the correct physical ID 2942 my_name = netutils.HostInfo().name 2943 for cf in disks: 2944 cf.SetPhysicalID(my_name, nodes_ip) 2945 2946 bdevs = [] 2947 2948 for cf in disks: 2949 rd = _RecursiveFindBD(cf) 2950 if rd is None: 2951 _Fail("Can't find device %s", cf) 2952 bdevs.append(rd) 2953 return bdevs
2954
2955 2956 -def DrbdDisconnectNet(nodes_ip, disks):
2957 """Disconnects the network on a list of drbd devices. 2958 2959 """ 2960 bdevs = _FindDisks(nodes_ip, disks) 2961 2962 # disconnect disks 2963 for rd in bdevs: 2964 try: 2965 rd.DisconnectNet() 2966 except errors.BlockDeviceError, err: 2967 _Fail("Can't change network configuration to standalone mode: %s", 2968 err, exc=True)
2969
2970 2971 -def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2972 """Attaches the network on a list of drbd devices. 2973 2974 """ 2975 bdevs = _FindDisks(nodes_ip, disks) 2976 2977 if multimaster: 2978 for idx, rd in enumerate(bdevs): 2979 try: 2980 _SymlinkBlockDev(instance_name, rd.dev_path, idx) 2981 except EnvironmentError, err: 2982 _Fail("Can't create symlink: %s", err) 2983 # reconnect disks, switch to new master configuration and if 2984 # needed primary mode 2985 for rd in bdevs: 2986 try: 2987 rd.AttachNet(multimaster) 2988 except errors.BlockDeviceError, err: 2989 _Fail("Can't change network configuration: %s", err) 2990 2991 # wait until the disks are connected; we need to retry the re-attach 2992 # if the device becomes standalone, as this might happen if the one 2993 # node disconnects and reconnects in a different mode before the 2994 # other node reconnects; in this case, one or both of the nodes will 2995 # decide it has wrong configuration and switch to standalone 2996 2997 def _Attach(): 2998 all_connected = True 2999 3000 for rd in bdevs: 3001 stats = rd.GetProcStatus() 3002 3003 all_connected = (all_connected and 3004 (stats.is_connected or stats.is_in_resync)) 3005 3006 if stats.is_standalone: 3007 # peer had different config info and this node became 3008 # standalone, even though this should not happen with the 3009 # new staged way of changing disk configs 3010 try: 3011 rd.AttachNet(multimaster) 3012 except errors.BlockDeviceError, err: 3013 _Fail("Can't change network configuration: %s", err) 3014 3015 if not all_connected: 3016 raise utils.RetryAgain()
3017 3018 try: 3019 # Start with a delay of 100 miliseconds and go up to 5 seconds 3020 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60) 3021 except utils.RetryTimeout: 3022 _Fail("Timeout in disk reconnecting") 3023 3024 if multimaster: 3025 # change to primary mode 3026 for rd in bdevs: 3027 try: 3028 rd.Open() 3029 except errors.BlockDeviceError, err: 3030 _Fail("Can't change to primary mode: %s", err) 3031
3032 3033 -def DrbdWaitSync(nodes_ip, disks):
3034 """Wait until DRBDs have synchronized. 3035 3036 """ 3037 def _helper(rd): 3038 stats = rd.GetProcStatus() 3039 if not (stats.is_connected or stats.is_in_resync): 3040 raise utils.RetryAgain() 3041 return stats
3042 3043 bdevs = _FindDisks(nodes_ip, disks) 3044 3045 min_resync = 100 3046 alldone = True 3047 for rd in bdevs: 3048 try: 3049 # poll each second for 15 seconds 3050 stats = utils.Retry(_helper, 1, 15, args=[rd]) 3051 except utils.RetryTimeout: 3052 stats = rd.GetProcStatus() 3053 # last check 3054 if not (stats.is_connected or stats.is_in_resync): 3055 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats) 3056 alldone = alldone and (not stats.is_in_resync) 3057 if stats.sync_percent is not None: 3058 min_resync = min(min_resync, stats.sync_percent) 3059 3060 return (alldone, min_resync) 3061
