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 = lu.cfg.GetInstanceList()
396 return utils.NiceSort(wanted)
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 nodes is offline
440 if lu.cfg.GetNodeInfo(node).offline:
441 raise errors.OpPrereqError("Can't use offline node %s" % node)
444 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
445 memory, vcpus, nics):
446 """Builds instance related env variables for hooks
448 This builds the hook environment from individual variables.
451 @param name: the name of the instance
452 @type primary_node: string
453 @param primary_node: the name of the instance's primary node
454 @type secondary_nodes: list
455 @param secondary_nodes: list of secondary nodes as strings
456 @type os_type: string
457 @param os_type: the name of the instance's OS
459 @param status: the desired status of the instances
461 @param memory: the memory size of the instance
463 @param vcpus: the count of VCPUs the instance has
465 @param nics: list of tuples (ip, bridge, mac) representing
466 the NICs the instance has
468 @return: the hook environment for this instance
473 "INSTANCE_NAME": name,
474 "INSTANCE_PRIMARY": primary_node,
475 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
476 "INSTANCE_OS_TYPE": os_type,
477 "INSTANCE_STATUS": status,
478 "INSTANCE_MEMORY": memory,
479 "INSTANCE_VCPUS": vcpus,
483 nic_count = len(nics)
484 for idx, (ip, bridge, mac) in enumerate(nics):
487 env["INSTANCE_NIC%d_IP" % idx] = ip
488 env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
489 env["INSTANCE_NIC%d_HWADDR" % idx] = mac
493 env["INSTANCE_NIC_COUNT"] = nic_count
498 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
499 """Builds instance related env variables for hooks from an object.
501 @type lu: L{LogicalUnit}
502 @param lu: the logical unit on whose behalf we execute
503 @type instance: L{objects.Instance}
504 @param instance: the instance for which we should build the
507 @param override: dictionary with key/values that will override
510 @return: the hook environment dictionary
513 bep = lu.cfg.GetClusterInfo().FillBE(instance)
515 'name': instance.name,
516 'primary_node': instance.primary_node,
517 'secondary_nodes': instance.secondary_nodes,
518 'os_type': instance.os,
519 'status': instance.os,
520 'memory': bep[constants.BE_MEMORY],
521 'vcpus': bep[constants.BE_VCPUS],
522 'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
525 args.update(override)
526 return _BuildInstanceHookEnv(**args)
529 def _AdjustCandidatePool(lu):
530 """Adjust the candidate pool after node operations.
533 mod_list = lu.cfg.MaintainCandidatePool()
535 lu.LogInfo("Promoted nodes to master candidate role: %s",
536 ", ".join(node.name for node in mod_list))
537 for name in mod_list:
538 lu.context.ReaddNode(name)
539 mc_now, mc_max = lu.cfg.GetMasterCandidateStats()
541 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
545 def _CheckInstanceBridgesExist(lu, instance):
546 """Check that the brigdes needed by an instance exist.
549 # check bridges existance
550 brlist = [nic.bridge for nic in instance.nics]
551 result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
554 raise errors.OpPrereqError("One or more target bridges %s does not"
555 " exist on destination node '%s'" %
556 (brlist, instance.primary_node))
559 class LUDestroyCluster(NoHooksLU):
560 """Logical unit for destroying the cluster.
565 def CheckPrereq(self):
566 """Check prerequisites.
568 This checks whether the cluster is empty.
570 Any errors are signalled by raising errors.OpPrereqError.
573 master = self.cfg.GetMasterNode()
575 nodelist = self.cfg.GetNodeList()
576 if len(nodelist) != 1 or nodelist[0] != master:
577 raise errors.OpPrereqError("There are still %d node(s) in"
578 " this cluster." % (len(nodelist) - 1))
579 instancelist = self.cfg.GetInstanceList()
581 raise errors.OpPrereqError("There are still %d instance(s) in"
582 " this cluster." % len(instancelist))
584 def Exec(self, feedback_fn):
585 """Destroys the cluster.
588 master = self.cfg.GetMasterNode()
589 result = self.rpc.call_node_stop_master(master, False)
592 raise errors.OpExecError("Could not disable the master role")
593 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
594 utils.CreateBackup(priv_key)
595 utils.CreateBackup(pub_key)
599 class LUVerifyCluster(LogicalUnit):
600 """Verifies the cluster status.
603 HPATH = "cluster-verify"
604 HTYPE = constants.HTYPE_CLUSTER
605 _OP_REQP = ["skip_checks"]
608 def ExpandNames(self):
609 self.needed_locks = {
610 locking.LEVEL_NODE: locking.ALL_SET,
611 locking.LEVEL_INSTANCE: locking.ALL_SET,
613 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
615 def _VerifyNode(self, nodeinfo, file_list, local_cksum,
616 node_result, feedback_fn, master_files):
617 """Run multiple tests against a node.
621 - compares ganeti version
622 - checks vg existance and size > 20G
623 - checks config file checksum
624 - checks ssh to other nodes
626 @type nodeinfo: L{objects.Node}
627 @param nodeinfo: the node to check
628 @param file_list: required list of files
629 @param local_cksum: dictionary of local files and their checksums
630 @param node_result: the results from the node
631 @param feedback_fn: function used to accumulate results
632 @param master_files: list of files that only masters should have
637 # main result, node_result should be a non-empty dict
638 if not node_result or not isinstance(node_result, dict):
639 feedback_fn(" - ERROR: unable to verify node %s." % (node,))
642 # compares ganeti version
643 local_version = constants.PROTOCOL_VERSION
644 remote_version = node_result.get('version', None)
645 if not remote_version:
646 feedback_fn(" - ERROR: connection to %s failed" % (node))
649 if local_version != remote_version:
650 feedback_fn(" - ERROR: sw version mismatch: master %s, node(%s) %s" %
651 (local_version, node, remote_version))
654 # checks vg existance and size > 20G
657 vglist = node_result.get(constants.NV_VGLIST, None)
659 feedback_fn(" - ERROR: unable to check volume groups on node %s." %
663 vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
664 constants.MIN_VG_SIZE)
666 feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
669 # checks config file checksum
671 remote_cksum = node_result.get(constants.NV_FILELIST, None)
672 if not isinstance(remote_cksum, dict):
674 feedback_fn(" - ERROR: node hasn't returned file checksum data")
676 for file_name in file_list:
677 node_is_mc = nodeinfo.master_candidate
678 must_have_file = file_name not in master_files
679 if file_name not in remote_cksum:
680 if node_is_mc or must_have_file:
682 feedback_fn(" - ERROR: file '%s' missing" % file_name)
683 elif remote_cksum[file_name] != local_cksum[file_name]:
684 if node_is_mc or must_have_file:
686 feedback_fn(" - ERROR: file '%s' has wrong checksum" % file_name)
688 # not candidate and this is not a must-have file
690 feedback_fn(" - ERROR: non master-candidate has old/wrong file"
693 # all good, except non-master/non-must have combination
694 if not node_is_mc and not must_have_file:
695 feedback_fn(" - ERROR: file '%s' should not exist on non master"
696 " candidates" % file_name)
700 if constants.NV_NODELIST not in node_result:
702 feedback_fn(" - ERROR: node hasn't returned node ssh connectivity data")
704 if node_result[constants.NV_NODELIST]:
706 for node in node_result[constants.NV_NODELIST]:
707 feedback_fn(" - ERROR: ssh communication with node '%s': %s" %
708 (node, node_result[constants.NV_NODELIST][node]))
710 if constants.NV_NODENETTEST not in node_result:
712 feedback_fn(" - ERROR: node hasn't returned node tcp connectivity data")
714 if node_result[constants.NV_NODENETTEST]:
716 nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
718 feedback_fn(" - ERROR: tcp communication with node '%s': %s" %
719 (node, node_result[constants.NV_NODENETTEST][node]))
721 hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
722 if isinstance(hyp_result, dict):
723 for hv_name, hv_result in hyp_result.iteritems():
724 if hv_result is not None:
725 feedback_fn(" - ERROR: hypervisor %s verify failure: '%s'" %
726 (hv_name, hv_result))
729 def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
730 node_instance, feedback_fn, n_offline):
731 """Verify an instance.
733 This function checks to see if the required block devices are
734 available on the instance's node.
739 node_current = instanceconfig.primary_node
742 instanceconfig.MapLVsByNode(node_vol_should)
744 for node in node_vol_should:
745 if node in n_offline:
746 # ignore missing volumes on offline nodes
748 for volume in node_vol_should[node]:
749 if node not in node_vol_is or volume not in node_vol_is[node]:
750 feedback_fn(" - ERROR: volume %s missing on node %s" %
754 if not instanceconfig.status == 'down':
755 if ((node_current not in node_instance or
756 not instance in node_instance[node_current]) and
757 node_current not in n_offline):
758 feedback_fn(" - ERROR: instance %s not running on node %s" %
759 (instance, node_current))
762 for node in node_instance:
763 if (not node == node_current):
764 if instance in node_instance[node]:
765 feedback_fn(" - ERROR: instance %s should not run on node %s" %
771 def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
772 """Verify if there are any unknown volumes in the cluster.
774 The .os, .swap and backup volumes are ignored. All other volumes are
780 for node in node_vol_is:
781 for volume in node_vol_is[node]:
782 if node not in node_vol_should or volume not in node_vol_should[node]:
783 feedback_fn(" - ERROR: volume %s on node %s should not exist" %
788 def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
789 """Verify the list of running instances.
791 This checks what instances are running but unknown to the cluster.
795 for node in node_instance:
796 for runninginstance in node_instance[node]:
797 if runninginstance not in instancelist:
798 feedback_fn(" - ERROR: instance %s on node %s should not exist" %
799 (runninginstance, node))
803 def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
804 """Verify N+1 Memory Resilience.
806 Check that if one single node dies we can still start all the instances it
812 for node, nodeinfo in node_info.iteritems():
813 # This code checks that every node which is now listed as secondary has
814 # enough memory to host all instances it is supposed to should a single
815 # other node in the cluster fail.
816 # FIXME: not ready for failover to an arbitrary node
817 # FIXME: does not support file-backed instances
818 # WARNING: we currently take into account down instances as well as up
819 # ones, considering that even if they're down someone might want to start
820 # them even in the event of a node failure.
821 for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
823 for instance in instances:
824 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
825 if bep[constants.BE_AUTO_BALANCE]:
826 needed_mem += bep[constants.BE_MEMORY]
827 if nodeinfo['mfree'] < needed_mem:
828 feedback_fn(" - ERROR: not enough memory on node %s to accomodate"
829 " failovers should node %s fail" % (node, prinode))
833 def CheckPrereq(self):
834 """Check prerequisites.
836 Transform the list of checks we're going to skip into a set and check that
837 all its members are valid.
840 self.skip_set = frozenset(self.op.skip_checks)
841 if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
842 raise errors.OpPrereqError("Invalid checks to be skipped specified")
844 def BuildHooksEnv(self):
847 Cluster-Verify hooks just rone in the post phase and their failure makes
848 the output be logged in the verify output and the verification to fail.
851 all_nodes = self.cfg.GetNodeList()
852 # TODO: populate the environment with useful information for verify hooks
854 return env, [], all_nodes
856 def Exec(self, feedback_fn):
857 """Verify integrity of cluster, performing various test on nodes.
861 feedback_fn("* Verifying global settings")
862 for msg in self.cfg.VerifyConfig():
863 feedback_fn(" - ERROR: %s" % msg)
865 vg_name = self.cfg.GetVGName()
866 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
867 nodelist = utils.NiceSort(self.cfg.GetNodeList())
868 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
869 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
870 i_non_redundant = [] # Non redundant instances
871 i_non_a_balanced = [] # Non auto-balanced instances
872 n_offline = [] # List of offline nodes
878 # FIXME: verify OS list
880 master_files = [constants.CLUSTER_CONF_FILE]
882 file_names = ssconf.SimpleStore().GetFileList()
883 file_names.append(constants.SSL_CERT_FILE)
884 file_names.append(constants.RAPI_CERT_FILE)
885 file_names.extend(master_files)
887 local_checksums = utils.FingerprintFiles(file_names)
889 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
890 node_verify_param = {
891 constants.NV_FILELIST: file_names,
892 constants.NV_NODELIST: [node.name for node in nodeinfo
893 if not node.offline],
894 constants.NV_HYPERVISOR: hypervisors,
895 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
896 node.secondary_ip) for node in nodeinfo
897 if not node.offline],
898 constants.NV_LVLIST: vg_name,
899 constants.NV_INSTANCELIST: hypervisors,
900 constants.NV_VGLIST: None,
901 constants.NV_VERSION: None,
902 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
904 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
905 self.cfg.GetClusterName())
907 cluster = self.cfg.GetClusterInfo()
908 master_node = self.cfg.GetMasterNode()
909 for node_i in nodeinfo:
911 nresult = all_nvinfo[node].data
914 feedback_fn("* Skipping offline node %s" % (node,))
915 n_offline.append(node)
918 if node == master_node:
920 elif node_i.master_candidate:
921 ntype = "master candidate"
924 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
926 if all_nvinfo[node].failed or not isinstance(nresult, dict):
927 feedback_fn(" - ERROR: connection to %s failed" % (node,))
931 result = self._VerifyNode(node_i, file_names, local_checksums,
932 nresult, feedback_fn, master_files)
935 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
936 if isinstance(lvdata, basestring):
937 feedback_fn(" - ERROR: LVM problem on node %s: %s" %
938 (node, lvdata.encode('string_escape')))
940 node_volume[node] = {}
941 elif not isinstance(lvdata, dict):
942 feedback_fn(" - ERROR: connection to %s failed (lvlist)" % (node,))
946 node_volume[node] = lvdata
949 idata = nresult.get(constants.NV_INSTANCELIST, None)
950 if not isinstance(idata, list):
951 feedback_fn(" - ERROR: connection to %s failed (instancelist)" %
956 node_instance[node] = idata
959 nodeinfo = nresult.get(constants.NV_HVINFO, None)
960 if not isinstance(nodeinfo, dict):
961 feedback_fn(" - ERROR: connection to %s failed (hvinfo)" % (node,))
967 "mfree": int(nodeinfo['memory_free']),
968 "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
971 # dictionary holding all instances this node is secondary for,
972 # grouped by their primary node. Each key is a cluster node, and each
973 # value is a list of instances which have the key as primary and the
974 # current node as secondary. this is handy to calculate N+1 memory
975 # availability if you can only failover from a primary to its
977 "sinst-by-pnode": {},
980 feedback_fn(" - ERROR: invalid value returned from node %s" % (node,))
986 for instance in instancelist:
987 feedback_fn("* Verifying instance %s" % instance)
988 inst_config = self.cfg.GetInstanceInfo(instance)
989 result = self._VerifyInstance(instance, inst_config, node_volume,
990 node_instance, feedback_fn, n_offline)
992 inst_nodes_offline = []
994 inst_config.MapLVsByNode(node_vol_should)
996 instance_cfg[instance] = inst_config
998 pnode = inst_config.primary_node
999 if pnode in node_info:
1000 node_info[pnode]['pinst'].append(instance)
1001 elif pnode not in n_offline:
1002 feedback_fn(" - ERROR: instance %s, connection to primary node"
1003 " %s failed" % (instance, pnode))
1006 if pnode in n_offline:
1007 inst_nodes_offline.append(pnode)
1009 # If the instance is non-redundant we cannot survive losing its primary
1010 # node, so we are not N+1 compliant. On the other hand we have no disk
1011 # templates with more than one secondary so that situation is not well
1013 # FIXME: does not support file-backed instances
1014 if len(inst_config.secondary_nodes) == 0:
1015 i_non_redundant.append(instance)
1016 elif len(inst_config.secondary_nodes) > 1:
1017 feedback_fn(" - WARNING: multiple secondaries for instance %s"
1020 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1021 i_non_a_balanced.append(instance)
1023 for snode in inst_config.secondary_nodes:
1024 if snode in node_info:
1025 node_info[snode]['sinst'].append(instance)
1026 if pnode not in node_info[snode]['sinst-by-pnode']:
1027 node_info[snode]['sinst-by-pnode'][pnode] = []
1028 node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1029 elif snode not in n_offline:
1030 feedback_fn(" - ERROR: instance %s, connection to secondary node"
1031 " %s failed" % (instance, snode))
1033 if snode in n_offline:
1034 inst_nodes_offline.append(snode)
1036 if inst_nodes_offline:
1037 # warn that the instance lives on offline nodes, and set bad=True
1038 feedback_fn(" - ERROR: instance lives on offline node(s) %s" %
1039 ", ".join(inst_nodes_offline))
1042 feedback_fn("* Verifying orphan volumes")
1043 result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
1047 feedback_fn("* Verifying remaining instances")
1048 result = self._VerifyOrphanInstances(instancelist, node_instance,
1052 if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1053 feedback_fn("* Verifying N+1 Memory redundancy")
1054 result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
1057 feedback_fn("* Other Notes")
1059 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
1060 % len(i_non_redundant))
1062 if i_non_a_balanced:
1063 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
1064 % len(i_non_a_balanced))
1067 feedback_fn(" - NOTICE: %d offline node(s) found." % len(n_offline))
1071 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1072 """Analize the post-hooks' result
1074 This method analyses the hook result, handles it, and sends some
1075 nicely-formatted feedback back to the user.
1077 @param phase: one of L{constants.HOOKS_PHASE_POST} or
1078 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1079 @param hooks_results: the results of the multi-node hooks rpc call
1080 @param feedback_fn: function used send feedback back to the caller
1081 @param lu_result: previous Exec result
1082 @return: the new Exec result, based on the previous result
1086 # We only really run POST phase hooks, and are only interested in
1088 if phase == constants.HOOKS_PHASE_POST:
1089 # Used to change hooks' output to proper indentation
1090 indent_re = re.compile('^', re.M)
1091 feedback_fn("* Hooks Results")
1092 if not hooks_results:
1093 feedback_fn(" - ERROR: general communication failure")
1096 for node_name in hooks_results:
1097 show_node_header = True
1098 res = hooks_results[node_name]
1099 if res.failed or res.data is False or not isinstance(res.data, list):
1101 # no need to warn or set fail return value
1103 feedback_fn(" Communication failure in hooks execution")
1106 for script, hkr, output in res.data:
1107 if hkr == constants.HKR_FAIL:
1108 # The node header is only shown once, if there are
1109 # failing hooks on that node
1110 if show_node_header:
1111 feedback_fn(" Node %s:" % node_name)
1112 show_node_header = False
1113 feedback_fn(" ERROR: Script %s failed, output:" % script)
1114 output = indent_re.sub(' ', output)
1115 feedback_fn("%s" % output)
1121 class LUVerifyDisks(NoHooksLU):
1122 """Verifies the cluster disks status.
1128 def ExpandNames(self):
1129 self.needed_locks = {
1130 locking.LEVEL_NODE: locking.ALL_SET,
1131 locking.LEVEL_INSTANCE: locking.ALL_SET,
1133 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1135 def CheckPrereq(self):
1136 """Check prerequisites.
1138 This has no prerequisites.
1143 def Exec(self, feedback_fn):
1144 """Verify integrity of cluster disks.
1147 result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1149 vg_name = self.cfg.GetVGName()
1150 nodes = utils.NiceSort(self.cfg.GetNodeList())
1151 instances = [self.cfg.GetInstanceInfo(name)
1152 for name in self.cfg.GetInstanceList()]
1155 for inst in instances:
1157 if (inst.status != "up" or
1158 inst.disk_template not in constants.DTS_NET_MIRROR):
1160 inst.MapLVsByNode(inst_lvs)
1161 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1162 for node, vol_list in inst_lvs.iteritems():
1163 for vol in vol_list:
1164 nv_dict[(node, vol)] = inst
1169 node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1174 lvs = node_lvs[node]
1177 self.LogWarning("Connection to node %s failed: %s" %
1181 if isinstance(lvs, basestring):
1182 logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1183 res_nlvm[node] = lvs
1184 elif not isinstance(lvs, dict):
1185 logging.warning("Connection to node %s failed or invalid data"
1187 res_nodes.append(node)
1190 for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1191 inst = nv_dict.pop((node, lv_name), None)
1192 if (not lv_online and inst is not None
1193 and inst.name not in res_instances):
1194 res_instances.append(inst.name)
1196 # any leftover items in nv_dict are missing LVs, let's arrange the
1198 for key, inst in nv_dict.iteritems():
1199 if inst.name not in res_missing:
1200 res_missing[inst.name] = []
1201 res_missing[inst.name].append(key)
1206 class LURenameCluster(LogicalUnit):
1207 """Rename the cluster.
1210 HPATH = "cluster-rename"
1211 HTYPE = constants.HTYPE_CLUSTER
1214 def BuildHooksEnv(self):
1219 "OP_TARGET": self.cfg.GetClusterName(),
1220 "NEW_NAME": self.op.name,
1222 mn = self.cfg.GetMasterNode()
1223 return env, [mn], [mn]
1225 def CheckPrereq(self):
1226 """Verify that the passed name is a valid one.
1229 hostname = utils.HostInfo(self.op.name)
1231 new_name = hostname.name
1232 self.ip = new_ip = hostname.ip
1233 old_name = self.cfg.GetClusterName()
1234 old_ip = self.cfg.GetMasterIP()
1235 if new_name == old_name and new_ip == old_ip:
1236 raise errors.OpPrereqError("Neither the name nor the IP address of the"
1237 " cluster has changed")
1238 if new_ip != old_ip:
1239 if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1240 raise errors.OpPrereqError("The given cluster IP address (%s) is"
1241 " reachable on the network. Aborting." %
1244 self.op.name = new_name
1246 def Exec(self, feedback_fn):
1247 """Rename the cluster.
