1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22 """Configuration management for Ganeti
23
24 This module provides the interface to the Ganeti cluster configuration.
25
26 The configuration data is stored on every node but is updated on the master
27 only. After each update, the master distributes the data to the other nodes.
28
29 Currently, the data storage format is JSON. YAML was slow and consuming too
30 much memory.
31
32 """
33
34
35
36
37 import os
38 import random
39 import logging
40 import time
41 import itertools
42
43 from ganeti import errors
44 from ganeti import locking
45 from ganeti import utils
46 from ganeti import constants
47 from ganeti import rpc
48 from ganeti import objects
49 from ganeti import serializer
50 from ganeti import uidpool
51 from ganeti import netutils
52 from ganeti import runtime
53
54
55 _config_lock = locking.SharedLock("ConfigWriter")
56
57
58 _UPGRADE_CONFIG_JID = "jid-cfg-upgrade"
72
75 """A temporary resource reservation manager.
76
77 This is used to reserve resources in a job, before using them, making sure
78 other jobs cannot get them in the meantime.
79
80 """
82 self._ec_reserved = {}
83
85 for holder_reserved in self._ec_reserved.values():
86 if resource in holder_reserved:
87 return True
88 return False
89
90 - def Reserve(self, ec_id, resource):
91 if self.Reserved(resource):
92 raise errors.ReservationError("Duplicate reservation for resource '%s'"
93 % str(resource))
94 if ec_id not in self._ec_reserved:
95 self._ec_reserved[ec_id] = set([resource])
96 else:
97 self._ec_reserved[ec_id].add(resource)
98
100 if ec_id in self._ec_reserved:
101 del self._ec_reserved[ec_id]
102
104 all_reserved = set()
105 for holder_reserved in self._ec_reserved.values():
106 all_reserved.update(holder_reserved)
107 return all_reserved
108
109 - def Generate(self, existing, generate_one_fn, ec_id):
110 """Generate a new resource of this type
111
112 """
113 assert callable(generate_one_fn)
114
115 all_elems = self.GetReserved()
116 all_elems.update(existing)
117 retries = 64
118 while retries > 0:
119 new_resource = generate_one_fn()
120 if new_resource is not None and new_resource not in all_elems:
121 break
122 else:
123 raise errors.ConfigurationError("Not able generate new resource"
124 " (last tried: %s)" % new_resource)
125 self.Reserve(ec_id, new_resource)
126 return new_resource
127
130 """Wrapper around L{utils.text.MatchNameComponent}.
131
132 """
133 return utils.MatchNameComponent(short_name, names, case_sensitive=False)
134
137 """Checks if instance's disks' C{iv_name} attributes are in order.
138
139 @type disks: list of L{objects.Disk}
140 @param disks: List of disks
141 @rtype: list of tuples; (int, string, string)
142 @return: List of wrongly named disks, each tuple contains disk index,
143 expected and actual name
144
145 """
146 result = []
147
148 for (idx, disk) in enumerate(disks):
149 exp_iv_name = "disk/%s" % idx
150 if disk.iv_name != exp_iv_name:
151 result.append((idx, exp_iv_name, disk.iv_name))
152
153 return result
154
157 """The interface to the cluster configuration.
158
159 @ivar _temporary_lvs: reservation manager for temporary LVs
160 @ivar _all_rms: a list of all temporary reservation managers
161
162 """
190
192 """Returns RPC runner for configuration.
193
194 """
195 return rpc.ConfigRunner(self._context, address_list)
196
197 - def SetContext(self, context):
198 """Sets Ganeti context.
199
200 """
201 self._context = context
202
203
204 @staticmethod
210
212 """Generate one mac address
213
214 """
215 prefix = self._config_data.cluster.mac_prefix
216 byte1 = random.randrange(0, 256)
217 byte2 = random.randrange(0, 256)
218 byte3 = random.randrange(0, 256)
219 mac = "%s:%02x:%02x:%02x" % (prefix, byte1, byte2, byte3)
220 return mac
221
222 @locking.ssynchronized(_config_lock, shared=1)
224 """Get the node params populated with cluster defaults.
225
226 @type node: L{objects.Node}
227 @param node: The node we want to know the params for
228 @return: A dict with the filled in node params
229
230 """
231 nodegroup = self._UnlockedGetNodeGroup(node.group)
232 return self._config_data.cluster.FillND(node, nodegroup)
233
234 @locking.ssynchronized(_config_lock, shared=1)
236 """Get the disk params populated with inherit chain.
237
238 @type instance: L{objects.Instance}
239 @param instance: The instance we want to know the params for
240 @return: A dict with the filled in disk params
241
242 """
243 node = self._UnlockedGetNodeInfo(instance.primary_node)
244 nodegroup = self._UnlockedGetNodeGroup(node.group)
245 return self._UnlockedGetGroupDiskParams(nodegroup)
246
247 @locking.ssynchronized(_config_lock, shared=1)
249 """Get the disk params populated with inherit chain.
250
251 @type group: L{objects.NodeGroup}
252 @param group: The group we want to know the params for
253 @return: A dict with the filled in disk params
254
255 """
256 return self._UnlockedGetGroupDiskParams(group)
257
259 """Get the disk params populated with inherit chain down to node-group.
260
261 @type group: L{objects.NodeGroup}
262 @param group: The group we want to know the params for
263 @return: A dict with the filled in disk params
264
265 """
266 return self._config_data.cluster.SimpleFillDP(group.diskparams)
267
268 @locking.ssynchronized(_config_lock, shared=1)
270 """Generate a MAC for an instance.
271
272 This should check the current instances for duplicates.
273
274 """
275 existing = self._AllMACs()
276 return self._temporary_ids.Generate(existing, self._GenerateOneMAC, ec_id)
277
278 @locking.ssynchronized(_config_lock, shared=1)
280 """Reserve a MAC for an instance.
281
282 This only checks instances managed by this cluster, it does not
283 check for potential collisions elsewhere.
284
285 """
286 all_macs = self._AllMACs()
287 if mac in all_macs:
288 raise errors.ReservationError("mac already in use")
289 else:
290 self._temporary_macs.Reserve(ec_id, mac)
291
292 @locking.ssynchronized(_config_lock, shared=1)
294 """Reserve an VG/LV pair for an instance.
295
296 @type lv_name: string
297 @param lv_name: the logical volume name to reserve
298
299 """
300 all_lvs = self._AllLVs()
301 if lv_name in all_lvs:
302 raise errors.ReservationError("LV already in use")
303 else:
304 self._temporary_lvs.Reserve(ec_id, lv_name)
305
306 @locking.ssynchronized(_config_lock, shared=1)
316
318 """Compute the list of all LVs.
319
320 """
321 lvnames = set()
322 for instance in self._config_data.instances.values():
323 node_data = instance.MapLVsByNode()
324 for lv_list in node_data.values():
325 lvnames.update(lv_list)
326 return lvnames
327
328 - def _AllIDs(self, include_temporary):
329 """Compute the list of all UUIDs and names we have.
330
331 @type include_temporary: boolean
332 @param include_temporary: whether to include the _temporary_ids set
333 @rtype: set
334 @return: a set of IDs
335
336 """
337 existing = set()
338 if include_temporary:
339 existing.update(self._temporary_ids.GetReserved())
340 existing.update(self._AllLVs())
341 existing.update(self._config_data.instances.keys())
342 existing.update(self._config_data.nodes.keys())
343 existing.update([i.uuid for i in self._AllUUIDObjects() if i.uuid])
344 return existing
345
347 """Generate an unique UUID.
348
349 This checks the current node, instances and disk names for
350 duplicates.
351
352 @rtype: string
353 @return: the unique id
354
355 """
356 existing = self._AllIDs(include_temporary=False)
357 return self._temporary_ids.Generate(existing, utils.NewUUID, ec_id)
358
359 @locking.ssynchronized(_config_lock, shared=1)
361 """Generate an unique ID.
