4 # Copyright (C) 2006, 2007, 2008 Google Inc.
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.
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.
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
22 """Module implementing the master-side code."""
24 # pylint: disable-msg=W0613,W0201
37 from ganeti import ssh
38 from ganeti import utils
39 from ganeti import errors
40 from ganeti import hypervisor
41 from ganeti import locking
42 from ganeti import constants
43 from ganeti import objects
44 from ganeti import opcodes
45 from ganeti import serializer
46 from ganeti import ssconf
49 class LogicalUnit(object):
50 """Logical Unit base class.
52 Subclasses must follow these rules:
53 - implement ExpandNames
54 - implement CheckPrereq
56 - implement BuildHooksEnv
57 - redefine HPATH and HTYPE
58 - optionally redefine their run requirements:
59 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
61 Note that all commands require root permissions.
69 def __init__(self, processor, op, context, rpc):
70 """Constructor for LogicalUnit.
72 This needs to be overriden in derived classes in order to check op
78 self.cfg = context.cfg
79 self.context = context
81 # Dicts used to declare locking needs to mcpu
82 self.needed_locks = None
83 self.acquired_locks = {}
84 self.share_locks = dict(((i, 0) for i in locking.LEVELS))
86 self.remove_locks = {}
87 # Used to force good behavior when calling helper functions
88 self.recalculate_locks = {}
91 self.LogWarning = processor.LogWarning
92 self.LogInfo = processor.LogInfo
94 for attr_name in self._OP_REQP:
95 attr_val = getattr(op, attr_name, None)
97 raise errors.OpPrereqError("Required parameter '%s' missing" %
102 """Returns the SshRunner object
106 self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
109 ssh = property(fget=__GetSSH)
111 def CheckArguments(self):
112 """Check syntactic validity for the opcode arguments.
114 This method is for doing a simple syntactic check and ensure
115 validity of opcode parameters, without any cluster-related
116 checks. While the same can be accomplished in ExpandNames and/or
117 CheckPrereq, doing these separate is better because:
119 - ExpandNames is left as as purely a lock-related function
120 - CheckPrereq is run after we have aquired locks (and possible
123 The function is allowed to change the self.op attribute so that
124 later methods can no longer worry about missing parameters.
129 def ExpandNames(self):
130 """Expand names for this LU.
132 This method is called before starting to execute the opcode, and it should
133 update all the parameters of the opcode to their canonical form (e.g. a
134 short node name must be fully expanded after this method has successfully
135 completed). This way locking, hooks, logging, ecc. can work correctly.
137 LUs which implement this method must also populate the self.needed_locks
138 member, as a dict with lock levels as keys, and a list of needed lock names
141 - use an empty dict if you don't need any lock
142 - if you don't need any lock at a particular level omit that level
143 - don't put anything for the BGL level
144 - if you want all locks at a level use locking.ALL_SET as a value
146 If you need to share locks (rather than acquire them exclusively) at one
147 level you can modify self.share_locks, setting a true value (usually 1) for
148 that level. By default locks are not shared.
152 # Acquire all nodes and one instance
153 self.needed_locks = {
154 locking.LEVEL_NODE: locking.ALL_SET,
155 locking.LEVEL_INSTANCE: ['instance1.example.tld'],
157 # Acquire just two nodes
158 self.needed_locks = {
159 locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
162 self.needed_locks = {} # No, you can't leave it to the default value None
165 # The implementation of this method is mandatory only if the new LU is
166 # concurrent, so that old LUs don't need to be changed all at the same
169 self.needed_locks = {} # Exclusive LUs don't need locks.
171 raise NotImplementedError
173 def DeclareLocks(self, level):
174 """Declare LU locking needs for a level
176 While most LUs can just declare their locking needs at ExpandNames time,
177 sometimes there's the need to calculate some locks after having acquired
178 the ones before. This function is called just before acquiring locks at a
179 particular level, but after acquiring the ones at lower levels, and permits
180 such calculations. It can be used to modify self.needed_locks, and by
181 default it does nothing.
183 This function is only called if you have something already set in
184 self.needed_locks for the level.
186 @param level: Locking level which is going to be locked
187 @type level: member of ganeti.locking.LEVELS
191 def CheckPrereq(self):
192 """Check prerequisites for this LU.
194 This method should check that the prerequisites for the execution
195 of this LU are fulfilled. It can do internode communication, but
196 it should be idempotent - no cluster or system changes are
199 The method should raise errors.OpPrereqError in case something is
200 not fulfilled. Its return value is ignored.
202 This method should also update all the parameters of the opcode to
203 their canonical form if it hasn't been done by ExpandNames before.
206 raise NotImplementedError
208 def Exec(self, feedback_fn):
211 This method should implement the actual work. It should raise
212 errors.OpExecError for failures that are somewhat dealt with in
216 raise NotImplementedError
218 def BuildHooksEnv(self):
219 """Build hooks environment for this LU.
221 This method should return a three-node tuple consisting of: a dict
222 containing the environment that will be used for running the
223 specific hook for this LU, a list of node names on which the hook
224 should run before the execution, and a list of node names on which
225 the hook should run after the execution.
227 The keys of the dict must not have 'GANETI_' prefixed as this will
228 be handled in the hooks runner. Also note additional keys will be
229 added by the hooks runner. If the LU doesn't define any
230 environment, an empty dict (and not None) should be returned.
232 No nodes should be returned as an empty list (and not None).
234 Note that if the HPATH for a LU class is None, this function will
238 raise NotImplementedError
240 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
241 """Notify the LU about the results of its hooks.
243 This method is called every time a hooks phase is executed, and notifies
244 the Logical Unit about the hooks' result. The LU can then use it to alter
245 its result based on the hooks. By default the method does nothing and the
246 previous result is passed back unchanged but any LU can define it if it
247 wants to use the local cluster hook-scripts somehow.
249 @param phase: one of L{constants.HOOKS_PHASE_POST} or
250 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
251 @param hook_results: the results of the multi-node hooks rpc call
252 @param feedback_fn: function used send feedback back to the caller
253 @param lu_result: the previous Exec result this LU had, or None
255 @return: the new Exec result, based on the previous result
261 def _ExpandAndLockInstance(self):
262 """Helper function to expand and lock an instance.
264 Many LUs that work on an instance take its name in self.op.instance_name
265 and need to expand it and then declare the expanded name for locking. This
266 function does it, and then updates self.op.instance_name to the expanded
267 name. It also initializes needed_locks as a dict, if this hasn't been done
271 if self.needed_locks is None:
272 self.needed_locks = {}
274 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
275 "_ExpandAndLockInstance called with instance-level locks set"
276 expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
277 if expanded_name is None:
278 raise errors.OpPrereqError("Instance '%s' not known" %
279 self.op.instance_name)
280 self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
281 self.op.instance_name = expanded_name
283 def _LockInstancesNodes(self, primary_only=False):
284 """Helper function to declare instances' nodes for locking.
286 This function should be called after locking one or more instances to lock
287 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
288 with all primary or secondary nodes for instances already locked and
289 present in self.needed_locks[locking.LEVEL_INSTANCE].
291 It should be called from DeclareLocks, and for safety only works if
292 self.recalculate_locks[locking.LEVEL_NODE] is set.
294 In the future it may grow parameters to just lock some instance's nodes, or
295 to just lock primaries or secondary nodes, if needed.
297 If should be called in DeclareLocks in a way similar to::
299 if level == locking.LEVEL_NODE:
300 self._LockInstancesNodes()
302 @type primary_only: boolean
303 @param primary_only: only lock primary nodes of locked instances
306 assert locking.LEVEL_NODE in self.recalculate_locks, \
307 "_LockInstancesNodes helper function called with no nodes to recalculate"
309 # TODO: check if we're really been called with the instance locks held
311 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
312 # future we might want to have different behaviors depending on the value
313 # of self.recalculate_locks[locking.LEVEL_NODE]
315 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
316 instance = self.context.cfg.GetInstanceInfo(instance_name)
317 wanted_nodes.append(instance.primary_node)
319 wanted_nodes.extend(instance.secondary_nodes)
321 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
322 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
323 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
324 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
326 del self.recalculate_locks[locking.LEVEL_NODE]
329 class NoHooksLU(LogicalUnit):
330 """Simple LU which runs no hooks.
332 This LU is intended as a parent for other LogicalUnits which will
333 run no hooks, in order to reduce duplicate code.
340 def _GetWantedNodes(lu, nodes):
341 """Returns list of checked and expanded node names.
343 @type lu: L{LogicalUnit}
344 @param lu: the logical unit on whose behalf we execute
346 @param nodes: list of node names or None for all nodes
348 @return: the list of nodes, sorted
349 @raise errors.OpProgrammerError: if the nodes parameter is wrong type
352 if not isinstance(nodes, list):
353 raise errors.OpPrereqError("Invalid argument type 'nodes'")
356 raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
357 " non-empty list of nodes whose name is to be expanded.")
361 node = lu.cfg.ExpandNodeName(name)
363 raise errors.OpPrereqError("No such node name '%s'" % name)
366 return utils.NiceSort(wanted)
369 def _GetWantedInstances(lu, instances):
370 """Returns list of checked and expanded instance names.
372 @type lu: L{LogicalUnit}
373 @param lu: the logical unit on whose behalf we execute
374 @type instances: list
375 @param instances: list of instance names or None for all instances
377 @return: the list of instances, sorted
378 @raise errors.OpPrereqError: if the instances parameter is wrong type
379 @raise errors.OpPrereqError: if any of the passed instances is not found
382 if not isinstance(instances, list):
383 raise errors.OpPrereqError("Invalid argument type 'instances'")
388 for name in instances:
389 instance = lu.cfg.ExpandInstanceName(name)
391 raise errors.OpPrereqError("No such instance name '%s'" % name)
392 wanted.append(instance)
395 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
399 def _CheckOutputFields(static, dynamic, selected):
400 """Checks whether all selected fields are valid.
402 @type static: L{utils.FieldSet}
403 @param static: static fields set
404 @type dynamic: L{utils.FieldSet}
405 @param dynamic: dynamic fields set
412 delta = f.NonMatching(selected)
414 raise errors.OpPrereqError("Unknown output fields selected: %s"
418 def _CheckBooleanOpField(op, name):
419 """Validates boolean opcode parameters.
421 This will ensure that an opcode parameter is either a boolean value,
422 or None (but that it always exists).
425 val = getattr(op, name, None)
426 if not (val is None or isinstance(val, bool)):
427 raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
429 setattr(op, name, val)
432 def _CheckNodeOnline(lu, node):
433 """Ensure that a given node is online.
435 @param lu: the LU on behalf of which we make the check
436 @param node: the node to check
437 @raise errors.OpPrereqError: if the node is offline
440 if lu.cfg.GetNodeInfo(node).offline:
441 raise errors.OpPrereqError("Can't use offline node %s" % node)
444 def _CheckNodeNotDrained(lu, node):
445 """Ensure that a given node is not drained.
447 @param lu: the LU on behalf of which we make the check
448 @param node: the node to check
449 @raise errors.OpPrereqError: if the node is drained
452 if lu.cfg.GetNodeInfo(node).drained:
453 raise errors.OpPrereqError("Can't use drained node %s" % node)
456 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
457 memory, vcpus, nics):
458 """Builds instance related env variables for hooks
460 This builds the hook environment from individual variables.
463 @param name: the name of the instance
464 @type primary_node: string
465 @param primary_node: the name of the instance's primary node
466 @type secondary_nodes: list
467 @param secondary_nodes: list of secondary nodes as strings
468 @type os_type: string
469 @param os_type: the name of the instance's OS
470 @type status: boolean
471 @param status: the should_run status of the instance
473 @param memory: the memory size of the instance
475 @param vcpus: the count of VCPUs the instance has
477 @param nics: list of tuples (ip, bridge, mac) representing
478 the NICs the instance has
480 @return: the hook environment for this instance
489 "INSTANCE_NAME": name,
490 "INSTANCE_PRIMARY": primary_node,
491 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
492 "INSTANCE_OS_TYPE": os_type,
493 "INSTANCE_STATUS": str_status,
494 "INSTANCE_MEMORY": memory,
495 "INSTANCE_VCPUS": vcpus,
499 nic_count = len(nics)
500 for idx, (ip, bridge, mac) in enumerate(nics):
503 env["INSTANCE_NIC%d_IP" % idx] = ip
504 env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
505 env["INSTANCE_NIC%d_HWADDR" % idx] = mac
509 env["INSTANCE_NIC_COUNT"] = nic_count
514 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
515 """Builds instance related env variables for hooks from an object.
517 @type lu: L{LogicalUnit}
518 @param lu: the logical unit on whose behalf we execute
519 @type instance: L{objects.Instance}
520 @param instance: the instance for which we should build the
523 @param override: dictionary with key/values that will override
526 @return: the hook environment dictionary
529 bep = lu.cfg.GetClusterInfo().FillBE(instance)
531 'name': instance.name,
532 'primary_node': instance.primary_node,
533 'secondary_nodes': instance.secondary_nodes,
534 'os_type': instance.os,
535 'status': instance.admin_up,
536 'memory': bep[constants.BE_MEMORY],
537 'vcpus': bep[constants.BE_VCPUS],
538 'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
541 args.update(override)
542 return _BuildInstanceHookEnv(**args)
545 def _AdjustCandidatePool(lu):
546 """Adjust the candidate pool after node operations.
549 mod_list = lu.cfg.MaintainCandidatePool()
551 lu.LogInfo("Promoted nodes to master candidate role: %s",
552 ", ".join(node.name for node in mod_list))
553 for name in mod_list:
554 lu.context.ReaddNode(name)
555 mc_now, mc_max = lu.cfg.GetMasterCandidateStats()
557 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
561 def _CheckInstanceBridgesExist(lu, instance):
562 """Check that the brigdes needed by an instance exist.
565 # check bridges existance
566 brlist = [nic.bridge for nic in instance.nics]
567 result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
570 raise errors.OpPrereqError("One or more target bridges %s does not"
571 " exist on destination node '%s'" %
572 (brlist, instance.primary_node))
575 class LUDestroyCluster(NoHooksLU):
576 """Logical unit for destroying the cluster.
581 def CheckPrereq(self):
582 """Check prerequisites.
584 This checks whether the cluster is empty.
586 Any errors are signalled by raising errors.OpPrereqError.
589 master = self.cfg.GetMasterNode()
591 nodelist = self.cfg.GetNodeList()
592 if len(nodelist) != 1 or nodelist[0] != master:
593 raise errors.OpPrereqError("There are still %d node(s) in"
594 " this cluster." % (len(nodelist) - 1))
595 instancelist = self.cfg.GetInstanceList()
597 raise errors.OpPrereqError("There are still %d instance(s) in"
598 " this cluster." % len(instancelist))
600 def Exec(self, feedback_fn):
601 """Destroys the cluster.
604 master = self.cfg.GetMasterNode()
605 result = self.rpc.call_node_stop_master(master, False)
608 raise errors.OpExecError("Could not disable the master role")
609 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
610 utils.CreateBackup(priv_key)
611 utils.CreateBackup(pub_key)
615 class LUVerifyCluster(LogicalUnit):
616 """Verifies the cluster status.
619 HPATH = "cluster-verify"
620 HTYPE = constants.HTYPE_CLUSTER
621 _OP_REQP = ["skip_checks"]
624 def ExpandNames(self):
625 self.needed_locks = {
626 locking.LEVEL_NODE: locking.ALL_SET,
627 locking.LEVEL_INSTANCE: locking.ALL_SET,
629 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
631 def _VerifyNode(self, nodeinfo, file_list, local_cksum,
632 node_result, feedback_fn, master_files,
634 """Run multiple tests against a node.
638 - compares ganeti version
639 - checks vg existance and size > 20G
640 - checks config file checksum
641 - checks ssh to other nodes
643 @type nodeinfo: L{objects.Node}
644 @param nodeinfo: the node to check
645 @param file_list: required list of files
646 @param local_cksum: dictionary of local files and their checksums
647 @param node_result: the results from the node
648 @param feedback_fn: function used to accumulate results
649 @param master_files: list of files that only masters should have
650 @param drbd_map: the useddrbd minors for this node, in
651 form of minor: (instance, must_exist) which correspond to instances
652 and their running status
657 # main result, node_result should be a non-empty dict
658 if not node_result or not isinstance(node_result, dict):
659 feedback_fn(" - ERROR: unable to verify node %s." % (node,))
662 # compares ganeti version
663 local_version = constants.PROTOCOL_VERSION
664 remote_version = node_result.get('version', None)
665 if not (remote_version and isinstance(remote_version, (list, tuple)) and
666 len(remote_version) == 2):
667 feedback_fn(" - ERROR: connection to %s failed" % (node))
670 if local_version != remote_version[0]:
671 feedback_fn(" - ERROR: incompatible protocol versions: master %s,"
672 " node %s %s" % (local_version, node, remote_version[0]))
675 # node seems compatible, we can actually try to look into its results
679 # full package version
680 if constants.RELEASE_VERSION != remote_version[1]:
681 feedback_fn(" - WARNING: software version mismatch: master %s,"
683 (constants.RELEASE_VERSION, node, remote_version[1]))
685 # checks vg existence and size > 20G
687 vglist = node_result.get(constants.NV_VGLIST, None)
689 feedback_fn(" - ERROR: unable to check volume groups on node %s." %
693 vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
694 constants.MIN_VG_SIZE)
696 feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
699 # checks config file checksum
701 remote_cksum = node_result.get(constants.NV_FILELIST, None)
702 if not isinstance(remote_cksum, dict):
704 feedback_fn(" - ERROR: node hasn't returned file checksum data")
706 for file_name in file_list:
707 node_is_mc = nodeinfo.master_candidate
708 must_have_file = file_name not in master_files
709 if file_name not in remote_cksum:
710 if node_is_mc or must_have_file:
712 feedback_fn(" - ERROR: file '%s' missing" % file_name)
713 elif remote_cksum[file_name] != local_cksum[file_name]:
714 if node_is_mc or must_have_file:
716 feedback_fn(" - ERROR: file '%s' has wrong checksum" % file_name)
718 # not candidate and this is not a must-have file
720 feedback_fn(" - ERROR: non master-candidate has old/wrong file"
723 # all good, except non-master/non-must have combination
724 if not node_is_mc and not must_have_file:
725 feedback_fn(" - ERROR: file '%s' should not exist on non master"
726 " candidates" % file_name)
730 if constants.NV_NODELIST not in node_result:
732 feedback_fn(" - ERROR: node hasn't returned node ssh connectivity data")
734 if node_result[constants.NV_NODELIST]:
736 for node in node_result[constants.NV_NODELIST]:
737 feedback_fn(" - ERROR: ssh communication with node '%s': %s" %
738 (node, node_result[constants.NV_NODELIST][node]))
740 if constants.NV_NODENETTEST not in node_result:
742 feedback_fn(" - ERROR: node hasn't returned node tcp connectivity data")
744 if node_result[constants.NV_NODENETTEST]:
746 nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
748 feedback_fn(" - ERROR: tcp communication with node '%s': %s" %
749 (node, node_result[constants.NV_NODENETTEST][node]))
751 hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
752 if isinstance(hyp_result, dict):
753 for hv_name, hv_result in hyp_result.iteritems():
754 if hv_result is not None:
755 feedback_fn(" - ERROR: hypervisor %s verify failure: '%s'" %
756 (hv_name, hv_result))
758 # check used drbd list
759 used_minors = node_result.get(constants.NV_DRBDLIST, [])
760 for minor, (iname, must_exist) in drbd_map.items():
761 if minor not in used_minors and must_exist:
762 feedback_fn(" - ERROR: drbd minor %d of instance %s is not active" %
765 for minor in used_minors:
766 if minor not in drbd_map:
767 feedback_fn(" - ERROR: unallocated drbd minor %d is in use" % minor)
772 def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
773 node_instance, feedback_fn, n_offline):
774 """Verify an instance.
776 This function checks to see if the required block devices are
777 available on the instance's node.
782 node_current = instanceconfig.primary_node
785 instanceconfig.MapLVsByNode(node_vol_should)
787 for node in node_vol_should:
788 if node in n_offline:
789 # ignore missing volumes on offline nodes
791 for volume in node_vol_should[node]:
792 if node not in node_vol_is or volume not in node_vol_is[node]:
793 feedback_fn(" - ERROR: volume %s missing on node %s" %
797 if instanceconfig.admin_up:
798 if ((node_current not in node_instance or
799 not instance in node_instance[node_current]) and
800 node_current not in n_offline):
801 feedback_fn(" - ERROR: instance %s not running on node %s" %
802 (instance, node_current))
805 for node in node_instance:
806 if (not node == node_current):
807 if instance in node_instance[node]:
808 feedback_fn(" - ERROR: instance %s should not run on node %s" %
814 def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
815 """Verify if there are any unknown volumes in the cluster.
817 The .os, .swap and backup volumes are ignored. All other volumes are
823 for node in node_vol_is:
824 for volume in node_vol_is[node]:
825 if node not in node_vol_should or volume not in node_vol_should[node]:
826 feedback_fn(" - ERROR: volume %s on node %s should not exist" %
831 def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
832 """Verify the list of running instances.
834 This checks what instances are running but unknown to the cluster.
838 for node in node_instance:
839 for runninginstance in node_instance[node]:
840 if runninginstance not in instancelist:
841 feedback_fn(" - ERROR: instance %s on node %s should not exist" %
842 (runninginstance, node))
846 def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
847 """Verify N+1 Memory Resilience.
849 Check that if one single node dies we can still start all the instances it
855 for node, nodeinfo in node_info.iteritems():
856 # This code checks that every node which is now listed as secondary has
857 # enough memory to host all instances it is supposed to should a single
858 # other node in the cluster fail.
859 # FIXME: not ready for failover to an arbitrary node
860 # FIXME: does not support file-backed instances
861 # WARNING: we currently take into account down instances as well as up
862 # ones, considering that even if they're down someone might want to start
863 # them even in the event of a node failure.
864 for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
866 for instance in instances:
867 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
868 if bep[constants.BE_AUTO_BALANCE]:
869 needed_mem += bep[constants.BE_MEMORY]
870 if nodeinfo['mfree'] < needed_mem:
871 feedback_fn(" - ERROR: not enough memory on node %s to accomodate"
872 " failovers should node %s fail" % (node, prinode))
876 def CheckPrereq(self):
877 """Check prerequisites.
879 Transform the list of checks we're going to skip into a set and check that
880 all its members are valid.
883 self.skip_set = frozenset(self.op.skip_checks)
884 if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
885 raise errors.OpPrereqError("Invalid checks to be skipped specified")
887 def BuildHooksEnv(self):
890 Cluster-Verify hooks just rone in the post phase and their failure makes
891 the output be logged in the verify output and the verification to fail.
894 all_nodes = self.cfg.GetNodeList()
895 # TODO: populate the environment with useful information for verify hooks
897 return env, [], all_nodes
899 def Exec(self, feedback_fn):
900 """Verify integrity of cluster, performing various test on nodes.