1250 clustername = self.op.name
1253 # shutdown the master IP
1254 master = self.cfg.GetMasterNode()
1255 result = self.rpc.call_node_stop_master(master, False)
1256 if result.failed or not result.data:
1257 raise errors.OpExecError("Could not disable the master role")
1260 cluster = self.cfg.GetClusterInfo()
1261 cluster.cluster_name = clustername
1262 cluster.master_ip = ip
1263 self.cfg.Update(cluster)
1265 # update the known hosts file
1266 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1267 node_list = self.cfg.GetNodeList()
1269 node_list.remove(master)
1272 result = self.rpc.call_upload_file(node_list,
1273 constants.SSH_KNOWN_HOSTS_FILE)
1274 for to_node, to_result in result.iteritems():
1275 if to_result.failed or not to_result.data:
1276 logging.error("Copy of file %s to node %s failed",
1277 constants.SSH_KNOWN_HOSTS_FILE, to_node)
1280 result = self.rpc.call_node_start_master(master, False)
1281 if result.failed or not result.data:
1282 self.LogWarning("Could not re-enable the master role on"
1283 " the master, please restart manually.")
1286 def _RecursiveCheckIfLVMBased(disk):
1287 """Check if the given disk or its children are lvm-based.
1289 @type disk: L{objects.Disk}
1290 @param disk: the disk to check
1292 @return: boolean indicating whether a LD_LV dev_type was found or not
1296 for chdisk in disk.children:
1297 if _RecursiveCheckIfLVMBased(chdisk):
1299 return disk.dev_type == constants.LD_LV
1302 class LUSetClusterParams(LogicalUnit):
1303 """Change the parameters of the cluster.
1306 HPATH = "cluster-modify"
1307 HTYPE = constants.HTYPE_CLUSTER
1311 def CheckParameters(self):
1315 if not hasattr(self.op, "candidate_pool_size"):
1316 self.op.candidate_pool_size = None
1317 if self.op.candidate_pool_size is not None:
1319 self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1320 except ValueError, err:
1321 raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1323 if self.op.candidate_pool_size < 1:
1324 raise errors.OpPrereqError("At least one master candidate needed")
1326 def ExpandNames(self):
1327 # FIXME: in the future maybe other cluster params won't require checking on
1328 # all nodes to be modified.
1329 self.needed_locks = {
1330 locking.LEVEL_NODE: locking.ALL_SET,
1332 self.share_locks[locking.LEVEL_NODE] = 1
1334 def BuildHooksEnv(self):
1339 "OP_TARGET": self.cfg.GetClusterName(),
1340 "NEW_VG_NAME": self.op.vg_name,
1342 mn = self.cfg.GetMasterNode()
1343 return env, [mn], [mn]
1345 def CheckPrereq(self):
1346 """Check prerequisites.
1348 This checks whether the given params don't conflict and
1349 if the given volume group is valid.
1352 # FIXME: This only works because there is only one parameter that can be
1353 # changed or removed.
1354 if self.op.vg_name is not None and not self.op.vg_name:
1355 instances = self.cfg.GetAllInstancesInfo().values()
1356 for inst in instances:
1357 for disk in inst.disks:
1358 if _RecursiveCheckIfLVMBased(disk):
1359 raise errors.OpPrereqError("Cannot disable lvm storage while"
1360 " lvm-based instances exist")
1362 node_list = self.acquired_locks[locking.LEVEL_NODE]
1364 # if vg_name not None, checks given volume group on all nodes
1366 vglist = self.rpc.call_vg_list(node_list)
1367 for node in node_list:
1368 if vglist[node].failed:
1369 # ignoring down node
1370 self.LogWarning("Node %s unreachable/error, ignoring" % node)
1372 vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
1374 constants.MIN_VG_SIZE)
1376 raise errors.OpPrereqError("Error on node '%s': %s" %
1379 self.cluster = cluster = self.cfg.GetClusterInfo()
1380 # validate beparams changes
1381 if self.op.beparams:
1382 utils.CheckBEParams(self.op.beparams)
1383 self.new_beparams = cluster.FillDict(
1384 cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1386 # hypervisor list/parameters
1387 self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1388 if self.op.hvparams:
1389 if not isinstance(self.op.hvparams, dict):
1390 raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1391 for hv_name, hv_dict in self.op.hvparams.items():
1392 if hv_name not in self.new_hvparams:
1393 self.new_hvparams[hv_name] = hv_dict
1395 self.new_hvparams[hv_name].update(hv_dict)
1397 if self.op.enabled_hypervisors is not None:
1398 self.hv_list = self.op.enabled_hypervisors
1400 self.hv_list = cluster.enabled_hypervisors
1402 if self.op.hvparams or self.op.enabled_hypervisors is not None:
1403 # either the enabled list has changed, or the parameters have, validate
1404 for hv_name, hv_params in self.new_hvparams.items():
1405 if ((self.op.hvparams and hv_name in self.op.hvparams) or
1406 (self.op.enabled_hypervisors and
1407 hv_name in self.op.enabled_hypervisors)):
1408 # either this is a new hypervisor, or its parameters have changed
1409 hv_class = hypervisor.GetHypervisor(hv_name)
1410 hv_class.CheckParameterSyntax(hv_params)
1411 _CheckHVParams(self, node_list, hv_name, hv_params)
1413 def Exec(self, feedback_fn):
1414 """Change the parameters of the cluster.
1417 if self.op.vg_name is not None:
1418 if self.op.vg_name != self.cfg.GetVGName():
1419 self.cfg.SetVGName(self.op.vg_name)
1421 feedback_fn("Cluster LVM configuration already in desired"
1422 " state, not changing")
1423 if self.op.hvparams:
1424 self.cluster.hvparams = self.new_hvparams
1425 if self.op.enabled_hypervisors is not None:
1426 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1427 if self.op.beparams:
1428 self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1429 if self.op.candidate_pool_size is not None:
1430 self.cluster.candidate_pool_size = self.op.candidate_pool_size
1432 self.cfg.Update(self.cluster)
1434 # we want to update nodes after the cluster so that if any errors
1435 # happen, we have recorded and saved the cluster info
1436 if self.op.candidate_pool_size is not None:
1437 _AdjustCandidatePool(self)
1440 class LURedistributeConfig(NoHooksLU):
1441 """Force the redistribution of cluster configuration.
1443 This is a very simple LU.
1449 def ExpandNames(self):
1450 self.needed_locks = {
1451 locking.LEVEL_NODE: locking.ALL_SET,
1453 self.share_locks[locking.LEVEL_NODE] = 1
1455 def CheckPrereq(self):
1456 """Check prerequisites.
1460 def Exec(self, feedback_fn):
1461 """Redistribute the configuration.
1464 self.cfg.Update(self.cfg.GetClusterInfo())
1467 def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1468 """Sleep and poll for an instance's disk to sync.
1471 if not instance.disks:
1475 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1477 node = instance.primary_node
1479 for dev in instance.disks:
1480 lu.cfg.SetDiskID(dev, node)
1486 cumul_degraded = False
1487 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1488 if rstats.failed or not rstats.data:
1489 lu.LogWarning("Can't get any data from node %s", node)
1492 raise errors.RemoteError("Can't contact node %s for mirror data,"
1493 " aborting." % node)
1496 rstats = rstats.data
1498 for i in range(len(rstats)):
1501 lu.LogWarning("Can't compute data for node %s/%s",
1502 node, instance.disks[i].iv_name)
1504 # we ignore the ldisk parameter
1505 perc_done, est_time, is_degraded, _ = mstat
1506 cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1507 if perc_done is not None:
1509 if est_time is not None:
1510 rem_time = "%d estimated seconds remaining" % est_time
1513 rem_time = "no time estimate"
1514 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1515 (instance.disks[i].iv_name, perc_done, rem_time))
1519 time.sleep(min(60, max_time))
1522 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1523 return not cumul_degraded
1526 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1527 """Check that mirrors are not degraded.
1529 The ldisk parameter, if True, will change the test from the
1530 is_degraded attribute (which represents overall non-ok status for
1531 the device(s)) to the ldisk (representing the local storage status).
1534 lu.cfg.SetDiskID(dev, node)
1541 if on_primary or dev.AssembleOnSecondary():
1542 rstats = lu.rpc.call_blockdev_find(node, dev)
1543 if rstats.failed or not rstats.data:
1544 logging.warning("Node %s: disk degraded, not found or node down", node)
1547 result = result and (not rstats.data[idx])
1549 for child in dev.children:
1550 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1555 class LUDiagnoseOS(NoHooksLU):
1556 """Logical unit for OS diagnose/query.
1559 _OP_REQP = ["output_fields", "names"]
1561 _FIELDS_STATIC = utils.FieldSet()
1562 _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1564 def ExpandNames(self):
1566 raise errors.OpPrereqError("Selective OS query not supported")
1568 _CheckOutputFields(static=self._FIELDS_STATIC,
1569 dynamic=self._FIELDS_DYNAMIC,
1570 selected=self.op.output_fields)
1572 # Lock all nodes, in shared mode
1573 self.needed_locks = {}
1574 self.share_locks[locking.LEVEL_NODE] = 1
1575 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1577 def CheckPrereq(self):
1578 """Check prerequisites.
1583 def _DiagnoseByOS(node_list, rlist):
1584 """Remaps a per-node return list into an a per-os per-node dictionary
1586 @param node_list: a list with the names of all nodes
1587 @param rlist: a map with node names as keys and OS objects as values
1590 @returns: a dictionary with osnames as keys and as value another map, with
1591 nodes as keys and list of OS objects as values, eg::
1593 {"debian-etch": {"node1": [<object>,...],
1594 "node2": [<object>,]}
1599 for node_name, nr in rlist.iteritems():
1600 if nr.failed or not nr.data:
1602 for os_obj in nr.data:
1603 if os_obj.name not in all_os:
1604 # build a list of nodes for this os containing empty lists
1605 # for each node in node_list
1606 all_os[os_obj.name] = {}
1607 for nname in node_list:
1608 all_os[os_obj.name][nname] = []
1609 all_os[os_obj.name][node_name].append(os_obj)
1612 def Exec(self, feedback_fn):
1613 """Compute the list of OSes.
1616 node_list = self.acquired_locks[locking.LEVEL_NODE]
1617 valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
1618 if node in node_list]
1619 node_data = self.rpc.call_os_diagnose(valid_nodes)
1620 if node_data == False:
1621 raise errors.OpExecError("Can't gather the list of OSes")
1622 pol = self._DiagnoseByOS(valid_nodes, node_data)
1624 for os_name, os_data in pol.iteritems():
1626 for field in self.op.output_fields:
1629 elif field == "valid":
1630 val = utils.all([osl and osl[0] for osl in os_data.values()])
1631 elif field == "node_status":
1633 for node_name, nos_list in os_data.iteritems():
1634 val[node_name] = [(v.status, v.path) for v in nos_list]
1636 raise errors.ParameterError(field)
1643 class LURemoveNode(LogicalUnit):
1644 """Logical unit for removing a node.
1647 HPATH = "node-remove"
1648 HTYPE = constants.HTYPE_NODE
1649 _OP_REQP = ["node_name"]
1651 def BuildHooksEnv(self):
1654 This doesn't run on the target node in the pre phase as a failed
1655 node would then be impossible to remove.
1659 "OP_TARGET": self.op.node_name,
1660 "NODE_NAME": self.op.node_name,
1662 all_nodes = self.cfg.GetNodeList()
1663 all_nodes.remove(self.op.node_name)
1664 return env, all_nodes, all_nodes
1666 def CheckPrereq(self):
1667 """Check prerequisites.
1670 - the node exists in the configuration
1671 - it does not have primary or secondary instances
1672 - it's not the master
1674 Any errors are signalled by raising errors.OpPrereqError.
1677 node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1679 raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1681 instance_list = self.cfg.GetInstanceList()
1683 masternode = self.cfg.GetMasterNode()
1684 if node.name == masternode:
1685 raise errors.OpPrereqError("Node is the master node,"
1686 " you need to failover first.")
1688 for instance_name in instance_list:
1689 instance = self.cfg.GetInstanceInfo(instance_name)
1690 if node.name in instance.all_nodes:
1691 raise errors.OpPrereqError("Instance %s is still running on the node,"
1692 " please remove first." % instance_name)
1693 self.op.node_name = node.name
1696 def Exec(self, feedback_fn):
1697 """Removes the node from the cluster.
1701 logging.info("Stopping the node daemon and removing configs from node %s",
1704 self.context.RemoveNode(node.name)
1706 self.rpc.call_node_leave_cluster(node.name)
1708 # Promote nodes to master candidate as needed
1709 _AdjustCandidatePool(self)
1712 class LUQueryNodes(NoHooksLU):
1713 """Logical unit for querying nodes.
1716 _OP_REQP = ["output_fields", "names"]
1718 _FIELDS_DYNAMIC = utils.FieldSet(
1720 "mtotal", "mnode", "mfree",
1725 _FIELDS_STATIC = utils.FieldSet(
1726 "name", "pinst_cnt", "sinst_cnt",
1727 "pinst_list", "sinst_list",
1728 "pip", "sip", "tags",
1735 def ExpandNames(self):
1736 _CheckOutputFields(static=self._FIELDS_STATIC,
1737 dynamic=self._FIELDS_DYNAMIC,
1738 selected=self.op.output_fields)
1740 self.needed_locks = {}
1741 self.share_locks[locking.LEVEL_NODE] = 1
1744 self.wanted = _GetWantedNodes(self, self.op.names)
1746 self.wanted = locking.ALL_SET
1748 self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1750 # if we don't request only static fields, we need to lock the nodes
1751 self.needed_locks[locking.LEVEL_NODE] = self.wanted
1754 def CheckPrereq(self):
1755 """Check prerequisites.
1758 # The validation of the node list is done in the _GetWantedNodes,
1759 # if non empty, and if empty, there's no validation to do
1762 def Exec(self, feedback_fn):
1763 """Computes the list of nodes and their attributes.
1766 all_info = self.cfg.GetAllNodesInfo()
1768 nodenames = self.acquired_locks[locking.LEVEL_NODE]
1769 elif self.wanted != locking.ALL_SET:
1770 nodenames = self.wanted
1771 missing = set(nodenames).difference(all_info.keys())
1773 raise errors.OpExecError(
1774 "Some nodes were removed before retrieving their data: %s" % missing)
1776 nodenames = all_info.keys()
1778 nodenames = utils.NiceSort(nodenames)
1779 nodelist = [all_info[name] for name in nodenames]
1781 # begin data gathering
1785 node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1786 self.cfg.GetHypervisorType())
1787 for name in nodenames:
1788 nodeinfo = node_data[name]
1789 if not nodeinfo.failed and nodeinfo.data:
1790 nodeinfo = nodeinfo.data
1791 fn = utils.TryConvert
1793 "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1794 "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1795 "mfree": fn(int, nodeinfo.get('memory_free', None)),
1796 "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1797 "dfree": fn(int, nodeinfo.get('vg_free', None)),
1798 "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1799 "bootid": nodeinfo.get('bootid', None),
1802 live_data[name] = {}
1804 live_data = dict.fromkeys(nodenames, {})
1806 node_to_primary = dict([(name, set()) for name in nodenames])
1807 node_to_secondary = dict([(name, set()) for name in nodenames])
1809 inst_fields = frozenset(("pinst_cnt", "pinst_list",
1810 "sinst_cnt", "sinst_list"))
1811 if inst_fields & frozenset(self.op.output_fields):
1812 instancelist = self.cfg.GetInstanceList()
1814 for instance_name in instancelist:
1815 inst = self.cfg.GetInstanceInfo(instance_name)
1816 if inst.primary_node in node_to_primary:
1817 node_to_primary[inst.primary_node].add(inst.name)
1818 for secnode in inst.secondary_nodes:
1819 if secnode in node_to_secondary:
1820 node_to_secondary[secnode].add(inst.name)
1822 master_node = self.cfg.GetMasterNode()
1824 # end data gathering
1827 for node in nodelist:
1829 for field in self.op.output_fields:
1832 elif field == "pinst_list":
1833 val = list(node_to_primary[node.name])
1834 elif field == "sinst_list":
1835 val = list(node_to_secondary[node.name])
1836 elif field == "pinst_cnt":
1837 val = len(node_to_primary[node.name])
1838 elif field == "sinst_cnt":
1839 val = len(node_to_secondary[node.name])
1840 elif field == "pip":
1841 val = node.primary_ip
1842 elif field == "sip":
1843 val = node.secondary_ip
1844 elif field == "tags":
1845 val = list(node.GetTags())
1846 elif field == "serial_no":
1847 val = node.serial_no
1848 elif field == "master_candidate":
1849 val = node.master_candidate
1850 elif field == "master":
1851 val = node.name == master_node
1852 elif field == "offline":
1854 elif self._FIELDS_DYNAMIC.Matches(field):
1855 val = live_data[node.name].get(field, None)
1857 raise errors.ParameterError(field)
1858 node_output.append(val)
1859 output.append(node_output)
1864 class LUQueryNodeVolumes(NoHooksLU):
1865 """Logical unit for getting volumes on node(s).
1868 _OP_REQP = ["nodes", "output_fields"]
1870 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1871 _FIELDS_STATIC = utils.FieldSet("node")
1873 def ExpandNames(self):
1874 _CheckOutputFields(static=self._FIELDS_STATIC,
1875 dynamic=self._FIELDS_DYNAMIC,
1876 selected=self.op.output_fields)
1878 self.needed_locks = {}
1879 self.share_locks[locking.LEVEL_NODE] = 1
1880 if not self.op.nodes:
1881 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1883 self.needed_locks[locking.LEVEL_NODE] = \
1884 _GetWantedNodes(self, self.op.nodes)
1886 def CheckPrereq(self):
1887 """Check prerequisites.
1889 This checks that the fields required are valid output fields.
1892 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1894 def Exec(self, feedback_fn):
1895 """Computes the list of nodes and their attributes.
1898 nodenames = self.nodes
1899 volumes = self.rpc.call_node_volumes(nodenames)
1901 ilist = [self.cfg.GetInstanceInfo(iname) for iname
1902 in self.cfg.GetInstanceList()]
1904 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1907 for node in nodenames:
1908 if node not in volumes or volumes[node].failed or not volumes[node].data:
1911 node_vols = volumes[node].data[:]
1912 node_vols.sort(key=lambda vol: vol['dev'])
1914 for vol in node_vols:
1916 for field in self.op.output_fields:
1919 elif field == "phys":
1923 elif field == "name":
1925 elif field == "size":
1926 val = int(float(vol['size']))
1927 elif field == "instance":
1929 if node not in lv_by_node[inst]:
1931 if vol['name'] in lv_by_node[inst][node]:
1937 raise errors.ParameterError(field)
1938 node_output.append(str(val))
1940 output.append(node_output)
1945 class LUAddNode(LogicalUnit):
1946 """Logical unit for adding node to the cluster.
1950 HTYPE = constants.HTYPE_NODE
1951 _OP_REQP = ["node_name"]
1953 def BuildHooksEnv(self):
1956 This will run on all nodes before, and on all nodes + the new node after.
1960 "OP_TARGET": self.op.node_name,
1961 "NODE_NAME": self.op.node_name,
1962 "NODE_PIP": self.op.primary_ip,
1963 "NODE_SIP": self.op.secondary_ip,
1965 nodes_0 = self.cfg.GetNodeList()
1966 nodes_1 = nodes_0 + [self.op.node_name, ]
1967 return env, nodes_0, nodes_1
1969 def CheckPrereq(self):
1970 """Check prerequisites.
1973 - the new node is not already in the config
1975 - its parameters (single/dual homed) matches the cluster
1977 Any errors are signalled by raising errors.OpPrereqError.
1980 node_name = self.op.node_name
1983 dns_data = utils.HostInfo(node_name)
1985 node = dns_data.name
1986 primary_ip = self.op.primary_ip = dns_data.ip
1987 secondary_ip = getattr(self.op, "secondary_ip", None)
1988 if secondary_ip is None:
1989 secondary_ip = primary_ip
1990 if not utils.IsValidIP(secondary_ip):
1991 raise errors.OpPrereqError("Invalid secondary IP given")
1992 self.op.secondary_ip = secondary_ip
1994 node_list = cfg.GetNodeList()
1995 if not self.op.readd and node in node_list:
1996 raise errors.OpPrereqError("Node %s is already in the configuration" %
1998 elif self.op.readd and node not in node_list:
1999 raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2001 for existing_node_name in node_list:
2002 existing_node = cfg.GetNodeInfo(existing_node_name)
2004 if self.op.readd and node == existing_node_name:
2005 if (existing_node.primary_ip != primary_ip or
2006 existing_node.secondary_ip != secondary_ip):
2007 raise errors.OpPrereqError("Readded node doesn't have the same IP"
2008 " address configuration as before")
2011 if (existing_node.primary_ip == primary_ip or
2012 existing_node.secondary_ip == primary_ip or
2013 existing_node.primary_ip == secondary_ip or
2014 existing_node.secondary_ip == secondary_ip):
2015 raise errors.OpPrereqError("New node ip address(es) conflict with"
2016 " existing node %s" % existing_node.name)
2018 # check that the type of the node (single versus dual homed) is the
2019 # same as for the master
2020 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2021 master_singlehomed = myself.secondary_ip == myself.primary_ip
2022 newbie_singlehomed = secondary_ip == primary_ip
2023 if master_singlehomed != newbie_singlehomed:
2024 if master_singlehomed:
2025 raise errors.OpPrereqError("The master has no private ip but the"
2026 " new node has one")
2028 raise errors.OpPrereqError("The master has a private ip but the"
2029 " new node doesn't have one")
2031 # checks reachablity
2032 if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2033 raise errors.OpPrereqError("Node not reachable by ping")
2035 if not newbie_singlehomed:
2036 # check reachability from my secondary ip to newbie's secondary ip
2037 if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2038 source=myself.secondary_ip):
2039 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2040 " based ping to noded port")
2042 cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2043 mc_now, _ = self.cfg.GetMasterCandidateStats()
2044 master_candidate = mc_now < cp_size
2046 self.new_node = objects.Node(name=node,
2047 primary_ip=primary_ip,
2048 secondary_ip=secondary_ip,
2049 master_candidate=master_candidate,
2052 def Exec(self, feedback_fn):
2053 """Adds the new node to the cluster.