3062 3063 -def GetDrbdUsermodeHelper():
3064 """Returns DRBD usermode helper currently configured. 3065 3066 """ 3067 try: 3068 return bdev.BaseDRBD.GetUsermodeHelper() 3069 except errors.BlockDeviceError, err: 3070 _Fail(str(err))
3071
3072 3073 -def PowercycleNode(hypervisor_type):
3074 """Hard-powercycle the node. 3075 3076 Because we need to return first, and schedule the powercycle in the 3077 background, we won't be able to report failures nicely. 3078 3079 """ 3080 hyper = hypervisor.GetHypervisor(hypervisor_type) 3081 try: 3082 pid = os.fork() 3083 except OSError: 3084 # if we can't fork, we'll pretend that we're in the child process 3085 pid = 0 3086 if pid > 0: 3087 return "Reboot scheduled in 5 seconds" 3088 # ensure the child is running on ram 3089 try: 3090 utils.Mlockall() 3091 except Exception: # pylint: disable-msg=W0703 3092 pass 3093 time.sleep(5) 3094 hyper.PowercycleNode()
3095
3096 3097 -class HooksRunner(object):
3098 """Hook runner. 3099 3100 This class is instantiated on the node side (ganeti-noded) and not 3101 on the master side. 3102 3103 """
3104 - def __init__(self, hooks_base_dir=None):
3105 """Constructor for hooks runner. 3106 3107 @type hooks_base_dir: str or None 3108 @param hooks_base_dir: if not None, this overrides the 3109 L{constants.HOOKS_BASE_DIR} (useful for unittests) 3110 3111 """ 3112 if hooks_base_dir is None: 3113 hooks_base_dir = constants.HOOKS_BASE_DIR 3114 # yeah, _BASE_DIR is not valid for attributes, we use it like a 3115 # constant 3116 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3117
3118 - def RunHooks(self, hpath, phase, env):
3119 """Run the scripts in the hooks directory. 3120 3121 @type hpath: str 3122 @param hpath: the path to the hooks directory which 3123 holds the scripts 3124 @type phase: str 3125 @param phase: either L{constants.HOOKS_PHASE_PRE} or 3126 L{constants.HOOKS_PHASE_POST} 3127 @type env: dict 3128 @param env: dictionary with the environment for the hook 3129 @rtype: list 3130 @return: list of 3-element tuples: 3131 - script path 3132 - script result, either L{constants.HKR_SUCCESS} or 3133 L{constants.HKR_FAIL} 3134 - output of the script 3135 3136 @raise errors.ProgrammerError: for invalid input 3137 parameters 3138 3139 """ 3140 if phase == constants.HOOKS_PHASE_PRE: 3141 suffix = "pre" 3142 elif phase == constants.HOOKS_PHASE_POST: 3143 suffix = "post" 3144 else: 3145 _Fail("Unknown hooks phase '%s'", phase) 3146 3147 3148 subdir = "%s-%s.d" % (hpath, suffix) 3149 dir_name = utils.PathJoin(self._BASE_DIR, subdir) 3150 3151 results = [] 3152 3153 if not os.path.isdir(dir_name): 3154 # for non-existing/non-dirs, we simply exit instead of logging a 3155 # warning at every operation 3156 return results 3157 3158 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True) 3159 3160 for (relname, relstatus, runresult) in runparts_results: 3161 if relstatus == constants.RUNPARTS_SKIP: 3162 rrval = constants.HKR_SKIP 3163 output = "" 3164 elif relstatus == constants.RUNPARTS_ERR: 3165 rrval = constants.HKR_FAIL 3166 output = "Hook script execution error: %s" % runresult 3167 elif relstatus == constants.RUNPARTS_RUN: 3168 if runresult.failed: 3169 rrval = constants.HKR_FAIL 3170 else: 3171 rrval = constants.HKR_SUCCESS 3172 output = utils.SafeEncode(runresult.output.strip()) 3173 results.append(("%s/%s" % (subdir, relname), rrval, output)) 3174 3175 return results
3176
3177 3178 -class IAllocatorRunner(object):