362
363 This is just a wrapper over the unlocked version.
364
365 @type ec_id: string
366 @param ec_id: unique id for the job to reserve the id to
367
368 """
369 return self._GenerateUniqueID(ec_id)
370
372 """Return all MACs present in the config.
373
374 @rtype: list
375 @return: the list of all MACs
376
377 """
378 result = []
379 for instance in self._config_data.instances.values():
380 for nic in instance.nics:
381 result.append(nic.mac)
382
383 return result
384
386 """Return all DRBD secrets present in the config.
387
388 @rtype: list
389 @return: the list of all DRBD secrets
390
391 """
392 def helper(disk, result):
393 """Recursively gather secrets from this disk."""
394 if disk.dev_type == constants.DT_DRBD8:
395 result.append(disk.logical_id[5])
396 if disk.children:
397 for child in disk.children:
398 helper(child, result)
399
400 result = []
401 for instance in self._config_data.instances.values():
402 for disk in instance.disks:
403 helper(disk, result)
404
405 return result
406
408 """Compute duplicate disk IDs
409
410 @type disk: L{objects.Disk}
411 @param disk: the disk at which to start searching
412 @type l_ids: list
413 @param l_ids: list of current logical ids
414 @type p_ids: list
415 @param p_ids: list of current physical ids
416 @rtype: list
417 @return: a list of error messages
418
419 """
420 result = []
421 if disk.logical_id is not None:
422 if disk.logical_id in l_ids:
423 result.append("duplicate logical id %s" % str(disk.logical_id))
424 else:
425 l_ids.append(disk.logical_id)
426 if disk.physical_id is not None:
427 if disk.physical_id in p_ids:
428 result.append("duplicate physical id %s" % str(disk.physical_id))
429 else:
430 p_ids.append(disk.physical_id)
431
432 if disk.children:
433 for child in disk.children:
434 result.extend(self._CheckDiskIDs(child, l_ids, p_ids))
435 return result
436
438 """Verify function.
439
440 @rtype: list
441 @return: a list of error messages; a non-empty list signifies
442 configuration errors
443
444 """
445
446 result = []
447 seen_macs = []
448 ports = {}
449 data = self._config_data
450 cluster = data.cluster
451 seen_lids = []
452 seen_pids = []
453
454
455 if not cluster.enabled_hypervisors:
456 result.append("enabled hypervisors list doesn't have any entries")
457 invalid_hvs = set(cluster.enabled_hypervisors) - constants.HYPER_TYPES
458 if invalid_hvs:
459 result.append("enabled hypervisors contains invalid entries: %s" %
460 invalid_hvs)
461 missing_hvp = (set(cluster.enabled_hypervisors) -
462 set(cluster.hvparams.keys()))
463 if missing_hvp:
464 result.append("hypervisor parameters missing for the enabled"
465 " hypervisor(s) %s" % utils.CommaJoin(missing_hvp))
466
467 if cluster.master_node not in data.nodes:
468 result.append("cluster has invalid primary node '%s'" %
469 cluster.master_node)
470
471 def _helper(owner, attr, value, template):
472 try:
473 utils.ForceDictType(value, template)
474 except errors.GenericError, err:
475 result.append("%s has invalid %s: %s" % (owner, attr, err))
476
477 def _helper_nic(owner, params):
478 try:
479 objects.NIC.CheckParameterSyntax(params)
480 except errors.ConfigurationError, err:
481 result.append("%s has invalid nicparams: %s" % (owner, err))
482
483 def _helper_ipolicy(owner, params, check_std):
484 try:
485 objects.InstancePolicy.CheckParameterSyntax(params, check_std)
486 except errors.ConfigurationError, err:
487 result.append("%s has invalid instance policy: %s" % (owner, err))
488
489 def _helper_ispecs(owner, params):
490 for key, value in params.items():
491 if key in constants.IPOLICY_ISPECS:
492 fullkey = "ipolicy/" + key
493 _helper(owner, fullkey, value, constants.ISPECS_PARAMETER_TYPES)
494 else:
495
496 if key in constants.IPOLICY_PARAMETERS:
497 exp_type = float
498 else:
499 exp_type = list
500 if not isinstance(value, exp_type):
501 result.append("%s has invalid instance policy: for %s,"
502 " expecting %s, got %s" %
503 (owner, key, exp_type.__name__, type(value)))
504
505
506 _helper("cluster", "beparams", cluster.SimpleFillBE({}),
507 constants.BES_PARAMETER_TYPES)
508 _helper("cluster", "nicparams", cluster.SimpleFillNIC({}),
509 constants.NICS_PARAMETER_TYPES)
510 _helper_nic("cluster", cluster.SimpleFillNIC({}))
511 _helper("cluster", "ndparams", cluster.SimpleFillND({}),
512 constants.NDS_PARAMETER_TYPES)
513 _helper_ipolicy("cluster", cluster.SimpleFillIPolicy({}), True)
514 _helper_ispecs("cluster", cluster.SimpleFillIPolicy({}))
515
516
517 for instance_name in data.instances:
518 instance = data.instances[instance_name]
519 if instance.name != instance_name:
520 result.append("instance '%s' is indexed by wrong name '%s'" %
521 (instance.name, instance_name))
522 if instance.primary_node not in data.nodes:
523 result.append("instance '%s' has invalid primary node '%s'" %
524 (instance_name, instance.primary_node))
525 for snode in instance.secondary_nodes:
526 if snode not in data.nodes:
527 result.append("instance '%s' has invalid secondary node '%s'" %
528 (instance_name, snode))
529 for idx, nic in enumerate(instance.nics):
530 if nic.mac in seen_macs:
531 result.append("instance '%s' has NIC %d mac %s duplicate" %
532 (instance_name, idx, nic.mac))
533 else:
534 seen_macs.append(nic.mac)
535 if nic.nicparams:
536 filled = cluster.SimpleFillNIC(nic.nicparams)
537 owner = "instance %s nic %d" % (instance.name, idx)
538 _helper(owner, "nicparams",
539 filled, constants.NICS_PARAMETER_TYPES)
540 _helper_nic(owner, filled)
541
542
543 if instance.beparams:
544 _helper("instance %s" % instance.name, "beparams",
545 cluster.FillBE(instance), constants.BES_PARAMETER_TYPES)
546
547
548 for (idx, dsk) in enumerate(instance.disks):
549 if dsk.dev_type in constants.LDS_DRBD:
550 tcp_port = dsk.logical_id[2]
551 if tcp_port not in ports:
552 ports[tcp_port] = []
553 ports[tcp_port].append((instance.name, "drbd disk %s" % idx))
554
555 net_port = getattr(instance, "network_port", None)
556 if net_port is not None:
557 if net_port not in ports:
558 ports[net_port] = []
559 ports[net_port].append((instance.name, "network port"))
560
561
562 for idx, disk in enumerate(instance.disks):
563 result.extend(["instance '%s' disk %d error: %s" %
564 (instance.name, idx, msg) for msg in disk.Verify()])
565 result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
566
567 wrong_names = _CheckInstanceDiskIvNames(instance.disks)
568 if wrong_names:
569 tmp = "; ".join(("name of disk %s should be '%s', but is '%s'" %
570 (idx, exp_name, actual_name))
571 for (idx, exp_name, actual_name) in wrong_names)
572
573 result.append("Instance '%s' has wrongly named disks: %s" %
574 (instance.name, tmp))
575
576
577 for free_port in cluster.tcpudp_port_pool:
578 if free_port not in ports:
579 ports[free_port] = []
580 ports[free_port].append(("cluster", "port marked as free"))
581
582
583 keys = ports.keys()
584 keys.sort()
585 for pnum in keys:
586 pdata = ports[pnum]
587 if len(pdata) > 1:
588 txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
589 result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
