4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 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=W0201,C0302
26 # W0201 since most LU attributes are defined in CheckPrereq or similar
29 # C0302: since we have waaaay to many lines in this module
45 from ganeti import ssh
46 from ganeti import utils
47 from ganeti import errors
48 from ganeti import hypervisor
49 from ganeti import locking
50 from ganeti import constants
51 from ganeti import objects
52 from ganeti import serializer
53 from ganeti import ssconf
54 from ganeti import uidpool
55 from ganeti import compat
56 from ganeti import masterd
57 from ganeti import netutils
58 from ganeti import query
59 from ganeti import qlang
60 from ganeti import opcodes
63 import ganeti.masterd.instance # pylint: disable=W0611
67 """Data container for LU results with jobs.
69 Instances of this class returned from L{LogicalUnit.Exec} will be recognized
70 by L{mcpu.Processor._ProcessResult}. The latter will then submit the jobs
71 contained in the C{jobs} attribute and include the job IDs in the opcode
75 def __init__(self, jobs, **kwargs):
76 """Initializes this class.
78 Additional return values can be specified as keyword arguments.
80 @type jobs: list of lists of L{opcode.OpCode}
81 @param jobs: A list of lists of opcode objects
88 class LogicalUnit(object):
89 """Logical Unit base class.
91 Subclasses must follow these rules:
92 - implement ExpandNames
93 - implement CheckPrereq (except when tasklets are used)
94 - implement Exec (except when tasklets are used)
95 - implement BuildHooksEnv
96 - implement BuildHooksNodes
97 - redefine HPATH and HTYPE
98 - optionally redefine their run requirements:
99 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
101 Note that all commands require root permissions.
103 @ivar dry_run_result: the value (if any) that will be returned to the caller
104 in dry-run mode (signalled by opcode dry_run parameter)
111 def __init__(self, processor, op, context, rpc):
112 """Constructor for LogicalUnit.
114 This needs to be overridden in derived classes in order to check op
118 self.proc = processor
120 self.cfg = context.cfg
121 self.glm = context.glm
123 self.owned_locks = context.glm.list_owned
124 self.context = context
126 # Dicts used to declare locking needs to mcpu
127 self.needed_locks = None
128 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
130 self.remove_locks = {}
131 # Used to force good behavior when calling helper functions
132 self.recalculate_locks = {}
134 self.Log = processor.Log # pylint: disable=C0103
135 self.LogWarning = processor.LogWarning # pylint: disable=C0103
136 self.LogInfo = processor.LogInfo # pylint: disable=C0103
137 self.LogStep = processor.LogStep # pylint: disable=C0103
138 # support for dry-run
139 self.dry_run_result = None
140 # support for generic debug attribute
141 if (not hasattr(self.op, "debug_level") or
142 not isinstance(self.op.debug_level, int)):
143 self.op.debug_level = 0
148 # Validate opcode parameters and set defaults
149 self.op.Validate(True)
151 self.CheckArguments()
153 def CheckArguments(self):
154 """Check syntactic validity for the opcode arguments.
156 This method is for doing a simple syntactic check and ensure
157 validity of opcode parameters, without any cluster-related
158 checks. While the same can be accomplished in ExpandNames and/or
159 CheckPrereq, doing these separate is better because:
161 - ExpandNames is left as as purely a lock-related function
162 - CheckPrereq is run after we have acquired locks (and possible
165 The function is allowed to change the self.op attribute so that
166 later methods can no longer worry about missing parameters.
171 def ExpandNames(self):
172 """Expand names for this LU.
174 This method is called before starting to execute the opcode, and it should
175 update all the parameters of the opcode to their canonical form (e.g. a
176 short node name must be fully expanded after this method has successfully
177 completed). This way locking, hooks, logging, etc. can work correctly.
179 LUs which implement this method must also populate the self.needed_locks
180 member, as a dict with lock levels as keys, and a list of needed lock names
183 - use an empty dict if you don't need any lock
184 - if you don't need any lock at a particular level omit that level
185 - don't put anything for the BGL level
186 - if you want all locks at a level use locking.ALL_SET as a value
188 If you need to share locks (rather than acquire them exclusively) at one
189 level you can modify self.share_locks, setting a true value (usually 1) for
190 that level. By default locks are not shared.
192 This function can also define a list of tasklets, which then will be
193 executed in order instead of the usual LU-level CheckPrereq and Exec
194 functions, if those are not defined by the LU.
198 # Acquire all nodes and one instance
199 self.needed_locks = {
200 locking.LEVEL_NODE: locking.ALL_SET,
201 locking.LEVEL_INSTANCE: ['instance1.example.com'],
203 # Acquire just two nodes
204 self.needed_locks = {
205 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
208 self.needed_locks = {} # No, you can't leave it to the default value None
211 # The implementation of this method is mandatory only if the new LU is
212 # concurrent, so that old LUs don't need to be changed all at the same
215 self.needed_locks = {} # Exclusive LUs don't need locks.
217 raise NotImplementedError
219 def DeclareLocks(self, level):
220 """Declare LU locking needs for a level
222 While most LUs can just declare their locking needs at ExpandNames time,
223 sometimes there's the need to calculate some locks after having acquired
224 the ones before. This function is called just before acquiring locks at a
225 particular level, but after acquiring the ones at lower levels, and permits
226 such calculations. It can be used to modify self.needed_locks, and by
227 default it does nothing.
229 This function is only called if you have something already set in
230 self.needed_locks for the level.
232 @param level: Locking level which is going to be locked
233 @type level: member of ganeti.locking.LEVELS
237 def CheckPrereq(self):
238 """Check prerequisites for this LU.
240 This method should check that the prerequisites for the execution
241 of this LU are fulfilled. It can do internode communication, but
242 it should be idempotent - no cluster or system changes are
245 The method should raise errors.OpPrereqError in case something is
246 not fulfilled. Its return value is ignored.
248 This method should also update all the parameters of the opcode to
249 their canonical form if it hasn't been done by ExpandNames before.
252 if self.tasklets is not None:
253 for (idx, tl) in enumerate(self.tasklets):
254 logging.debug("Checking prerequisites for tasklet %s/%s",
255 idx + 1, len(self.tasklets))
260 def Exec(self, feedback_fn):
263 This method should implement the actual work. It should raise
264 errors.OpExecError for failures that are somewhat dealt with in
268 if self.tasklets is not None:
269 for (idx, tl) in enumerate(self.tasklets):
270 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
273 raise NotImplementedError
275 def BuildHooksEnv(self):
276 """Build hooks environment for this LU.
279 @return: Dictionary containing the environment that will be used for
280 running the hooks for this LU. The keys of the dict must not be prefixed
281 with "GANETI_"--that'll be added by the hooks runner. The hooks runner
282 will extend the environment with additional variables. If no environment
283 should be defined, an empty dictionary should be returned (not C{None}).
284 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
288 raise NotImplementedError
290 def BuildHooksNodes(self):
291 """Build list of nodes to run LU's hooks.
293 @rtype: tuple; (list, list)
294 @return: Tuple containing a list of node names on which the hook
295 should run before the execution and a list of node names on which the
296 hook should run after the execution. No nodes should be returned as an
297 empty list (and not None).
298 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
302 raise NotImplementedError
304 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
305 """Notify the LU about the results of its hooks.
307 This method is called every time a hooks phase is executed, and notifies
308 the Logical Unit about the hooks' result. The LU can then use it to alter
309 its result based on the hooks. By default the method does nothing and the
310 previous result is passed back unchanged but any LU can define it if it
311 wants to use the local cluster hook-scripts somehow.
313 @param phase: one of L{constants.HOOKS_PHASE_POST} or
314 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
315 @param hook_results: the results of the multi-node hooks rpc call
316 @param feedback_fn: function used send feedback back to the caller
317 @param lu_result: the previous Exec result this LU had, or None
319 @return: the new Exec result, based on the previous result
323 # API must be kept, thus we ignore the unused argument and could
324 # be a function warnings
325 # pylint: disable=W0613,R0201
328 def _ExpandAndLockInstance(self):
329 """Helper function to expand and lock an instance.
331 Many LUs that work on an instance take its name in self.op.instance_name
332 and need to expand it and then declare the expanded name for locking. This
333 function does it, and then updates self.op.instance_name to the expanded
334 name. It also initializes needed_locks as a dict, if this hasn't been done
338 if self.needed_locks is None:
339 self.needed_locks = {}
341 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
342 "_ExpandAndLockInstance called with instance-level locks set"
343 self.op.instance_name = _ExpandInstanceName(self.cfg,
344 self.op.instance_name)
345 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
347 def _LockInstancesNodes(self, primary_only=False):
348 """Helper function to declare instances' nodes for locking.
350 This function should be called after locking one or more instances to lock
351 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
352 with all primary or secondary nodes for instances already locked and
353 present in self.needed_locks[locking.LEVEL_INSTANCE].
355 It should be called from DeclareLocks, and for safety only works if
356 self.recalculate_locks[locking.LEVEL_NODE] is set.
358 In the future it may grow parameters to just lock some instance's nodes, or
359 to just lock primaries or secondary nodes, if needed.
361 If should be called in DeclareLocks in a way similar to::
363 if level == locking.LEVEL_NODE:
364 self._LockInstancesNodes()
366 @type primary_only: boolean
367 @param primary_only: only lock primary nodes of locked instances
370 assert locking.LEVEL_NODE in self.recalculate_locks, \
371 "_LockInstancesNodes helper function called with no nodes to recalculate"
373 # TODO: check if we're really been called with the instance locks held
375 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
376 # future we might want to have different behaviors depending on the value
377 # of self.recalculate_locks[locking.LEVEL_NODE]
379 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
380 for _, instance in self.cfg.GetMultiInstanceInfo(locked_i):
381 wanted_nodes.append(instance.primary_node)
383 wanted_nodes.extend(instance.secondary_nodes)
385 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
386 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
387 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
388 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
390 del self.recalculate_locks[locking.LEVEL_NODE]
393 class NoHooksLU(LogicalUnit): # pylint: disable=W0223
394 """Simple LU which runs no hooks.
396 This LU is intended as a parent for other LogicalUnits which will
397 run no hooks, in order to reduce duplicate code.
403 def BuildHooksEnv(self):
404 """Empty BuildHooksEnv for NoHooksLu.
406 This just raises an error.
409 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
411 def BuildHooksNodes(self):
412 """Empty BuildHooksNodes for NoHooksLU.
415 raise AssertionError("BuildHooksNodes called for NoHooksLU")
419 """Tasklet base class.
421 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
422 they can mix legacy code with tasklets. Locking needs to be done in the LU,
423 tasklets know nothing about locks.
425 Subclasses must follow these rules:
426 - Implement CheckPrereq
430 def __init__(self, lu):
437 def CheckPrereq(self):
438 """Check prerequisites for this tasklets.
440 This method should check whether the prerequisites for the execution of
441 this tasklet are fulfilled. It can do internode communication, but it
442 should be idempotent - no cluster or system changes are allowed.
444 The method should raise errors.OpPrereqError in case something is not
445 fulfilled. Its return value is ignored.
447 This method should also update all parameters to their canonical form if it
448 hasn't been done before.
453 def Exec(self, feedback_fn):
454 """Execute the tasklet.
456 This method should implement the actual work. It should raise
457 errors.OpExecError for failures that are somewhat dealt with in code, or
461 raise NotImplementedError
465 """Base for query utility classes.
468 #: Attribute holding field definitions
471 def __init__(self, filter_, fields, use_locking):
472 """Initializes this class.
475 self.use_locking = use_locking
477 self.query = query.Query(self.FIELDS, fields, filter_=filter_,
479 self.requested_data = self.query.RequestedData()
480 self.names = self.query.RequestedNames()
482 # Sort only if no names were requested
483 self.sort_by_name = not self.names
485 self.do_locking = None
488 def _GetNames(self, lu, all_names, lock_level):
489 """Helper function to determine names asked for in the query.
493 names = lu.owned_locks(lock_level)
497 if self.wanted == locking.ALL_SET:
498 assert not self.names
499 # caller didn't specify names, so ordering is not important
500 return utils.NiceSort(names)
502 # caller specified names and we must keep the same order
504 assert not self.do_locking or lu.glm.is_owned(lock_level)
506 missing = set(self.wanted).difference(names)
508 raise errors.OpExecError("Some items were removed before retrieving"
509 " their data: %s" % missing)
511 # Return expanded names
514 def ExpandNames(self, lu):
515 """Expand names for this query.
517 See L{LogicalUnit.ExpandNames}.
520 raise NotImplementedError()
522 def DeclareLocks(self, lu, level):
523 """Declare locks for this query.
525 See L{LogicalUnit.DeclareLocks}.
528 raise NotImplementedError()
530 def _GetQueryData(self, lu):
531 """Collects all data for this query.
533 @return: Query data object
536 raise NotImplementedError()
538 def NewStyleQuery(self, lu):
539 """Collect data and execute query.
542 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
543 sort_by_name=self.sort_by_name)
545 def OldStyleQuery(self, lu):
546 """Collect data and execute query.
549 return self.query.OldStyleQuery(self._GetQueryData(lu),
550 sort_by_name=self.sort_by_name)
554 """Returns a dict declaring all lock levels shared.
557 return dict.fromkeys(locking.LEVELS, 1)
560 def _CheckInstanceNodeGroups(cfg, instance_name, owned_groups):
561 """Checks if the owned node groups are still correct for an instance.
563 @type cfg: L{config.ConfigWriter}
564 @param cfg: The cluster configuration
565 @type instance_name: string
566 @param instance_name: Instance name
567 @type owned_groups: set or frozenset
568 @param owned_groups: List of currently owned node groups
571 inst_groups = cfg.GetInstanceNodeGroups(instance_name)
573 if not owned_groups.issuperset(inst_groups):
574 raise errors.OpPrereqError("Instance %s's node groups changed since"
575 " locks were acquired, current groups are"
576 " are '%s', owning groups '%s'; retry the"
579 utils.CommaJoin(inst_groups),
580 utils.CommaJoin(owned_groups)),
586 def _CheckNodeGroupInstances(cfg, group_uuid, owned_instances):
587 """Checks if the instances in a node group are still correct.
589 @type cfg: L{config.ConfigWriter}
590 @param cfg: The cluster configuration
591 @type group_uuid: string
592 @param group_uuid: Node group UUID
593 @type owned_instances: set or frozenset
594 @param owned_instances: List of currently owned instances
597 wanted_instances = cfg.GetNodeGroupInstances(group_uuid)
598 if owned_instances != wanted_instances:
599 raise errors.OpPrereqError("Instances in node group '%s' changed since"
600 " locks were acquired, wanted '%s', have '%s';"
601 " retry the operation" %
603 utils.CommaJoin(wanted_instances),
604 utils.CommaJoin(owned_instances)),
607 return wanted_instances
610 def _SupportsOob(cfg, node):
611 """Tells if node supports OOB.
613 @type cfg: L{config.ConfigWriter}
614 @param cfg: The cluster configuration
615 @type node: L{objects.Node}
616 @param node: The node
617 @return: The OOB script if supported or an empty string otherwise
620 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
623 def _GetWantedNodes(lu, nodes):
624 """Returns list of checked and expanded node names.
626 @type lu: L{LogicalUnit}
627 @param lu: the logical unit on whose behalf we execute
629 @param nodes: list of node names or None for all nodes
631 @return: the list of nodes, sorted
632 @raise errors.ProgrammerError: if the nodes parameter is wrong type
636 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
638 return utils.NiceSort(lu.cfg.GetNodeList())
641 def _GetWantedInstances(lu, instances):
642 """Returns list of checked and expanded instance names.
644 @type lu: L{LogicalUnit}
645 @param lu: the logical unit on whose behalf we execute
646 @type instances: list
647 @param instances: list of instance names or None for all instances
649 @return: the list of instances, sorted
650 @raise errors.OpPrereqError: if the instances parameter is wrong type
651 @raise errors.OpPrereqError: if any of the passed instances is not found
655 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
657 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
661 def _GetUpdatedParams(old_params, update_dict,
662 use_default=True, use_none=False):
663 """Return the new version of a parameter dictionary.
665 @type old_params: dict
666 @param old_params: old parameters
667 @type update_dict: dict
668 @param update_dict: dict containing new parameter values, or
669 constants.VALUE_DEFAULT to reset the parameter to its default
671 @param use_default: boolean
672 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
673 values as 'to be deleted' values
674 @param use_none: boolean
675 @type use_none: whether to recognise C{None} values as 'to be
678 @return: the new parameter dictionary
681 params_copy = copy.deepcopy(old_params)
682 for key, val in update_dict.iteritems():
683 if ((use_default and val == constants.VALUE_DEFAULT) or
684 (use_none and val is None)):
690 params_copy[key] = val
694 def _ReleaseLocks(lu, level, names=None, keep=None):
695 """Releases locks owned by an LU.
697 @type lu: L{LogicalUnit}
698 @param level: Lock level
699 @type names: list or None
700 @param names: Names of locks to release
701 @type keep: list or None
702 @param keep: Names of locks to retain
705 assert not (keep is not None and names is not None), \
706 "Only one of the 'names' and the 'keep' parameters can be given"
708 if names is not None:
709 should_release = names.__contains__
711 should_release = lambda name: name not in keep
713 should_release = None
719 # Determine which locks to release
720 for name in lu.owned_locks(level):
721 if should_release(name):
726 assert len(lu.owned_locks(level)) == (len(retain) + len(release))
728 # Release just some locks
729 lu.glm.release(level, names=release)
731 assert frozenset(lu.owned_locks(level)) == frozenset(retain)
734 lu.glm.release(level)
736 assert not lu.glm.is_owned(level), "No locks should be owned"
739 def _MapInstanceDisksToNodes(instances):
740 """Creates a map from (node, volume) to instance name.
742 @type instances: list of L{objects.Instance}
743 @rtype: dict; tuple of (node name, volume name) as key, instance name as value
746 return dict(((node, vol), inst.name)
747 for inst in instances
748 for (node, vols) in inst.MapLVsByNode().items()
752 def _RunPostHook(lu, node_name):
753 """Runs the post-hook for an opcode on a single node.
756 hm = lu.proc.hmclass(lu.rpc.call_hooks_runner, lu)
758 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
760 # pylint: disable=W0702
761 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
764 def _CheckOutputFields(static, dynamic, selected):
765 """Checks whether all selected fields are valid.
767 @type static: L{utils.FieldSet}
768 @param static: static fields set
769 @type dynamic: L{utils.FieldSet}
770 @param dynamic: dynamic fields set
777 delta = f.NonMatching(selected)
779 raise errors.OpPrereqError("Unknown output fields selected: %s"
780 % ",".join(delta), errors.ECODE_INVAL)
783 def _CheckGlobalHvParams(params):
784 """Validates that given hypervisor params are not global ones.
786 This will ensure that instances don't get customised versions of
790 used_globals = constants.HVC_GLOBALS.intersection(params)
792 msg = ("The following hypervisor parameters are global and cannot"
793 " be customized at instance level, please modify them at"
794 " cluster level: %s" % utils.CommaJoin(used_globals))
795 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
798 def _CheckNodeOnline(lu, node, msg=None):
799 """Ensure that a given node is online.
801 @param lu: the LU on behalf of which we make the check
802 @param node: the node to check
803 @param msg: if passed, should be a message to replace the default one
804 @raise errors.OpPrereqError: if the node is offline
808 msg = "Can't use offline node"
809 if lu.cfg.GetNodeInfo(node).offline:
810 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
813 def _CheckNodeNotDrained(lu, node):
814 """Ensure that a given node is not drained.
816 @param lu: the LU on behalf of which we make the check
817 @param node: the node to check
818 @raise errors.OpPrereqError: if the node is drained
821 if lu.cfg.GetNodeInfo(node).drained:
822 raise errors.OpPrereqError("Can't use drained node %s" % node,
826 def _CheckNodeVmCapable(lu, node):
827 """Ensure that a given node is vm capable.
829 @param lu: the LU on behalf of which we make the check
830 @param node: the node to check
831 @raise errors.OpPrereqError: if the node is not vm capable
834 if not lu.cfg.GetNodeInfo(node).vm_capable:
835 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
839 def _CheckNodeHasOS(lu, node, os_name, force_variant):
840 """Ensure that a node supports a given OS.
842 @param lu: the LU on behalf of which we make the check
843 @param node: the node to check
844 @param os_name: the OS to query about
845 @param force_variant: whether to ignore variant errors
846 @raise errors.OpPrereqError: if the node is not supporting the OS
849 result = lu.rpc.call_os_get(node, os_name)
850 result.Raise("OS '%s' not in supported OS list for node %s" %
852 prereq=True, ecode=errors.ECODE_INVAL)
853 if not force_variant:
854 _CheckOSVariant(result.payload, os_name)
857 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
858 """Ensure that a node has the given secondary ip.
860 @type lu: L{LogicalUnit}
861 @param lu: the LU on behalf of which we make the check
863 @param node: the node to check
864 @type secondary_ip: string
865 @param secondary_ip: the ip to check
866 @type prereq: boolean
867 @param prereq: whether to throw a prerequisite or an execute error
868 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
869 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
872 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
873 result.Raise("Failure checking secondary ip on node %s" % node,
874 prereq=prereq, ecode=errors.ECODE_ENVIRON)
875 if not result.payload:
876 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
877 " please fix and re-run this command" % secondary_ip)
879 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
881 raise errors.OpExecError(msg)
884 def _GetClusterDomainSecret():
885 """Reads the cluster domain secret.
888 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
892 def _CheckInstanceDown(lu, instance, reason):
893 """Ensure that an instance is not running."""
894 if instance.admin_up:
895 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
896 (instance.name, reason), errors.ECODE_STATE)
898 pnode = instance.primary_node
899 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
900 ins_l.Raise("Can't contact node %s for instance information" % pnode,
901 prereq=True, ecode=errors.ECODE_ENVIRON)
903 if instance.name in ins_l.payload:
904 raise errors.OpPrereqError("Instance %s is running, %s" %
905 (instance.name, reason), errors.ECODE_STATE)
908 def _ExpandItemName(fn, name, kind):
909 """Expand an item name.
911 @param fn: the function to use for expansion
912 @param name: requested item name
913 @param kind: text description ('Node' or 'Instance')
914 @return: the resolved (full) name
915 @raise errors.OpPrereqError: if the item is not found
919 if full_name is None:
920 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
925 def _ExpandNodeName(cfg, name):
926 """Wrapper over L{_ExpandItemName} for nodes."""
927 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
930 def _ExpandInstanceName(cfg, name):
931 """Wrapper over L{_ExpandItemName} for instance."""
932 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
935 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
936 memory, vcpus, nics, disk_template, disks,
937 bep, hvp, hypervisor_name, tags):
938 """Builds instance related env variables for hooks
940 This builds the hook environment from individual variables.
943 @param name: the name of the instance
944 @type primary_node: string
945 @param primary_node: the name of the instance's primary node
946 @type secondary_nodes: list
947 @param secondary_nodes: list of secondary nodes as strings
948 @type os_type: string
949 @param os_type: the name of the instance's OS
950 @type status: boolean
951 @param status: the should_run status of the instance
953 @param memory: the memory size of the instance
955 @param vcpus: the count of VCPUs the instance has
957 @param nics: list of tuples (ip, mac, mode, link) representing
958 the NICs the instance has
959 @type disk_template: string
960 @param disk_template: the disk template of the instance
962 @param disks: the list of (size, mode) pairs
964 @param bep: the backend parameters for the instance
966 @param hvp: the hypervisor parameters for the instance
967 @type hypervisor_name: string
968 @param hypervisor_name: the hypervisor for the instance
970 @param tags: list of instance tags as strings
972 @return: the hook environment for this instance
981 "INSTANCE_NAME": name,
982 "INSTANCE_PRIMARY": primary_node,
983 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
984 "INSTANCE_OS_TYPE": os_type,
985 "INSTANCE_STATUS": str_status,
986 "INSTANCE_MEMORY": memory,
987 "INSTANCE_VCPUS": vcpus,
988 "INSTANCE_DISK_TEMPLATE": disk_template,
989 "INSTANCE_HYPERVISOR": hypervisor_name,
993 nic_count = len(nics)
994 for idx, (ip, mac, mode, link) in enumerate(nics):
997 env["INSTANCE_NIC%d_IP" % idx] = ip
998 env["INSTANCE_NIC%d_MAC" % idx] = mac
999 env["INSTANCE_NIC%d_MODE" % idx] = mode
1000 env["INSTANCE_NIC%d_LINK" % idx] = link
1001 if mode == constants.NIC_MODE_BRIDGED:
1002 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
1006 env["INSTANCE_NIC_COUNT"] = nic_count
1009 disk_count = len(disks)
1010 for idx, (size, mode) in enumerate(disks):
1011 env["INSTANCE_DISK%d_SIZE" % idx] = size
1012 env["INSTANCE_DISK%d_MODE" % idx] = mode
1016 env["INSTANCE_DISK_COUNT"] = disk_count
1021 env["INSTANCE_TAGS"] = " ".join(tags)
1023 for source, kind in [(bep, "BE"), (hvp, "HV")]:
1024 for key, value in source.items():
1025 env["INSTANCE_%s_%s" % (kind, key)] = value
1030 def _NICListToTuple(lu, nics):
1031 """Build a list of nic information tuples.
1033 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
1034 value in LUInstanceQueryData.
1036 @type lu: L{LogicalUnit}
1037 @param lu: the logical unit on whose behalf we execute
1038 @type nics: list of L{objects.NIC}
1039 @param nics: list of nics to convert to hooks tuples
1043 cluster = lu.cfg.GetClusterInfo()
1047 filled_params = cluster.SimpleFillNIC(nic.nicparams)
1048 mode = filled_params[constants.NIC_MODE]
1049 link = filled_params[constants.NIC_LINK]
1050 hooks_nics.append((ip, mac, mode, link))
1054 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1055 """Builds instance related env variables for hooks from an object.
1057 @type lu: L{LogicalUnit}
1058 @param lu: the logical unit on whose behalf we execute
1059 @type instance: L{objects.Instance}
1060 @param instance: the instance for which we should build the
1062 @type override: dict
1063 @param override: dictionary with key/values that will override
1066 @return: the hook environment dictionary
1069 cluster = lu.cfg.GetClusterInfo()
1070 bep = cluster.FillBE(instance)
1071 hvp = cluster.FillHV(instance)
1073 "name": instance.name,
1074 "primary_node": instance.primary_node,
1075 "secondary_nodes": instance.secondary_nodes,
1076 "os_type": instance.os,
1077 "status": instance.admin_up,
1078 "memory": bep[constants.BE_MEMORY],
1079 "vcpus": bep[constants.BE_VCPUS],
1080 "nics": _NICListToTuple(lu, instance.nics),
1081 "disk_template": instance.disk_template,
1082 "disks": [(disk.size, disk.mode) for disk in instance.disks],
1085 "hypervisor_name": instance.hypervisor,
1086 "tags": instance.tags,
1089 args.update(override)
1090 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
1093 def _AdjustCandidatePool(lu, exceptions):
1094 """Adjust the candidate pool after node operations.
1097 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1099 lu.LogInfo("Promoted nodes to master candidate role: %s",
1100 utils.CommaJoin(node.name for node in mod_list))
1101 for name in mod_list:
1102 lu.context.ReaddNode(name)
1103 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1105 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1109 def _DecideSelfPromotion(lu, exceptions=None):
1110 """Decide whether I should promote myself as a master candidate.
1113 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1114 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1115 # the new node will increase mc_max with one, so:
1116 mc_should = min(mc_should + 1, cp_size)
1117 return mc_now < mc_should
1120 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1121 """Check that the brigdes needed by a list of nics exist.
1124 cluster = lu.cfg.GetClusterInfo()
1125 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1126 brlist = [params[constants.NIC_LINK] for params in paramslist
1127 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1129 result = lu.rpc.call_bridges_exist(target_node, brlist)
1130 result.Raise("Error checking bridges on destination node '%s'" %
1131 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1134 def _CheckInstanceBridgesExist(lu, instance, node=None):
1135 """Check that the brigdes needed by an instance exist.
1139 node = instance.primary_node
1140 _CheckNicsBridgesExist(lu, instance.nics, node)
1143 def _CheckOSVariant(os_obj, name):
1144 """Check whether an OS name conforms to the os variants specification.
1146 @type os_obj: L{objects.OS}
1147 @param os_obj: OS object to check
1149 @param name: OS name passed by the user, to check for validity
1152 variant = objects.OS.GetVariant(name)
1153 if not os_obj.supported_variants:
1155 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'"
1156 " passed)" % (os_obj.name, variant),
1160 raise errors.OpPrereqError("OS name must include a variant",
1163 if variant not in os_obj.supported_variants:
1164 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1167 def _GetNodeInstancesInner(cfg, fn):
1168 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1171 def _GetNodeInstances(cfg, node_name):
1172 """Returns a list of all primary and secondary instances on a node.
1176 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1179 def _GetNodePrimaryInstances(cfg, node_name):
1180 """Returns primary instances on a node.
1183 return _GetNodeInstancesInner(cfg,
1184 lambda inst: node_name == inst.primary_node)
1187 def _GetNodeSecondaryInstances(cfg, node_name):
1188 """Returns secondary instances on a node.
1191 return _GetNodeInstancesInner(cfg,
1192 lambda inst: node_name in inst.secondary_nodes)
1195 def _GetStorageTypeArgs(cfg, storage_type):
1196 """Returns the arguments for a storage type.
1199 # Special case for file storage
1200 if storage_type == constants.ST_FILE:
1201 # storage.FileStorage wants a list of storage directories
1202 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1207 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1210 for dev in instance.disks:
1211 cfg.SetDiskID(dev, node_name)
1213 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1214 result.Raise("Failed to get disk status from node %s" % node_name,
1215 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1217 for idx, bdev_status in enumerate(result.payload):
1218 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1224 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1225 """Check the sanity of iallocator and node arguments and use the
1226 cluster-wide iallocator if appropriate.
1228 Check that at most one of (iallocator, node) is specified. If none is
1229 specified, then the LU's opcode's iallocator slot is filled with the
1230 cluster-wide default iallocator.
1232 @type iallocator_slot: string
1233 @param iallocator_slot: the name of the opcode iallocator slot
1234 @type node_slot: string
1235 @param node_slot: the name of the opcode target node slot
1238 node = getattr(lu.op, node_slot, None)
1239 iallocator = getattr(lu.op, iallocator_slot, None)
1241 if node is not None and iallocator is not None:
1242 raise errors.OpPrereqError("Do not specify both, iallocator and node",
1244 elif node is None and iallocator is None:
1245 default_iallocator = lu.cfg.GetDefaultIAllocator()
1246 if default_iallocator:
1247 setattr(lu.op, iallocator_slot, default_iallocator)
1249 raise errors.OpPrereqError("No iallocator or node given and no"
1250 " cluster-wide default iallocator found;"
1251 " please specify either an iallocator or a"
1252 " node, or set a cluster-wide default"
1256 def _GetDefaultIAllocator(cfg, iallocator):
1257 """Decides on which iallocator to use.