904 feedback_fn("* Verifying global settings")
905 for msg in self.cfg.VerifyConfig():
906 feedback_fn(" - ERROR: %s" % msg)
908 vg_name = self.cfg.GetVGName()
909 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
910 nodelist = utils.NiceSort(self.cfg.GetNodeList())
911 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
912 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
913 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
914 for iname in instancelist)
915 i_non_redundant = [] # Non redundant instances
916 i_non_a_balanced = [] # Non auto-balanced instances
917 n_offline = [] # List of offline nodes
918 n_drained = [] # List of nodes being drained
924 # FIXME: verify OS list
926 master_files = [constants.CLUSTER_CONF_FILE]
928 file_names = ssconf.SimpleStore().GetFileList()
929 file_names.append(constants.SSL_CERT_FILE)
930 file_names.append(constants.RAPI_CERT_FILE)
931 file_names.extend(master_files)
933 local_checksums = utils.FingerprintFiles(file_names)
935 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
936 node_verify_param = {
937 constants.NV_FILELIST: file_names,
938 constants.NV_NODELIST: [node.name for node in nodeinfo
939 if not node.offline],
940 constants.NV_HYPERVISOR: hypervisors,
941 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
942 node.secondary_ip) for node in nodeinfo
943 if not node.offline],
944 constants.NV_LVLIST: vg_name,
945 constants.NV_INSTANCELIST: hypervisors,
946 constants.NV_VGLIST: None,
947 constants.NV_VERSION: None,
948 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
949 constants.NV_DRBDLIST: None,
951 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
952 self.cfg.GetClusterName())
954 cluster = self.cfg.GetClusterInfo()
955 master_node = self.cfg.GetMasterNode()
956 all_drbd_map = self.cfg.ComputeDRBDMap()
958 for node_i in nodeinfo:
960 nresult = all_nvinfo[node].data
963 feedback_fn("* Skipping offline node %s" % (node,))
964 n_offline.append(node)
967 if node == master_node:
969 elif node_i.master_candidate:
970 ntype = "master candidate"
973 n_drained.append(node)
976 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
978 if all_nvinfo[node].failed or not isinstance(nresult, dict):
979 feedback_fn(" - ERROR: connection to %s failed" % (node,))
984 for minor, instance in all_drbd_map[node].items():
985 instance = instanceinfo[instance]
986 node_drbd[minor] = (instance.name, instance.admin_up)
987 result = self._VerifyNode(node_i, file_names, local_checksums,
988 nresult, feedback_fn, master_files,
992 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
993 if isinstance(lvdata, basestring):
994 feedback_fn(" - ERROR: LVM problem on node %s: %s" %
995 (node, utils.SafeEncode(lvdata)))
997 node_volume[node] = {}
998 elif not isinstance(lvdata, dict):
999 feedback_fn(" - ERROR: connection to %s failed (lvlist)" % (node,))
1003 node_volume[node] = lvdata
1006 idata = nresult.get(constants.NV_INSTANCELIST, None)
1007 if not isinstance(idata, list):
1008 feedback_fn(" - ERROR: connection to %s failed (instancelist)" %
1013 node_instance[node] = idata
1016 nodeinfo = nresult.get(constants.NV_HVINFO, None)
1017 if not isinstance(nodeinfo, dict):
1018 feedback_fn(" - ERROR: connection to %s failed (hvinfo)" % (node,))
1024 "mfree": int(nodeinfo['memory_free']),
1025 "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
1028 # dictionary holding all instances this node is secondary for,
1029 # grouped by their primary node. Each key is a cluster node, and each
1030 # value is a list of instances which have the key as primary and the
1031 # current node as secondary. this is handy to calculate N+1 memory
1032 # availability if you can only failover from a primary to its
1034 "sinst-by-pnode": {},
1037 feedback_fn(" - ERROR: invalid value returned from node %s" % (node,))
1041 node_vol_should = {}
1043 for instance in instancelist:
1044 feedback_fn("* Verifying instance %s" % instance)
1045 inst_config = instanceinfo[instance]
1046 result = self._VerifyInstance(instance, inst_config, node_volume,
1047 node_instance, feedback_fn, n_offline)
1049 inst_nodes_offline = []
1051 inst_config.MapLVsByNode(node_vol_should)
1053 instance_cfg[instance] = inst_config
1055 pnode = inst_config.primary_node
1056 if pnode in node_info:
1057 node_info[pnode]['pinst'].append(instance)
1058 elif pnode not in n_offline:
1059 feedback_fn(" - ERROR: instance %s, connection to primary node"
1060 " %s failed" % (instance, pnode))
1063 if pnode in n_offline:
1064 inst_nodes_offline.append(pnode)
1066 # If the instance is non-redundant we cannot survive losing its primary
1067 # node, so we are not N+1 compliant. On the other hand we have no disk
1068 # templates with more than one secondary so that situation is not well
1070 # FIXME: does not support file-backed instances
1071 if len(inst_config.secondary_nodes) == 0:
1072 i_non_redundant.append(instance)
1073 elif len(inst_config.secondary_nodes) > 1:
1074 feedback_fn(" - WARNING: multiple secondaries for instance %s"
1077 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1078 i_non_a_balanced.append(instance)
1080 for snode in inst_config.secondary_nodes:
1081 if snode in node_info:
1082 node_info[snode]['sinst'].append(instance)
1083 if pnode not in node_info[snode]['sinst-by-pnode']:
1084 node_info[snode]['sinst-by-pnode'][pnode] = []
1085 node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1086 elif snode not in n_offline:
1087 feedback_fn(" - ERROR: instance %s, connection to secondary node"
1088 " %s failed" % (instance, snode))
1090 if snode in n_offline:
1091 inst_nodes_offline.append(snode)
1093 if inst_nodes_offline:
1094 # warn that the instance lives on offline nodes, and set bad=True
1095 feedback_fn(" - ERROR: instance lives on offline node(s) %s" %
1096 ", ".join(inst_nodes_offline))
1099 feedback_fn("* Verifying orphan volumes")
1100 result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
1104 feedback_fn("* Verifying remaining instances")
1105 result = self._VerifyOrphanInstances(instancelist, node_instance,
1109 if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1110 feedback_fn("* Verifying N+1 Memory redundancy")
1111 result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
1114 feedback_fn("* Other Notes")
1116 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
1117 % len(i_non_redundant))
1119 if i_non_a_balanced:
1120 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
1121 % len(i_non_a_balanced))
1124 feedback_fn(" - NOTICE: %d offline node(s) found." % len(n_offline))
1127 feedback_fn(" - NOTICE: %d drained node(s) found." % len(n_drained))
1131 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1132 """Analize the post-hooks' result
1134 This method analyses the hook result, handles it, and sends some
1135 nicely-formatted feedback back to the user.
1137 @param phase: one of L{constants.HOOKS_PHASE_POST} or
1138 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1139 @param hooks_results: the results of the multi-node hooks rpc call
1140 @param feedback_fn: function used send feedback back to the caller
1141 @param lu_result: previous Exec result
1142 @return: the new Exec result, based on the previous result
1146 # We only really run POST phase hooks, and are only interested in
1148 if phase == constants.HOOKS_PHASE_POST:
1149 # Used to change hooks' output to proper indentation
1150 indent_re = re.compile('^', re.M)
1151 feedback_fn("* Hooks Results")
1152 if not hooks_results:
1153 feedback_fn(" - ERROR: general communication failure")
1156 for node_name in hooks_results:
1157 show_node_header = True
1158 res = hooks_results[node_name]
1159 if res.failed or res.data is False or not isinstance(res.data, list):
1161 # no need to warn or set fail return value
1163 feedback_fn(" Communication failure in hooks execution")
1166 for script, hkr, output in res.data:
1167 if hkr == constants.HKR_FAIL:
1168 # The node header is only shown once, if there are
1169 # failing hooks on that node
1170 if show_node_header:
1171 feedback_fn(" Node %s:" % node_name)
1172 show_node_header = False
1173 feedback_fn(" ERROR: Script %s failed, output:" % script)
1174 output = indent_re.sub(' ', output)
1175 feedback_fn("%s" % output)
1181 class LUVerifyDisks(NoHooksLU):
1182 """Verifies the cluster disks status.
1188 def ExpandNames(self):
1189 self.needed_locks = {
1190 locking.LEVEL_NODE: locking.ALL_SET,
1191 locking.LEVEL_INSTANCE: locking.ALL_SET,
1193 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1195 def CheckPrereq(self):
1196 """Check prerequisites.
1198 This has no prerequisites.
1203 def Exec(self, feedback_fn):
1204 """Verify integrity of cluster disks.
1207 result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1209 vg_name = self.cfg.GetVGName()
1210 nodes = utils.NiceSort(self.cfg.GetNodeList())
1211 instances = [self.cfg.GetInstanceInfo(name)
1212 for name in self.cfg.GetInstanceList()]
1215 for inst in instances:
1217 if (not inst.admin_up or
1218 inst.disk_template not in constants.DTS_NET_MIRROR):
1220 inst.MapLVsByNode(inst_lvs)
1221 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1222 for node, vol_list in inst_lvs.iteritems():
1223 for vol in vol_list:
1224 nv_dict[(node, vol)] = inst
1229 node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1234 lvs = node_lvs[node]
1237 self.LogWarning("Connection to node %s failed: %s" %
1241 if isinstance(lvs, basestring):
1242 logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1243 res_nlvm[node] = lvs
1244 elif not isinstance(lvs, dict):
1245 logging.warning("Connection to node %s failed or invalid data"
1247 res_nodes.append(node)
1250 for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1251 inst = nv_dict.pop((node, lv_name), None)
1252 if (not lv_online and inst is not None
1253 and inst.name not in res_instances):
1254 res_instances.append(inst.name)
1256 # any leftover items in nv_dict are missing LVs, let's arrange the
1258 for key, inst in nv_dict.iteritems():
1259 if inst.name not in res_missing:
1260 res_missing[inst.name] = []
1261 res_missing[inst.name].append(key)
1266 class LURenameCluster(LogicalUnit):
1267 """Rename the cluster.
1270 HPATH = "cluster-rename"
1271 HTYPE = constants.HTYPE_CLUSTER
1274 def BuildHooksEnv(self):
1279 "OP_TARGET": self.cfg.GetClusterName(),
1280 "NEW_NAME": self.op.name,
1282 mn = self.cfg.GetMasterNode()
1283 return env, [mn], [mn]
1285 def CheckPrereq(self):
1286 """Verify that the passed name is a valid one.
1289 hostname = utils.HostInfo(self.op.name)
1291 new_name = hostname.name
1292 self.ip = new_ip = hostname.ip
1293 old_name = self.cfg.GetClusterName()
1294 old_ip = self.cfg.GetMasterIP()
1295 if new_name == old_name and new_ip == old_ip:
1296 raise errors.OpPrereqError("Neither the name nor the IP address of the"
1297 " cluster has changed")
1298 if new_ip != old_ip:
1299 if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1300 raise errors.OpPrereqError("The given cluster IP address (%s) is"
1301 " reachable on the network. Aborting." %
1304 self.op.name = new_name
1306 def Exec(self, feedback_fn):
1307 """Rename the cluster.
1310 clustername = self.op.name
1313 # shutdown the master IP
1314 master = self.cfg.GetMasterNode()
1315 result = self.rpc.call_node_stop_master(master, False)
1316 if result.failed or not result.data:
1317 raise errors.OpExecError("Could not disable the master role")
1320 cluster = self.cfg.GetClusterInfo()
1321 cluster.cluster_name = clustername
1322 cluster.master_ip = ip
1323 self.cfg.Update(cluster)
1325 # update the known hosts file
1326 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1327 node_list = self.cfg.GetNodeList()
1329 node_list.remove(master)
1332 result = self.rpc.call_upload_file(node_list,
1333 constants.SSH_KNOWN_HOSTS_FILE)
1334 for to_node, to_result in result.iteritems():
1335 if to_result.failed or not to_result.data:
1336 logging.error("Copy of file %s to node %s failed",
1337 constants.SSH_KNOWN_HOSTS_FILE, to_node)
1340 result = self.rpc.call_node_start_master(master, False)
1341 if result.failed or not result.data:
1342 self.LogWarning("Could not re-enable the master role on"
1343 " the master, please restart manually.")
1346 def _RecursiveCheckIfLVMBased(disk):
1347 """Check if the given disk or its children are lvm-based.
1349 @type disk: L{objects.Disk}
1350 @param disk: the disk to check
1352 @return: boolean indicating whether a LD_LV dev_type was found or not
1356 for chdisk in disk.children:
1357 if _RecursiveCheckIfLVMBased(chdisk):
1359 return disk.dev_type == constants.LD_LV
1362 class LUSetClusterParams(LogicalUnit):
1363 """Change the parameters of the cluster.
1366 HPATH = "cluster-modify"
1367 HTYPE = constants.HTYPE_CLUSTER
1371 def CheckParameters(self):
1375 if not hasattr(self.op, "candidate_pool_size"):
1376 self.op.candidate_pool_size = None
1377 if self.op.candidate_pool_size is not None:
1379 self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1380 except ValueError, err:
1381 raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1383 if self.op.candidate_pool_size < 1:
1384 raise errors.OpPrereqError("At least one master candidate needed")
1386 def ExpandNames(self):
1387 # FIXME: in the future maybe other cluster params won't require checking on
1388 # all nodes to be modified.
1389 self.needed_locks = {
1390 locking.LEVEL_NODE: locking.ALL_SET,
1392 self.share_locks[locking.LEVEL_NODE] = 1
1394 def BuildHooksEnv(self):
1399 "OP_TARGET": self.cfg.GetClusterName(),
1400 "NEW_VG_NAME": self.op.vg_name,
1402 mn = self.cfg.GetMasterNode()
1403 return env, [mn], [mn]
1405 def CheckPrereq(self):
1406 """Check prerequisites.
1408 This checks whether the given params don't conflict and
1409 if the given volume group is valid.
1412 if self.op.vg_name is not None and not self.op.vg_name:
1413 instances = self.cfg.GetAllInstancesInfo().values()
1414 for inst in instances:
1415 for disk in inst.disks:
1416 if _RecursiveCheckIfLVMBased(disk):
1417 raise errors.OpPrereqError("Cannot disable lvm storage while"
1418 " lvm-based instances exist")
1420 node_list = self.acquired_locks[locking.LEVEL_NODE]
1422 # if vg_name not None, checks given volume group on all nodes
1424 vglist = self.rpc.call_vg_list(node_list)
1425 for node in node_list:
1426 if vglist[node].failed:
1427 # ignoring down node
1428 self.LogWarning("Node %s unreachable/error, ignoring" % node)
1430 vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
1432 constants.MIN_VG_SIZE)
1434 raise errors.OpPrereqError("Error on node '%s': %s" %
1437 self.cluster = cluster = self.cfg.GetClusterInfo()
1438 # validate beparams changes
1439 if self.op.beparams:
1440 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
1441 self.new_beparams = cluster.FillDict(
1442 cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1444 # hypervisor list/parameters
1445 self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1446 if self.op.hvparams:
1447 if not isinstance(self.op.hvparams, dict):
1448 raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1449 for hv_name, hv_dict in self.op.hvparams.items():
1450 if hv_name not in self.new_hvparams:
1451 self.new_hvparams[hv_name] = hv_dict
1453 self.new_hvparams[hv_name].update(hv_dict)
1455 if self.op.enabled_hypervisors is not None:
1456 self.hv_list = self.op.enabled_hypervisors
1458 self.hv_list = cluster.enabled_hypervisors
1460 if self.op.hvparams or self.op.enabled_hypervisors is not None:
1461 # either the enabled list has changed, or the parameters have, validate
1462 for hv_name, hv_params in self.new_hvparams.items():
1463 if ((self.op.hvparams and hv_name in self.op.hvparams) or
1464 (self.op.enabled_hypervisors and
1465 hv_name in self.op.enabled_hypervisors)):
1466 # either this is a new hypervisor, or its parameters have changed
1467 hv_class = hypervisor.GetHypervisor(hv_name)
1468 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1469 hv_class.CheckParameterSyntax(hv_params)
1470 _CheckHVParams(self, node_list, hv_name, hv_params)
1472 def Exec(self, feedback_fn):
1473 """Change the parameters of the cluster.
1476 if self.op.vg_name is not None:
1477 if self.op.vg_name != self.cfg.GetVGName():
1478 self.cfg.SetVGName(self.op.vg_name)
1480 feedback_fn("Cluster LVM configuration already in desired"
1481 " state, not changing")
1482 if self.op.hvparams:
1483 self.cluster.hvparams = self.new_hvparams
1484 if self.op.enabled_hypervisors is not None:
1485 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1486 if self.op.beparams:
1487 self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1488 if self.op.candidate_pool_size is not None:
1489 self.cluster.candidate_pool_size = self.op.candidate_pool_size
1491 self.cfg.Update(self.cluster)
1493 # we want to update nodes after the cluster so that if any errors
1494 # happen, we have recorded and saved the cluster info
1495 if self.op.candidate_pool_size is not None:
1496 _AdjustCandidatePool(self)
1499 class LURedistributeConfig(NoHooksLU):
1500 """Force the redistribution of cluster configuration.
1502 This is a very simple LU.
1508 def ExpandNames(self):
1509 self.needed_locks = {
1510 locking.LEVEL_NODE: locking.ALL_SET,
1512 self.share_locks[locking.LEVEL_NODE] = 1
1514 def CheckPrereq(self):
1515 """Check prerequisites.
1519 def Exec(self, feedback_fn):
1520 """Redistribute the configuration.
1523 self.cfg.Update(self.cfg.GetClusterInfo())
1526 def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1527 """Sleep and poll for an instance's disk to sync.
1530 if not instance.disks:
1534 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1536 node = instance.primary_node
1538 for dev in instance.disks:
1539 lu.cfg.SetDiskID(dev, node)
1545 cumul_degraded = False
1546 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1547 if rstats.failed or not rstats.data:
1548 lu.LogWarning("Can't get any data from node %s", node)
1551 raise errors.RemoteError("Can't contact node %s for mirror data,"
1552 " aborting." % node)
1555 rstats = rstats.data
1557 for i, mstat in enumerate(rstats):
1559 lu.LogWarning("Can't compute data for node %s/%s",
1560 node, instance.disks[i].iv_name)
1562 # we ignore the ldisk parameter
1563 perc_done, est_time, is_degraded, _ = mstat
1564 cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1565 if perc_done is not None:
1567 if est_time is not None:
1568 rem_time = "%d estimated seconds remaining" % est_time
1571 rem_time = "no time estimate"
1572 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1573 (instance.disks[i].iv_name, perc_done, rem_time))
1577 time.sleep(min(60, max_time))
1580 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1581 return not cumul_degraded
1584 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1585 """Check that mirrors are not degraded.
1587 The ldisk parameter, if True, will change the test from the
1588 is_degraded attribute (which represents overall non-ok status for
1589 the device(s)) to the ldisk (representing the local storage status).
1592 lu.cfg.SetDiskID(dev, node)
1599 if on_primary or dev.AssembleOnSecondary():
1600 rstats = lu.rpc.call_blockdev_find(node, dev)
1601 msg = rstats.RemoteFailMsg()
1603 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
1605 elif not rstats.payload:
1606 lu.LogWarning("Can't find disk on node %s", node)
1609 result = result and (not rstats.payload[idx])
1611 for child in dev.children:
1612 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1617 class LUDiagnoseOS(NoHooksLU):
1618 """Logical unit for OS diagnose/query.
1621 _OP_REQP = ["output_fields", "names"]
1623 _FIELDS_STATIC = utils.FieldSet()
1624 _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1626 def ExpandNames(self):
1628 raise errors.OpPrereqError("Selective OS query not supported")
1630 _CheckOutputFields(static=self._FIELDS_STATIC,
1631 dynamic=self._FIELDS_DYNAMIC,
1632 selected=self.op.output_fields)
1634 # Lock all nodes, in shared mode
1635 self.needed_locks = {}
1636 self.share_locks[locking.LEVEL_NODE] = 1
1637 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1639 def CheckPrereq(self):
1640 """Check prerequisites.
1645 def _DiagnoseByOS(node_list, rlist):
1646 """Remaps a per-node return list into an a per-os per-node dictionary
1648 @param node_list: a list with the names of all nodes
1649 @param rlist: a map with node names as keys and OS objects as values
1652 @returns: a dictionary with osnames as keys and as value another map, with
1653 nodes as keys and list of OS objects as values, eg::
1655 {"debian-etch": {"node1": [<object>,...],
1656 "node2": [<object>,]}
1661 for node_name, nr in rlist.iteritems():
1662 if nr.failed or not nr.data:
1664 for os_obj in nr.data:
1665 if os_obj.name not in all_os:
1666 # build a list of nodes for this os containing empty lists
1667 # for each node in node_list
1668 all_os[os_obj.name] = {}
1669 for nname in node_list:
1670 all_os[os_obj.name][nname] = []
1671 all_os[os_obj.name][node_name].append(os_obj)
1674 def Exec(self, feedback_fn):
1675 """Compute the list of OSes.
1678 node_list = self.acquired_locks[locking.LEVEL_NODE]
1679 valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
1680 if node in node_list]
1681 node_data = self.rpc.call_os_diagnose(valid_nodes)
1682 if node_data == False:
1683 raise errors.OpExecError("Can't gather the list of OSes")
1684 pol = self._DiagnoseByOS(valid_nodes, node_data)
1686 for os_name, os_data in pol.iteritems():
1688 for field in self.op.output_fields:
1691 elif field == "valid":
1692 val = utils.all([osl and osl[0] for osl in os_data.values()])
1693 elif field == "node_status":
1695 for node_name, nos_list in os_data.iteritems():
1696 val[node_name] = [(v.status, v.path) for v in nos_list]
1698 raise errors.ParameterError(field)
1705 class LURemoveNode(LogicalUnit):
1706 """Logical unit for removing a node.
1709 HPATH = "node-remove"
1710 HTYPE = constants.HTYPE_NODE
1711 _OP_REQP = ["node_name"]
1713 def BuildHooksEnv(self):
1716 This doesn't run on the target node in the pre phase as a failed
1717 node would then be impossible to remove.
1721 "OP_TARGET": self.op.node_name,
1722 "NODE_NAME": self.op.node_name,
1724 all_nodes = self.cfg.GetNodeList()
1725 all_nodes.remove(self.op.node_name)
1726 return env, all_nodes, all_nodes
1728 def CheckPrereq(self):
1729 """Check prerequisites.
1732 - the node exists in the configuration
1733 - it does not have primary or secondary instances
1734 - it's not the master
1736 Any errors are signalled by raising errors.OpPrereqError.
1739 node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1741 raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1743 instance_list = self.cfg.GetInstanceList()
1745 masternode = self.cfg.GetMasterNode()
1746 if node.name == masternode:
1747 raise errors.OpPrereqError("Node is the master node,"
1748 " you need to failover first.")
1750 for instance_name in instance_list:
1751 instance = self.cfg.GetInstanceInfo(instance_name)
1752 if node.name in instance.all_nodes:
1753 raise errors.OpPrereqError("Instance %s is still running on the node,"
1754 " please remove first." % instance_name)
1755 self.op.node_name = node.name
1758 def Exec(self, feedback_fn):
1759 """Removes the node from the cluster.
1763 logging.info("Stopping the node daemon and removing configs from node %s",
1766 self.context.RemoveNode(node.name)
1768 self.rpc.call_node_leave_cluster(node.name)
1770 # Promote nodes to master candidate as needed
1771 _AdjustCandidatePool(self)
1774 class LUQueryNodes(NoHooksLU):
1775 """Logical unit for querying nodes.
1778 _OP_REQP = ["output_fields", "names", "use_locking"]
1780 _FIELDS_DYNAMIC = utils.FieldSet(
1782 "mtotal", "mnode", "mfree",
1784 "ctotal", "cnodes", "csockets",
1787 _FIELDS_STATIC = utils.FieldSet(
1788 "name", "pinst_cnt", "sinst_cnt",
1789 "pinst_list", "sinst_list",
1790 "pip", "sip", "tags",
1798 def ExpandNames(self):
1799 _CheckOutputFields(static=self._FIELDS_STATIC,
1800 dynamic=self._FIELDS_DYNAMIC,
1801 selected=self.op.output_fields)
1803 self.needed_locks = {}
1804 self.share_locks[locking.LEVEL_NODE] = 1
1807 self.wanted = _GetWantedNodes(self, self.op.names)
1809 self.wanted = locking.ALL_SET
1811 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1812 self.do_locking = self.do_node_query and self.op.use_locking
1814 # if we don't request only static fields, we need to lock the nodes
1815 self.needed_locks[locking.LEVEL_NODE] = self.wanted
1818 def CheckPrereq(self):
1819 """Check prerequisites.