590
591
592 if keys:
593 if keys[-1] > cluster.highest_used_port:
594 result.append("Highest used port mismatch, saved %s, computed %s" %
595 (cluster.highest_used_port, keys[-1]))
596
597 if not data.nodes[cluster.master_node].master_candidate:
598 result.append("Master node is not a master candidate")
599
600
601 mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
602 if mc_now < mc_max:
603 result.append("Not enough master candidates: actual %d, target %d" %
604 (mc_now, mc_max))
605
606
607 for node_name, node in data.nodes.items():
608 if node.name != node_name:
609 result.append("Node '%s' is indexed by wrong name '%s'" %
610 (node.name, node_name))
611 if [node.master_candidate, node.drained, node.offline].count(True) > 1:
612 result.append("Node %s state is invalid: master_candidate=%s,"
613 " drain=%s, offline=%s" %
614 (node.name, node.master_candidate, node.drained,
615 node.offline))
616 if node.group not in data.nodegroups:
617 result.append("Node '%s' has invalid group '%s'" %
618 (node.name, node.group))
619 else:
620 _helper("node %s" % node.name, "ndparams",
621 cluster.FillND(node, data.nodegroups[node.group]),
622 constants.NDS_PARAMETER_TYPES)
623
624
625 nodegroups_names = set()
626 for nodegroup_uuid in data.nodegroups:
627 nodegroup = data.nodegroups[nodegroup_uuid]
628 if nodegroup.uuid != nodegroup_uuid:
629 result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'"
630 % (nodegroup.name, nodegroup.uuid, nodegroup_uuid))
631 if utils.UUID_RE.match(nodegroup.name.lower()):
632 result.append("node group '%s' (uuid: '%s') has uuid-like name" %
633 (nodegroup.name, nodegroup.uuid))
634 if nodegroup.name in nodegroups_names:
635 result.append("duplicate node group name '%s'" % nodegroup.name)
636 else:
637 nodegroups_names.add(nodegroup.name)
638 group_name = "group %s" % nodegroup.name
639 _helper_ipolicy(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy),
640 False)
641 _helper_ispecs(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy))
642 if nodegroup.ndparams:
643 _helper(group_name, "ndparams",
644 cluster.SimpleFillND(nodegroup.ndparams),
645 constants.NDS_PARAMETER_TYPES)
646
647
648 _, duplicates = self._UnlockedComputeDRBDMap()
649 for node, minor, instance_a, instance_b in duplicates:
650 result.append("DRBD minor %d on node %s is assigned twice to instances"
651 " %s and %s" % (minor, node, instance_a, instance_b))
652
653
654 default_nicparams = cluster.nicparams[constants.PP_DEFAULT]
655 ips = {}
656
657 def _AddIpAddress(ip, name):
658 ips.setdefault(ip, []).append(name)
659
660 _AddIpAddress(cluster.master_ip, "cluster_ip")
661
662 for node in data.nodes.values():
663 _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
664 if node.secondary_ip != node.primary_ip:
665 _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
666
667 for instance in data.instances.values():
668 for idx, nic in enumerate(instance.nics):
669 if nic.ip is None:
670 continue
671
672 nicparams = objects.FillDict(default_nicparams, nic.nicparams)
673 nic_mode = nicparams[constants.NIC_MODE]
674 nic_link = nicparams[constants.NIC_LINK]
675
676 if nic_mode == constants.NIC_MODE_BRIDGED:
677 link = "bridge:%s" % nic_link
678 elif nic_mode == constants.NIC_MODE_ROUTED:
679 link = "route:%s" % nic_link
680 else:
681 raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
682
683 _AddIpAddress("%s/%s" % (link, nic.ip),
684 "instance:%s/nic:%d" % (instance.name, idx))
685
686 for ip, owners in ips.items():
687 if len(owners) > 1:
688 result.append("IP address %s is used by multiple owners: %s" %
689 (ip, utils.CommaJoin(owners)))
690
691 return result
692
693 @locking.ssynchronized(_config_lock, shared=1)
695 """Verify function.
696
697 This is just a wrapper over L{_UnlockedVerifyConfig}.
698
699 @rtype: list
700 @return: a list of error messages; a non-empty list signifies
701 configuration errors
702
703 """
704 return self._UnlockedVerifyConfig()
705
707 """Convert the unique ID to the ID needed on the target nodes.
708
709 This is used only for drbd, which needs ip/port configuration.
710
711 The routine descends down and updates its children also, because
712 this helps when the only the top device is passed to the remote
713 node.
714
715 This function is for internal use, when the config lock is already held.
716
717 """
718 if disk.children:
719 for child in disk.children:
720 self._UnlockedSetDiskID(child, node_name)
721
722 if disk.logical_id is None and disk.physical_id is not None:
723 return
724 if disk.dev_type == constants.LD_DRBD8:
725 pnode, snode, port, pminor, sminor, secret = disk.logical_id
726 if node_name not in (pnode, snode):
727 raise errors.ConfigurationError("DRBD device not knowing node %s" %
728 node_name)
729 pnode_info = self._UnlockedGetNodeInfo(pnode)
730 snode_info = self._UnlockedGetNodeInfo(snode)
731 if pnode_info is None or snode_info is None:
732 raise errors.ConfigurationError("Can't find primary or secondary node"
733 " for %s" % str(disk))
734 p_data = (pnode_info.secondary_ip, port)
735 s_data = (snode_info.secondary_ip, port)
736 if pnode == node_name:
737 disk.physical_id = p_data + s_data + (pminor, secret)
738 else:
739 disk.physical_id = s_data + p_data + (sminor, secret)
740 else:
741 disk.physical_id = disk.logical_id
742 return
743
744 @locking.ssynchronized(_config_lock)
746 """Convert the unique ID to the ID needed on the target nodes.
747
748 This is used only for drbd, which needs ip/port configuration.
749
750 The routine descends down and updates its children also, because
751 this helps when the only the top device is passed to the remote
752 node.
753
754 """
755 return self._UnlockedSetDiskID(disk, node_name)
756
757 @locking.ssynchronized(_config_lock)
759 """Adds a new port to the available port pool.
760
761 @warning: this method does not "flush" the configuration (via
762 L{_WriteConfig}); callers should do that themselves once the
763 configuration is stable
764
765 """
766 if not isinstance(port, int):
767 raise errors.ProgrammerError("Invalid type passed for port")
768
769 self._config_data.cluster.tcpudp_port_pool.add(port)
770
771 @locking.ssynchronized(_config_lock, shared=1)
773 """Returns a copy of the current port list.
774
775 """
776 return self._config_data.cluster.tcpudp_port_pool.copy()
777
778 @locking.ssynchronized(_config_lock)
780 """Allocate a port.
781
782 The port will be taken from the available port pool or from the
783 default port range (and in this case we increase
784 highest_used_port).
785
786 """
787
788 if self._config_data.cluster.tcpudp_port_pool:
789 port = self._config_data.cluster.tcpudp_port_pool.pop()
790 else:
791 port = self._config_data.cluster.highest_used_port + 1
792 if port >= constants.LAST_DRBD_PORT:
793 raise errors.ConfigurationError("The highest used port is greater"
794 " than %s. Aborting." %
795 constants.LAST_DRBD_PORT)
796 self._config_data.cluster.highest_used_port = port
797
798 self._WriteConfig()
799 return port
800
802 """Compute the used DRBD minor/nodes.