1259 @type cfg: L{config.ConfigWriter}
1260 @param cfg: Cluster configuration object
1261 @type iallocator: string or None
1262 @param iallocator: Iallocator specified in opcode
1264 @return: Iallocator name
1268 # Use default iallocator
1269 iallocator = cfg.GetDefaultIAllocator()
1272 raise errors.OpPrereqError("No iallocator was specified, neither in the"
1273 " opcode nor as a cluster-wide default",
1279 class LUClusterPostInit(LogicalUnit):
1280 """Logical unit for running hooks after cluster initialization.
1283 HPATH = "cluster-init"
1284 HTYPE = constants.HTYPE_CLUSTER
1286 def BuildHooksEnv(self):
1291 "OP_TARGET": self.cfg.GetClusterName(),
1294 def BuildHooksNodes(self):
1295 """Build hooks nodes.
1298 return ([], [self.cfg.GetMasterNode()])
1300 def Exec(self, feedback_fn):
1307 class LUClusterDestroy(LogicalUnit):
1308 """Logical unit for destroying the cluster.
1311 HPATH = "cluster-destroy"
1312 HTYPE = constants.HTYPE_CLUSTER
1314 def BuildHooksEnv(self):
1319 "OP_TARGET": self.cfg.GetClusterName(),
1322 def BuildHooksNodes(self):
1323 """Build hooks nodes.
1328 def CheckPrereq(self):
1329 """Check prerequisites.
1331 This checks whether the cluster is empty.
1333 Any errors are signaled by raising errors.OpPrereqError.
1336 master = self.cfg.GetMasterNode()
1338 nodelist = self.cfg.GetNodeList()
1339 if len(nodelist) != 1 or nodelist[0] != master:
1340 raise errors.OpPrereqError("There are still %d node(s) in"
1341 " this cluster." % (len(nodelist) - 1),
1343 instancelist = self.cfg.GetInstanceList()
1345 raise errors.OpPrereqError("There are still %d instance(s) in"
1346 " this cluster." % len(instancelist),
1349 def Exec(self, feedback_fn):
1350 """Destroys the cluster.
1353 master = self.cfg.GetMasterNode()
1355 # Run post hooks on master node before it's removed
1356 _RunPostHook(self, master)
1358 result = self.rpc.call_node_stop_master(master, False)
1359 result.Raise("Could not disable the master role")
1364 def _VerifyCertificate(filename):
1365 """Verifies a certificate for L{LUClusterVerifyConfig}.
1367 @type filename: string
1368 @param filename: Path to PEM file
1372 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1373 utils.ReadFile(filename))
1374 except Exception, err: # pylint: disable=W0703
1375 return (LUClusterVerifyConfig.ETYPE_ERROR,
1376 "Failed to load X509 certificate %s: %s" % (filename, err))
1379 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1380 constants.SSL_CERT_EXPIRATION_ERROR)
1383 fnamemsg = "While verifying %s: %s" % (filename, msg)
1388 return (None, fnamemsg)
1389 elif errcode == utils.CERT_WARNING:
1390 return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg)
1391 elif errcode == utils.CERT_ERROR:
1392 return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg)
1394 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1397 def _GetAllHypervisorParameters(cluster, instances):
1398 """Compute the set of all hypervisor parameters.
1400 @type cluster: L{objects.Cluster}
1401 @param cluster: the cluster object
1402 @param instances: list of L{objects.Instance}
1403 @param instances: additional instances from which to obtain parameters
1404 @rtype: list of (origin, hypervisor, parameters)
1405 @return: a list with all parameters found, indicating the hypervisor they
1406 apply to, and the origin (can be "cluster", "os X", or "instance Y")
1411 for hv_name in cluster.enabled_hypervisors:
1412 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
1414 for os_name, os_hvp in cluster.os_hvp.items():
1415 for hv_name, hv_params in os_hvp.items():
1417 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
1418 hvp_data.append(("os %s" % os_name, hv_name, full_params))
1420 # TODO: collapse identical parameter values in a single one
1421 for instance in instances:
1422 if instance.hvparams:
1423 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
1424 cluster.FillHV(instance)))
1429 class _VerifyErrors(object):
1430 """Mix-in for cluster/group verify LUs.
1432 It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects
1433 self.op and self._feedback_fn to be available.)
1436 TCLUSTER = "cluster"
1438 TINSTANCE = "instance"
1440 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1441 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1442 ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK")
1443 ECLUSTERDANGLINGNODES = (TNODE, "ECLUSTERDANGLINGNODES")
1444 ECLUSTERDANGLINGINST = (TNODE, "ECLUSTERDANGLINGINST")
1445 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1446 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1447 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1448 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1449 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1450 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1451 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1452 ENODEDRBD = (TNODE, "ENODEDRBD")
1453 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1454 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1455 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1456 ENODEHV = (TNODE, "ENODEHV")
1457 ENODELVM = (TNODE, "ENODELVM")
1458 ENODEN1 = (TNODE, "ENODEN1")
1459 ENODENET = (TNODE, "ENODENET")
1460 ENODEOS = (TNODE, "ENODEOS")
1461 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1462 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1463 ENODERPC = (TNODE, "ENODERPC")
1464 ENODESSH = (TNODE, "ENODESSH")
1465 ENODEVERSION = (TNODE, "ENODEVERSION")
1466 ENODESETUP = (TNODE, "ENODESETUP")
1467 ENODETIME = (TNODE, "ENODETIME")
1468 ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1470 ETYPE_FIELD = "code"
1471 ETYPE_ERROR = "ERROR"
1472 ETYPE_WARNING = "WARNING"
1474 def _Error(self, ecode, item, msg, *args, **kwargs):
1475 """Format an error message.
1477 Based on the opcode's error_codes parameter, either format a
1478 parseable error code, or a simpler error string.
1480 This must be called only from Exec and functions called from Exec.
1483 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1485 # first complete the msg
1488 # then format the whole message
1489 if self.op.error_codes: # This is a mix-in. pylint: disable=E1101
1490 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1496 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1497 # and finally report it via the feedback_fn
1498 self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable=E1101
1500 def _ErrorIf(self, cond, *args, **kwargs):
1501 """Log an error message if the passed condition is True.
1505 or self.op.debug_simulate_errors) # pylint: disable=E1101
1507 self._Error(*args, **kwargs)
1508 # do not mark the operation as failed for WARN cases only
1509 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1510 self.bad = self.bad or cond
1513 class LUClusterVerify(NoHooksLU):
1514 """Submits all jobs necessary to verify the cluster.
1519 def ExpandNames(self):
1520 self.needed_locks = {}
1522 def Exec(self, feedback_fn):
1525 if self.op.group_name:
1526 groups = [self.op.group_name]
1527 depends_fn = lambda: None
1529 groups = self.cfg.GetNodeGroupList()
1531 # Verify global configuration
1532 jobs.append([opcodes.OpClusterVerifyConfig()])
1534 # Always depend on global verification
1535 depends_fn = lambda: [(-len(jobs), [])]
1537 jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,
1538 depends=depends_fn())]
1539 for group in groups)
1541 # Fix up all parameters
1542 for op in itertools.chain(*jobs): # pylint: disable=W0142
1543 op.debug_simulate_errors = self.op.debug_simulate_errors
1544 op.verbose = self.op.verbose
1545 op.error_codes = self.op.error_codes
1547 op.skip_checks = self.op.skip_checks
1548 except AttributeError:
1549 assert not isinstance(op, opcodes.OpClusterVerifyGroup)
1551 return ResultWithJobs(jobs)
1554 class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
1555 """Verifies the cluster config.
1560 def _VerifyHVP(self, hvp_data):
1561 """Verifies locally the syntax of the hypervisor parameters.
1564 for item, hv_name, hv_params in hvp_data:
1565 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
1568 hv_class = hypervisor.GetHypervisor(hv_name)
1569 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1570 hv_class.CheckParameterSyntax(hv_params)
1571 except errors.GenericError, err:
1572 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
1574 def ExpandNames(self):
1575 # Information can be safely retrieved as the BGL is acquired in exclusive
1577 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER)
1578 self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
1579 self.all_node_info = self.cfg.GetAllNodesInfo()
1580 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1581 self.needed_locks = {}
1583 def Exec(self, feedback_fn):
1584 """Verify integrity of cluster, performing various test on nodes.
1588 self._feedback_fn = feedback_fn
1590 feedback_fn("* Verifying cluster config")
1592 for msg in self.cfg.VerifyConfig():
1593 self._ErrorIf(True, self.ECLUSTERCFG, None, msg)
1595 feedback_fn("* Verifying cluster certificate files")
1597 for cert_filename in constants.ALL_CERT_FILES:
1598 (errcode, msg) = _VerifyCertificate(cert_filename)
1599 self._ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
1601 feedback_fn("* Verifying hypervisor parameters")
1603 self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(),
1604 self.all_inst_info.values()))
1606 feedback_fn("* Verifying all nodes belong to an existing group")
1608 # We do this verification here because, should this bogus circumstance
1609 # occur, it would never be caught by VerifyGroup, which only acts on
1610 # nodes/instances reachable from existing node groups.
1612 dangling_nodes = set(node.name for node in self.all_node_info.values()
1613 if node.group not in self.all_group_info)
1615 dangling_instances = {}
1616 no_node_instances = []
1618 for inst in self.all_inst_info.values():
1619 if inst.primary_node in dangling_nodes:
1620 dangling_instances.setdefault(inst.primary_node, []).append(inst.name)
1621 elif inst.primary_node not in self.all_node_info:
1622 no_node_instances.append(inst.name)
1627 utils.CommaJoin(dangling_instances.get(node.name,
1629 for node in dangling_nodes]
1631 self._ErrorIf(bool(dangling_nodes), self.ECLUSTERDANGLINGNODES, None,
1632 "the following nodes (and their instances) belong to a non"
1633 " existing group: %s", utils.CommaJoin(pretty_dangling))
1635 self._ErrorIf(bool(no_node_instances), self.ECLUSTERDANGLINGINST, None,
1636 "the following instances have a non-existing primary-node:"
1637 " %s", utils.CommaJoin(no_node_instances))
1642 class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
1643 """Verifies the status of a node group.
1646 HPATH = "cluster-verify"
1647 HTYPE = constants.HTYPE_CLUSTER
1650 _HOOKS_INDENT_RE = re.compile("^", re.M)
1652 class NodeImage(object):
1653 """A class representing the logical and physical status of a node.
1656 @ivar name: the node name to which this object refers
1657 @ivar volumes: a structure as returned from
1658 L{ganeti.backend.GetVolumeList} (runtime)
1659 @ivar instances: a list of running instances (runtime)
1660 @ivar pinst: list of configured primary instances (config)
1661 @ivar sinst: list of configured secondary instances (config)
1662 @ivar sbp: dictionary of {primary-node: list of instances} for all
1663 instances for which this node is secondary (config)
1664 @ivar mfree: free memory, as reported by hypervisor (runtime)
1665 @ivar dfree: free disk, as reported by the node (runtime)
1666 @ivar offline: the offline status (config)
1667 @type rpc_fail: boolean
1668 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1669 not whether the individual keys were correct) (runtime)
1670 @type lvm_fail: boolean
1671 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1672 @type hyp_fail: boolean
1673 @ivar hyp_fail: whether the RPC call didn't return the instance list
1674 @type ghost: boolean
1675 @ivar ghost: whether this is a known node or not (config)
1676 @type os_fail: boolean
1677 @ivar os_fail: whether the RPC call didn't return valid OS data
1679 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1680 @type vm_capable: boolean
1681 @ivar vm_capable: whether the node can host instances
1684 def __init__(self, offline=False, name=None, vm_capable=True):
1693 self.offline = offline
1694 self.vm_capable = vm_capable
1695 self.rpc_fail = False
1696 self.lvm_fail = False
1697 self.hyp_fail = False
1699 self.os_fail = False
1702 def ExpandNames(self):
1703 # This raises errors.OpPrereqError on its own:
1704 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
1706 # Get instances in node group; this is unsafe and needs verification later
1707 inst_names = self.cfg.GetNodeGroupInstances(self.group_uuid)
1709 self.needed_locks = {
1710 locking.LEVEL_INSTANCE: inst_names,
1711 locking.LEVEL_NODEGROUP: [self.group_uuid],
1712 locking.LEVEL_NODE: [],
1715 self.share_locks = _ShareAll()
1717 def DeclareLocks(self, level):
1718 if level == locking.LEVEL_NODE:
1719 # Get members of node group; this is unsafe and needs verification later
1720 nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members)
1722 all_inst_info = self.cfg.GetAllInstancesInfo()
1724 # In Exec(), we warn about mirrored instances that have primary and
1725 # secondary living in separate node groups. To fully verify that
1726 # volumes for these instances are healthy, we will need to do an
1727 # extra call to their secondaries. We ensure here those nodes will
1729 for inst in self.owned_locks(locking.LEVEL_INSTANCE):
1730 # Important: access only the instances whose lock is owned
1731 if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR:
1732 nodes.update(all_inst_info[inst].secondary_nodes)
1734 self.needed_locks[locking.LEVEL_NODE] = nodes
1736 def CheckPrereq(self):
1737 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
1738 self.group_info = self.cfg.GetNodeGroup(self.group_uuid)
1740 group_nodes = set(self.group_info.members)
1741 group_instances = self.cfg.GetNodeGroupInstances(self.group_uuid)
1744 group_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
1746 unlocked_instances = \
1747 group_instances.difference(self.owned_locks(locking.LEVEL_INSTANCE))
1750 raise errors.OpPrereqError("Missing lock for nodes: %s" %
1751 utils.CommaJoin(unlocked_nodes))
1753 if unlocked_instances:
1754 raise errors.OpPrereqError("Missing lock for instances: %s" %
1755 utils.CommaJoin(unlocked_instances))
1757 self.all_node_info = self.cfg.GetAllNodesInfo()
1758 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1760 self.my_node_names = utils.NiceSort(group_nodes)
1761 self.my_inst_names = utils.NiceSort(group_instances)
1763 self.my_node_info = dict((name, self.all_node_info[name])
1764 for name in self.my_node_names)
1766 self.my_inst_info = dict((name, self.all_inst_info[name])
1767 for name in self.my_inst_names)
1769 # We detect here the nodes that will need the extra RPC calls for verifying
1770 # split LV volumes; they should be locked.
1771 extra_lv_nodes = set()
1773 for inst in self.my_inst_info.values():
1774 if inst.disk_template in constants.DTS_INT_MIRROR:
1775 group = self.my_node_info[inst.primary_node].group
1776 for nname in inst.secondary_nodes:
1777 if self.all_node_info[nname].group != group:
1778 extra_lv_nodes.add(nname)
1780 unlocked_lv_nodes = \
1781 extra_lv_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
1783 if unlocked_lv_nodes:
1784 raise errors.OpPrereqError("these nodes could be locked: %s" %
1785 utils.CommaJoin(unlocked_lv_nodes))
1786 self.extra_lv_nodes = list(extra_lv_nodes)
1788 def _VerifyNode(self, ninfo, nresult):
1789 """Perform some basic validation on data returned from a node.
1791 - check the result data structure is well formed and has all the
1793 - check ganeti version
1795 @type ninfo: L{objects.Node}
1796 @param ninfo: the node to check
1797 @param nresult: the results from the node
1799 @return: whether overall this call was successful (and we can expect
1800 reasonable values in the respose)
1804 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1806 # main result, nresult should be a non-empty dict
1807 test = not nresult or not isinstance(nresult, dict)
1808 _ErrorIf(test, self.ENODERPC, node,
1809 "unable to verify node: no data returned")
1813 # compares ganeti version
1814 local_version = constants.PROTOCOL_VERSION
1815 remote_version = nresult.get("version", None)
1816 test = not (remote_version and
1817 isinstance(remote_version, (list, tuple)) and
1818 len(remote_version) == 2)
1819 _ErrorIf(test, self.ENODERPC, node,
1820 "connection to node returned invalid data")
1824 test = local_version != remote_version[0]
1825 _ErrorIf(test, self.ENODEVERSION, node,
1826 "incompatible protocol versions: master %s,"
1827 " node %s", local_version, remote_version[0])
1831 # node seems compatible, we can actually try to look into its results
1833 # full package version
1834 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1835 self.ENODEVERSION, node,
1836 "software version mismatch: master %s, node %s",
1837 constants.RELEASE_VERSION, remote_version[1],
1838 code=self.ETYPE_WARNING)
1840 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1841 if ninfo.vm_capable and isinstance(hyp_result, dict):
1842 for hv_name, hv_result in hyp_result.iteritems():
1843 test = hv_result is not None
1844 _ErrorIf(test, self.ENODEHV, node,
1845 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1847 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1848 if ninfo.vm_capable and isinstance(hvp_result, list):
1849 for item, hv_name, hv_result in hvp_result:
1850 _ErrorIf(True, self.ENODEHV, node,
1851 "hypervisor %s parameter verify failure (source %s): %s",
1852 hv_name, item, hv_result)
1854 test = nresult.get(constants.NV_NODESETUP,
1855 ["Missing NODESETUP results"])
1856 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1861 def _VerifyNodeTime(self, ninfo, nresult,
1862 nvinfo_starttime, nvinfo_endtime):
1863 """Check the node time.
1865 @type ninfo: L{objects.Node}
1866 @param ninfo: the node to check
1867 @param nresult: the remote results for the node
1868 @param nvinfo_starttime: the start time of the RPC call
1869 @param nvinfo_endtime: the end time of the RPC call
1873 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1875 ntime = nresult.get(constants.NV_TIME, None)
1877 ntime_merged = utils.MergeTime(ntime)
1878 except (ValueError, TypeError):
1879 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1882 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1883 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1884 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1885 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1889 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1890 "Node time diverges by at least %s from master node time",
1893 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1894 """Check the node LVM results.
1896 @type ninfo: L{objects.Node}
1897 @param ninfo: the node to check
1898 @param nresult: the remote results for the node
1899 @param vg_name: the configured VG name
1906 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1908 # checks vg existence and size > 20G
1909 vglist = nresult.get(constants.NV_VGLIST, None)
1911 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1913 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1914 constants.MIN_VG_SIZE)
1915 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1918 pvlist = nresult.get(constants.NV_PVLIST, None)
1919 test = pvlist is None
1920 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1922 # check that ':' is not present in PV names, since it's a
1923 # special character for lvcreate (denotes the range of PEs to
1925 for _, pvname, owner_vg in pvlist:
1926 test = ":" in pvname
1927 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1928 " '%s' of VG '%s'", pvname, owner_vg)
1930 def _VerifyNodeBridges(self, ninfo, nresult, bridges):
1931 """Check the node bridges.
1933 @type ninfo: L{objects.Node}
1934 @param ninfo: the node to check
1935 @param nresult: the remote results for the node
1936 @param bridges: the expected list of bridges
1943 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1945 missing = nresult.get(constants.NV_BRIDGES, None)
1946 test = not isinstance(missing, list)
1947 _ErrorIf(test, self.ENODENET, node,
1948 "did not return valid bridge information")
1950 _ErrorIf(bool(missing), self.ENODENET, node, "missing bridges: %s" %
1951 utils.CommaJoin(sorted(missing)))
1953 def _VerifyNodeNetwork(self, ninfo, nresult):
1954 """Check the node network connectivity results.
1956 @type ninfo: L{objects.Node}
1957 @param ninfo: the node to check
1958 @param nresult: the remote results for the node
1962 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1964 test = constants.NV_NODELIST not in nresult
1965 _ErrorIf(test, self.ENODESSH, node,
1966 "node hasn't returned node ssh connectivity data")
1968 if nresult[constants.NV_NODELIST]:
1969 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1970 _ErrorIf(True, self.ENODESSH, node,
1971 "ssh communication with node '%s': %s", a_node, a_msg)
1973 test = constants.NV_NODENETTEST not in nresult
1974 _ErrorIf(test, self.ENODENET, node,
1975 "node hasn't returned node tcp connectivity data")
1977 if nresult[constants.NV_NODENETTEST]:
1978 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1980 _ErrorIf(True, self.ENODENET, node,
1981 "tcp communication with node '%s': %s",
1982 anode, nresult[constants.NV_NODENETTEST][anode])
1984 test = constants.NV_MASTERIP not in nresult
1985 _ErrorIf(test, self.ENODENET, node,
1986 "node hasn't returned node master IP reachability data")
1988 if not nresult[constants.NV_MASTERIP]:
1989 if node == self.master_node:
1990 msg = "the master node cannot reach the master IP (not configured?)"
1992 msg = "cannot reach the master IP"
1993 _ErrorIf(True, self.ENODENET, node, msg)
1995 def _VerifyInstance(self, instance, instanceconfig, node_image,
1997 """Verify an instance.
1999 This function checks to see if the required block devices are
2000 available on the instance's node.
2003 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2004 node_current = instanceconfig.primary_node
2006 node_vol_should = {}
2007 instanceconfig.MapLVsByNode(node_vol_should)
2009 for node in node_vol_should:
2010 n_img = node_image[node]
2011 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2012 # ignore missing volumes on offline or broken nodes
2014 for volume in node_vol_should[node]:
2015 test = volume not in n_img.volumes
2016 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
2017 "volume %s missing on node %s", volume, node)
2019 if instanceconfig.admin_up:
2020 pri_img = node_image[node_current]
2021 test = instance not in pri_img.instances and not pri_img.offline
2022 _ErrorIf(test, self.EINSTANCEDOWN, instance,
2023 "instance not running on its primary node %s",
2026 diskdata = [(nname, success, status, idx)
2027 for (nname, disks) in diskstatus.items()
2028 for idx, (success, status) in enumerate(disks)]
2030 for nname, success, bdev_status, idx in diskdata:
2031 # the 'ghost node' construction in Exec() ensures that we have a
2033 snode = node_image[nname]
2034 bad_snode = snode.ghost or snode.offline
2035 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
2036 self.EINSTANCEFAULTYDISK, instance,
2037 "couldn't retrieve status for disk/%s on %s: %s",
2038 idx, nname, bdev_status)
2039 _ErrorIf((instanceconfig.admin_up and success and
2040 bdev_status.ldisk_status == constants.LDS_FAULTY),
2041 self.EINSTANCEFAULTYDISK, instance,
2042 "disk/%s on %s is faulty", idx, nname)
2044 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
2045 """Verify if there are any unknown volumes in the cluster.
2047 The .os, .swap and backup volumes are ignored. All other volumes are
2048 reported as unknown.
2050 @type reserved: L{ganeti.utils.FieldSet}
2051 @param reserved: a FieldSet of reserved volume names
2054 for node, n_img in node_image.items():
2055 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2056 # skip non-healthy nodes
2058 for volume in n_img.volumes:
2059 test = ((node not in node_vol_should or
2060 volume not in node_vol_should[node]) and
2061 not reserved.Matches(volume))
2062 self._ErrorIf(test, self.ENODEORPHANLV, node,
2063 "volume %s is unknown", volume)
2065 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
2066 """Verify N+1 Memory Resilience.
2068 Check that if one single node dies we can still start all the
2069 instances it was primary for.
2072 cluster_info = self.cfg.GetClusterInfo()
2073 for node, n_img in node_image.items():
2074 # This code checks that every node which is now listed as
2075 # secondary has enough memory to host all instances it is
2076 # supposed to should a single other node in the cluster fail.
2077 # FIXME: not ready for failover to an arbitrary node
2078 # FIXME: does not support file-backed instances
2079 # WARNING: we currently take into account down instances as well
2080 # as up ones, considering that even if they're down someone
2081 # might want to start them even in the event of a node failure.
2083 # we're skipping offline nodes from the N+1 warning, since
2084 # most likely we don't have good memory infromation from them;
2085 # we already list instances living on such nodes, and that's
2088 for prinode, instances in n_img.sbp.items():
2090 for instance in instances:
2091 bep = cluster_info.FillBE(instance_cfg[instance])
2092 if bep[constants.BE_AUTO_BALANCE]:
2093 needed_mem += bep[constants.BE_MEMORY]
2094 test = n_img.mfree < needed_mem
2095 self._ErrorIf(test, self.ENODEN1, node,
2096 "not enough memory to accomodate instance failovers"
2097 " should node %s fail (%dMiB needed, %dMiB available)",
2098 prinode, needed_mem, n_img.mfree)
2101 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
2102 (files_all, files_all_opt, files_mc, files_vm)):
2103 """Verifies file checksums collected from all nodes.
2105 @param errorif: Callback for reporting errors
2106 @param nodeinfo: List of L{objects.Node} objects
2107 @param master_node: Name of master node
2108 @param all_nvinfo: RPC results
2111 node_names = frozenset(node.name for node in nodeinfo if not node.offline)
2113 assert master_node in node_names
2114 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
2115 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
2116 "Found file listed in more than one file list"
2118 # Define functions determining which nodes to consider for a file
2119 file2nodefn = dict([(filename, fn)
2120 for (files, fn) in [(files_all, None),
2121 (files_all_opt, None),
2122 (files_mc, lambda node: (node.master_candidate or
2123 node.name == master_node)),
2124 (files_vm, lambda node: node.vm_capable)]
2125 for filename in files])
2127 fileinfo = dict((filename, {}) for filename in file2nodefn.keys())
2129 for node in nodeinfo:
2133 nresult = all_nvinfo[node.name]
2135 if nresult.fail_msg or not nresult.payload:
2138 node_files = nresult.payload.get(constants.NV_FILELIST, None)
2140 test = not (node_files and isinstance(node_files, dict))
2141 errorif(test, cls.ENODEFILECHECK, node.name,
2142 "Node did not return file checksum data")
2146 for (filename, checksum) in node_files.items():
2147 # Check if the file should be considered for a node
2148 fn = file2nodefn[filename]
2149 if fn is None or fn(node):
2150 fileinfo[filename].setdefault(checksum, set()).add(node.name)
2152 for (filename, checksums) in fileinfo.items():
2153 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
2155 # Nodes having the file
2156 with_file = frozenset(node_name
2157 for nodes in fileinfo[filename].values()
2158 for node_name in nodes)
2160 # Nodes missing file
2161 missing_file = node_names - with_file
2163 if filename in files_all_opt:
2165 errorif(missing_file and missing_file != node_names,
2166 cls.ECLUSTERFILECHECK, None,
2167 "File %s is optional, but it must exist on all or no"
2168 " nodes (not found on %s)",
2169 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
2171 errorif(missing_file, cls.ECLUSTERFILECHECK, None,
2172 "File %s is missing from node(s) %s", filename,
2173 utils.CommaJoin(utils.NiceSort(missing_file)))
2175 # See if there are multiple versions of the file
2176 test = len(checksums) > 1
2178 variants = ["variant %s on %s" %
2179 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
2180 for (idx, (checksum, nodes)) in
2181 enumerate(sorted(checksums.items()))]
2185 errorif(test, cls.ECLUSTERFILECHECK, None,
2186 "File %s found with %s different checksums (%s)",
2187 filename, len(checksums), "; ".join(variants))
2189 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
2191 """Verifies and the node DRBD status.
2193 @type ninfo: L{objects.Node}
2194 @param ninfo: the node to check
2195 @param nresult: the remote results for the node
2196 @param instanceinfo: the dict of instances
2197 @param drbd_helper: the configured DRBD usermode helper
2198 @param drbd_map: the DRBD map as returned by
2199 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
2203 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2206 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
2207 test = (helper_result == None)
2208 _ErrorIf(test, self.ENODEDRBDHELPER, node,
2209 "no drbd usermode helper returned")
2211 status, payload = helper_result
2213 _ErrorIf(test, self.ENODEDRBDHELPER, node,
2214 "drbd usermode helper check unsuccessful: %s", payload)
2215 test = status and (payload != drbd_helper)
2216 _ErrorIf(test, self.ENODEDRBDHELPER, node,
2217 "wrong drbd usermode helper: %s", payload)
2219 # compute the DRBD minors
2221 for minor, instance in drbd_map[node].items():
2222 test = instance not in instanceinfo
2223 _ErrorIf(test, self.ECLUSTERCFG, None,
2224 "ghost instance '%s' in temporary DRBD map", instance)
2225 # ghost instance should not be running, but otherwise we
2226 # don't give double warnings (both ghost instance and
2227 # unallocated minor in use)
2229 node_drbd[minor] = (instance, False)
2231 instance = instanceinfo[instance]
2232 node_drbd[minor] = (instance.name, instance.admin_up)
2234 # and now check them
2235 used_minors = nresult.get(constants.NV_DRBDLIST, [])
2236 test = not isinstance(used_minors, (tuple, list))
2237 _ErrorIf(test, self.ENODEDRBD, node,
2238 "cannot parse drbd status file: %s", str(used_minors))
2240 # we cannot check drbd status
2243 for minor, (iname, must_exist) in node_drbd.items():
2244 test = minor not in used_minors and must_exist
2245 _ErrorIf(test, self.ENODEDRBD, node,
2246 "drbd minor %d of instance %s is not active", minor, iname)
2247 for minor in used_minors:
2248 test = minor not in node_drbd
2249 _ErrorIf(test, self.ENODEDRBD, node,
2250 "unallocated drbd minor %d is in use", minor)
2252 def _UpdateNodeOS(self, ninfo, nresult, nimg):
2253 """Builds the node OS structures.
2255 @type ninfo: L{objects.Node}
2256 @param ninfo: the node to check
2257 @param nresult: the remote results for the node
2258 @param nimg: the node image object
2262 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2264 remote_os = nresult.get(constants.NV_OSLIST, None)
2265 test = (not isinstance(remote_os, list) or
2266 not compat.all(isinstance(v, list) and len(v) == 7
2267 for v in remote_os))
2269 _ErrorIf(test, self.ENODEOS, node,
2270 "node hasn't returned valid OS data")
2279 for (name, os_path, status, diagnose,
2280 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
2282 if name not in os_dict:
2285 # parameters is a list of lists instead of list of tuples due to
2286 # JSON lacking a real tuple type, fix it:
2287 parameters = [tuple(v) for v in parameters]
2288 os_dict[name].append((os_path, status, diagnose,
2289 set(variants), set(parameters), set(api_ver)))
2291 nimg.oslist = os_dict
2293 def _VerifyNodeOS(self, ninfo, nimg, base):
2294 """Verifies the node OS list.