2056 new_node = self.new_node
2057 node = new_node.name
2059 # check connectivity
2060 result = self.rpc.call_version([node])[node]
2063 if constants.PROTOCOL_VERSION == result.data:
2064 logging.info("Communication to node %s fine, sw version %s match",
2067 raise errors.OpExecError("Version mismatch master version %s,"
2068 " node version %s" %
2069 (constants.PROTOCOL_VERSION, result.data))
2071 raise errors.OpExecError("Cannot get version from the new node")
2074 logging.info("Copy ssh key to node %s", node)
2075 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2077 keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2078 constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2084 keyarray.append(f.read())
2088 result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2090 keyarray[3], keyarray[4], keyarray[5])
2092 if result.failed or not result.data:
2093 raise errors.OpExecError("Cannot transfer ssh keys to the new node")
2095 # Add node to our /etc/hosts, and add key to known_hosts
2096 utils.AddHostToEtcHosts(new_node.name)
2098 if new_node.secondary_ip != new_node.primary_ip:
2099 result = self.rpc.call_node_has_ip_address(new_node.name,
2100 new_node.secondary_ip)
2101 if result.failed or not result.data:
2102 raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2103 " you gave (%s). Please fix and re-run this"
2104 " command." % new_node.secondary_ip)
2106 node_verify_list = [self.cfg.GetMasterNode()]
2107 node_verify_param = {
2109 # TODO: do a node-net-test as well?
2112 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2113 self.cfg.GetClusterName())
2114 for verifier in node_verify_list:
2115 if result[verifier].failed or not result[verifier].data:
2116 raise errors.OpExecError("Cannot communicate with %s's node daemon"
2117 " for remote verification" % verifier)
2118 if result[verifier].data['nodelist']:
2119 for failed in result[verifier].data['nodelist']:
2120 feedback_fn("ssh/hostname verification failed %s -> %s" %
2121 (verifier, result[verifier]['nodelist'][failed]))
2122 raise errors.OpExecError("ssh/hostname verification failed.")
2124 # Distribute updated /etc/hosts and known_hosts to all nodes,
2125 # including the node just added
2126 myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2127 dist_nodes = self.cfg.GetNodeList()
2128 if not self.op.readd:
2129 dist_nodes.append(node)
2130 if myself.name in dist_nodes:
2131 dist_nodes.remove(myself.name)
2133 logging.debug("Copying hosts and known_hosts to all nodes")
2134 for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2135 result = self.rpc.call_upload_file(dist_nodes, fname)
2136 for to_node, to_result in result.iteritems():
2137 if to_result.failed or not to_result.data:
2138 logging.error("Copy of file %s to node %s failed", fname, to_node)
2141 if constants.HT_XEN_HVM in self.cfg.GetClusterInfo().enabled_hypervisors:
2142 to_copy.append(constants.VNC_PASSWORD_FILE)
2143 for fname in to_copy:
2144 result = self.rpc.call_upload_file([node], fname)
2145 if result[node].failed or not result[node]:
2146 logging.error("Could not copy file %s to node %s", fname, node)
2149 self.context.ReaddNode(new_node)
2151 self.context.AddNode(new_node)
2154 class LUSetNodeParams(LogicalUnit):
2155 """Modifies the parameters of a node.
2158 HPATH = "node-modify"
2159 HTYPE = constants.HTYPE_NODE
2160 _OP_REQP = ["node_name"]
2163 def CheckArguments(self):
2164 node_name = self.cfg.ExpandNodeName(self.op.node_name)
2165 if node_name is None:
2166 raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2167 self.op.node_name = node_name
2168 _CheckBooleanOpField(self.op, 'master_candidate')
2169 _CheckBooleanOpField(self.op, 'offline')
2170 if self.op.master_candidate is None and self.op.offline is None:
2171 raise errors.OpPrereqError("Please pass at least one modification")
2172 if self.op.offline == True and self.op.master_candidate == True:
2173 raise errors.OpPrereqError("Can't set the node into offline and"
2174 " master_candidate at the same time")
2176 def ExpandNames(self):
2177 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2179 def BuildHooksEnv(self):
2182 This runs on the master node.
2186 "OP_TARGET": self.op.node_name,
2187 "MASTER_CANDIDATE": str(self.op.master_candidate),
2188 "OFFLINE": str(self.op.offline),
2190 nl = [self.cfg.GetMasterNode(),
2194 def CheckPrereq(self):
2195 """Check prerequisites.
2197 This only checks the instance list against the existing names.
2200 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2202 if ((self.op.master_candidate == False or self.op.offline == True)
2203 and node.master_candidate):
2204 # we will demote the node from master_candidate
2205 if self.op.node_name == self.cfg.GetMasterNode():
2206 raise errors.OpPrereqError("The master node has to be a"
2207 " master candidate and online")
2208 cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2209 num_candidates, _ = self.cfg.GetMasterCandidateStats()
2210 if num_candidates <= cp_size:
2211 msg = ("Not enough master candidates (desired"
2212 " %d, new value will be %d)" % (cp_size, num_candidates-1))
2214 self.LogWarning(msg)
2216 raise errors.OpPrereqError(msg)
2218 if (self.op.master_candidate == True and node.offline and
2219 not self.op.offline == False):
2220 raise errors.OpPrereqError("Can't set an offline node to"
2221 " master_candidate")
2225 def Exec(self, feedback_fn):
2233 if self.op.offline is not None:
2234 node.offline = self.op.offline
2235 result.append(("offline", str(self.op.offline)))
2236 if self.op.offline == True and node.master_candidate:
2237 node.master_candidate = False
2238 result.append(("master_candidate", "auto-demotion due to offline"))
2240 if self.op.master_candidate is not None:
2241 node.master_candidate = self.op.master_candidate
2242 result.append(("master_candidate", str(self.op.master_candidate)))
2243 if self.op.master_candidate == False:
2244 rrc = self.rpc.call_node_demote_from_mc(node.name)
2245 if (rrc.failed or not isinstance(rrc.data, (tuple, list))
2246 or len(rrc.data) != 2):
2247 self.LogWarning("Node rpc error: %s" % rrc.error)
2248 elif not rrc.data[0]:
2249 self.LogWarning("Node failed to demote itself: %s" % rrc.data[1])
2251 # this will trigger configuration file update, if needed
2252 self.cfg.Update(node)
2253 # this will trigger job queue propagation or cleanup
2254 if self.op.node_name != self.cfg.GetMasterNode():
2255 self.context.ReaddNode(node)
2260 class LUQueryClusterInfo(NoHooksLU):
2261 """Query cluster configuration.
2267 def ExpandNames(self):
2268 self.needed_locks = {}
2270 def CheckPrereq(self):
2271 """No prerequsites needed for this LU.
2276 def Exec(self, feedback_fn):
2277 """Return cluster config.
2280 cluster = self.cfg.GetClusterInfo()
2282 "software_version": constants.RELEASE_VERSION,
2283 "protocol_version": constants.PROTOCOL_VERSION,
2284 "config_version": constants.CONFIG_VERSION,
2285 "os_api_version": constants.OS_API_VERSION,
2286 "export_version": constants.EXPORT_VERSION,
2287 "architecture": (platform.architecture()[0], platform.machine()),
2288 "name": cluster.cluster_name,
2289 "master": cluster.master_node,
2290 "default_hypervisor": cluster.default_hypervisor,
2291 "enabled_hypervisors": cluster.enabled_hypervisors,
2292 "hvparams": cluster.hvparams,
2293 "beparams": cluster.beparams,
2294 "candidate_pool_size": cluster.candidate_pool_size,
2300 class LUQueryConfigValues(NoHooksLU):
2301 """Return configuration values.
2306 _FIELDS_DYNAMIC = utils.FieldSet()
2307 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2309 def ExpandNames(self):
2310 self.needed_locks = {}
2312 _CheckOutputFields(static=self._FIELDS_STATIC,
2313 dynamic=self._FIELDS_DYNAMIC,
2314 selected=self.op.output_fields)
2316 def CheckPrereq(self):
2317 """No prerequisites.
2322 def Exec(self, feedback_fn):
2323 """Dump a representation of the cluster config to the standard output.
2327 for field in self.op.output_fields:
2328 if field == "cluster_name":
2329 entry = self.cfg.GetClusterName()
2330 elif field == "master_node":
2331 entry = self.cfg.GetMasterNode()
2332 elif field == "drain_flag":
2333 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2335 raise errors.ParameterError(field)
2336 values.append(entry)
2340 class LUActivateInstanceDisks(NoHooksLU):
2341 """Bring up an instance's disks.
2344 _OP_REQP = ["instance_name"]
2347 def ExpandNames(self):
2348 self._ExpandAndLockInstance()
2349 self.needed_locks[locking.LEVEL_NODE] = []
2350 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2352 def DeclareLocks(self, level):
2353 if level == locking.LEVEL_NODE:
2354 self._LockInstancesNodes()
2356 def CheckPrereq(self):
2357 """Check prerequisites.
2359 This checks that the instance is in the cluster.
2362 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2363 assert self.instance is not None, \
2364 "Cannot retrieve locked instance %s" % self.op.instance_name
2365 _CheckNodeOnline(self, self.instance.primary_node)
2367 def Exec(self, feedback_fn):
2368 """Activate the disks.
2371 disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2373 raise errors.OpExecError("Cannot activate block devices")
2378 def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2379 """Prepare the block devices for an instance.
2381 This sets up the block devices on all nodes.
2383 @type lu: L{LogicalUnit}
2384 @param lu: the logical unit on whose behalf we execute
2385 @type instance: L{objects.Instance}
2386 @param instance: the instance for whose disks we assemble
2387 @type ignore_secondaries: boolean
2388 @param ignore_secondaries: if true, errors on secondary nodes
2389 won't result in an error return from the function
2390 @return: False if the operation failed, otherwise a list of
2391 (host, instance_visible_name, node_visible_name)
2392 with the mapping from node devices to instance devices
2397 iname = instance.name
2398 # With the two passes mechanism we try to reduce the window of
2399 # opportunity for the race condition of switching DRBD to primary
2400 # before handshaking occured, but we do not eliminate it
2402 # The proper fix would be to wait (with some limits) until the
2403 # connection has been made and drbd transitions from WFConnection
2404 # into any other network-connected state (Connected, SyncTarget,
2407 # 1st pass, assemble on all nodes in secondary mode
2408 for inst_disk in instance.disks:
2409 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2410 lu.cfg.SetDiskID(node_disk, node)
2411 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2412 if result.failed or not result:
2413 lu.proc.LogWarning("Could not prepare block device %s on node %s"
2414 " (is_primary=False, pass=1)",
2415 inst_disk.iv_name, node)
2416 if not ignore_secondaries:
2419 # FIXME: race condition on drbd migration to primary
2421 # 2nd pass, do only the primary node
2422 for inst_disk in instance.disks:
2423 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2424 if node != instance.primary_node:
2426 lu.cfg.SetDiskID(node_disk, node)
2427 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2428 if result.failed or not result:
2429 lu.proc.LogWarning("Could not prepare block device %s on node %s"
2430 " (is_primary=True, pass=2)",
2431 inst_disk.iv_name, node)
2433 device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
2435 # leave the disks configured for the primary node
2436 # this is a workaround that would be fixed better by
2437 # improving the logical/physical id handling
2438 for disk in instance.disks:
2439 lu.cfg.SetDiskID(disk, instance.primary_node)
2441 return disks_ok, device_info
2444 def _StartInstanceDisks(lu, instance, force):
2445 """Start the disks of an instance.
2448 disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2449 ignore_secondaries=force)
2451 _ShutdownInstanceDisks(lu, instance)
2452 if force is not None and not force:
2453 lu.proc.LogWarning("", hint="If the message above refers to a"
2455 " you can retry the operation using '--force'.")
2456 raise errors.OpExecError("Disk consistency error")
2459 class LUDeactivateInstanceDisks(NoHooksLU):
2460 """Shutdown an instance's disks.
2463 _OP_REQP = ["instance_name"]
2466 def ExpandNames(self):
2467 self._ExpandAndLockInstance()
2468 self.needed_locks[locking.LEVEL_NODE] = []
2469 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2471 def DeclareLocks(self, level):
2472 if level == locking.LEVEL_NODE:
2473 self._LockInstancesNodes()
2475 def CheckPrereq(self):
2476 """Check prerequisites.
2478 This checks that the instance is in the cluster.
2481 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2482 assert self.instance is not None, \
2483 "Cannot retrieve locked instance %s" % self.op.instance_name
2485 def Exec(self, feedback_fn):
2486 """Deactivate the disks
2489 instance = self.instance
2490 _SafeShutdownInstanceDisks(self, instance)
2493 def _SafeShutdownInstanceDisks(lu, instance):
2494 """Shutdown block devices of an instance.
2496 This function checks if an instance is running, before calling
2497 _ShutdownInstanceDisks.
2500 ins_l = lu.rpc.call_instance_list([instance.primary_node],
2501 [instance.hypervisor])
2502 ins_l = ins_l[instance.primary_node]
2503 if ins_l.failed or not isinstance(ins_l.data, list):
2504 raise errors.OpExecError("Can't contact node '%s'" %
2505 instance.primary_node)
2507 if instance.name in ins_l.data:
2508 raise errors.OpExecError("Instance is running, can't shutdown"
2511 _ShutdownInstanceDisks(lu, instance)
2514 def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2515 """Shutdown block devices of an instance.
2517 This does the shutdown on all nodes of the instance.
2519 If the ignore_primary is false, errors on the primary node are
2524 for disk in instance.disks:
2525 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2526 lu.cfg.SetDiskID(top_disk, node)
2527 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2528 if result.failed or not result.data:
2529 logging.error("Could not shutdown block device %s on node %s",
2531 if not ignore_primary or node != instance.primary_node:
2536 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2537 """Checks if a node has enough free memory.
2539 This function check if a given node has the needed amount of free
2540 memory. In case the node has less memory or we cannot get the
2541 information from the node, this function raise an OpPrereqError
2544 @type lu: C{LogicalUnit}
2545 @param lu: a logical unit from which we get configuration data
2547 @param node: the node to check
2548 @type reason: C{str}
2549 @param reason: string to use in the error message
2550 @type requested: C{int}
2551 @param requested: the amount of memory in MiB to check for
2552 @type hypervisor_name: C{str}
2553 @param hypervisor_name: the hypervisor to ask for memory stats
2554 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2555 we cannot check the node
2558 nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2559 nodeinfo[node].Raise()
2560 free_mem = nodeinfo[node].data.get('memory_free')
2561 if not isinstance(free_mem, int):
2562 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2563 " was '%s'" % (node, free_mem))
2564 if requested > free_mem:
2565 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2566 " needed %s MiB, available %s MiB" %
2567 (node, reason, requested, free_mem))
2570 class LUStartupInstance(LogicalUnit):
2571 """Starts an instance.
2574 HPATH = "instance-start"
2575 HTYPE = constants.HTYPE_INSTANCE
2576 _OP_REQP = ["instance_name", "force"]
2579 def ExpandNames(self):
2580 self._ExpandAndLockInstance()
2582 def BuildHooksEnv(self):
2585 This runs on master, primary and secondary nodes of the instance.
2589 "FORCE": self.op.force,
2591 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2592 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2595 def CheckPrereq(self):
2596 """Check prerequisites.
2598 This checks that the instance is in the cluster.
2601 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2602 assert self.instance is not None, \
2603 "Cannot retrieve locked instance %s" % self.op.instance_name
2605 _CheckNodeOnline(self, instance.primary_node)
2607 bep = self.cfg.GetClusterInfo().FillBE(instance)
2608 # check bridges existance
2609 _CheckInstanceBridgesExist(self, instance)
2611 _CheckNodeFreeMemory(self, instance.primary_node,
2612 "starting instance %s" % instance.name,
2613 bep[constants.BE_MEMORY], instance.hypervisor)
2615 def Exec(self, feedback_fn):
2616 """Start the instance.
2619 instance = self.instance
2620 force = self.op.force
2621 extra_args = getattr(self.op, "extra_args", "")
2623 self.cfg.MarkInstanceUp(instance.name)
2625 node_current = instance.primary_node
2627 _StartInstanceDisks(self, instance, force)
2629 result = self.rpc.call_instance_start(node_current, instance, extra_args)
2630 if result.failed or not result.data:
2631 _ShutdownInstanceDisks(self, instance)
2632 raise errors.OpExecError("Could not start instance")
2635 class LURebootInstance(LogicalUnit):
2636 """Reboot an instance.
2639 HPATH = "instance-reboot"
2640 HTYPE = constants.HTYPE_INSTANCE
2641 _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2644 def ExpandNames(self):
2645 if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2646 constants.INSTANCE_REBOOT_HARD,
2647 constants.INSTANCE_REBOOT_FULL]:
2648 raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2649 (constants.INSTANCE_REBOOT_SOFT,
2650 constants.INSTANCE_REBOOT_HARD,
2651 constants.INSTANCE_REBOOT_FULL))
2652 self._ExpandAndLockInstance()
2654 def BuildHooksEnv(self):
2657 This runs on master, primary and secondary nodes of the instance.
2661 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2663 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2664 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2667 def CheckPrereq(self):
2668 """Check prerequisites.
2670 This checks that the instance is in the cluster.
2673 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2674 assert self.instance is not None, \
2675 "Cannot retrieve locked instance %s" % self.op.instance_name
2677 _CheckNodeOnline(self, instance.primary_node)
2679 # check bridges existance
2680 _CheckInstanceBridgesExist(self, instance)
2682 def Exec(self, feedback_fn):
2683 """Reboot the instance.
2686 instance = self.instance
2687 ignore_secondaries = self.op.ignore_secondaries
2688 reboot_type = self.op.reboot_type
2689 extra_args = getattr(self.op, "extra_args", "")
2691 node_current = instance.primary_node
2693 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2694 constants.INSTANCE_REBOOT_HARD]:
2695 result = self.rpc.call_instance_reboot(node_current, instance,
2696 reboot_type, extra_args)
2697 if result.failed or not result.data:
2698 raise errors.OpExecError("Could not reboot instance")
2700 if not self.rpc.call_instance_shutdown(node_current, instance):
2701 raise errors.OpExecError("could not shutdown instance for full reboot")
2702 _ShutdownInstanceDisks(self, instance)
2703 _StartInstanceDisks(self, instance, ignore_secondaries)
2704 result = self.rpc.call_instance_start(node_current, instance, extra_args)
2705 if result.failed or not result.data:
2706 _ShutdownInstanceDisks(self, instance)
2707 raise errors.OpExecError("Could not start instance for full reboot")
2709 self.cfg.MarkInstanceUp(instance.name)
2712 class LUShutdownInstance(LogicalUnit):
2713 """Shutdown an instance.
2716 HPATH = "instance-stop"
2717 HTYPE = constants.HTYPE_INSTANCE
2718 _OP_REQP = ["instance_name"]
2721 def ExpandNames(self):
2722 self._ExpandAndLockInstance()
2724 def BuildHooksEnv(self):
2727 This runs on master, primary and secondary nodes of the instance.
2730 env = _BuildInstanceHookEnvByObject(self, self.instance)
2731 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2734 def CheckPrereq(self):
2735 """Check prerequisites.
2737 This checks that the instance is in the cluster.
2740 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2741 assert self.instance is not None, \
2742 "Cannot retrieve locked instance %s" % self.op.instance_name
2743 _CheckNodeOnline(self, self.instance.primary_node)
2745 def Exec(self, feedback_fn):
2746 """Shutdown the instance.
2749 instance = self.instance
2750 node_current = instance.primary_node
2751 self.cfg.MarkInstanceDown(instance.name)
2752 result = self.rpc.call_instance_shutdown(node_current, instance)
2753 if result.failed or not result.data:
2754 self.proc.LogWarning("Could not shutdown instance")
2756 _ShutdownInstanceDisks(self, instance)
2759 class LUReinstallInstance(LogicalUnit):
2760 """Reinstall an instance.
2763 HPATH = "instance-reinstall"
2764 HTYPE = constants.HTYPE_INSTANCE
2765 _OP_REQP = ["instance_name"]
2768 def ExpandNames(self):
2769 self._ExpandAndLockInstance()
2771 def BuildHooksEnv(self):
2774 This runs on master, primary and secondary nodes of the instance.
2777 env = _BuildInstanceHookEnvByObject(self, self.instance)
2778 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2781 def CheckPrereq(self):
2782 """Check prerequisites.
2784 This checks that the instance is in the cluster and is not running.
2787 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2788 assert instance is not None, \
2789 "Cannot retrieve locked instance %s" % self.op.instance_name
2790 _CheckNodeOnline(self, instance.primary_node)
2792 if instance.disk_template == constants.DT_DISKLESS:
2793 raise errors.OpPrereqError("Instance '%s' has no disks" %
2794 self.op.instance_name)
2795 if instance.status != "down":
2796 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2797 self.op.instance_name)
2798 remote_info = self.rpc.call_instance_info(instance.primary_node,
2800 instance.hypervisor)
2801 if remote_info.failed or remote_info.data:
2802 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2803 (self.op.instance_name,
2804 instance.primary_node))
2806 self.op.os_type = getattr(self.op, "os_type", None)
2807 if self.op.os_type is not None:
2809 pnode = self.cfg.GetNodeInfo(
2810 self.cfg.ExpandNodeName(instance.primary_node))
2812 raise errors.OpPrereqError("Primary node '%s' is unknown" %
2814 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2816 if not isinstance(result.data, objects.OS):
2817 raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2818 " primary node" % self.op.os_type)
2820 self.instance = instance
2822 def Exec(self, feedback_fn):
2823 """Reinstall the instance.
2826 inst = self.instance
2828 if self.op.os_type is not None:
2829 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2830 inst.os = self.op.os_type
2831 self.cfg.Update(inst)
2833 _StartInstanceDisks(self, inst, None)
2835 feedback_fn("Running the instance OS create scripts...")