1822 # The validation of the node list is done in the _GetWantedNodes,
1823 # if non empty, and if empty, there's no validation to do
1826 def Exec(self, feedback_fn):
1827 """Computes the list of nodes and their attributes.
1830 all_info = self.cfg.GetAllNodesInfo()
1832 nodenames = self.acquired_locks[locking.LEVEL_NODE]
1833 elif self.wanted != locking.ALL_SET:
1834 nodenames = self.wanted
1835 missing = set(nodenames).difference(all_info.keys())
1837 raise errors.OpExecError(
1838 "Some nodes were removed before retrieving their data: %s" % missing)
1840 nodenames = all_info.keys()
1842 nodenames = utils.NiceSort(nodenames)
1843 nodelist = [all_info[name] for name in nodenames]
1845 # begin data gathering
1847 if self.do_node_query:
1849 node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1850 self.cfg.GetHypervisorType())
1851 for name in nodenames:
1852 nodeinfo = node_data[name]
1853 if not nodeinfo.failed and nodeinfo.data:
1854 nodeinfo = nodeinfo.data
1855 fn = utils.TryConvert
1857 "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1858 "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1859 "mfree": fn(int, nodeinfo.get('memory_free', None)),
1860 "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1861 "dfree": fn(int, nodeinfo.get('vg_free', None)),
1862 "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1863 "bootid": nodeinfo.get('bootid', None),
1864 "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
1865 "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
1868 live_data[name] = {}
1870 live_data = dict.fromkeys(nodenames, {})
1872 node_to_primary = dict([(name, set()) for name in nodenames])
1873 node_to_secondary = dict([(name, set()) for name in nodenames])
1875 inst_fields = frozenset(("pinst_cnt", "pinst_list",
1876 "sinst_cnt", "sinst_list"))
1877 if inst_fields & frozenset(self.op.output_fields):
1878 instancelist = self.cfg.GetInstanceList()
1880 for instance_name in instancelist:
1881 inst = self.cfg.GetInstanceInfo(instance_name)
1882 if inst.primary_node in node_to_primary:
1883 node_to_primary[inst.primary_node].add(inst.name)
1884 for secnode in inst.secondary_nodes:
1885 if secnode in node_to_secondary:
1886 node_to_secondary[secnode].add(inst.name)
1888 master_node = self.cfg.GetMasterNode()
1890 # end data gathering
1893 for node in nodelist:
1895 for field in self.op.output_fields:
1898 elif field == "pinst_list":
1899 val = list(node_to_primary[node.name])
1900 elif field == "sinst_list":
1901 val = list(node_to_secondary[node.name])
1902 elif field == "pinst_cnt":
1903 val = len(node_to_primary[node.name])
1904 elif field == "sinst_cnt":
1905 val = len(node_to_secondary[node.name])
1906 elif field == "pip":
1907 val = node.primary_ip
1908 elif field == "sip":
1909 val = node.secondary_ip
1910 elif field == "tags":
1911 val = list(node.GetTags())
1912 elif field == "serial_no":
1913 val = node.serial_no
1914 elif field == "master_candidate":
1915 val = node.master_candidate
1916 elif field == "master":
1917 val = node.name == master_node
1918 elif field == "offline":
1920 elif field == "drained":
1922 elif self._FIELDS_DYNAMIC.Matches(field):
1923 val = live_data[node.name].get(field, None)
1925 raise errors.ParameterError(field)
1926 node_output.append(val)
1927 output.append(node_output)
1932 class LUQueryNodeVolumes(NoHooksLU):
1933 """Logical unit for getting volumes on node(s).
1936 _OP_REQP = ["nodes", "output_fields"]
1938 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1939 _FIELDS_STATIC = utils.FieldSet("node")
1941 def ExpandNames(self):
1942 _CheckOutputFields(static=self._FIELDS_STATIC,
1943 dynamic=self._FIELDS_DYNAMIC,
1944 selected=self.op.output_fields)
1946 self.needed_locks = {}
1947 self.share_locks[locking.LEVEL_NODE] = 1
1948 if not self.op.nodes:
1949 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1951 self.needed_locks[locking.LEVEL_NODE] = \
1952 _GetWantedNodes(self, self.op.nodes)
1954 def CheckPrereq(self):
1955 """Check prerequisites.
1957 This checks that the fields required are valid output fields.
1960 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1962 def Exec(self, feedback_fn):
1963 """Computes the list of nodes and their attributes.
1966 nodenames = self.nodes
1967 volumes = self.rpc.call_node_volumes(nodenames)
1969 ilist = [self.cfg.GetInstanceInfo(iname) for iname
1970 in self.cfg.GetInstanceList()]
1972 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1975 for node in nodenames:
1976 if node not in volumes or volumes[node].failed or not volumes[node].data:
1979 node_vols = volumes[node].data[:]
1980 node_vols.sort(key=lambda vol: vol['dev'])
1982 for vol in node_vols:
1984 for field in self.op.output_fields:
1987 elif field == "phys":
1991 elif field == "name":
1993 elif field == "size":
1994 val = int(float(vol['size']))
1995 elif field == "instance":
1997 if node not in lv_by_node[inst]:
1999 if vol['name'] in lv_by_node[inst][node]:
2005 raise errors.ParameterError(field)
2006 node_output.append(str(val))
2008 output.append(node_output)
2013 class LUAddNode(LogicalUnit):
2014 """Logical unit for adding node to the cluster.
2018 HTYPE = constants.HTYPE_NODE
2019 _OP_REQP = ["node_name"]
2021 def BuildHooksEnv(self):
2024 This will run on all nodes before, and on all nodes + the new node after.
2028 "OP_TARGET": self.op.node_name,
2029 "NODE_NAME": self.op.node_name,
2030 "NODE_PIP": self.op.primary_ip,
2031 "NODE_SIP": self.op.secondary_ip,
2033 nodes_0 = self.cfg.GetNodeList()
2034 nodes_1 = nodes_0 + [self.op.node_name, ]
2035 return env, nodes_0, nodes_1
2037 def CheckPrereq(self):
2038 """Check prerequisites.
2041 - the new node is not already in the config
2043 - its parameters (single/dual homed) matches the cluster
2045 Any errors are signalled by raising errors.OpPrereqError.
2048 node_name = self.op.node_name
2051 dns_data = utils.HostInfo(node_name)
2053 node = dns_data.name
2054 primary_ip = self.op.primary_ip = dns_data.ip
2055 secondary_ip = getattr(self.op, "secondary_ip", None)
2056 if secondary_ip is None:
2057 secondary_ip = primary_ip
2058 if not utils.IsValidIP(secondary_ip):
2059 raise errors.OpPrereqError("Invalid secondary IP given")
2060 self.op.secondary_ip = secondary_ip
2062 node_list = cfg.GetNodeList()
2063 if not self.op.readd and node in node_list:
2064 raise errors.OpPrereqError("Node %s is already in the configuration" %
2066 elif self.op.readd and node not in node_list:
2067 raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2069 for existing_node_name in node_list:
2070 existing_node = cfg.GetNodeInfo(existing_node_name)
2072 if self.op.readd and node == existing_node_name:
2073 if (existing_node.primary_ip != primary_ip or
2074 existing_node.secondary_ip != secondary_ip):
2075 raise errors.OpPrereqError("Readded node doesn't have the same IP"
2076 " address configuration as before")
2079 if (existing_node.primary_ip == primary_ip or
2080 existing_node.secondary_ip == primary_ip or
2081 existing_node.primary_ip == secondary_ip or
2082 existing_node.secondary_ip == secondary_ip):
2083 raise errors.OpPrereqError("New node ip address(es) conflict with"
2084 " existing node %s" % existing_node.name)
2086 # check that the type of the node (single versus dual homed) is the
2087 # same as for the master
2088 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2089 master_singlehomed = myself.secondary_ip == myself.primary_ip
2090 newbie_singlehomed = secondary_ip == primary_ip
2091 if master_singlehomed != newbie_singlehomed:
2092 if master_singlehomed:
2093 raise errors.OpPrereqError("The master has no private ip but the"
2094 " new node has one")
2096 raise errors.OpPrereqError("The master has a private ip but the"
2097 " new node doesn't have one")
2099 # checks reachablity
2100 if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2101 raise errors.OpPrereqError("Node not reachable by ping")
2103 if not newbie_singlehomed:
2104 # check reachability from my secondary ip to newbie's secondary ip
2105 if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2106 source=myself.secondary_ip):
2107 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2108 " based ping to noded port")
2110 cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2111 mc_now, _ = self.cfg.GetMasterCandidateStats()
2112 master_candidate = mc_now < cp_size
2114 self.new_node = objects.Node(name=node,
2115 primary_ip=primary_ip,
2116 secondary_ip=secondary_ip,
2117 master_candidate=master_candidate,
2118 offline=False, drained=False)
2120 def Exec(self, feedback_fn):
2121 """Adds the new node to the cluster.
2124 new_node = self.new_node
2125 node = new_node.name
2127 # check connectivity
2128 result = self.rpc.call_version([node])[node]
2131 if constants.PROTOCOL_VERSION == result.data:
2132 logging.info("Communication to node %s fine, sw version %s match",
2135 raise errors.OpExecError("Version mismatch master version %s,"
2136 " node version %s" %
2137 (constants.PROTOCOL_VERSION, result.data))
2139 raise errors.OpExecError("Cannot get version from the new node")
2142 logging.info("Copy ssh key to node %s", node)
2143 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2145 keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2146 constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2152 keyarray.append(f.read())
2156 result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2158 keyarray[3], keyarray[4], keyarray[5])
2160 msg = result.RemoteFailMsg()
2162 raise errors.OpExecError("Cannot transfer ssh keys to the"
2163 " new node: %s" % msg)
2165 # Add node to our /etc/hosts, and add key to known_hosts
2166 utils.AddHostToEtcHosts(new_node.name)
2168 if new_node.secondary_ip != new_node.primary_ip:
2169 result = self.rpc.call_node_has_ip_address(new_node.name,
2170 new_node.secondary_ip)
2171 if result.failed or not result.data:
2172 raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2173 " you gave (%s). Please fix and re-run this"
2174 " command." % new_node.secondary_ip)
2176 node_verify_list = [self.cfg.GetMasterNode()]
2177 node_verify_param = {
2179 # TODO: do a node-net-test as well?
2182 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2183 self.cfg.GetClusterName())
2184 for verifier in node_verify_list:
2185 if result[verifier].failed or not result[verifier].data:
2186 raise errors.OpExecError("Cannot communicate with %s's node daemon"
2187 " for remote verification" % verifier)
2188 if result[verifier].data['nodelist']:
2189 for failed in result[verifier].data['nodelist']:
2190 feedback_fn("ssh/hostname verification failed %s -> %s" %
2191 (verifier, result[verifier].data['nodelist'][failed]))
2192 raise errors.OpExecError("ssh/hostname verification failed.")
2194 # Distribute updated /etc/hosts and known_hosts to all nodes,
2195 # including the node just added
2196 myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2197 dist_nodes = self.cfg.GetNodeList()
2198 if not self.op.readd:
2199 dist_nodes.append(node)
2200 if myself.name in dist_nodes:
2201 dist_nodes.remove(myself.name)
2203 logging.debug("Copying hosts and known_hosts to all nodes")
2204 for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2205 result = self.rpc.call_upload_file(dist_nodes, fname)
2206 for to_node, to_result in result.iteritems():
2207 if to_result.failed or not to_result.data:
2208 logging.error("Copy of file %s to node %s failed", fname, to_node)
2211 enabled_hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2212 if constants.HTS_COPY_VNC_PASSWORD.intersection(enabled_hypervisors):
2213 to_copy.append(constants.VNC_PASSWORD_FILE)
2215 for fname in to_copy:
2216 result = self.rpc.call_upload_file([node], fname)
2217 if result[node].failed or not result[node]:
2218 logging.error("Could not copy file %s to node %s", fname, node)
2221 self.context.ReaddNode(new_node)
2223 self.context.AddNode(new_node)
2226 class LUSetNodeParams(LogicalUnit):
2227 """Modifies the parameters of a node.
2230 HPATH = "node-modify"
2231 HTYPE = constants.HTYPE_NODE
2232 _OP_REQP = ["node_name"]
2235 def CheckArguments(self):
2236 node_name = self.cfg.ExpandNodeName(self.op.node_name)
2237 if node_name is None:
2238 raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2239 self.op.node_name = node_name
2240 _CheckBooleanOpField(self.op, 'master_candidate')
2241 _CheckBooleanOpField(self.op, 'offline')
2242 _CheckBooleanOpField(self.op, 'drained')
2243 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
2244 if all_mods.count(None) == 3:
2245 raise errors.OpPrereqError("Please pass at least one modification")
2246 if all_mods.count(True) > 1:
2247 raise errors.OpPrereqError("Can't set the node into more than one"
2248 " state at the same time")
2250 def ExpandNames(self):
2251 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2253 def BuildHooksEnv(self):
2256 This runs on the master node.
2260 "OP_TARGET": self.op.node_name,
2261 "MASTER_CANDIDATE": str(self.op.master_candidate),
2262 "OFFLINE": str(self.op.offline),
2263 "DRAINED": str(self.op.drained),
2265 nl = [self.cfg.GetMasterNode(),
2269 def CheckPrereq(self):
2270 """Check prerequisites.
2272 This only checks the instance list against the existing names.
2275 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2277 if ((self.op.master_candidate == False or self.op.offline == True or
2278 self.op.drained == True) and node.master_candidate):
2279 # we will demote the node from master_candidate
2280 if self.op.node_name == self.cfg.GetMasterNode():
2281 raise errors.OpPrereqError("The master node has to be a"
2282 " master candidate, online and not drained")
2283 cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2284 num_candidates, _ = self.cfg.GetMasterCandidateStats()
2285 if num_candidates <= cp_size:
2286 msg = ("Not enough master candidates (desired"
2287 " %d, new value will be %d)" % (cp_size, num_candidates-1))
2289 self.LogWarning(msg)
2291 raise errors.OpPrereqError(msg)
2293 if (self.op.master_candidate == True and
2294 ((node.offline and not self.op.offline == False) or
2295 (node.drained and not self.op.drained == False))):
2296 raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
2297 " to master_candidate")
2301 def Exec(self, feedback_fn):
2310 if self.op.offline is not None:
2311 node.offline = self.op.offline
2312 result.append(("offline", str(self.op.offline)))
2313 if self.op.offline == True:
2314 if node.master_candidate:
2315 node.master_candidate = False
2317 result.append(("master_candidate", "auto-demotion due to offline"))
2319 node.drained = False
2320 result.append(("drained", "clear drained status due to offline"))
2322 if self.op.master_candidate is not None:
2323 node.master_candidate = self.op.master_candidate
2325 result.append(("master_candidate", str(self.op.master_candidate)))
2326 if self.op.master_candidate == False:
2327 rrc = self.rpc.call_node_demote_from_mc(node.name)
2328 msg = rrc.RemoteFailMsg()
2330 self.LogWarning("Node failed to demote itself: %s" % msg)
2332 if self.op.drained is not None:
2333 node.drained = self.op.drained
2334 result.append(("drained", str(self.op.drained)))
2335 if self.op.drained == True:
2336 if node.master_candidate:
2337 node.master_candidate = False
2339 result.append(("master_candidate", "auto-demotion due to drain"))
2341 node.offline = False
2342 result.append(("offline", "clear offline status due to drain"))
2344 # this will trigger configuration file update, if needed
2345 self.cfg.Update(node)
2346 # this will trigger job queue propagation or cleanup
2348 self.context.ReaddNode(node)
2353 class LUQueryClusterInfo(NoHooksLU):
2354 """Query cluster configuration.
2360 def ExpandNames(self):
2361 self.needed_locks = {}
2363 def CheckPrereq(self):
2364 """No prerequsites needed for this LU.
2369 def Exec(self, feedback_fn):
2370 """Return cluster config.
2373 cluster = self.cfg.GetClusterInfo()
2375 "software_version": constants.RELEASE_VERSION,
2376 "protocol_version": constants.PROTOCOL_VERSION,
2377 "config_version": constants.CONFIG_VERSION,
2378 "os_api_version": constants.OS_API_VERSION,
2379 "export_version": constants.EXPORT_VERSION,
2380 "architecture": (platform.architecture()[0], platform.machine()),
2381 "name": cluster.cluster_name,
2382 "master": cluster.master_node,
2383 "default_hypervisor": cluster.default_hypervisor,
2384 "enabled_hypervisors": cluster.enabled_hypervisors,
2385 "hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
2386 for hypervisor in cluster.enabled_hypervisors]),
2387 "beparams": cluster.beparams,
2388 "candidate_pool_size": cluster.candidate_pool_size,
2394 class LUQueryConfigValues(NoHooksLU):
2395 """Return configuration values.
2400 _FIELDS_DYNAMIC = utils.FieldSet()
2401 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2403 def ExpandNames(self):
2404 self.needed_locks = {}
2406 _CheckOutputFields(static=self._FIELDS_STATIC,
2407 dynamic=self._FIELDS_DYNAMIC,
2408 selected=self.op.output_fields)
2410 def CheckPrereq(self):
2411 """No prerequisites.
2416 def Exec(self, feedback_fn):
2417 """Dump a representation of the cluster config to the standard output.
2421 for field in self.op.output_fields:
2422 if field == "cluster_name":
2423 entry = self.cfg.GetClusterName()
2424 elif field == "master_node":
2425 entry = self.cfg.GetMasterNode()
2426 elif field == "drain_flag":
2427 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2429 raise errors.ParameterError(field)
2430 values.append(entry)
2434 class LUActivateInstanceDisks(NoHooksLU):
2435 """Bring up an instance's disks.
2438 _OP_REQP = ["instance_name"]
2441 def ExpandNames(self):
2442 self._ExpandAndLockInstance()
2443 self.needed_locks[locking.LEVEL_NODE] = []
2444 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2446 def DeclareLocks(self, level):
2447 if level == locking.LEVEL_NODE:
2448 self._LockInstancesNodes()
2450 def CheckPrereq(self):
2451 """Check prerequisites.
2453 This checks that the instance is in the cluster.
2456 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2457 assert self.instance is not None, \
2458 "Cannot retrieve locked instance %s" % self.op.instance_name
2459 _CheckNodeOnline(self, self.instance.primary_node)
2461 def Exec(self, feedback_fn):
2462 """Activate the disks.
2465 disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2467 raise errors.OpExecError("Cannot activate block devices")
2472 def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2473 """Prepare the block devices for an instance.
2475 This sets up the block devices on all nodes.
2477 @type lu: L{LogicalUnit}
2478 @param lu: the logical unit on whose behalf we execute
2479 @type instance: L{objects.Instance}
2480 @param instance: the instance for whose disks we assemble
2481 @type ignore_secondaries: boolean
2482 @param ignore_secondaries: if true, errors on secondary nodes
2483 won't result in an error return from the function
2484 @return: False if the operation failed, otherwise a list of
2485 (host, instance_visible_name, node_visible_name)
2486 with the mapping from node devices to instance devices
2491 iname = instance.name
2492 # With the two passes mechanism we try to reduce the window of
2493 # opportunity for the race condition of switching DRBD to primary
2494 # before handshaking occured, but we do not eliminate it
2496 # The proper fix would be to wait (with some limits) until the
2497 # connection has been made and drbd transitions from WFConnection
2498 # into any other network-connected state (Connected, SyncTarget,
2501 # 1st pass, assemble on all nodes in secondary mode
2502 for inst_disk in instance.disks:
2503 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2504 lu.cfg.SetDiskID(node_disk, node)
2505 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2506 msg = result.RemoteFailMsg()
2508 lu.proc.LogWarning("Could not prepare block device %s on node %s"
2509 " (is_primary=False, pass=1): %s",
2510 inst_disk.iv_name, node, msg)
2511 if not ignore_secondaries:
2514 # FIXME: race condition on drbd migration to primary
2516 # 2nd pass, do only the primary node
2517 for inst_disk in instance.disks:
2518 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2519 if node != instance.primary_node:
2521 lu.cfg.SetDiskID(node_disk, node)
2522 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2523 msg = result.RemoteFailMsg()
2525 lu.proc.LogWarning("Could not prepare block device %s on node %s"
2526 " (is_primary=True, pass=2): %s",
2527 inst_disk.iv_name, node, msg)
2529 device_info.append((instance.primary_node, inst_disk.iv_name,
2532 # leave the disks configured for the primary node
2533 # this is a workaround that would be fixed better by
2534 # improving the logical/physical id handling
2535 for disk in instance.disks:
2536 lu.cfg.SetDiskID(disk, instance.primary_node)
2538 return disks_ok, device_info
2541 def _StartInstanceDisks(lu, instance, force):
2542 """Start the disks of an instance.
2545 disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2546 ignore_secondaries=force)
2548 _ShutdownInstanceDisks(lu, instance)
2549 if force is not None and not force:
2550 lu.proc.LogWarning("", hint="If the message above refers to a"
2552 " you can retry the operation using '--force'.")
2553 raise errors.OpExecError("Disk consistency error")
2556 class LUDeactivateInstanceDisks(NoHooksLU):
2557 """Shutdown an instance's disks.
2560 _OP_REQP = ["instance_name"]
2563 def ExpandNames(self):
2564 self._ExpandAndLockInstance()
2565 self.needed_locks[locking.LEVEL_NODE] = []
2566 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2568 def DeclareLocks(self, level):
2569 if level == locking.LEVEL_NODE:
2570 self._LockInstancesNodes()
2572 def CheckPrereq(self):
2573 """Check prerequisites.
2575 This checks that the instance is in the cluster.
2578 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2579 assert self.instance is not None, \
2580 "Cannot retrieve locked instance %s" % self.op.instance_name
2582 def Exec(self, feedback_fn):
2583 """Deactivate the disks
2586 instance = self.instance
2587 _SafeShutdownInstanceDisks(self, instance)
2590 def _SafeShutdownInstanceDisks(lu, instance):
2591 """Shutdown block devices of an instance.
2593 This function checks if an instance is running, before calling
2594 _ShutdownInstanceDisks.
2597 ins_l = lu.rpc.call_instance_list([instance.primary_node],
2598 [instance.hypervisor])
2599 ins_l = ins_l[instance.primary_node]
2600 if ins_l.failed or not isinstance(ins_l.data, list):
2601 raise errors.OpExecError("Can't contact node '%s'" %
2602 instance.primary_node)
2604 if instance.name in ins_l.data:
2605 raise errors.OpExecError("Instance is running, can't shutdown"
2608 _ShutdownInstanceDisks(lu, instance)
2611 def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2612 """Shutdown block devices of an instance.
2614 This does the shutdown on all nodes of the instance.
2616 If the ignore_primary is false, errors on the primary node are
2621 for disk in instance.disks:
2622 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2623 lu.cfg.SetDiskID(top_disk, node)
2624 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2625 msg = result.RemoteFailMsg()
2627 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
2628 disk.iv_name, node, msg)
2629 if not ignore_primary or node != instance.primary_node:
2634 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2635 """Checks if a node has enough free memory.
2637 This function check if a given node has the needed amount of free
2638 memory. In case the node has less memory or we cannot get the
2639 information from the node, this function raise an OpPrereqError
2642 @type lu: C{LogicalUnit}
2643 @param lu: a logical unit from which we get configuration data
2645 @param node: the node to check
2646 @type reason: C{str}
2647 @param reason: string to use in the error message
2648 @type requested: C{int}
2649 @param requested: the amount of memory in MiB to check for
2650 @type hypervisor_name: C{str}
2651 @param hypervisor_name: the hypervisor to ask for memory stats
2652 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2653 we cannot check the node
2656 nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2657 nodeinfo[node].Raise()
2658 free_mem = nodeinfo[node].data.get('memory_free')
2659 if not isinstance(free_mem, int):
2660 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2661 " was '%s'" % (node, free_mem))
2662 if requested > free_mem:
2663 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2664 " needed %s MiB, available %s MiB" %
2665 (node, reason, requested, free_mem))
2668 class LUStartupInstance(LogicalUnit):
2669 """Starts an instance.