2296 @type ninfo: L{objects.Node}
2297 @param ninfo: the node to check
2298 @param nimg: the node image object
2299 @param base: the 'template' node we match against (e.g. from the master)
2303 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2305 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
2307 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
2308 for os_name, os_data in nimg.oslist.items():
2309 assert os_data, "Empty OS status for OS %s?!" % os_name
2310 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
2311 _ErrorIf(not f_status, self.ENODEOS, node,
2312 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
2313 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
2314 "OS '%s' has multiple entries (first one shadows the rest): %s",
2315 os_name, utils.CommaJoin([v[0] for v in os_data]))
2316 # comparisons with the 'base' image
2317 test = os_name not in base.oslist
2318 _ErrorIf(test, self.ENODEOS, node,
2319 "Extra OS %s not present on reference node (%s)",
2323 assert base.oslist[os_name], "Base node has empty OS status?"
2324 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
2326 # base OS is invalid, skipping
2328 for kind, a, b in [("API version", f_api, b_api),
2329 ("variants list", f_var, b_var),
2330 ("parameters", beautify_params(f_param),
2331 beautify_params(b_param))]:
2332 _ErrorIf(a != b, self.ENODEOS, node,
2333 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
2334 kind, os_name, base.name,
2335 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
2337 # check any missing OSes
2338 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
2339 _ErrorIf(missing, self.ENODEOS, node,
2340 "OSes present on reference node %s but missing on this node: %s",
2341 base.name, utils.CommaJoin(missing))
2343 def _VerifyOob(self, ninfo, nresult):
2344 """Verifies out of band functionality of a node.
2346 @type ninfo: L{objects.Node}
2347 @param ninfo: the node to check
2348 @param nresult: the remote results for the node
2352 # We just have to verify the paths on master and/or master candidates
2353 # as the oob helper is invoked on the master
2354 if ((ninfo.master_candidate or ninfo.master_capable) and
2355 constants.NV_OOB_PATHS in nresult):
2356 for path_result in nresult[constants.NV_OOB_PATHS]:
2357 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
2359 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
2360 """Verifies and updates the node volume data.
2362 This function will update a L{NodeImage}'s internal structures
2363 with data from the remote call.
2365 @type ninfo: L{objects.Node}
2366 @param ninfo: the node to check
2367 @param nresult: the remote results for the node
2368 @param nimg: the node image object
2369 @param vg_name: the configured VG name
2373 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2375 nimg.lvm_fail = True
2376 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2379 elif isinstance(lvdata, basestring):
2380 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
2381 utils.SafeEncode(lvdata))
2382 elif not isinstance(lvdata, dict):
2383 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
2385 nimg.volumes = lvdata
2386 nimg.lvm_fail = False
2388 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2389 """Verifies and updates the node instance list.
2391 If the listing was successful, then updates this node's instance
2392 list. Otherwise, it marks the RPC call as failed for the instance
2395 @type ninfo: L{objects.Node}
2396 @param ninfo: the node to check
2397 @param nresult: the remote results for the node
2398 @param nimg: the node image object
2401 idata = nresult.get(constants.NV_INSTANCELIST, None)
2402 test = not isinstance(idata, list)
2403 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
2404 " (instancelist): %s", utils.SafeEncode(str(idata)))
2406 nimg.hyp_fail = True
2408 nimg.instances = idata
2410 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2411 """Verifies and computes a node information map
2413 @type ninfo: L{objects.Node}
2414 @param ninfo: the node to check
2415 @param nresult: the remote results for the node
2416 @param nimg: the node image object
2417 @param vg_name: the configured VG name
2421 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2423 # try to read free memory (from the hypervisor)
2424 hv_info = nresult.get(constants.NV_HVINFO, None)
2425 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2426 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
2429 nimg.mfree = int(hv_info["memory_free"])
2430 except (ValueError, TypeError):
2431 _ErrorIf(True, self.ENODERPC, node,
2432 "node returned invalid nodeinfo, check hypervisor")
2434 # FIXME: devise a free space model for file based instances as well
2435 if vg_name is not None:
2436 test = (constants.NV_VGLIST not in nresult or
2437 vg_name not in nresult[constants.NV_VGLIST])
2438 _ErrorIf(test, self.ENODELVM, node,
2439 "node didn't return data for the volume group '%s'"
2440 " - it is either missing or broken", vg_name)
2443 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2444 except (ValueError, TypeError):
2445 _ErrorIf(True, self.ENODERPC, node,
2446 "node returned invalid LVM info, check LVM status")
2448 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2449 """Gets per-disk status information for all instances.
2451 @type nodelist: list of strings
2452 @param nodelist: Node names
2453 @type node_image: dict of (name, L{objects.Node})
2454 @param node_image: Node objects
2455 @type instanceinfo: dict of (name, L{objects.Instance})
2456 @param instanceinfo: Instance objects
2457 @rtype: {instance: {node: [(succes, payload)]}}
2458 @return: a dictionary of per-instance dictionaries with nodes as
2459 keys and disk information as values; the disk information is a
2460 list of tuples (success, payload)
2463 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2466 node_disks_devonly = {}
2467 diskless_instances = set()
2468 diskless = constants.DT_DISKLESS
2470 for nname in nodelist:
2471 node_instances = list(itertools.chain(node_image[nname].pinst,
2472 node_image[nname].sinst))
2473 diskless_instances.update(inst for inst in node_instances
2474 if instanceinfo[inst].disk_template == diskless)
2475 disks = [(inst, disk)
2476 for inst in node_instances
2477 for disk in instanceinfo[inst].disks]
2480 # No need to collect data
2483 node_disks[nname] = disks
2485 # Creating copies as SetDiskID below will modify the objects and that can
2486 # lead to incorrect data returned from nodes
2487 devonly = [dev.Copy() for (_, dev) in disks]
2490 self.cfg.SetDiskID(dev, nname)
2492 node_disks_devonly[nname] = devonly
2494 assert len(node_disks) == len(node_disks_devonly)
2496 # Collect data from all nodes with disks
2497 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2500 assert len(result) == len(node_disks)
2504 for (nname, nres) in result.items():
2505 disks = node_disks[nname]
2508 # No data from this node
2509 data = len(disks) * [(False, "node offline")]
2512 _ErrorIf(msg, self.ENODERPC, nname,
2513 "while getting disk information: %s", msg)
2515 # No data from this node
2516 data = len(disks) * [(False, msg)]
2519 for idx, i in enumerate(nres.payload):
2520 if isinstance(i, (tuple, list)) and len(i) == 2:
2523 logging.warning("Invalid result from node %s, entry %d: %s",
2525 data.append((False, "Invalid result from the remote node"))
2527 for ((inst, _), status) in zip(disks, data):
2528 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2530 # Add empty entries for diskless instances.
2531 for inst in diskless_instances:
2532 assert inst not in instdisk
2535 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2536 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2537 compat.all(isinstance(s, (tuple, list)) and
2538 len(s) == 2 for s in statuses)
2539 for inst, nnames in instdisk.items()
2540 for nname, statuses in nnames.items())
2541 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2545 def BuildHooksEnv(self):
2548 Cluster-Verify hooks just ran in the post phase and their failure makes
2549 the output be logged in the verify output and the verification to fail.
2553 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2556 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
2557 for node in self.my_node_info.values())
2561 def BuildHooksNodes(self):
2562 """Build hooks nodes.
2565 return ([], self.my_node_names)
2567 def Exec(self, feedback_fn):
2568 """Verify integrity of the node group, performing various test on nodes.
2571 # This method has too many local variables. pylint: disable=R0914
2572 feedback_fn("* Verifying group '%s'" % self.group_info.name)
2574 if not self.my_node_names:
2576 feedback_fn("* Empty node group, skipping verification")
2580 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2581 verbose = self.op.verbose
2582 self._feedback_fn = feedback_fn
2584 vg_name = self.cfg.GetVGName()
2585 drbd_helper = self.cfg.GetDRBDHelper()
2586 cluster = self.cfg.GetClusterInfo()
2587 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2588 hypervisors = cluster.enabled_hypervisors
2589 node_data_list = [self.my_node_info[name] for name in self.my_node_names]
2591 i_non_redundant = [] # Non redundant instances
2592 i_non_a_balanced = [] # Non auto-balanced instances
2593 n_offline = 0 # Count of offline nodes
2594 n_drained = 0 # Count of nodes being drained
2595 node_vol_should = {}
2597 # FIXME: verify OS list
2600 filemap = _ComputeAncillaryFiles(cluster, False)
2602 # do local checksums
2603 master_node = self.master_node = self.cfg.GetMasterNode()
2604 master_ip = self.cfg.GetMasterIP()
2606 feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
2608 # We will make nodes contact all nodes in their group, and one node from
2609 # every other group.
2610 # TODO: should it be a *random* node, different every time?
2611 online_nodes = [node.name for node in node_data_list if not node.offline]
2612 other_group_nodes = {}
2614 for name in sorted(self.all_node_info):
2615 node = self.all_node_info[name]
2616 if (node.group not in other_group_nodes
2617 and node.group != self.group_uuid
2618 and not node.offline):
2619 other_group_nodes[node.group] = node.name
2621 node_verify_param = {
2622 constants.NV_FILELIST:
2623 utils.UniqueSequence(filename
2624 for files in filemap
2625 for filename in files),
2626 constants.NV_NODELIST: online_nodes + other_group_nodes.values(),
2627 constants.NV_HYPERVISOR: hypervisors,
2628 constants.NV_HVPARAMS:
2629 _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
2630 constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip)
2631 for node in node_data_list
2632 if not node.offline],
2633 constants.NV_INSTANCELIST: hypervisors,
2634 constants.NV_VERSION: None,
2635 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2636 constants.NV_NODESETUP: None,
2637 constants.NV_TIME: None,
2638 constants.NV_MASTERIP: (master_node, master_ip),
2639 constants.NV_OSLIST: None,
2640 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2643 if vg_name is not None:
2644 node_verify_param[constants.NV_VGLIST] = None
2645 node_verify_param[constants.NV_LVLIST] = vg_name
2646 node_verify_param[constants.NV_PVLIST] = [vg_name]
2647 node_verify_param[constants.NV_DRBDLIST] = None
2650 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2653 # FIXME: this needs to be changed per node-group, not cluster-wide
2655 default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
2656 if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2657 bridges.add(default_nicpp[constants.NIC_LINK])
2658 for instance in self.my_inst_info.values():
2659 for nic in instance.nics:
2660 full_nic = cluster.SimpleFillNIC(nic.nicparams)
2661 if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2662 bridges.add(full_nic[constants.NIC_LINK])
2665 node_verify_param[constants.NV_BRIDGES] = list(bridges)
2667 # Build our expected cluster state
2668 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2670 vm_capable=node.vm_capable))
2671 for node in node_data_list)
2675 for node in self.all_node_info.values():
2676 path = _SupportsOob(self.cfg, node)
2677 if path and path not in oob_paths:
2678 oob_paths.append(path)
2681 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2683 for instance in self.my_inst_names:
2684 inst_config = self.my_inst_info[instance]
2686 for nname in inst_config.all_nodes:
2687 if nname not in node_image:
2688 gnode = self.NodeImage(name=nname)
2689 gnode.ghost = (nname not in self.all_node_info)
2690 node_image[nname] = gnode
2692 inst_config.MapLVsByNode(node_vol_should)
2694 pnode = inst_config.primary_node
2695 node_image[pnode].pinst.append(instance)
2697 for snode in inst_config.secondary_nodes:
2698 nimg = node_image[snode]
2699 nimg.sinst.append(instance)
2700 if pnode not in nimg.sbp:
2701 nimg.sbp[pnode] = []
2702 nimg.sbp[pnode].append(instance)
2704 # At this point, we have the in-memory data structures complete,
2705 # except for the runtime information, which we'll gather next
2707 # Due to the way our RPC system works, exact response times cannot be
2708 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2709 # time before and after executing the request, we can at least have a time
2711 nvinfo_starttime = time.time()
2712 all_nvinfo = self.rpc.call_node_verify(self.my_node_names,
2714 self.cfg.GetClusterName())
2715 nvinfo_endtime = time.time()
2717 if self.extra_lv_nodes and vg_name is not None:
2719 self.rpc.call_node_verify(self.extra_lv_nodes,
2720 {constants.NV_LVLIST: vg_name},
2721 self.cfg.GetClusterName())
2723 extra_lv_nvinfo = {}
2725 all_drbd_map = self.cfg.ComputeDRBDMap()
2727 feedback_fn("* Gathering disk information (%s nodes)" %
2728 len(self.my_node_names))
2729 instdisk = self._CollectDiskInfo(self.my_node_names, node_image,
2732 feedback_fn("* Verifying configuration file consistency")
2734 # If not all nodes are being checked, we need to make sure the master node
2735 # and a non-checked vm_capable node are in the list.
2736 absent_nodes = set(self.all_node_info).difference(self.my_node_info)
2738 vf_nvinfo = all_nvinfo.copy()
2739 vf_node_info = list(self.my_node_info.values())
2740 additional_nodes = []
2741 if master_node not in self.my_node_info:
2742 additional_nodes.append(master_node)
2743 vf_node_info.append(self.all_node_info[master_node])
2744 # Add the first vm_capable node we find which is not included
2745 for node in absent_nodes:
2746 nodeinfo = self.all_node_info[node]
2747 if nodeinfo.vm_capable and not nodeinfo.offline:
2748 additional_nodes.append(node)
2749 vf_node_info.append(self.all_node_info[node])
2751 key = constants.NV_FILELIST
2752 vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes,
2753 {key: node_verify_param[key]},
2754 self.cfg.GetClusterName()))
2756 vf_nvinfo = all_nvinfo
2757 vf_node_info = self.my_node_info.values()
2759 self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap)
2761 feedback_fn("* Verifying node status")
2765 for node_i in node_data_list:
2767 nimg = node_image[node]
2771 feedback_fn("* Skipping offline node %s" % (node,))
2775 if node == master_node:
2777 elif node_i.master_candidate:
2778 ntype = "master candidate"
2779 elif node_i.drained:
2785 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2787 msg = all_nvinfo[node].fail_msg
2788 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2790 nimg.rpc_fail = True
2793 nresult = all_nvinfo[node].payload
2795 nimg.call_ok = self._VerifyNode(node_i, nresult)
2796 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2797 self._VerifyNodeNetwork(node_i, nresult)
2798 self._VerifyOob(node_i, nresult)
2801 self._VerifyNodeLVM(node_i, nresult, vg_name)
2802 self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper,
2805 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2806 self._UpdateNodeInstances(node_i, nresult, nimg)
2807 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2808 self._UpdateNodeOS(node_i, nresult, nimg)
2810 if not nimg.os_fail:
2811 if refos_img is None:
2813 self._VerifyNodeOS(node_i, nimg, refos_img)
2814 self._VerifyNodeBridges(node_i, nresult, bridges)
2816 # Check whether all running instancies are primary for the node. (This
2817 # can no longer be done from _VerifyInstance below, since some of the
2818 # wrong instances could be from other node groups.)
2819 non_primary_inst = set(nimg.instances).difference(nimg.pinst)
2821 for inst in non_primary_inst:
2822 test = inst in self.all_inst_info
2823 _ErrorIf(test, self.EINSTANCEWRONGNODE, inst,
2824 "instance should not run on node %s", node_i.name)
2825 _ErrorIf(not test, self.ENODEORPHANINSTANCE, node_i.name,
2826 "node is running unknown instance %s", inst)
2828 for node, result in extra_lv_nvinfo.items():
2829 self._UpdateNodeVolumes(self.all_node_info[node], result.payload,
2830 node_image[node], vg_name)
2832 feedback_fn("* Verifying instance status")
2833 for instance in self.my_inst_names:
2835 feedback_fn("* Verifying instance %s" % instance)
2836 inst_config = self.my_inst_info[instance]
2837 self._VerifyInstance(instance, inst_config, node_image,
2839 inst_nodes_offline = []
2841 pnode = inst_config.primary_node
2842 pnode_img = node_image[pnode]
2843 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2844 self.ENODERPC, pnode, "instance %s, connection to"
2845 " primary node failed", instance)
2847 _ErrorIf(inst_config.admin_up and pnode_img.offline,
2848 self.EINSTANCEBADNODE, instance,
2849 "instance is marked as running and lives on offline node %s",
2850 inst_config.primary_node)
2852 # If the instance is non-redundant we cannot survive losing its primary
2853 # node, so we are not N+1 compliant. On the other hand we have no disk
2854 # templates with more than one secondary so that situation is not well
2856 # FIXME: does not support file-backed instances
2857 if not inst_config.secondary_nodes:
2858 i_non_redundant.append(instance)
2860 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2861 instance, "instance has multiple secondary nodes: %s",
2862 utils.CommaJoin(inst_config.secondary_nodes),
2863 code=self.ETYPE_WARNING)
2865 if inst_config.disk_template in constants.DTS_INT_MIRROR:
2866 pnode = inst_config.primary_node
2867 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2868 instance_groups = {}
2870 for node in instance_nodes:
2871 instance_groups.setdefault(self.all_node_info[node].group,
2875 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2876 # Sort so that we always list the primary node first.
2877 for group, nodes in sorted(instance_groups.items(),
2878 key=lambda (_, nodes): pnode in nodes,
2881 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2882 instance, "instance has primary and secondary nodes in"
2883 " different groups: %s", utils.CommaJoin(pretty_list),
2884 code=self.ETYPE_WARNING)
2886 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2887 i_non_a_balanced.append(instance)
2889 for snode in inst_config.secondary_nodes:
2890 s_img = node_image[snode]
2891 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2892 "instance %s, connection to secondary node failed", instance)
2895 inst_nodes_offline.append(snode)
2897 # warn that the instance lives on offline nodes
2898 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2899 "instance has offline secondary node(s) %s",
2900 utils.CommaJoin(inst_nodes_offline))
2901 # ... or ghost/non-vm_capable nodes
2902 for node in inst_config.all_nodes:
2903 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2904 "instance lives on ghost node %s", node)
2905 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2906 instance, "instance lives on non-vm_capable node %s", node)
2908 feedback_fn("* Verifying orphan volumes")
2909 reserved = utils.FieldSet(*cluster.reserved_lvs)
2911 # We will get spurious "unknown volume" warnings if any node of this group
2912 # is secondary for an instance whose primary is in another group. To avoid
2913 # them, we find these instances and add their volumes to node_vol_should.
2914 for inst in self.all_inst_info.values():
2915 for secondary in inst.secondary_nodes:
2916 if (secondary in self.my_node_info
2917 and inst.name not in self.my_inst_info):
2918 inst.MapLVsByNode(node_vol_should)
2921 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2923 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2924 feedback_fn("* Verifying N+1 Memory redundancy")
2925 self._VerifyNPlusOneMemory(node_image, self.my_inst_info)
2927 feedback_fn("* Other Notes")
2929 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2930 % len(i_non_redundant))
2932 if i_non_a_balanced:
2933 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2934 % len(i_non_a_balanced))
2937 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2940 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2944 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2945 """Analyze the post-hooks' result
2947 This method analyses the hook result, handles it, and sends some
2948 nicely-formatted feedback back to the user.
2950 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2951 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2952 @param hooks_results: the results of the multi-node hooks rpc call
2953 @param feedback_fn: function used send feedback back to the caller
2954 @param lu_result: previous Exec result
2955 @return: the new Exec result, based on the previous result
2959 # We only really run POST phase hooks, only for non-empty groups,
2960 # and are only interested in their results
2961 if not self.my_node_names:
2964 elif phase == constants.HOOKS_PHASE_POST:
2965 # Used to change hooks' output to proper indentation
2966 feedback_fn("* Hooks Results")
2967 assert hooks_results, "invalid result from hooks"
2969 for node_name in hooks_results:
2970 res = hooks_results[node_name]
2972 test = msg and not res.offline
2973 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2974 "Communication failure in hooks execution: %s", msg)
2975 if res.offline or msg:
2976 # No need to investigate payload if node is offline or gave an error.
2977 # override manually lu_result here as _ErrorIf only
2978 # overrides self.bad
2981 for script, hkr, output in res.payload:
2982 test = hkr == constants.HKR_FAIL
2983 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2984 "Script %s failed, output:", script)
2986 output = self._HOOKS_INDENT_RE.sub(" ", output)
2987 feedback_fn("%s" % output)
2993 class LUClusterVerifyDisks(NoHooksLU):
2994 """Verifies the cluster disks status.
2999 def ExpandNames(self):
3000 self.share_locks = _ShareAll()
3001 self.needed_locks = {
3002 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3005 def Exec(self, feedback_fn):
3006 group_names = self.owned_locks(locking.LEVEL_NODEGROUP)
3008 # Submit one instance of L{opcodes.OpGroupVerifyDisks} per node group
3009 return ResultWithJobs([[opcodes.OpGroupVerifyDisks(group_name=group)]
3010 for group in group_names])
3013 class LUGroupVerifyDisks(NoHooksLU):
3014 """Verifies the status of all disks in a node group.
3019 def ExpandNames(self):
3020 # Raises errors.OpPrereqError on its own if group can't be found
3021 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
3023 self.share_locks = _ShareAll()
3024 self.needed_locks = {
3025 locking.LEVEL_INSTANCE: [],
3026 locking.LEVEL_NODEGROUP: [],
3027 locking.LEVEL_NODE: [],
3030 def DeclareLocks(self, level):
3031 if level == locking.LEVEL_INSTANCE:
3032 assert not self.needed_locks[locking.LEVEL_INSTANCE]
3034 # Lock instances optimistically, needs verification once node and group
3035 # locks have been acquired
3036 self.needed_locks[locking.LEVEL_INSTANCE] = \
3037 self.cfg.GetNodeGroupInstances(self.group_uuid)
3039 elif level == locking.LEVEL_NODEGROUP:
3040 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
3042 self.needed_locks[locking.LEVEL_NODEGROUP] = \
3043 set([self.group_uuid] +
3044 # Lock all groups used by instances optimistically; this requires
3045 # going via the node before it's locked, requiring verification
3048 for instance_name in self.owned_locks(locking.LEVEL_INSTANCE)
3049 for group_uuid in self.cfg.GetInstanceNodeGroups(instance_name)])
3051 elif level == locking.LEVEL_NODE:
3052 # This will only lock the nodes in the group to be verified which contain
3054 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
3055 self._LockInstancesNodes()
3057 # Lock all nodes in group to be verified
3058 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
3059 member_nodes = self.cfg.GetNodeGroup(self.group_uuid).members
3060 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
3062 def CheckPrereq(self):
3063 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
3064 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
3065 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
3067 assert self.group_uuid in owned_groups
3069 # Check if locked instances are still correct
3070 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
3072 # Get instance information
3073 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
3075 # Check if node groups for locked instances are still correct
3076 for (instance_name, inst) in self.instances.items():
3077 assert owned_nodes.issuperset(inst.all_nodes), \
3078 "Instance %s's nodes changed while we kept the lock" % instance_name
3080 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
3083 assert self.group_uuid in inst_groups, \
3084 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
3086 def Exec(self, feedback_fn):
3087 """Verify integrity of cluster disks.
3089 @rtype: tuple of three items
3090 @return: a tuple of (dict of node-to-node_error, list of instances
3091 which need activate-disks, dict of instance: (node, volume) for
3096 res_instances = set()
3099 nv_dict = _MapInstanceDisksToNodes([inst
3100 for inst in self.instances.values()
3104 nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &
3105 set(self.cfg.GetVmCapableNodeList()))
3107 node_lvs = self.rpc.call_lv_list(nodes, [])
3109 for (node, node_res) in node_lvs.items():
3110 if node_res.offline:
3113 msg = node_res.fail_msg
3115 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
3116 res_nodes[node] = msg
3119 for lv_name, (_, _, lv_online) in node_res.payload.items():
3120 inst = nv_dict.pop((node, lv_name), None)
3121 if not (lv_online or inst is None):
3122 res_instances.add(inst)
3124 # any leftover items in nv_dict are missing LVs, let's arrange the data
3126 for key, inst in nv_dict.iteritems():
3127 res_missing.setdefault(inst, []).append(key)
3129 return (res_nodes, list(res_instances), res_missing)
3132 class LUClusterRepairDiskSizes(NoHooksLU):
3133 """Verifies the cluster disks sizes.
3138 def ExpandNames(self):
3139 if self.op.instances:
3140 self.wanted_names = _GetWantedInstances(self, self.op.instances)
3141 self.needed_locks = {
3142 locking.LEVEL_NODE: [],
3143 locking.LEVEL_INSTANCE: self.wanted_names,
3145 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3147 self.wanted_names = None
3148 self.needed_locks = {
3149 locking.LEVEL_NODE: locking.ALL_SET,
3150 locking.LEVEL_INSTANCE: locking.ALL_SET,
3152 self.share_locks = _ShareAll()
3154 def DeclareLocks(self, level):
3155 if level == locking.LEVEL_NODE and self.wanted_names is not None:
3156 self._LockInstancesNodes(primary_only=True)
3158 def CheckPrereq(self):
3159 """Check prerequisites.
3161 This only checks the optional instance list against the existing names.
3164 if self.wanted_names is None:
3165 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
3167 self.wanted_instances = \
3168 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
3170 def _EnsureChildSizes(self, disk):
3171 """Ensure children of the disk have the needed disk size.
3173 This is valid mainly for DRBD8 and fixes an issue where the
3174 children have smaller disk size.
3176 @param disk: an L{ganeti.objects.Disk} object
3179 if disk.dev_type == constants.LD_DRBD8:
3180 assert disk.children, "Empty children for DRBD8?"
3181 fchild = disk.children[0]
3182 mismatch = fchild.size < disk.size
3184 self.LogInfo("Child disk has size %d, parent %d, fixing",
3185 fchild.size, disk.size)
3186 fchild.size = disk.size
3188 # and we recurse on this child only, not on the metadev
3189 return self._EnsureChildSizes(fchild) or mismatch
3193 def Exec(self, feedback_fn):
3194 """Verify the size of cluster disks.
3197 # TODO: check child disks too
3198 # TODO: check differences in size between primary/secondary nodes
3200 for instance in self.wanted_instances:
3201 pnode = instance.primary_node
3202 if pnode not in per_node_disks:
3203 per_node_disks[pnode] = []
3204 for idx, disk in enumerate(instance.disks):
3205 per_node_disks[pnode].append((instance, idx, disk))
3208 for node, dskl in per_node_disks.items():
3209 newl = [v[2].Copy() for v in dskl]
3211 self.cfg.SetDiskID(dsk, node)
3212 result = self.rpc.call_blockdev_getsize(node, newl)
3214 self.LogWarning("Failure in blockdev_getsize call to node"
3215 " %s, ignoring", node)
3217 if len(result.payload) != len(dskl):
3218 logging.warning("Invalid result from node %s: len(dksl)=%d,"
3219 " result.payload=%s", node, len(dskl), result.payload)
3220 self.LogWarning("Invalid result from node %s, ignoring node results",
3223 for ((instance, idx, disk), size) in zip(dskl, result.payload):
3225 self.LogWarning("Disk %d of instance %s did not return size"
3226 " information, ignoring", idx, instance.name)
3228 if not isinstance(size, (int, long)):
3229 self.LogWarning("Disk %d of instance %s did not return valid"
3230 " size information, ignoring", idx, instance.name)
3233 if size != disk.size:
3234 self.LogInfo("Disk %d of instance %s has mismatched size,"
3235 " correcting: recorded %d, actual %d", idx,
3236 instance.name, disk.size, size)
3238 self.cfg.Update(instance, feedback_fn)
3239 changed.append((instance.name, idx, size))
3240 if self._EnsureChildSizes(disk):
3241 self.cfg.Update(instance, feedback_fn)
3242 changed.append((instance.name, idx, disk.size))
3246 class LUClusterRename(LogicalUnit):
3247 """Rename the cluster.
3250 HPATH = "cluster-rename"
3251 HTYPE = constants.HTYPE_CLUSTER
3253 def BuildHooksEnv(self):
3258 "OP_TARGET": self.cfg.GetClusterName(),
3259 "NEW_NAME": self.op.name,
3262 def BuildHooksNodes(self):
3263 """Build hooks nodes.
3266 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
3268 def CheckPrereq(self):
3269 """Verify that the passed name is a valid one.
3272 hostname = netutils.GetHostname(name=self.op.name,
3273 family=self.cfg.GetPrimaryIPFamily())
3275 new_name = hostname.name
3276 self.ip = new_ip = hostname.ip
3277 old_name = self.cfg.GetClusterName()
3278 old_ip = self.cfg.GetMasterIP()
3279 if new_name == old_name and new_ip == old_ip:
3280 raise errors.OpPrereqError("Neither the name nor the IP address of the"
3281 " cluster has changed",
3283 if new_ip != old_ip:
3284 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
3285 raise errors.OpPrereqError("The given cluster IP address (%s) is"
3286 " reachable on the network" %
3287 new_ip, errors.ECODE_NOTUNIQUE)
3289 self.op.name = new_name
3291 def Exec(self, feedback_fn):
3292 """Rename the cluster.
3295 clustername = self.op.name
3298 # shutdown the master IP
3299 master = self.cfg.GetMasterNode()
3300 result = self.rpc.call_node_stop_master(master, False)
3301 result.Raise("Could not disable the master role")
3304 cluster = self.cfg.GetClusterInfo()
3305 cluster.cluster_name = clustername
3306 cluster.master_ip = ip
3307 self.cfg.Update(cluster, feedback_fn)
3309 # update the known hosts file
3310 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
3311 node_list = self.cfg.GetOnlineNodeList()
3313 node_list.remove(master)
3316 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
3318 result = self.rpc.call_node_start_master(master, False, False)
3319 msg = result.fail_msg
3321 self.LogWarning("Could not re-enable the master role on"
3322 " the master, please restart manually: %s", msg)
3327 class LUClusterSetParams(LogicalUnit):
3328 """Change the parameters of the cluster.
3331 HPATH = "cluster-modify"
3332 HTYPE = constants.HTYPE_CLUSTER
3335 def CheckArguments(self):
3339 if self.op.uid_pool:
3340 uidpool.CheckUidPool(self.op.uid_pool)
3342 if self.op.add_uids:
3343 uidpool.CheckUidPool(self.op.add_uids)
3345 if self.op.remove_uids:
3346 uidpool.CheckUidPool(self.op.remove_uids)
3348 def ExpandNames(self):
3349 # FIXME: in the future maybe other cluster params won't require checking on
3350 # all nodes to be modified.
3351 self.needed_locks = {
3352 locking.LEVEL_NODE: locking.ALL_SET,
3354 self.share_locks[locking.LEVEL_NODE] = 1
3356 def BuildHooksEnv(self):
3361 "OP_TARGET": self.cfg.GetClusterName(),
3362 "NEW_VG_NAME": self.op.vg_name,
3365 def BuildHooksNodes(self):
3366 """Build hooks nodes.
3369 mn = self.cfg.GetMasterNode()
3372 def CheckPrereq(self):
3373 """Check prerequisites.
3375 This checks whether the given params don't conflict and
3376 if the given volume group is valid.