2836 result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2839 raise errors.OpExecError("Could not install OS for instance %s"
2841 (inst.name, inst.primary_node))
2843 _ShutdownInstanceDisks(self, inst)
2846 class LURenameInstance(LogicalUnit):
2847 """Rename an instance.
2850 HPATH = "instance-rename"
2851 HTYPE = constants.HTYPE_INSTANCE
2852 _OP_REQP = ["instance_name", "new_name"]
2854 def BuildHooksEnv(self):
2857 This runs on master, primary and secondary nodes of the instance.
2860 env = _BuildInstanceHookEnvByObject(self, self.instance)
2861 env["INSTANCE_NEW_NAME"] = self.op.new_name
2862 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2865 def CheckPrereq(self):
2866 """Check prerequisites.
2868 This checks that the instance is in the cluster and is not running.
2871 instance = self.cfg.GetInstanceInfo(
2872 self.cfg.ExpandInstanceName(self.op.instance_name))
2873 if instance is None:
2874 raise errors.OpPrereqError("Instance '%s' not known" %
2875 self.op.instance_name)
2876 _CheckNodeOnline(self, instance.primary_node)
2878 if instance.status != "down":
2879 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2880 self.op.instance_name)
2881 remote_info = self.rpc.call_instance_info(instance.primary_node,
2883 instance.hypervisor)
2885 if remote_info.data:
2886 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2887 (self.op.instance_name,
2888 instance.primary_node))
2889 self.instance = instance
2891 # new name verification
2892 name_info = utils.HostInfo(self.op.new_name)
2894 self.op.new_name = new_name = name_info.name
2895 instance_list = self.cfg.GetInstanceList()
2896 if new_name in instance_list:
2897 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2900 if not getattr(self.op, "ignore_ip", False):
2901 if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2902 raise errors.OpPrereqError("IP %s of instance %s already in use" %
2903 (name_info.ip, new_name))
2906 def Exec(self, feedback_fn):
2907 """Reinstall the instance.
2910 inst = self.instance
2911 old_name = inst.name
2913 if inst.disk_template == constants.DT_FILE:
2914 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2916 self.cfg.RenameInstance(inst.name, self.op.new_name)
2917 # Change the instance lock. This is definitely safe while we hold the BGL
2918 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2919 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2921 # re-read the instance from the configuration after rename
2922 inst = self.cfg.GetInstanceInfo(self.op.new_name)
2924 if inst.disk_template == constants.DT_FILE:
2925 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2926 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2927 old_file_storage_dir,
2928 new_file_storage_dir)
2931 raise errors.OpExecError("Could not connect to node '%s' to rename"
2932 " directory '%s' to '%s' (but the instance"
2933 " has been renamed in Ganeti)" % (
2934 inst.primary_node, old_file_storage_dir,
2935 new_file_storage_dir))
2937 if not result.data[0]:
2938 raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2939 " (but the instance has been renamed in"
2940 " Ganeti)" % (old_file_storage_dir,
2941 new_file_storage_dir))
2943 _StartInstanceDisks(self, inst, None)
2945 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
2947 if result.failed or not result.data:
2948 msg = ("Could not run OS rename script for instance %s on node %s"
2949 " (but the instance has been renamed in Ganeti)" %
2950 (inst.name, inst.primary_node))
2951 self.proc.LogWarning(msg)
2953 _ShutdownInstanceDisks(self, inst)
2956 class LURemoveInstance(LogicalUnit):
2957 """Remove an instance.
2960 HPATH = "instance-remove"
2961 HTYPE = constants.HTYPE_INSTANCE
2962 _OP_REQP = ["instance_name", "ignore_failures"]
2965 def ExpandNames(self):
2966 self._ExpandAndLockInstance()
2967 self.needed_locks[locking.LEVEL_NODE] = []
2968 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2970 def DeclareLocks(self, level):
2971 if level == locking.LEVEL_NODE:
2972 self._LockInstancesNodes()
2974 def BuildHooksEnv(self):
2977 This runs on master, primary and secondary nodes of the instance.
2980 env = _BuildInstanceHookEnvByObject(self, self.instance)
2981 nl = [self.cfg.GetMasterNode()]
2984 def CheckPrereq(self):
2985 """Check prerequisites.
2987 This checks that the instance is in the cluster.
2990 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2991 assert self.instance is not None, \
2992 "Cannot retrieve locked instance %s" % self.op.instance_name
2994 def Exec(self, feedback_fn):
2995 """Remove the instance.
2998 instance = self.instance
2999 logging.info("Shutting down instance %s on node %s",
3000 instance.name, instance.primary_node)
3002 result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3003 if result.failed or not result.data:
3004 if self.op.ignore_failures:
3005 feedback_fn("Warning: can't shutdown instance")
3007 raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3008 (instance.name, instance.primary_node))
3010 logging.info("Removing block devices for instance %s", instance.name)
3012 if not _RemoveDisks(self, instance):
3013 if self.op.ignore_failures:
3014 feedback_fn("Warning: can't remove instance's disks")
3016 raise errors.OpExecError("Can't remove instance's disks")
3018 logging.info("Removing instance %s out of cluster config", instance.name)
3020 self.cfg.RemoveInstance(instance.name)
3021 self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3024 class LUQueryInstances(NoHooksLU):
3025 """Logical unit for querying instances.
3028 _OP_REQP = ["output_fields", "names"]
3030 _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3031 "admin_state", "admin_ram",
3032 "disk_template", "ip", "mac", "bridge",
3033 "sda_size", "sdb_size", "vcpus", "tags",
3034 "network_port", "beparams",
3035 "(disk).(size)/([0-9]+)",
3037 "(nic).(mac|ip|bridge)/([0-9]+)",
3038 "(nic).(macs|ips|bridges)",
3039 "(disk|nic).(count)",
3040 "serial_no", "hypervisor", "hvparams",] +
3042 for name in constants.HVS_PARAMETERS] +
3044 for name in constants.BES_PARAMETERS])
3045 _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3048 def ExpandNames(self):
3049 _CheckOutputFields(static=self._FIELDS_STATIC,
3050 dynamic=self._FIELDS_DYNAMIC,
3051 selected=self.op.output_fields)
3053 self.needed_locks = {}
3054 self.share_locks[locking.LEVEL_INSTANCE] = 1
3055 self.share_locks[locking.LEVEL_NODE] = 1
3058 self.wanted = _GetWantedInstances(self, self.op.names)
3060 self.wanted = locking.ALL_SET
3062 self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3064 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3065 self.needed_locks[locking.LEVEL_NODE] = []
3066 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3068 def DeclareLocks(self, level):
3069 if level == locking.LEVEL_NODE and self.do_locking:
3070 self._LockInstancesNodes()
3072 def CheckPrereq(self):
3073 """Check prerequisites.
3078 def Exec(self, feedback_fn):
3079 """Computes the list of nodes and their attributes.
3082 all_info = self.cfg.GetAllInstancesInfo()
3084 instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3085 elif self.wanted != locking.ALL_SET:
3086 instance_names = self.wanted
3087 missing = set(instance_names).difference(all_info.keys())
3089 raise errors.OpExecError(
3090 "Some instances were removed before retrieving their data: %s"
3093 instance_names = all_info.keys()
3095 instance_names = utils.NiceSort(instance_names)
3096 instance_list = [all_info[iname] for iname in instance_names]
3098 # begin data gathering
3100 nodes = frozenset([inst.primary_node for inst in instance_list])
3101 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3107 node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3109 result = node_data[name]
3111 # offline nodes will be in both lists
3112 off_nodes.append(name)
3114 bad_nodes.append(name)
3117 live_data.update(result.data)
3118 # else no instance is alive
3120 live_data = dict([(name, {}) for name in instance_names])
3122 # end data gathering
3127 for instance in instance_list:
3129 i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3130 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3131 for field in self.op.output_fields:
3132 st_match = self._FIELDS_STATIC.Matches(field)
3137 elif field == "pnode":
3138 val = instance.primary_node
3139 elif field == "snodes":
3140 val = list(instance.secondary_nodes)
3141 elif field == "admin_state":
3142 val = (instance.status != "down")
3143 elif field == "oper_state":
3144 if instance.primary_node in bad_nodes:
3147 val = bool(live_data.get(instance.name))
3148 elif field == "status":
3149 if instance.primary_node in off_nodes:
3150 val = "ERROR_nodeoffline"
3151 elif instance.primary_node in bad_nodes:
3152 val = "ERROR_nodedown"
3154 running = bool(live_data.get(instance.name))
3156 if instance.status != "down":
3161 if instance.status != "down":
3165 elif field == "oper_ram":
3166 if instance.primary_node in bad_nodes:
3168 elif instance.name in live_data:
3169 val = live_data[instance.name].get("memory", "?")
3172 elif field == "disk_template":
3173 val = instance.disk_template
3175 val = instance.nics[0].ip
3176 elif field == "bridge":
3177 val = instance.nics[0].bridge
3178 elif field == "mac":
3179 val = instance.nics[0].mac
3180 elif field == "sda_size" or field == "sdb_size":
3181 idx = ord(field[2]) - ord('a')
3183 val = instance.FindDisk(idx).size
3184 except errors.OpPrereqError:
3186 elif field == "tags":
3187 val = list(instance.GetTags())
3188 elif field == "serial_no":
3189 val = instance.serial_no
3190 elif field == "network_port":
3191 val = instance.network_port
3192 elif field == "hypervisor":
3193 val = instance.hypervisor
3194 elif field == "hvparams":
3196 elif (field.startswith(HVPREFIX) and
3197 field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3198 val = i_hv.get(field[len(HVPREFIX):], None)
3199 elif field == "beparams":
3201 elif (field.startswith(BEPREFIX) and
3202 field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3203 val = i_be.get(field[len(BEPREFIX):], None)
3204 elif st_match and st_match.groups():
3205 # matches a variable list
3206 st_groups = st_match.groups()
3207 if st_groups and st_groups[0] == "disk":
3208 if st_groups[1] == "count":
3209 val = len(instance.disks)
3210 elif st_groups[1] == "sizes":
3211 val = [disk.size for disk in instance.disks]
3212 elif st_groups[1] == "size":
3214 val = instance.FindDisk(st_groups[2]).size
3215 except errors.OpPrereqError:
3218 assert False, "Unhandled disk parameter"
3219 elif st_groups[0] == "nic":
3220 if st_groups[1] == "count":
3221 val = len(instance.nics)
3222 elif st_groups[1] == "macs":
3223 val = [nic.mac for nic in instance.nics]
3224 elif st_groups[1] == "ips":
3225 val = [nic.ip for nic in instance.nics]
3226 elif st_groups[1] == "bridges":
3227 val = [nic.bridge for nic in instance.nics]
3230 nic_idx = int(st_groups[2])
3231 if nic_idx >= len(instance.nics):
3234 if st_groups[1] == "mac":
3235 val = instance.nics[nic_idx].mac
3236 elif st_groups[1] == "ip":
3237 val = instance.nics[nic_idx].ip
3238 elif st_groups[1] == "bridge":
3239 val = instance.nics[nic_idx].bridge
3241 assert False, "Unhandled NIC parameter"
3243 assert False, "Unhandled variable parameter"
3245 raise errors.ParameterError(field)
3252 class LUFailoverInstance(LogicalUnit):
3253 """Failover an instance.
3256 HPATH = "instance-failover"
3257 HTYPE = constants.HTYPE_INSTANCE
3258 _OP_REQP = ["instance_name", "ignore_consistency"]
3261 def ExpandNames(self):
3262 self._ExpandAndLockInstance()
3263 self.needed_locks[locking.LEVEL_NODE] = []
3264 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3266 def DeclareLocks(self, level):
3267 if level == locking.LEVEL_NODE:
3268 self._LockInstancesNodes()
3270 def BuildHooksEnv(self):
3273 This runs on master, primary and secondary nodes of the instance.
3277 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3279 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3280 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3283 def CheckPrereq(self):
3284 """Check prerequisites.
3286 This checks that the instance is in the cluster.
3289 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3290 assert self.instance is not None, \
3291 "Cannot retrieve locked instance %s" % self.op.instance_name
3293 bep = self.cfg.GetClusterInfo().FillBE(instance)
3294 if instance.disk_template not in constants.DTS_NET_MIRROR:
3295 raise errors.OpPrereqError("Instance's disk layout is not"
3296 " network mirrored, cannot failover.")
3298 secondary_nodes = instance.secondary_nodes
3299 if not secondary_nodes:
3300 raise errors.ProgrammerError("no secondary node but using "
3301 "a mirrored disk template")
3303 target_node = secondary_nodes[0]
3304 _CheckNodeOnline(self, target_node)
3305 # check memory requirements on the secondary node
3306 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3307 instance.name, bep[constants.BE_MEMORY],
3308 instance.hypervisor)
3310 # check bridge existance
3311 brlist = [nic.bridge for nic in instance.nics]
3312 result = self.rpc.call_bridges_exist(target_node, brlist)
3315 raise errors.OpPrereqError("One or more target bridges %s does not"
3316 " exist on destination node '%s'" %
3317 (brlist, target_node))
3319 def Exec(self, feedback_fn):
3320 """Failover an instance.
3322 The failover is done by shutting it down on its present node and
3323 starting it on the secondary.
3326 instance = self.instance
3328 source_node = instance.primary_node
3329 target_node = instance.secondary_nodes[0]
3331 feedback_fn("* checking disk consistency between source and target")
3332 for dev in instance.disks:
3333 # for drbd, these are drbd over lvm
3334 if not _CheckDiskConsistency(self, dev, target_node, False):
3335 if instance.status == "up" and not self.op.ignore_consistency:
3336 raise errors.OpExecError("Disk %s is degraded on target node,"
3337 " aborting failover." % dev.iv_name)
3339 feedback_fn("* shutting down instance on source node")
3340 logging.info("Shutting down instance %s on node %s",
3341 instance.name, source_node)
3343 result = self.rpc.call_instance_shutdown(source_node, instance)
3344 if result.failed or not result.data:
3345 if self.op.ignore_consistency:
3346 self.proc.LogWarning("Could not shutdown instance %s on node %s."
3348 " anyway. Please make sure node %s is down",
3349 instance.name, source_node, source_node)
3351 raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3352 (instance.name, source_node))
3354 feedback_fn("* deactivating the instance's disks on source node")
3355 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3356 raise errors.OpExecError("Can't shut down the instance's disks.")
3358 instance.primary_node = target_node
3359 # distribute new instance config to the other nodes
3360 self.cfg.Update(instance)
3362 # Only start the instance if it's marked as up
3363 if instance.status == "up":
3364 feedback_fn("* activating the instance's disks on target node")
3365 logging.info("Starting instance %s on node %s",
3366 instance.name, target_node)
3368 disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3369 ignore_secondaries=True)
3371 _ShutdownInstanceDisks(self, instance)
3372 raise errors.OpExecError("Can't activate the instance's disks")
3374 feedback_fn("* starting the instance on the target node")
3375 result = self.rpc.call_instance_start(target_node, instance, None)
3376 if result.failed or not result.data:
3377 _ShutdownInstanceDisks(self, instance)
3378 raise errors.OpExecError("Could not start instance %s on node %s." %
3379 (instance.name, target_node))
3382 class LUMigrateInstance(LogicalUnit):
3383 """Migrate an instance.
3385 This is migration without shutting down, compared to the failover,
3386 which is done with shutdown.
3389 HPATH = "instance-migrate"
3390 HTYPE = constants.HTYPE_INSTANCE
3391 _OP_REQP = ["instance_name", "live", "cleanup"]
3395 def ExpandNames(self):
3396 self._ExpandAndLockInstance()
3397 self.needed_locks[locking.LEVEL_NODE] = []
3398 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3400 def DeclareLocks(self, level):
3401 if level == locking.LEVEL_NODE:
3402 self._LockInstancesNodes()
3404 def BuildHooksEnv(self):
3407 This runs on master, primary and secondary nodes of the instance.
3410 env = _BuildInstanceHookEnvByObject(self, self.instance)
3411 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3414 def CheckPrereq(self):
3415 """Check prerequisites.
3417 This checks that the instance is in the cluster.
3420 instance = self.cfg.GetInstanceInfo(
3421 self.cfg.ExpandInstanceName(self.op.instance_name))
3422 if instance is None:
3423 raise errors.OpPrereqError("Instance '%s' not known" %
3424 self.op.instance_name)
3426 if instance.disk_template != constants.DT_DRBD8:
3427 raise errors.OpPrereqError("Instance's disk layout is not"
3428 " drbd8, cannot migrate.")
3430 secondary_nodes = instance.secondary_nodes
3431 if not secondary_nodes:
3432 raise errors.ProgrammerError("no secondary node but using "
3433 "drbd8 disk template")
3435 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3437 target_node = secondary_nodes[0]
3438 # check memory requirements on the secondary node
3439 _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3440 instance.name, i_be[constants.BE_MEMORY],
3441 instance.hypervisor)
3443 # check bridge existance
3444 brlist = [nic.bridge for nic in instance.nics]
3445 result = self.rpc.call_bridges_exist(target_node, brlist)
3446 if result.failed or not result.data:
3447 raise errors.OpPrereqError("One or more target bridges %s does not"
3448 " exist on destination node '%s'" %
3449 (brlist, target_node))
3451 if not self.op.cleanup:
3452 result = self.rpc.call_instance_migratable(instance.primary_node,
3454 msg = result.RemoteFailMsg()
3456 raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3459 self.instance = instance
3461 def _WaitUntilSync(self):
3462 """Poll with custom rpc for disk sync.
3464 This uses our own step-based rpc call.
3467 self.feedback_fn("* wait until resync is done")
3471 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3473 self.instance.disks)
3475 for node, nres in result.items():
3476 msg = nres.RemoteFailMsg()
3478 raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3480 node_done, node_percent = nres.data[1]
3481 all_done = all_done and node_done
3482 if node_percent is not None:
3483 min_percent = min(min_percent, node_percent)
3485 if min_percent < 100:
3486 self.feedback_fn(" - progress: %.1f%%" % min_percent)
3489 def _EnsureSecondary(self, node):
3490 """Demote a node to secondary.
3493 self.feedback_fn("* switching node %s to secondary mode" % node)
3495 for dev in self.instance.disks:
3496 self.cfg.SetDiskID(dev, node)
3498 result = self.rpc.call_blockdev_close(node, self.instance.name,
3499 self.instance.disks)
3500 msg = result.RemoteFailMsg()
3502 raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3503 " error %s" % (node, msg))
3505 def _GoStandalone(self):
3506 """Disconnect from the network.
3509 self.feedback_fn("* changing into standalone mode")
3510 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3511 self.instance.disks)
3512 for node, nres in result.items():
3513 msg = nres.RemoteFailMsg()
3515 raise errors.OpExecError("Cannot disconnect disks node %s,"
3516 " error %s" % (node, msg))
3518 def _GoReconnect(self, multimaster):
3519 """Reconnect to the network.
3525 msg = "single-master"
3526 self.feedback_fn("* changing disks into %s mode" % msg)
3527 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3528 self.instance.disks,
3529 self.instance.name, multimaster)
3530 for node, nres in result.items():
3531 msg = nres.RemoteFailMsg()
3533 raise errors.OpExecError("Cannot change disks config on node %s,"
3534 " error: %s" % (node, msg))
3536 def _ExecCleanup(self):
3537 """Try to cleanup after a failed migration.
3539 The cleanup is done by:
3540 - check that the instance is running only on one node
3541 (and update the config if needed)
3542 - change disks on its secondary node to secondary
3543 - wait until disks are fully synchronized
3544 - disconnect from the network
3545 - change disks into single-master mode
3546 - wait again until disks are fully synchronized
3549 instance = self.instance
3550 target_node = self.target_node
3551 source_node = self.source_node
3553 # check running on only one node
3554 self.feedback_fn("* checking where the instance actually runs"
3555 " (if this hangs, the hypervisor might be in"
3557 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3558 for node, result in ins_l.items():
3560 if not isinstance(result.data, list):
3561 raise errors.OpExecError("Can't contact node '%s'" % node)
3563 runningon_source = instance.name in ins_l[source_node].data
3564 runningon_target = instance.name in ins_l[target_node].data
3566 if runningon_source and runningon_target:
3567 raise errors.OpExecError("Instance seems to be running on two nodes,"
3568 " or the hypervisor is confused. You will have"
3569 " to ensure manually that it runs only on one"
3570 " and restart this operation.")
3572 if not (runningon_source or runningon_target):
3573 raise errors.OpExecError("Instance does not seem to be running at all."
3574 " In this case, it's safer to repair by"
3575 " running 'gnt-instance stop' to ensure disk"
3576 " shutdown, and then restarting it.")
3578 if runningon_target:
3579 # the migration has actually succeeded, we need to update the config
3580 self.feedback_fn("* instance running on secondary node (%s),"
3581 " updating config" % target_node)
3582 instance.primary_node = target_node
3583 self.cfg.Update(instance)
3584 demoted_node = source_node
3586 self.feedback_fn("* instance confirmed to be running on its"
3587 " primary node (%s)" % source_node)
3588 demoted_node = target_node
3590 self._EnsureSecondary(demoted_node)
3592 self._WaitUntilSync()
3593 except errors.OpExecError:
3594 # we ignore here errors, since if the device is standalone, it
3595 # won't be able to sync
3597 self._GoStandalone()
3598 self._GoReconnect(False)
3599 self._WaitUntilSync()
3601 self.feedback_fn("* done")
3603 def _ExecMigration(self):
3604 """Migrate an instance.