2672 HPATH = "instance-start"
2673 HTYPE = constants.HTYPE_INSTANCE
2674 _OP_REQP = ["instance_name", "force"]
2677 def ExpandNames(self):
2678 self._ExpandAndLockInstance()
2680 def BuildHooksEnv(self):
2683 This runs on master, primary and secondary nodes of the instance.
2687 "FORCE": self.op.force,
2689 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2690 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2693 def CheckPrereq(self):
2694 """Check prerequisites.
2696 This checks that the instance is in the cluster.
2699 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2700 assert self.instance is not None, \
2701 "Cannot retrieve locked instance %s" % self.op.instance_name
2703 _CheckNodeOnline(self, instance.primary_node)
2705 bep = self.cfg.GetClusterInfo().FillBE(instance)
2706 # check bridges existance
2707 _CheckInstanceBridgesExist(self, instance)
2709 _CheckNodeFreeMemory(self, instance.primary_node,
2710 "starting instance %s" % instance.name,
2711 bep[constants.BE_MEMORY], instance.hypervisor)
2713 def Exec(self, feedback_fn):
2714 """Start the instance.
2717 instance = self.instance
2718 force = self.op.force
2719 extra_args = getattr(self.op, "extra_args", "")
2721 self.cfg.MarkInstanceUp(instance.name)
2723 node_current = instance.primary_node
2725 _StartInstanceDisks(self, instance, force)
2727 result = self.rpc.call_instance_start(node_current, instance, extra_args)
2728 msg = result.RemoteFailMsg()
2730 _ShutdownInstanceDisks(self, instance)
2731 raise errors.OpExecError("Could not start instance: %s" % msg)
2734 class LURebootInstance(LogicalUnit):
2735 """Reboot an instance.
2738 HPATH = "instance-reboot"
2739 HTYPE = constants.HTYPE_INSTANCE
2740 _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2743 def ExpandNames(self):
2744 if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2745 constants.INSTANCE_REBOOT_HARD,
2746 constants.INSTANCE_REBOOT_FULL]:
2747 raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2748 (constants.INSTANCE_REBOOT_SOFT,
2749 constants.INSTANCE_REBOOT_HARD,
2750 constants.INSTANCE_REBOOT_FULL))
2751 self._ExpandAndLockInstance()
2753 def BuildHooksEnv(self):
2756 This runs on master, primary and secondary nodes of the instance.
2760 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2762 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2763 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2766 def CheckPrereq(self):
2767 """Check prerequisites.
2769 This checks that the instance is in the cluster.
2772 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2773 assert self.instance is not None, \
2774 "Cannot retrieve locked instance %s" % self.op.instance_name
2776 _CheckNodeOnline(self, instance.primary_node)
2778 # check bridges existance
2779 _CheckInstanceBridgesExist(self, instance)
2781 def Exec(self, feedback_fn):
2782 """Reboot the instance.
2785 instance = self.instance
2786 ignore_secondaries = self.op.ignore_secondaries
2787 reboot_type = self.op.reboot_type
2788 extra_args = getattr(self.op, "extra_args", "")
2790 node_current = instance.primary_node
2792 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2793 constants.INSTANCE_REBOOT_HARD]:
2794 result = self.rpc.call_instance_reboot(node_current, instance,
2795 reboot_type, extra_args)
2796 msg = result.RemoteFailMsg()
2798 raise errors.OpExecError("Could not reboot instance: %s" % msg)
2800 result = self.rpc.call_instance_shutdown(node_current, instance)
2801 msg = result.RemoteFailMsg()
2803 raise errors.OpExecError("Could not shutdown instance for"
2804 " full reboot: %s" % msg)
2805 _ShutdownInstanceDisks(self, instance)
2806 _StartInstanceDisks(self, instance, ignore_secondaries)
2807 result = self.rpc.call_instance_start(node_current, instance, extra_args)
2808 msg = result.RemoteFailMsg()
2810 _ShutdownInstanceDisks(self, instance)
2811 raise errors.OpExecError("Could not start instance for"
2812 " full reboot: %s" % msg)
2814 self.cfg.MarkInstanceUp(instance.name)
2817 class LUShutdownInstance(LogicalUnit):
2818 """Shutdown an instance.
2821 HPATH = "instance-stop"
2822 HTYPE = constants.HTYPE_INSTANCE
2823 _OP_REQP = ["instance_name"]
2826 def ExpandNames(self):
2827 self._ExpandAndLockInstance()
2829 def BuildHooksEnv(self):
2832 This runs on master, primary and secondary nodes of the instance.
2835 env = _BuildInstanceHookEnvByObject(self, self.instance)
2836 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2839 def CheckPrereq(self):
2840 """Check prerequisites.
2842 This checks that the instance is in the cluster.
2845 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2846 assert self.instance is not None, \
2847 "Cannot retrieve locked instance %s" % self.op.instance_name
2848 _CheckNodeOnline(self, self.instance.primary_node)
2850 def Exec(self, feedback_fn):
2851 """Shutdown the instance.
2854 instance = self.instance
2855 node_current = instance.primary_node
2856 self.cfg.MarkInstanceDown(instance.name)
2857 result = self.rpc.call_instance_shutdown(node_current, instance)
2858 msg = result.RemoteFailMsg()
2860 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
2862 _ShutdownInstanceDisks(self, instance)
2865 class LUReinstallInstance(LogicalUnit):
2866 """Reinstall an instance.
2869 HPATH = "instance-reinstall"
2870 HTYPE = constants.HTYPE_INSTANCE
2871 _OP_REQP = ["instance_name"]
2874 def ExpandNames(self):
2875 self._ExpandAndLockInstance()
2877 def BuildHooksEnv(self):
2880 This runs on master, primary and secondary nodes of the instance.
2883 env = _BuildInstanceHookEnvByObject(self, self.instance)
2884 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2887 def CheckPrereq(self):
2888 """Check prerequisites.
2890 This checks that the instance is in the cluster and is not running.
2893 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2894 assert instance is not None, \
2895 "Cannot retrieve locked instance %s" % self.op.instance_name
2896 _CheckNodeOnline(self, instance.primary_node)
2898 if instance.disk_template == constants.DT_DISKLESS:
2899 raise errors.OpPrereqError("Instance '%s' has no disks" %
2900 self.op.instance_name)
2901 if instance.admin_up:
2902 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2903 self.op.instance_name)
2904 remote_info = self.rpc.call_instance_info(instance.primary_node,
2906 instance.hypervisor)
2907 if remote_info.failed or remote_info.data:
2908 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2909 (self.op.instance_name,
2910 instance.primary_node))
2912 self.op.os_type = getattr(self.op, "os_type", None)
2913 if self.op.os_type is not None:
2915 pnode = self.cfg.GetNodeInfo(
2916 self.cfg.ExpandNodeName(instance.primary_node))
2918 raise errors.OpPrereqError("Primary node '%s' is unknown" %
2920 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2922 if not isinstance(result.data, objects.OS):
2923 raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2924 " primary node" % self.op.os_type)
2926 self.instance = instance
2928 def Exec(self, feedback_fn):
2929 """Reinstall the instance.
2932 inst = self.instance
2934 if self.op.os_type is not None:
2935 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2936 inst.os = self.op.os_type
2937 self.cfg.Update(inst)
2939 _StartInstanceDisks(self, inst, None)
2941 feedback_fn("Running the instance OS create scripts...")
2942 result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2943 msg = result.RemoteFailMsg()
2945 raise errors.OpExecError("Could not install OS for instance %s"
2947 (inst.name, inst.primary_node, msg))
2949 _ShutdownInstanceDisks(self, inst)
2952 class LURenameInstance(LogicalUnit):
2953 """Rename an instance.
2956 HPATH = "instance-rename"
2957 HTYPE = constants.HTYPE_INSTANCE
2958 _OP_REQP = ["instance_name", "new_name"]
2960 def BuildHooksEnv(self):
2963 This runs on master, primary and secondary nodes of the instance.
2966 env = _BuildInstanceHookEnvByObject(self, self.instance)
2967 env["INSTANCE_NEW_NAME"] = self.op.new_name
2968 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2971 def CheckPrereq(self):
2972 """Check prerequisites.
2974 This checks that the instance is in the cluster and is not running.
2977 instance = self.cfg.GetInstanceInfo(
2978 self.cfg.ExpandInstanceName(self.op.instance_name))
2979 if instance is None:
2980 raise errors.OpPrereqError("Instance '%s' not known" %
2981 self.op.instance_name)
2982 _CheckNodeOnline(self, instance.primary_node)
2984 if instance.admin_up:
2985 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2986 self.op.instance_name)
2987 remote_info = self.rpc.call_instance_info(instance.primary_node,
2989 instance.hypervisor)
2991 if remote_info.data:
2992 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2993 (self.op.instance_name,
2994 instance.primary_node))
2995 self.instance = instance
2997 # new name verification
2998 name_info = utils.HostInfo(self.op.new_name)
3000 self.op.new_name = new_name = name_info.name
3001 instance_list = self.cfg.GetInstanceList()
3002 if new_name in instance_list:
3003 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
3006 if not getattr(self.op, "ignore_ip", False):
3007 if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
3008 raise errors.OpPrereqError("IP %s of instance %s already in use" %
3009 (name_info.ip, new_name))
3012 def Exec(self, feedback_fn):
3013 """Reinstall the instance.
3016 inst = self.instance
3017 old_name = inst.name
3019 if inst.disk_template == constants.DT_FILE:
3020 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
3022 self.cfg.RenameInstance(inst.name, self.op.new_name)
3023 # Change the instance lock. This is definitely safe while we hold the BGL
3024 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
3025 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
3027 # re-read the instance from the configuration after rename
3028 inst = self.cfg.GetInstanceInfo(self.op.new_name)
3030 if inst.disk_template == constants.DT_FILE:
3031 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
3032 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
3033 old_file_storage_dir,
3034 new_file_storage_dir)
3037 raise errors.OpExecError("Could not connect to node '%s' to rename"
3038 " directory '%s' to '%s' (but the instance"
3039 " has been renamed in Ganeti)" % (
3040 inst.primary_node, old_file_storage_dir,
3041 new_file_storage_dir))
3043 if not result.data[0]:
3044 raise errors.OpExecError("Could not rename directory '%s' to '%s'"
3045 " (but the instance has been renamed in"
3046 " Ganeti)" % (old_file_storage_dir,
3047 new_file_storage_dir))
3049 _StartInstanceDisks(self, inst, None)
3051 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
3053 msg = result.RemoteFailMsg()
3055 msg = ("Could not run OS rename script for instance %s on node %s"
3056 " (but the instance has been renamed in Ganeti): %s" %
3057 (inst.name, inst.primary_node, msg))
3058 self.proc.LogWarning(msg)
3060 _ShutdownInstanceDisks(self, inst)
3063 class LURemoveInstance(LogicalUnit):
3064 """Remove an instance.
3067 HPATH = "instance-remove"
3068 HTYPE = constants.HTYPE_INSTANCE
3069 _OP_REQP = ["instance_name", "ignore_failures"]
3072 def ExpandNames(self):
3073 self._ExpandAndLockInstance()
3074 self.needed_locks[locking.LEVEL_NODE] = []
3075 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3077 def DeclareLocks(self, level):
3078 if level == locking.LEVEL_NODE:
3079 self._LockInstancesNodes()
3081 def BuildHooksEnv(self):
3084 This runs on master, primary and secondary nodes of the instance.
3087 env = _BuildInstanceHookEnvByObject(self, self.instance)
3088 nl = [self.cfg.GetMasterNode()]
3091 def CheckPrereq(self):
3092 """Check prerequisites.
3094 This checks that the instance is in the cluster.
3097 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3098 assert self.instance is not None, \
3099 "Cannot retrieve locked instance %s" % self.op.instance_name
3101 def Exec(self, feedback_fn):
3102 """Remove the instance.
3105 instance = self.instance
3106 logging.info("Shutting down instance %s on node %s",
3107 instance.name, instance.primary_node)
3109 result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3110 msg = result.RemoteFailMsg()
3112 if self.op.ignore_failures:
3113 feedback_fn("Warning: can't shutdown instance: %s" % msg)
3115 raise errors.OpExecError("Could not shutdown instance %s on"
3117 (instance.name, instance.primary_node, msg))
3119 logging.info("Removing block devices for instance %s", instance.name)
3121 if not _RemoveDisks(self, instance):
3122 if self.op.ignore_failures:
3123 feedback_fn("Warning: can't remove instance's disks")
3125 raise errors.OpExecError("Can't remove instance's disks")
3127 logging.info("Removing instance %s out of cluster config", instance.name)
3129 self.cfg.RemoveInstance(instance.name)
3130 self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3133 class LUQueryInstances(NoHooksLU):
3134 """Logical unit for querying instances.
3137 _OP_REQP = ["output_fields", "names", "use_locking"]
3139 _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3140 "admin_state", "admin_ram",
3141 "disk_template", "ip", "mac", "bridge",
3142 "sda_size", "sdb_size", "vcpus", "tags",
3143 "network_port", "beparams",
3144 "(disk).(size)/([0-9]+)",
3145 "(disk).(sizes)", "disk_usage",
3146 "(nic).(mac|ip|bridge)/([0-9]+)",
3147 "(nic).(macs|ips|bridges)",
3148 "(disk|nic).(count)",
3149 "serial_no", "hypervisor", "hvparams",] +
3151 for name in constants.HVS_PARAMETERS] +
3153 for name in constants.BES_PARAMETERS])
3154 _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3157 def ExpandNames(self):
3158 _CheckOutputFields(static=self._FIELDS_STATIC,
3159 dynamic=self._FIELDS_DYNAMIC,
3160 selected=self.op.output_fields)
3162 self.needed_locks = {}
3163 self.share_locks[locking.LEVEL_INSTANCE] = 1
3164 self.share_locks[locking.LEVEL_NODE] = 1
3167 self.wanted = _GetWantedInstances(self, self.op.names)
3169 self.wanted = locking.ALL_SET
3171 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3172 self.do_locking = self.do_node_query and self.op.use_locking
3174 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3175 self.needed_locks[locking.LEVEL_NODE] = []
3176 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3178 def DeclareLocks(self, level):
3179 if level == locking.LEVEL_NODE and self.do_locking:
3180 self._LockInstancesNodes()
3182 def CheckPrereq(self):
3183 """Check prerequisites.
3188 def Exec(self, feedback_fn):
3189 """Computes the list of nodes and their attributes.
3192 all_info = self.cfg.GetAllInstancesInfo()
3193 if self.wanted == locking.ALL_SET:
3194 # caller didn't specify instance names, so ordering is not important
3196 instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3198 instance_names = all_info.keys()
3199 instance_names = utils.NiceSort(instance_names)
3201 # caller did specify names, so we must keep the ordering
3203 tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3205 tgt_set = all_info.keys()
3206 missing = set(self.wanted).difference(tgt_set)
3208 raise errors.OpExecError("Some instances were removed before"
3209 " retrieving their data: %s" % missing)
3210 instance_names = self.wanted
3212 instance_list = [all_info[iname] for iname in instance_names]
3214 # begin data gathering
3216 nodes = frozenset([inst.primary_node for inst in instance_list])
3217 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3221 if self.do_node_query:
3223 node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3225 result = node_data[name]
3227 # offline nodes will be in both lists
3228 off_nodes.append(name)
3230 bad_nodes.append(name)
3233 live_data.update(result.data)
3234 # else no instance is alive
3236 live_data = dict([(name, {}) for name in instance_names])
3238 # end data gathering
3243 for instance in instance_list:
3245 i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3246 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3247 for field in self.op.output_fields:
3248 st_match = self._FIELDS_STATIC.Matches(field)
3253 elif field == "pnode":
3254 val = instance.primary_node
3255 elif field == "snodes":
3256 val = list(instance.secondary_nodes)
3257 elif field == "admin_state":
3258 val = instance.admin_up
3259 elif field == "oper_state":
3260 if instance.primary_node in bad_nodes:
3263 val = bool(live_data.get(instance.name))
3264 elif field == "status":
3265 if instance.primary_node in off_nodes:
3266 val = "ERROR_nodeoffline"
3267 elif instance.primary_node in bad_nodes:
3268 val = "ERROR_nodedown"
3270 running = bool(live_data.get(instance.name))
3272 if instance.admin_up:
3277 if instance.admin_up:
3281 elif field == "oper_ram":
3282 if instance.primary_node in bad_nodes:
3284 elif instance.name in live_data:
3285 val = live_data[instance.name].get("memory", "?")
3288 elif field == "disk_template":
3289 val = instance.disk_template
3291 val = instance.nics[0].ip
3292 elif field == "bridge":
3293 val = instance.nics[0].bridge
3294 elif field == "mac":
3295 val = instance.nics[0].mac
3296 elif field == "sda_size" or field == "sdb_size":
3297 idx = ord(field[2]) - ord('a')
3299 val = instance.FindDisk(idx).size
3300 except errors.OpPrereqError:
3302 elif field == "disk_usage": # total disk usage per node
3303 disk_sizes = [{'size': disk.size} for disk in instance.disks]
3304 val = _ComputeDiskSize(instance.disk_template, disk_sizes)
3305 elif field == "tags":
3306 val = list(instance.GetTags())
3307 elif field == "serial_no":
3308 val = instance.serial_no
3309 elif field == "network_port":
3310 val = instance.network_port
3311 elif field == "hypervisor":
3312 val = instance.hypervisor
3313 elif field == "hvparams":
3315 elif (field.startswith(HVPREFIX) and
3316 field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3317 val = i_hv.get(field[len(HVPREFIX):], None)
3318 elif field == "beparams":
3320 elif (field.startswith(BEPREFIX) and
3321 field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3322 val = i_be.get(field[len(BEPREFIX):], None)
3323 elif st_match and st_match.groups():
3324 # matches a variable list
3325 st_groups = st_match.groups()
3326 if st_groups and st_groups[0] == "disk":
3327 if st_groups[1] == "count":
3328 val = len(instance.disks)
3329 elif st_groups[1] == "sizes":
3330 val = [disk.size for disk in instance.disks]
3331 elif st_groups[1] == "size":
3333 val = instance.FindDisk(st_groups[2]).size
3334 except errors.OpPrereqError:
3337 assert False, "Unhandled disk parameter"
3338 elif st_groups[0] == "nic":
3339 if st_groups[1] == "count":
3340 val = len(instance.nics)
3341 elif st_groups[1] == "macs":
3342 val = [nic.mac for nic in instance.nics]
3343 elif st_groups[1] == "ips":
3344 val = [nic.ip for nic in instance.nics]
3345 elif st_groups[1] == "bridges":
3346 val = [nic.bridge for nic in instance.nics]
3349 nic_idx = int(st_groups[2])
3350 if nic_idx >= len(instance.nics):
3353 if st_groups[1] == "mac":
3354 val = instance.nics[nic_idx].mac
3355 elif st_groups[1] == "ip":
3356 val = instance.nics[nic_idx].ip
3357 elif st_groups[1] == "bridge":
3358 val = instance.nics[nic_idx].bridge
3360 assert False, "Unhandled NIC parameter"
3362 assert False, "Unhandled variable parameter"
3364 raise errors.ParameterError(field)
3371 class LUFailoverInstance(LogicalUnit):
3372 """Failover an instance.
3375 HPATH = "instance-failover"
3376 HTYPE = constants.HTYPE_INSTANCE
3377 _OP_REQP = ["instance_name", "ignore_consistency"]
3380 def ExpandNames(self):
3381 self._ExpandAndLockInstance()
3382 self.needed_locks[locking.LEVEL_NODE] = []
3383 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3385 def DeclareLocks(self, level):
3386 if level == locking.LEVEL_NODE:
3387 self._LockInstancesNodes()
3389 def BuildHooksEnv(self):
3392 This runs on master, primary and secondary nodes of the instance.
3396 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3398 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3399 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3402 def CheckPrereq(self):
3403 """Check prerequisites.
3405 This checks that the instance is in the cluster.
3408 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3409 assert self.instance is not None, \
3410 "Cannot retrieve locked instance %s" % self.op.instance_name
3412 bep = self.cfg.GetClusterInfo().FillBE(instance)
3413 if instance.disk_template not in constants.DTS_NET_MIRROR:
3414 raise errors.OpPrereqError("Instance's disk layout is not"
3415 " network mirrored, cannot failover.")
3417 secondary_nodes = instance.secondary_nodes
3418 if not secondary_nodes:
3419 raise errors.ProgrammerError("no secondary node but using "
3420 "a mirrored disk template")
3422 target_node = secondary_nodes[0]
3423 _CheckNodeOnline(self, target_node)
3424 _CheckNodeNotDrained(self, target_node)
3425 # check memory requirements on the secondary node
3426 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3427 instance.name, bep[constants.BE_MEMORY],
3428 instance.hypervisor)
3430 # check bridge existance
3431 brlist = [nic.bridge for nic in instance.nics]
3432 result = self.rpc.call_bridges_exist(target_node, brlist)
3435 raise errors.OpPrereqError("One or more target bridges %s does not"
3436 " exist on destination node '%s'" %
3437 (brlist, target_node))
3439 def Exec(self, feedback_fn):
3440 """Failover an instance.
3442 The failover is done by shutting it down on its present node and
3443 starting it on the secondary.
3446 instance = self.instance
3448 source_node = instance.primary_node
3449 target_node = instance.secondary_nodes[0]
3451 feedback_fn("* checking disk consistency between source and target")
3452 for dev in instance.disks:
3453 # for drbd, these are drbd over lvm
3454 if not _CheckDiskConsistency(self, dev, target_node, False):
3455 if instance.admin_up and not self.op.ignore_consistency:
3456 raise errors.OpExecError("Disk %s is degraded on target node,"
3457 " aborting failover." % dev.iv_name)
3459 feedback_fn("* shutting down instance on source node")
3460 logging.info("Shutting down instance %s on node %s",
3461 instance.name, source_node)
3463 result = self.rpc.call_instance_shutdown(source_node, instance)
3464 msg = result.RemoteFailMsg()
3466 if self.op.ignore_consistency:
3467 self.proc.LogWarning("Could not shutdown instance %s on node %s."
3468 " Proceeding anyway. Please make sure node"
3469 " %s is down. Error details: %s",
3470 instance.name, source_node, source_node, msg)
3472 raise errors.OpExecError("Could not shutdown instance %s on"
3474 (instance.name, source_node, msg))
3476 feedback_fn("* deactivating the instance's disks on source node")
3477 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3478 raise errors.OpExecError("Can't shut down the instance's disks.")
3480 instance.primary_node = target_node
3481 # distribute new instance config to the other nodes
3482 self.cfg.Update(instance)
3484 # Only start the instance if it's marked as up
3485 if instance.admin_up:
3486 feedback_fn("* activating the instance's disks on target node")
3487 logging.info("Starting instance %s on node %s",
3488 instance.name, target_node)
3490 disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3491 ignore_secondaries=True)
3493 _ShutdownInstanceDisks(self, instance)
3494 raise errors.OpExecError("Can't activate the instance's disks")
3496 feedback_fn("* starting the instance on the target node")
3497 result = self.rpc.call_instance_start(target_node, instance, None)
3498 msg = result.RemoteFailMsg()
3500 _ShutdownInstanceDisks(self, instance)
3501 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3502 (instance.name, target_node, msg))
3505 class LUMigrateInstance(LogicalUnit):
3506 """Migrate an instance.
3508 This is migration without shutting down, compared to the failover,
3509 which is done with shutdown.
3512 HPATH = "instance-migrate"
3513 HTYPE = constants.HTYPE_INSTANCE
3514 _OP_REQP = ["instance_name", "live", "cleanup"]
3518 def ExpandNames(self):
3519 self._ExpandAndLockInstance()
3520 self.needed_locks[locking.LEVEL_NODE] = []
3521 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3523 def DeclareLocks(self, level):
3524 if level == locking.LEVEL_NODE:
3525 self._LockInstancesNodes()
3527 def BuildHooksEnv(self):
3530 This runs on master, primary and secondary nodes of the instance.
3533 env = _BuildInstanceHookEnvByObject(self, self.instance)
3534 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3537 def CheckPrereq(self):
3538 """Check prerequisites.