803
804 @rtype: (dict, list)
805 @return: dictionary of node_name: dict of minor: instance_name;
806 the returned dict will have all the nodes in it (even if with
807 an empty list), and a list of duplicates; if the duplicates
808 list is not empty, the configuration is corrupted and its caller
809 should raise an exception
810
811 """
812 def _AppendUsedPorts(instance_name, disk, used):
813 duplicates = []
814 if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:
815 node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
816 for node, port in ((node_a, minor_a), (node_b, minor_b)):
817 assert node in used, ("Node '%s' of instance '%s' not found"
818 " in node list" % (node, instance_name))
819 if port in used[node]:
820 duplicates.append((node, port, instance_name, used[node][port]))
821 else:
822 used[node][port] = instance_name
823 if disk.children:
824 for child in disk.children:
825 duplicates.extend(_AppendUsedPorts(instance_name, child, used))
826 return duplicates
827
828 duplicates = []
829 my_dict = dict((node, {}) for node in self._config_data.nodes)
830 for instance in self._config_data.instances.itervalues():
831 for disk in instance.disks:
832 duplicates.extend(_AppendUsedPorts(instance.name, disk, my_dict))
833 for (node, minor), instance in self._temporary_drbds.iteritems():
834 if minor in my_dict[node] and my_dict[node][minor] != instance:
835 duplicates.append((node, minor, instance, my_dict[node][minor]))
836 else:
837 my_dict[node][minor] = instance
838 return my_dict, duplicates
839
840 @locking.ssynchronized(_config_lock)
842 """Compute the used DRBD minor/nodes.
843
844 This is just a wrapper over L{_UnlockedComputeDRBDMap}.
845
846 @return: dictionary of node_name: dict of minor: instance_name;
847 the returned dict will have all the nodes in it (even if with
848 an empty list).
849
850 """
851 d_map, duplicates = self._UnlockedComputeDRBDMap()
852 if duplicates:
853 raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
854 str(duplicates))
855 return d_map
856
857 @locking.ssynchronized(_config_lock)
859 """Allocate a drbd minor.
860
861 The free minor will be automatically computed from the existing
862 devices. A node can be given multiple times in order to allocate
863 multiple minors. The result is the list of minors, in the same
864 order as the passed nodes.
865
866 @type instance: string
867 @param instance: the instance for which we allocate minors
868
869 """
870 assert isinstance(instance, basestring), \
871 "Invalid argument '%s' passed to AllocateDRBDMinor" % instance
872
873 d_map, duplicates = self._UnlockedComputeDRBDMap()
874 if duplicates:
875 raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
876 str(duplicates))
877 result = []
878 for nname in nodes:
879 ndata = d_map[nname]
880 if not ndata:
881
882 result.append(0)
883 ndata[0] = instance
884 self._temporary_drbds[(nname, 0)] = instance
885 continue
886 keys = ndata.keys()
887 keys.sort()
888 ffree = utils.FirstFree(keys)
889 if ffree is None:
890
891
892 minor = keys[-1] + 1
893 else:
894 minor = ffree
895
896 assert minor not in d_map[nname], \
897 ("Attempt to reuse allocated DRBD minor %d on node %s,"
898 " already allocated to instance %s" %
899 (minor, nname, d_map[nname][minor]))
900 ndata[minor] = instance
901
902 r_key = (nname, minor)
903 assert r_key not in self._temporary_drbds, \
904 ("Attempt to reuse reserved DRBD minor %d on node %s,"
905 " reserved for instance %s" %
906 (minor, nname, self._temporary_drbds[r_key]))
907 self._temporary_drbds[r_key] = instance
908 result.append(minor)
909 logging.debug("Request to allocate drbd minors, input: %s, returning %s",
910 nodes, result)
911 return result
912
914 """Release temporary drbd minors allocated for a given instance.
915
916 @type instance: string
917 @param instance: the instance for which temporary minors should be
918 released
919
920 """
921 assert isinstance(instance, basestring), \
922 "Invalid argument passed to ReleaseDRBDMinors"
923 for key, name in self._temporary_drbds.items():
924 if name == instance:
925 del self._temporary_drbds[key]
926
927 @locking.ssynchronized(_config_lock)
929 """Release temporary drbd minors allocated for a given instance.
930
931 This should be called on the error paths, on the success paths
932 it's automatically called by the ConfigWriter add and update
933 functions.
934
935 This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
936
937 @type instance: string
938 @param instance: the instance for which temporary minors should be
939 released
940
941 """
942 self._UnlockedReleaseDRBDMinors(instance)
943
944 @locking.ssynchronized(_config_lock, shared=1)
946 """Get the configuration version.
947
948 @return: Config version
949
950 """
951 return self._config_data.version
952
953 @locking.ssynchronized(_config_lock, shared=1)
955 """Get cluster name.
956
957 @return: Cluster name
958
959 """
960 return self._config_data.cluster.cluster_name
961
962 @locking.ssynchronized(_config_lock, shared=1)
964 """Get the hostname of the master node for this cluster.
965
966 @return: Master hostname
967
968 """
969 return self._config_data.cluster.master_node
970
971 @locking.ssynchronized(_config_lock, shared=1)
973 """Get the IP of the master node for this cluster.
974
975 @return: Master IP
976
977 """
978 return self._config_data.cluster.master_ip
979
980 @locking.ssynchronized(_config_lock, shared=1)
982 """Get the master network device for this cluster.
983
984 """
985 return self._config_data.cluster.master_netdev
986
987 @locking.ssynchronized(_config_lock, shared=1)
989 """Get the netmask of the master node for this cluster.
990
991 """
992 return self._config_data.cluster.master_netmask
993
994 @locking.ssynchronized(_config_lock, shared=1)
996 """Get flag representing whether to use the external master IP setup script.
997
998 """
999 return self._config_data.cluster.use_external_mip_script
1000
1001 @locking.ssynchronized(_config_lock, shared=1)
1003 """Get the file storage dir for this cluster.
1004
1005 """
1006 return self._config_data.cluster.file_storage_dir
1007
1008 @locking.ssynchronized(_config_lock, shared=1)
1010 """Get the shared file storage dir for this cluster.
1011
1012 """
1013 return self._config_data.cluster.shared_file_storage_dir
1014
1015 @locking.ssynchronized(_config_lock, shared=1)
1017 """Get the hypervisor type for this cluster.
1018
1019 """
1020 return self._config_data.cluster.enabled_hypervisors[0]
1021
1022 @locking.ssynchronized(_config_lock, shared=1)
1024 """Return the rsa hostkey from the config.
1025
1026 @rtype: string
1027 @return: the rsa hostkey
1028
1029 """
1030 return self._config_data.cluster.rsahostkeypub
1031
1032 @locking.ssynchronized(_config_lock, shared=1)
1034 """Get the default instance allocator for this cluster.
1035
1036 """
1037 return self._config_data.cluster.default_iallocator
1038
1039 @locking.ssynchronized(_config_lock, shared=1)
1041 """Get cluster primary ip family.
1042
1043 @return: primary ip family
1044
1045 """
1046 return self._config_data.cluster.primary_ip_family
1047
1048 @locking.ssynchronized(_config_lock, shared=1)
1050 """Get network parameters of the master node.
1051
1052 @rtype: L{object.MasterNetworkParameters}
1053 @return: network parameters of the master node
1054
1055 """
1056 cluster = self._config_data.cluster
1057 result = objects.MasterNetworkParameters(name=cluster.master_node,
1058 ip=cluster.master_ip,
1059 netmask=cluster.master_netmask,
1060 netdev=cluster.master_netdev,
1061 ip_family=cluster.primary_ip_family)
1062
1063 return result
1064
1065 @locking.ssynchronized(_config_lock)
1067 """Add a node group to the configuration.
1068
1069 This method calls group.UpgradeConfig() to fill any missing attributes
1070 according to their default values.
1071
1072 @type group: L{objects.NodeGroup}
1073 @param group: the NodeGroup object to add
1074 @type ec_id: string
1075 @param ec_id: unique id for the job to use when creating a missing UUID
1076 @type check_uuid: bool
1077 @param check_uuid: add an UUID to the group if it doesn't have one or, if
1078 it does, ensure that it does not exist in the
1079 configuration already
1080
1081 """
1082 self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1083 self._WriteConfig()
1084
1086 """Add a node group to the configuration.