3606 The migrate is done by:
3607 - change the disks into dual-master mode
3608 - wait until disks are fully synchronized again
3609 - migrate the instance
3610 - change disks on the new secondary node (the old primary) to secondary
3611 - wait until disks are fully synchronized
3612 - change disks into single-master mode
3615 instance = self.instance
3616 target_node = self.target_node
3617 source_node = self.source_node
3619 self.feedback_fn("* checking disk consistency between source and target")
3620 for dev in instance.disks:
3621 if not _CheckDiskConsistency(self, dev, target_node, False):
3622 raise errors.OpExecError("Disk %s is degraded or not fully"
3623 " synchronized on target node,"
3624 " aborting migrate." % dev.iv_name)
3626 self._EnsureSecondary(target_node)
3627 self._GoStandalone()
3628 self._GoReconnect(True)
3629 self._WaitUntilSync()
3631 self.feedback_fn("* migrating instance to %s" % target_node)
3633 result = self.rpc.call_instance_migrate(source_node, instance,
3634 self.nodes_ip[target_node],
3636 msg = result.RemoteFailMsg()
3638 logging.error("Instance migration failed, trying to revert"
3639 " disk status: %s", msg)
3641 self._EnsureSecondary(target_node)
3642 self._GoStandalone()
3643 self._GoReconnect(False)
3644 self._WaitUntilSync()
3645 except errors.OpExecError, err:
3646 self.LogWarning("Migration failed and I can't reconnect the"
3647 " drives: error '%s'\n"
3648 "Please look and recover the instance status" %
3651 raise errors.OpExecError("Could not migrate instance %s: %s" %
3652 (instance.name, msg))
3655 instance.primary_node = target_node
3656 # distribute new instance config to the other nodes
3657 self.cfg.Update(instance)
3659 self._EnsureSecondary(source_node)
3660 self._WaitUntilSync()
3661 self._GoStandalone()
3662 self._GoReconnect(False)
3663 self._WaitUntilSync()
3665 self.feedback_fn("* done")
3667 def Exec(self, feedback_fn):
3668 """Perform the migration.
3671 self.feedback_fn = feedback_fn
3673 self.source_node = self.instance.primary_node
3674 self.target_node = self.instance.secondary_nodes[0]
3675 self.all_nodes = [self.source_node, self.target_node]
3677 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3678 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3681 return self._ExecCleanup()
3683 return self._ExecMigration()
3686 def _CreateBlockDev(lu, node, instance, device, force_create,
3688 """Create a tree of block devices on a given node.
3690 If this device type has to be created on secondaries, create it and
3693 If not, just recurse to children keeping the same 'force' value.
3695 @param lu: the lu on whose behalf we execute
3696 @param node: the node on which to create the device
3697 @type instance: L{objects.Instance}
3698 @param instance: the instance which owns the device
3699 @type device: L{objects.Disk}
3700 @param device: the device to create
3701 @type force_create: boolean
3702 @param force_create: whether to force creation of this device; this
3703 will be change to True whenever we find a device which has
3704 CreateOnSecondary() attribute
3705 @param info: the extra 'metadata' we should attach to the device
3706 (this will be represented as a LVM tag)
3707 @type force_open: boolean
3708 @param force_open: this parameter will be passes to the
3709 L{backend.CreateBlockDevice} function where it specifies
3710 whether we run on primary or not, and it affects both
3711 the child assembly and the device own Open() execution
3714 if device.CreateOnSecondary():
3718 for child in device.children:
3719 _CreateBlockDev(lu, node, instance, child, force_create,
3722 if not force_create:
3725 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3728 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3729 """Create a single block device on a given node.
3731 This will not recurse over children of the device, so they must be
3734 @param lu: the lu on whose behalf we execute
3735 @param node: the node on which to create the device
3736 @type instance: L{objects.Instance}
3737 @param instance: the instance which owns the device
3738 @type device: L{objects.Disk}
3739 @param device: the device to create
3740 @param info: the extra 'metadata' we should attach to the device
3741 (this will be represented as a LVM tag)
3742 @type force_open: boolean
3743 @param force_open: this parameter will be passes to the
3744 L{backend.CreateBlockDevice} function where it specifies
3745 whether we run on primary or not, and it affects both
3746 the child assembly and the device own Open() execution
3749 lu.cfg.SetDiskID(device, node)
3750 result = lu.rpc.call_blockdev_create(node, device, device.size,
3751 instance.name, force_open, info)
3752 msg = result.RemoteFailMsg()
3754 raise errors.OpExecError("Can't create block device %s on"
3755 " node %s for instance %s: %s" %
3756 (device, node, instance.name, msg))
3757 if device.physical_id is None:
3758 device.physical_id = result.data[1]
3761 def _GenerateUniqueNames(lu, exts):
3762 """Generate a suitable LV name.
3764 This will generate a logical volume name for the given instance.
3769 new_id = lu.cfg.GenerateUniqueID()
3770 results.append("%s%s" % (new_id, val))
3774 def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3776 """Generate a drbd8 device complete with its children.
3779 port = lu.cfg.AllocatePort()
3780 vgname = lu.cfg.GetVGName()
3781 shared_secret = lu.cfg.GenerateDRBDSecret()
3782 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3783 logical_id=(vgname, names[0]))
3784 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3785 logical_id=(vgname, names[1]))
3786 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3787 logical_id=(primary, secondary, port,
3790 children=[dev_data, dev_meta],
3795 def _GenerateDiskTemplate(lu, template_name,
3796 instance_name, primary_node,
3797 secondary_nodes, disk_info,
3798 file_storage_dir, file_driver,
3800 """Generate the entire disk layout for a given template type.
3803 #TODO: compute space requirements
3805 vgname = lu.cfg.GetVGName()
3806 disk_count = len(disk_info)
3808 if template_name == constants.DT_DISKLESS:
3810 elif template_name == constants.DT_PLAIN:
3811 if len(secondary_nodes) != 0:
3812 raise errors.ProgrammerError("Wrong template configuration")
3814 names = _GenerateUniqueNames(lu, [".disk%d" % i
3815 for i in range(disk_count)])
3816 for idx, disk in enumerate(disk_info):
3817 disk_index = idx + base_index
3818 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3819 logical_id=(vgname, names[idx]),
3820 iv_name="disk/%d" % disk_index)
3821 disks.append(disk_dev)
3822 elif template_name == constants.DT_DRBD8:
3823 if len(secondary_nodes) != 1:
3824 raise errors.ProgrammerError("Wrong template configuration")
3825 remote_node = secondary_nodes[0]
3826 minors = lu.cfg.AllocateDRBDMinor(
3827 [primary_node, remote_node] * len(disk_info), instance_name)
3830 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
3831 for i in range(disk_count)]):
3832 names.append(lv_prefix + "_data")
3833 names.append(lv_prefix + "_meta")
3834 for idx, disk in enumerate(disk_info):
3835 disk_index = idx + base_index
3836 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3837 disk["size"], names[idx*2:idx*2+2],
3838 "disk/%d" % disk_index,
3839 minors[idx*2], minors[idx*2+1])
3840 disks.append(disk_dev)
3841 elif template_name == constants.DT_FILE:
3842 if len(secondary_nodes) != 0:
3843 raise errors.ProgrammerError("Wrong template configuration")
3845 for idx, disk in enumerate(disk_info):
3846 disk_index = idx + base_index
3847 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3848 iv_name="disk/%d" % disk_index,
3849 logical_id=(file_driver,
3850 "%s/disk%d" % (file_storage_dir,
3852 disks.append(disk_dev)
3854 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3858 def _GetInstanceInfoText(instance):
3859 """Compute that text that should be added to the disk's metadata.
3862 return "originstname+%s" % instance.name
3865 def _CreateDisks(lu, instance):
3866 """Create all disks for an instance.
3868 This abstracts away some work from AddInstance.
3870 @type lu: L{LogicalUnit}
3871 @param lu: the logical unit on whose behalf we execute
3872 @type instance: L{objects.Instance}
3873 @param instance: the instance whose disks we should create
3875 @return: the success of the creation
3878 info = _GetInstanceInfoText(instance)
3879 pnode = instance.primary_node
3881 if instance.disk_template == constants.DT_FILE:
3882 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3883 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
3885 if result.failed or not result.data:
3886 raise errors.OpExecError("Could not connect to node '%s'" % pnode)
3888 if not result.data[0]:
3889 raise errors.OpExecError("Failed to create directory '%s'" %
3892 # Note: this needs to be kept in sync with adding of disks in
3893 # LUSetInstanceParams
3894 for device in instance.disks:
3895 logging.info("Creating volume %s for instance %s",
3896 device.iv_name, instance.name)
3898 for node in instance.all_nodes:
3899 f_create = node == pnode
3900 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
3903 def _RemoveDisks(lu, instance):
3904 """Remove all disks for an instance.
3906 This abstracts away some work from `AddInstance()` and
3907 `RemoveInstance()`. Note that in case some of the devices couldn't
3908 be removed, the removal will continue with the other ones (compare
3909 with `_CreateDisks()`).
3911 @type lu: L{LogicalUnit}
3912 @param lu: the logical unit on whose behalf we execute
3913 @type instance: L{objects.Instance}
3914 @param instance: the instance whose disks we should remove
3916 @return: the success of the removal
3919 logging.info("Removing block devices for instance %s", instance.name)
3922 for device in instance.disks:
3923 for node, disk in device.ComputeNodeTree(instance.primary_node):
3924 lu.cfg.SetDiskID(disk, node)
3925 result = lu.rpc.call_blockdev_remove(node, disk)
3926 if result.failed or not result.data:
3927 lu.proc.LogWarning("Could not remove block device %s on node %s,"
3928 " continuing anyway", device.iv_name, node)
3931 if instance.disk_template == constants.DT_FILE:
3932 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3933 result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
3935 if result.failed or not result.data:
3936 logging.error("Could not remove directory '%s'", file_storage_dir)
3942 def _ComputeDiskSize(disk_template, disks):
3943 """Compute disk size requirements in the volume group
3946 # Required free disk space as a function of disk and swap space
3948 constants.DT_DISKLESS: None,
3949 constants.DT_PLAIN: sum(d["size"] for d in disks),
3950 # 128 MB are added for drbd metadata for each disk
3951 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
3952 constants.DT_FILE: None,
3955 if disk_template not in req_size_dict:
3956 raise errors.ProgrammerError("Disk template '%s' size requirement"
3957 " is unknown" % disk_template)
3959 return req_size_dict[disk_template]
3962 def _CheckHVParams(lu, nodenames, hvname, hvparams):
3963 """Hypervisor parameter validation.
3965 This function abstract the hypervisor parameter validation to be
3966 used in both instance create and instance modify.
3968 @type lu: L{LogicalUnit}
3969 @param lu: the logical unit for which we check
3970 @type nodenames: list
3971 @param nodenames: the list of nodes on which we should check
3972 @type hvname: string
3973 @param hvname: the name of the hypervisor we should use
3974 @type hvparams: dict
3975 @param hvparams: the parameters which we need to check
3976 @raise errors.OpPrereqError: if the parameters are not valid
3979 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
3982 for node in nodenames:
3985 if not info.data or not isinstance(info.data, (tuple, list)):
3986 raise errors.OpPrereqError("Cannot get current information"
3987 " from node '%s' (%s)" % (node, info.data))
3988 if not info.data[0]:
3989 raise errors.OpPrereqError("Hypervisor parameter validation failed:"
3990 " %s" % info.data[1])
3993 class LUCreateInstance(LogicalUnit):
3994 """Create an instance.
3997 HPATH = "instance-add"
3998 HTYPE = constants.HTYPE_INSTANCE
3999 _OP_REQP = ["instance_name", "disks", "disk_template",
4001 "wait_for_sync", "ip_check", "nics",
4002 "hvparams", "beparams"]
4005 def _ExpandNode(self, node):
4006 """Expands and checks one node name.
4009 node_full = self.cfg.ExpandNodeName(node)
4010 if node_full is None:
4011 raise errors.OpPrereqError("Unknown node %s" % node)
4014 def ExpandNames(self):
4015 """ExpandNames for CreateInstance.
4017 Figure out the right locks for instance creation.
4020 self.needed_locks = {}
4022 # set optional parameters to none if they don't exist
4023 for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4024 if not hasattr(self.op, attr):
4025 setattr(self.op, attr, None)
4027 # cheap checks, mostly valid constants given
4029 # verify creation mode
4030 if self.op.mode not in (constants.INSTANCE_CREATE,
4031 constants.INSTANCE_IMPORT):
4032 raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4035 # disk template and mirror node verification
4036 if self.op.disk_template not in constants.DISK_TEMPLATES:
4037 raise errors.OpPrereqError("Invalid disk template name")
4039 if self.op.hypervisor is None:
4040 self.op.hypervisor = self.cfg.GetHypervisorType()
4042 cluster = self.cfg.GetClusterInfo()
4043 enabled_hvs = cluster.enabled_hypervisors
4044 if self.op.hypervisor not in enabled_hvs:
4045 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4046 " cluster (%s)" % (self.op.hypervisor,
4047 ",".join(enabled_hvs)))
4049 # check hypervisor parameter syntax (locally)
4051 filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
4053 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4054 hv_type.CheckParameterSyntax(filled_hvp)
4056 # fill and remember the beparams dict
4057 utils.CheckBEParams(self.op.beparams)
4058 self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4061 #### instance parameters check
4063 # instance name verification
4064 hostname1 = utils.HostInfo(self.op.instance_name)
4065 self.op.instance_name = instance_name = hostname1.name
4067 # this is just a preventive check, but someone might still add this
4068 # instance in the meantime, and creation will fail at lock-add time
4069 if instance_name in self.cfg.GetInstanceList():
4070 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4073 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4077 for nic in self.op.nics:
4078 # ip validity checks
4079 ip = nic.get("ip", None)
4080 if ip is None or ip.lower() == "none":
4082 elif ip.lower() == constants.VALUE_AUTO:
4083 nic_ip = hostname1.ip
4085 if not utils.IsValidIP(ip):
4086 raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4087 " like a valid IP" % ip)
4090 # MAC address verification
4091 mac = nic.get("mac", constants.VALUE_AUTO)
4092 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4093 if not utils.IsValidMac(mac.lower()):
4094 raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4096 # bridge verification
4097 bridge = nic.get("bridge", self.cfg.GetDefBridge())
4098 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
4100 # disk checks/pre-build
4102 for disk in self.op.disks:
4103 mode = disk.get("mode", constants.DISK_RDWR)
4104 if mode not in constants.DISK_ACCESS_SET:
4105 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4107 size = disk.get("size", None)
4109 raise errors.OpPrereqError("Missing disk size")
4113 raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4114 self.disks.append({"size": size, "mode": mode})
4116 # used in CheckPrereq for ip ping check
4117 self.check_ip = hostname1.ip
4119 # file storage checks
4120 if (self.op.file_driver and
4121 not self.op.file_driver in constants.FILE_DRIVER):
4122 raise errors.OpPrereqError("Invalid file driver name '%s'" %
4123 self.op.file_driver)
4125 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4126 raise errors.OpPrereqError("File storage directory path not absolute")
4128 ### Node/iallocator related checks
4129 if [self.op.iallocator, self.op.pnode].count(None) != 1:
4130 raise errors.OpPrereqError("One and only one of iallocator and primary"
4131 " node must be given")
4133 if self.op.iallocator:
4134 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4136 self.op.pnode = self._ExpandNode(self.op.pnode)
4137 nodelist = [self.op.pnode]
4138 if self.op.snode is not None:
4139 self.op.snode = self._ExpandNode(self.op.snode)
4140 nodelist.append(self.op.snode)
4141 self.needed_locks[locking.LEVEL_NODE] = nodelist
4143 # in case of import lock the source node too
4144 if self.op.mode == constants.INSTANCE_IMPORT:
4145 src_node = getattr(self.op, "src_node", None)
4146 src_path = getattr(self.op, "src_path", None)
4148 if src_path is None:
4149 self.op.src_path = src_path = self.op.instance_name
4151 if src_node is None:
4152 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4153 self.op.src_node = None
4154 if os.path.isabs(src_path):
4155 raise errors.OpPrereqError("Importing an instance from an absolute"
4156 " path requires a source node option.")
4158 self.op.src_node = src_node = self._ExpandNode(src_node)
4159 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4160 self.needed_locks[locking.LEVEL_NODE].append(src_node)
4161 if not os.path.isabs(src_path):
4162 self.op.src_path = src_path = \
4163 os.path.join(constants.EXPORT_DIR, src_path)
4165 else: # INSTANCE_CREATE
4166 if getattr(self.op, "os_type", None) is None:
4167 raise errors.OpPrereqError("No guest OS specified")
4169 def _RunAllocator(self):
4170 """Run the allocator based on input opcode.
4173 nics = [n.ToDict() for n in self.nics]
4174 ial = IAllocator(self,
4175 mode=constants.IALLOCATOR_MODE_ALLOC,
4176 name=self.op.instance_name,
4177 disk_template=self.op.disk_template,
4180 vcpus=self.be_full[constants.BE_VCPUS],
4181 mem_size=self.be_full[constants.BE_MEMORY],
4184 hypervisor=self.op.hypervisor,
4187 ial.Run(self.op.iallocator)
4190 raise errors.OpPrereqError("Can't compute nodes using"
4191 " iallocator '%s': %s" % (self.op.iallocator,
4193 if len(ial.nodes) != ial.required_nodes:
4194 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4195 " of nodes (%s), required %s" %
4196 (self.op.iallocator, len(ial.nodes),
4197 ial.required_nodes))
4198 self.op.pnode = ial.nodes[0]
4199 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4200 self.op.instance_name, self.op.iallocator,
4201 ", ".join(ial.nodes))
4202 if ial.required_nodes == 2:
4203 self.op.snode = ial.nodes[1]
4205 def BuildHooksEnv(self):
4208 This runs on master, primary and secondary nodes of the instance.
4212 "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
4213 "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
4214 "INSTANCE_ADD_MODE": self.op.mode,
4216 if self.op.mode == constants.INSTANCE_IMPORT:
4217 env["INSTANCE_SRC_NODE"] = self.op.src_node
4218 env["INSTANCE_SRC_PATH"] = self.op.src_path
4219 env["INSTANCE_SRC_IMAGES"] = self.src_images
4221 env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
4222 primary_node=self.op.pnode,
4223 secondary_nodes=self.secondaries,
4224 status=self.instance_status,
4225 os_type=self.op.os_type,
4226 memory=self.be_full[constants.BE_MEMORY],
4227 vcpus=self.be_full[constants.BE_VCPUS],
4228 nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
4231 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4236 def CheckPrereq(self):
4237 """Check prerequisites.
4240 if (not self.cfg.GetVGName() and
4241 self.op.disk_template not in constants.DTS_NOT_LVM):
4242 raise errors.OpPrereqError("Cluster does not support lvm-based"
4246 if self.op.mode == constants.INSTANCE_IMPORT:
4247 src_node = self.op.src_node
4248 src_path = self.op.src_path
4250 if src_node is None:
4251 exp_list = self.rpc.call_export_list(
4252 self.acquired_locks[locking.LEVEL_NODE])
4254 for node in exp_list:
4255 if not exp_list[node].failed and src_path in exp_list[node].data:
4257 self.op.src_node = src_node = node
4258 self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4262 raise errors.OpPrereqError("No export found for relative path %s" %
4265 _CheckNodeOnline(self, src_node)
4266 result = self.rpc.call_export_info(src_node, src_path)
4269 raise errors.OpPrereqError("No export found in dir %s" % src_path)
4271 export_info = result.data
4272 if not export_info.has_section(constants.INISECT_EXP):
4273 raise errors.ProgrammerError("Corrupted export config")
4275 ei_version = export_info.get(constants.INISECT_EXP, 'version')
4276 if (int(ei_version) != constants.EXPORT_VERSION):
4277 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4278 (ei_version, constants.EXPORT_VERSION))
4280 # Check that the new instance doesn't have less disks than the export
4281 instance_disks = len(self.disks)
4282 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4283 if instance_disks < export_disks:
4284 raise errors.OpPrereqError("Not enough disks to import."
4285 " (instance: %d, export: %d)" %
4286 (instance_disks, export_disks))
4288 self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4290 for idx in range(export_disks):
4291 option = 'disk%d_dump' % idx
4292 if export_info.has_option(constants.INISECT_INS, option):
4293 # FIXME: are the old os-es, disk sizes, etc. useful?
4294 export_name = export_info.get(constants.INISECT_INS, option)
4295 image = os.path.join(src_path, export_name)
4296 disk_images.append(image)
4298 disk_images.append(False)
4300 self.src_images = disk_images
4302 old_name = export_info.get(constants.INISECT_INS, 'name')
4303 # FIXME: int() here could throw a ValueError on broken exports
4304 exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4305 if self.op.instance_name == old_name:
4306 for idx, nic in enumerate(self.nics):
4307 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4308 nic_mac_ini = 'nic%d_mac' % idx
4309 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4311 # ip ping checks (we use the same ip that was resolved in ExpandNames)
4312 if self.op.start and not self.op.ip_check:
4313 raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4314 " adding an instance in start mode")
4316 if self.op.ip_check:
4317 if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4318 raise errors.OpPrereqError("IP %s of instance %s already in use" %
4319 (self.check_ip, self.op.instance_name))
4323 if self.op.iallocator is not None:
4324 self._RunAllocator()
4326 #### node related checks
4328 # check primary node
4329 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4330 assert self.pnode is not None, \
4331 "Cannot retrieve locked node %s" % self.op.pnode
4333 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4336 self.secondaries = []
4338 # mirror node verification
4339 if self.op.disk_template in constants.DTS_NET_MIRROR:
4340 if self.op.snode is None:
4341 raise errors.OpPrereqError("The networked disk templates need"
4343 if self.op.snode == pnode.name:
4344 raise errors.OpPrereqError("The secondary node cannot be"
4345 " the primary node.")