3540 This checks that the instance is in the cluster.
3543 instance = self.cfg.GetInstanceInfo(
3544 self.cfg.ExpandInstanceName(self.op.instance_name))
3545 if instance is None:
3546 raise errors.OpPrereqError("Instance '%s' not known" %
3547 self.op.instance_name)
3549 if instance.disk_template != constants.DT_DRBD8:
3550 raise errors.OpPrereqError("Instance's disk layout is not"
3551 " drbd8, cannot migrate.")
3553 secondary_nodes = instance.secondary_nodes
3554 if not secondary_nodes:
3555 raise errors.ConfigurationError("No secondary node but using"
3556 " drbd8 disk template")
3558 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3560 target_node = secondary_nodes[0]
3561 # check memory requirements on the secondary node
3562 _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3563 instance.name, i_be[constants.BE_MEMORY],
3564 instance.hypervisor)
3566 # check bridge existance
3567 brlist = [nic.bridge for nic in instance.nics]
3568 result = self.rpc.call_bridges_exist(target_node, brlist)
3569 if result.failed or not result.data:
3570 raise errors.OpPrereqError("One or more target bridges %s does not"
3571 " exist on destination node '%s'" %
3572 (brlist, target_node))
3574 if not self.op.cleanup:
3575 _CheckNodeNotDrained(self, target_node)
3576 result = self.rpc.call_instance_migratable(instance.primary_node,
3578 msg = result.RemoteFailMsg()
3580 raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3583 self.instance = instance
3585 def _WaitUntilSync(self):
3586 """Poll with custom rpc for disk sync.
3588 This uses our own step-based rpc call.
3591 self.feedback_fn("* wait until resync is done")
3595 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3597 self.instance.disks)
3599 for node, nres in result.items():
3600 msg = nres.RemoteFailMsg()
3602 raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3604 node_done, node_percent = nres.payload
3605 all_done = all_done and node_done
3606 if node_percent is not None:
3607 min_percent = min(min_percent, node_percent)
3609 if min_percent < 100:
3610 self.feedback_fn(" - progress: %.1f%%" % min_percent)
3613 def _EnsureSecondary(self, node):
3614 """Demote a node to secondary.
3617 self.feedback_fn("* switching node %s to secondary mode" % node)
3619 for dev in self.instance.disks:
3620 self.cfg.SetDiskID(dev, node)
3622 result = self.rpc.call_blockdev_close(node, self.instance.name,
3623 self.instance.disks)
3624 msg = result.RemoteFailMsg()
3626 raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3627 " error %s" % (node, msg))
3629 def _GoStandalone(self):
3630 """Disconnect from the network.
3633 self.feedback_fn("* changing into standalone mode")
3634 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3635 self.instance.disks)
3636 for node, nres in result.items():
3637 msg = nres.RemoteFailMsg()
3639 raise errors.OpExecError("Cannot disconnect disks node %s,"
3640 " error %s" % (node, msg))
3642 def _GoReconnect(self, multimaster):
3643 """Reconnect to the network.
3649 msg = "single-master"
3650 self.feedback_fn("* changing disks into %s mode" % msg)
3651 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3652 self.instance.disks,
3653 self.instance.name, multimaster)
3654 for node, nres in result.items():
3655 msg = nres.RemoteFailMsg()
3657 raise errors.OpExecError("Cannot change disks config on node %s,"
3658 " error: %s" % (node, msg))
3660 def _ExecCleanup(self):
3661 """Try to cleanup after a failed migration.
3663 The cleanup is done by:
3664 - check that the instance is running only on one node
3665 (and update the config if needed)
3666 - change disks on its secondary node to secondary
3667 - wait until disks are fully synchronized
3668 - disconnect from the network
3669 - change disks into single-master mode
3670 - wait again until disks are fully synchronized
3673 instance = self.instance
3674 target_node = self.target_node
3675 source_node = self.source_node
3677 # check running on only one node
3678 self.feedback_fn("* checking where the instance actually runs"
3679 " (if this hangs, the hypervisor might be in"
3681 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3682 for node, result in ins_l.items():
3684 if not isinstance(result.data, list):
3685 raise errors.OpExecError("Can't contact node '%s'" % node)
3687 runningon_source = instance.name in ins_l[source_node].data
3688 runningon_target = instance.name in ins_l[target_node].data
3690 if runningon_source and runningon_target:
3691 raise errors.OpExecError("Instance seems to be running on two nodes,"
3692 " or the hypervisor is confused. You will have"
3693 " to ensure manually that it runs only on one"
3694 " and restart this operation.")
3696 if not (runningon_source or runningon_target):
3697 raise errors.OpExecError("Instance does not seem to be running at all."
3698 " In this case, it's safer to repair by"
3699 " running 'gnt-instance stop' to ensure disk"
3700 " shutdown, and then restarting it.")
3702 if runningon_target:
3703 # the migration has actually succeeded, we need to update the config
3704 self.feedback_fn("* instance running on secondary node (%s),"
3705 " updating config" % target_node)
3706 instance.primary_node = target_node
3707 self.cfg.Update(instance)
3708 demoted_node = source_node
3710 self.feedback_fn("* instance confirmed to be running on its"
3711 " primary node (%s)" % source_node)
3712 demoted_node = target_node
3714 self._EnsureSecondary(demoted_node)
3716 self._WaitUntilSync()
3717 except errors.OpExecError:
3718 # we ignore here errors, since if the device is standalone, it
3719 # won't be able to sync
3721 self._GoStandalone()
3722 self._GoReconnect(False)
3723 self._WaitUntilSync()
3725 self.feedback_fn("* done")
3727 def _RevertDiskStatus(self):
3728 """Try to revert the disk status after a failed migration.
3731 target_node = self.target_node
3733 self._EnsureSecondary(target_node)
3734 self._GoStandalone()
3735 self._GoReconnect(False)
3736 self._WaitUntilSync()
3737 except errors.OpExecError, err:
3738 self.LogWarning("Migration failed and I can't reconnect the"
3739 " drives: error '%s'\n"
3740 "Please look and recover the instance status" %
3743 def _AbortMigration(self):
3744 """Call the hypervisor code to abort a started migration.
3747 instance = self.instance
3748 target_node = self.target_node
3749 migration_info = self.migration_info
3751 abort_result = self.rpc.call_finalize_migration(target_node,
3755 abort_msg = abort_result.RemoteFailMsg()
3757 logging.error("Aborting migration failed on target node %s: %s" %
3758 (target_node, abort_msg))
3759 # Don't raise an exception here, as we stil have to try to revert the
3760 # disk status, even if this step failed.
3762 def _ExecMigration(self):
3763 """Migrate an instance.
3765 The migrate is done by:
3766 - change the disks into dual-master mode
3767 - wait until disks are fully synchronized again
3768 - migrate the instance
3769 - change disks on the new secondary node (the old primary) to secondary
3770 - wait until disks are fully synchronized
3771 - change disks into single-master mode
3774 instance = self.instance
3775 target_node = self.target_node
3776 source_node = self.source_node
3778 self.feedback_fn("* checking disk consistency between source and target")
3779 for dev in instance.disks:
3780 if not _CheckDiskConsistency(self, dev, target_node, False):
3781 raise errors.OpExecError("Disk %s is degraded or not fully"
3782 " synchronized on target node,"
3783 " aborting migrate." % dev.iv_name)
3785 # First get the migration information from the remote node
3786 result = self.rpc.call_migration_info(source_node, instance)
3787 msg = result.RemoteFailMsg()
3789 log_err = ("Failed fetching source migration information from %s: %s" %
3791 logging.error(log_err)
3792 raise errors.OpExecError(log_err)
3794 self.migration_info = migration_info = result.payload
3796 # Then switch the disks to master/master mode
3797 self._EnsureSecondary(target_node)
3798 self._GoStandalone()
3799 self._GoReconnect(True)
3800 self._WaitUntilSync()
3802 self.feedback_fn("* preparing %s to accept the instance" % target_node)
3803 result = self.rpc.call_accept_instance(target_node,
3806 self.nodes_ip[target_node])
3808 msg = result.RemoteFailMsg()
3810 logging.error("Instance pre-migration failed, trying to revert"
3811 " disk status: %s", msg)
3812 self._AbortMigration()
3813 self._RevertDiskStatus()
3814 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
3815 (instance.name, msg))
3817 self.feedback_fn("* migrating instance to %s" % target_node)
3819 result = self.rpc.call_instance_migrate(source_node, instance,
3820 self.nodes_ip[target_node],
3822 msg = result.RemoteFailMsg()
3824 logging.error("Instance migration failed, trying to revert"
3825 " disk status: %s", msg)
3826 self._AbortMigration()
3827 self._RevertDiskStatus()
3828 raise errors.OpExecError("Could not migrate instance %s: %s" %
3829 (instance.name, msg))
3832 instance.primary_node = target_node
3833 # distribute new instance config to the other nodes
3834 self.cfg.Update(instance)
3836 result = self.rpc.call_finalize_migration(target_node,
3840 msg = result.RemoteFailMsg()
3842 logging.error("Instance migration succeeded, but finalization failed:"
3844 raise errors.OpExecError("Could not finalize instance migration: %s" %
3847 self._EnsureSecondary(source_node)
3848 self._WaitUntilSync()
3849 self._GoStandalone()
3850 self._GoReconnect(False)
3851 self._WaitUntilSync()
3853 self.feedback_fn("* done")
3855 def Exec(self, feedback_fn):
3856 """Perform the migration.
3859 self.feedback_fn = feedback_fn
3861 self.source_node = self.instance.primary_node
3862 self.target_node = self.instance.secondary_nodes[0]
3863 self.all_nodes = [self.source_node, self.target_node]
3865 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3866 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3869 return self._ExecCleanup()
3871 return self._ExecMigration()
3874 def _CreateBlockDev(lu, node, instance, device, force_create,
3876 """Create a tree of block devices on a given node.
3878 If this device type has to be created on secondaries, create it and
3881 If not, just recurse to children keeping the same 'force' value.
3883 @param lu: the lu on whose behalf we execute
3884 @param node: the node on which to create the device
3885 @type instance: L{objects.Instance}
3886 @param instance: the instance which owns the device
3887 @type device: L{objects.Disk}
3888 @param device: the device to create
3889 @type force_create: boolean
3890 @param force_create: whether to force creation of this device; this
3891 will be change to True whenever we find a device which has
3892 CreateOnSecondary() attribute
3893 @param info: the extra 'metadata' we should attach to the device
3894 (this will be represented as a LVM tag)
3895 @type force_open: boolean
3896 @param force_open: this parameter will be passes to the
3897 L{backend.BlockdevCreate} function where it specifies
3898 whether we run on primary or not, and it affects both
3899 the child assembly and the device own Open() execution
3902 if device.CreateOnSecondary():
3906 for child in device.children:
3907 _CreateBlockDev(lu, node, instance, child, force_create,
3910 if not force_create:
3913 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3916 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3917 """Create a single block device on a given node.
3919 This will not recurse over children of the device, so they must be
3922 @param lu: the lu on whose behalf we execute
3923 @param node: the node on which to create the device
3924 @type instance: L{objects.Instance}
3925 @param instance: the instance which owns the device
3926 @type device: L{objects.Disk}
3927 @param device: the device to create
3928 @param info: the extra 'metadata' we should attach to the device
3929 (this will be represented as a LVM tag)
3930 @type force_open: boolean
3931 @param force_open: this parameter will be passes to the
3932 L{backend.BlockdevCreate} function where it specifies
3933 whether we run on primary or not, and it affects both
3934 the child assembly and the device own Open() execution
3937 lu.cfg.SetDiskID(device, node)
3938 result = lu.rpc.call_blockdev_create(node, device, device.size,
3939 instance.name, force_open, info)
3940 msg = result.RemoteFailMsg()
3942 raise errors.OpExecError("Can't create block device %s on"
3943 " node %s for instance %s: %s" %
3944 (device, node, instance.name, msg))
3945 if device.physical_id is None:
3946 device.physical_id = result.payload
3949 def _GenerateUniqueNames(lu, exts):
3950 """Generate a suitable LV name.
3952 This will generate a logical volume name for the given instance.
3957 new_id = lu.cfg.GenerateUniqueID()
3958 results.append("%s%s" % (new_id, val))
3962 def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3964 """Generate a drbd8 device complete with its children.
3967 port = lu.cfg.AllocatePort()
3968 vgname = lu.cfg.GetVGName()
3969 shared_secret = lu.cfg.GenerateDRBDSecret()
3970 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3971 logical_id=(vgname, names[0]))
3972 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3973 logical_id=(vgname, names[1]))
3974 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3975 logical_id=(primary, secondary, port,
3978 children=[dev_data, dev_meta],
3983 def _GenerateDiskTemplate(lu, template_name,
3984 instance_name, primary_node,
3985 secondary_nodes, disk_info,
3986 file_storage_dir, file_driver,
3988 """Generate the entire disk layout for a given template type.
3991 #TODO: compute space requirements
3993 vgname = lu.cfg.GetVGName()
3994 disk_count = len(disk_info)
3996 if template_name == constants.DT_DISKLESS:
3998 elif template_name == constants.DT_PLAIN:
3999 if len(secondary_nodes) != 0:
4000 raise errors.ProgrammerError("Wrong template configuration")
4002 names = _GenerateUniqueNames(lu, [".disk%d" % i
4003 for i in range(disk_count)])
4004 for idx, disk in enumerate(disk_info):
4005 disk_index = idx + base_index
4006 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
4007 logical_id=(vgname, names[idx]),
4008 iv_name="disk/%d" % disk_index,
4010 disks.append(disk_dev)
4011 elif template_name == constants.DT_DRBD8:
4012 if len(secondary_nodes) != 1:
4013 raise errors.ProgrammerError("Wrong template configuration")
4014 remote_node = secondary_nodes[0]
4015 minors = lu.cfg.AllocateDRBDMinor(
4016 [primary_node, remote_node] * len(disk_info), instance_name)
4019 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
4020 for i in range(disk_count)]):
4021 names.append(lv_prefix + "_data")
4022 names.append(lv_prefix + "_meta")
4023 for idx, disk in enumerate(disk_info):
4024 disk_index = idx + base_index
4025 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
4026 disk["size"], names[idx*2:idx*2+2],
4027 "disk/%d" % disk_index,
4028 minors[idx*2], minors[idx*2+1])
4029 disk_dev.mode = disk["mode"]
4030 disks.append(disk_dev)
4031 elif template_name == constants.DT_FILE:
4032 if len(secondary_nodes) != 0:
4033 raise errors.ProgrammerError("Wrong template configuration")
4035 for idx, disk in enumerate(disk_info):
4036 disk_index = idx + base_index
4037 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
4038 iv_name="disk/%d" % disk_index,
4039 logical_id=(file_driver,
4040 "%s/disk%d" % (file_storage_dir,
4043 disks.append(disk_dev)
4045 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
4049 def _GetInstanceInfoText(instance):
4050 """Compute that text that should be added to the disk's metadata.
4053 return "originstname+%s" % instance.name
4056 def _CreateDisks(lu, instance):
4057 """Create all disks for an instance.
4059 This abstracts away some work from AddInstance.
4061 @type lu: L{LogicalUnit}
4062 @param lu: the logical unit on whose behalf we execute
4063 @type instance: L{objects.Instance}
4064 @param instance: the instance whose disks we should create
4066 @return: the success of the creation
4069 info = _GetInstanceInfoText(instance)
4070 pnode = instance.primary_node
4072 if instance.disk_template == constants.DT_FILE:
4073 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4074 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
4076 if result.failed or not result.data:
4077 raise errors.OpExecError("Could not connect to node '%s'" % pnode)
4079 if not result.data[0]:
4080 raise errors.OpExecError("Failed to create directory '%s'" %
4083 # Note: this needs to be kept in sync with adding of disks in
4084 # LUSetInstanceParams
4085 for device in instance.disks:
4086 logging.info("Creating volume %s for instance %s",
4087 device.iv_name, instance.name)
4089 for node in instance.all_nodes:
4090 f_create = node == pnode
4091 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4094 def _RemoveDisks(lu, instance):
4095 """Remove all disks for an instance.
4097 This abstracts away some work from `AddInstance()` and
4098 `RemoveInstance()`. Note that in case some of the devices couldn't
4099 be removed, the removal will continue with the other ones (compare
4100 with `_CreateDisks()`).
4102 @type lu: L{LogicalUnit}
4103 @param lu: the logical unit on whose behalf we execute
4104 @type instance: L{objects.Instance}
4105 @param instance: the instance whose disks we should remove
4107 @return: the success of the removal
4110 logging.info("Removing block devices for instance %s", instance.name)
4113 for device in instance.disks:
4114 for node, disk in device.ComputeNodeTree(instance.primary_node):
4115 lu.cfg.SetDiskID(disk, node)
4116 msg = lu.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
4118 lu.LogWarning("Could not remove block device %s on node %s,"
4119 " continuing anyway: %s", device.iv_name, node, msg)
4122 if instance.disk_template == constants.DT_FILE:
4123 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4124 result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4126 if result.failed or not result.data:
4127 logging.error("Could not remove directory '%s'", file_storage_dir)
4133 def _ComputeDiskSize(disk_template, disks):
4134 """Compute disk size requirements in the volume group
4137 # Required free disk space as a function of disk and swap space
4139 constants.DT_DISKLESS: None,
4140 constants.DT_PLAIN: sum(d["size"] for d in disks),
4141 # 128 MB are added for drbd metadata for each disk
4142 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4143 constants.DT_FILE: None,
4146 if disk_template not in req_size_dict:
4147 raise errors.ProgrammerError("Disk template '%s' size requirement"
4148 " is unknown" % disk_template)
4150 return req_size_dict[disk_template]
4153 def _CheckHVParams(lu, nodenames, hvname, hvparams):
4154 """Hypervisor parameter validation.
4156 This function abstract the hypervisor parameter validation to be
4157 used in both instance create and instance modify.
4159 @type lu: L{LogicalUnit}
4160 @param lu: the logical unit for which we check
4161 @type nodenames: list
4162 @param nodenames: the list of nodes on which we should check
4163 @type hvname: string
4164 @param hvname: the name of the hypervisor we should use
4165 @type hvparams: dict
4166 @param hvparams: the parameters which we need to check
4167 @raise errors.OpPrereqError: if the parameters are not valid
4170 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4173 for node in nodenames:
4177 msg = info.RemoteFailMsg()
4179 raise errors.OpPrereqError("Hypervisor parameter validation failed:"
4183 class LUCreateInstance(LogicalUnit):
4184 """Create an instance.
4187 HPATH = "instance-add"
4188 HTYPE = constants.HTYPE_INSTANCE
4189 _OP_REQP = ["instance_name", "disks", "disk_template",
4191 "wait_for_sync", "ip_check", "nics",
4192 "hvparams", "beparams"]
4195 def _ExpandNode(self, node):
4196 """Expands and checks one node name.
4199 node_full = self.cfg.ExpandNodeName(node)
4200 if node_full is None:
4201 raise errors.OpPrereqError("Unknown node %s" % node)
4204 def ExpandNames(self):
4205 """ExpandNames for CreateInstance.
4207 Figure out the right locks for instance creation.
4210 self.needed_locks = {}
4212 # set optional parameters to none if they don't exist
4213 for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4214 if not hasattr(self.op, attr):
4215 setattr(self.op, attr, None)
4217 # cheap checks, mostly valid constants given
4219 # verify creation mode
4220 if self.op.mode not in (constants.INSTANCE_CREATE,
4221 constants.INSTANCE_IMPORT):
4222 raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4225 # disk template and mirror node verification
4226 if self.op.disk_template not in constants.DISK_TEMPLATES:
4227 raise errors.OpPrereqError("Invalid disk template name")
4229 if self.op.hypervisor is None:
4230 self.op.hypervisor = self.cfg.GetHypervisorType()
4232 cluster = self.cfg.GetClusterInfo()
4233 enabled_hvs = cluster.enabled_hypervisors
4234 if self.op.hypervisor not in enabled_hvs:
4235 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4236 " cluster (%s)" % (self.op.hypervisor,
4237 ",".join(enabled_hvs)))
4239 # check hypervisor parameter syntax (locally)
4240 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4241 filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
4243 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4244 hv_type.CheckParameterSyntax(filled_hvp)
4246 # fill and remember the beparams dict
4247 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4248 self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4251 #### instance parameters check
4253 # instance name verification
4254 hostname1 = utils.HostInfo(self.op.instance_name)
4255 self.op.instance_name = instance_name = hostname1.name
4257 # this is just a preventive check, but someone might still add this
4258 # instance in the meantime, and creation will fail at lock-add time
4259 if instance_name in self.cfg.GetInstanceList():
4260 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4263 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4267 for nic in self.op.nics:
4268 # ip validity checks
4269 ip = nic.get("ip", None)
4270 if ip is None or ip.lower() == "none":
4272 elif ip.lower() == constants.VALUE_AUTO:
4273 nic_ip = hostname1.ip
4275 if not utils.IsValidIP(ip):
4276 raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4277 " like a valid IP" % ip)
4280 # MAC address verification
4281 mac = nic.get("mac", constants.VALUE_AUTO)
4282 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4283 if not utils.IsValidMac(mac.lower()):
4284 raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4286 # bridge verification
4287 bridge = nic.get("bridge", None)
4289 bridge = self.cfg.GetDefBridge()
4290 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
4292 # disk checks/pre-build
4294 for disk in self.op.disks:
4295 mode = disk.get("mode", constants.DISK_RDWR)
4296 if mode not in constants.DISK_ACCESS_SET:
4297 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4299 size = disk.get("size", None)
4301 raise errors.OpPrereqError("Missing disk size")
4305 raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4306 self.disks.append({"size": size, "mode": mode})
4308 # used in CheckPrereq for ip ping check
4309 self.check_ip = hostname1.ip
4311 # file storage checks
4312 if (self.op.file_driver and
4313 not self.op.file_driver in constants.FILE_DRIVER):
4314 raise errors.OpPrereqError("Invalid file driver name '%s'" %
4315 self.op.file_driver)
4317 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4318 raise errors.OpPrereqError("File storage directory path not absolute")
4320 ### Node/iallocator related checks
4321 if [self.op.iallocator, self.op.pnode].count(None) != 1:
4322 raise errors.OpPrereqError("One and only one of iallocator and primary"
4323 " node must be given")
4325 if self.op.iallocator:
4326 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4328 self.op.pnode = self._ExpandNode(self.op.pnode)
4329 nodelist = [self.op.pnode]
4330 if self.op.snode is not None:
4331 self.op.snode = self._ExpandNode(self.op.snode)
4332 nodelist.append(self.op.snode)
4333 self.needed_locks[locking.LEVEL_NODE] = nodelist
4335 # in case of import lock the source node too
4336 if self.op.mode == constants.INSTANCE_IMPORT:
4337 src_node = getattr(self.op, "src_node", None)
4338 src_path = getattr(self.op, "src_path", None)
4340 if src_path is None:
4341 self.op.src_path = src_path = self.op.instance_name
4343 if src_node is None:
4344 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4345 self.op.src_node = None
4346 if os.path.isabs(src_path):
4347 raise errors.OpPrereqError("Importing an instance from an absolute"
4348 " path requires a source node option.")
4350 self.op.src_node = src_node = self._ExpandNode(src_node)
4351 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4352 self.needed_locks[locking.LEVEL_NODE].append(src_node)
4353 if not os.path.isabs(src_path):
4354 self.op.src_path = src_path = \
4355 os.path.join(constants.EXPORT_DIR, src_path)
4357 else: # INSTANCE_CREATE
4358 if getattr(self.op, "os_type", None) is None:
4359 raise errors.OpPrereqError("No guest OS specified")
4361 def _RunAllocator(self):
4362 """Run the allocator based on input opcode.