1087
1088 """
1089 logging.info("Adding node group %s to configuration", group.name)
1090
1091
1092
1093
1094 if check_uuid:
1095 self._EnsureUUID(group, ec_id)
1096
1097 try:
1098 existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1099 except errors.OpPrereqError:
1100 pass
1101 else:
1102 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1103 " node group (UUID: %s)" %
1104 (group.name, existing_uuid),
1105 errors.ECODE_EXISTS)
1106
1107 group.serial_no = 1
1108 group.ctime = group.mtime = time.time()
1109 group.UpgradeConfig()
1110
1111 self._config_data.nodegroups[group.uuid] = group
1112 self._config_data.cluster.serial_no += 1
1113
1114 @locking.ssynchronized(_config_lock)
1116 """Remove a node group from the configuration.
1117
1118 @type group_uuid: string
1119 @param group_uuid: the UUID of the node group to remove
1120
1121 """
1122 logging.info("Removing node group %s from configuration", group_uuid)
1123
1124 if group_uuid not in self._config_data.nodegroups:
1125 raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1126
1127 assert len(self._config_data.nodegroups) != 1, \
1128 "Group '%s' is the only group, cannot be removed" % group_uuid
1129
1130 del self._config_data.nodegroups[group_uuid]
1131 self._config_data.cluster.serial_no += 1
1132 self._WriteConfig()
1133
1135 """Lookup a node group's UUID.
1136
1137 @type target: string or None
1138 @param target: group name or UUID or None to look for the default
1139 @rtype: string
1140 @return: nodegroup UUID
1141 @raises errors.OpPrereqError: when the target group cannot be found
1142
1143 """
1144 if target is None:
1145 if len(self._config_data.nodegroups) != 1:
1146 raise errors.OpPrereqError("More than one node group exists. Target"
1147 " group must be specified explicitly.")
1148 else:
1149 return self._config_data.nodegroups.keys()[0]
1150 if target in self._config_data.nodegroups:
1151 return target
1152 for nodegroup in self._config_data.nodegroups.values():
1153 if nodegroup.name == target:
1154 return nodegroup.uuid
1155 raise errors.OpPrereqError("Node group '%s' not found" % target,
1156 errors.ECODE_NOENT)
1157
1158 @locking.ssynchronized(_config_lock, shared=1)
1160 """Lookup a node group's UUID.
1161
1162 This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1163
1164 @type target: string or None
1165 @param target: group name or UUID or None to look for the default
1166 @rtype: string
1167 @return: nodegroup UUID
1168
1169 """
1170 return self._UnlockedLookupNodeGroup(target)
1171
1173 """Lookup a node group.
1174
1175 @type uuid: string
1176 @param uuid: group UUID
1177 @rtype: L{objects.NodeGroup} or None
1178 @return: nodegroup object, or None if not found
1179
1180 """
1181 if uuid not in self._config_data.nodegroups:
1182 return None
1183
1184 return self._config_data.nodegroups[uuid]
1185
1186 @locking.ssynchronized(_config_lock, shared=1)
1188 """Lookup a node group.
1189
1190 @type uuid: string
1191 @param uuid: group UUID
1192 @rtype: L{objects.NodeGroup} or None
1193 @return: nodegroup object, or None if not found
1194
1195 """
1196 return self._UnlockedGetNodeGroup(uuid)
1197
1198 @locking.ssynchronized(_config_lock, shared=1)
1200 """Get the configuration of all node groups.
1201
1202 """
1203 return dict(self._config_data.nodegroups)
1204
1205 @locking.ssynchronized(_config_lock, shared=1)
1207 """Get a list of node groups.
1208
1209 """
1210 return self._config_data.nodegroups.keys()
1211
1212 @locking.ssynchronized(_config_lock, shared=1)
1214 """Get nodes which are member in the same nodegroups as the given nodes.
1215
1216 """
1217 ngfn = lambda node_name: self._UnlockedGetNodeInfo(node_name).group
1218 return frozenset(member_name
1219 for node_name in nodes
1220 for member_name in
1221 self._UnlockedGetNodeGroup(ngfn(node_name)).members)
1222
1223 @locking.ssynchronized(_config_lock, shared=1)
1225 """Get the configuration of multiple node groups.
1226
1227 @param group_uuids: List of node group UUIDs
1228 @rtype: list
1229 @return: List of tuples of (group_uuid, group_info)
1230
1231 """
1232 return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1233
1234 @locking.ssynchronized(_config_lock)
1266
1268 """Ensures a given object has a valid UUID.
1269
1270 @param item: the instance or node to be checked
1271 @param ec_id: the execution context id for the uuid reservation
1272
1273 """
1274 if not item.uuid:
1275 item.uuid = self._GenerateUniqueID(ec_id)
1276 elif item.uuid in self._AllIDs(include_temporary=True):
1277 raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1278 " in use" % (item.name, item.uuid))
1279
1281 """Set the instance's status to a given value.
1282
1283 """
1284 assert status in constants.ADMINST_ALL, \
1285 "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1286
1287 if instance_name not in self._config_data.instances:
1288 raise errors.ConfigurationError("Unknown instance '%s'" %
1289 instance_name)
1290 instance = self._config_data.instances[instance_name]
1291 if instance.admin_state != status:
1292 instance.admin_state = status
1293 instance.serial_no += 1
1294 instance.mtime = time.time()
1295 self._WriteConfig()
1296
1297 @locking.ssynchronized(_config_lock)
1303
1304 @locking.ssynchronized(_config_lock)
1310
1311 @locking.ssynchronized(_config_lock)
1313 """Remove the instance from the configuration.
1314
1315 """
1316 if instance_name not in self._config_data.instances:
1317 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1318
1319
1320
1321 inst = self._config_data.instances[instance_name]
1322 network_port = getattr(inst, "network_port", None)
1323 if network_port is not None:
1324 self._config_data.cluster.tcpudp_port_pool.add(network_port)
1325
1326 del self._config_data.instances[instance_name]
1327 self._config_data.cluster.serial_no += 1
1328 self._WriteConfig()
1329
1330 @locking.ssynchronized(_config_lock)
1332 """Rename an instance.
1333
1334 This needs to be done in ConfigWriter and not by RemoveInstance
1335 combined with AddInstance as only we can guarantee an atomic
1336 rename.
1337
1338 """
1339 if old_name not in self._config_data.instances:
1340 raise errors.ConfigurationError("Unknown instance '%s'" % old_name)
1341
1342
1343 inst = self._config_data.instances[old_name].Copy()
1344 inst.name = new_name
1345
1346 for (idx, disk) in enumerate(inst.disks):
1347 if disk.dev_type == constants.LD_FILE:
1348
1349 file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1350 disk.logical_id = (disk.logical_id[0],
1351 utils.PathJoin(file_storage_dir, inst.name,
1352 "disk%s" % idx))
1353 disk.physical_id = disk.logical_id
1354
1355
1356 del self._config_data.instances[old_name]
1357 self._config_data.instances[inst.name] = inst
1358
1359
1360 self._config_data.cluster.serial_no += 1
1361
1362 self._WriteConfig()
1363
1364 @locking.ssynchronized(_config_lock)
1370
1372 """Get the list of instances.
1373
1374 This function is for internal use, when the config lock is already held.
1375
1376 """
1377 return self._config_data.instances.keys()
1378
1379 @locking.ssynchronized(_config_lock, shared=1)
1381 """Get the list of instances.
1382
1383 @return: array of instances, ex. ['instance2.example.com',
1384 'instance1.example.com']
1385
1386 """
1387 return self._UnlockedGetInstanceList()
1388
1395
1397 """Returns information about an instance.
1398
1399 This function is for internal use, when the config lock is already held.
1400
1401 """
1402 if instance_name not in self._config_data.instances:
1403 return None
1404
1405 return self._config_data.instances[instance_name]
1406
1407 @locking.ssynchronized(_config_lock, shared=1)
1409 """Returns information about an instance.
1410
1411 It takes the information from the configuration file. Other information of
1412 an instance are taken from the live systems.
1413
1414 @param instance_name: name of the instance, e.g.