4346 self.secondaries.append(self.op.snode)
4347 _CheckNodeOnline(self, self.op.snode)
4349 nodenames = [pnode.name] + self.secondaries
4351 req_size = _ComputeDiskSize(self.op.disk_template,
4354 # Check lv size requirements
4355 if req_size is not None:
4356 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4358 for node in nodenames:
4359 info = nodeinfo[node]
4363 raise errors.OpPrereqError("Cannot get current information"
4364 " from node '%s'" % node)
4365 vg_free = info.get('vg_free', None)
4366 if not isinstance(vg_free, int):
4367 raise errors.OpPrereqError("Can't compute free disk space on"
4369 if req_size > info['vg_free']:
4370 raise errors.OpPrereqError("Not enough disk space on target node %s."
4371 " %d MB available, %d MB required" %
4372 (node, info['vg_free'], req_size))
4374 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4377 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4379 if not isinstance(result.data, objects.OS):
4380 raise errors.OpPrereqError("OS '%s' not in supported os list for"
4381 " primary node" % self.op.os_type)
4383 # bridge check on primary node
4384 bridges = [n.bridge for n in self.nics]
4385 result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
4388 raise errors.OpPrereqError("One of the target bridges '%s' does not"
4389 " exist on destination node '%s'" %
4390 (",".join(bridges), pnode.name))
4392 # memory check on primary node
4394 _CheckNodeFreeMemory(self, self.pnode.name,
4395 "creating instance %s" % self.op.instance_name,
4396 self.be_full[constants.BE_MEMORY],
4400 self.instance_status = 'up'
4402 self.instance_status = 'down'
4404 def Exec(self, feedback_fn):
4405 """Create and add the instance to the cluster.
4408 instance = self.op.instance_name
4409 pnode_name = self.pnode.name
4411 for nic in self.nics:
4412 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4413 nic.mac = self.cfg.GenerateMAC()
4415 ht_kind = self.op.hypervisor
4416 if ht_kind in constants.HTS_REQ_PORT:
4417 network_port = self.cfg.AllocatePort()
4421 ##if self.op.vnc_bind_address is None:
4422 ## self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4424 # this is needed because os.path.join does not accept None arguments
4425 if self.op.file_storage_dir is None:
4426 string_file_storage_dir = ""
4428 string_file_storage_dir = self.op.file_storage_dir
4430 # build the full file storage dir path
4431 file_storage_dir = os.path.normpath(os.path.join(
4432 self.cfg.GetFileStorageDir(),
4433 string_file_storage_dir, instance))
4436 disks = _GenerateDiskTemplate(self,
4437 self.op.disk_template,
4438 instance, pnode_name,
4442 self.op.file_driver,
4445 iobj = objects.Instance(name=instance, os=self.op.os_type,
4446 primary_node=pnode_name,
4447 nics=self.nics, disks=disks,
4448 disk_template=self.op.disk_template,
4449 status=self.instance_status,
4450 network_port=network_port,
4451 beparams=self.op.beparams,
4452 hvparams=self.op.hvparams,
4453 hypervisor=self.op.hypervisor,
4456 feedback_fn("* creating instance disks...")
4458 _CreateDisks(self, iobj)
4459 except errors.OpExecError:
4460 self.LogWarning("Device creation failed, reverting...")
4462 _RemoveDisks(self, iobj)
4464 self.cfg.ReleaseDRBDMinors(instance)
4467 feedback_fn("adding instance %s to cluster config" % instance)
4469 self.cfg.AddInstance(iobj)
4470 # Declare that we don't want to remove the instance lock anymore, as we've
4471 # added the instance to the config
4472 del self.remove_locks[locking.LEVEL_INSTANCE]
4473 # Remove the temp. assignements for the instance's drbds
4474 self.cfg.ReleaseDRBDMinors(instance)
4475 # Unlock all the nodes
4476 if self.op.mode == constants.INSTANCE_IMPORT:
4477 nodes_keep = [self.op.src_node]
4478 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4479 if node != self.op.src_node]
4480 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4481 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4483 self.context.glm.release(locking.LEVEL_NODE)
4484 del self.acquired_locks[locking.LEVEL_NODE]
4486 if self.op.wait_for_sync:
4487 disk_abort = not _WaitForSync(self, iobj)
4488 elif iobj.disk_template in constants.DTS_NET_MIRROR:
4489 # make sure the disks are not degraded (still sync-ing is ok)
4491 feedback_fn("* checking mirrors status")
4492 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4497 _RemoveDisks(self, iobj)
4498 self.cfg.RemoveInstance(iobj.name)
4499 # Make sure the instance lock gets removed
4500 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4501 raise errors.OpExecError("There are some degraded disks for"
4504 feedback_fn("creating os for instance %s on node %s" %
4505 (instance, pnode_name))
4507 if iobj.disk_template != constants.DT_DISKLESS:
4508 if self.op.mode == constants.INSTANCE_CREATE:
4509 feedback_fn("* running the instance OS create scripts...")
4510 result = self.rpc.call_instance_os_add(pnode_name, iobj)
4513 raise errors.OpExecError("Could not add os for instance %s"
4515 (instance, pnode_name))
4517 elif self.op.mode == constants.INSTANCE_IMPORT:
4518 feedback_fn("* running the instance OS import scripts...")
4519 src_node = self.op.src_node
4520 src_images = self.src_images
4521 cluster_name = self.cfg.GetClusterName()
4522 import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4523 src_node, src_images,
4525 import_result.Raise()
4526 for idx, result in enumerate(import_result.data):
4528 self.LogWarning("Could not import the image %s for instance"
4529 " %s, disk %d, on node %s" %
4530 (src_images[idx], instance, idx, pnode_name))
4532 # also checked in the prereq part
4533 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4537 logging.info("Starting instance %s on node %s", instance, pnode_name)
4538 feedback_fn("* starting instance...")
4539 result = self.rpc.call_instance_start(pnode_name, iobj, None)
4542 raise errors.OpExecError("Could not start instance")
4545 class LUConnectConsole(NoHooksLU):
4546 """Connect to an instance's console.
4548 This is somewhat special in that it returns the command line that
4549 you need to run on the master node in order to connect to the
4553 _OP_REQP = ["instance_name"]
4556 def ExpandNames(self):
4557 self._ExpandAndLockInstance()
4559 def CheckPrereq(self):
4560 """Check prerequisites.
4562 This checks that the instance is in the cluster.
4565 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4566 assert self.instance is not None, \
4567 "Cannot retrieve locked instance %s" % self.op.instance_name
4568 _CheckNodeOnline(self, self.instance.primary_node)
4570 def Exec(self, feedback_fn):
4571 """Connect to the console of an instance
4574 instance = self.instance
4575 node = instance.primary_node
4577 node_insts = self.rpc.call_instance_list([node],
4578 [instance.hypervisor])[node]
4581 if instance.name not in node_insts.data:
4582 raise errors.OpExecError("Instance %s is not running." % instance.name)
4584 logging.debug("Connecting to console of %s on %s", instance.name, node)
4586 hyper = hypervisor.GetHypervisor(instance.hypervisor)
4587 console_cmd = hyper.GetShellCommandForConsole(instance)
4590 return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4593 class LUReplaceDisks(LogicalUnit):
4594 """Replace the disks of an instance.
4597 HPATH = "mirrors-replace"
4598 HTYPE = constants.HTYPE_INSTANCE
4599 _OP_REQP = ["instance_name", "mode", "disks"]
4602 def CheckArguments(self):
4603 if not hasattr(self.op, "remote_node"):
4604 self.op.remote_node = None
4605 if not hasattr(self.op, "iallocator"):
4606 self.op.iallocator = None
4608 # check for valid parameter combination
4609 cnt = [self.op.remote_node, self.op.iallocator].count(None)
4610 if self.op.mode == constants.REPLACE_DISK_CHG:
4612 raise errors.OpPrereqError("When changing the secondary either an"
4613 " iallocator script must be used or the"
4616 raise errors.OpPrereqError("Give either the iallocator or the new"
4617 " secondary, not both")
4618 else: # not replacing the secondary
4620 raise errors.OpPrereqError("The iallocator and new node options can"
4621 " be used only when changing the"
4624 def ExpandNames(self):
4625 self._ExpandAndLockInstance()
4627 if self.op.iallocator is not None:
4628 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4629 elif self.op.remote_node is not None:
4630 remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4631 if remote_node is None:
4632 raise errors.OpPrereqError("Node '%s' not known" %
4633 self.op.remote_node)
4634 self.op.remote_node = remote_node
4635 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4636 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4638 self.needed_locks[locking.LEVEL_NODE] = []
4639 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4641 def DeclareLocks(self, level):
4642 # If we're not already locking all nodes in the set we have to declare the
4643 # instance's primary/secondary nodes.
4644 if (level == locking.LEVEL_NODE and
4645 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4646 self._LockInstancesNodes()
4648 def _RunAllocator(self):
4649 """Compute a new secondary node using an IAllocator.
4652 ial = IAllocator(self,
4653 mode=constants.IALLOCATOR_MODE_RELOC,
4654 name=self.op.instance_name,
4655 relocate_from=[self.sec_node])
4657 ial.Run(self.op.iallocator)
4660 raise errors.OpPrereqError("Can't compute nodes using"
4661 " iallocator '%s': %s" % (self.op.iallocator,
4663 if len(ial.nodes) != ial.required_nodes:
4664 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4665 " of nodes (%s), required %s" %
4666 (len(ial.nodes), ial.required_nodes))
4667 self.op.remote_node = ial.nodes[0]
4668 self.LogInfo("Selected new secondary for the instance: %s",
4669 self.op.remote_node)
4671 def BuildHooksEnv(self):
4674 This runs on the master, the primary and all the secondaries.
4678 "MODE": self.op.mode,
4679 "NEW_SECONDARY": self.op.remote_node,
4680 "OLD_SECONDARY": self.instance.secondary_nodes[0],
4682 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4684 self.cfg.GetMasterNode(),
4685 self.instance.primary_node,
4687 if self.op.remote_node is not None:
4688 nl.append(self.op.remote_node)
4691 def CheckPrereq(self):
4692 """Check prerequisites.
4694 This checks that the instance is in the cluster.
4697 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4698 assert instance is not None, \
4699 "Cannot retrieve locked instance %s" % self.op.instance_name
4700 self.instance = instance
4702 if instance.disk_template != constants.DT_DRBD8:
4703 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
4706 if len(instance.secondary_nodes) != 1:
4707 raise errors.OpPrereqError("The instance has a strange layout,"
4708 " expected one secondary but found %d" %
4709 len(instance.secondary_nodes))
4711 self.sec_node = instance.secondary_nodes[0]
4713 if self.op.iallocator is not None:
4714 self._RunAllocator()
4716 remote_node = self.op.remote_node
4717 if remote_node is not None:
4718 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4719 assert self.remote_node_info is not None, \
4720 "Cannot retrieve locked node %s" % remote_node
4722 self.remote_node_info = None
4723 if remote_node == instance.primary_node:
4724 raise errors.OpPrereqError("The specified node is the primary node of"
4726 elif remote_node == self.sec_node:
4727 raise errors.OpPrereqError("The specified node is already the"
4728 " secondary node of the instance.")
4730 if self.op.mode == constants.REPLACE_DISK_PRI:
4731 n1 = self.tgt_node = instance.primary_node
4732 n2 = self.oth_node = self.sec_node
4733 elif self.op.mode == constants.REPLACE_DISK_SEC:
4734 n1 = self.tgt_node = self.sec_node
4735 n2 = self.oth_node = instance.primary_node
4736 elif self.op.mode == constants.REPLACE_DISK_CHG:
4737 n1 = self.new_node = remote_node
4738 n2 = self.oth_node = instance.primary_node
4739 self.tgt_node = self.sec_node
4741 raise errors.ProgrammerError("Unhandled disk replace mode")
4743 _CheckNodeOnline(self, n1)
4744 _CheckNodeOnline(self, n2)
4746 if not self.op.disks:
4747 self.op.disks = range(len(instance.disks))
4749 for disk_idx in self.op.disks:
4750 instance.FindDisk(disk_idx)
4752 def _ExecD8DiskOnly(self, feedback_fn):
4753 """Replace a disk on the primary or secondary for dbrd8.
4755 The algorithm for replace is quite complicated:
4757 1. for each disk to be replaced:
4759 1. create new LVs on the target node with unique names
4760 1. detach old LVs from the drbd device
4761 1. rename old LVs to name_replaced.<time_t>
4762 1. rename new LVs to old LVs
4763 1. attach the new LVs (with the old names now) to the drbd device
4765 1. wait for sync across all devices
4767 1. for each modified disk:
4769 1. remove old LVs (which have the name name_replaces.<time_t>)
4771 Failures are not very well handled.
4775 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4776 instance = self.instance
4778 vgname = self.cfg.GetVGName()
4781 tgt_node = self.tgt_node
4782 oth_node = self.oth_node
4784 # Step: check device activation
4785 self.proc.LogStep(1, steps_total, "check device existence")
4786 info("checking volume groups")
4787 my_vg = cfg.GetVGName()
4788 results = self.rpc.call_vg_list([oth_node, tgt_node])
4790 raise errors.OpExecError("Can't list volume groups on the nodes")
4791 for node in oth_node, tgt_node:
4793 if res.failed or not res.data or my_vg not in res.data:
4794 raise errors.OpExecError("Volume group '%s' not found on %s" %
4796 for idx, dev in enumerate(instance.disks):
4797 if idx not in self.op.disks:
4799 for node in tgt_node, oth_node:
4800 info("checking disk/%d on %s" % (idx, node))
4801 cfg.SetDiskID(dev, node)
4802 if not self.rpc.call_blockdev_find(node, dev):
4803 raise errors.OpExecError("Can't find disk/%d on node %s" %
4806 # Step: check other node consistency
4807 self.proc.LogStep(2, steps_total, "check peer consistency")
4808 for idx, dev in enumerate(instance.disks):
4809 if idx not in self.op.disks:
4811 info("checking disk/%d consistency on %s" % (idx, oth_node))
4812 if not _CheckDiskConsistency(self, dev, oth_node,
4813 oth_node==instance.primary_node):
4814 raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4815 " to replace disks on this node (%s)" %
4816 (oth_node, tgt_node))
4818 # Step: create new storage
4819 self.proc.LogStep(3, steps_total, "allocate new storage")
4820 for idx, dev in enumerate(instance.disks):
4821 if idx not in self.op.disks:
4824 cfg.SetDiskID(dev, tgt_node)
4825 lv_names = [".disk%d_%s" % (idx, suf)
4826 for suf in ["data", "meta"]]
4827 names = _GenerateUniqueNames(self, lv_names)
4828 lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4829 logical_id=(vgname, names[0]))
4830 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4831 logical_id=(vgname, names[1]))
4832 new_lvs = [lv_data, lv_meta]
4833 old_lvs = dev.children
4834 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4835 info("creating new local storage on %s for %s" %
4836 (tgt_node, dev.iv_name))
4837 # we pass force_create=True to force the LVM creation
4838 for new_lv in new_lvs:
4839 _CreateBlockDev(self, tgt_node, instance, new_lv, True,
4840 _GetInstanceInfoText(instance), False)
4842 # Step: for each lv, detach+rename*2+attach
4843 self.proc.LogStep(4, steps_total, "change drbd configuration")
4844 for dev, old_lvs, new_lvs in iv_names.itervalues():
4845 info("detaching %s drbd from local storage" % dev.iv_name)
4846 result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
4849 raise errors.OpExecError("Can't detach drbd from local storage on node"
4850 " %s for device %s" % (tgt_node, dev.iv_name))
4852 #cfg.Update(instance)
4854 # ok, we created the new LVs, so now we know we have the needed
4855 # storage; as such, we proceed on the target node to rename
4856 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
4857 # using the assumption that logical_id == physical_id (which in
4858 # turn is the unique_id on that node)
4860 # FIXME(iustin): use a better name for the replaced LVs
4861 temp_suffix = int(time.time())
4862 ren_fn = lambda d, suff: (d.physical_id[0],
4863 d.physical_id[1] + "_replaced-%s" % suff)
4864 # build the rename list based on what LVs exist on the node
4866 for to_ren in old_lvs:
4867 find_res = self.rpc.call_blockdev_find(tgt_node, to_ren)
4868 if not find_res.failed and find_res.data is not None: # device exists
4869 rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
4871 info("renaming the old LVs on the target node")
4872 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
4875 raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
4876 # now we rename the new LVs to the old LVs
4877 info("renaming the new LVs on the target node")
4878 rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
4879 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
4882 raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
4884 for old, new in zip(old_lvs, new_lvs):
4885 new.logical_id = old.logical_id
4886 cfg.SetDiskID(new, tgt_node)
4888 for disk in old_lvs:
4889 disk.logical_id = ren_fn(disk, temp_suffix)
4890 cfg.SetDiskID(disk, tgt_node)
4892 # now that the new lvs have the old name, we can add them to the device
4893 info("adding new mirror component on %s" % tgt_node)
4894 result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
4895 if result.failed or not result.data:
4896 for new_lv in new_lvs:
4897 result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
4898 if result.failed or not result.data:
4899 warning("Can't rollback device %s", hint="manually cleanup unused"
4901 raise errors.OpExecError("Can't add local storage to drbd")
4903 dev.children = new_lvs
4904 cfg.Update(instance)
4906 # Step: wait for sync
4908 # this can fail as the old devices are degraded and _WaitForSync
4909 # does a combined result over all disks, so we don't check its
4911 self.proc.LogStep(5, steps_total, "sync devices")
4912 _WaitForSync(self, instance, unlock=True)
4914 # so check manually all the devices
4915 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
4916 cfg.SetDiskID(dev, instance.primary_node)
4917 result = self.rpc.call_blockdev_find(instance.primary_node, dev)
4918 if result.failed or result.data[5]:
4919 raise errors.OpExecError("DRBD device %s is degraded!" % name)
4921 # Step: remove old storage
4922 self.proc.LogStep(6, steps_total, "removing old storage")
4923 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
4924 info("remove logical volumes for %s" % name)
4926 cfg.SetDiskID(lv, tgt_node)
4927 result = self.rpc.call_blockdev_remove(tgt_node, lv)
4928 if result.failed or not result.data:
4929 warning("Can't remove old LV", hint="manually remove unused LVs")
4932 def _ExecD8Secondary(self, feedback_fn):
4933 """Replace the secondary node for drbd8.
4935 The algorithm for replace is quite complicated:
4936 - for all disks of the instance:
4937 - create new LVs on the new node with same names
4938 - shutdown the drbd device on the old secondary
4939 - disconnect the drbd network on the primary
4940 - create the drbd device on the new secondary
4941 - network attach the drbd on the primary, using an artifice:
4942 the drbd code for Attach() will connect to the network if it
4943 finds a device which is connected to the good local disks but
4945 - wait for sync across all devices
4946 - remove all disks from the old secondary
4948 Failures are not very well handled.
4952 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4953 instance = self.instance
4957 old_node = self.tgt_node
4958 new_node = self.new_node
4959 pri_node = instance.primary_node
4961 old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
4962 new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
4963 pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
4966 # Step: check device activation
4967 self.proc.LogStep(1, steps_total, "check device existence")
4968 info("checking volume groups")
4969 my_vg = cfg.GetVGName()
4970 results = self.rpc.call_vg_list([pri_node, new_node])
4971 for node in pri_node, new_node:
4973 if res.failed or not res.data or my_vg not in res.data:
4974 raise errors.OpExecError("Volume group '%s' not found on %s" %
4976 for idx, dev in enumerate(instance.disks):
4977 if idx not in self.op.disks:
4979 info("checking disk/%d on %s" % (idx, pri_node))
4980 cfg.SetDiskID(dev, pri_node)
4981 result = self.rpc.call_blockdev_find(pri_node, dev)
4984 raise errors.OpExecError("Can't find disk/%d on node %s" %
4987 # Step: check other node consistency
4988 self.proc.LogStep(2, steps_total, "check peer consistency")
4989 for idx, dev in enumerate(instance.disks):
4990 if idx not in self.op.disks:
4992 info("checking disk/%d consistency on %s" % (idx, pri_node))
4993 if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
4994 raise errors.OpExecError("Primary node (%s) has degraded storage,"
4995 " unsafe to replace the secondary" %
4998 # Step: create new storage
4999 self.proc.LogStep(3, steps_total, "allocate new storage")
5000 for idx, dev in enumerate(instance.disks):
5001 info("adding new local storage on %s for disk/%d" %
5003 # we pass force_create=True to force LVM creation
5004 for new_lv in dev.children:
5005 _CreateBlockDev(self, new_node, instance, new_lv, True,
5006 _GetInstanceInfoText(instance), False)
5008 # Step 4: dbrd minors and drbd setups changes
5009 # after this, we must manually remove the drbd minors on both the
5010 # error and the success paths
5011 minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5013 logging.debug("Allocated minors %s" % (minors,))
5014 self.proc.LogStep(4, steps_total, "changing drbd configuration")
5015 for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5017 info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5018 # create new devices on new_node; note that we create two IDs:
5019 # one without port, so the drbd will be activated without
5020 # networking information on the new node at this stage, and one
5021 # with network, for the latter activation in step 4
5022 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5023 if pri_node == o_node1:
5028 new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5029 new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5031 iv_names[idx] = (dev, dev.children, new_net_id)
5032 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5034 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5035 logical_id=new_alone_id,
5036 children=dev.children)
5038 _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5039 _GetInstanceInfoText(instance), False)
5040 except error.BlockDeviceError:
5041 self.cfg.ReleaseDRBDMinors(instance.name)
5044 for idx, dev in enumerate(instance.disks):
5045 # we have new devices, shutdown the drbd on the old secondary
5046 info("shutting down drbd for disk/%d on old node" % idx)
5047 cfg.SetDiskID(dev, old_node)
5048 result = self.rpc.call_blockdev_shutdown(old_node, dev)
5049 if result.failed or not result.data:
5050 warning("Failed to shutdown drbd for disk/%d on old node" % idx,
5051 hint="Please cleanup this device manually as soon as possible")
5053 info("detaching primary drbds from the network (=> standalone)")
5054 result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5055 instance.disks)[pri_node]
5057 msg = result.RemoteFailMsg()
5059 # detaches didn't succeed (unlikely)
5060 self.cfg.ReleaseDRBDMinors(instance.name)
5061 raise errors.OpExecError("Can't detach the disks from the network on"
5062 " old node: %s" % (msg,))
5064 # if we managed to detach at least one, we update all the disks of
5065 # the instance to point to the new secondary
5066 info("updating instance configuration")
5067 for dev, _, new_logical_id in iv_names.itervalues():
5068 dev.logical_id = new_logical_id
5069 cfg.SetDiskID(dev, pri_node)
5070 cfg.Update(instance)
5071 # we can remove now the temp minors as now the new values are
5072 # written to the config file (and therefore stable)
5073 self.cfg.ReleaseDRBDMinors(instance.name)
5075 # and now perform the drbd attach
5076 info("attaching primary drbds to new secondary (standalone => connected)")
5077 result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5078 instance.disks, instance.name,
5080 for to_node, to_result in result.items():
5081 msg = to_result.RemoteFailMsg()
5083 warning("can't attach drbd disks on node %s: %s", to_node, msg,
5084 hint="please do a gnt-instance info to see the"
5087 # this can fail as the old devices are degraded and _WaitForSync
5088 # does a combined result over all disks, so we don't check its
5090 self.proc.LogStep(5, steps_total, "sync devices")
5091 _WaitForSync(self, instance, unlock=True)
5093 # so check manually all the devices
5094 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5095 cfg.SetDiskID(dev, pri_node)
5096 result = self.rpc.call_blockdev_find(pri_node, dev)
5099 raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5101 self.proc.LogStep(6, steps_total, "removing old storage")
5102 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5103 info("remove logical volumes for disk/%d" % idx)
5105 cfg.SetDiskID(lv, old_node)
5106 result = self.rpc.call_blockdev_remove(old_node, lv)
5107 if result.failed or not result.data:
5108 warning("Can't remove LV on old secondary",
5109 hint="Cleanup stale volumes by hand")
5111 def Exec(self, feedback_fn):
5112 """Execute disk replacement.