4365 nics = [n.ToDict() for n in self.nics]
4366 ial = IAllocator(self,
4367 mode=constants.IALLOCATOR_MODE_ALLOC,
4368 name=self.op.instance_name,
4369 disk_template=self.op.disk_template,
4372 vcpus=self.be_full[constants.BE_VCPUS],
4373 mem_size=self.be_full[constants.BE_MEMORY],
4376 hypervisor=self.op.hypervisor,
4379 ial.Run(self.op.iallocator)
4382 raise errors.OpPrereqError("Can't compute nodes using"
4383 " iallocator '%s': %s" % (self.op.iallocator,
4385 if len(ial.nodes) != ial.required_nodes:
4386 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4387 " of nodes (%s), required %s" %
4388 (self.op.iallocator, len(ial.nodes),
4389 ial.required_nodes))
4390 self.op.pnode = ial.nodes[0]
4391 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4392 self.op.instance_name, self.op.iallocator,
4393 ", ".join(ial.nodes))
4394 if ial.required_nodes == 2:
4395 self.op.snode = ial.nodes[1]
4397 def BuildHooksEnv(self):
4400 This runs on master, primary and secondary nodes of the instance.
4404 "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
4405 "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
4406 "INSTANCE_ADD_MODE": self.op.mode,
4408 if self.op.mode == constants.INSTANCE_IMPORT:
4409 env["INSTANCE_SRC_NODE"] = self.op.src_node
4410 env["INSTANCE_SRC_PATH"] = self.op.src_path
4411 env["INSTANCE_SRC_IMAGES"] = self.src_images
4413 env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
4414 primary_node=self.op.pnode,
4415 secondary_nodes=self.secondaries,
4416 status=self.op.start,
4417 os_type=self.op.os_type,
4418 memory=self.be_full[constants.BE_MEMORY],
4419 vcpus=self.be_full[constants.BE_VCPUS],
4420 nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
4423 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4428 def CheckPrereq(self):
4429 """Check prerequisites.
4432 if (not self.cfg.GetVGName() and
4433 self.op.disk_template not in constants.DTS_NOT_LVM):
4434 raise errors.OpPrereqError("Cluster does not support lvm-based"
4438 if self.op.mode == constants.INSTANCE_IMPORT:
4439 src_node = self.op.src_node
4440 src_path = self.op.src_path
4442 if src_node is None:
4443 exp_list = self.rpc.call_export_list(
4444 self.acquired_locks[locking.LEVEL_NODE])
4446 for node in exp_list:
4447 if not exp_list[node].failed and src_path in exp_list[node].data:
4449 self.op.src_node = src_node = node
4450 self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4454 raise errors.OpPrereqError("No export found for relative path %s" %
4457 _CheckNodeOnline(self, src_node)
4458 result = self.rpc.call_export_info(src_node, src_path)
4461 raise errors.OpPrereqError("No export found in dir %s" % src_path)
4463 export_info = result.data
4464 if not export_info.has_section(constants.INISECT_EXP):
4465 raise errors.ProgrammerError("Corrupted export config")
4467 ei_version = export_info.get(constants.INISECT_EXP, 'version')
4468 if (int(ei_version) != constants.EXPORT_VERSION):
4469 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4470 (ei_version, constants.EXPORT_VERSION))
4472 # Check that the new instance doesn't have less disks than the export
4473 instance_disks = len(self.disks)
4474 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4475 if instance_disks < export_disks:
4476 raise errors.OpPrereqError("Not enough disks to import."
4477 " (instance: %d, export: %d)" %
4478 (instance_disks, export_disks))
4480 self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4482 for idx in range(export_disks):
4483 option = 'disk%d_dump' % idx
4484 if export_info.has_option(constants.INISECT_INS, option):
4485 # FIXME: are the old os-es, disk sizes, etc. useful?
4486 export_name = export_info.get(constants.INISECT_INS, option)
4487 image = os.path.join(src_path, export_name)
4488 disk_images.append(image)
4490 disk_images.append(False)
4492 self.src_images = disk_images
4494 old_name = export_info.get(constants.INISECT_INS, 'name')
4495 # FIXME: int() here could throw a ValueError on broken exports
4496 exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4497 if self.op.instance_name == old_name:
4498 for idx, nic in enumerate(self.nics):
4499 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4500 nic_mac_ini = 'nic%d_mac' % idx
4501 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4503 # ip ping checks (we use the same ip that was resolved in ExpandNames)
4504 if self.op.start and not self.op.ip_check:
4505 raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4506 " adding an instance in start mode")
4508 if self.op.ip_check:
4509 if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4510 raise errors.OpPrereqError("IP %s of instance %s already in use" %
4511 (self.check_ip, self.op.instance_name))
4515 if self.op.iallocator is not None:
4516 self._RunAllocator()
4518 #### node related checks
4520 # check primary node
4521 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4522 assert self.pnode is not None, \
4523 "Cannot retrieve locked node %s" % self.op.pnode
4525 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4528 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
4531 self.secondaries = []
4533 # mirror node verification
4534 if self.op.disk_template in constants.DTS_NET_MIRROR:
4535 if self.op.snode is None:
4536 raise errors.OpPrereqError("The networked disk templates need"
4538 if self.op.snode == pnode.name:
4539 raise errors.OpPrereqError("The secondary node cannot be"
4540 " the primary node.")
4541 _CheckNodeOnline(self, self.op.snode)
4542 _CheckNodeNotDrained(self, self.op.snode)
4543 self.secondaries.append(self.op.snode)
4545 nodenames = [pnode.name] + self.secondaries
4547 req_size = _ComputeDiskSize(self.op.disk_template,
4550 # Check lv size requirements
4551 if req_size is not None:
4552 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4554 for node in nodenames:
4555 info = nodeinfo[node]
4559 raise errors.OpPrereqError("Cannot get current information"
4560 " from node '%s'" % node)
4561 vg_free = info.get('vg_free', None)
4562 if not isinstance(vg_free, int):
4563 raise errors.OpPrereqError("Can't compute free disk space on"
4565 if req_size > info['vg_free']:
4566 raise errors.OpPrereqError("Not enough disk space on target node %s."
4567 " %d MB available, %d MB required" %
4568 (node, info['vg_free'], req_size))
4570 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4573 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4575 if not isinstance(result.data, objects.OS):
4576 raise errors.OpPrereqError("OS '%s' not in supported os list for"
4577 " primary node" % self.op.os_type)
4579 # bridge check on primary node
4580 bridges = [n.bridge for n in self.nics]
4581 result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
4584 raise errors.OpPrereqError("One of the target bridges '%s' does not"
4585 " exist on destination node '%s'" %
4586 (",".join(bridges), pnode.name))
4588 # memory check on primary node
4590 _CheckNodeFreeMemory(self, self.pnode.name,
4591 "creating instance %s" % self.op.instance_name,
4592 self.be_full[constants.BE_MEMORY],
4595 def Exec(self, feedback_fn):
4596 """Create and add the instance to the cluster.
4599 instance = self.op.instance_name
4600 pnode_name = self.pnode.name
4602 for nic in self.nics:
4603 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4604 nic.mac = self.cfg.GenerateMAC()
4606 ht_kind = self.op.hypervisor
4607 if ht_kind in constants.HTS_REQ_PORT:
4608 network_port = self.cfg.AllocatePort()
4612 ##if self.op.vnc_bind_address is None:
4613 ## self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4615 # this is needed because os.path.join does not accept None arguments
4616 if self.op.file_storage_dir is None:
4617 string_file_storage_dir = ""
4619 string_file_storage_dir = self.op.file_storage_dir
4621 # build the full file storage dir path
4622 file_storage_dir = os.path.normpath(os.path.join(
4623 self.cfg.GetFileStorageDir(),
4624 string_file_storage_dir, instance))
4627 disks = _GenerateDiskTemplate(self,
4628 self.op.disk_template,
4629 instance, pnode_name,
4633 self.op.file_driver,
4636 iobj = objects.Instance(name=instance, os=self.op.os_type,
4637 primary_node=pnode_name,
4638 nics=self.nics, disks=disks,
4639 disk_template=self.op.disk_template,
4641 network_port=network_port,
4642 beparams=self.op.beparams,
4643 hvparams=self.op.hvparams,
4644 hypervisor=self.op.hypervisor,
4647 feedback_fn("* creating instance disks...")
4649 _CreateDisks(self, iobj)
4650 except errors.OpExecError:
4651 self.LogWarning("Device creation failed, reverting...")
4653 _RemoveDisks(self, iobj)
4655 self.cfg.ReleaseDRBDMinors(instance)
4658 feedback_fn("adding instance %s to cluster config" % instance)
4660 self.cfg.AddInstance(iobj)
4661 # Declare that we don't want to remove the instance lock anymore, as we've
4662 # added the instance to the config
4663 del self.remove_locks[locking.LEVEL_INSTANCE]
4664 # Unlock all the nodes
4665 if self.op.mode == constants.INSTANCE_IMPORT:
4666 nodes_keep = [self.op.src_node]
4667 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4668 if node != self.op.src_node]
4669 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4670 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4672 self.context.glm.release(locking.LEVEL_NODE)
4673 del self.acquired_locks[locking.LEVEL_NODE]
4675 if self.op.wait_for_sync:
4676 disk_abort = not _WaitForSync(self, iobj)
4677 elif iobj.disk_template in constants.DTS_NET_MIRROR:
4678 # make sure the disks are not degraded (still sync-ing is ok)
4680 feedback_fn("* checking mirrors status")
4681 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4686 _RemoveDisks(self, iobj)
4687 self.cfg.RemoveInstance(iobj.name)
4688 # Make sure the instance lock gets removed
4689 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4690 raise errors.OpExecError("There are some degraded disks for"
4693 feedback_fn("creating os for instance %s on node %s" %
4694 (instance, pnode_name))
4696 if iobj.disk_template != constants.DT_DISKLESS:
4697 if self.op.mode == constants.INSTANCE_CREATE:
4698 feedback_fn("* running the instance OS create scripts...")
4699 result = self.rpc.call_instance_os_add(pnode_name, iobj)
4700 msg = result.RemoteFailMsg()
4702 raise errors.OpExecError("Could not add os for instance %s"
4704 (instance, pnode_name, msg))
4706 elif self.op.mode == constants.INSTANCE_IMPORT:
4707 feedback_fn("* running the instance OS import scripts...")
4708 src_node = self.op.src_node
4709 src_images = self.src_images
4710 cluster_name = self.cfg.GetClusterName()
4711 import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4712 src_node, src_images,
4714 import_result.Raise()
4715 for idx, result in enumerate(import_result.data):
4717 self.LogWarning("Could not import the image %s for instance"
4718 " %s, disk %d, on node %s" %
4719 (src_images[idx], instance, idx, pnode_name))
4721 # also checked in the prereq part
4722 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4726 iobj.admin_up = True
4727 self.cfg.Update(iobj)
4728 logging.info("Starting instance %s on node %s", instance, pnode_name)
4729 feedback_fn("* starting instance...")
4730 result = self.rpc.call_instance_start(pnode_name, iobj, None)
4731 msg = result.RemoteFailMsg()
4733 raise errors.OpExecError("Could not start instance: %s" % msg)
4736 class LUConnectConsole(NoHooksLU):
4737 """Connect to an instance's console.
4739 This is somewhat special in that it returns the command line that
4740 you need to run on the master node in order to connect to the
4744 _OP_REQP = ["instance_name"]
4747 def ExpandNames(self):
4748 self._ExpandAndLockInstance()
4750 def CheckPrereq(self):
4751 """Check prerequisites.
4753 This checks that the instance is in the cluster.
4756 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4757 assert self.instance is not None, \
4758 "Cannot retrieve locked instance %s" % self.op.instance_name
4759 _CheckNodeOnline(self, self.instance.primary_node)
4761 def Exec(self, feedback_fn):
4762 """Connect to the console of an instance
4765 instance = self.instance
4766 node = instance.primary_node
4768 node_insts = self.rpc.call_instance_list([node],
4769 [instance.hypervisor])[node]
4772 if instance.name not in node_insts.data:
4773 raise errors.OpExecError("Instance %s is not running." % instance.name)
4775 logging.debug("Connecting to console of %s on %s", instance.name, node)
4777 hyper = hypervisor.GetHypervisor(instance.hypervisor)
4778 cluster = self.cfg.GetClusterInfo()
4779 # beparams and hvparams are passed separately, to avoid editing the
4780 # instance and then saving the defaults in the instance itself.
4781 hvparams = cluster.FillHV(instance)
4782 beparams = cluster.FillBE(instance)
4783 console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
4786 return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4789 class LUReplaceDisks(LogicalUnit):
4790 """Replace the disks of an instance.
4793 HPATH = "mirrors-replace"
4794 HTYPE = constants.HTYPE_INSTANCE
4795 _OP_REQP = ["instance_name", "mode", "disks"]
4798 def CheckArguments(self):
4799 if not hasattr(self.op, "remote_node"):
4800 self.op.remote_node = None
4801 if not hasattr(self.op, "iallocator"):
4802 self.op.iallocator = None
4804 # check for valid parameter combination
4805 cnt = [self.op.remote_node, self.op.iallocator].count(None)
4806 if self.op.mode == constants.REPLACE_DISK_CHG:
4808 raise errors.OpPrereqError("When changing the secondary either an"
4809 " iallocator script must be used or the"
4812 raise errors.OpPrereqError("Give either the iallocator or the new"
4813 " secondary, not both")
4814 else: # not replacing the secondary
4816 raise errors.OpPrereqError("The iallocator and new node options can"
4817 " be used only when changing the"
4820 def ExpandNames(self):
4821 self._ExpandAndLockInstance()
4823 if self.op.iallocator is not None:
4824 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4825 elif self.op.remote_node is not None:
4826 remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4827 if remote_node is None:
4828 raise errors.OpPrereqError("Node '%s' not known" %
4829 self.op.remote_node)
4830 self.op.remote_node = remote_node
4831 # Warning: do not remove the locking of the new secondary here
4832 # unless DRBD8.AddChildren is changed to work in parallel;
4833 # currently it doesn't since parallel invocations of
4834 # FindUnusedMinor will conflict
4835 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4836 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4838 self.needed_locks[locking.LEVEL_NODE] = []
4839 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4841 def DeclareLocks(self, level):
4842 # If we're not already locking all nodes in the set we have to declare the
4843 # instance's primary/secondary nodes.
4844 if (level == locking.LEVEL_NODE and
4845 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4846 self._LockInstancesNodes()
4848 def _RunAllocator(self):
4849 """Compute a new secondary node using an IAllocator.
4852 ial = IAllocator(self,
4853 mode=constants.IALLOCATOR_MODE_RELOC,
4854 name=self.op.instance_name,
4855 relocate_from=[self.sec_node])
4857 ial.Run(self.op.iallocator)
4860 raise errors.OpPrereqError("Can't compute nodes using"
4861 " iallocator '%s': %s" % (self.op.iallocator,
4863 if len(ial.nodes) != ial.required_nodes:
4864 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4865 " of nodes (%s), required %s" %
4866 (len(ial.nodes), ial.required_nodes))
4867 self.op.remote_node = ial.nodes[0]
4868 self.LogInfo("Selected new secondary for the instance: %s",
4869 self.op.remote_node)
4871 def BuildHooksEnv(self):
4874 This runs on the master, the primary and all the secondaries.
4878 "MODE": self.op.mode,
4879 "NEW_SECONDARY": self.op.remote_node,
4880 "OLD_SECONDARY": self.instance.secondary_nodes[0],
4882 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4884 self.cfg.GetMasterNode(),
4885 self.instance.primary_node,
4887 if self.op.remote_node is not None:
4888 nl.append(self.op.remote_node)
4891 def CheckPrereq(self):
4892 """Check prerequisites.
4894 This checks that the instance is in the cluster.
4897 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4898 assert instance is not None, \
4899 "Cannot retrieve locked instance %s" % self.op.instance_name
4900 self.instance = instance
4902 if instance.disk_template != constants.DT_DRBD8:
4903 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
4906 if len(instance.secondary_nodes) != 1:
4907 raise errors.OpPrereqError("The instance has a strange layout,"
4908 " expected one secondary but found %d" %
4909 len(instance.secondary_nodes))
4911 self.sec_node = instance.secondary_nodes[0]
4913 if self.op.iallocator is not None:
4914 self._RunAllocator()
4916 remote_node = self.op.remote_node
4917 if remote_node is not None:
4918 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4919 assert self.remote_node_info is not None, \
4920 "Cannot retrieve locked node %s" % remote_node
4922 self.remote_node_info = None
4923 if remote_node == instance.primary_node:
4924 raise errors.OpPrereqError("The specified node is the primary node of"
4926 elif remote_node == self.sec_node:
4927 raise errors.OpPrereqError("The specified node is already the"
4928 " secondary node of the instance.")
4930 if self.op.mode == constants.REPLACE_DISK_PRI:
4931 n1 = self.tgt_node = instance.primary_node
4932 n2 = self.oth_node = self.sec_node
4933 elif self.op.mode == constants.REPLACE_DISK_SEC:
4934 n1 = self.tgt_node = self.sec_node
4935 n2 = self.oth_node = instance.primary_node
4936 elif self.op.mode == constants.REPLACE_DISK_CHG:
4937 n1 = self.new_node = remote_node
4938 n2 = self.oth_node = instance.primary_node
4939 self.tgt_node = self.sec_node
4940 _CheckNodeNotDrained(self, remote_node)
4942 raise errors.ProgrammerError("Unhandled disk replace mode")
4944 _CheckNodeOnline(self, n1)
4945 _CheckNodeOnline(self, n2)
4947 if not self.op.disks:
4948 self.op.disks = range(len(instance.disks))
4950 for disk_idx in self.op.disks:
4951 instance.FindDisk(disk_idx)
4953 def _ExecD8DiskOnly(self, feedback_fn):
4954 """Replace a disk on the primary or secondary for dbrd8.
4956 The algorithm for replace is quite complicated:
4958 1. for each disk to be replaced:
4960 1. create new LVs on the target node with unique names
4961 1. detach old LVs from the drbd device
4962 1. rename old LVs to name_replaced.<time_t>
4963 1. rename new LVs to old LVs
4964 1. attach the new LVs (with the old names now) to the drbd device
4966 1. wait for sync across all devices
4968 1. for each modified disk:
4970 1. remove old LVs (which have the name name_replaces.<time_t>)
4972 Failures are not very well handled.
4976 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4977 instance = self.instance
4979 vgname = self.cfg.GetVGName()
4982 tgt_node = self.tgt_node
4983 oth_node = self.oth_node
4985 # Step: check device activation
4986 self.proc.LogStep(1, steps_total, "check device existence")
4987 info("checking volume groups")
4988 my_vg = cfg.GetVGName()
4989 results = self.rpc.call_vg_list([oth_node, tgt_node])
4991 raise errors.OpExecError("Can't list volume groups on the nodes")
4992 for node in oth_node, tgt_node:
4994 if res.failed or not res.data or my_vg not in res.data:
4995 raise errors.OpExecError("Volume group '%s' not found on %s" %
4997 for idx, dev in enumerate(instance.disks):
4998 if idx not in self.op.disks:
5000 for node in tgt_node, oth_node:
5001 info("checking disk/%d on %s" % (idx, node))
5002 cfg.SetDiskID(dev, node)
5003 result = self.rpc.call_blockdev_find(node, dev)
5004 msg = result.RemoteFailMsg()
5005 if not msg and not result.payload:
5006 msg = "disk not found"
5008 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5011 # Step: check other node consistency
5012 self.proc.LogStep(2, steps_total, "check peer consistency")
5013 for idx, dev in enumerate(instance.disks):
5014 if idx not in self.op.disks:
5016 info("checking disk/%d consistency on %s" % (idx, oth_node))
5017 if not _CheckDiskConsistency(self, dev, oth_node,
5018 oth_node==instance.primary_node):
5019 raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
5020 " to replace disks on this node (%s)" %
5021 (oth_node, tgt_node))
5023 # Step: create new storage
5024 self.proc.LogStep(3, steps_total, "allocate new storage")
5025 for idx, dev in enumerate(instance.disks):
5026 if idx not in self.op.disks:
5029 cfg.SetDiskID(dev, tgt_node)
5030 lv_names = [".disk%d_%s" % (idx, suf)
5031 for suf in ["data", "meta"]]
5032 names = _GenerateUniqueNames(self, lv_names)
5033 lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5034 logical_id=(vgname, names[0]))
5035 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5036 logical_id=(vgname, names[1]))
5037 new_lvs = [lv_data, lv_meta]
5038 old_lvs = dev.children
5039 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
5040 info("creating new local storage on %s for %s" %
5041 (tgt_node, dev.iv_name))
5042 # we pass force_create=True to force the LVM creation
5043 for new_lv in new_lvs:
5044 _CreateBlockDev(self, tgt_node, instance, new_lv, True,
5045 _GetInstanceInfoText(instance), False)
5047 # Step: for each lv, detach+rename*2+attach
5048 self.proc.LogStep(4, steps_total, "change drbd configuration")
5049 for dev, old_lvs, new_lvs in iv_names.itervalues():
5050 info("detaching %s drbd from local storage" % dev.iv_name)
5051 result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
5054 raise errors.OpExecError("Can't detach drbd from local storage on node"
5055 " %s for device %s" % (tgt_node, dev.iv_name))
5057 #cfg.Update(instance)
5059 # ok, we created the new LVs, so now we know we have the needed
5060 # storage; as such, we proceed on the target node to rename
5061 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
5062 # using the assumption that logical_id == physical_id (which in
5063 # turn is the unique_id on that node)
5065 # FIXME(iustin): use a better name for the replaced LVs
5066 temp_suffix = int(time.time())
5067 ren_fn = lambda d, suff: (d.physical_id[0],
5068 d.physical_id[1] + "_replaced-%s" % suff)
5069 # build the rename list based on what LVs exist on the node
5071 for to_ren in old_lvs:
5072 result = self.rpc.call_blockdev_find(tgt_node, to_ren)
5073 if not result.RemoteFailMsg() and result.payload:
5075 rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
5077 info("renaming the old LVs on the target node")
5078 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5081 raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
5082 # now we rename the new LVs to the old LVs
5083 info("renaming the new LVs on the target node")
5084 rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
5085 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5088 raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
5090 for old, new in zip(old_lvs, new_lvs):
5091 new.logical_id = old.logical_id
5092 cfg.SetDiskID(new, tgt_node)
5094 for disk in old_lvs:
5095 disk.logical_id = ren_fn(disk, temp_suffix)
5096 cfg.SetDiskID(disk, tgt_node)
5098 # now that the new lvs have the old name, we can add them to the device
5099 info("adding new mirror component on %s" % tgt_node)
5100 result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
5101 if result.failed or not result.data:
5102 for new_lv in new_lvs:
5103 msg = self.rpc.call_blockdev_remove(tgt_node, new_lv).RemoteFailMsg()
5105 warning("Can't rollback device %s: %s", dev, msg,
5106 hint="cleanup manually the unused logical volumes")
5107 raise errors.OpExecError("Can't add local storage to drbd")
5109 dev.children = new_lvs
5110 cfg.Update(instance)
5112 # Step: wait for sync
5114 # this can fail as the old devices are degraded and _WaitForSync
5115 # does a combined result over all disks, so we don't check its
5117 self.proc.LogStep(5, steps_total, "sync devices")
5118 _WaitForSync(self, instance, unlock=True)
5120 # so check manually all the devices
5121 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5122 cfg.SetDiskID(dev, instance.primary_node)
5123 result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5124 msg = result.RemoteFailMsg()
5125 if not msg and not result.payload:
5126 msg = "disk not found"
5128 raise errors.OpExecError("Can't find DRBD device %s: %s" %
5130 if result.payload[5]:
5131 raise errors.OpExecError("DRBD device %s is degraded!" % name)
5133 # Step: remove old storage
5134 self.proc.LogStep(6, steps_total, "removing old storage")
5135 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5136 info("remove logical volumes for %s" % name)
5138 cfg.SetDiskID(lv, tgt_node)
5139 msg = self.rpc.call_blockdev_remove(tgt_node, lv).RemoteFailMsg()
5141 warning("Can't remove old LV: %s" % msg,
5142 hint="manually remove unused LVs")
5145 def _ExecD8Secondary(self, feedback_fn):
5146 """Replace the secondary node for drbd8.