1415 I{instance1.example.com}
1416
1417 @rtype: L{objects.Instance}
1418 @return: the instance object
1419
1420 """
1421 return self._UnlockedGetInstanceInfo(instance_name)
1422
1423 @locking.ssynchronized(_config_lock, shared=1)
1441
1442 @locking.ssynchronized(_config_lock, shared=1)
1444 """Get the configuration of multiple instances.
1445
1446 @param instances: list of instance names
1447 @rtype: list
1448 @return: list of tuples (instance, instance_info), where
1449 instance_info is what would GetInstanceInfo return for the
1450 node, while keeping the original order
1451
1452 """
1453 return [(name, self._UnlockedGetInstanceInfo(name)) for name in instances]
1454
1455 @locking.ssynchronized(_config_lock, shared=1)
1467
1468 @locking.ssynchronized(_config_lock, shared=1)
1470 """Get instance configuration with a filter.
1471
1472 @type filter_fn: callable
1473 @param filter_fn: Filter function receiving instance object as parameter,
1474 returning boolean. Important: this function is called while the
1475 configuration locks is held. It must not do any complex work or call
1476 functions potentially leading to a deadlock. Ideally it doesn't call any
1477 other functions and just compares instance attributes.
1478
1479 """
1480 return dict((name, inst)
1481 for (name, inst) in self._config_data.instances.items()
1482 if filter_fn(inst))
1483
1484 @locking.ssynchronized(_config_lock)
1486 """Add a node to the configuration.
1487
1488 @type node: L{objects.Node}
1489 @param node: a Node instance
1490
1491 """
1492 logging.info("Adding node %s to configuration", node.name)
1493
1494 self._EnsureUUID(node, ec_id)
1495
1496 node.serial_no = 1
1497 node.ctime = node.mtime = time.time()
1498 self._UnlockedAddNodeToGroup(node.name, node.group)
1499 self._config_data.nodes[node.name] = node
1500 self._config_data.cluster.serial_no += 1
1501 self._WriteConfig()
1502
1503 @locking.ssynchronized(_config_lock)
1505 """Remove a node from the configuration.
1506
1507 """
1508 logging.info("Removing node %s from configuration", node_name)
1509
1510 if node_name not in self._config_data.nodes:
1511 raise errors.ConfigurationError("Unknown node '%s'" % node_name)
1512
1513 self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_name])
1514 del self._config_data.nodes[node_name]
1515 self._config_data.cluster.serial_no += 1
1516 self._WriteConfig()
1517
1524
1526 """Get the configuration of a node, as stored in the config.
1527
1528 This function is for internal use, when the config lock is already
1529 held.
1530
1531 @param node_name: the node name, e.g. I{node1.example.com}
1532
1533 @rtype: L{objects.Node}
1534 @return: the node object
1535
1536 """
1537 if node_name not in self._config_data.nodes:
1538 return None
1539
1540 return self._config_data.nodes[node_name]
1541
1542 @locking.ssynchronized(_config_lock, shared=1)
1544 """Get the configuration of a node, as stored in the config.
1545
1546 This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1547
1548 @param node_name: the node name, e.g. I{node1.example.com}
1549
1550 @rtype: L{objects.Node}
1551 @return: the node object
1552
1553 """
1554 return self._UnlockedGetNodeInfo(node_name)
1555
1556 @locking.ssynchronized(_config_lock, shared=1)
1558 """Get the instances of a node, as stored in the config.
1559
1560 @param node_name: the node name, e.g. I{node1.example.com}
1561
1562 @rtype: (list, list)
1563 @return: a tuple with two lists: the primary and the secondary instances
1564
1565 """
1566 pri = []
1567 sec = []
1568 for inst in self._config_data.instances.values():
1569 if inst.primary_node == node_name:
1570 pri.append(inst.name)
1571 if node_name in inst.secondary_nodes:
1572 sec.append(inst.name)
1573 return (pri, sec)
1574
1575 @locking.ssynchronized(_config_lock, shared=1)
1577 """Get the instances of a node group.
1578
1579 @param uuid: Node group UUID
1580 @param primary_only: Whether to only consider primary nodes
1581 @rtype: frozenset
1582 @return: List of instance names in node group
1583
1584 """
1585 if primary_only:
1586 nodes_fn = lambda inst: [inst.primary_node]
1587 else:
1588 nodes_fn = lambda inst: inst.all_nodes
1589
1590 return frozenset(inst.name
1591 for inst in self._config_data.instances.values()
1592 for node_name in nodes_fn(inst)
1593 if self._UnlockedGetNodeInfo(node_name).group == uuid)
1594
1596 """Return the list of nodes which are in the configuration.
1597
1598 This function is for internal use, when the config lock is already
1599 held.
1600
1601 @rtype: list
1602
1603 """
1604 return self._config_data.nodes.keys()
1605
1606 @locking.ssynchronized(_config_lock, shared=1)
1608 """Return the list of nodes which are in the configuration.
1609
1610 """
1611 return self._UnlockedGetNodeList()
1612
1620
1621 @locking.ssynchronized(_config_lock, shared=1)
1627
1628 @locking.ssynchronized(_config_lock, shared=1)
1636
1637 @locking.ssynchronized(_config_lock, shared=1)
1645
1646 @locking.ssynchronized(_config_lock, shared=1)
1648 """Get the configuration of multiple nodes.
1649
1650 @param nodes: list of node names
1651 @rtype: list
1652 @return: list of tuples of (node, node_info), where node_info is
1653 what would GetNodeInfo return for the node, in the original
1654 order
1655
1656 """
1657 return [(name, self._UnlockedGetNodeInfo(name)) for name in nodes]
1658
1659 @locking.ssynchronized(_config_lock, shared=1)
1661 """Get the configuration of all nodes.
1662
1663 @rtype: dict
1664 @return: dict of (node, node_info), where node_info is what
1665 would GetNodeInfo return for the node
1666
1667 """
1668 return self._UnlockedGetAllNodesInfo()
1669
1678
1679 @locking.ssynchronized(_config_lock, shared=1)
1681 """Returns groups for a list of nodes.
1682
1683 @type nodes: list of string
1684 @param nodes: List of node names
1685 @rtype: frozenset
1686
1687 """
1688 return frozenset(self._UnlockedGetNodeInfo(name).group for name in nodes)
1689
1691 """Get the number of current and maximum desired and possible candidates.
1692
1693 @type exceptions: list
1694 @param exceptions: if passed, list of nodes that should be ignored
1695 @rtype: tuple
1696 @return: tuple of (current, desired and possible, possible)
1697
1698 """
1699 mc_now = mc_should = mc_max = 0
1700 for node in self._config_data.nodes.values():
1701 if exceptions and node.name in exceptions:
1702 continue
1703 if not (node.offline or node.drained) and node.master_capable:
1704 mc_max += 1
1705 if node.master_candidate:
1706 mc_now += 1
1707 mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
1708 return (mc_now, mc_should, mc_max)
1709
1710 @locking.ssynchronized(_config_lock, shared=1)
1712 """Get the number of current and maximum possible candidates.
1713
1714 This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
1715
1716 @type exceptions: list
1717 @param exceptions: if passed, list of nodes that should be ignored
1718 @rtype: tuple
1719 @return: tuple of (current, max)
1720
1721 """
1722 return self._UnlockedGetMasterCandidateStats(exceptions)
1723
1724 @locking.ssynchronized(_config_lock)
1725 - def MaintainCandidatePool(self, exceptions):
1726 """Try to grow the candidate pool to the desired size.
1727
1728 @type exceptions: list
1729 @param exceptions: if passed, list of nodes that should be ignored
1730 @rtype: list
1731 @return: list with the adjusted nodes (L{objects.Node} instances)
1732
1733 """
1734 mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(exceptions)
1735 mod_list = []
1736 if mc_now < mc_max:
1737 node_list = self._config_data.nodes.keys()
1738 random.shuffle(node_list)
1739 for name in node_list:
1740 if mc_now >= mc_max:
1741 break
1742 node = self._config_data.nodes[name]
1743 if (node.master_candidate or node.offline or node.drained or
1744 node.name in exceptions or not node.master_capable):
1745 continue
1746 mod_list.append(node)
1747 node.master_candidate = True
1748 node.serial_no += 1
1749 mc_now += 1
1750 if mc_now != mc_max:
1751
1752 logging.warning("Warning: MaintainCandidatePool didn't manage to"
1753 " fill the candidate pool (%d/%d)", mc_now, mc_max)
1754 if mod_list:
1755 self._config_data.cluster.serial_no += 1
1756 self._WriteConfig()
1757
1758 return mod_list
1759
1761 """Add a given node to the specified group.