5114 This dispatches the disk replacement to the appropriate handler.
5117 instance = self.instance
5119 # Activate the instance disks if we're replacing them on a down instance
5120 if instance.status == "down":
5121 _StartInstanceDisks(self, instance, True)
5123 if self.op.mode == constants.REPLACE_DISK_CHG:
5124 fn = self._ExecD8Secondary
5126 fn = self._ExecD8DiskOnly
5128 ret = fn(feedback_fn)
5130 # Deactivate the instance disks if we're replacing them on a down instance
5131 if instance.status == "down":
5132 _SafeShutdownInstanceDisks(self, instance)
5137 class LUGrowDisk(LogicalUnit):
5138 """Grow a disk of an instance.
5142 HTYPE = constants.HTYPE_INSTANCE
5143 _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5146 def ExpandNames(self):
5147 self._ExpandAndLockInstance()
5148 self.needed_locks[locking.LEVEL_NODE] = []
5149 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5151 def DeclareLocks(self, level):
5152 if level == locking.LEVEL_NODE:
5153 self._LockInstancesNodes()
5155 def BuildHooksEnv(self):
5158 This runs on the master, the primary and all the secondaries.
5162 "DISK": self.op.disk,
5163 "AMOUNT": self.op.amount,
5165 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5167 self.cfg.GetMasterNode(),
5168 self.instance.primary_node,
5172 def CheckPrereq(self):
5173 """Check prerequisites.
5175 This checks that the instance is in the cluster.
5178 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5179 assert instance is not None, \
5180 "Cannot retrieve locked instance %s" % self.op.instance_name
5181 nodenames = list(instance.all_nodes)
5182 for node in nodenames:
5183 _CheckNodeOnline(self, node)
5186 self.instance = instance
5188 if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5189 raise errors.OpPrereqError("Instance's disk layout does not support"
5192 self.disk = instance.FindDisk(self.op.disk)
5194 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5195 instance.hypervisor)
5196 for node in nodenames:
5197 info = nodeinfo[node]
5198 if info.failed or not info.data:
5199 raise errors.OpPrereqError("Cannot get current information"
5200 " from node '%s'" % node)
5201 vg_free = info.data.get('vg_free', None)
5202 if not isinstance(vg_free, int):
5203 raise errors.OpPrereqError("Can't compute free disk space on"
5205 if self.op.amount > vg_free:
5206 raise errors.OpPrereqError("Not enough disk space on target node %s:"
5207 " %d MiB available, %d MiB required" %
5208 (node, vg_free, self.op.amount))
5210 def Exec(self, feedback_fn):
5211 """Execute disk grow.
5214 instance = self.instance
5216 for node in instance.all_nodes:
5217 self.cfg.SetDiskID(disk, node)
5218 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5220 if (not result.data or not isinstance(result.data, (list, tuple)) or
5221 len(result.data) != 2):
5222 raise errors.OpExecError("Grow request failed to node %s" % node)
5223 elif not result.data[0]:
5224 raise errors.OpExecError("Grow request failed to node %s: %s" %
5225 (node, result.data[1]))
5226 disk.RecordGrow(self.op.amount)
5227 self.cfg.Update(instance)
5228 if self.op.wait_for_sync:
5229 disk_abort = not _WaitForSync(self, instance)
5231 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5232 " status.\nPlease check the instance.")
5235 class LUQueryInstanceData(NoHooksLU):
5236 """Query runtime instance data.
5239 _OP_REQP = ["instances", "static"]
5242 def ExpandNames(self):
5243 self.needed_locks = {}
5244 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5246 if not isinstance(self.op.instances, list):
5247 raise errors.OpPrereqError("Invalid argument type 'instances'")
5249 if self.op.instances:
5250 self.wanted_names = []
5251 for name in self.op.instances:
5252 full_name = self.cfg.ExpandInstanceName(name)
5253 if full_name is None:
5254 raise errors.OpPrereqError("Instance '%s' not known" % name)
5255 self.wanted_names.append(full_name)
5256 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5258 self.wanted_names = None
5259 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5261 self.needed_locks[locking.LEVEL_NODE] = []
5262 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5264 def DeclareLocks(self, level):
5265 if level == locking.LEVEL_NODE:
5266 self._LockInstancesNodes()
5268 def CheckPrereq(self):
5269 """Check prerequisites.
5271 This only checks the optional instance list against the existing names.
5274 if self.wanted_names is None:
5275 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5277 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5278 in self.wanted_names]
5281 def _ComputeDiskStatus(self, instance, snode, dev):
5282 """Compute block device status.
5285 static = self.op.static
5287 self.cfg.SetDiskID(dev, instance.primary_node)
5288 dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5290 dev_pstatus = dev_pstatus.data
5294 if dev.dev_type in constants.LDS_DRBD:
5295 # we change the snode then (otherwise we use the one passed in)
5296 if dev.logical_id[0] == instance.primary_node:
5297 snode = dev.logical_id[1]
5299 snode = dev.logical_id[0]
5301 if snode and not static:
5302 self.cfg.SetDiskID(dev, snode)
5303 dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5305 dev_sstatus = dev_sstatus.data
5310 dev_children = [self._ComputeDiskStatus(instance, snode, child)
5311 for child in dev.children]
5316 "iv_name": dev.iv_name,
5317 "dev_type": dev.dev_type,
5318 "logical_id": dev.logical_id,
5319 "physical_id": dev.physical_id,
5320 "pstatus": dev_pstatus,
5321 "sstatus": dev_sstatus,
5322 "children": dev_children,
5328 def Exec(self, feedback_fn):
5329 """Gather and return data"""
5332 cluster = self.cfg.GetClusterInfo()
5334 for instance in self.wanted_instances:
5335 if not self.op.static:
5336 remote_info = self.rpc.call_instance_info(instance.primary_node,
5338 instance.hypervisor)
5340 remote_info = remote_info.data
5341 if remote_info and "state" in remote_info:
5344 remote_state = "down"
5347 if instance.status == "down":
5348 config_state = "down"
5352 disks = [self._ComputeDiskStatus(instance, None, device)
5353 for device in instance.disks]
5356 "name": instance.name,
5357 "config_state": config_state,
5358 "run_state": remote_state,
5359 "pnode": instance.primary_node,
5360 "snodes": instance.secondary_nodes,
5362 "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5364 "hypervisor": instance.hypervisor,
5365 "network_port": instance.network_port,
5366 "hv_instance": instance.hvparams,
5367 "hv_actual": cluster.FillHV(instance),
5368 "be_instance": instance.beparams,
5369 "be_actual": cluster.FillBE(instance),
5372 result[instance.name] = idict
5377 class LUSetInstanceParams(LogicalUnit):
5378 """Modifies an instances's parameters.
5381 HPATH = "instance-modify"
5382 HTYPE = constants.HTYPE_INSTANCE
5383 _OP_REQP = ["instance_name"]
5386 def CheckArguments(self):
5387 if not hasattr(self.op, 'nics'):
5389 if not hasattr(self.op, 'disks'):
5391 if not hasattr(self.op, 'beparams'):
5392 self.op.beparams = {}
5393 if not hasattr(self.op, 'hvparams'):
5394 self.op.hvparams = {}
5395 self.op.force = getattr(self.op, "force", False)
5396 if not (self.op.nics or self.op.disks or
5397 self.op.hvparams or self.op.beparams):
5398 raise errors.OpPrereqError("No changes submitted")
5400 utils.CheckBEParams(self.op.beparams)
5404 for disk_op, disk_dict in self.op.disks:
5405 if disk_op == constants.DDM_REMOVE:
5408 elif disk_op == constants.DDM_ADD:
5411 if not isinstance(disk_op, int):
5412 raise errors.OpPrereqError("Invalid disk index")
5413 if disk_op == constants.DDM_ADD:
5414 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5415 if mode not in (constants.DISK_RDONLY, constants.DISK_RDWR):
5416 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5417 size = disk_dict.get('size', None)
5419 raise errors.OpPrereqError("Required disk parameter size missing")
5422 except ValueError, err:
5423 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5425 disk_dict['size'] = size
5427 # modification of disk
5428 if 'size' in disk_dict:
5429 raise errors.OpPrereqError("Disk size change not possible, use"
5432 if disk_addremove > 1:
5433 raise errors.OpPrereqError("Only one disk add or remove operation"
5434 " supported at a time")
5438 for nic_op, nic_dict in self.op.nics:
5439 if nic_op == constants.DDM_REMOVE:
5442 elif nic_op == constants.DDM_ADD:
5445 if not isinstance(nic_op, int):
5446 raise errors.OpPrereqError("Invalid nic index")
5448 # nic_dict should be a dict
5449 nic_ip = nic_dict.get('ip', None)
5450 if nic_ip is not None:
5451 if nic_ip.lower() == "none":
5452 nic_dict['ip'] = None
5454 if not utils.IsValidIP(nic_ip):
5455 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5456 # we can only check None bridges and assign the default one
5457 nic_bridge = nic_dict.get('bridge', None)
5458 if nic_bridge is None:
5459 nic_dict['bridge'] = self.cfg.GetDefBridge()
5460 # but we can validate MACs
5461 nic_mac = nic_dict.get('mac', None)
5462 if nic_mac is not None:
5463 if self.cfg.IsMacInUse(nic_mac):
5464 raise errors.OpPrereqError("MAC address %s already in use"
5465 " in cluster" % nic_mac)
5466 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5467 if not utils.IsValidMac(nic_mac):
5468 raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5469 if nic_addremove > 1:
5470 raise errors.OpPrereqError("Only one NIC add or remove operation"
5471 " supported at a time")
5473 def ExpandNames(self):
5474 self._ExpandAndLockInstance()
5475 self.needed_locks[locking.LEVEL_NODE] = []
5476 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5478 def DeclareLocks(self, level):
5479 if level == locking.LEVEL_NODE:
5480 self._LockInstancesNodes()
5482 def BuildHooksEnv(self):
5485 This runs on the master, primary and secondaries.
5489 if constants.BE_MEMORY in self.be_new:
5490 args['memory'] = self.be_new[constants.BE_MEMORY]
5491 if constants.BE_VCPUS in self.be_new:
5492 args['vcpus'] = self.be_new[constants.BE_VCPUS]
5493 # FIXME: readd disk/nic changes
5494 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5495 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5498 def CheckPrereq(self):
5499 """Check prerequisites.
5501 This only checks the instance list against the existing names.
5504 force = self.force = self.op.force
5506 # checking the new params on the primary/secondary nodes
5508 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5509 assert self.instance is not None, \
5510 "Cannot retrieve locked instance %s" % self.op.instance_name
5511 pnode = instance.primary_node
5512 nodelist = list(instance.all_nodes)
5514 # hvparams processing
5515 if self.op.hvparams:
5516 i_hvdict = copy.deepcopy(instance.hvparams)
5517 for key, val in self.op.hvparams.iteritems():
5518 if val == constants.VALUE_DEFAULT:
5523 elif val == constants.VALUE_NONE:
5524 i_hvdict[key] = None
5527 cluster = self.cfg.GetClusterInfo()
5528 hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5531 hypervisor.GetHypervisor(
5532 instance.hypervisor).CheckParameterSyntax(hv_new)
5533 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5534 self.hv_new = hv_new # the new actual values
5535 self.hv_inst = i_hvdict # the new dict (without defaults)
5537 self.hv_new = self.hv_inst = {}
5539 # beparams processing
5540 if self.op.beparams:
5541 i_bedict = copy.deepcopy(instance.beparams)
5542 for key, val in self.op.beparams.iteritems():
5543 if val == constants.VALUE_DEFAULT:
5550 cluster = self.cfg.GetClusterInfo()
5551 be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5553 self.be_new = be_new # the new actual values
5554 self.be_inst = i_bedict # the new dict (without defaults)
5556 self.be_new = self.be_inst = {}
5560 if constants.BE_MEMORY in self.op.beparams and not self.force:
5561 mem_check_list = [pnode]
5562 if be_new[constants.BE_AUTO_BALANCE]:
5563 # either we changed auto_balance to yes or it was from before
5564 mem_check_list.extend(instance.secondary_nodes)
5565 instance_info = self.rpc.call_instance_info(pnode, instance.name,
5566 instance.hypervisor)
5567 nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5568 instance.hypervisor)
5569 if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5570 # Assume the primary node is unreachable and go ahead
5571 self.warn.append("Can't get info from primary node %s" % pnode)
5573 if not instance_info.failed and instance_info.data:
5574 current_mem = instance_info.data['memory']
5576 # Assume instance not running
5577 # (there is a slight race condition here, but it's not very probable,
5578 # and we have no other way to check)
5580 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5581 nodeinfo[pnode].data['memory_free'])
5583 raise errors.OpPrereqError("This change will prevent the instance"
5584 " from starting, due to %d MB of memory"
5585 " missing on its primary node" % miss_mem)
5587 if be_new[constants.BE_AUTO_BALANCE]:
5588 for node, nres in nodeinfo.iteritems():
5589 if node not in instance.secondary_nodes:
5591 if nres.failed or not isinstance(nres.data, dict):
5592 self.warn.append("Can't get info from secondary node %s" % node)
5593 elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5594 self.warn.append("Not enough memory to failover instance to"
5595 " secondary node %s" % node)
5598 for nic_op, nic_dict in self.op.nics:
5599 if nic_op == constants.DDM_REMOVE:
5600 if not instance.nics:
5601 raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5603 if nic_op != constants.DDM_ADD:
5605 if nic_op < 0 or nic_op >= len(instance.nics):
5606 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5608 (nic_op, len(instance.nics)))
5609 nic_bridge = nic_dict.get('bridge', None)
5610 if nic_bridge is not None:
5611 if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5612 msg = ("Bridge '%s' doesn't exist on one of"
5613 " the instance nodes" % nic_bridge)
5615 self.warn.append(msg)
5617 raise errors.OpPrereqError(msg)
5620 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5621 raise errors.OpPrereqError("Disk operations not supported for"
5622 " diskless instances")
5623 for disk_op, disk_dict in self.op.disks:
5624 if disk_op == constants.DDM_REMOVE:
5625 if len(instance.disks) == 1:
5626 raise errors.OpPrereqError("Cannot remove the last disk of"
5628 ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5629 ins_l = ins_l[pnode]
5630 if ins_l.failed or not isinstance(ins_l.data, list):
5631 raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5632 if instance.name in ins_l.data:
5633 raise errors.OpPrereqError("Instance is running, can't remove"
5636 if (disk_op == constants.DDM_ADD and
5637 len(instance.nics) >= constants.MAX_DISKS):
5638 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5639 " add more" % constants.MAX_DISKS)
5640 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5642 if disk_op < 0 or disk_op >= len(instance.disks):
5643 raise errors.OpPrereqError("Invalid disk index %s, valid values"
5645 (disk_op, len(instance.disks)))
5649 def Exec(self, feedback_fn):
5650 """Modifies an instance.
5652 All parameters take effect only at the next restart of the instance.
5655 # Process here the warnings from CheckPrereq, as we don't have a
5656 # feedback_fn there.
5657 for warn in self.warn:
5658 feedback_fn("WARNING: %s" % warn)
5661 instance = self.instance
5663 for disk_op, disk_dict in self.op.disks:
5664 if disk_op == constants.DDM_REMOVE:
5665 # remove the last disk
5666 device = instance.disks.pop()
5667 device_idx = len(instance.disks)
5668 for node, disk in device.ComputeNodeTree(instance.primary_node):
5669 self.cfg.SetDiskID(disk, node)
5670 rpc_result = self.rpc.call_blockdev_remove(node, disk)
5671 if rpc_result.failed or not rpc_result.data:
5672 self.proc.LogWarning("Could not remove disk/%d on node %s,"
5673 " continuing anyway", device_idx, node)
5674 result.append(("disk/%d" % device_idx, "remove"))
5675 elif disk_op == constants.DDM_ADD:
5677 if instance.disk_template == constants.DT_FILE:
5678 file_driver, file_path = instance.disks[0].logical_id
5679 file_path = os.path.dirname(file_path)
5681 file_driver = file_path = None
5682 disk_idx_base = len(instance.disks)
5683 new_disk = _GenerateDiskTemplate(self,
5684 instance.disk_template,
5685 instance, instance.primary_node,
5686 instance.secondary_nodes,
5691 new_disk.mode = disk_dict['mode']
5692 instance.disks.append(new_disk)
5693 info = _GetInstanceInfoText(instance)
5695 logging.info("Creating volume %s for instance %s",
5696 new_disk.iv_name, instance.name)
5697 # Note: this needs to be kept in sync with _CreateDisks
5699 for node in instance.all_nodes:
5700 f_create = node == instance.primary_node
5702 _CreateBlockDev(self, node, instance, new_disk,
5703 f_create, info, f_create)
5704 except error.OpExecError, err:
5705 self.LogWarning("Failed to create volume %s (%s) on"
5707 new_disk.iv_name, new_disk, node, err)
5708 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5709 (new_disk.size, new_disk.mode)))
5711 # change a given disk
5712 instance.disks[disk_op].mode = disk_dict['mode']
5713 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5715 for nic_op, nic_dict in self.op.nics:
5716 if nic_op == constants.DDM_REMOVE:
5717 # remove the last nic
5718 del instance.nics[-1]
5719 result.append(("nic.%d" % len(instance.nics), "remove"))
5720 elif nic_op == constants.DDM_ADD:
5722 if 'mac' not in nic_dict:
5723 mac = constants.VALUE_GENERATE
5725 mac = nic_dict['mac']
5726 if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5727 mac = self.cfg.GenerateMAC()
5728 new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5729 bridge=nic_dict.get('bridge', None))
5730 instance.nics.append(new_nic)
5731 result.append(("nic.%d" % (len(instance.nics) - 1),
5732 "add:mac=%s,ip=%s,bridge=%s" %
5733 (new_nic.mac, new_nic.ip, new_nic.bridge)))
5735 # change a given nic
5736 for key in 'mac', 'ip', 'bridge':
5738 setattr(instance.nics[nic_op], key, nic_dict[key])
5739 result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5742 if self.op.hvparams:
5743 instance.hvparams = self.hv_new
5744 for key, val in self.op.hvparams.iteritems():
5745 result.append(("hv/%s" % key, val))
5748 if self.op.beparams:
5749 instance.beparams = self.be_inst
5750 for key, val in self.op.beparams.iteritems():
5751 result.append(("be/%s" % key, val))
5753 self.cfg.Update(instance)
5758 class LUQueryExports(NoHooksLU):
5759 """Query the exports list
5762 _OP_REQP = ['nodes']
5765 def ExpandNames(self):
5766 self.needed_locks = {}
5767 self.share_locks[locking.LEVEL_NODE] = 1
5768 if not self.op.nodes:
5769 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5771 self.needed_locks[locking.LEVEL_NODE] = \
5772 _GetWantedNodes(self, self.op.nodes)
5774 def CheckPrereq(self):
5775 """Check prerequisites.
5778 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5780 def Exec(self, feedback_fn):
5781 """Compute the list of all the exported system images.
5784 @return: a dictionary with the structure node->(export-list)
5785 where export-list is a list of the instances exported on
5789 rpcresult = self.rpc.call_export_list(self.nodes)
5791 for node in rpcresult:
5792 if rpcresult[node].failed:
5793 result[node] = False
5795 result[node] = rpcresult[node].data
5800 class LUExportInstance(LogicalUnit):
5801 """Export an instance to an image in the cluster.
5804 HPATH = "instance-export"
5805 HTYPE = constants.HTYPE_INSTANCE
5806 _OP_REQP = ["instance_name", "target_node", "shutdown"]
5809 def ExpandNames(self):
5810 self._ExpandAndLockInstance()
5811 # FIXME: lock only instance primary and destination node
5813 # Sad but true, for now we have do lock all nodes, as we don't know where
5814 # the previous export might be, and and in this LU we search for it and
5815 # remove it from its current node. In the future we could fix this by:
5816 # - making a tasklet to search (share-lock all), then create the new one,
5817 # then one to remove, after
5818 # - removing the removal operation altoghether
5819 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5821 def DeclareLocks(self, level):
5822 """Last minute lock declaration."""