5148 The algorithm for replace is quite complicated:
5149 - for all disks of the instance:
5150 - create new LVs on the new node with same names
5151 - shutdown the drbd device on the old secondary
5152 - disconnect the drbd network on the primary
5153 - create the drbd device on the new secondary
5154 - network attach the drbd on the primary, using an artifice:
5155 the drbd code for Attach() will connect to the network if it
5156 finds a device which is connected to the good local disks but
5158 - wait for sync across all devices
5159 - remove all disks from the old secondary
5161 Failures are not very well handled.
5165 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5166 instance = self.instance
5170 old_node = self.tgt_node
5171 new_node = self.new_node
5172 pri_node = instance.primary_node
5174 old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5175 new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5176 pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5179 # Step: check device activation
5180 self.proc.LogStep(1, steps_total, "check device existence")
5181 info("checking volume groups")
5182 my_vg = cfg.GetVGName()
5183 results = self.rpc.call_vg_list([pri_node, new_node])
5184 for node in pri_node, new_node:
5186 if res.failed or not res.data or my_vg not in res.data:
5187 raise errors.OpExecError("Volume group '%s' not found on %s" %
5189 for idx, dev in enumerate(instance.disks):
5190 if idx not in self.op.disks:
5192 info("checking disk/%d on %s" % (idx, pri_node))
5193 cfg.SetDiskID(dev, pri_node)
5194 result = self.rpc.call_blockdev_find(pri_node, dev)
5195 msg = result.RemoteFailMsg()
5196 if not msg and not result.payload:
5197 msg = "disk not found"
5199 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5200 (idx, pri_node, msg))
5202 # Step: check other node consistency
5203 self.proc.LogStep(2, steps_total, "check peer consistency")
5204 for idx, dev in enumerate(instance.disks):
5205 if idx not in self.op.disks:
5207 info("checking disk/%d consistency on %s" % (idx, pri_node))
5208 if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5209 raise errors.OpExecError("Primary node (%s) has degraded storage,"
5210 " unsafe to replace the secondary" %
5213 # Step: create new storage
5214 self.proc.LogStep(3, steps_total, "allocate new storage")
5215 for idx, dev in enumerate(instance.disks):
5216 info("adding new local storage on %s for disk/%d" %
5218 # we pass force_create=True to force LVM creation
5219 for new_lv in dev.children:
5220 _CreateBlockDev(self, new_node, instance, new_lv, True,
5221 _GetInstanceInfoText(instance), False)
5223 # Step 4: dbrd minors and drbd setups changes
5224 # after this, we must manually remove the drbd minors on both the
5225 # error and the success paths
5226 minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5228 logging.debug("Allocated minors %s" % (minors,))
5229 self.proc.LogStep(4, steps_total, "changing drbd configuration")
5230 for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5232 info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5233 # create new devices on new_node; note that we create two IDs:
5234 # one without port, so the drbd will be activated without
5235 # networking information on the new node at this stage, and one
5236 # with network, for the latter activation in step 4
5237 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5238 if pri_node == o_node1:
5243 new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5244 new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5246 iv_names[idx] = (dev, dev.children, new_net_id)
5247 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5249 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5250 logical_id=new_alone_id,
5251 children=dev.children)
5253 _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5254 _GetInstanceInfoText(instance), False)
5255 except errors.BlockDeviceError:
5256 self.cfg.ReleaseDRBDMinors(instance.name)
5259 for idx, dev in enumerate(instance.disks):
5260 # we have new devices, shutdown the drbd on the old secondary
5261 info("shutting down drbd for disk/%d on old node" % idx)
5262 cfg.SetDiskID(dev, old_node)
5263 msg = self.rpc.call_blockdev_shutdown(old_node, dev).RemoteFailMsg()
5265 warning("Failed to shutdown drbd for disk/%d on old node: %s" %
5267 hint="Please cleanup this device manually as soon as possible")
5269 info("detaching primary drbds from the network (=> standalone)")
5270 result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5271 instance.disks)[pri_node]
5273 msg = result.RemoteFailMsg()
5275 # detaches didn't succeed (unlikely)
5276 self.cfg.ReleaseDRBDMinors(instance.name)
5277 raise errors.OpExecError("Can't detach the disks from the network on"
5278 " old node: %s" % (msg,))
5280 # if we managed to detach at least one, we update all the disks of
5281 # the instance to point to the new secondary
5282 info("updating instance configuration")
5283 for dev, _, new_logical_id in iv_names.itervalues():
5284 dev.logical_id = new_logical_id
5285 cfg.SetDiskID(dev, pri_node)
5286 cfg.Update(instance)
5288 # and now perform the drbd attach
5289 info("attaching primary drbds to new secondary (standalone => connected)")
5290 result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5291 instance.disks, instance.name,
5293 for to_node, to_result in result.items():
5294 msg = to_result.RemoteFailMsg()
5296 warning("can't attach drbd disks on node %s: %s", to_node, msg,
5297 hint="please do a gnt-instance info to see the"
5300 # this can fail as the old devices are degraded and _WaitForSync
5301 # does a combined result over all disks, so we don't check its
5303 self.proc.LogStep(5, steps_total, "sync devices")
5304 _WaitForSync(self, instance, unlock=True)
5306 # so check manually all the devices
5307 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5308 cfg.SetDiskID(dev, pri_node)
5309 result = self.rpc.call_blockdev_find(pri_node, dev)
5310 msg = result.RemoteFailMsg()
5311 if not msg and not result.payload:
5312 msg = "disk not found"
5314 raise errors.OpExecError("Can't find DRBD device disk/%d: %s" %
5316 if result.payload[5]:
5317 raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5319 self.proc.LogStep(6, steps_total, "removing old storage")
5320 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5321 info("remove logical volumes for disk/%d" % idx)
5323 cfg.SetDiskID(lv, old_node)
5324 msg = self.rpc.call_blockdev_remove(old_node, lv).RemoteFailMsg()
5326 warning("Can't remove LV on old secondary: %s", msg,
5327 hint="Cleanup stale volumes by hand")
5329 def Exec(self, feedback_fn):
5330 """Execute disk replacement.
5332 This dispatches the disk replacement to the appropriate handler.
5335 instance = self.instance
5337 # Activate the instance disks if we're replacing them on a down instance
5338 if not instance.admin_up:
5339 _StartInstanceDisks(self, instance, True)
5341 if self.op.mode == constants.REPLACE_DISK_CHG:
5342 fn = self._ExecD8Secondary
5344 fn = self._ExecD8DiskOnly
5346 ret = fn(feedback_fn)
5348 # Deactivate the instance disks if we're replacing them on a down instance
5349 if not instance.admin_up:
5350 _SafeShutdownInstanceDisks(self, instance)
5355 class LUGrowDisk(LogicalUnit):
5356 """Grow a disk of an instance.
5360 HTYPE = constants.HTYPE_INSTANCE
5361 _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5364 def ExpandNames(self):
5365 self._ExpandAndLockInstance()
5366 self.needed_locks[locking.LEVEL_NODE] = []
5367 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5369 def DeclareLocks(self, level):
5370 if level == locking.LEVEL_NODE:
5371 self._LockInstancesNodes()
5373 def BuildHooksEnv(self):
5376 This runs on the master, the primary and all the secondaries.
5380 "DISK": self.op.disk,
5381 "AMOUNT": self.op.amount,
5383 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5385 self.cfg.GetMasterNode(),
5386 self.instance.primary_node,
5390 def CheckPrereq(self):
5391 """Check prerequisites.
5393 This checks that the instance is in the cluster.
5396 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5397 assert instance is not None, \
5398 "Cannot retrieve locked instance %s" % self.op.instance_name
5399 nodenames = list(instance.all_nodes)
5400 for node in nodenames:
5401 _CheckNodeOnline(self, node)
5404 self.instance = instance
5406 if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5407 raise errors.OpPrereqError("Instance's disk layout does not support"
5410 self.disk = instance.FindDisk(self.op.disk)
5412 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5413 instance.hypervisor)
5414 for node in nodenames:
5415 info = nodeinfo[node]
5416 if info.failed or not info.data:
5417 raise errors.OpPrereqError("Cannot get current information"
5418 " from node '%s'" % node)
5419 vg_free = info.data.get('vg_free', None)
5420 if not isinstance(vg_free, int):
5421 raise errors.OpPrereqError("Can't compute free disk space on"
5423 if self.op.amount > vg_free:
5424 raise errors.OpPrereqError("Not enough disk space on target node %s:"
5425 " %d MiB available, %d MiB required" %
5426 (node, vg_free, self.op.amount))
5428 def Exec(self, feedback_fn):
5429 """Execute disk grow.
5432 instance = self.instance
5434 for node in instance.all_nodes:
5435 self.cfg.SetDiskID(disk, node)
5436 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5437 msg = result.RemoteFailMsg()
5439 raise errors.OpExecError("Grow request failed to node %s: %s" %
5441 disk.RecordGrow(self.op.amount)
5442 self.cfg.Update(instance)
5443 if self.op.wait_for_sync:
5444 disk_abort = not _WaitForSync(self, instance)
5446 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5447 " status.\nPlease check the instance.")
5450 class LUQueryInstanceData(NoHooksLU):
5451 """Query runtime instance data.
5454 _OP_REQP = ["instances", "static"]
5457 def ExpandNames(self):
5458 self.needed_locks = {}
5459 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5461 if not isinstance(self.op.instances, list):
5462 raise errors.OpPrereqError("Invalid argument type 'instances'")
5464 if self.op.instances:
5465 self.wanted_names = []
5466 for name in self.op.instances:
5467 full_name = self.cfg.ExpandInstanceName(name)
5468 if full_name is None:
5469 raise errors.OpPrereqError("Instance '%s' not known" % name)
5470 self.wanted_names.append(full_name)
5471 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5473 self.wanted_names = None
5474 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5476 self.needed_locks[locking.LEVEL_NODE] = []
5477 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5479 def DeclareLocks(self, level):
5480 if level == locking.LEVEL_NODE:
5481 self._LockInstancesNodes()
5483 def CheckPrereq(self):
5484 """Check prerequisites.
5486 This only checks the optional instance list against the existing names.
5489 if self.wanted_names is None:
5490 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5492 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5493 in self.wanted_names]
5496 def _ComputeDiskStatus(self, instance, snode, dev):
5497 """Compute block device status.
5500 static = self.op.static
5502 self.cfg.SetDiskID(dev, instance.primary_node)
5503 dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5504 msg = dev_pstatus.RemoteFailMsg()
5506 raise errors.OpExecError("Can't compute disk status for %s: %s" %
5507 (instance.name, msg))
5508 dev_pstatus = dev_pstatus.payload
5512 if dev.dev_type in constants.LDS_DRBD:
5513 # we change the snode then (otherwise we use the one passed in)
5514 if dev.logical_id[0] == instance.primary_node:
5515 snode = dev.logical_id[1]
5517 snode = dev.logical_id[0]
5519 if snode and not static:
5520 self.cfg.SetDiskID(dev, snode)
5521 dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5522 msg = dev_sstatus.RemoteFailMsg()
5524 raise errors.OpExecError("Can't compute disk status for %s: %s" %
5525 (instance.name, msg))
5526 dev_sstatus = dev_sstatus.payload
5531 dev_children = [self._ComputeDiskStatus(instance, snode, child)
5532 for child in dev.children]
5537 "iv_name": dev.iv_name,
5538 "dev_type": dev.dev_type,
5539 "logical_id": dev.logical_id,
5540 "physical_id": dev.physical_id,
5541 "pstatus": dev_pstatus,
5542 "sstatus": dev_sstatus,
5543 "children": dev_children,
5549 def Exec(self, feedback_fn):
5550 """Gather and return data"""
5553 cluster = self.cfg.GetClusterInfo()
5555 for instance in self.wanted_instances:
5556 if not self.op.static:
5557 remote_info = self.rpc.call_instance_info(instance.primary_node,
5559 instance.hypervisor)
5561 remote_info = remote_info.data
5562 if remote_info and "state" in remote_info:
5565 remote_state = "down"
5568 if instance.admin_up:
5571 config_state = "down"
5573 disks = [self._ComputeDiskStatus(instance, None, device)
5574 for device in instance.disks]
5577 "name": instance.name,
5578 "config_state": config_state,
5579 "run_state": remote_state,
5580 "pnode": instance.primary_node,
5581 "snodes": instance.secondary_nodes,
5583 "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5585 "hypervisor": instance.hypervisor,
5586 "network_port": instance.network_port,
5587 "hv_instance": instance.hvparams,
5588 "hv_actual": cluster.FillHV(instance),
5589 "be_instance": instance.beparams,
5590 "be_actual": cluster.FillBE(instance),
5593 result[instance.name] = idict
5598 class LUSetInstanceParams(LogicalUnit):
5599 """Modifies an instances's parameters.
5602 HPATH = "instance-modify"
5603 HTYPE = constants.HTYPE_INSTANCE
5604 _OP_REQP = ["instance_name"]
5607 def CheckArguments(self):
5608 if not hasattr(self.op, 'nics'):
5610 if not hasattr(self.op, 'disks'):
5612 if not hasattr(self.op, 'beparams'):
5613 self.op.beparams = {}
5614 if not hasattr(self.op, 'hvparams'):
5615 self.op.hvparams = {}
5616 self.op.force = getattr(self.op, "force", False)
5617 if not (self.op.nics or self.op.disks or
5618 self.op.hvparams or self.op.beparams):
5619 raise errors.OpPrereqError("No changes submitted")
5623 for disk_op, disk_dict in self.op.disks:
5624 if disk_op == constants.DDM_REMOVE:
5627 elif disk_op == constants.DDM_ADD:
5630 if not isinstance(disk_op, int):
5631 raise errors.OpPrereqError("Invalid disk index")
5632 if disk_op == constants.DDM_ADD:
5633 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5634 if mode not in constants.DISK_ACCESS_SET:
5635 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5636 size = disk_dict.get('size', None)
5638 raise errors.OpPrereqError("Required disk parameter size missing")
5641 except ValueError, err:
5642 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5644 disk_dict['size'] = size
5646 # modification of disk
5647 if 'size' in disk_dict:
5648 raise errors.OpPrereqError("Disk size change not possible, use"
5651 if disk_addremove > 1:
5652 raise errors.OpPrereqError("Only one disk add or remove operation"
5653 " supported at a time")
5657 for nic_op, nic_dict in self.op.nics:
5658 if nic_op == constants.DDM_REMOVE:
5661 elif nic_op == constants.DDM_ADD:
5664 if not isinstance(nic_op, int):
5665 raise errors.OpPrereqError("Invalid nic index")
5667 # nic_dict should be a dict
5668 nic_ip = nic_dict.get('ip', None)
5669 if nic_ip is not None:
5670 if nic_ip.lower() == "none":
5671 nic_dict['ip'] = None
5673 if not utils.IsValidIP(nic_ip):
5674 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5675 # we can only check None bridges and assign the default one
5676 nic_bridge = nic_dict.get('bridge', None)
5677 if nic_bridge is None:
5678 nic_dict['bridge'] = self.cfg.GetDefBridge()
5679 # but we can validate MACs
5680 nic_mac = nic_dict.get('mac', None)
5681 if nic_mac is not None:
5682 if self.cfg.IsMacInUse(nic_mac):
5683 raise errors.OpPrereqError("MAC address %s already in use"
5684 " in cluster" % nic_mac)
5685 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5686 if not utils.IsValidMac(nic_mac):
5687 raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5688 if nic_addremove > 1:
5689 raise errors.OpPrereqError("Only one NIC add or remove operation"
5690 " supported at a time")
5692 def ExpandNames(self):
5693 self._ExpandAndLockInstance()
5694 self.needed_locks[locking.LEVEL_NODE] = []
5695 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5697 def DeclareLocks(self, level):
5698 if level == locking.LEVEL_NODE:
5699 self._LockInstancesNodes()
5701 def BuildHooksEnv(self):
5704 This runs on the master, primary and secondaries.
5708 if constants.BE_MEMORY in self.be_new:
5709 args['memory'] = self.be_new[constants.BE_MEMORY]
5710 if constants.BE_VCPUS in self.be_new:
5711 args['vcpus'] = self.be_new[constants.BE_VCPUS]
5712 # FIXME: readd disk/nic changes
5713 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5714 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5717 def CheckPrereq(self):
5718 """Check prerequisites.
5720 This only checks the instance list against the existing names.
5723 force = self.force = self.op.force
5725 # checking the new params on the primary/secondary nodes
5727 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5728 assert self.instance is not None, \
5729 "Cannot retrieve locked instance %s" % self.op.instance_name
5730 pnode = instance.primary_node
5731 nodelist = list(instance.all_nodes)
5733 # hvparams processing
5734 if self.op.hvparams:
5735 i_hvdict = copy.deepcopy(instance.hvparams)
5736 for key, val in self.op.hvparams.iteritems():
5737 if val == constants.VALUE_DEFAULT:
5744 cluster = self.cfg.GetClusterInfo()
5745 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
5746 hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5749 hypervisor.GetHypervisor(
5750 instance.hypervisor).CheckParameterSyntax(hv_new)
5751 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5752 self.hv_new = hv_new # the new actual values
5753 self.hv_inst = i_hvdict # the new dict (without defaults)
5755 self.hv_new = self.hv_inst = {}
5757 # beparams processing
5758 if self.op.beparams:
5759 i_bedict = copy.deepcopy(instance.beparams)
5760 for key, val in self.op.beparams.iteritems():
5761 if val == constants.VALUE_DEFAULT:
5768 cluster = self.cfg.GetClusterInfo()
5769 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
5770 be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5772 self.be_new = be_new # the new actual values
5773 self.be_inst = i_bedict # the new dict (without defaults)
5775 self.be_new = self.be_inst = {}
5779 if constants.BE_MEMORY in self.op.beparams and not self.force:
5780 mem_check_list = [pnode]
5781 if be_new[constants.BE_AUTO_BALANCE]:
5782 # either we changed auto_balance to yes or it was from before
5783 mem_check_list.extend(instance.secondary_nodes)
5784 instance_info = self.rpc.call_instance_info(pnode, instance.name,
5785 instance.hypervisor)
5786 nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5787 instance.hypervisor)
5788 if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5789 # Assume the primary node is unreachable and go ahead
5790 self.warn.append("Can't get info from primary node %s" % pnode)
5792 if not instance_info.failed and instance_info.data:
5793 current_mem = instance_info.data['memory']
5795 # Assume instance not running
5796 # (there is a slight race condition here, but it's not very probable,
5797 # and we have no other way to check)
5799 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5800 nodeinfo[pnode].data['memory_free'])
5802 raise errors.OpPrereqError("This change will prevent the instance"
5803 " from starting, due to %d MB of memory"
5804 " missing on its primary node" % miss_mem)
5806 if be_new[constants.BE_AUTO_BALANCE]:
5807 for node, nres in nodeinfo.iteritems():
5808 if node not in instance.secondary_nodes:
5810 if nres.failed or not isinstance(nres.data, dict):
5811 self.warn.append("Can't get info from secondary node %s" % node)
5812 elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5813 self.warn.append("Not enough memory to failover instance to"
5814 " secondary node %s" % node)
5817 for nic_op, nic_dict in self.op.nics:
5818 if nic_op == constants.DDM_REMOVE:
5819 if not instance.nics:
5820 raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5822 if nic_op != constants.DDM_ADD:
5824 if nic_op < 0 or nic_op >= len(instance.nics):
5825 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5827 (nic_op, len(instance.nics)))
5828 nic_bridge = nic_dict.get('bridge', None)
5829 if nic_bridge is not None:
5830 if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5831 msg = ("Bridge '%s' doesn't exist on one of"
5832 " the instance nodes" % nic_bridge)
5834 self.warn.append(msg)
5836 raise errors.OpPrereqError(msg)
5839 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5840 raise errors.OpPrereqError("Disk operations not supported for"
5841 " diskless instances")
5842 for disk_op, disk_dict in self.op.disks:
5843 if disk_op == constants.DDM_REMOVE:
5844 if len(instance.disks) == 1:
5845 raise errors.OpPrereqError("Cannot remove the last disk of"
5847 ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5848 ins_l = ins_l[pnode]
5849 if ins_l.failed or not isinstance(ins_l.data, list):
5850 raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5851 if instance.name in ins_l.data:
5852 raise errors.OpPrereqError("Instance is running, can't remove"
5855 if (disk_op == constants.DDM_ADD and
5856 len(instance.nics) >= constants.MAX_DISKS):
5857 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5858 " add more" % constants.MAX_DISKS)
5859 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5861 if disk_op < 0 or disk_op >= len(instance.disks):
5862 raise errors.OpPrereqError("Invalid disk index %s, valid values"
5864 (disk_op, len(instance.disks)))
5868 def Exec(self, feedback_fn):
5869 """Modifies an instance.
5871 All parameters take effect only at the next restart of the instance.
5874 # Process here the warnings from CheckPrereq, as we don't have a
5875 # feedback_fn there.
5876 for warn in self.warn:
5877 feedback_fn("WARNING: %s" % warn)
5880 instance = self.instance
5882 for disk_op, disk_dict in self.op.disks:
5883 if disk_op == constants.DDM_REMOVE:
5884 # remove the last disk
5885 device = instance.disks.pop()
5886 device_idx = len(instance.disks)
5887 for node, disk in device.ComputeNodeTree(instance.primary_node):
5888 self.cfg.SetDiskID(disk, node)
5889 msg = self.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
5891 self.LogWarning("Could not remove disk/%d on node %s: %s,"
5892 " continuing anyway", device_idx, node, msg)
5893 result.append(("disk/%d" % device_idx, "remove"))
5894 elif disk_op == constants.DDM_ADD:
5896 if instance.disk_template == constants.DT_FILE:
5897 file_driver, file_path = instance.disks[0].logical_id
5898 file_path = os.path.dirname(file_path)
5900 file_driver = file_path = None
5901 disk_idx_base = len(instance.disks)
5902 new_disk = _GenerateDiskTemplate(self,
5903 instance.disk_template,
5904 instance.name, instance.primary_node,
5905 instance.secondary_nodes,
5910 instance.disks.append(new_disk)
5911 info = _GetInstanceInfoText(instance)
5913 logging.info("Creating volume %s for instance %s",
5914 new_disk.iv_name, instance.name)
5915 # Note: this needs to be kept in sync with _CreateDisks
5917 for node in instance.all_nodes:
5918 f_create = node == instance.primary_node
5920 _CreateBlockDev(self, node, instance, new_disk,
5921 f_create, info, f_create)
5922 except errors.OpExecError, err:
5923 self.LogWarning("Failed to create volume %s (%s) on"
5925 new_disk.iv_name, new_disk, node, err)
5926 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5927 (new_disk.size, new_disk.mode)))
5929 # change a given disk
5930 instance.disks[disk_op].mode = disk_dict['mode']
5931 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5933 for nic_op, nic_dict in self.op.nics:
5934 if nic_op == constants.DDM_REMOVE:
5935 # remove the last nic
5936 del instance.nics[-1]
5937 result.append(("nic.%d" % len(instance.nics), "remove"))
5938 elif nic_op == constants.DDM_ADD:
5940 if 'mac' not in nic_dict:
5941 mac = constants.VALUE_GENERATE
5943 mac = nic_dict['mac']
5944 if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5945 mac = self.cfg.GenerateMAC()
5946 new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5947 bridge=nic_dict.get('bridge', None))
5948 instance.nics.append(new_nic)
5949 result.append(("nic.%d" % (len(instance.nics) - 1),
5950 "add:mac=%s,ip=%s,bridge=%s" %
5951 (new_nic.mac, new_nic.ip, new_nic.bridge)))
5953 # change a given nic
5954 for key in 'mac', 'ip', 'bridge':
5956 setattr(instance.nics[nic_op], key, nic_dict[key])
5957 result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5960 if self.op.hvparams:
5961 instance.hvparams = self.hv_inst
5962 for key, val in self.op.hvparams.iteritems():
5963 result.append(("hv/%s" % key, val))
5966 if self.op.beparams:
5967 instance.beparams = self.be_inst
5968 for key, val in self.op.beparams.iteritems():
5969 result.append(("be/%s" % key, val))
5971 self.cfg.Update(instance)
5976 class LUQueryExports(NoHooksLU):
5977 """Query the exports list
5980 _OP_REQP = ['nodes']
5983 def ExpandNames(self):
5984 self.needed_locks = {}
5985 self.share_locks[locking.LEVEL_NODE] = 1
5986 if not self.op.nodes:
5987 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5989 self.needed_locks[locking.LEVEL_NODE] = \
5990 _GetWantedNodes(self, self.op.nodes)
5992 def CheckPrereq(self):
5993 """Check prerequisites.