3379 if self.op.vg_name is not None and not self.op.vg_name:
3380 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
3381 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
3382 " instances exist", errors.ECODE_INVAL)
3384 if self.op.drbd_helper is not None and not self.op.drbd_helper:
3385 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
3386 raise errors.OpPrereqError("Cannot disable drbd helper while"
3387 " drbd-based instances exist",
3390 node_list = self.owned_locks(locking.LEVEL_NODE)
3392 # if vg_name not None, checks given volume group on all nodes
3394 vglist = self.rpc.call_vg_list(node_list)
3395 for node in node_list:
3396 msg = vglist[node].fail_msg
3398 # ignoring down node
3399 self.LogWarning("Error while gathering data on node %s"
3400 " (ignoring node): %s", node, msg)
3402 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
3404 constants.MIN_VG_SIZE)
3406 raise errors.OpPrereqError("Error on node '%s': %s" %
3407 (node, vgstatus), errors.ECODE_ENVIRON)
3409 if self.op.drbd_helper:
3410 # checks given drbd helper on all nodes
3411 helpers = self.rpc.call_drbd_helper(node_list)
3412 for (node, ninfo) in self.cfg.GetMultiNodeInfo(node_list):
3414 self.LogInfo("Not checking drbd helper on offline node %s", node)
3416 msg = helpers[node].fail_msg
3418 raise errors.OpPrereqError("Error checking drbd helper on node"
3419 " '%s': %s" % (node, msg),
3420 errors.ECODE_ENVIRON)
3421 node_helper = helpers[node].payload
3422 if node_helper != self.op.drbd_helper:
3423 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
3424 (node, node_helper), errors.ECODE_ENVIRON)
3426 self.cluster = cluster = self.cfg.GetClusterInfo()
3427 # validate params changes
3428 if self.op.beparams:
3429 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
3430 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
3432 if self.op.ndparams:
3433 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
3434 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
3436 # TODO: we need a more general way to handle resetting
3437 # cluster-level parameters to default values
3438 if self.new_ndparams["oob_program"] == "":
3439 self.new_ndparams["oob_program"] = \
3440 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
3442 if self.op.nicparams:
3443 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
3444 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
3445 objects.NIC.CheckParameterSyntax(self.new_nicparams)
3448 # check all instances for consistency
3449 for instance in self.cfg.GetAllInstancesInfo().values():
3450 for nic_idx, nic in enumerate(instance.nics):
3451 params_copy = copy.deepcopy(nic.nicparams)
3452 params_filled = objects.FillDict(self.new_nicparams, params_copy)
3454 # check parameter syntax
3456 objects.NIC.CheckParameterSyntax(params_filled)
3457 except errors.ConfigurationError, err:
3458 nic_errors.append("Instance %s, nic/%d: %s" %
3459 (instance.name, nic_idx, err))
3461 # if we're moving instances to routed, check that they have an ip
3462 target_mode = params_filled[constants.NIC_MODE]
3463 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
3464 nic_errors.append("Instance %s, nic/%d: routed NIC with no ip"
3465 " address" % (instance.name, nic_idx))
3467 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
3468 "\n".join(nic_errors))
3470 # hypervisor list/parameters
3471 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
3472 if self.op.hvparams:
3473 for hv_name, hv_dict in self.op.hvparams.items():
3474 if hv_name not in self.new_hvparams:
3475 self.new_hvparams[hv_name] = hv_dict
3477 self.new_hvparams[hv_name].update(hv_dict)
3479 # os hypervisor parameters
3480 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
3482 for os_name, hvs in self.op.os_hvp.items():
3483 if os_name not in self.new_os_hvp:
3484 self.new_os_hvp[os_name] = hvs
3486 for hv_name, hv_dict in hvs.items():
3487 if hv_name not in self.new_os_hvp[os_name]:
3488 self.new_os_hvp[os_name][hv_name] = hv_dict
3490 self.new_os_hvp[os_name][hv_name].update(hv_dict)
3493 self.new_osp = objects.FillDict(cluster.osparams, {})
3494 if self.op.osparams:
3495 for os_name, osp in self.op.osparams.items():
3496 if os_name not in self.new_osp:
3497 self.new_osp[os_name] = {}
3499 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
3502 if not self.new_osp[os_name]:
3503 # we removed all parameters
3504 del self.new_osp[os_name]
3506 # check the parameter validity (remote check)
3507 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
3508 os_name, self.new_osp[os_name])
3510 # changes to the hypervisor list
3511 if self.op.enabled_hypervisors is not None:
3512 self.hv_list = self.op.enabled_hypervisors
3513 for hv in self.hv_list:
3514 # if the hypervisor doesn't already exist in the cluster
3515 # hvparams, we initialize it to empty, and then (in both
3516 # cases) we make sure to fill the defaults, as we might not
3517 # have a complete defaults list if the hypervisor wasn't
3519 if hv not in new_hvp:
3521 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
3522 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
3524 self.hv_list = cluster.enabled_hypervisors
3526 if self.op.hvparams or self.op.enabled_hypervisors is not None:
3527 # either the enabled list has changed, or the parameters have, validate
3528 for hv_name, hv_params in self.new_hvparams.items():
3529 if ((self.op.hvparams and hv_name in self.op.hvparams) or
3530 (self.op.enabled_hypervisors and
3531 hv_name in self.op.enabled_hypervisors)):
3532 # either this is a new hypervisor, or its parameters have changed
3533 hv_class = hypervisor.GetHypervisor(hv_name)
3534 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3535 hv_class.CheckParameterSyntax(hv_params)
3536 _CheckHVParams(self, node_list, hv_name, hv_params)
3539 # no need to check any newly-enabled hypervisors, since the
3540 # defaults have already been checked in the above code-block
3541 for os_name, os_hvp in self.new_os_hvp.items():
3542 for hv_name, hv_params in os_hvp.items():
3543 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3544 # we need to fill in the new os_hvp on top of the actual hv_p
3545 cluster_defaults = self.new_hvparams.get(hv_name, {})
3546 new_osp = objects.FillDict(cluster_defaults, hv_params)
3547 hv_class = hypervisor.GetHypervisor(hv_name)
3548 hv_class.CheckParameterSyntax(new_osp)
3549 _CheckHVParams(self, node_list, hv_name, new_osp)
3551 if self.op.default_iallocator:
3552 alloc_script = utils.FindFile(self.op.default_iallocator,
3553 constants.IALLOCATOR_SEARCH_PATH,
3555 if alloc_script is None:
3556 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
3557 " specified" % self.op.default_iallocator,
3560 def Exec(self, feedback_fn):
3561 """Change the parameters of the cluster.
3564 if self.op.vg_name is not None:
3565 new_volume = self.op.vg_name
3568 if new_volume != self.cfg.GetVGName():
3569 self.cfg.SetVGName(new_volume)
3571 feedback_fn("Cluster LVM configuration already in desired"
3572 " state, not changing")
3573 if self.op.drbd_helper is not None:
3574 new_helper = self.op.drbd_helper
3577 if new_helper != self.cfg.GetDRBDHelper():
3578 self.cfg.SetDRBDHelper(new_helper)
3580 feedback_fn("Cluster DRBD helper already in desired state,"
3582 if self.op.hvparams:
3583 self.cluster.hvparams = self.new_hvparams
3585 self.cluster.os_hvp = self.new_os_hvp
3586 if self.op.enabled_hypervisors is not None:
3587 self.cluster.hvparams = self.new_hvparams
3588 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
3589 if self.op.beparams:
3590 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
3591 if self.op.nicparams:
3592 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
3593 if self.op.osparams:
3594 self.cluster.osparams = self.new_osp
3595 if self.op.ndparams:
3596 self.cluster.ndparams = self.new_ndparams
3598 if self.op.candidate_pool_size is not None:
3599 self.cluster.candidate_pool_size = self.op.candidate_pool_size
3600 # we need to update the pool size here, otherwise the save will fail
3601 _AdjustCandidatePool(self, [])
3603 if self.op.maintain_node_health is not None:
3604 self.cluster.maintain_node_health = self.op.maintain_node_health
3606 if self.op.prealloc_wipe_disks is not None:
3607 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
3609 if self.op.add_uids is not None:
3610 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
3612 if self.op.remove_uids is not None:
3613 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
3615 if self.op.uid_pool is not None:
3616 self.cluster.uid_pool = self.op.uid_pool
3618 if self.op.default_iallocator is not None:
3619 self.cluster.default_iallocator = self.op.default_iallocator
3621 if self.op.reserved_lvs is not None:
3622 self.cluster.reserved_lvs = self.op.reserved_lvs
3624 def helper_os(aname, mods, desc):
3626 lst = getattr(self.cluster, aname)
3627 for key, val in mods:
3628 if key == constants.DDM_ADD:
3630 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3633 elif key == constants.DDM_REMOVE:
3637 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3639 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3641 if self.op.hidden_os:
3642 helper_os("hidden_os", self.op.hidden_os, "hidden")
3644 if self.op.blacklisted_os:
3645 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3647 if self.op.master_netdev:
3648 master = self.cfg.GetMasterNode()
3649 feedback_fn("Shutting down master ip on the current netdev (%s)" %
3650 self.cluster.master_netdev)
3651 result = self.rpc.call_node_stop_master(master, False)
3652 result.Raise("Could not disable the master ip")
3653 feedback_fn("Changing master_netdev from %s to %s" %
3654 (self.cluster.master_netdev, self.op.master_netdev))
3655 self.cluster.master_netdev = self.op.master_netdev
3657 self.cfg.Update(self.cluster, feedback_fn)
3659 if self.op.master_netdev:
3660 feedback_fn("Starting the master ip on the new master netdev (%s)" %
3661 self.op.master_netdev)
3662 result = self.rpc.call_node_start_master(master, False, False)
3664 self.LogWarning("Could not re-enable the master ip on"
3665 " the master, please restart manually: %s",
3669 def _UploadHelper(lu, nodes, fname):
3670 """Helper for uploading a file and showing warnings.
3673 if os.path.exists(fname):
3674 result = lu.rpc.call_upload_file(nodes, fname)
3675 for to_node, to_result in result.items():
3676 msg = to_result.fail_msg
3678 msg = ("Copy of file %s to node %s failed: %s" %
3679 (fname, to_node, msg))
3680 lu.proc.LogWarning(msg)
3683 def _ComputeAncillaryFiles(cluster, redist):
3684 """Compute files external to Ganeti which need to be consistent.
3686 @type redist: boolean
3687 @param redist: Whether to include files which need to be redistributed
3690 # Compute files for all nodes
3692 constants.SSH_KNOWN_HOSTS_FILE,
3693 constants.CONFD_HMAC_KEY,
3694 constants.CLUSTER_DOMAIN_SECRET_FILE,
3698 files_all.update(constants.ALL_CERT_FILES)
3699 files_all.update(ssconf.SimpleStore().GetFileList())
3701 if cluster.modify_etc_hosts:
3702 files_all.add(constants.ETC_HOSTS)
3704 # Files which must either exist on all nodes or on none
3705 files_all_opt = set([
3706 constants.RAPI_USERS_FILE,
3709 # Files which should only be on master candidates
3712 files_mc.add(constants.CLUSTER_CONF_FILE)
3714 # Files which should only be on VM-capable nodes
3715 files_vm = set(filename
3716 for hv_name in cluster.enabled_hypervisors
3717 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles())
3719 # Filenames must be unique
3720 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
3721 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
3722 "Found file listed in more than one file list"
3724 return (files_all, files_all_opt, files_mc, files_vm)
3727 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3728 """Distribute additional files which are part of the cluster configuration.
3730 ConfigWriter takes care of distributing the config and ssconf files, but
3731 there are more files which should be distributed to all nodes. This function
3732 makes sure those are copied.
3734 @param lu: calling logical unit
3735 @param additional_nodes: list of nodes not in the config to distribute to
3736 @type additional_vm: boolean
3737 @param additional_vm: whether the additional nodes are vm-capable or not
3740 # Gather target nodes
3741 cluster = lu.cfg.GetClusterInfo()
3742 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3744 online_nodes = lu.cfg.GetOnlineNodeList()
3745 vm_nodes = lu.cfg.GetVmCapableNodeList()
3747 if additional_nodes is not None:
3748 online_nodes.extend(additional_nodes)
3750 vm_nodes.extend(additional_nodes)
3752 # Never distribute to master node
3753 for nodelist in [online_nodes, vm_nodes]:
3754 if master_info.name in nodelist:
3755 nodelist.remove(master_info.name)
3758 (files_all, files_all_opt, files_mc, files_vm) = \
3759 _ComputeAncillaryFiles(cluster, True)
3761 # Never re-distribute configuration file from here
3762 assert not (constants.CLUSTER_CONF_FILE in files_all or
3763 constants.CLUSTER_CONF_FILE in files_vm)
3764 assert not files_mc, "Master candidates not handled in this function"
3767 (online_nodes, files_all),
3768 (online_nodes, files_all_opt),
3769 (vm_nodes, files_vm),
3773 for (node_list, files) in filemap:
3775 _UploadHelper(lu, node_list, fname)
3778 class LUClusterRedistConf(NoHooksLU):
3779 """Force the redistribution of cluster configuration.
3781 This is a very simple LU.
3786 def ExpandNames(self):
3787 self.needed_locks = {
3788 locking.LEVEL_NODE: locking.ALL_SET,
3790 self.share_locks[locking.LEVEL_NODE] = 1
3792 def Exec(self, feedback_fn):
3793 """Redistribute the configuration.
3796 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3797 _RedistributeAncillaryFiles(self)
3800 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3801 """Sleep and poll for an instance's disk to sync.
3804 if not instance.disks or disks is not None and not disks:
3807 disks = _ExpandCheckDisks(instance, disks)
3810 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3812 node = instance.primary_node
3815 lu.cfg.SetDiskID(dev, node)
3817 # TODO: Convert to utils.Retry
3820 degr_retries = 10 # in seconds, as we sleep 1 second each time
3824 cumul_degraded = False
3825 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3826 msg = rstats.fail_msg
3828 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3831 raise errors.RemoteError("Can't contact node %s for mirror data,"
3832 " aborting." % node)
3835 rstats = rstats.payload
3837 for i, mstat in enumerate(rstats):
3839 lu.LogWarning("Can't compute data for node %s/%s",
3840 node, disks[i].iv_name)
3843 cumul_degraded = (cumul_degraded or
3844 (mstat.is_degraded and mstat.sync_percent is None))
3845 if mstat.sync_percent is not None:
3847 if mstat.estimated_time is not None:
3848 rem_time = ("%s remaining (estimated)" %
3849 utils.FormatSeconds(mstat.estimated_time))
3850 max_time = mstat.estimated_time
3852 rem_time = "no time estimate"
3853 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3854 (disks[i].iv_name, mstat.sync_percent, rem_time))
3856 # if we're done but degraded, let's do a few small retries, to
3857 # make sure we see a stable and not transient situation; therefore
3858 # we force restart of the loop
3859 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3860 logging.info("Degraded disks found, %d retries left", degr_retries)
3868 time.sleep(min(60, max_time))
3871 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3872 return not cumul_degraded
3875 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3876 """Check that mirrors are not degraded.
3878 The ldisk parameter, if True, will change the test from the
3879 is_degraded attribute (which represents overall non-ok status for
3880 the device(s)) to the ldisk (representing the local storage status).
3883 lu.cfg.SetDiskID(dev, node)
3887 if on_primary or dev.AssembleOnSecondary():
3888 rstats = lu.rpc.call_blockdev_find(node, dev)
3889 msg = rstats.fail_msg
3891 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3893 elif not rstats.payload:
3894 lu.LogWarning("Can't find disk on node %s", node)
3898 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3900 result = result and not rstats.payload.is_degraded
3903 for child in dev.children:
3904 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3909 class LUOobCommand(NoHooksLU):
3910 """Logical unit for OOB handling.
3914 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
3916 def ExpandNames(self):
3917 """Gather locks we need.
3920 if self.op.node_names:
3921 self.op.node_names = _GetWantedNodes(self, self.op.node_names)
3922 lock_names = self.op.node_names
3924 lock_names = locking.ALL_SET
3926 self.needed_locks = {
3927 locking.LEVEL_NODE: lock_names,
3930 def CheckPrereq(self):
3931 """Check prerequisites.
3934 - the node exists in the configuration
3937 Any errors are signaled by raising errors.OpPrereqError.
3941 self.master_node = self.cfg.GetMasterNode()
3943 assert self.op.power_delay >= 0.0
3945 if self.op.node_names:
3946 if (self.op.command in self._SKIP_MASTER and
3947 self.master_node in self.op.node_names):
3948 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
3949 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
3951 if master_oob_handler:
3952 additional_text = ("run '%s %s %s' if you want to operate on the"
3953 " master regardless") % (master_oob_handler,
3957 additional_text = "it does not support out-of-band operations"
3959 raise errors.OpPrereqError(("Operating on the master node %s is not"
3960 " allowed for %s; %s") %
3961 (self.master_node, self.op.command,
3962 additional_text), errors.ECODE_INVAL)
3964 self.op.node_names = self.cfg.GetNodeList()
3965 if self.op.command in self._SKIP_MASTER:
3966 self.op.node_names.remove(self.master_node)
3968 if self.op.command in self._SKIP_MASTER:
3969 assert self.master_node not in self.op.node_names
3971 for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names):
3973 raise errors.OpPrereqError("Node %s not found" % node_name,
3976 self.nodes.append(node)
3978 if (not self.op.ignore_status and
3979 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
3980 raise errors.OpPrereqError(("Cannot power off node %s because it is"
3981 " not marked offline") % node_name,
3984 def Exec(self, feedback_fn):
3985 """Execute OOB and return result if we expect any.
3988 master_node = self.master_node
3991 for idx, node in enumerate(utils.NiceSort(self.nodes,
3992 key=lambda node: node.name)):
3993 node_entry = [(constants.RS_NORMAL, node.name)]
3994 ret.append(node_entry)
3996 oob_program = _SupportsOob(self.cfg, node)
3999 node_entry.append((constants.RS_UNAVAIL, None))
4002 logging.info("Executing out-of-band command '%s' using '%s' on %s",
4003 self.op.command, oob_program, node.name)
4004 result = self.rpc.call_run_oob(master_node, oob_program,
4005 self.op.command, node.name,
4009 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
4010 node.name, result.fail_msg)
4011 node_entry.append((constants.RS_NODATA, None))
4014 self._CheckPayload(result)
4015 except errors.OpExecError, err:
4016 self.LogWarning("Payload returned by node '%s' is not valid: %s",
4018 node_entry.append((constants.RS_NODATA, None))
4020 if self.op.command == constants.OOB_HEALTH:
4021 # For health we should log important events
4022 for item, status in result.payload:
4023 if status in [constants.OOB_STATUS_WARNING,
4024 constants.OOB_STATUS_CRITICAL]:
4025 self.LogWarning("Item '%s' on node '%s' has status '%s'",
4026 item, node.name, status)
4028 if self.op.command == constants.OOB_POWER_ON:
4030 elif self.op.command == constants.OOB_POWER_OFF:
4031 node.powered = False
4032 elif self.op.command == constants.OOB_POWER_STATUS:
4033 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
4034 if powered != node.powered:
4035 logging.warning(("Recorded power state (%s) of node '%s' does not"
4036 " match actual power state (%s)"), node.powered,
4039 # For configuration changing commands we should update the node
4040 if self.op.command in (constants.OOB_POWER_ON,
4041 constants.OOB_POWER_OFF):
4042 self.cfg.Update(node, feedback_fn)
4044 node_entry.append((constants.RS_NORMAL, result.payload))
4046 if (self.op.command == constants.OOB_POWER_ON and
4047 idx < len(self.nodes) - 1):
4048 time.sleep(self.op.power_delay)
4052 def _CheckPayload(self, result):
4053 """Checks if the payload is valid.
4055 @param result: RPC result
4056 @raises errors.OpExecError: If payload is not valid
4060 if self.op.command == constants.OOB_HEALTH:
4061 if not isinstance(result.payload, list):
4062 errs.append("command 'health' is expected to return a list but got %s" %
4063 type(result.payload))
4065 for item, status in result.payload:
4066 if status not in constants.OOB_STATUSES:
4067 errs.append("health item '%s' has invalid status '%s'" %
4070 if self.op.command == constants.OOB_POWER_STATUS:
4071 if not isinstance(result.payload, dict):
4072 errs.append("power-status is expected to return a dict but got %s" %
4073 type(result.payload))
4075 if self.op.command in [
4076 constants.OOB_POWER_ON,
4077 constants.OOB_POWER_OFF,
4078 constants.OOB_POWER_CYCLE,
4080 if result.payload is not None:
4081 errs.append("%s is expected to not return payload but got '%s'" %
4082 (self.op.command, result.payload))
4085 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
4086 utils.CommaJoin(errs))
4089 class _OsQuery(_QueryBase):
4090 FIELDS = query.OS_FIELDS
4092 def ExpandNames(self, lu):
4093 # Lock all nodes in shared mode
4094 # Temporary removal of locks, should be reverted later
4095 # TODO: reintroduce locks when they are lighter-weight
4096 lu.needed_locks = {}
4097 #self.share_locks[locking.LEVEL_NODE] = 1
4098 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4100 # The following variables interact with _QueryBase._GetNames
4102 self.wanted = self.names
4104 self.wanted = locking.ALL_SET
4106 self.do_locking = self.use_locking
4108 def DeclareLocks(self, lu, level):
4112 def _DiagnoseByOS(rlist):
4113 """Remaps a per-node return list into an a per-os per-node dictionary
4115 @param rlist: a map with node names as keys and OS objects as values
4118 @return: a dictionary with osnames as keys and as value another
4119 map, with nodes as keys and tuples of (path, status, diagnose,
4120 variants, parameters, api_versions) as values, eg::
4122 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
4123 (/srv/..., False, "invalid api")],
4124 "node2": [(/srv/..., True, "", [], [])]}
4129 # we build here the list of nodes that didn't fail the RPC (at RPC
4130 # level), so that nodes with a non-responding node daemon don't
4131 # make all OSes invalid
4132 good_nodes = [node_name for node_name in rlist
4133 if not rlist[node_name].fail_msg]
4134 for node_name, nr in rlist.items():
4135 if nr.fail_msg or not nr.payload:
4137 for (name, path, status, diagnose, variants,
4138 params, api_versions) in nr.payload:
4139 if name not in all_os:
4140 # build a list of nodes for this os containing empty lists
4141 # for each node in node_list
4143 for nname in good_nodes:
4144 all_os[name][nname] = []
4145 # convert params from [name, help] to (name, help)
4146 params = [tuple(v) for v in params]
4147 all_os[name][node_name].append((path, status, diagnose,
4148 variants, params, api_versions))
4151 def _GetQueryData(self, lu):
4152 """Computes the list of nodes and their attributes.
4155 # Locking is not used
4156 assert not (compat.any(lu.glm.is_owned(level)
4157 for level in locking.LEVELS
4158 if level != locking.LEVEL_CLUSTER) or
4159 self.do_locking or self.use_locking)
4161 valid_nodes = [node.name
4162 for node in lu.cfg.GetAllNodesInfo().values()
4163 if not node.offline and node.vm_capable]
4164 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
4165 cluster = lu.cfg.GetClusterInfo()
4169 for (os_name, os_data) in pol.items():
4170 info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
4171 hidden=(os_name in cluster.hidden_os),
4172 blacklisted=(os_name in cluster.blacklisted_os))
4176 api_versions = set()
4178 for idx, osl in enumerate(os_data.values()):
4179 info.valid = bool(info.valid and osl and osl[0][1])
4183 (node_variants, node_params, node_api) = osl[0][3:6]
4186 variants.update(node_variants)
4187 parameters.update(node_params)
4188 api_versions.update(node_api)
4190 # Filter out inconsistent values
4191 variants.intersection_update(node_variants)
4192 parameters.intersection_update(node_params)
4193 api_versions.intersection_update(node_api)
4195 info.variants = list(variants)
4196 info.parameters = list(parameters)
4197 info.api_versions = list(api_versions)
4199 data[os_name] = info
4201 # Prepare data in requested order
4202 return [data[name] for name in self._GetNames(lu, pol.keys(), None)
4206 class LUOsDiagnose(NoHooksLU):
4207 """Logical unit for OS diagnose/query.
4213 def _BuildFilter(fields, names):
4214 """Builds a filter for querying OSes.
4217 name_filter = qlang.MakeSimpleFilter("name", names)
4219 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
4220 # respective field is not requested
4221 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]]
4222 for fname in ["hidden", "blacklisted"]
4223 if fname not in fields]
4224 if "valid" not in fields:
4225 status_filter.append([qlang.OP_TRUE, "valid"])
4228 status_filter.insert(0, qlang.OP_AND)
4230 status_filter = None
4232 if name_filter and status_filter:
4233 return [qlang.OP_AND, name_filter, status_filter]
4237 return status_filter
4239 def CheckArguments(self):
4240 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names),
4241 self.op.output_fields, False)
4243 def ExpandNames(self):
4244 self.oq.ExpandNames(self)
4246 def Exec(self, feedback_fn):
4247 return self.oq.OldStyleQuery(self)
4250 class LUNodeRemove(LogicalUnit):
4251 """Logical unit for removing a node.
4254 HPATH = "node-remove"
4255 HTYPE = constants.HTYPE_NODE
4257 def BuildHooksEnv(self):
4260 This doesn't run on the target node in the pre phase as a failed
4261 node would then be impossible to remove.
4265 "OP_TARGET": self.op.node_name,
4266 "NODE_NAME": self.op.node_name,
4269 def BuildHooksNodes(self):
4270 """Build hooks nodes.
4273 all_nodes = self.cfg.GetNodeList()
4275 all_nodes.remove(self.op.node_name)
4277 logging.warning("Node '%s', which is about to be removed, was not found"
4278 " in the list of all nodes", self.op.node_name)
4279 return (all_nodes, all_nodes)
4281 def CheckPrereq(self):
4282 """Check prerequisites.
4285 - the node exists in the configuration
4286 - it does not have primary or secondary instances
4287 - it's not the master
4289 Any errors are signaled by raising errors.OpPrereqError.
4292 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4293 node = self.cfg.GetNodeInfo(self.op.node_name)
4294 assert node is not None
4296 masternode = self.cfg.GetMasterNode()
4297 if node.name == masternode:
4298 raise errors.OpPrereqError("Node is the master node, failover to another"
4299 " node is required", errors.ECODE_INVAL)
4301 for instance_name, instance in self.cfg.GetAllInstancesInfo():
4302 if node.name in instance.all_nodes:
4303 raise errors.OpPrereqError("Instance %s is still running on the node,"
4304 " please remove first" % instance_name,
4306 self.op.node_name = node.name
4309 def Exec(self, feedback_fn):
4310 """Removes the node from the cluster.
4314 logging.info("Stopping the node daemon and removing configs from node %s",
4317 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
4319 # Promote nodes to master candidate as needed
4320 _AdjustCandidatePool(self, exceptions=[node.name])
4321 self.context.RemoveNode(node.name)
4323 # Run post hooks on the node before it's removed
4324 _RunPostHook(self, node.name)
4326 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
4327 msg = result.fail_msg
4329 self.LogWarning("Errors encountered on the remote node while leaving"
4330 " the cluster: %s", msg)
4332 # Remove node from our /etc/hosts
4333 if self.cfg.GetClusterInfo().modify_etc_hosts:
4334 master_node = self.cfg.GetMasterNode()
4335 result = self.rpc.call_etc_hosts_modify(master_node,
4336 constants.ETC_HOSTS_REMOVE,
4338 result.Raise("Can't update hosts file with new host data")
4339 _RedistributeAncillaryFiles(self)
4342 class _NodeQuery(_QueryBase):
4343 FIELDS = query.NODE_FIELDS
4345 def ExpandNames(self, lu):
4346 lu.needed_locks = {}
4347 lu.share_locks = _ShareAll()
4350 self.wanted = _GetWantedNodes(lu, self.names)
4352 self.wanted = locking.ALL_SET
4354 self.do_locking = (self.use_locking and
4355 query.NQ_LIVE in self.requested_data)
4358 # If any non-static field is requested we need to lock the nodes
4359 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
4361 def DeclareLocks(self, lu, level):
4364 def _GetQueryData(self, lu):
4365 """Computes the list of nodes and their attributes.
4368 all_info = lu.cfg.GetAllNodesInfo()
4370 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
4372 # Gather data as requested
4373 if query.NQ_LIVE in self.requested_data:
4374 # filter out non-vm_capable nodes
4375 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
4377 node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
4378 lu.cfg.GetHypervisorType())
4379 live_data = dict((name, nresult.payload)
4380 for (name, nresult) in node_data.items()
4381 if not nresult.fail_msg and nresult.payload)
4385 if query.NQ_INST in self.requested_data:
4386 node_to_primary = dict([(name, set()) for name in nodenames])
4387 node_to_secondary = dict([(name, set()) for name in nodenames])
4389 inst_data = lu.cfg.GetAllInstancesInfo()
4391 for inst in inst_data.values():
4392 if inst.primary_node in node_to_primary:
4393 node_to_primary[inst.primary_node].add(inst.name)
4394 for secnode in inst.secondary_nodes:
4395 if secnode in node_to_secondary:
4396 node_to_secondary[secnode].add(inst.name)
4398 node_to_primary = None
4399 node_to_secondary = None
4401 if query.NQ_OOB in self.requested_data:
4402 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
4403 for name, node in all_info.iteritems())
4407 if query.NQ_GROUP in self.requested_data:
4408 groups = lu.cfg.GetAllNodeGroupsInfo()
4412 return query.NodeQueryData([all_info[name] for name in nodenames],
4413 live_data, lu.cfg.GetMasterNode(),
4414 node_to_primary, node_to_secondary, groups,
4415 oob_support, lu.cfg.GetClusterInfo())
4418 class LUNodeQuery(NoHooksLU):
4419 """Logical unit for querying nodes.
4422 # pylint: disable=W0142
4425 def CheckArguments(self):
4426 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
4427 self.op.output_fields, self.op.use_locking)
4429 def ExpandNames(self):
4430 self.nq.ExpandNames(self)
4432 def Exec(self, feedback_fn):
4433 return self.nq.OldStyleQuery(self)
4436 class LUNodeQueryvols(NoHooksLU):
4437 """Logical unit for getting volumes on node(s).
4441 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
4442 _FIELDS_STATIC = utils.FieldSet("node")
4444 def CheckArguments(self):
4445 _CheckOutputFields(static=self._FIELDS_STATIC,
4446 dynamic=self._FIELDS_DYNAMIC,
4447 selected=self.op.output_fields)
4449 def ExpandNames(self):
4450 self.needed_locks = {}
4451 self.share_locks[locking.LEVEL_NODE] = 1
4452 if not self.op.nodes:
4453 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4455 self.needed_locks[locking.LEVEL_NODE] = \
4456 _GetWantedNodes(self, self.op.nodes)
4458 def Exec(self, feedback_fn):
4459 """Computes the list of nodes and their attributes.