1762
1763 """
1764 if nodegroup_uuid not in self._config_data.nodegroups:
1765
1766
1767
1768
1769 raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
1770 if node_name not in self._config_data.nodegroups[nodegroup_uuid].members:
1771 self._config_data.nodegroups[nodegroup_uuid].members.append(node_name)
1772
1774 """Remove a given node from its group.
1775
1776 """
1777 nodegroup = node.group
1778 if nodegroup not in self._config_data.nodegroups:
1779 logging.warning("Warning: node '%s' has unknown node group '%s'"
1780 " (while being removed from it)", node.name, nodegroup)
1781 nodegroup_obj = self._config_data.nodegroups[nodegroup]
1782 if node.name not in nodegroup_obj.members:
1783 logging.warning("Warning: node '%s' not a member of its node group '%s'"
1784 " (while being removed from it)", node.name, nodegroup)
1785 else:
1786 nodegroup_obj.members.remove(node.name)
1787
1788 @locking.ssynchronized(_config_lock)
1790 """Changes the group of a number of nodes.
1791
1792 @type mods: list of tuples; (node name, new group UUID)
1793 @param mods: Node membership modifications
1794
1795 """
1796 groups = self._config_data.nodegroups
1797 nodes = self._config_data.nodes
1798
1799 resmod = []
1800
1801
1802 for (node_name, new_group_uuid) in mods:
1803 try:
1804 node = nodes[node_name]
1805 except KeyError:
1806 raise errors.ConfigurationError("Unable to find node '%s'" % node_name)
1807
1808 if node.group == new_group_uuid:
1809
1810 logging.debug("Node '%s' was assigned to its current group (%s)",
1811 node_name, node.group)
1812 continue
1813
1814
1815 try:
1816 old_group = groups[node.group]
1817 except KeyError:
1818 raise errors.ConfigurationError("Unable to find old group '%s'" %
1819 node.group)
1820
1821
1822 try:
1823 new_group = groups[new_group_uuid]
1824 except KeyError:
1825 raise errors.ConfigurationError("Unable to find new group '%s'" %
1826 new_group_uuid)
1827
1828 assert node.name in old_group.members, \
1829 ("Inconsistent configuration: node '%s' not listed in members for its"
1830 " old group '%s'" % (node.name, old_group.uuid))
1831 assert node.name not in new_group.members, \
1832 ("Inconsistent configuration: node '%s' already listed in members for"
1833 " its new group '%s'" % (node.name, new_group.uuid))
1834
1835 resmod.append((node, old_group, new_group))
1836
1837
1838 for (node, old_group, new_group) in resmod:
1839 assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
1840 "Assigning to current group is not possible"
1841
1842 node.group = new_group.uuid
1843
1844
1845 if node.name in old_group.members:
1846 old_group.members.remove(node.name)
1847 if node.name not in new_group.members:
1848 new_group.members.append(node.name)
1849
1850
1851 now = time.time()
1852 for obj in frozenset(itertools.chain(*resmod)):
1853 obj.serial_no += 1
1854 obj.mtime = now
1855
1856
1857 self._config_data.cluster.serial_no += 1
1858
1859 self._WriteConfig()
1860
1862 """Bump up the serial number of the config.
1863
1864 """
1865 self._config_data.serial_no += 1
1866 self._config_data.mtime = time.time()
1867
1869 """Returns all objects with uuid attributes.
1870
1871 """
1872 return (self._config_data.instances.values() +
1873 self._config_data.nodes.values() +
1874 self._config_data.nodegroups.values() +
1875 [self._config_data.cluster])
1876
1878 """Read the config data from disk.
1879
1880 """
1881 raw_data = utils.ReadFile(self._cfg_file)
1882
1883 try:
1884 data = objects.ConfigData.FromDict(serializer.Load(raw_data))
1885 except Exception, err:
1886 raise errors.ConfigurationError(err)
1887
1888
1889 _ValidateConfig(data)
1890
1891 if (not hasattr(data, "cluster") or
1892 not hasattr(data.cluster, "rsahostkeypub")):
1893 raise errors.ConfigurationError("Incomplete configuration"
1894 " (missing cluster.rsahostkeypub)")
1895
1896 if data.cluster.master_node != self._my_hostname and not accept_foreign:
1897 msg = ("The configuration denotes node %s as master, while my"
1898 " hostname is %s; opening a foreign configuration is only"
1899 " possible in accept_foreign mode" %
1900 (data.cluster.master_node, self._my_hostname))
1901 raise errors.ConfigurationError(msg)
1902
1903
1904 data.UpgradeConfig()
1905
1906 self._config_data = data
1907
1908
1909 self._last_cluster_serial = -1
1910
1911
1912 self._UpgradeConfig()
1913
1914 self._cfg_id = utils.GetFileID(path=self._cfg_file)
1915
1917 """Run upgrade steps that cannot be done purely in the objects.
1918
1919 This is because some data elements need uniqueness across the
1920 whole configuration, etc.
1921
1922 @warning: this function will call L{_WriteConfig()}, but also
1923 L{DropECReservations} so it needs to be called only from a
1924 "safe" place (the constructor). If one wanted to call it with
1925 the lock held, a DropECReservationUnlocked would need to be
1926 created first, to avoid causing deadlock.
1927
1928 """
1929 modified = False
1930 for item in self._AllUUIDObjects():
1931 if item.uuid is None:
1932 item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
1933 modified = True
1934 if not self._config_data.nodegroups:
1935 default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
1936 default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
1937 members=[])
1938 self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
1939 modified = True
1940 for node in self._config_data.nodes.values():
1941 if not node.group:
1942 node.group = self.LookupNodeGroup(None)
1943 modified = True
1944
1945
1946
1947
1948 self._UnlockedAddNodeToGroup(node.name, node.group)
1949 if modified:
1950 self._WriteConfig()
1951
1952
1953 self.DropECReservations(_UPGRADE_CONFIG_JID)
1954
1956 """Distribute the configuration to the other nodes.
1957
1958 Currently, this only copies the configuration file. In the future,
1959 it could be used to encapsulate the 2/3-phase update mechanism.
1960
1961 """
1962 if self._offline:
1963 return True
1964
1965 bad = False
1966
1967 node_list = []
1968 addr_list = []
1969 myhostname = self._my_hostname
1970
1971
1972
1973
1974 for node_name in self._UnlockedGetNodeList():
1975 if node_name == myhostname:
1976 continue
1977 node_info = self._UnlockedGetNodeInfo(node_name)
1978 if not node_info.master_candidate:
1979 continue
1980 node_list.append(node_info.name)
1981 addr_list.append(node_info.primary_ip)
1982
1983
1984 result = \
1985 self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
1986 for to_node, to_result in result.items():
1987 msg = to_result.fail_msg
1988 if msg:
1989 msg = ("Copy of file %s to node %s failed: %s" %
1990 (self._cfg_file, to_node, msg))
1991 logging.error(msg)
1992
1993 if feedback_fn:
1994 feedback_fn(msg)
1995
1996 bad = True
1997
1998 return not bad
1999
2000 - def _WriteConfig(self, destination=None, feedback_fn=None):
2001 """Write the configuration data to persistent storage.