5823 # All nodes are locked anyway, so nothing to do here.
5825 def BuildHooksEnv(self):
5828 This will run on the master, primary node and target node.
5832 "EXPORT_NODE": self.op.target_node,
5833 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
5835 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5836 nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
5837 self.op.target_node]
5840 def CheckPrereq(self):
5841 """Check prerequisites.
5843 This checks that the instance and node names are valid.
5846 instance_name = self.op.instance_name
5847 self.instance = self.cfg.GetInstanceInfo(instance_name)
5848 assert self.instance is not None, \
5849 "Cannot retrieve locked instance %s" % self.op.instance_name
5850 _CheckNodeOnline(self, self.instance.primary_node)
5852 self.dst_node = self.cfg.GetNodeInfo(
5853 self.cfg.ExpandNodeName(self.op.target_node))
5855 if self.dst_node is None:
5856 # This is wrong node name, not a non-locked node
5857 raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
5858 _CheckNodeOnline(self, self.dst_node.name)
5860 # instance disk type verification
5861 for disk in self.instance.disks:
5862 if disk.dev_type == constants.LD_FILE:
5863 raise errors.OpPrereqError("Export not supported for instances with"
5864 " file-based disks")
5866 def Exec(self, feedback_fn):
5867 """Export an instance to an image in the cluster.
5870 instance = self.instance
5871 dst_node = self.dst_node
5872 src_node = instance.primary_node
5873 if self.op.shutdown:
5874 # shutdown the instance, but not the disks
5875 result = self.rpc.call_instance_shutdown(src_node, instance)
5878 raise errors.OpExecError("Could not shutdown instance %s on node %s" %
5879 (instance.name, src_node))
5881 vgname = self.cfg.GetVGName()
5885 # set the disks ID correctly since call_instance_start needs the
5886 # correct drbd minor to create the symlinks
5887 for disk in instance.disks:
5888 self.cfg.SetDiskID(disk, src_node)
5891 for disk in instance.disks:
5892 # new_dev_name will be a snapshot of an lvm leaf of the one we passed
5893 new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
5894 if new_dev_name.failed or not new_dev_name.data:
5895 self.LogWarning("Could not snapshot block device %s on node %s",
5896 disk.logical_id[1], src_node)
5897 snap_disks.append(False)
5899 new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
5900 logical_id=(vgname, new_dev_name.data),
5901 physical_id=(vgname, new_dev_name.data),
5902 iv_name=disk.iv_name)
5903 snap_disks.append(new_dev)
5906 if self.op.shutdown and instance.status == "up":
5907 result = self.rpc.call_instance_start(src_node, instance, None)
5908 if result.failed or not result.data:
5909 _ShutdownInstanceDisks(self, instance)
5910 raise errors.OpExecError("Could not start instance")
5912 # TODO: check for size
5914 cluster_name = self.cfg.GetClusterName()
5915 for idx, dev in enumerate(snap_disks):
5917 result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
5918 instance, cluster_name, idx)
5919 if result.failed or not result.data:
5920 self.LogWarning("Could not export block device %s from node %s to"
5921 " node %s", dev.logical_id[1], src_node,
5923 result = self.rpc.call_blockdev_remove(src_node, dev)
5924 if result.failed or not result.data:
5925 self.LogWarning("Could not remove snapshot block device %s from node"
5926 " %s", dev.logical_id[1], src_node)
5928 result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
5929 if result.failed or not result.data:
5930 self.LogWarning("Could not finalize export for instance %s on node %s",
5931 instance.name, dst_node.name)
5933 nodelist = self.cfg.GetNodeList()
5934 nodelist.remove(dst_node.name)
5936 # on one-node clusters nodelist will be empty after the removal
5937 # if we proceed the backup would be removed because OpQueryExports
5938 # substitutes an empty list with the full cluster node list.
5940 exportlist = self.rpc.call_export_list(nodelist)
5941 for node in exportlist:
5942 if exportlist[node].failed:
5944 if instance.name in exportlist[node].data:
5945 if not self.rpc.call_export_remove(node, instance.name):
5946 self.LogWarning("Could not remove older export for instance %s"
5947 " on node %s", instance.name, node)
5950 class LURemoveExport(NoHooksLU):
5951 """Remove exports related to the named instance.
5954 _OP_REQP = ["instance_name"]
5957 def ExpandNames(self):
5958 self.needed_locks = {}
5959 # We need all nodes to be locked in order for RemoveExport to work, but we
5960 # don't need to lock the instance itself, as nothing will happen to it (and
5961 # we can remove exports also for a removed instance)
5962 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5964 def CheckPrereq(self):
5965 """Check prerequisites.
5969 def Exec(self, feedback_fn):
5970 """Remove any export.
5973 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
5974 # If the instance was not found we'll try with the name that was passed in.
5975 # This will only work if it was an FQDN, though.
5977 if not instance_name:
5979 instance_name = self.op.instance_name
5981 exportlist = self.rpc.call_export_list(self.acquired_locks[
5982 locking.LEVEL_NODE])
5984 for node in exportlist:
5985 if exportlist[node].failed:
5986 self.LogWarning("Failed to query node %s, continuing" % node)
5988 if instance_name in exportlist[node].data:
5990 result = self.rpc.call_export_remove(node, instance_name)
5991 if result.failed or not result.data:
5992 logging.error("Could not remove export for instance %s"
5993 " on node %s", instance_name, node)
5995 if fqdn_warn and not found:
5996 feedback_fn("Export not found. If trying to remove an export belonging"
5997 " to a deleted instance please use its Fully Qualified"
6001 class TagsLU(NoHooksLU):
6004 This is an abstract class which is the parent of all the other tags LUs.
6008 def ExpandNames(self):
6009 self.needed_locks = {}
6010 if self.op.kind == constants.TAG_NODE:
6011 name = self.cfg.ExpandNodeName(self.op.name)
6013 raise errors.OpPrereqError("Invalid node name (%s)" %
6016 self.needed_locks[locking.LEVEL_NODE] = name
6017 elif self.op.kind == constants.TAG_INSTANCE:
6018 name = self.cfg.ExpandInstanceName(self.op.name)
6020 raise errors.OpPrereqError("Invalid instance name (%s)" %
6023 self.needed_locks[locking.LEVEL_INSTANCE] = name
6025 def CheckPrereq(self):
6026 """Check prerequisites.
6029 if self.op.kind == constants.TAG_CLUSTER:
6030 self.target = self.cfg.GetClusterInfo()
6031 elif self.op.kind == constants.TAG_NODE:
6032 self.target = self.cfg.GetNodeInfo(self.op.name)
6033 elif self.op.kind == constants.TAG_INSTANCE:
6034 self.target = self.cfg.GetInstanceInfo(self.op.name)
6036 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6040 class LUGetTags(TagsLU):
6041 """Returns the tags of a given object.
6044 _OP_REQP = ["kind", "name"]
6047 def Exec(self, feedback_fn):
6048 """Returns the tag list.
6051 return list(self.target.GetTags())
6054 class LUSearchTags(NoHooksLU):
6055 """Searches the tags for a given pattern.
6058 _OP_REQP = ["pattern"]
6061 def ExpandNames(self):
6062 self.needed_locks = {}
6064 def CheckPrereq(self):
6065 """Check prerequisites.
6067 This checks the pattern passed for validity by compiling it.
6071 self.re = re.compile(self.op.pattern)
6072 except re.error, err:
6073 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6074 (self.op.pattern, err))
6076 def Exec(self, feedback_fn):
6077 """Returns the tag list.
6081 tgts = [("/cluster", cfg.GetClusterInfo())]
6082 ilist = cfg.GetAllInstancesInfo().values()
6083 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6084 nlist = cfg.GetAllNodesInfo().values()
6085 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6087 for path, target in tgts:
6088 for tag in target.GetTags():
6089 if self.re.search(tag):
6090 results.append((path, tag))
6094 class LUAddTags(TagsLU):
6095 """Sets a tag on a given object.
6098 _OP_REQP = ["kind", "name", "tags"]
6101 def CheckPrereq(self):
6102 """Check prerequisites.
6104 This checks the type and length of the tag name and value.
6107 TagsLU.CheckPrereq(self)
6108 for tag in self.op.tags:
6109 objects.TaggableObject.ValidateTag(tag)
6111 def Exec(self, feedback_fn):
6116 for tag in self.op.tags:
6117 self.target.AddTag(tag)
6118 except errors.TagError, err:
6119 raise errors.OpExecError("Error while setting tag: %s" % str(err))
6121 self.cfg.Update(self.target)
6122 except errors.ConfigurationError:
6123 raise errors.OpRetryError("There has been a modification to the"
6124 " config file and the operation has been"
6125 " aborted. Please retry.")
6128 class LUDelTags(TagsLU):
6129 """Delete a list of tags from a given object.
6132 _OP_REQP = ["kind", "name", "tags"]
6135 def CheckPrereq(self):
6136 """Check prerequisites.
6138 This checks that we have the given tag.
6141 TagsLU.CheckPrereq(self)
6142 for tag in self.op.tags:
6143 objects.TaggableObject.ValidateTag(tag)
6144 del_tags = frozenset(self.op.tags)
6145 cur_tags = self.target.GetTags()
6146 if not del_tags <= cur_tags:
6147 diff_tags = del_tags - cur_tags
6148 diff_names = ["'%s'" % tag for tag in diff_tags]
6150 raise errors.OpPrereqError("Tag(s) %s not found" %
6151 (",".join(diff_names)))
6153 def Exec(self, feedback_fn):
6154 """Remove the tag from the object.
6157 for tag in self.op.tags:
6158 self.target.RemoveTag(tag)
6160 self.cfg.Update(self.target)
6161 except errors.ConfigurationError:
6162 raise errors.OpRetryError("There has been a modification to the"
6163 " config file and the operation has been"
6164 " aborted. Please retry.")
6167 class LUTestDelay(NoHooksLU):
6168 """Sleep for a specified amount of time.
6170 This LU sleeps on the master and/or nodes for a specified amount of
6174 _OP_REQP = ["duration", "on_master", "on_nodes"]
6177 def ExpandNames(self):
6178 """Expand names and set required locks.
6180 This expands the node list, if any.
6183 self.needed_locks = {}
6184 if self.op.on_nodes:
6185 # _GetWantedNodes can be used here, but is not always appropriate to use
6186 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6188 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6189 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6191 def CheckPrereq(self):
6192 """Check prerequisites.
6196 def Exec(self, feedback_fn):
6197 """Do the actual sleep.
6200 if self.op.on_master:
6201 if not utils.TestDelay(self.op.duration):
6202 raise errors.OpExecError("Error during master delay test")
6203 if self.op.on_nodes:
6204 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6206 raise errors.OpExecError("Complete failure from rpc call")
6207 for node, node_result in result.items():
6209 if not node_result.data:
6210 raise errors.OpExecError("Failure during rpc call to node %s,"
6211 " result: %s" % (node, node_result.data))
6214 class IAllocator(object):
6215 """IAllocator framework.
6217 An IAllocator instance has three sets of attributes:
6218 - cfg that is needed to query the cluster
6219 - input data (all members of the _KEYS class attribute are required)
6220 - four buffer attributes (in|out_data|text), that represent the
6221 input (to the external script) in text and data structure format,
6222 and the output from it, again in two formats
6223 - the result variables from the script (success, info, nodes) for
6228 "mem_size", "disks", "disk_template",
6229 "os", "tags", "nics", "vcpus", "hypervisor",
6235 def __init__(self, lu, mode, name, **kwargs):
6237 # init buffer variables
6238 self.in_text = self.out_text = self.in_data = self.out_data = None
6239 # init all input fields so that pylint is happy
6242 self.mem_size = self.disks = self.disk_template = None
6243 self.os = self.tags = self.nics = self.vcpus = None
6244 self.hypervisor = None
6245 self.relocate_from = None
6247 self.required_nodes = None
6248 # init result fields
6249 self.success = self.info = self.nodes = None
6250 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6251 keyset = self._ALLO_KEYS
6252 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6253 keyset = self._RELO_KEYS
6255 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6256 " IAllocator" % self.mode)
6258 if key not in keyset:
6259 raise errors.ProgrammerError("Invalid input parameter '%s' to"
6260 " IAllocator" % key)
6261 setattr(self, key, kwargs[key])
6263 if key not in kwargs:
6264 raise errors.ProgrammerError("Missing input parameter '%s' to"
6265 " IAllocator" % key)
6266 self._BuildInputData()
6268 def _ComputeClusterData(self):
6269 """Compute the generic allocator input data.
6271 This is the data that is independent of the actual operation.
6275 cluster_info = cfg.GetClusterInfo()
6279 "cluster_name": cfg.GetClusterName(),
6280 "cluster_tags": list(cluster_info.GetTags()),
6281 "enable_hypervisors": list(cluster_info.enabled_hypervisors),
6282 # we don't have job IDs
6284 iinfo = cfg.GetAllInstancesInfo().values()
6285 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6289 node_list = cfg.GetNodeList()
6291 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6292 hypervisor_name = self.hypervisor
6293 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6294 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6296 node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6298 node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6299 cluster_info.enabled_hypervisors)
6300 for nname in node_list:
6301 ninfo = cfg.GetNodeInfo(nname)
6302 node_data[nname].Raise()
6303 if not isinstance(node_data[nname].data, dict):
6304 raise errors.OpExecError("Can't get data for node %s" % nname)
6305 remote_info = node_data[nname].data
6306 for attr in ['memory_total', 'memory_free', 'memory_dom0',
6307 'vg_size', 'vg_free', 'cpu_total']:
6308 if attr not in remote_info:
6309 raise errors.OpExecError("Node '%s' didn't return attribute '%s'" %
6312 remote_info[attr] = int(remote_info[attr])
6313 except ValueError, err:
6314 raise errors.OpExecError("Node '%s' returned invalid value for '%s':"
6315 " %s" % (nname, attr, str(err)))
6316 # compute memory used by primary instances
6317 i_p_mem = i_p_up_mem = 0
6318 for iinfo, beinfo in i_list:
6319 if iinfo.primary_node == nname:
6320 i_p_mem += beinfo[constants.BE_MEMORY]
6321 if iinfo.name not in node_iinfo[nname]:
6324 i_used_mem = int(node_iinfo[nname][iinfo.name]['memory'])
6325 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6326 remote_info['memory_free'] -= max(0, i_mem_diff)
6328 if iinfo.status == "up":
6329 i_p_up_mem += beinfo[constants.BE_MEMORY]
6331 # compute memory used by instances
6333 "tags": list(ninfo.GetTags()),
6334 "total_memory": remote_info['memory_total'],
6335 "reserved_memory": remote_info['memory_dom0'],
6336 "free_memory": remote_info['memory_free'],
6337 "i_pri_memory": i_p_mem,
6338 "i_pri_up_memory": i_p_up_mem,
6339 "total_disk": remote_info['vg_size'],
6340 "free_disk": remote_info['vg_free'],
6341 "primary_ip": ninfo.primary_ip,
6342 "secondary_ip": ninfo.secondary_ip,
6343 "total_cpus": remote_info['cpu_total'],
6344 "offline": ninfo.offline,
6346 node_results[nname] = pnr
6347 data["nodes"] = node_results
6351 for iinfo, beinfo in i_list:
6352 nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
6353 for n in iinfo.nics]
6355 "tags": list(iinfo.GetTags()),
6356 "should_run": iinfo.status == "up",
6357 "vcpus": beinfo[constants.BE_VCPUS],
6358 "memory": beinfo[constants.BE_MEMORY],
6360 "nodes": list(iinfo.all_nodes),
6362 "disks": [{"size": dsk.size, "mode": "w"} for dsk in iinfo.disks],
6363 "disk_template": iinfo.disk_template,
6364 "hypervisor": iinfo.hypervisor,
6366 instance_data[iinfo.name] = pir
6368 data["instances"] = instance_data
6372 def _AddNewInstance(self):
6373 """Add new instance data to allocator structure.
6375 This in combination with _AllocatorGetClusterData will create the
6376 correct structure needed as input for the allocator.
6378 The checks for the completeness of the opcode must have already been
6383 if len(self.disks) != 2:
6384 raise errors.OpExecError("Only two-disk configurations supported")
6386 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6388 if self.disk_template in constants.DTS_NET_MIRROR:
6389 self.required_nodes = 2
6391 self.required_nodes = 1
6395 "disk_template": self.disk_template,
6398 "vcpus": self.vcpus,
6399 "memory": self.mem_size,
6400 "disks": self.disks,
6401 "disk_space_total": disk_space,
6403 "required_nodes": self.required_nodes,
6405 data["request"] = request
6407 def _AddRelocateInstance(self):
6408 """Add relocate instance data to allocator structure.
6410 This in combination with _IAllocatorGetClusterData will create the
6411 correct structure needed as input for the allocator.
6413 The checks for the completeness of the opcode must have already been
6417 instance = self.lu.cfg.GetInstanceInfo(self.name)
6418 if instance is None:
6419 raise errors.ProgrammerError("Unknown instance '%s' passed to"
6420 " IAllocator" % self.name)
6422 if instance.disk_template not in constants.DTS_NET_MIRROR:
6423 raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6425 if len(instance.secondary_nodes) != 1:
6426 raise errors.OpPrereqError("Instance has not exactly one secondary node")
6428 self.required_nodes = 1
6429 disk_sizes = [{'size': disk.size} for disk in instance.disks]
6430 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6435 "disk_space_total": disk_space,
6436 "required_nodes": self.required_nodes,
6437 "relocate_from": self.relocate_from,
6439 self.in_data["request"] = request
6441 def _BuildInputData(self):
6442 """Build input data structures.
6445 self._ComputeClusterData()
6447 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6448 self._AddNewInstance()
6450 self._AddRelocateInstance()
6452 self.in_text = serializer.Dump(self.in_data)
6454 def Run(self, name, validate=True, call_fn=None):
6455 """Run an instance allocator and return the results.
6459 call_fn = self.lu.rpc.call_iallocator_runner
6462 result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6465 if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6466 raise errors.OpExecError("Invalid result from master iallocator runner")
6468 rcode, stdout, stderr, fail = result.data
6470 if rcode == constants.IARUN_NOTFOUND:
6471 raise errors.OpExecError("Can't find allocator '%s'" % name)
6472 elif rcode == constants.IARUN_FAILURE:
6473 raise errors.OpExecError("Instance allocator call failed: %s,"
6474 " output: %s" % (fail, stdout+stderr))
6475 self.out_text = stdout
6477 self._ValidateResult()
6479 def _ValidateResult(self):
6480 """Process the allocator results.
6482 This will process and if successful save the result in
6483 self.out_data and the other parameters.
6487 rdict = serializer.Load(self.out_text)
6488 except Exception, err:
6489 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6491 if not isinstance(rdict, dict):
6492 raise errors.OpExecError("Can't parse iallocator results: not a dict")
6494 for key in "success", "info", "nodes":
6495 if key not in rdict:
6496 raise errors.OpExecError("Can't parse iallocator results:"
6497 " missing key '%s'" % key)
6498 setattr(self, key, rdict[key])
6500 if not isinstance(rdict["nodes"], list):
6501 raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6503 self.out_data = rdict
6506 class LUTestAllocator(NoHooksLU):
6507 """Run allocator tests.
6509 This LU runs the allocator tests
6512 _OP_REQP = ["direction", "mode", "name"]
6514 def CheckPrereq(self):
6515 """Check prerequisites.
6517 This checks the opcode parameters depending on the director and mode test.
6520 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6521 for attr in ["name", "mem_size", "disks", "disk_template",
6522 "os", "tags", "nics", "vcpus"]:
6523 if not hasattr(self.op, attr):
6524 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6526 iname = self.cfg.ExpandInstanceName(self.op.name)
6527 if iname is not None:
6528 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6530 if not isinstance(self.op.nics, list):
6531 raise errors.OpPrereqError("Invalid parameter 'nics'")
6532 for row in self.op.nics:
6533 if (not isinstance(row, dict) or
6536 "bridge" not in row):
6537 raise errors.OpPrereqError("Invalid contents of the"
6538 " 'nics' parameter")
6539 if not isinstance(self.op.disks, list):
6540 raise errors.OpPrereqError("Invalid parameter 'disks'")
6541 if len(self.op.disks) != 2:
6542 raise errors.OpPrereqError("Only two-disk configurations supported")
6543 for row in self.op.disks:
6544 if (not isinstance(row, dict) or
6545 "size" not in row or
6546 not isinstance(row["size"], int) or
6547 "mode" not in row or
6548 row["mode"] not in ['r', 'w']):
6549 raise errors.OpPrereqError("Invalid contents of the"
6550 " 'disks' parameter")
6551 if self.op.hypervisor is None:
6552 self.op.hypervisor = self.cfg.GetHypervisorType()
6553 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6554 if not hasattr(self.op, "name"):
6555 raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6556 fname = self.cfg.ExpandInstanceName(self.op.name)
6558 raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6560 self.op.name = fname
6561 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6563 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6566 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6567 if not hasattr(self.op, "allocator") or self.op.allocator is None:
6568 raise errors.OpPrereqError("Missing allocator name")
6569 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6570 raise errors.OpPrereqError("Wrong allocator test '%s'" %
6573 def Exec(self, feedback_fn):
6574 """Run the allocator test.
6577 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6578 ial = IAllocator(self,
6581 mem_size=self.op.mem_size,
6582 disks=self.op.disks,
6583 disk_template=self.op.disk_template,
6587 vcpus=self.op.vcpus,
6588 hypervisor=self.op.hypervisor,
6591 ial = IAllocator(self,
6594 relocate_from=list(self.relocate_from),
6597 if self.op.direction == constants.IALLOCATOR_DIR_IN:
6598 result = ial.in_text
6600 ial.Run(self.op.allocator, validate=False)
6601 result = ial.out_text