4462 nodenames = self.owned_locks(locking.LEVEL_NODE)
4463 volumes = self.rpc.call_node_volumes(nodenames)
4465 ilist = self.cfg.GetAllInstancesInfo()
4466 vol2inst = _MapInstanceDisksToNodes(ilist.values())
4469 for node in nodenames:
4470 nresult = volumes[node]
4473 msg = nresult.fail_msg
4475 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
4478 node_vols = sorted(nresult.payload,
4479 key=operator.itemgetter("dev"))
4481 for vol in node_vols:
4483 for field in self.op.output_fields:
4486 elif field == "phys":
4490 elif field == "name":
4492 elif field == "size":
4493 val = int(float(vol["size"]))
4494 elif field == "instance":
4495 val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-")
4497 raise errors.ParameterError(field)
4498 node_output.append(str(val))
4500 output.append(node_output)
4505 class LUNodeQueryStorage(NoHooksLU):
4506 """Logical unit for getting information on storage units on node(s).
4509 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
4512 def CheckArguments(self):
4513 _CheckOutputFields(static=self._FIELDS_STATIC,
4514 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
4515 selected=self.op.output_fields)
4517 def ExpandNames(self):
4518 self.needed_locks = {}
4519 self.share_locks[locking.LEVEL_NODE] = 1
4522 self.needed_locks[locking.LEVEL_NODE] = \
4523 _GetWantedNodes(self, self.op.nodes)
4525 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4527 def Exec(self, feedback_fn):
4528 """Computes the list of nodes and their attributes.
4531 self.nodes = self.owned_locks(locking.LEVEL_NODE)
4533 # Always get name to sort by
4534 if constants.SF_NAME in self.op.output_fields:
4535 fields = self.op.output_fields[:]
4537 fields = [constants.SF_NAME] + self.op.output_fields
4539 # Never ask for node or type as it's only known to the LU
4540 for extra in [constants.SF_NODE, constants.SF_TYPE]:
4541 while extra in fields:
4542 fields.remove(extra)
4544 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
4545 name_idx = field_idx[constants.SF_NAME]
4547 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4548 data = self.rpc.call_storage_list(self.nodes,
4549 self.op.storage_type, st_args,
4550 self.op.name, fields)
4554 for node in utils.NiceSort(self.nodes):
4555 nresult = data[node]
4559 msg = nresult.fail_msg
4561 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
4564 rows = dict([(row[name_idx], row) for row in nresult.payload])
4566 for name in utils.NiceSort(rows.keys()):
4571 for field in self.op.output_fields:
4572 if field == constants.SF_NODE:
4574 elif field == constants.SF_TYPE:
4575 val = self.op.storage_type
4576 elif field in field_idx:
4577 val = row[field_idx[field]]
4579 raise errors.ParameterError(field)
4588 class _InstanceQuery(_QueryBase):
4589 FIELDS = query.INSTANCE_FIELDS
4591 def ExpandNames(self, lu):
4592 lu.needed_locks = {}
4593 lu.share_locks = _ShareAll()
4596 self.wanted = _GetWantedInstances(lu, self.names)
4598 self.wanted = locking.ALL_SET
4600 self.do_locking = (self.use_locking and
4601 query.IQ_LIVE in self.requested_data)
4603 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4604 lu.needed_locks[locking.LEVEL_NODEGROUP] = []
4605 lu.needed_locks[locking.LEVEL_NODE] = []
4606 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4608 self.do_grouplocks = (self.do_locking and
4609 query.IQ_NODES in self.requested_data)
4611 def DeclareLocks(self, lu, level):
4613 if level == locking.LEVEL_NODEGROUP and self.do_grouplocks:
4614 assert not lu.needed_locks[locking.LEVEL_NODEGROUP]
4616 # Lock all groups used by instances optimistically; this requires going
4617 # via the node before it's locked, requiring verification later on
4618 lu.needed_locks[locking.LEVEL_NODEGROUP] = \
4620 for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
4621 for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
4622 elif level == locking.LEVEL_NODE:
4623 lu._LockInstancesNodes() # pylint: disable=W0212
4626 def _CheckGroupLocks(lu):
4627 owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE))
4628 owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP))
4630 # Check if node groups for locked instances are still correct
4631 for instance_name in owned_instances:
4632 _CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups)
4634 def _GetQueryData(self, lu):
4635 """Computes the list of instances and their attributes.
4638 if self.do_grouplocks:
4639 self._CheckGroupLocks(lu)
4641 cluster = lu.cfg.GetClusterInfo()
4642 all_info = lu.cfg.GetAllInstancesInfo()
4644 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
4646 instance_list = [all_info[name] for name in instance_names]
4647 nodes = frozenset(itertools.chain(*(inst.all_nodes
4648 for inst in instance_list)))
4649 hv_list = list(set([inst.hypervisor for inst in instance_list]))
4652 wrongnode_inst = set()
4654 # Gather data as requested
4655 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
4657 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
4659 result = node_data[name]
4661 # offline nodes will be in both lists
4662 assert result.fail_msg
4663 offline_nodes.append(name)
4665 bad_nodes.append(name)
4666 elif result.payload:
4667 for inst in result.payload:
4668 if inst in all_info:
4669 if all_info[inst].primary_node == name:
4670 live_data.update(result.payload)
4672 wrongnode_inst.add(inst)
4674 # orphan instance; we don't list it here as we don't
4675 # handle this case yet in the output of instance listing
4676 logging.warning("Orphan instance '%s' found on node %s",
4678 # else no instance is alive
4682 if query.IQ_DISKUSAGE in self.requested_data:
4683 disk_usage = dict((inst.name,
4684 _ComputeDiskSize(inst.disk_template,
4685 [{constants.IDISK_SIZE: disk.size}
4686 for disk in inst.disks]))
4687 for inst in instance_list)
4691 if query.IQ_CONSOLE in self.requested_data:
4693 for inst in instance_list:
4694 if inst.name in live_data:
4695 # Instance is running
4696 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
4698 consinfo[inst.name] = None
4699 assert set(consinfo.keys()) == set(instance_names)
4703 if query.IQ_NODES in self.requested_data:
4704 node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"),
4706 nodes = dict(lu.cfg.GetMultiNodeInfo(node_names))
4707 groups = dict((uuid, lu.cfg.GetNodeGroup(uuid))
4708 for uuid in set(map(operator.attrgetter("group"),
4714 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
4715 disk_usage, offline_nodes, bad_nodes,
4716 live_data, wrongnode_inst, consinfo,
4720 class LUQuery(NoHooksLU):
4721 """Query for resources/items of a certain kind.
4724 # pylint: disable=W0142
4727 def CheckArguments(self):
4728 qcls = _GetQueryImplementation(self.op.what)
4730 self.impl = qcls(self.op.filter, self.op.fields, self.op.use_locking)
4732 def ExpandNames(self):
4733 self.impl.ExpandNames(self)
4735 def DeclareLocks(self, level):
4736 self.impl.DeclareLocks(self, level)
4738 def Exec(self, feedback_fn):
4739 return self.impl.NewStyleQuery(self)
4742 class LUQueryFields(NoHooksLU):
4743 """Query for resources/items of a certain kind.
4746 # pylint: disable=W0142
4749 def CheckArguments(self):
4750 self.qcls = _GetQueryImplementation(self.op.what)
4752 def ExpandNames(self):
4753 self.needed_locks = {}
4755 def Exec(self, feedback_fn):
4756 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
4759 class LUNodeModifyStorage(NoHooksLU):
4760 """Logical unit for modifying a storage volume on a node.
4765 def CheckArguments(self):
4766 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4768 storage_type = self.op.storage_type
4771 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4773 raise errors.OpPrereqError("Storage units of type '%s' can not be"
4774 " modified" % storage_type,
4777 diff = set(self.op.changes.keys()) - modifiable
4779 raise errors.OpPrereqError("The following fields can not be modified for"
4780 " storage units of type '%s': %r" %
4781 (storage_type, list(diff)),
4784 def ExpandNames(self):
4785 self.needed_locks = {
4786 locking.LEVEL_NODE: self.op.node_name,
4789 def Exec(self, feedback_fn):
4790 """Computes the list of nodes and their attributes.
4793 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4794 result = self.rpc.call_storage_modify(self.op.node_name,
4795 self.op.storage_type, st_args,
4796 self.op.name, self.op.changes)
4797 result.Raise("Failed to modify storage unit '%s' on %s" %
4798 (self.op.name, self.op.node_name))
4801 class LUNodeAdd(LogicalUnit):
4802 """Logical unit for adding node to the cluster.
4806 HTYPE = constants.HTYPE_NODE
4807 _NFLAGS = ["master_capable", "vm_capable"]
4809 def CheckArguments(self):
4810 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4811 # validate/normalize the node name
4812 self.hostname = netutils.GetHostname(name=self.op.node_name,
4813 family=self.primary_ip_family)
4814 self.op.node_name = self.hostname.name
4816 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
4817 raise errors.OpPrereqError("Cannot readd the master node",
4820 if self.op.readd and self.op.group:
4821 raise errors.OpPrereqError("Cannot pass a node group when a node is"
4822 " being readded", errors.ECODE_INVAL)
4824 def BuildHooksEnv(self):
4827 This will run on all nodes before, and on all nodes + the new node after.
4831 "OP_TARGET": self.op.node_name,
4832 "NODE_NAME": self.op.node_name,
4833 "NODE_PIP": self.op.primary_ip,
4834 "NODE_SIP": self.op.secondary_ip,
4835 "MASTER_CAPABLE": str(self.op.master_capable),
4836 "VM_CAPABLE": str(self.op.vm_capable),
4839 def BuildHooksNodes(self):
4840 """Build hooks nodes.
4843 # Exclude added node
4844 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name]))
4845 post_nodes = pre_nodes + [self.op.node_name, ]
4847 return (pre_nodes, post_nodes)
4849 def CheckPrereq(self):
4850 """Check prerequisites.
4853 - the new node is not already in the config
4855 - its parameters (single/dual homed) matches the cluster
4857 Any errors are signaled by raising errors.OpPrereqError.
4861 hostname = self.hostname
4862 node = hostname.name
4863 primary_ip = self.op.primary_ip = hostname.ip
4864 if self.op.secondary_ip is None:
4865 if self.primary_ip_family == netutils.IP6Address.family:
4866 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4867 " IPv4 address must be given as secondary",
4869 self.op.secondary_ip = primary_ip
4871 secondary_ip = self.op.secondary_ip
4872 if not netutils.IP4Address.IsValid(secondary_ip):
4873 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4874 " address" % secondary_ip, errors.ECODE_INVAL)
4876 node_list = cfg.GetNodeList()
4877 if not self.op.readd and node in node_list:
4878 raise errors.OpPrereqError("Node %s is already in the configuration" %
4879 node, errors.ECODE_EXISTS)
4880 elif self.op.readd and node not in node_list:
4881 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4884 self.changed_primary_ip = False
4886 for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list):
4887 if self.op.readd and node == existing_node_name:
4888 if existing_node.secondary_ip != secondary_ip:
4889 raise errors.OpPrereqError("Readded node doesn't have the same IP"
4890 " address configuration as before",
4892 if existing_node.primary_ip != primary_ip:
4893 self.changed_primary_ip = True
4897 if (existing_node.primary_ip == primary_ip or
4898 existing_node.secondary_ip == primary_ip or
4899 existing_node.primary_ip == secondary_ip or
4900 existing_node.secondary_ip == secondary_ip):
4901 raise errors.OpPrereqError("New node ip address(es) conflict with"
4902 " existing node %s" % existing_node.name,
4903 errors.ECODE_NOTUNIQUE)
4905 # After this 'if' block, None is no longer a valid value for the
4906 # _capable op attributes
4908 old_node = self.cfg.GetNodeInfo(node)
4909 assert old_node is not None, "Can't retrieve locked node %s" % node
4910 for attr in self._NFLAGS:
4911 if getattr(self.op, attr) is None:
4912 setattr(self.op, attr, getattr(old_node, attr))
4914 for attr in self._NFLAGS:
4915 if getattr(self.op, attr) is None:
4916 setattr(self.op, attr, True)
4918 if self.op.readd and not self.op.vm_capable:
4919 pri, sec = cfg.GetNodeInstances(node)
4921 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4922 " flag set to false, but it already holds"
4923 " instances" % node,
4926 # check that the type of the node (single versus dual homed) is the
4927 # same as for the master
4928 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4929 master_singlehomed = myself.secondary_ip == myself.primary_ip
4930 newbie_singlehomed = secondary_ip == primary_ip
4931 if master_singlehomed != newbie_singlehomed:
4932 if master_singlehomed:
4933 raise errors.OpPrereqError("The master has no secondary ip but the"
4934 " new node has one",
4937 raise errors.OpPrereqError("The master has a secondary ip but the"
4938 " new node doesn't have one",
4941 # checks reachability
4942 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
4943 raise errors.OpPrereqError("Node not reachable by ping",
4944 errors.ECODE_ENVIRON)
4946 if not newbie_singlehomed:
4947 # check reachability from my secondary ip to newbie's secondary ip
4948 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
4949 source=myself.secondary_ip):
4950 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4951 " based ping to node daemon port",
4952 errors.ECODE_ENVIRON)
4959 if self.op.master_capable:
4960 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
4962 self.master_candidate = False
4965 self.new_node = old_node
4967 node_group = cfg.LookupNodeGroup(self.op.group)
4968 self.new_node = objects.Node(name=node,
4969 primary_ip=primary_ip,
4970 secondary_ip=secondary_ip,
4971 master_candidate=self.master_candidate,
4972 offline=False, drained=False,
4975 if self.op.ndparams:
4976 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
4978 def Exec(self, feedback_fn):
4979 """Adds the new node to the cluster.
4982 new_node = self.new_node
4983 node = new_node.name
4985 # We adding a new node so we assume it's powered
4986 new_node.powered = True
4988 # for re-adds, reset the offline/drained/master-candidate flags;
4989 # we need to reset here, otherwise offline would prevent RPC calls
4990 # later in the procedure; this also means that if the re-add
4991 # fails, we are left with a non-offlined, broken node
4993 new_node.drained = new_node.offline = False # pylint: disable=W0201
4994 self.LogInfo("Readding a node, the offline/drained flags were reset")
4995 # if we demote the node, we do cleanup later in the procedure
4996 new_node.master_candidate = self.master_candidate
4997 if self.changed_primary_ip:
4998 new_node.primary_ip = self.op.primary_ip
5000 # copy the master/vm_capable flags
5001 for attr in self._NFLAGS:
5002 setattr(new_node, attr, getattr(self.op, attr))
5004 # notify the user about any possible mc promotion
5005 if new_node.master_candidate:
5006 self.LogInfo("Node will be a master candidate")
5008 if self.op.ndparams:
5009 new_node.ndparams = self.op.ndparams
5011 new_node.ndparams = {}
5013 # check connectivity
5014 result = self.rpc.call_version([node])[node]
5015 result.Raise("Can't get version information from node %s" % node)
5016 if constants.PROTOCOL_VERSION == result.payload:
5017 logging.info("Communication to node %s fine, sw version %s match",
5018 node, result.payload)
5020 raise errors.OpExecError("Version mismatch master version %s,"
5021 " node version %s" %
5022 (constants.PROTOCOL_VERSION, result.payload))
5024 # Add node to our /etc/hosts, and add key to known_hosts
5025 if self.cfg.GetClusterInfo().modify_etc_hosts:
5026 master_node = self.cfg.GetMasterNode()
5027 result = self.rpc.call_etc_hosts_modify(master_node,
5028 constants.ETC_HOSTS_ADD,
5031 result.Raise("Can't update hosts file with new host data")
5033 if new_node.secondary_ip != new_node.primary_ip:
5034 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
5037 node_verify_list = [self.cfg.GetMasterNode()]
5038 node_verify_param = {
5039 constants.NV_NODELIST: [node],
5040 # TODO: do a node-net-test as well?
5043 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
5044 self.cfg.GetClusterName())
5045 for verifier in node_verify_list:
5046 result[verifier].Raise("Cannot communicate with node %s" % verifier)
5047 nl_payload = result[verifier].payload[constants.NV_NODELIST]
5049 for failed in nl_payload:
5050 feedback_fn("ssh/hostname verification failed"
5051 " (checking from %s): %s" %
5052 (verifier, nl_payload[failed]))
5053 raise errors.OpExecError("ssh/hostname verification failed")
5056 _RedistributeAncillaryFiles(self)
5057 self.context.ReaddNode(new_node)
5058 # make sure we redistribute the config
5059 self.cfg.Update(new_node, feedback_fn)
5060 # and make sure the new node will not have old files around
5061 if not new_node.master_candidate:
5062 result = self.rpc.call_node_demote_from_mc(new_node.name)
5063 msg = result.fail_msg
5065 self.LogWarning("Node failed to demote itself from master"
5066 " candidate status: %s" % msg)
5068 _RedistributeAncillaryFiles(self, additional_nodes=[node],
5069 additional_vm=self.op.vm_capable)
5070 self.context.AddNode(new_node, self.proc.GetECId())
5073 class LUNodeSetParams(LogicalUnit):
5074 """Modifies the parameters of a node.
5076 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
5077 to the node role (as _ROLE_*)
5078 @cvar _R2F: a dictionary from node role to tuples of flags
5079 @cvar _FLAGS: a list of attribute names corresponding to the flags
5082 HPATH = "node-modify"
5083 HTYPE = constants.HTYPE_NODE
5085 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
5087 (True, False, False): _ROLE_CANDIDATE,
5088 (False, True, False): _ROLE_DRAINED,
5089 (False, False, True): _ROLE_OFFLINE,
5090 (False, False, False): _ROLE_REGULAR,
5092 _R2F = dict((v, k) for k, v in _F2R.items())
5093 _FLAGS = ["master_candidate", "drained", "offline"]
5095 def CheckArguments(self):
5096 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5097 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
5098 self.op.master_capable, self.op.vm_capable,
5099 self.op.secondary_ip, self.op.ndparams]
5100 if all_mods.count(None) == len(all_mods):
5101 raise errors.OpPrereqError("Please pass at least one modification",
5103 if all_mods.count(True) > 1:
5104 raise errors.OpPrereqError("Can't set the node into more than one"
5105 " state at the same time",
5108 # Boolean value that tells us whether we might be demoting from MC
5109 self.might_demote = (self.op.master_candidate == False or
5110 self.op.offline == True or
5111 self.op.drained == True or
5112 self.op.master_capable == False)
5114 if self.op.secondary_ip:
5115 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
5116 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5117 " address" % self.op.secondary_ip,
5120 self.lock_all = self.op.auto_promote and self.might_demote
5121 self.lock_instances = self.op.secondary_ip is not None
5123 def ExpandNames(self):
5125 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
5127 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
5129 if self.lock_instances:
5130 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5132 def DeclareLocks(self, level):
5133 # If we have locked all instances, before waiting to lock nodes, release
5134 # all the ones living on nodes unrelated to the current operation.
5135 if level == locking.LEVEL_NODE and self.lock_instances:
5136 self.affected_instances = []
5137 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
5140 # Build list of instances to release
5141 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
5142 for instance_name, instance in self.cfg.GetMultiInstanceInfo(locked_i):
5143 if (instance.disk_template in constants.DTS_INT_MIRROR and
5144 self.op.node_name in instance.all_nodes):
5145 instances_keep.append(instance_name)
5146 self.affected_instances.append(instance)
5148 _ReleaseLocks(self, locking.LEVEL_INSTANCE, keep=instances_keep)
5150 assert (set(self.owned_locks(locking.LEVEL_INSTANCE)) ==
5151 set(instances_keep))
5153 def BuildHooksEnv(self):
5156 This runs on the master node.
5160 "OP_TARGET": self.op.node_name,
5161 "MASTER_CANDIDATE": str(self.op.master_candidate),
5162 "OFFLINE": str(self.op.offline),
5163 "DRAINED": str(self.op.drained),
5164 "MASTER_CAPABLE": str(self.op.master_capable),
5165 "VM_CAPABLE": str(self.op.vm_capable),
5168 def BuildHooksNodes(self):
5169 """Build hooks nodes.
5172 nl = [self.cfg.GetMasterNode(), self.op.node_name]
5175 def CheckPrereq(self):
5176 """Check prerequisites.
5178 This only checks the instance list against the existing names.
5181 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
5183 if (self.op.master_candidate is not None or
5184 self.op.drained is not None or
5185 self.op.offline is not None):
5186 # we can't change the master's node flags
5187 if self.op.node_name == self.cfg.GetMasterNode():
5188 raise errors.OpPrereqError("The master role can be changed"
5189 " only via master-failover",
5192 if self.op.master_candidate and not node.master_capable:
5193 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
5194 " it a master candidate" % node.name,
5197 if self.op.vm_capable == False:
5198 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
5200 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
5201 " the vm_capable flag" % node.name,
5204 if node.master_candidate and self.might_demote and not self.lock_all:
5205 assert not self.op.auto_promote, "auto_promote set but lock_all not"
5206 # check if after removing the current node, we're missing master
5208 (mc_remaining, mc_should, _) = \
5209 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
5210 if mc_remaining < mc_should:
5211 raise errors.OpPrereqError("Not enough master candidates, please"
5212 " pass auto promote option to allow"
5213 " promotion", errors.ECODE_STATE)
5215 self.old_flags = old_flags = (node.master_candidate,
5216 node.drained, node.offline)
5217 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
5218 self.old_role = old_role = self._F2R[old_flags]
5220 # Check for ineffective changes
5221 for attr in self._FLAGS:
5222 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
5223 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
5224 setattr(self.op, attr, None)
5226 # Past this point, any flag change to False means a transition
5227 # away from the respective state, as only real changes are kept
5229 # TODO: We might query the real power state if it supports OOB
5230 if _SupportsOob(self.cfg, node):
5231 if self.op.offline is False and not (node.powered or
5232 self.op.powered == True):
5233 raise errors.OpPrereqError(("Node %s needs to be turned on before its"
5234 " offline status can be reset") %
5236 elif self.op.powered is not None:
5237 raise errors.OpPrereqError(("Unable to change powered state for node %s"
5238 " as it does not support out-of-band"
5239 " handling") % self.op.node_name)
5241 # If we're being deofflined/drained, we'll MC ourself if needed
5242 if (self.op.drained == False or self.op.offline == False or
5243 (self.op.master_capable and not node.master_capable)):
5244 if _DecideSelfPromotion(self):
5245 self.op.master_candidate = True
5246 self.LogInfo("Auto-promoting node to master candidate")
5248 # If we're no longer master capable, we'll demote ourselves from MC
5249 if self.op.master_capable == False and node.master_candidate:
5250 self.LogInfo("Demoting from master candidate")
5251 self.op.master_candidate = False
5254 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
5255 if self.op.master_candidate:
5256 new_role = self._ROLE_CANDIDATE
5257 elif self.op.drained:
5258 new_role = self._ROLE_DRAINED
5259 elif self.op.offline:
5260 new_role = self._ROLE_OFFLINE
5261 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
5262 # False is still in new flags, which means we're un-setting (the
5264 new_role = self._ROLE_REGULAR
5265 else: # no new flags, nothing, keep old role
5268 self.new_role = new_role
5270 if old_role == self._ROLE_OFFLINE and new_role != old_role:
5271 # Trying to transition out of offline status
5272 result = self.rpc.call_version([node.name])[node.name]
5274 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
5275 " to report its version: %s" %
5276 (node.name, result.fail_msg),
5279 self.LogWarning("Transitioning node from offline to online state"
5280 " without using re-add. Please make sure the node"
5283 if self.op.secondary_ip:
5284 # Ok even without locking, because this can't be changed by any LU
5285 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
5286 master_singlehomed = master.secondary_ip == master.primary_ip
5287 if master_singlehomed and self.op.secondary_ip:
5288 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
5289 " homed cluster", errors.ECODE_INVAL)
5292 if self.affected_instances:
5293 raise errors.OpPrereqError("Cannot change secondary ip: offline"
5294 " node has instances (%s) configured"
5295 " to use it" % self.affected_instances)
5297 # On online nodes, check that no instances are running, and that
5298 # the node has the new ip and we can reach it.
5299 for instance in self.affected_instances:
5300 _CheckInstanceDown(self, instance, "cannot change secondary ip")
5302 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
5303 if master.name != node.name:
5304 # check reachability from master secondary ip to new secondary ip
5305 if not netutils.TcpPing(self.op.secondary_ip,
5306 constants.DEFAULT_NODED_PORT,
5307 source=master.secondary_ip):
5308 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5309 " based ping to node daemon port",
5310 errors.ECODE_ENVIRON)
5312 if self.op.ndparams:
5313 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
5314 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
5315 self.new_ndparams = new_ndparams
5317 def Exec(self, feedback_fn):
5322 old_role = self.old_role
5323 new_role = self.new_role
5327 if self.op.ndparams:
5328 node.ndparams = self.new_ndparams
5330 if self.op.powered is not None:
5331 node.powered = self.op.powered
5333 for attr in ["master_capable", "vm_capable"]:
5334 val = getattr(self.op, attr)
5336 setattr(node, attr, val)
5337 result.append((attr, str(val)))
5339 if new_role != old_role:
5340 # Tell the node to demote itself, if no longer MC and not offline
5341 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
5342 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
5344 self.LogWarning("Node failed to demote itself: %s", msg)
5346 new_flags = self._R2F[new_role]
5347 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
5349 result.append((desc, str(nf)))
5350 (node.master_candidate, node.drained, node.offline) = new_flags
5352 # we locked all nodes, we adjust the CP before updating this node
5354 _AdjustCandidatePool(self, [node.name])
5356 if self.op.secondary_ip:
5357 node.secondary_ip = self.op.secondary_ip
5358 result.append(("secondary_ip", self.op.secondary_ip))
5360 # this will trigger configuration file update, if needed
5361 self.cfg.Update(node, feedback_fn)
5363 # this will trigger job queue propagation or cleanup if the mc
5365 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
5366 self.context.ReaddNode(node)
5371 class LUNodePowercycle(NoHooksLU):
5372 """Powercycles a node.
5377 def CheckArguments(self):
5378 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5379 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
5380 raise errors.OpPrereqError("The node is the master and the force"
5381 " parameter was not set",
5384 def ExpandNames(self):
5385 """Locking for PowercycleNode.
5387 This is a last-resort option and shouldn't block on other
5388 jobs. Therefore, we grab no locks.
5391 self.needed_locks = {}
5393 def Exec(self, feedback_fn):
5397 result = self.rpc.call_node_powercycle(self.op.node_name,
5398 self.cfg.GetHypervisorType())
5399 result.Raise("Failed to schedule the reboot")
5400 return result.payload
5403 class LUClusterQuery(NoHooksLU):
5404 """Query cluster configuration.
5409 def ExpandNames(self):
5410 self.needed_locks = {}
5412 def Exec(self, feedback_fn):
5413 """Return cluster config.
5416 cluster = self.cfg.GetClusterInfo()
5419 # Filter just for enabled hypervisors
5420 for os_name, hv_dict in cluster.os_hvp.items():
5421 os_hvp[os_name] = {}
5422 for hv_name, hv_params in hv_dict.items():
5423 if hv_name in cluster.enabled_hypervisors:
5424 os_hvp[os_name][hv_name] = hv_params
5426 # Convert ip_family to ip_version
5427 primary_ip_version = constants.IP4_VERSION
5428 if cluster.primary_ip_family == netutils.IP6Address.family:
5429 primary_ip_version = constants.IP6_VERSION
5432 "software_version": constants.RELEASE_VERSION,
5433 "protocol_version": constants.PROTOCOL_VERSION,
5434 "config_version": constants.CONFIG_VERSION,
5435 "os_api_version": max(constants.OS_API_VERSIONS),
5436 "export_version": constants.EXPORT_VERSION,
5437 "architecture": (platform.architecture()[0], platform.machine()),
5438 "name": cluster.cluster_name,
5439 "master": cluster.master_node,
5440 "default_hypervisor": cluster.enabled_hypervisors[0],
5441 "enabled_hypervisors": cluster.enabled_hypervisors,
5442 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
5443 for hypervisor_name in cluster.enabled_hypervisors]),
5445 "beparams": cluster.beparams,
5446 "osparams": cluster.osparams,
5447 "nicparams": cluster.nicparams,
5448 "ndparams": cluster.ndparams,
5449 "candidate_pool_size": cluster.candidate_pool_size,
5450 "master_netdev": cluster.master_netdev,
5451 "volume_group_name": cluster.volume_group_name,
5452 "drbd_usermode_helper": cluster.drbd_usermode_helper,
5453 "file_storage_dir": cluster.file_storage_dir,
5454 "shared_file_storage_dir": cluster.shared_file_storage_dir,
5455 "maintain_node_health": cluster.maintain_node_health,
5456 "ctime": cluster.ctime,
5457 "mtime": cluster.mtime,
5458 "uuid": cluster.uuid,
5459 "tags": list(cluster.GetTags()),
5460 "uid_pool": cluster.uid_pool,
5461 "default_iallocator": cluster.default_iallocator,
5462 "reserved_lvs": cluster.reserved_lvs,
5463 "primary_ip_version": primary_ip_version,
5464 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
5465 "hidden_os": cluster.hidden_os,
5466 "blacklisted_os": cluster.blacklisted_os,
5472 class LUClusterConfigQuery(NoHooksLU):
5473 """Return configuration values.
5477 _FIELDS_DYNAMIC = utils.FieldSet()
5478 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
5479 "watcher_pause", "volume_group_name")
5481 def CheckArguments(self):
5482 _CheckOutputFields(static=self._FIELDS_STATIC,
5483 dynamic=self._FIELDS_DYNAMIC,
5484 selected=self.op.output_fields)
5486 def ExpandNames(self):
5487 self.needed_locks = {}
5489 def Exec(self, feedback_fn):
5490 """Dump a representation of the cluster config to the standard output.
5494 for field in self.op.output_fields:
5495 if field == "cluster_name":
5496 entry = self.cfg.GetClusterName()
5497 elif field == "master_node":
5498 entry = self.cfg.GetMasterNode()
5499 elif field == "drain_flag":
5500 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
5501 elif field == "watcher_pause":
5502 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
5503 elif field == "volume_group_name":
5504 entry = self.cfg.GetVGName()
5506 raise errors.ParameterError(field)
5507 values.append(entry)
5511 class LUInstanceActivateDisks(NoHooksLU):
5512 """Bring up an instance's disks.