3179 """IAllocator runner. 3180 3181 This class is instantiated on the node side (ganeti-noded) and not on 3182 the master side. 3183 3184 """ 3185 @staticmethod
3186 - def Run(name, idata):
3187 """Run an iallocator script. 3188 3189 @type name: str 3190 @param name: the iallocator script name 3191 @type idata: str 3192 @param idata: the allocator input data 3193 3194 @rtype: tuple 3195 @return: two element tuple of: 3196 - status 3197 - either error message or stdout of allocator (for success) 3198 3199 """ 3200 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH, 3201 os.path.isfile) 3202 if alloc_script is None: 3203 _Fail("iallocator module '%s' not found in the search path", name) 3204 3205 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.") 3206 try: 3207 os.write(fd, idata) 3208 os.close(fd) 3209 result = utils.RunCmd([alloc_script, fin_name]) 3210 if result.failed: 3211 _Fail("iallocator module '%s' failed: %s, output '%s'", 3212 name, result.fail_reason, result.output) 3213 finally: 3214 os.unlink(fin_name) 3215 3216 return result.stdout
3217
3218 3219 -class DevCacheManager(object):
3220 """Simple class for managing a cache of block device information. 3221 3222 """ 3223 _DEV_PREFIX = "/dev/" 3224 _ROOT_DIR = constants.BDEV_CACHE_DIR 3225 3226 @classmethod
3227 - def _ConvertPath(cls, dev_path):
3228 """Converts a /dev/name path to the cache file name. 3229 3230 This replaces slashes with underscores and strips the /dev 3231 prefix. It then returns the full path to the cache file. 3232 3233 @type dev_path: str 3234 @param dev_path: the C{/dev/} path name 3235 @rtype: str 3236 @return: the converted path name 3237 3238 """ 3239 if dev_path.startswith(cls._DEV_PREFIX): 3240 dev_path = dev_path[len(cls._DEV_PREFIX):] 3241 dev_path = dev_path.replace("/", "_") 3242 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path) 3243 return fpath
3244 3245 @classmethod
3246 - def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3247 """Updates the cache information for a given device. 3248 3249 @type dev_path: str 3250 @param dev_path: the pathname of the device 3251 @type owner: str 3252 @param owner: the owner (instance name) of the device 3253 @type on_primary: bool 3254 @param on_primary: whether this is the primary 3255 node nor not 3256 @type iv_name: str 3257 @param iv_name: the instance-visible name of the 3258 device, as in objects.Disk.iv_name 3259 3260 @rtype: None 3261 3262 """ 3263 if dev_path is None: 3264 logging.error("DevCacheManager.UpdateCache got a None dev_path") 3265 return 3266 fpath = cls._ConvertPath(dev_path) 3267 if on_primary: 3268 state = "primary" 3269 else: 3270 state = "secondary" 3271 if iv_name is None: 3272 iv_name = "not_visible" 3273 fdata = "%s %s %s\n" % (str(owner), state, iv_name) 3274 try: 3275 utils.WriteFile(fpath, data=fdata) 3276 except EnvironmentError, err: 3277 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3278 3279 @classmethod
3280 - def RemoveCache(cls, dev_path):
3281 """Remove data for a dev_path. 3282 3283 This is just a wrapper over L{utils.RemoveFile} with a converted 3284 path name and logging. 3285 3286 @type dev_path: str 3287 @param dev_path: the pathname of the device 3288 3289 @rtype: None 3290 3291 """ 3292 if dev_path is None: 3293 logging.error("DevCacheManager.RemoveCache got a None dev_path") 3294 return 3295 fpath = cls._ConvertPath(dev_path) 3296 try: 3297 utils.RemoveFile(fpath) 3298 except EnvironmentError, err: 3299 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3300