5996 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5998 def Exec(self, feedback_fn):
5999 """Compute the list of all the exported system images.
6002 @return: a dictionary with the structure node->(export-list)
6003 where export-list is a list of the instances exported on
6007 rpcresult = self.rpc.call_export_list(self.nodes)
6009 for node in rpcresult:
6010 if rpcresult[node].failed:
6011 result[node] = False
6013 result[node] = rpcresult[node].data
6018 class LUExportInstance(LogicalUnit):
6019 """Export an instance to an image in the cluster.
6022 HPATH = "instance-export"
6023 HTYPE = constants.HTYPE_INSTANCE
6024 _OP_REQP = ["instance_name", "target_node", "shutdown"]
6027 def ExpandNames(self):
6028 self._ExpandAndLockInstance()
6029 # FIXME: lock only instance primary and destination node
6031 # Sad but true, for now we have do lock all nodes, as we don't know where
6032 # the previous export might be, and and in this LU we search for it and
6033 # remove it from its current node. In the future we could fix this by:
6034 # - making a tasklet to search (share-lock all), then create the new one,
6035 # then one to remove, after
6036 # - removing the removal operation altoghether
6037 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6039 def DeclareLocks(self, level):
6040 """Last minute lock declaration."""
6041 # All nodes are locked anyway, so nothing to do here.
6043 def BuildHooksEnv(self):
6046 This will run on the master, primary node and target node.
6050 "EXPORT_NODE": self.op.target_node,
6051 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
6053 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6054 nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
6055 self.op.target_node]
6058 def CheckPrereq(self):
6059 """Check prerequisites.
6061 This checks that the instance and node names are valid.
6064 instance_name = self.op.instance_name
6065 self.instance = self.cfg.GetInstanceInfo(instance_name)
6066 assert self.instance is not None, \
6067 "Cannot retrieve locked instance %s" % self.op.instance_name
6068 _CheckNodeOnline(self, self.instance.primary_node)
6070 self.dst_node = self.cfg.GetNodeInfo(
6071 self.cfg.ExpandNodeName(self.op.target_node))
6073 if self.dst_node is None:
6074 # This is wrong node name, not a non-locked node
6075 raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
6076 _CheckNodeOnline(self, self.dst_node.name)
6077 _CheckNodeNotDrained(self, self.dst_node.name)
6079 # instance disk type verification
6080 for disk in self.instance.disks:
6081 if disk.dev_type == constants.LD_FILE:
6082 raise errors.OpPrereqError("Export not supported for instances with"
6083 " file-based disks")
6085 def Exec(self, feedback_fn):
6086 """Export an instance to an image in the cluster.
6089 instance = self.instance
6090 dst_node = self.dst_node
6091 src_node = instance.primary_node
6092 if self.op.shutdown:
6093 # shutdown the instance, but not the disks
6094 result = self.rpc.call_instance_shutdown(src_node, instance)
6095 msg = result.RemoteFailMsg()
6097 raise errors.OpExecError("Could not shutdown instance %s on"
6099 (instance.name, src_node, msg))
6101 vgname = self.cfg.GetVGName()
6105 # set the disks ID correctly since call_instance_start needs the
6106 # correct drbd minor to create the symlinks
6107 for disk in instance.disks:
6108 self.cfg.SetDiskID(disk, src_node)
6111 for disk in instance.disks:
6112 # new_dev_name will be a snapshot of an lvm leaf of the one we passed
6113 new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
6114 if new_dev_name.failed or not new_dev_name.data:
6115 self.LogWarning("Could not snapshot block device %s on node %s",
6116 disk.logical_id[1], src_node)
6117 snap_disks.append(False)
6119 new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
6120 logical_id=(vgname, new_dev_name.data),
6121 physical_id=(vgname, new_dev_name.data),
6122 iv_name=disk.iv_name)
6123 snap_disks.append(new_dev)
6126 if self.op.shutdown and instance.admin_up:
6127 result = self.rpc.call_instance_start(src_node, instance, None)
6128 msg = result.RemoteFailMsg()
6130 _ShutdownInstanceDisks(self, instance)
6131 raise errors.OpExecError("Could not start instance: %s" % msg)
6133 # TODO: check for size
6135 cluster_name = self.cfg.GetClusterName()
6136 for idx, dev in enumerate(snap_disks):
6138 result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6139 instance, cluster_name, idx)
6140 if result.failed or not result.data:
6141 self.LogWarning("Could not export block device %s from node %s to"
6142 " node %s", dev.logical_id[1], src_node,
6144 msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
6146 self.LogWarning("Could not remove snapshot block device %s from node"
6147 " %s: %s", dev.logical_id[1], src_node, msg)
6149 result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6150 if result.failed or not result.data:
6151 self.LogWarning("Could not finalize export for instance %s on node %s",
6152 instance.name, dst_node.name)
6154 nodelist = self.cfg.GetNodeList()
6155 nodelist.remove(dst_node.name)
6157 # on one-node clusters nodelist will be empty after the removal
6158 # if we proceed the backup would be removed because OpQueryExports
6159 # substitutes an empty list with the full cluster node list.
6161 exportlist = self.rpc.call_export_list(nodelist)
6162 for node in exportlist:
6163 if exportlist[node].failed:
6165 if instance.name in exportlist[node].data:
6166 if not self.rpc.call_export_remove(node, instance.name):
6167 self.LogWarning("Could not remove older export for instance %s"
6168 " on node %s", instance.name, node)
6171 class LURemoveExport(NoHooksLU):
6172 """Remove exports related to the named instance.
6175 _OP_REQP = ["instance_name"]
6178 def ExpandNames(self):
6179 self.needed_locks = {}
6180 # We need all nodes to be locked in order for RemoveExport to work, but we
6181 # don't need to lock the instance itself, as nothing will happen to it (and
6182 # we can remove exports also for a removed instance)
6183 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6185 def CheckPrereq(self):
6186 """Check prerequisites.
6190 def Exec(self, feedback_fn):
6191 """Remove any export.
6194 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6195 # If the instance was not found we'll try with the name that was passed in.
6196 # This will only work if it was an FQDN, though.
6198 if not instance_name:
6200 instance_name = self.op.instance_name
6202 exportlist = self.rpc.call_export_list(self.acquired_locks[
6203 locking.LEVEL_NODE])
6205 for node in exportlist:
6206 if exportlist[node].failed:
6207 self.LogWarning("Failed to query node %s, continuing" % node)
6209 if instance_name in exportlist[node].data:
6211 result = self.rpc.call_export_remove(node, instance_name)
6212 if result.failed or not result.data:
6213 logging.error("Could not remove export for instance %s"
6214 " on node %s", instance_name, node)
6216 if fqdn_warn and not found:
6217 feedback_fn("Export not found. If trying to remove an export belonging"
6218 " to a deleted instance please use its Fully Qualified"
6222 class TagsLU(NoHooksLU):
6225 This is an abstract class which is the parent of all the other tags LUs.
6229 def ExpandNames(self):
6230 self.needed_locks = {}
6231 if self.op.kind == constants.TAG_NODE:
6232 name = self.cfg.ExpandNodeName(self.op.name)
6234 raise errors.OpPrereqError("Invalid node name (%s)" %
6237 self.needed_locks[locking.LEVEL_NODE] = name
6238 elif self.op.kind == constants.TAG_INSTANCE:
6239 name = self.cfg.ExpandInstanceName(self.op.name)
6241 raise errors.OpPrereqError("Invalid instance name (%s)" %
6244 self.needed_locks[locking.LEVEL_INSTANCE] = name
6246 def CheckPrereq(self):
6247 """Check prerequisites.
6250 if self.op.kind == constants.TAG_CLUSTER:
6251 self.target = self.cfg.GetClusterInfo()
6252 elif self.op.kind == constants.TAG_NODE:
6253 self.target = self.cfg.GetNodeInfo(self.op.name)
6254 elif self.op.kind == constants.TAG_INSTANCE:
6255 self.target = self.cfg.GetInstanceInfo(self.op.name)
6257 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6261 class LUGetTags(TagsLU):
6262 """Returns the tags of a given object.
6265 _OP_REQP = ["kind", "name"]
6268 def Exec(self, feedback_fn):
6269 """Returns the tag list.
6272 return list(self.target.GetTags())
6275 class LUSearchTags(NoHooksLU):
6276 """Searches the tags for a given pattern.
6279 _OP_REQP = ["pattern"]
6282 def ExpandNames(self):
6283 self.needed_locks = {}
6285 def CheckPrereq(self):
6286 """Check prerequisites.
6288 This checks the pattern passed for validity by compiling it.
6292 self.re = re.compile(self.op.pattern)
6293 except re.error, err:
6294 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6295 (self.op.pattern, err))
6297 def Exec(self, feedback_fn):
6298 """Returns the tag list.
6302 tgts = [("/cluster", cfg.GetClusterInfo())]
6303 ilist = cfg.GetAllInstancesInfo().values()
6304 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6305 nlist = cfg.GetAllNodesInfo().values()
6306 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6308 for path, target in tgts:
6309 for tag in target.GetTags():
6310 if self.re.search(tag):
6311 results.append((path, tag))
6315 class LUAddTags(TagsLU):
6316 """Sets a tag on a given object.
6319 _OP_REQP = ["kind", "name", "tags"]
6322 def CheckPrereq(self):
6323 """Check prerequisites.
6325 This checks the type and length of the tag name and value.
6328 TagsLU.CheckPrereq(self)
6329 for tag in self.op.tags:
6330 objects.TaggableObject.ValidateTag(tag)
6332 def Exec(self, feedback_fn):
6337 for tag in self.op.tags:
6338 self.target.AddTag(tag)
6339 except errors.TagError, err:
6340 raise errors.OpExecError("Error while setting tag: %s" % str(err))
6342 self.cfg.Update(self.target)
6343 except errors.ConfigurationError:
6344 raise errors.OpRetryError("There has been a modification to the"
6345 " config file and the operation has been"
6346 " aborted. Please retry.")
6349 class LUDelTags(TagsLU):
6350 """Delete a list of tags from a given object.
6353 _OP_REQP = ["kind", "name", "tags"]
6356 def CheckPrereq(self):
6357 """Check prerequisites.
6359 This checks that we have the given tag.
6362 TagsLU.CheckPrereq(self)
6363 for tag in self.op.tags:
6364 objects.TaggableObject.ValidateTag(tag)
6365 del_tags = frozenset(self.op.tags)
6366 cur_tags = self.target.GetTags()
6367 if not del_tags <= cur_tags:
6368 diff_tags = del_tags - cur_tags
6369 diff_names = ["'%s'" % tag for tag in diff_tags]
6371 raise errors.OpPrereqError("Tag(s) %s not found" %
6372 (",".join(diff_names)))
6374 def Exec(self, feedback_fn):
6375 """Remove the tag from the object.
6378 for tag in self.op.tags:
6379 self.target.RemoveTag(tag)
6381 self.cfg.Update(self.target)
6382 except errors.ConfigurationError:
6383 raise errors.OpRetryError("There has been a modification to the"
6384 " config file and the operation has been"
6385 " aborted. Please retry.")
6388 class LUTestDelay(NoHooksLU):
6389 """Sleep for a specified amount of time.
6391 This LU sleeps on the master and/or nodes for a specified amount of
6395 _OP_REQP = ["duration", "on_master", "on_nodes"]
6398 def ExpandNames(self):
6399 """Expand names and set required locks.
6401 This expands the node list, if any.
6404 self.needed_locks = {}
6405 if self.op.on_nodes:
6406 # _GetWantedNodes can be used here, but is not always appropriate to use
6407 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6409 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6410 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6412 def CheckPrereq(self):
6413 """Check prerequisites.
6417 def Exec(self, feedback_fn):
6418 """Do the actual sleep.
6421 if self.op.on_master:
6422 if not utils.TestDelay(self.op.duration):
6423 raise errors.OpExecError("Error during master delay test")
6424 if self.op.on_nodes:
6425 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6427 raise errors.OpExecError("Complete failure from rpc call")
6428 for node, node_result in result.items():
6430 if not node_result.data:
6431 raise errors.OpExecError("Failure during rpc call to node %s,"
6432 " result: %s" % (node, node_result.data))
6435 class IAllocator(object):
6436 """IAllocator framework.
6438 An IAllocator instance has three sets of attributes:
6439 - cfg that is needed to query the cluster
6440 - input data (all members of the _KEYS class attribute are required)
6441 - four buffer attributes (in|out_data|text), that represent the
6442 input (to the external script) in text and data structure format,
6443 and the output from it, again in two formats
6444 - the result variables from the script (success, info, nodes) for
6449 "mem_size", "disks", "disk_template",
6450 "os", "tags", "nics", "vcpus", "hypervisor",
6456 def __init__(self, lu, mode, name, **kwargs):
6458 # init buffer variables
6459 self.in_text = self.out_text = self.in_data = self.out_data = None
6460 # init all input fields so that pylint is happy
6463 self.mem_size = self.disks = self.disk_template = None
6464 self.os = self.tags = self.nics = self.vcpus = None
6465 self.hypervisor = None
6466 self.relocate_from = None
6468 self.required_nodes = None
6469 # init result fields
6470 self.success = self.info = self.nodes = None
6471 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6472 keyset = self._ALLO_KEYS
6473 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6474 keyset = self._RELO_KEYS
6476 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6477 " IAllocator" % self.mode)
6479 if key not in keyset:
6480 raise errors.ProgrammerError("Invalid input parameter '%s' to"
6481 " IAllocator" % key)
6482 setattr(self, key, kwargs[key])
6484 if key not in kwargs:
6485 raise errors.ProgrammerError("Missing input parameter '%s' to"
6486 " IAllocator" % key)
6487 self._BuildInputData()
6489 def _ComputeClusterData(self):
6490 """Compute the generic allocator input data.
6492 This is the data that is independent of the actual operation.
6496 cluster_info = cfg.GetClusterInfo()
6500 "cluster_name": cfg.GetClusterName(),
6501 "cluster_tags": list(cluster_info.GetTags()),
6502 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
6503 # we don't have job IDs
6505 iinfo = cfg.GetAllInstancesInfo().values()
6506 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6510 node_list = cfg.GetNodeList()
6512 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6513 hypervisor_name = self.hypervisor
6514 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6515 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6517 node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6519 node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6520 cluster_info.enabled_hypervisors)
6521 for nname, nresult in node_data.items():
6522 # first fill in static (config-based) values
6523 ninfo = cfg.GetNodeInfo(nname)
6525 "tags": list(ninfo.GetTags()),
6526 "primary_ip": ninfo.primary_ip,
6527 "secondary_ip": ninfo.secondary_ip,
6528 "offline": ninfo.offline,
6529 "drained": ninfo.drained,
6530 "master_candidate": ninfo.master_candidate,
6533 if not ninfo.offline:
6535 if not isinstance(nresult.data, dict):
6536 raise errors.OpExecError("Can't get data for node %s" % nname)
6537 remote_info = nresult.data
6538 for attr in ['memory_total', 'memory_free', 'memory_dom0',
6539 'vg_size', 'vg_free', 'cpu_total']:
6540 if attr not in remote_info:
6541 raise errors.OpExecError("Node '%s' didn't return attribute"
6542 " '%s'" % (nname, attr))
6544 remote_info[attr] = int(remote_info[attr])
6545 except ValueError, err:
6546 raise errors.OpExecError("Node '%s' returned invalid value"
6547 " for '%s': %s" % (nname, attr, err))
6548 # compute memory used by primary instances
6549 i_p_mem = i_p_up_mem = 0
6550 for iinfo, beinfo in i_list:
6551 if iinfo.primary_node == nname:
6552 i_p_mem += beinfo[constants.BE_MEMORY]
6553 if iinfo.name not in node_iinfo[nname].data:
6556 i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
6557 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6558 remote_info['memory_free'] -= max(0, i_mem_diff)
6561 i_p_up_mem += beinfo[constants.BE_MEMORY]
6563 # compute memory used by instances
6565 "total_memory": remote_info['memory_total'],
6566 "reserved_memory": remote_info['memory_dom0'],
6567 "free_memory": remote_info['memory_free'],
6568 "total_disk": remote_info['vg_size'],
6569 "free_disk": remote_info['vg_free'],
6570 "total_cpus": remote_info['cpu_total'],
6571 "i_pri_memory": i_p_mem,
6572 "i_pri_up_memory": i_p_up_mem,
6576 node_results[nname] = pnr
6577 data["nodes"] = node_results
6581 for iinfo, beinfo in i_list:
6582 nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
6583 for n in iinfo.nics]
6585 "tags": list(iinfo.GetTags()),
6586 "admin_up": iinfo.admin_up,
6587 "vcpus": beinfo[constants.BE_VCPUS],
6588 "memory": beinfo[constants.BE_MEMORY],
6590 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
6592 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
6593 "disk_template": iinfo.disk_template,
6594 "hypervisor": iinfo.hypervisor,
6596 instance_data[iinfo.name] = pir
6598 data["instances"] = instance_data
6602 def _AddNewInstance(self):
6603 """Add new instance data to allocator structure.
6605 This in combination with _AllocatorGetClusterData will create the
6606 correct structure needed as input for the allocator.
6608 The checks for the completeness of the opcode must have already been
6614 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6616 if self.disk_template in constants.DTS_NET_MIRROR:
6617 self.required_nodes = 2
6619 self.required_nodes = 1
6623 "disk_template": self.disk_template,
6626 "vcpus": self.vcpus,
6627 "memory": self.mem_size,
6628 "disks": self.disks,
6629 "disk_space_total": disk_space,
6631 "required_nodes": self.required_nodes,
6633 data["request"] = request
6635 def _AddRelocateInstance(self):
6636 """Add relocate instance data to allocator structure.
6638 This in combination with _IAllocatorGetClusterData will create the
6639 correct structure needed as input for the allocator.
6641 The checks for the completeness of the opcode must have already been
6645 instance = self.lu.cfg.GetInstanceInfo(self.name)
6646 if instance is None:
6647 raise errors.ProgrammerError("Unknown instance '%s' passed to"
6648 " IAllocator" % self.name)
6650 if instance.disk_template not in constants.DTS_NET_MIRROR:
6651 raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6653 if len(instance.secondary_nodes) != 1:
6654 raise errors.OpPrereqError("Instance has not exactly one secondary node")
6656 self.required_nodes = 1
6657 disk_sizes = [{'size': disk.size} for disk in instance.disks]
6658 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6663 "disk_space_total": disk_space,
6664 "required_nodes": self.required_nodes,
6665 "relocate_from": self.relocate_from,
6667 self.in_data["request"] = request
6669 def _BuildInputData(self):
6670 """Build input data structures.
6673 self._ComputeClusterData()
6675 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6676 self._AddNewInstance()
6678 self._AddRelocateInstance()
6680 self.in_text = serializer.Dump(self.in_data)
6682 def Run(self, name, validate=True, call_fn=None):
6683 """Run an instance allocator and return the results.
6687 call_fn = self.lu.rpc.call_iallocator_runner
6690 result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6693 if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6694 raise errors.OpExecError("Invalid result from master iallocator runner")
6696 rcode, stdout, stderr, fail = result.data
6698 if rcode == constants.IARUN_NOTFOUND:
6699 raise errors.OpExecError("Can't find allocator '%s'" % name)
6700 elif rcode == constants.IARUN_FAILURE:
6701 raise errors.OpExecError("Instance allocator call failed: %s,"
6702 " output: %s" % (fail, stdout+stderr))
6703 self.out_text = stdout
6705 self._ValidateResult()
6707 def _ValidateResult(self):
6708 """Process the allocator results.
6710 This will process and if successful save the result in
6711 self.out_data and the other parameters.
6715 rdict = serializer.Load(self.out_text)
6716 except Exception, err:
6717 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6719 if not isinstance(rdict, dict):
6720 raise errors.OpExecError("Can't parse iallocator results: not a dict")
6722 for key in "success", "info", "nodes":
6723 if key not in rdict:
6724 raise errors.OpExecError("Can't parse iallocator results:"
6725 " missing key '%s'" % key)
6726 setattr(self, key, rdict[key])
6728 if not isinstance(rdict["nodes"], list):
6729 raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6731 self.out_data = rdict
6734 class LUTestAllocator(NoHooksLU):
6735 """Run allocator tests.
6737 This LU runs the allocator tests
6740 _OP_REQP = ["direction", "mode", "name"]
6742 def CheckPrereq(self):
6743 """Check prerequisites.
6745 This checks the opcode parameters depending on the director and mode test.
6748 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6749 for attr in ["name", "mem_size", "disks", "disk_template",
6750 "os", "tags", "nics", "vcpus"]:
6751 if not hasattr(self.op, attr):
6752 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6754 iname = self.cfg.ExpandInstanceName(self.op.name)
6755 if iname is not None:
6756 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6758 if not isinstance(self.op.nics, list):
6759 raise errors.OpPrereqError("Invalid parameter 'nics'")
6760 for row in self.op.nics:
6761 if (not isinstance(row, dict) or
6764 "bridge" not in row):
6765 raise errors.OpPrereqError("Invalid contents of the"
6766 " 'nics' parameter")
6767 if not isinstance(self.op.disks, list):
6768 raise errors.OpPrereqError("Invalid parameter 'disks'")
6769 for row in self.op.disks:
6770 if (not isinstance(row, dict) or
6771 "size" not in row or
6772 not isinstance(row["size"], int) or
6773 "mode" not in row or
6774 row["mode"] not in ['r', 'w']):
6775 raise errors.OpPrereqError("Invalid contents of the"
6776 " 'disks' parameter")
6777 if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
6778 self.op.hypervisor = self.cfg.GetHypervisorType()
6779 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6780 if not hasattr(self.op, "name"):
6781 raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6782 fname = self.cfg.ExpandInstanceName(self.op.name)
6784 raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6786 self.op.name = fname
6787 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6789 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6792 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6793 if not hasattr(self.op, "allocator") or self.op.allocator is None:
6794 raise errors.OpPrereqError("Missing allocator name")
6795 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6796 raise errors.OpPrereqError("Wrong allocator test '%s'" %
6799 def Exec(self, feedback_fn):
6800 """Run the allocator test.
6803 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6804 ial = IAllocator(self,
6807 mem_size=self.op.mem_size,
6808 disks=self.op.disks,
6809 disk_template=self.op.disk_template,
6813 vcpus=self.op.vcpus,
6814 hypervisor=self.op.hypervisor,
6817 ial = IAllocator(self,
6820 relocate_from=list(self.relocate_from),
6823 if self.op.direction == constants.IALLOCATOR_DIR_IN:
6824 result = ial.in_text
6826 ial.Run(self.op.allocator, validate=False)
6827 result = ial.out_text