5517 def ExpandNames(self):
5518 self._ExpandAndLockInstance()
5519 self.needed_locks[locking.LEVEL_NODE] = []
5520 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5522 def DeclareLocks(self, level):
5523 if level == locking.LEVEL_NODE:
5524 self._LockInstancesNodes()
5526 def CheckPrereq(self):
5527 """Check prerequisites.
5529 This checks that the instance is in the cluster.
5532 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5533 assert self.instance is not None, \
5534 "Cannot retrieve locked instance %s" % self.op.instance_name
5535 _CheckNodeOnline(self, self.instance.primary_node)
5537 def Exec(self, feedback_fn):
5538 """Activate the disks.
5541 disks_ok, disks_info = \
5542 _AssembleInstanceDisks(self, self.instance,
5543 ignore_size=self.op.ignore_size)
5545 raise errors.OpExecError("Cannot activate block devices")
5550 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
5552 """Prepare the block devices for an instance.
5554 This sets up the block devices on all nodes.
5556 @type lu: L{LogicalUnit}
5557 @param lu: the logical unit on whose behalf we execute
5558 @type instance: L{objects.Instance}
5559 @param instance: the instance for whose disks we assemble
5560 @type disks: list of L{objects.Disk} or None
5561 @param disks: which disks to assemble (or all, if None)
5562 @type ignore_secondaries: boolean
5563 @param ignore_secondaries: if true, errors on secondary nodes
5564 won't result in an error return from the function
5565 @type ignore_size: boolean
5566 @param ignore_size: if true, the current known size of the disk
5567 will not be used during the disk activation, useful for cases
5568 when the size is wrong
5569 @return: False if the operation failed, otherwise a list of
5570 (host, instance_visible_name, node_visible_name)
5571 with the mapping from node devices to instance devices
5576 iname = instance.name
5577 disks = _ExpandCheckDisks(instance, disks)
5579 # With the two passes mechanism we try to reduce the window of
5580 # opportunity for the race condition of switching DRBD to primary
5581 # before handshaking occured, but we do not eliminate it
5583 # The proper fix would be to wait (with some limits) until the
5584 # connection has been made and drbd transitions from WFConnection
5585 # into any other network-connected state (Connected, SyncTarget,
5588 # 1st pass, assemble on all nodes in secondary mode
5589 for idx, inst_disk in enumerate(disks):
5590 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
5592 node_disk = node_disk.Copy()
5593 node_disk.UnsetSize()
5594 lu.cfg.SetDiskID(node_disk, node)
5595 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
5596 msg = result.fail_msg
5598 lu.proc.LogWarning("Could not prepare block device %s on node %s"
5599 " (is_primary=False, pass=1): %s",
5600 inst_disk.iv_name, node, msg)
5601 if not ignore_secondaries:
5604 # FIXME: race condition on drbd migration to primary
5606 # 2nd pass, do only the primary node
5607 for idx, inst_disk in enumerate(disks):
5610 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
5611 if node != instance.primary_node:
5614 node_disk = node_disk.Copy()
5615 node_disk.UnsetSize()
5616 lu.cfg.SetDiskID(node_disk, node)
5617 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
5618 msg = result.fail_msg
5620 lu.proc.LogWarning("Could not prepare block device %s on node %s"
5621 " (is_primary=True, pass=2): %s",
5622 inst_disk.iv_name, node, msg)
5625 dev_path = result.payload
5627 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
5629 # leave the disks configured for the primary node
5630 # this is a workaround that would be fixed better by
5631 # improving the logical/physical id handling
5633 lu.cfg.SetDiskID(disk, instance.primary_node)
5635 return disks_ok, device_info
5638 def _StartInstanceDisks(lu, instance, force):
5639 """Start the disks of an instance.
5642 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
5643 ignore_secondaries=force)
5645 _ShutdownInstanceDisks(lu, instance)
5646 if force is not None and not force:
5647 lu.proc.LogWarning("", hint="If the message above refers to a"
5649 " you can retry the operation using '--force'.")
5650 raise errors.OpExecError("Disk consistency error")
5653 class LUInstanceDeactivateDisks(NoHooksLU):
5654 """Shutdown an instance's disks.
5659 def ExpandNames(self):
5660 self._ExpandAndLockInstance()
5661 self.needed_locks[locking.LEVEL_NODE] = []
5662 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5664 def DeclareLocks(self, level):
5665 if level == locking.LEVEL_NODE:
5666 self._LockInstancesNodes()
5668 def CheckPrereq(self):
5669 """Check prerequisites.
5671 This checks that the instance is in the cluster.
5674 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5675 assert self.instance is not None, \
5676 "Cannot retrieve locked instance %s" % self.op.instance_name
5678 def Exec(self, feedback_fn):
5679 """Deactivate the disks
5682 instance = self.instance
5684 _ShutdownInstanceDisks(self, instance)
5686 _SafeShutdownInstanceDisks(self, instance)
5689 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
5690 """Shutdown block devices of an instance.
5692 This function checks if an instance is running, before calling
5693 _ShutdownInstanceDisks.
5696 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
5697 _ShutdownInstanceDisks(lu, instance, disks=disks)
5700 def _ExpandCheckDisks(instance, disks):
5701 """Return the instance disks selected by the disks list
5703 @type disks: list of L{objects.Disk} or None
5704 @param disks: selected disks
5705 @rtype: list of L{objects.Disk}
5706 @return: selected instance disks to act on
5710 return instance.disks
5712 if not set(disks).issubset(instance.disks):
5713 raise errors.ProgrammerError("Can only act on disks belonging to the"
5718 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
5719 """Shutdown block devices of an instance.
5721 This does the shutdown on all nodes of the instance.
5723 If the ignore_primary is false, errors on the primary node are
5728 disks = _ExpandCheckDisks(instance, disks)
5731 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
5732 lu.cfg.SetDiskID(top_disk, node)
5733 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
5734 msg = result.fail_msg
5736 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
5737 disk.iv_name, node, msg)
5738 if ((node == instance.primary_node and not ignore_primary) or
5739 (node != instance.primary_node and not result.offline)):
5744 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
5745 """Checks if a node has enough free memory.
5747 This function check if a given node has the needed amount of free
5748 memory. In case the node has less memory or we cannot get the
5749 information from the node, this function raise an OpPrereqError
5752 @type lu: C{LogicalUnit}
5753 @param lu: a logical unit from which we get configuration data
5755 @param node: the node to check
5756 @type reason: C{str}
5757 @param reason: string to use in the error message
5758 @type requested: C{int}
5759 @param requested: the amount of memory in MiB to check for
5760 @type hypervisor_name: C{str}
5761 @param hypervisor_name: the hypervisor to ask for memory stats
5762 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
5763 we cannot check the node
5766 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
5767 nodeinfo[node].Raise("Can't get data from node %s" % node,
5768 prereq=True, ecode=errors.ECODE_ENVIRON)
5769 free_mem = nodeinfo[node].payload.get("memory_free", None)
5770 if not isinstance(free_mem, int):
5771 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
5772 " was '%s'" % (node, free_mem),
5773 errors.ECODE_ENVIRON)
5774 if requested > free_mem:
5775 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
5776 " needed %s MiB, available %s MiB" %
5777 (node, reason, requested, free_mem),
5781 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5782 """Checks if nodes have enough free disk space in the all VGs.
5784 This function check if all given nodes have the needed amount of
5785 free disk. In case any node has less disk or we cannot get the
5786 information from the node, this function raise an OpPrereqError
5789 @type lu: C{LogicalUnit}
5790 @param lu: a logical unit from which we get configuration data
5791 @type nodenames: C{list}
5792 @param nodenames: the list of node names to check
5793 @type req_sizes: C{dict}
5794 @param req_sizes: the hash of vg and corresponding amount of disk in
5796 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5797 or we cannot check the node
5800 for vg, req_size in req_sizes.items():
5801 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5804 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5805 """Checks if nodes have enough free disk space in the specified VG.
5807 This function check if all given nodes have the needed amount of
5808 free disk. In case any node has less disk or we cannot get the
5809 information from the node, this function raise an OpPrereqError
5812 @type lu: C{LogicalUnit}
5813 @param lu: a logical unit from which we get configuration data
5814 @type nodenames: C{list}
5815 @param nodenames: the list of node names to check
5817 @param vg: the volume group to check
5818 @type requested: C{int}
5819 @param requested: the amount of disk in MiB to check for
5820 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5821 or we cannot check the node
5824 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
5825 for node in nodenames:
5826 info = nodeinfo[node]
5827 info.Raise("Cannot get current information from node %s" % node,
5828 prereq=True, ecode=errors.ECODE_ENVIRON)
5829 vg_free = info.payload.get("vg_free", None)
5830 if not isinstance(vg_free, int):
5831 raise errors.OpPrereqError("Can't compute free disk space on node"
5832 " %s for vg %s, result was '%s'" %
5833 (node, vg, vg_free), errors.ECODE_ENVIRON)
5834 if requested > vg_free:
5835 raise errors.OpPrereqError("Not enough disk space on target node %s"
5836 " vg %s: required %d MiB, available %d MiB" %
5837 (node, vg, requested, vg_free),
5841 class LUInstanceStartup(LogicalUnit):
5842 """Starts an instance.
5845 HPATH = "instance-start"
5846 HTYPE = constants.HTYPE_INSTANCE
5849 def CheckArguments(self):
5851 if self.op.beparams:
5852 # fill the beparams dict
5853 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5855 def ExpandNames(self):
5856 self._ExpandAndLockInstance()
5858 def BuildHooksEnv(self):
5861 This runs on master, primary and secondary nodes of the instance.
5865 "FORCE": self.op.force,
5868 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5872 def BuildHooksNodes(self):
5873 """Build hooks nodes.
5876 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5879 def CheckPrereq(self):
5880 """Check prerequisites.
5882 This checks that the instance is in the cluster.
5885 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5886 assert self.instance is not None, \
5887 "Cannot retrieve locked instance %s" % self.op.instance_name
5890 if self.op.hvparams:
5891 # check hypervisor parameter syntax (locally)
5892 cluster = self.cfg.GetClusterInfo()
5893 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5894 filled_hvp = cluster.FillHV(instance)
5895 filled_hvp.update(self.op.hvparams)
5896 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5897 hv_type.CheckParameterSyntax(filled_hvp)
5898 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5900 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5902 if self.primary_offline and self.op.ignore_offline_nodes:
5903 self.proc.LogWarning("Ignoring offline primary node")
5905 if self.op.hvparams or self.op.beparams:
5906 self.proc.LogWarning("Overridden parameters are ignored")
5908 _CheckNodeOnline(self, instance.primary_node)
5910 bep = self.cfg.GetClusterInfo().FillBE(instance)
5912 # check bridges existence
5913 _CheckInstanceBridgesExist(self, instance)
5915 remote_info = self.rpc.call_instance_info(instance.primary_node,
5917 instance.hypervisor)
5918 remote_info.Raise("Error checking node %s" % instance.primary_node,
5919 prereq=True, ecode=errors.ECODE_ENVIRON)
5920 if not remote_info.payload: # not running already
5921 _CheckNodeFreeMemory(self, instance.primary_node,
5922 "starting instance %s" % instance.name,
5923 bep[constants.BE_MEMORY], instance.hypervisor)
5925 def Exec(self, feedback_fn):
5926 """Start the instance.
5929 instance = self.instance
5930 force = self.op.force
5932 if not self.op.no_remember:
5933 self.cfg.MarkInstanceUp(instance.name)
5935 if self.primary_offline:
5936 assert self.op.ignore_offline_nodes
5937 self.proc.LogInfo("Primary node offline, marked instance as started")
5939 node_current = instance.primary_node
5941 _StartInstanceDisks(self, instance, force)
5943 result = self.rpc.call_instance_start(node_current, instance,
5944 self.op.hvparams, self.op.beparams,
5945 self.op.startup_paused)
5946 msg = result.fail_msg
5948 _ShutdownInstanceDisks(self, instance)
5949 raise errors.OpExecError("Could not start instance: %s" % msg)
5952 class LUInstanceReboot(LogicalUnit):
5953 """Reboot an instance.
5956 HPATH = "instance-reboot"
5957 HTYPE = constants.HTYPE_INSTANCE
5960 def ExpandNames(self):
5961 self._ExpandAndLockInstance()
5963 def BuildHooksEnv(self):
5966 This runs on master, primary and secondary nodes of the instance.
5970 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
5971 "REBOOT_TYPE": self.op.reboot_type,
5972 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5975 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5979 def BuildHooksNodes(self):
5980 """Build hooks nodes.
5983 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5986 def CheckPrereq(self):
5987 """Check prerequisites.
5989 This checks that the instance is in the cluster.
5992 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5993 assert self.instance is not None, \
5994 "Cannot retrieve locked instance %s" % self.op.instance_name
5996 _CheckNodeOnline(self, instance.primary_node)
5998 # check bridges existence
5999 _CheckInstanceBridgesExist(self, instance)
6001 def Exec(self, feedback_fn):
6002 """Reboot the instance.
6005 instance = self.instance
6006 ignore_secondaries = self.op.ignore_secondaries
6007 reboot_type = self.op.reboot_type
6009 remote_info = self.rpc.call_instance_info(instance.primary_node,
6011 instance.hypervisor)
6012 remote_info.Raise("Error checking node %s" % instance.primary_node)
6013 instance_running = bool(remote_info.payload)
6015 node_current = instance.primary_node
6017 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
6018 constants.INSTANCE_REBOOT_HARD]:
6019 for disk in instance.disks:
6020 self.cfg.SetDiskID(disk, node_current)
6021 result = self.rpc.call_instance_reboot(node_current, instance,
6023 self.op.shutdown_timeout)
6024 result.Raise("Could not reboot instance")
6026 if instance_running:
6027 result = self.rpc.call_instance_shutdown(node_current, instance,
6028 self.op.shutdown_timeout)
6029 result.Raise("Could not shutdown instance for full reboot")
6030 _ShutdownInstanceDisks(self, instance)
6032 self.LogInfo("Instance %s was already stopped, starting now",
6034 _StartInstanceDisks(self, instance, ignore_secondaries)
6035 result = self.rpc.call_instance_start(node_current, instance,
6037 msg = result.fail_msg
6039 _ShutdownInstanceDisks(self, instance)
6040 raise errors.OpExecError("Could not start instance for"
6041 " full reboot: %s" % msg)
6043 self.cfg.MarkInstanceUp(instance.name)
6046 class LUInstanceShutdown(LogicalUnit):
6047 """Shutdown an instance.
6050 HPATH = "instance-stop"
6051 HTYPE = constants.HTYPE_INSTANCE
6054 def ExpandNames(self):
6055 self._ExpandAndLockInstance()
6057 def BuildHooksEnv(self):
6060 This runs on master, primary and secondary nodes of the instance.
6063 env = _BuildInstanceHookEnvByObject(self, self.instance)
6064 env["TIMEOUT"] = self.op.timeout
6067 def BuildHooksNodes(self):
6068 """Build hooks nodes.
6071 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6074 def CheckPrereq(self):
6075 """Check prerequisites.
6077 This checks that the instance is in the cluster.
6080 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6081 assert self.instance is not None, \
6082 "Cannot retrieve locked instance %s" % self.op.instance_name
6084 self.primary_offline = \
6085 self.cfg.GetNodeInfo(self.instance.primary_node).offline
6087 if self.primary_offline and self.op.ignore_offline_nodes:
6088 self.proc.LogWarning("Ignoring offline primary node")
6090 _CheckNodeOnline(self, self.instance.primary_node)
6092 def Exec(self, feedback_fn):
6093 """Shutdown the instance.
6096 instance = self.instance
6097 node_current = instance.primary_node
6098 timeout = self.op.timeout
6100 if not self.op.no_remember:
6101 self.cfg.MarkInstanceDown(instance.name)
6103 if self.primary_offline:
6104 assert self.op.ignore_offline_nodes
6105 self.proc.LogInfo("Primary node offline, marked instance as stopped")
6107 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
6108 msg = result.fail_msg
6110 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
6112 _ShutdownInstanceDisks(self, instance)
6115 class LUInstanceReinstall(LogicalUnit):
6116 """Reinstall an instance.
6119 HPATH = "instance-reinstall"
6120 HTYPE = constants.HTYPE_INSTANCE
6123 def ExpandNames(self):
6124 self._ExpandAndLockInstance()
6126 def BuildHooksEnv(self):
6129 This runs on master, primary and secondary nodes of the instance.
6132 return _BuildInstanceHookEnvByObject(self, self.instance)
6134 def BuildHooksNodes(self):
6135 """Build hooks nodes.
6138 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6141 def CheckPrereq(self):
6142 """Check prerequisites.
6144 This checks that the instance is in the cluster and is not running.
6147 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6148 assert instance is not None, \
6149 "Cannot retrieve locked instance %s" % self.op.instance_name
6150 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
6151 " offline, cannot reinstall")
6152 for node in instance.secondary_nodes:
6153 _CheckNodeOnline(self, node, "Instance secondary node offline,"
6154 " cannot reinstall")
6156 if instance.disk_template == constants.DT_DISKLESS:
6157 raise errors.OpPrereqError("Instance '%s' has no disks" %
6158 self.op.instance_name,
6160 _CheckInstanceDown(self, instance, "cannot reinstall")
6162 if self.op.os_type is not None:
6164 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
6165 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
6166 instance_os = self.op.os_type
6168 instance_os = instance.os
6170 nodelist = list(instance.all_nodes)
6172 if self.op.osparams:
6173 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
6174 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
6175 self.os_inst = i_osdict # the new dict (without defaults)
6179 self.instance = instance
6181 def Exec(self, feedback_fn):
6182 """Reinstall the instance.
6185 inst = self.instance
6187 if self.op.os_type is not None:
6188 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
6189 inst.os = self.op.os_type
6190 # Write to configuration
6191 self.cfg.Update(inst, feedback_fn)
6193 _StartInstanceDisks(self, inst, None)
6195 feedback_fn("Running the instance OS create scripts...")
6196 # FIXME: pass debug option from opcode to backend
6197 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
6198 self.op.debug_level,
6199 osparams=self.os_inst)
6200 result.Raise("Could not install OS for instance %s on node %s" %
6201 (inst.name, inst.primary_node))
6203 _ShutdownInstanceDisks(self, inst)
6206 class LUInstanceRecreateDisks(LogicalUnit):
6207 """Recreate an instance's missing disks.
6210 HPATH = "instance-recreate-disks"
6211 HTYPE = constants.HTYPE_INSTANCE
6214 def CheckArguments(self):
6215 # normalise the disk list
6216 self.op.disks = sorted(frozenset(self.op.disks))
6218 def ExpandNames(self):
6219 self._ExpandAndLockInstance()
6220 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6222 self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes]
6223 self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes)
6225 self.needed_locks[locking.LEVEL_NODE] = []
6227 def DeclareLocks(self, level):
6228 if level == locking.LEVEL_NODE:
6229 # if we replace the nodes, we only need to lock the old primary,
6230 # otherwise we need to lock all nodes for disk re-creation
6231 primary_only = bool(self.op.nodes)
6232 self._LockInstancesNodes(primary_only=primary_only)
6234 def BuildHooksEnv(self):
6237 This runs on master, primary and secondary nodes of the instance.
6240 return _BuildInstanceHookEnvByObject(self, self.instance)
6242 def BuildHooksNodes(self):
6243 """Build hooks nodes.
6246 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6249 def CheckPrereq(self):
6250 """Check prerequisites.
6252 This checks that the instance is in the cluster and is not running.
6255 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6256 assert instance is not None, \
6257 "Cannot retrieve locked instance %s" % self.op.instance_name
6259 if len(self.op.nodes) != len(instance.all_nodes):
6260 raise errors.OpPrereqError("Instance %s currently has %d nodes, but"
6261 " %d replacement nodes were specified" %
6262 (instance.name, len(instance.all_nodes),
6263 len(self.op.nodes)),
6265 assert instance.disk_template != constants.DT_DRBD8 or \
6266 len(self.op.nodes) == 2
6267 assert instance.disk_template != constants.DT_PLAIN or \
6268 len(self.op.nodes) == 1
6269 primary_node = self.op.nodes[0]
6271 primary_node = instance.primary_node
6272 _CheckNodeOnline(self, primary_node)
6274 if instance.disk_template == constants.DT_DISKLESS:
6275 raise errors.OpPrereqError("Instance '%s' has no disks" %
6276 self.op.instance_name, errors.ECODE_INVAL)
6277 # if we replace nodes *and* the old primary is offline, we don't
6279 assert instance.primary_node in self.needed_locks[locking.LEVEL_NODE]
6280 old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
6281 if not (self.op.nodes and old_pnode.offline):
6282 _CheckInstanceDown(self, instance, "cannot recreate disks")
6284 if not self.op.disks:
6285 self.op.disks = range(len(instance.disks))
6287 for idx in self.op.disks:
6288 if idx >= len(instance.disks):
6289 raise errors.OpPrereqError("Invalid disk index '%s'" % idx,
6291 if self.op.disks != range(len(instance.disks)) and self.op.nodes:
6292 raise errors.OpPrereqError("Can't recreate disks partially and"
6293 " change the nodes at the same time",
6295 self.instance = instance
6297 def Exec(self, feedback_fn):
6298 """Recreate the disks.
6301 instance = self.instance
6304 mods = [] # keeps track of needed logical_id changes
6306 for idx, disk in enumerate(instance.disks):
6307 if idx not in self.op.disks: # disk idx has not been passed in
6310 # update secondaries for disks, if needed
6312 if disk.dev_type == constants.LD_DRBD8:
6313 # need to update the nodes and minors
6314 assert len(self.op.nodes) == 2
6315 assert len(disk.logical_id) == 6 # otherwise disk internals
6317 (_, _, old_port, _, _, old_secret) = disk.logical_id
6318 new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name)
6319 new_id = (self.op.nodes[0], self.op.nodes[1], old_port,
6320 new_minors[0], new_minors[1], old_secret)
6321 assert len(disk.logical_id) == len(new_id)
6322 mods.append((idx, new_id))
6324 # now that we have passed all asserts above, we can apply the mods
6325 # in a single run (to avoid partial changes)
6326 for idx, new_id in mods:
6327 instance.disks[idx].logical_id = new_id
6329 # change primary node, if needed
6331 instance.primary_node = self.op.nodes[0]
6332 self.LogWarning("Changing the instance's nodes, you will have to"
6333 " remove any disks left on the older nodes manually")
6336 self.cfg.Update(instance, feedback_fn)
6338 _CreateDisks(self, instance, to_skip=to_skip)
6341 class LUInstanceRename(LogicalUnit):
6342 """Rename an instance.
6345 HPATH = "instance-rename"
6346 HTYPE = constants.HTYPE_INSTANCE
6348 def CheckArguments(self):
6352 if self.op.ip_check and not self.op.name_check:
6353 # TODO: make the ip check more flexible and not depend on the name check
6354 raise errors.OpPrereqError("IP address check requires a name check",
6357 def BuildHooksEnv(self):
6360 This runs on master, primary and secondary nodes of the instance.
6363 env = _BuildInstanceHookEnvByObject(self, self.instance)
6364 env["INSTANCE_NEW_NAME"] = self.op.new_name
6367 def BuildHooksNodes(self):
6368 """Build hooks nodes.
6371 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6374 def CheckPrereq(self):
6375 """Check prerequisites.
6377 This checks that the instance is in the cluster and is not running.
6380 self.op.instance_name = _ExpandInstanceName(self.cfg,
6381 self.op.instance_name)
6382 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6383 assert instance is not None
6384 _CheckNodeOnline(self, instance.primary_node)
6385 _CheckInstanceDown(self, instance, "cannot rename")
6386 self.instance = instance
6388 new_name = self.op.new_name
6389 if self.op.name_check:
6390 hostname = netutils.GetHostname(name=new_name)
6391 if hostname != new_name:
6392 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
6394 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
6395 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
6396 " same as given hostname '%s'") %
6397 (hostname.name, self.op.new_name),
6399 new_name = self.op.new_name = hostname.name
6400 if (self.op.ip_check and
6401 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
6402 raise errors.OpPrereqError("IP %s of instance %s already in use" %
6403 (hostname.ip, new_name),
6404 errors.ECODE_NOTUNIQUE)
6406 instance_list = self.cfg.GetInstanceList()
6407 if new_name in instance_list and new_name != instance.name:
6408 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6409 new_name, errors.ECODE_EXISTS)
6411 def Exec(self, feedback_fn):
6412 """Rename the instance.
6415 inst = self.instance
6416 old_name = inst.name
6418 rename_file_storage = False
6419 if (inst.disk_template in constants.DTS_FILEBASED and
6420 self.op.new_name != inst.name):
6421 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
6422 rename_file_storage = True
6424 self.cfg.RenameInstance(inst.name, self.op.new_name)
6425 # Change the instance lock. This is definitely safe while we hold the BGL.
6426 # Otherwise the new lock would have to be added in acquired mode.
6428 self.glm.remove(locking.LEVEL_INSTANCE, old_name)
6429 self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
6431 # re-read the instance from the configuration after rename
6432 inst = self.cfg.GetInstanceInfo(self.op.new_name)
6434 if rename_file_storage:
6435 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
6436 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
6437 old_file_storage_dir,
6438 new_file_storage_dir)
6439 result.Raise("Could not rename on node %s directory '%s' to '%s'"
6440 " (but the instance has been renamed in Ganeti)" %
6441 (inst.primary_node, old_file_storage_dir,
6442 new_file_storage_dir))
6444 _StartInstanceDisks(self, inst, None)
6446 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
6447 old_name, self.op.debug_level)
6448 msg = result.fail_msg
6450 msg = ("Could not run OS rename script for instance %s on node %s"
6451 " (but the instance has been renamed in Ganeti): %s" %
6452 (inst.name, inst.primary_node, msg))
6453 self.proc.LogWarning(msg)
6455 _ShutdownInstanceDisks(self, inst)
6460 class LUInstanceRemove(LogicalUnit):
6461 """Remove an instance.
6464 HPATH = "instance-remove"
6465 HTYPE = constants.HTYPE_INSTANCE
6468 def ExpandNames(self):
6469 self._ExpandAndLockInstance()
6470 self.needed_locks[locking.LEVEL_NODE] = []
6471 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6473 def DeclareLocks(self, level):
6474 if level == locking.LEVEL_NODE:
6475 self._LockInstancesNodes()
6477 def BuildHooksEnv(self):
6480 This runs on master, primary and secondary nodes of the instance.
6483 env = _BuildInstanceHookEnvByObject(self, self.instance)
6484 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
6487 def BuildHooksNodes(self):
6488 """Build hooks nodes.
6491 nl = [self.cfg.GetMasterNode()]
6492 nl_post = list(self.instance.all_nodes) + nl
6493 return (nl, nl_post)
6495 def CheckPrereq(self):
6496 """Check prerequisites.
6498 This checks that the instance is in the cluster.
6501 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6502 assert self.instance is not None, \
6503 "Cannot retrieve locked instance %s" % self.op.instance_name
6505 def Exec(self, feedback_fn):
6506 """Remove the instance.
6509 instance = self.instance
6510 logging.info("Shutting down instance %s on node %s",
6511 instance.name, instance.primary_node)
6513 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
6514 self.op.shutdown_timeout)
6515 msg = result.fail_msg
6517 if self.op.ignore_failures:
6518 feedback_fn("Warning: can't shutdown instance: %s" % msg)
6520 raise errors.OpExecError("Could not shutdown instance %s on"
6522 (instance.name, instance.primary_node, msg))
6524 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
6527 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
6528 """Utility function to remove an instance.
6531 logging.info("Removing block devices for instance %s", instance.name)
6533 if not _RemoveDisks(lu, instance):
6534 if not ignore_failures:
6535 raise errors.OpExecError("Can't remove instance's disks")
6536 feedback_fn("Warning: can't remove instance's disks")
6538 logging.info("Removing instance %s out of cluster config", instance.name)
6540 lu.cfg.RemoveInstance(instance.name)
6542 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
6543 "Instance lock removal conflict"
6545 # Remove lock for the instance
6546 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
6549 class LUInstanceQuery(NoHooksLU):
6550 """Logical unit for querying instances.
6553 # pylint: disable=W0142
6556 def CheckArguments(self):
6557 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
6558 self.op.output_fields, self.op.use_locking)
6560 def ExpandNames(self):
6561 self.iq.ExpandNames(self)
6563 def DeclareLocks(self, level):
6564 self.iq.DeclareLocks(self, level)
6566 def Exec(self, feedback_fn):
6567 return self.iq.OldStyleQuery(self)
6570 class LUInstanceFailover(LogicalUnit):
6571 """Failover an instance.
6574 HPATH = "instance-failover"
6575 HTYPE = constants.HTYPE_INSTANCE
6578 def CheckArguments(self):
6579 """Check the arguments.
6582 self.iallocator = getattr(self.op, "iallocator", None)
6583 self.target_node = getattr(self.op, "target_node", None)
6585 def ExpandNames(self):
6586 self._ExpandAndLockInstance()
6588 if self.op.target_node is not None:
6589 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6591 self.needed_locks[locking.LEVEL_NODE] = []
6592 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6594 ignore_consistency = self.op.ignore_consistency
6595 shutdown_timeout = self.op.shutdown_timeout
6596 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6599 ignore_consistency=ignore_consistency,
6600 shutdown_timeout=shutdown_timeout)
6601 self.tasklets = [self._migrater]
6603 def DeclareLocks(self, level):
6604 if level == locking.LEVEL_NODE:
6605 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6606 if instance.disk_template in constants.DTS_EXT_MIRROR:
6607 if self.op.target_node is None:
6608 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6610 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6611 self.op.target_node]
6612 del self.recalculate_locks[locking.LEVEL_NODE]
6614 self._LockInstancesNodes()
6616 def BuildHooksEnv(self):
6619 This runs on master, primary and secondary nodes of the instance.
6622 instance = self._migrater.instance
6623 source_node = instance.primary_node
6624 target_node = self.op.target_node
6626 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
6627 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6628 "OLD_PRIMARY": source_node,
6629 "NEW_PRIMARY": target_node,
6632 if instance.disk_template in constants.DTS_INT_MIRROR:
6633 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
6634 env["NEW_SECONDARY"] = source_node
6636 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
6638 env.update(_BuildInstanceHookEnvByObject(self, instance))
6642 def BuildHooksNodes(self):
6643 """Build hooks nodes.
6646 instance = self._migrater.instance
6647 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6648 return (nl, nl + [instance.primary_node])
6651 class LUInstanceMigrate(LogicalUnit):
6652 """Migrate an instance.
6654 This is migration without shutting down, compared to the failover,
6655 which is done with shutdown.