2002
2003 """
2004 assert feedback_fn is None or callable(feedback_fn)
2005
2006
2007
2008
2009
2010 config_errors = self._UnlockedVerifyConfig()
2011 if config_errors:
2012 errmsg = ("Configuration data is not consistent: %s" %
2013 (utils.CommaJoin(config_errors)))
2014 logging.critical(errmsg)
2015 if feedback_fn:
2016 feedback_fn(errmsg)
2017
2018 if destination is None:
2019 destination = self._cfg_file
2020 self._BumpSerialNo()
2021 txt = serializer.Dump(self._config_data.ToDict())
2022
2023 getents = self._getents()
2024 try:
2025 fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2026 close=False, gid=getents.confd_gid, mode=0640)
2027 except errors.LockError:
2028 raise errors.ConfigurationError("The configuration file has been"
2029 " modified since the last write, cannot"
2030 " update")
2031 try:
2032 self._cfg_id = utils.GetFileID(fd=fd)
2033 finally:
2034 os.close(fd)
2035
2036 self.write_count += 1
2037
2038
2039 self._DistributeConfig(feedback_fn)
2040
2041
2042 if self._last_cluster_serial < self._config_data.cluster.serial_no:
2043 if not self._offline:
2044 result = self._GetRpc(None).call_write_ssconf_files(
2045 self._UnlockedGetOnlineNodeList(),
2046 self._UnlockedGetSsconfValues())
2047
2048 for nname, nresu in result.items():
2049 msg = nresu.fail_msg
2050 if msg:
2051 errmsg = ("Error while uploading ssconf files to"
2052 " node %s: %s" % (nname, msg))
2053 logging.warning(errmsg)
2054
2055 if feedback_fn:
2056 feedback_fn(errmsg)
2057
2058 self._last_cluster_serial = self._config_data.cluster.serial_no
2059
2061 """Return the values needed by ssconf.
2062
2063 @rtype: dict
2064 @return: a dictionary with keys the ssconf names and values their
2065 associated value
2066
2067 """
2068 fn = "\n".join
2069 instance_names = utils.NiceSort(self._UnlockedGetInstanceList())
2070 node_names = utils.NiceSort(self._UnlockedGetNodeList())
2071 node_info = [self._UnlockedGetNodeInfo(name) for name in node_names]
2072 node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2073 for ninfo in node_info]
2074 node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2075 for ninfo in node_info]
2076
2077 instance_data = fn(instance_names)
2078 off_data = fn(node.name for node in node_info if node.offline)
2079 on_data = fn(node.name for node in node_info if not node.offline)
2080 mc_data = fn(node.name for node in node_info if node.master_candidate)
2081 mc_ips_data = fn(node.primary_ip for node in node_info
2082 if node.master_candidate)
2083 node_data = fn(node_names)
2084 node_pri_ips_data = fn(node_pri_ips)
2085 node_snd_ips_data = fn(node_snd_ips)
2086
2087 cluster = self._config_data.cluster
2088 cluster_tags = fn(cluster.GetTags())
2089
2090 hypervisor_list = fn(cluster.enabled_hypervisors)
2091
2092 uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2093
2094 nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2095 self._config_data.nodegroups.values()]
2096 nodegroups_data = fn(utils.NiceSort(nodegroups))
2097
2098 ssconf_values = {
2099 constants.SS_CLUSTER_NAME: cluster.cluster_name,
2100 constants.SS_CLUSTER_TAGS: cluster_tags,
2101 constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2102 constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2103 constants.SS_MASTER_CANDIDATES: mc_data,
2104 constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2105 constants.SS_MASTER_IP: cluster.master_ip,
2106 constants.SS_MASTER_NETDEV: cluster.master_netdev,
2107 constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2108 constants.SS_MASTER_NODE: cluster.master_node,
2109 constants.SS_NODE_LIST: node_data,
2110 constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2111 constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2112 constants.SS_OFFLINE_NODES: off_data,
2113 constants.SS_ONLINE_NODES: on_data,
2114 constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2115 constants.SS_INSTANCE_LIST: instance_data,
2116 constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2117 constants.SS_HYPERVISOR_LIST: hypervisor_list,
2118 constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2119 constants.SS_UID_POOL: uid_pool,
2120 constants.SS_NODEGROUPS: nodegroups_data,
2121 }
2122 bad_values = [(k, v) for k, v in ssconf_values.items()
2123 if not isinstance(v, (str, basestring))]
2124 if bad_values:
2125 err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2126 raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2127 " values: %s" % err)
2128 return ssconf_values
2129
2130 @locking.ssynchronized(_config_lock, shared=1)
2136
2137 @locking.ssynchronized(_config_lock, shared=1)
2139 """Return the volume group name.
2140
2141 """
2142 return self._config_data.cluster.volume_group_name
2143
2144 @locking.ssynchronized(_config_lock)
2146 """Set the volume group name.
2147
2148 """
2149 self._config_data.cluster.volume_group_name = vg_name
2150 self._config_data.cluster.serial_no += 1
2151 self._WriteConfig()
2152
2153 @locking.ssynchronized(_config_lock, shared=1)
2155 """Return DRBD usermode helper.
2156
2157 """
2158 return self._config_data.cluster.drbd_usermode_helper
2159
2160 @locking.ssynchronized(_config_lock)
2162 """Set DRBD usermode helper.
2163
2164 """
2165 self._config_data.cluster.drbd_usermode_helper = drbd_helper
2166 self._config_data.cluster.serial_no += 1
2167 self._WriteConfig()
2168
2169 @locking.ssynchronized(_config_lock, shared=1)
2171 """Return the mac prefix.
2172
2173 """
2174 return self._config_data.cluster.mac_prefix
2175
2176 @locking.ssynchronized(_config_lock, shared=1)
2178 """Returns information about the cluster
2179
2180 @rtype: L{objects.Cluster}
2181 @return: the cluster object
2182
2183 """
2184 return self._config_data.cluster
2185
2186 @locking.ssynchronized(_config_lock, shared=1)
2188 """Check if in there is at disk of the given type in the configuration.
2189
2190 """
2191 return self._config_data.HasAnyDiskOfType(dev_type)
2192
2193 @locking.ssynchronized(_config_lock)
2194 - def Update(self, target, feedback_fn):
2195 """Notify function to be called after updates.
2196
2197 This function must be called when an object (as returned by
2198 GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2199 caller wants the modifications saved to the backing store. Note
2200 that all modified objects will be saved, but the target argument
2201 is the one the caller wants to ensure that it's saved.
2202
2203 @param target: an instance of either L{objects.Cluster},
2204 L{objects.Node} or L{objects.Instance} which is existing in
2205 the cluster
2206 @param feedback_fn: Callable feedback function
2207
2208 """
2209 if self._config_data is None:
2210 raise errors.ProgrammerError("Configuration file not read,"
2211 " cannot save.")
2212 update_serial = False
2213 if isinstance(target, objects.Cluster):
2214 test = target == self._config_data.cluster
2215 elif isinstance(target, objects.Node):
2216 test = target in self._config_data.nodes.values()
2217 update_serial = True
2218 elif isinstance(target, objects.Instance):
2219 test = target in self._config_data.instances.values()
2220 elif isinstance(target, objects.NodeGroup):
2221 test = target in self._config_data.nodegroups.values()
2222 else:
2223 raise errors.ProgrammerError("Invalid object type (%s) passed to"
2224 " ConfigWriter.Update" % type(target))
2225 if not test:
2226 raise errors.ConfigurationError("Configuration updated since object"
2227 " has been read or unknown object")
2228 target.serial_no += 1
2229 target.mtime = now = time.time()
2230
2231 if update_serial:
2232
2233 self._config_data.cluster.serial_no += 1
2234 self._config_data.cluster.mtime = now
2235
2236 if isinstance(target, objects.Instance):
2237 self._UnlockedReleaseDRBDMinors(target.name)
2238
2239 self._WriteConfig(feedback_fn=feedback_fn)
2240
2241 @locking.ssynchronized(_config_lock)
2243 """Drop per-execution-context reservations
2244
2245 """
2246 for rm in self._all_rms:
2247 rm.DropECReservations(ec_id)
2248