6658 HPATH = "instance-migrate"
6659 HTYPE = constants.HTYPE_INSTANCE
6662 def ExpandNames(self):
6663 self._ExpandAndLockInstance()
6665 if self.op.target_node is not None:
6666 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6668 self.needed_locks[locking.LEVEL_NODE] = []
6669 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6671 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6672 cleanup=self.op.cleanup,
6674 fallback=self.op.allow_failover)
6675 self.tasklets = [self._migrater]
6677 def DeclareLocks(self, level):
6678 if level == locking.LEVEL_NODE:
6679 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6680 if instance.disk_template in constants.DTS_EXT_MIRROR:
6681 if self.op.target_node is None:
6682 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6684 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6685 self.op.target_node]
6686 del self.recalculate_locks[locking.LEVEL_NODE]
6688 self._LockInstancesNodes()
6690 def BuildHooksEnv(self):
6693 This runs on master, primary and secondary nodes of the instance.
6696 instance = self._migrater.instance
6697 source_node = instance.primary_node
6698 target_node = self.op.target_node
6699 env = _BuildInstanceHookEnvByObject(self, instance)
6701 "MIGRATE_LIVE": self._migrater.live,
6702 "MIGRATE_CLEANUP": self.op.cleanup,
6703 "OLD_PRIMARY": source_node,
6704 "NEW_PRIMARY": target_node,
6707 if instance.disk_template in constants.DTS_INT_MIRROR:
6708 env["OLD_SECONDARY"] = target_node
6709 env["NEW_SECONDARY"] = source_node
6711 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
6715 def BuildHooksNodes(self):
6716 """Build hooks nodes.
6719 instance = self._migrater.instance
6720 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6721 return (nl, nl + [instance.primary_node])
6724 class LUInstanceMove(LogicalUnit):
6725 """Move an instance by data-copying.
6728 HPATH = "instance-move"
6729 HTYPE = constants.HTYPE_INSTANCE
6732 def ExpandNames(self):
6733 self._ExpandAndLockInstance()
6734 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6735 self.op.target_node = target_node
6736 self.needed_locks[locking.LEVEL_NODE] = [target_node]
6737 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6739 def DeclareLocks(self, level):
6740 if level == locking.LEVEL_NODE:
6741 self._LockInstancesNodes(primary_only=True)
6743 def BuildHooksEnv(self):
6746 This runs on master, primary and secondary nodes of the instance.
6750 "TARGET_NODE": self.op.target_node,
6751 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6753 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6756 def BuildHooksNodes(self):
6757 """Build hooks nodes.
6761 self.cfg.GetMasterNode(),
6762 self.instance.primary_node,
6763 self.op.target_node,
6767 def CheckPrereq(self):
6768 """Check prerequisites.
6770 This checks that the instance is in the cluster.
6773 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6774 assert self.instance is not None, \
6775 "Cannot retrieve locked instance %s" % self.op.instance_name
6777 node = self.cfg.GetNodeInfo(self.op.target_node)
6778 assert node is not None, \
6779 "Cannot retrieve locked node %s" % self.op.target_node
6781 self.target_node = target_node = node.name
6783 if target_node == instance.primary_node:
6784 raise errors.OpPrereqError("Instance %s is already on the node %s" %
6785 (instance.name, target_node),
6788 bep = self.cfg.GetClusterInfo().FillBE(instance)
6790 for idx, dsk in enumerate(instance.disks):
6791 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
6792 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
6793 " cannot copy" % idx, errors.ECODE_STATE)
6795 _CheckNodeOnline(self, target_node)
6796 _CheckNodeNotDrained(self, target_node)
6797 _CheckNodeVmCapable(self, target_node)
6799 if instance.admin_up:
6800 # check memory requirements on the secondary node
6801 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
6802 instance.name, bep[constants.BE_MEMORY],
6803 instance.hypervisor)
6805 self.LogInfo("Not checking memory on the secondary node as"
6806 " instance will not be started")
6808 # check bridge existance
6809 _CheckInstanceBridgesExist(self, instance, node=target_node)
6811 def Exec(self, feedback_fn):
6812 """Move an instance.
6814 The move is done by shutting it down on its present node, copying
6815 the data over (slow) and starting it on the new node.
6818 instance = self.instance
6820 source_node = instance.primary_node
6821 target_node = self.target_node
6823 self.LogInfo("Shutting down instance %s on source node %s",
6824 instance.name, source_node)
6826 result = self.rpc.call_instance_shutdown(source_node, instance,
6827 self.op.shutdown_timeout)
6828 msg = result.fail_msg
6830 if self.op.ignore_consistency:
6831 self.proc.LogWarning("Could not shutdown instance %s on node %s."
6832 " Proceeding anyway. Please make sure node"
6833 " %s is down. Error details: %s",
6834 instance.name, source_node, source_node, msg)
6836 raise errors.OpExecError("Could not shutdown instance %s on"
6838 (instance.name, source_node, msg))
6840 # create the target disks
6842 _CreateDisks(self, instance, target_node=target_node)
6843 except errors.OpExecError:
6844 self.LogWarning("Device creation failed, reverting...")
6846 _RemoveDisks(self, instance, target_node=target_node)
6848 self.cfg.ReleaseDRBDMinors(instance.name)
6851 cluster_name = self.cfg.GetClusterInfo().cluster_name
6854 # activate, get path, copy the data over
6855 for idx, disk in enumerate(instance.disks):
6856 self.LogInfo("Copying data for disk %d", idx)
6857 result = self.rpc.call_blockdev_assemble(target_node, disk,
6858 instance.name, True, idx)
6860 self.LogWarning("Can't assemble newly created disk %d: %s",
6861 idx, result.fail_msg)
6862 errs.append(result.fail_msg)
6864 dev_path = result.payload
6865 result = self.rpc.call_blockdev_export(source_node, disk,
6866 target_node, dev_path,
6869 self.LogWarning("Can't copy data over for disk %d: %s",
6870 idx, result.fail_msg)
6871 errs.append(result.fail_msg)
6875 self.LogWarning("Some disks failed to copy, aborting")
6877 _RemoveDisks(self, instance, target_node=target_node)
6879 self.cfg.ReleaseDRBDMinors(instance.name)
6880 raise errors.OpExecError("Errors during disk copy: %s" %
6883 instance.primary_node = target_node
6884 self.cfg.Update(instance, feedback_fn)
6886 self.LogInfo("Removing the disks on the original node")
6887 _RemoveDisks(self, instance, target_node=source_node)
6889 # Only start the instance if it's marked as up
6890 if instance.admin_up:
6891 self.LogInfo("Starting instance %s on node %s",
6892 instance.name, target_node)
6894 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6895 ignore_secondaries=True)
6897 _ShutdownInstanceDisks(self, instance)
6898 raise errors.OpExecError("Can't activate the instance's disks")
6900 result = self.rpc.call_instance_start(target_node, instance,
6902 msg = result.fail_msg
6904 _ShutdownInstanceDisks(self, instance)
6905 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6906 (instance.name, target_node, msg))
6909 class LUNodeMigrate(LogicalUnit):
6910 """Migrate all instances from a node.
6913 HPATH = "node-migrate"
6914 HTYPE = constants.HTYPE_NODE
6917 def CheckArguments(self):
6920 def ExpandNames(self):
6921 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6923 self.share_locks = _ShareAll()
6924 self.needed_locks = {
6925 locking.LEVEL_NODE: [self.op.node_name],
6928 def BuildHooksEnv(self):
6931 This runs on the master, the primary and all the secondaries.
6935 "NODE_NAME": self.op.node_name,
6938 def BuildHooksNodes(self):
6939 """Build hooks nodes.
6942 nl = [self.cfg.GetMasterNode()]
6945 def CheckPrereq(self):
6948 def Exec(self, feedback_fn):
6949 # Prepare jobs for migration instances
6951 [opcodes.OpInstanceMigrate(instance_name=inst.name,
6954 iallocator=self.op.iallocator,
6955 target_node=self.op.target_node)]
6956 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name)
6959 # TODO: Run iallocator in this opcode and pass correct placement options to
6960 # OpInstanceMigrate. Since other jobs can modify the cluster between
6961 # running the iallocator and the actual migration, a good consistency model
6962 # will have to be found.
6964 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
6965 frozenset([self.op.node_name]))
6967 return ResultWithJobs(jobs)
6970 class TLMigrateInstance(Tasklet):
6971 """Tasklet class for instance migration.
6974 @ivar live: whether the migration will be done live or non-live;
6975 this variable is initalized only after CheckPrereq has run
6976 @type cleanup: boolean
6977 @ivar cleanup: Wheater we cleanup from a failed migration
6978 @type iallocator: string
6979 @ivar iallocator: The iallocator used to determine target_node
6980 @type target_node: string
6981 @ivar target_node: If given, the target_node to reallocate the instance to
6982 @type failover: boolean
6983 @ivar failover: Whether operation results in failover or migration
6984 @type fallback: boolean
6985 @ivar fallback: Whether fallback to failover is allowed if migration not
6987 @type ignore_consistency: boolean
6988 @ivar ignore_consistency: Wheter we should ignore consistency between source
6990 @type shutdown_timeout: int
6991 @ivar shutdown_timeout: In case of failover timeout of the shutdown
6994 def __init__(self, lu, instance_name, cleanup=False,
6995 failover=False, fallback=False,
6996 ignore_consistency=False,
6997 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT):
6998 """Initializes this class.
7001 Tasklet.__init__(self, lu)
7004 self.instance_name = instance_name
7005 self.cleanup = cleanup
7006 self.live = False # will be overridden later
7007 self.failover = failover
7008 self.fallback = fallback
7009 self.ignore_consistency = ignore_consistency
7010 self.shutdown_timeout = shutdown_timeout
7012 def CheckPrereq(self):
7013 """Check prerequisites.
7015 This checks that the instance is in the cluster.
7018 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
7019 instance = self.cfg.GetInstanceInfo(instance_name)
7020 assert instance is not None
7021 self.instance = instance
7023 if (not self.cleanup and not instance.admin_up and not self.failover and
7025 self.lu.LogInfo("Instance is marked down, fallback allowed, switching"
7027 self.failover = True
7029 if instance.disk_template not in constants.DTS_MIRRORED:
7034 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
7035 " %s" % (instance.disk_template, text),
7038 if instance.disk_template in constants.DTS_EXT_MIRROR:
7039 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
7041 if self.lu.op.iallocator:
7042 self._RunAllocator()
7044 # We set set self.target_node as it is required by
7046 self.target_node = self.lu.op.target_node
7048 # self.target_node is already populated, either directly or by the
7050 target_node = self.target_node
7051 if self.target_node == instance.primary_node:
7052 raise errors.OpPrereqError("Cannot migrate instance %s"
7053 " to its primary (%s)" %
7054 (instance.name, instance.primary_node))
7056 if len(self.lu.tasklets) == 1:
7057 # It is safe to release locks only when we're the only tasklet
7059 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
7060 keep=[instance.primary_node, self.target_node])
7063 secondary_nodes = instance.secondary_nodes
7064 if not secondary_nodes:
7065 raise errors.ConfigurationError("No secondary node but using"
7066 " %s disk template" %
7067 instance.disk_template)
7068 target_node = secondary_nodes[0]
7069 if self.lu.op.iallocator or (self.lu.op.target_node and
7070 self.lu.op.target_node != target_node):
7072 text = "failed over"
7075 raise errors.OpPrereqError("Instances with disk template %s cannot"
7076 " be %s to arbitrary nodes"
7077 " (neither an iallocator nor a target"
7078 " node can be passed)" %
7079 (instance.disk_template, text),
7082 i_be = self.cfg.GetClusterInfo().FillBE(instance)
7084 # check memory requirements on the secondary node
7085 if not self.failover or instance.admin_up:
7086 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
7087 instance.name, i_be[constants.BE_MEMORY],
7088 instance.hypervisor)
7090 self.lu.LogInfo("Not checking memory on the secondary node as"
7091 " instance will not be started")
7093 # check bridge existance
7094 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
7096 if not self.cleanup:
7097 _CheckNodeNotDrained(self.lu, target_node)
7098 if not self.failover:
7099 result = self.rpc.call_instance_migratable(instance.primary_node,
7101 if result.fail_msg and self.fallback:
7102 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
7104 self.failover = True
7106 result.Raise("Can't migrate, please use failover",
7107 prereq=True, ecode=errors.ECODE_STATE)
7109 assert not (self.failover and self.cleanup)
7111 if not self.failover:
7112 if self.lu.op.live is not None and self.lu.op.mode is not None:
7113 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
7114 " parameters are accepted",
7116 if self.lu.op.live is not None:
7118 self.lu.op.mode = constants.HT_MIGRATION_LIVE
7120 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
7121 # reset the 'live' parameter to None so that repeated
7122 # invocations of CheckPrereq do not raise an exception
7123 self.lu.op.live = None
7124 elif self.lu.op.mode is None:
7125 # read the default value from the hypervisor
7126 i_hv = self.cfg.GetClusterInfo().FillHV(self.instance,
7128 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
7130 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
7132 # Failover is never live
7135 def _RunAllocator(self):
7136 """Run the allocator based on input opcode.
7139 ial = IAllocator(self.cfg, self.rpc,
7140 mode=constants.IALLOCATOR_MODE_RELOC,
7141 name=self.instance_name,
7142 # TODO See why hail breaks with a single node below
7143 relocate_from=[self.instance.primary_node,
7144 self.instance.primary_node],
7147 ial.Run(self.lu.op.iallocator)
7150 raise errors.OpPrereqError("Can't compute nodes using"
7151 " iallocator '%s': %s" %
7152 (self.lu.op.iallocator, ial.info),
7154 if len(ial.result) != ial.required_nodes:
7155 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7156 " of nodes (%s), required %s" %
7157 (self.lu.op.iallocator, len(ial.result),
7158 ial.required_nodes), errors.ECODE_FAULT)
7159 self.target_node = ial.result[0]
7160 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7161 self.instance_name, self.lu.op.iallocator,
7162 utils.CommaJoin(ial.result))
7164 def _WaitUntilSync(self):
7165 """Poll with custom rpc for disk sync.
7167 This uses our own step-based rpc call.
7170 self.feedback_fn("* wait until resync is done")
7174 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
7176 self.instance.disks)
7178 for node, nres in result.items():
7179 nres.Raise("Cannot resync disks on node %s" % node)
7180 node_done, node_percent = nres.payload
7181 all_done = all_done and node_done
7182 if node_percent is not None:
7183 min_percent = min(min_percent, node_percent)
7185 if min_percent < 100:
7186 self.feedback_fn(" - progress: %.1f%%" % min_percent)
7189 def _EnsureSecondary(self, node):
7190 """Demote a node to secondary.
7193 self.feedback_fn("* switching node %s to secondary mode" % node)
7195 for dev in self.instance.disks:
7196 self.cfg.SetDiskID(dev, node)
7198 result = self.rpc.call_blockdev_close(node, self.instance.name,
7199 self.instance.disks)
7200 result.Raise("Cannot change disk to secondary on node %s" % node)
7202 def _GoStandalone(self):
7203 """Disconnect from the network.
7206 self.feedback_fn("* changing into standalone mode")
7207 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
7208 self.instance.disks)
7209 for node, nres in result.items():
7210 nres.Raise("Cannot disconnect disks node %s" % node)
7212 def _GoReconnect(self, multimaster):
7213 """Reconnect to the network.
7219 msg = "single-master"
7220 self.feedback_fn("* changing disks into %s mode" % msg)
7221 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
7222 self.instance.disks,
7223 self.instance.name, multimaster)
7224 for node, nres in result.items():
7225 nres.Raise("Cannot change disks config on node %s" % node)
7227 def _ExecCleanup(self):
7228 """Try to cleanup after a failed migration.
7230 The cleanup is done by:
7231 - check that the instance is running only on one node
7232 (and update the config if needed)
7233 - change disks on its secondary node to secondary
7234 - wait until disks are fully synchronized
7235 - disconnect from the network
7236 - change disks into single-master mode
7237 - wait again until disks are fully synchronized
7240 instance = self.instance
7241 target_node = self.target_node
7242 source_node = self.source_node
7244 # check running on only one node
7245 self.feedback_fn("* checking where the instance actually runs"
7246 " (if this hangs, the hypervisor might be in"
7248 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
7249 for node, result in ins_l.items():
7250 result.Raise("Can't contact node %s" % node)
7252 runningon_source = instance.name in ins_l[source_node].payload
7253 runningon_target = instance.name in ins_l[target_node].payload
7255 if runningon_source and runningon_target:
7256 raise errors.OpExecError("Instance seems to be running on two nodes,"
7257 " or the hypervisor is confused; you will have"
7258 " to ensure manually that it runs only on one"
7259 " and restart this operation")
7261 if not (runningon_source or runningon_target):
7262 raise errors.OpExecError("Instance does not seem to be running at all;"
7263 " in this case it's safer to repair by"
7264 " running 'gnt-instance stop' to ensure disk"
7265 " shutdown, and then restarting it")
7267 if runningon_target:
7268 # the migration has actually succeeded, we need to update the config
7269 self.feedback_fn("* instance running on secondary node (%s),"
7270 " updating config" % target_node)
7271 instance.primary_node = target_node
7272 self.cfg.Update(instance, self.feedback_fn)
7273 demoted_node = source_node
7275 self.feedback_fn("* instance confirmed to be running on its"
7276 " primary node (%s)" % source_node)
7277 demoted_node = target_node
7279 if instance.disk_template in constants.DTS_INT_MIRROR:
7280 self._EnsureSecondary(demoted_node)
7282 self._WaitUntilSync()
7283 except errors.OpExecError:
7284 # we ignore here errors, since if the device is standalone, it
7285 # won't be able to sync
7287 self._GoStandalone()
7288 self._GoReconnect(False)
7289 self._WaitUntilSync()
7291 self.feedback_fn("* done")
7293 def _RevertDiskStatus(self):
7294 """Try to revert the disk status after a failed migration.
7297 target_node = self.target_node
7298 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
7302 self._EnsureSecondary(target_node)
7303 self._GoStandalone()
7304 self._GoReconnect(False)
7305 self._WaitUntilSync()
7306 except errors.OpExecError, err:
7307 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
7308 " please try to recover the instance manually;"
7309 " error '%s'" % str(err))
7311 def _AbortMigration(self):
7312 """Call the hypervisor code to abort a started migration.
7315 instance = self.instance
7316 target_node = self.target_node
7317 migration_info = self.migration_info
7319 abort_result = self.rpc.call_finalize_migration(target_node,
7323 abort_msg = abort_result.fail_msg
7325 logging.error("Aborting migration failed on target node %s: %s",
7326 target_node, abort_msg)
7327 # Don't raise an exception here, as we stil have to try to revert the
7328 # disk status, even if this step failed.
7330 def _ExecMigration(self):
7331 """Migrate an instance.
7333 The migrate is done by:
7334 - change the disks into dual-master mode
7335 - wait until disks are fully synchronized again
7336 - migrate the instance
7337 - change disks on the new secondary node (the old primary) to secondary
7338 - wait until disks are fully synchronized
7339 - change disks into single-master mode
7342 instance = self.instance
7343 target_node = self.target_node
7344 source_node = self.source_node
7346 self.feedback_fn("* checking disk consistency between source and target")
7347 for dev in instance.disks:
7348 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
7349 raise errors.OpExecError("Disk %s is degraded or not fully"
7350 " synchronized on target node,"
7351 " aborting migration" % dev.iv_name)
7353 # First get the migration information from the remote node
7354 result = self.rpc.call_migration_info(source_node, instance)
7355 msg = result.fail_msg
7357 log_err = ("Failed fetching source migration information from %s: %s" %
7359 logging.error(log_err)
7360 raise errors.OpExecError(log_err)
7362 self.migration_info = migration_info = result.payload
7364 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
7365 # Then switch the disks to master/master mode
7366 self._EnsureSecondary(target_node)
7367 self._GoStandalone()
7368 self._GoReconnect(True)
7369 self._WaitUntilSync()
7371 self.feedback_fn("* preparing %s to accept the instance" % target_node)
7372 result = self.rpc.call_accept_instance(target_node,
7375 self.nodes_ip[target_node])
7377 msg = result.fail_msg
7379 logging.error("Instance pre-migration failed, trying to revert"
7380 " disk status: %s", msg)
7381 self.feedback_fn("Pre-migration failed, aborting")
7382 self._AbortMigration()
7383 self._RevertDiskStatus()
7384 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
7385 (instance.name, msg))
7387 self.feedback_fn("* migrating instance to %s" % target_node)
7388 result = self.rpc.call_instance_migrate(source_node, instance,
7389 self.nodes_ip[target_node],
7391 msg = result.fail_msg
7393 logging.error("Instance migration failed, trying to revert"
7394 " disk status: %s", msg)
7395 self.feedback_fn("Migration failed, aborting")
7396 self._AbortMigration()
7397 self._RevertDiskStatus()
7398 raise errors.OpExecError("Could not migrate instance %s: %s" %
7399 (instance.name, msg))
7401 instance.primary_node = target_node
7402 # distribute new instance config to the other nodes
7403 self.cfg.Update(instance, self.feedback_fn)
7405 result = self.rpc.call_finalize_migration(target_node,
7409 msg = result.fail_msg
7411 logging.error("Instance migration succeeded, but finalization failed:"
7413 raise errors.OpExecError("Could not finalize instance migration: %s" %
7416 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
7417 self._EnsureSecondary(source_node)
7418 self._WaitUntilSync()
7419 self._GoStandalone()
7420 self._GoReconnect(False)
7421 self._WaitUntilSync()
7423 self.feedback_fn("* done")
7425 def _ExecFailover(self):
7426 """Failover an instance.
7428 The failover is done by shutting it down on its present node and
7429 starting it on the secondary.
7432 instance = self.instance
7433 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
7435 source_node = instance.primary_node
7436 target_node = self.target_node
7438 if instance.admin_up:
7439 self.feedback_fn("* checking disk consistency between source and target")
7440 for dev in instance.disks:
7441 # for drbd, these are drbd over lvm
7442 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
7443 if primary_node.offline:
7444 self.feedback_fn("Node %s is offline, ignoring degraded disk %s on"
7446 (primary_node.name, dev.iv_name, target_node))
7447 elif not self.ignore_consistency:
7448 raise errors.OpExecError("Disk %s is degraded on target node,"
7449 " aborting failover" % dev.iv_name)
7451 self.feedback_fn("* not checking disk consistency as instance is not"
7454 self.feedback_fn("* shutting down instance on source node")
7455 logging.info("Shutting down instance %s on node %s",
7456 instance.name, source_node)
7458 result = self.rpc.call_instance_shutdown(source_node, instance,
7459 self.shutdown_timeout)
7460 msg = result.fail_msg
7462 if self.ignore_consistency or primary_node.offline:
7463 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
7464 " proceeding anyway; please make sure node"
7465 " %s is down; error details: %s",
7466 instance.name, source_node, source_node, msg)
7468 raise errors.OpExecError("Could not shutdown instance %s on"
7470 (instance.name, source_node, msg))
7472 self.feedback_fn("* deactivating the instance's disks on source node")
7473 if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True):
7474 raise errors.OpExecError("Can't shut down the instance's disks")
7476 instance.primary_node = target_node
7477 # distribute new instance config to the other nodes
7478 self.cfg.Update(instance, self.feedback_fn)
7480 # Only start the instance if it's marked as up
7481 if instance.admin_up:
7482 self.feedback_fn("* activating the instance's disks on target node %s" %
7484 logging.info("Starting instance %s on node %s",
7485 instance.name, target_node)
7487 disks_ok, _ = _AssembleInstanceDisks(self.lu, instance,
7488 ignore_secondaries=True)
7490 _ShutdownInstanceDisks(self.lu, instance)
7491 raise errors.OpExecError("Can't activate the instance's disks")
7493 self.feedback_fn("* starting the instance on the target node %s" %
7495 result = self.rpc.call_instance_start(target_node, instance, None, None,
7497 msg = result.fail_msg
7499 _ShutdownInstanceDisks(self.lu, instance)
7500 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
7501 (instance.name, target_node, msg))
7503 def Exec(self, feedback_fn):
7504 """Perform the migration.
7507 self.feedback_fn = feedback_fn
7508 self.source_node = self.instance.primary_node
7510 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
7511 if self.instance.disk_template in constants.DTS_INT_MIRROR:
7512 self.target_node = self.instance.secondary_nodes[0]
7513 # Otherwise self.target_node has been populated either
7514 # directly, or through an iallocator.
7516 self.all_nodes = [self.source_node, self.target_node]
7517 self.nodes_ip = dict((name, node.secondary_ip) for (name, node)
7518 in self.cfg.GetMultiNodeInfo(self.all_nodes))
7521 feedback_fn("Failover instance %s" % self.instance.name)
7522 self._ExecFailover()
7524 feedback_fn("Migrating instance %s" % self.instance.name)
7527 return self._ExecCleanup()
7529 return self._ExecMigration()
7532 def _CreateBlockDev(lu, node, instance, device, force_create,
7534 """Create a tree of block devices on a given node.
7536 If this device type has to be created on secondaries, create it and
7539 If not, just recurse to children keeping the same 'force' value.
7541 @param lu: the lu on whose behalf we execute
7542 @param node: the node on which to create the device
7543 @type instance: L{objects.Instance}
7544 @param instance: the instance which owns the device
7545 @type device: L{objects.Disk}
7546 @param device: the device to create
7547 @type force_create: boolean
7548 @param force_create: whether to force creation of this device; this
7549 will be change to True whenever we find a device which has
7550 CreateOnSecondary() attribute
7551 @param info: the extra 'metadata' we should attach to the device
7552 (this will be represented as a LVM tag)
7553 @type force_open: boolean
7554 @param force_open: this parameter will be passes to the
7555 L{backend.BlockdevCreate} function where it specifies
7556 whether we run on primary or not, and it affects both
7557 the child assembly and the device own Open() execution
7560 if device.CreateOnSecondary():
7564 for child in device.children:
7565 _CreateBlockDev(lu, node, instance, child, force_create,
7568 if not force_create:
7571 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
7574 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
7575 """Create a single block device on a given node.
7577 This will not recurse over children of the device, so they must be
7580 @param lu: the lu on whose behalf we execute
7581 @param node: the node on which to create the device
7582 @type instance: L{objects.Instance}
7583 @param instance: the instance which owns the device
7584 @type device: L{objects.Disk}
7585 @param device: the device to create
7586 @param info: the extra 'metadata' we should attach to the device
7587 (this will be represented as a LVM tag)
7588 @type force_open: boolean
7589 @param force_open: this parameter will be passes to the
7590 L{backend.BlockdevCreate} function where it specifies
7591 whether we run on primary or not, and it affects both
7592 the child assembly and the device own Open() execution
7595 lu.cfg.SetDiskID(device, node)
7596 result = lu.rpc.call_blockdev_create(node, device, device.size,
7597 instance.name, force_open, info)
7598 result.Raise("Can't create block device %s on"
7599 " node %s for instance %s" % (device, node, instance.name))
7600 if device.physical_id is None:
7601 device.physical_id = result.payload
7604 def _GenerateUniqueNames(lu, exts):
7605 """Generate a suitable LV name.
7607 This will generate a logical volume name for the given instance.
7612 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
7613 results.append("%s%s" % (new_id, val))
7617 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
7618 iv_name, p_minor, s_minor):
7619 """Generate a drbd8 device complete with its children.
7622 assert len(vgnames) == len(names) == 2
7623 port = lu.cfg.AllocatePort()
7624 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
7625 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
7626 logical_id=(vgnames[0], names[0]))
7627 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7628 logical_id=(vgnames[1], names[1]))
7629 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
7630 logical_id=(primary, secondary, port,
7633 children=[dev_data, dev_meta],
7638 def _GenerateDiskTemplate(lu, template_name,
7639 instance_name, primary_node,
7640 secondary_nodes, disk_info,
7641 file_storage_dir, file_driver,
7642 base_index, feedback_fn):
7643 """Generate the entire disk layout for a given template type.
7646 #TODO: compute space requirements
7648 vgname = lu.cfg.GetVGName()
7649 disk_count = len(disk_info)
7651 if template_name == constants.DT_DISKLESS:
7653 elif template_name == constants.DT_PLAIN:
7654 if len(secondary_nodes) != 0:
7655 raise errors.ProgrammerError("Wrong template configuration")
7657 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7658 for i in range(disk_count)])
7659 for idx, disk in enumerate(disk_info):
7660 disk_index = idx + base_index
7661 vg = disk.get(constants.IDISK_VG, vgname)
7662 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
7663 disk_dev = objects.Disk(dev_type=constants.LD_LV,
7664 size=disk[constants.IDISK_SIZE],
7665 logical_id=(vg, names[idx]),
7666 iv_name="disk/%d" % disk_index,
7667 mode=disk[constants.IDISK_MODE])
7668 disks.append(disk_dev)
7669 elif template_name == constants.DT_DRBD8:
7670 if len(secondary_nodes) != 1:
7671 raise errors.ProgrammerError("Wrong template configuration")
7672 remote_node = secondary_nodes[0]
7673 minors = lu.cfg.AllocateDRBDMinor(
7674 [primary_node, remote_node] * len(disk_info), instance_name)
7677 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7678 for i in range(disk_count)]):
7679 names.append(lv_prefix + "_data")
7680 names.append(lv_prefix + "_meta")
7681 for idx, disk in enumerate(disk_info):
7682 disk_index = idx + base_index
7683 data_vg = disk.get(constants.IDISK_VG, vgname)
7684 meta_vg = disk.get(constants.IDISK_METAVG, data_vg)
7685 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
7686 disk[constants.IDISK_SIZE],
7688 names[idx * 2:idx * 2 + 2],
7689 "disk/%d" % disk_index,
7690 minors[idx * 2], minors[idx * 2 + 1])
7691 disk_dev.mode = disk[constants.IDISK_MODE]
7692 disks.append(disk_dev)
7693 elif template_name == constants.DT_FILE:
7694 if len(secondary_nodes) != 0:
7695 raise errors.ProgrammerError("Wrong template configuration")
7697 opcodes.RequireFileStorage()
7699 for idx, disk in enumerate(disk_info):
7700 disk_index = idx + base_index
7701 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7702 size=disk[constants.IDISK_SIZE],
7703 iv_name="disk/%d" % disk_index,
7704 logical_id=(file_driver,
7705 "%s/disk%d" % (file_storage_dir,
7707 mode=disk[constants.IDISK_MODE])
7708 disks.append(disk_dev)
7709 elif template_name == constants.DT_SHARED_FILE:
7710 if len(secondary_nodes) != 0:
7711 raise errors.ProgrammerError("Wrong template configuration")
7713 opcodes.RequireSharedFileStorage()
7715 for idx, disk in enumerate(disk_info):
7716 disk_index = idx + base_index
7717 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7718 size=disk[constants.IDISK_SIZE],
7719 iv_name="disk/%d" % disk_index,
7720 logical_id=(file_driver,
7721 "%s/disk%d" % (file_storage_dir,
7723 mode=disk[constants.IDISK_MODE])
7724 disks.append(disk_dev)
7725 elif template_name == constants.DT_BLOCK:
7726 if len(secondary_nodes) != 0:
7727 raise errors.ProgrammerError("Wrong template configuration")
7729 for idx, disk in enumerate(disk_info):
7730 disk_index = idx + base_index
7731 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV,
7732 size=disk[constants.IDISK_SIZE],
7733 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL,
7734 disk[constants.IDISK_ADOPT]),
7735 iv_name="disk/%d" % disk_index,
7736 mode=disk[constants.IDISK_MODE])
7737 disks.append(disk_dev)
7740 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
7744 def _GetInstanceInfoText(instance):
7745 """Compute that text that should be added to the disk's metadata.
7748 return "originstname+%s" % instance.name
7751 def _CalcEta(time_taken, written, total_size):
7752 """Calculates the ETA based on size written and total size.
7754 @param time_taken: The time taken so far
7755 @param written: amount written so far
7756 @param total_size: The total size of data to be written
7757 @return: The remaining time in seconds
7760 avg_time = time_taken / float(written)
7761 return (total_size - written) * avg_time
7764 def _WipeDisks(lu, instance):
7765 """Wipes instance disks.
7767 @type lu: L{LogicalUnit}
7768 @param lu: the logical unit on whose behalf we execute
7769 @type instance: L{objects.Instance}
7770 @param instance: the instance whose disks we should create
7771 @return: the success of the wipe
7774 node = instance.primary_node
7776 for device in instance.disks:
7777 lu.cfg.SetDiskID(device, node)
7779 logging.info("Pause sync of instance %s disks", instance.name)
7780 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
7782 for idx, success in enumerate(result.payload):
7784 logging.warn("pause-sync of instance %s for disks %d failed",
7788 for idx, device in enumerate(instance.disks):
7789 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
7790 # MAX_WIPE_CHUNK at max
7791 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
7792 constants.MIN_WIPE_CHUNK_PERCENT)
7793 # we _must_ make this an int, otherwise rounding errors will
7795 wipe_chunk_size = int(wipe_chunk_size)
7797 lu.LogInfo("* Wiping disk %d", idx)
7798 logging.info("Wiping disk %d for instance %s, node %s using"
7799 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
7804 start_time = time.time()
7806 while offset < size:
7807 wipe_size = min(wipe_chunk_size, size - offset)
7808 logging.debug("Wiping disk %d, offset %s, chunk %s",
7809 idx, offset, wipe_size)
7810 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
7811 result.Raise("Could not wipe disk %d at offset %d for size %d" %
7812 (idx, offset, wipe_size))
7815 if now - last_output >= 60:
7816 eta = _CalcEta(now - start_time, offset, size)
7817 lu.LogInfo(" - done: %.1f%% ETA: %s" %
7818 (offset / float(size) * 100, utils.FormatSeconds(eta)))
7821 logging.info("Resume sync of instance %s disks", instance.name)
7823 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
7825 for idx, success in enumerate(result.payload):
7827 lu.LogWarning("Resume sync of disk %d failed, please have a"
7828 " look at the status and troubleshoot the issue", idx)
7829 logging.warn("resume-sync of instance %s for disks %d failed",
7833 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
7834 """Create all disks for an instance.
7836 This abstracts away some work from AddInstance.
7838 @type lu: L{LogicalUnit}
7839 @param lu: the logical unit on whose behalf we execute
7840 @type instance: L{objects.Instance}
7841 @param instance: the instance whose disks we should create
7843 @param to_skip: list of indices to skip
7844 @type target_node: string
7845 @param target_node: if passed, overrides the target node for creation
7847 @return: the success of the creation
7850 info = _GetInstanceInfoText(instance)
7851 if target_node is None:
7852 pnode = instance.primary_node
7853 all_nodes = instance.all_nodes
7858 if instance.disk_template in constants.DTS_FILEBASED:
7859 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7860 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
7862 result.Raise("Failed to create directory '%s' on"
7863 " node %s" % (file_storage_dir, pnode))
7865 # Note: this needs to be kept in sync with adding of disks in
7866 # LUInstanceSetParams
7867 for idx, device in enumerate(instance.disks):
7868 if to_skip and idx in to_skip:
7870 logging.info("Creating volume %s for instance %s",
7871 device.iv_name, instance.name)
7873 for node in all_nodes:
7874 f_create = node == pnode
7875 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
7878 def _RemoveDisks(lu, instance, target_node=None):
7879 """Remove all disks for an instance.
7881 This abstracts away some work from `AddInstance()` and
7882 `RemoveInstance()`. Note that in case some of the devices couldn't
7883 be removed, the removal will continue with the other ones (compare
7884 with `_CreateDisks()`).
7886 @type lu: L{LogicalUnit}
7887 @param lu: the logical unit on whose behalf we execute
7888 @type instance: L{objects.Instance}
7889 @param instance: the instance whose disks we should remove
7890 @type target_node: string
7891 @param target_node: used to override the node on which to remove the disks
7893 @return: the success of the removal
7896 logging.info("Removing block devices for instance %s", instance.name)
7899 for device in instance.disks:
7901 edata = [(target_node, device)]
7903 edata = device.ComputeNodeTree(instance.primary_node)
7904 for node, disk in edata:
7905 lu.cfg.SetDiskID(disk, node)
7906 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
7908 lu.LogWarning("Could not remove block device %s on node %s,"
7909 " continuing anyway: %s", device.iv_name, node, msg)
7912 if instance.disk_template == constants.DT_FILE:
7913 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7917 tgt = instance.primary_node
7918 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
7920 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
7921 file_storage_dir, instance.primary_node, result.fail_msg)
7927 def _ComputeDiskSizePerVG(disk_template, disks):
7928 """Compute disk size requirements in the volume group
7931 def _compute(disks, payload):
7932 """Universal algorithm.
7937 vgs[disk[constants.IDISK_VG]] = \
7938 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
7942 # Required free disk space as a function of disk and swap space
7944 constants.DT_DISKLESS: {},
7945 constants.DT_PLAIN: _compute(disks, 0),
7946 # 128 MB are added for drbd metadata for each disk
7947 constants.DT_DRBD8: _compute(disks, 128),
7948 constants.DT_FILE: {},
7949 constants.DT_SHARED_FILE: {},
7952 if disk_template not in req_size_dict:
7953 raise errors.ProgrammerError("Disk template '%s' size requirement"
7954 " is unknown" % disk_template)
7956 return req_size_dict[disk_template]
7959 def _ComputeDiskSize(disk_template, disks):
7960 """Compute disk size requirements in the volume group
7963 # Required free disk space as a function of disk and swap space
7965 constants.DT_DISKLESS: None,
7966 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
7967 # 128 MB are added for drbd metadata for each disk
7968 constants.DT_DRBD8: sum(d[constants.IDISK_SIZE] + 128 for d in disks),
7969 constants.DT_FILE: None,
7970 constants.DT_SHARED_FILE: 0,
7971 constants.DT_BLOCK: 0,
7974 if disk_template not in req_size_dict:
7975 raise errors.ProgrammerError("Disk template '%s' size requirement"
7976 " is unknown" % disk_template)
7978 return req_size_dict[disk_template]
7981 def _FilterVmNodes(lu, nodenames):
7982 """Filters out non-vm_capable nodes from a list.
7984 @type lu: L{LogicalUnit}
7985 @param lu: the logical unit for which we check
7986 @type nodenames: list
7987 @param nodenames: the list of nodes on which we should check
7989 @return: the list of vm-capable nodes
7992 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
7993 return [name for name in nodenames if name not in vm_nodes]
7996 def _CheckHVParams(lu, nodenames, hvname, hvparams):
7997 """Hypervisor parameter validation.
7999 This function abstract the hypervisor parameter validation to be
8000 used in both instance create and instance modify.
8002 @type lu: L{LogicalUnit}
8003 @param lu: the logical unit for which we check
8004 @type nodenames: list
8005 @param nodenames: the list of nodes on which we should check
8006 @type hvname: string
8007 @param hvname: the name of the hypervisor we should use
8008 @type hvparams: dict
8009 @param hvparams: the parameters which we need to check
8010 @raise errors.OpPrereqError: if the parameters are not valid
8013 nodenames = _FilterVmNodes(lu, nodenames)
8014 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
8017 for node in nodenames:
8021 info.Raise("Hypervisor parameter validation failed on node %s" % node)
8024 def _CheckOSParams(lu, required, nodenames, osname, osparams):
8025 """OS parameters validation.
8027 @type lu: L{LogicalUnit}
8028 @param lu: the logical unit for which we check
8029 @type required: boolean
8030 @param required: whether the validation should fail if the OS is not
8032 @type nodenames: list
8033 @param nodenames: the list of nodes on which we should check
8034 @type osname: string
8035 @param osname: the name of the hypervisor we should use
8036 @type osparams: dict
8037 @param osparams: the parameters which we need to check
8038 @raise errors.OpPrereqError: if the parameters are not valid
8041 nodenames = _FilterVmNodes(lu, nodenames)
8042 result = lu.rpc.call_os_validate(required, nodenames, osname,
8043 [constants.OS_VALIDATE_PARAMETERS],
8045 for node, nres in result.items():
8046 # we don't check for offline cases since this should be run only
8047 # against the master node and/or an instance's nodes
8048 nres.Raise("OS Parameters validation failed on node %s" % node)
8049 if not nres.payload:
8050 lu.LogInfo("OS %s not found on node %s, validation skipped",
8054 class LUInstanceCreate(LogicalUnit):
8055 """Create an instance.
8058 HPATH = "instance-add"
8059 HTYPE = constants.HTYPE_INSTANCE
8062 def CheckArguments(self):
8066 # do not require name_check to ease forward/backward compatibility
8068 if self.op.no_install and self.op.start:
8069 self.LogInfo("No-installation mode selected, disabling startup")
8070 self.op.start = False
8071 # validate/normalize the instance name
8072 self.op.instance_name = \
8073 netutils.Hostname.GetNormalizedName(self.op.instance_name)
8075 if self.op.ip_check and not self.op.name_check:
8076 # TODO: make the ip check more flexible and not depend on the name check
8077 raise errors.OpPrereqError("Cannot do IP address check without a name"
8078 " check", errors.ECODE_INVAL)
8080 # check nics' parameter names
8081 for nic in self.op.nics:
8082 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
8084 # check disks. parameter names and consistent adopt/no-adopt strategy
8085 has_adopt = has_no_adopt = False
8086 for disk in self.op.disks:
8087 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
8088 if constants.IDISK_ADOPT in disk:
8092 if has_adopt and has_no_adopt:
8093 raise errors.OpPrereqError("Either all disks are adopted or none is",
8096 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
8097 raise errors.OpPrereqError("Disk adoption is not supported for the"
8098 " '%s' disk template" %
8099 self.op.disk_template,
8101 if self.op.iallocator is not None:
8102 raise errors.OpPrereqError("Disk adoption not allowed with an"
8103 " iallocator script", errors.ECODE_INVAL)
8104 if self.op.mode == constants.INSTANCE_IMPORT:
8105 raise errors.OpPrereqError("Disk adoption not allowed for"
8106 " instance import", errors.ECODE_INVAL)
8108 if self.op.disk_template in constants.DTS_MUST_ADOPT:
8109 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
8110 " but no 'adopt' parameter given" %
8111 self.op.disk_template,
8114 self.adopt_disks = has_adopt
8116 # instance name verification
8117 if self.op.name_check:
8118 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
8119 self.op.instance_name = self.hostname1.name
8120 # used in CheckPrereq for ip ping check
8121 self.check_ip = self.hostname1.ip
8123 self.check_ip = None
8125 # file storage checks
8126 if (self.op.file_driver and
8127 not self.op.file_driver in constants.FILE_DRIVER):
8128 raise errors.OpPrereqError("Invalid file driver name '%s'" %
8129 self.op.file_driver, errors.ECODE_INVAL)
8131 if self.op.disk_template == constants.DT_FILE:
8132 opcodes.RequireFileStorage()
8133 elif self.op.disk_template == constants.DT_SHARED_FILE:
8134 opcodes.RequireSharedFileStorage()
8136 ### Node/iallocator related checks
8137 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
8139 if self.op.pnode is not None:
8140 if self.op.disk_template in constants.DTS_INT_MIRROR:
8141 if self.op.snode is None:
8142 raise errors.OpPrereqError("The networked disk templates need"
8143 " a mirror node", errors.ECODE_INVAL)
8145 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
8147 self.op.snode = None
8149 self._cds = _GetClusterDomainSecret()
8151 if self.op.mode == constants.INSTANCE_IMPORT:
8152 # On import force_variant must be True, because if we forced it at
8153 # initial install, our only chance when importing it back is that it
8155 self.op.force_variant = True
8157 if self.op.no_install:
8158 self.LogInfo("No-installation mode has no effect during import")
8160 elif self.op.mode == constants.INSTANCE_CREATE:
8161 if self.op.os_type is None:
8162 raise errors.OpPrereqError("No guest OS specified",
8164 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
8165 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
8166 " installation" % self.op.os_type,
8168 if self.op.disk_template is None:
8169 raise errors.OpPrereqError("No disk template specified",
8172 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
8173 # Check handshake to ensure both clusters have the same domain secret
8174 src_handshake = self.op.source_handshake
8175 if not src_handshake:
8176 raise errors.OpPrereqError("Missing source handshake",
8179 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
8182 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
8185 # Load and check source CA
8186 self.source_x509_ca_pem = self.op.source_x509_ca
8187 if not self.source_x509_ca_pem:
8188 raise errors.OpPrereqError("Missing source X509 CA",
8192 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
8194 except OpenSSL.crypto.Error, err:
8195 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
8196 (err, ), errors.ECODE_INVAL)
8198 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
8199 if errcode is not None:
8200 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
8203 self.source_x509_ca = cert
8205 src_instance_name = self.op.source_instance_name
8206 if not src_instance_name:
8207 raise errors.OpPrereqError("Missing source instance name",
8210 self.source_instance_name = \
8211 netutils.GetHostname(name=src_instance_name).name
8214 raise errors.OpPrereqError("Invalid instance creation mode %r" %
8215 self.op.mode, errors.ECODE_INVAL)
8217 def ExpandNames(self):
8218 """ExpandNames for CreateInstance.
8220 Figure out the right locks for instance creation.
8223 self.needed_locks = {}
8225 instance_name = self.op.instance_name
8226 # this is just a preventive check, but someone might still add this
8227 # instance in the meantime, and creation will fail at lock-add time
8228 if instance_name in self.cfg.GetInstanceList():
8229 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
8230 instance_name, errors.ECODE_EXISTS)
8232 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
8234 if self.op.iallocator:
8235 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8237 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
8238 nodelist = [self.op.pnode]
8239 if self.op.snode is not None:
8240 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
8241 nodelist.append(self.op.snode)
8242 self.needed_locks[locking.LEVEL_NODE] = nodelist
8244 # in case of import lock the source node too
8245 if self.op.mode == constants.INSTANCE_IMPORT:
8246 src_node = self.op.src_node
8247 src_path = self.op.src_path
8249 if src_path is None:
8250 self.op.src_path = src_path = self.op.instance_name
8252 if src_node is None:
8253 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8254 self.op.src_node = None
8255 if os.path.isabs(src_path):
8256 raise errors.OpPrereqError("Importing an instance from a path"
8257 " requires a source node option",
8260 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
8261 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
8262 self.needed_locks[locking.LEVEL_NODE].append(src_node)
8263 if not os.path.isabs(src_path):
8264 self.op.src_path = src_path = \
8265 utils.PathJoin(constants.EXPORT_DIR, src_path)
8267 def _RunAllocator(self):
8268 """Run the allocator based on input opcode.
8271 nics = [n.ToDict() for n in self.nics]
8272 ial = IAllocator(self.cfg, self.rpc,
8273 mode=constants.IALLOCATOR_MODE_ALLOC,
8274 name=self.op.instance_name,
8275 disk_template=self.op.disk_template,
8278 vcpus=self.be_full[constants.BE_VCPUS],
8279 memory=self.be_full[constants.BE_MEMORY],
8282 hypervisor=self.op.hypervisor,
8285 ial.Run(self.op.iallocator)
8288 raise errors.OpPrereqError("Can't compute nodes using"
8289 " iallocator '%s': %s" %
8290 (self.op.iallocator, ial.info),
8292 if len(ial.result) != ial.required_nodes:
8293 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8294 " of nodes (%s), required %s" %
8295 (self.op.iallocator, len(ial.result),
8296 ial.required_nodes), errors.ECODE_FAULT)
8297 self.op.pnode = ial.result[0]
8298 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
8299 self.op.instance_name, self.op.iallocator,
8300 utils.CommaJoin(ial.result))
8301 if ial.required_nodes == 2:
8302 self.op.snode = ial.result[1]
8304 def BuildHooksEnv(self):
8307 This runs on master, primary and secondary nodes of the instance.
8311 "ADD_MODE": self.op.mode,
8313 if self.op.mode == constants.INSTANCE_IMPORT:
8314 env["SRC_NODE"] = self.op.src_node
8315 env["SRC_PATH"] = self.op.src_path
8316 env["SRC_IMAGES"] = self.src_images
8318 env.update(_BuildInstanceHookEnv(
8319 name=self.op.instance_name,
8320 primary_node=self.op.pnode,
8321 secondary_nodes=self.secondaries,
8322 status=self.op.start,
8323 os_type=self.op.os_type,
8324 memory=self.be_full[constants.BE_MEMORY],
8325 vcpus=self.be_full[constants.BE_VCPUS],
8326 nics=_NICListToTuple(self, self.nics),
8327 disk_template=self.op.disk_template,
8328 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
8329 for d in self.disks],
8332 hypervisor_name=self.op.hypervisor,
8338 def BuildHooksNodes(self):
8339 """Build hooks nodes.
8342 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
8345 def _ReadExportInfo(self):
8346 """Reads the export information from disk.
8348 It will override the opcode source node and path with the actual
8349 information, if these two were not specified before.
8351 @return: the export information
8354 assert self.op.mode == constants.INSTANCE_IMPORT
8356 src_node = self.op.src_node
8357 src_path = self.op.src_path
8359 if src_node is None:
8360 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
8361 exp_list = self.rpc.call_export_list(locked_nodes)
8363 for node in exp_list:
8364 if exp_list[node].fail_msg:
8366 if src_path in exp_list[node].payload:
8368 self.op.src_node = src_node = node
8369 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
8373 raise errors.OpPrereqError("No export found for relative path %s" %
8374 src_path, errors.ECODE_INVAL)
8376 _CheckNodeOnline(self, src_node)
8377 result = self.rpc.call_export_info(src_node, src_path)
8378 result.Raise("No export or invalid export found in dir %s" % src_path)
8380 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
8381 if not export_info.has_section(constants.INISECT_EXP):
8382 raise errors.ProgrammerError("Corrupted export config",
8383 errors.ECODE_ENVIRON)
8385 ei_version = export_info.get(constants.INISECT_EXP, "version")
8386 if (int(ei_version) != constants.EXPORT_VERSION):
8387 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
8388 (ei_version, constants.EXPORT_VERSION),
8389 errors.ECODE_ENVIRON)
8392 def _ReadExportParams(self, einfo):
8393 """Use export parameters as defaults.
8395 In case the opcode doesn't specify (as in override) some instance
8396 parameters, then try to use them from the export information, if
8400 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
8402 if self.op.disk_template is None:
8403 if einfo.has_option(constants.INISECT_INS, "disk_template"):
8404 self.op.disk_template = einfo.get(constants.INISECT_INS,
8407 raise errors.OpPrereqError("No disk template specified and the export"
8408 " is missing the disk_template information",
8411 if not self.op.disks:
8412 if einfo.has_option(constants.INISECT_INS, "disk_count"):
8414 # TODO: import the disk iv_name too
8415 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
8416 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
8417 disks.append({constants.IDISK_SIZE: disk_sz})
8418 self.op.disks = disks
8420 raise errors.OpPrereqError("No disk info specified and the export"
8421 " is missing the disk information",
8424 if (not self.op.nics and
8425 einfo.has_option(constants.INISECT_INS, "nic_count")):
8427 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
8429 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
8430 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
8435 if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):
8436 self.op.tags = einfo.get(constants.INISECT_INS, "tags").split()
8438 if (self.op.hypervisor is None and
8439 einfo.has_option(constants.INISECT_INS, "hypervisor")):
8440 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
8442 if einfo.has_section(constants.INISECT_HYP):
8443 # use the export parameters but do not override the ones
8444 # specified by the user
8445 for name, value in einfo.items(constants.INISECT_HYP):
8446 if name not in self.op.hvparams:
8447 self.op.hvparams[name] = value
8449 if einfo.has_section(constants.INISECT_BEP):
8450 # use the parameters, without overriding
8451 for name, value in einfo.items(constants.INISECT_BEP):
8452 if name not in self.op.beparams:
8453 self.op.beparams[name] = value
8455 # try to read the parameters old style, from the main section
8456 for name in constants.BES_PARAMETERS:
8457 if (name not in self.op.beparams and
8458 einfo.has_option(constants.INISECT_INS, name)):
8459 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
8461 if einfo.has_section(constants.INISECT_OSP):
8462 # use the parameters, without overriding
8463 for name, value in einfo.items(constants.INISECT_OSP):
8464 if name not in self.op.osparams:
8465 self.op.osparams[name] = value
8467 def _RevertToDefaults(self, cluster):
8468 """Revert the instance parameters to the default values.
8472 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
8473 for name in self.op.hvparams.keys():
8474 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
8475 del self.op.hvparams[name]
8477 be_defs = cluster.SimpleFillBE({})
8478 for name in self.op.beparams.keys():
8479 if name in be_defs and be_defs[name] == self.op.beparams[name]:
8480 del self.op.beparams[name]
8482 nic_defs = cluster.SimpleFillNIC({})
8483 for nic in self.op.nics:
8484 for name in constants.NICS_PARAMETERS:
8485 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
8488 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
8489 for name in self.op.osparams.keys():
8490 if name in os_defs and os_defs[name] == self.op.osparams[name]:
8491 del self.op.osparams[name]
8493 def _CalculateFileStorageDir(self):
8494 """Calculate final instance file storage dir.
8497 # file storage dir calculation/check
8498 self.instance_file_storage_dir = None
8499 if self.op.disk_template in constants.DTS_FILEBASED:
8500 # build the full file storage dir path
8503 if self.op.disk_template == constants.DT_SHARED_FILE:
8504 get_fsd_fn = self.cfg.GetSharedFileStorageDir
8506 get_fsd_fn = self.cfg.GetFileStorageDir
8508 cfg_storagedir = get_fsd_fn()
8509 if not cfg_storagedir:
8510 raise errors.OpPrereqError("Cluster file storage dir not defined")
8511 joinargs.append(cfg_storagedir)
8513 if self.op.file_storage_dir is not None:
8514 joinargs.append(self.op.file_storage_dir)
8516 joinargs.append(self.op.instance_name)
8518 # pylint: disable=W0142
8519 self.instance_file_storage_dir = utils.PathJoin(*joinargs)
8521 def CheckPrereq(self):
8522 """Check prerequisites.
8525 self._CalculateFileStorageDir()
8527 if self.op.mode == constants.INSTANCE_IMPORT:
8528 export_info = self._ReadExportInfo()
8529 self._ReadExportParams(export_info)
8531 if (not self.cfg.GetVGName() and
8532 self.op.disk_template not in constants.DTS_NOT_LVM):
8533 raise errors.OpPrereqError("Cluster does not support lvm-based"
8534 " instances", errors.ECODE_STATE)
8536 if self.op.hypervisor is None:
8537 self.op.hypervisor = self.cfg.GetHypervisorType()
8539 cluster = self.cfg.GetClusterInfo()
8540 enabled_hvs = cluster.enabled_hypervisors
8541 if self.op.hypervisor not in enabled_hvs:
8542 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
8543 " cluster (%s)" % (self.op.hypervisor,
8544 ",".join(enabled_hvs)),
8547 # Check tag validity
8548 for tag in self.op.tags:
8549 objects.TaggableObject.ValidateTag(tag)
8551 # check hypervisor parameter syntax (locally)
8552 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
8553 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
8555 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
8556 hv_type.CheckParameterSyntax(filled_hvp)
8557 self.hv_full = filled_hvp
8558 # check that we don't specify global parameters on an instance
8559 _CheckGlobalHvParams(self.op.hvparams)
8561 # fill and remember the beparams dict
8562 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
8563 self.be_full = cluster.SimpleFillBE(self.op.beparams)
8565 # build os parameters
8566 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
8568 # now that hvp/bep are in final format, let's reset to defaults,
8570 if self.op.identify_defaults:
8571 self._RevertToDefaults(cluster)
8575 for idx, nic in enumerate(self.op.nics):
8576 nic_mode_req = nic.get(constants.INIC_MODE, None)
8577 nic_mode = nic_mode_req
8578 if nic_mode is None:
8579 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
8581 # in routed mode, for the first nic, the default ip is 'auto'
8582 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
8583 default_ip_mode = constants.VALUE_AUTO
8585 default_ip_mode = constants.VALUE_NONE
8587 # ip validity checks
8588 ip = nic.get(constants.INIC_IP, default_ip_mode)
8589 if ip is None or ip.lower() == constants.VALUE_NONE:
8591 elif ip.lower() == constants.VALUE_AUTO:
8592 if not self.op.name_check:
8593 raise errors.OpPrereqError("IP address set to auto but name checks"
8594 " have been skipped",
8596 nic_ip = self.hostname1.ip
8598 if not netutils.IPAddress.IsValid(ip):
8599 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
8603 # TODO: check the ip address for uniqueness
8604 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
8605 raise errors.OpPrereqError("Routed nic mode requires an ip address",
8608 # MAC address verification
8609 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
8610 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8611 mac = utils.NormalizeAndValidateMac(mac)
8614 self.cfg.ReserveMAC(mac, self.proc.GetECId())
8615 except errors.ReservationError:
8616 raise errors.OpPrereqError("MAC address %s already in use"
8617 " in cluster" % mac,
8618 errors.ECODE_NOTUNIQUE)
8620 # Build nic parameters
8621 link = nic.get(constants.INIC_LINK, None)
8624 nicparams[constants.NIC_MODE] = nic_mode_req
8626 nicparams[constants.NIC_LINK] = link
8628 check_params = cluster.SimpleFillNIC(nicparams)
8629 objects.NIC.CheckParameterSyntax(check_params)
8630 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
8632 # disk checks/pre-build
8633 default_vg = self.cfg.GetVGName()
8635 for disk in self.op.disks:
8636 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
8637 if mode not in constants.DISK_ACCESS_SET:
8638 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
8639 mode, errors.ECODE_INVAL)
8640 size = disk.get(constants.IDISK_SIZE, None)
8642 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
8645 except (TypeError, ValueError):
8646 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
8649 data_vg = disk.get(constants.IDISK_VG, default_vg)
8651 constants.IDISK_SIZE: size,
8652 constants.IDISK_MODE: mode,
8653 constants.IDISK_VG: data_vg,
8654 constants.IDISK_METAVG: disk.get(constants.IDISK_METAVG, data_vg),
8656 if constants.IDISK_ADOPT in disk:
8657 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
8658 self.disks.append(new_disk)
8660 if self.op.mode == constants.INSTANCE_IMPORT:
8662 # Check that the new instance doesn't have less disks than the export
8663 instance_disks = len(self.disks)
8664 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
8665 if instance_disks < export_disks:
8666 raise errors.OpPrereqError("Not enough disks to import."
8667 " (instance: %d, export: %d)" %
8668 (instance_disks, export_disks),
8672 for idx in range(export_disks):
8673 option = "disk%d_dump" % idx
8674 if export_info.has_option(constants.INISECT_INS, option):
8675 # FIXME: are the old os-es, disk sizes, etc. useful?
8676 export_name = export_info.get(constants.INISECT_INS, option)
8677 image = utils.PathJoin(self.op.src_path, export_name)
8678 disk_images.append(image)
8680 disk_images.append(False)
8682 self.src_images = disk_images
8684 old_name = export_info.get(constants.INISECT_INS, "name")
8686 exp_nic_count = export_info.getint(constants.INISECT_INS, "nic_count")
8687 except (TypeError, ValueError), err:
8688 raise errors.OpPrereqError("Invalid export file, nic_count is not"
8689 " an integer: %s" % str(err),
8691 if self.op.instance_name == old_name:
8692 for idx, nic in enumerate(self.nics):
8693 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
8694 nic_mac_ini = "nic%d_mac" % idx
8695 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
8697 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
8699 # ip ping checks (we use the same ip that was resolved in ExpandNames)
8700 if self.op.ip_check:
8701 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
8702 raise errors.OpPrereqError("IP %s of instance %s already in use" %
8703 (self.check_ip, self.op.instance_name),
8704 errors.ECODE_NOTUNIQUE)
8706 #### mac address generation
8707 # By generating here the mac address both the allocator and the hooks get
8708 # the real final mac address rather than the 'auto' or 'generate' value.
8709 # There is a race condition between the generation and the instance object
8710 # creation, which means that we know the mac is valid now, but we're not
8711 # sure it will be when we actually add the instance. If things go bad
8712 # adding the instance will abort because of a duplicate mac, and the
8713 # creation job will fail.
8714 for nic in self.nics:
8715 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8716 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
8720 if self.op.iallocator is not None:
8721 self._RunAllocator()
8723 #### node related checks
8725 # check primary node
8726 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
8727 assert self.pnode is not None, \
8728 "Cannot retrieve locked node %s" % self.op.pnode
8730 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
8731 pnode.name, errors.ECODE_STATE)
8733 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
8734 pnode.name, errors.ECODE_STATE)
8735 if not pnode.vm_capable:
8736 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
8737 " '%s'" % pnode.name, errors.ECODE_STATE)
8739 self.secondaries = []
8741 # mirror node verification
8742 if self.op.disk_template in constants.DTS_INT_MIRROR:
8743 if self.op.snode == pnode.name:
8744 raise errors.OpPrereqError("The secondary node cannot be the"
8745 " primary node", errors.ECODE_INVAL)
8746 _CheckNodeOnline(self, self.op.snode)
8747 _CheckNodeNotDrained(self, self.op.snode)
8748 _CheckNodeVmCapable(self, self.op.snode)
8749 self.secondaries.append(self.op.snode)
8751 nodenames = [pnode.name] + self.secondaries
8753 if not self.adopt_disks:
8754 # Check lv size requirements, if not adopting
8755 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
8756 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
8758 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
8759 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
8760 disk[constants.IDISK_ADOPT])
8761 for disk in self.disks])
8762 if len(all_lvs) != len(self.disks):
8763 raise errors.OpPrereqError("Duplicate volume names given for adoption",
8765 for lv_name in all_lvs:
8767 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
8768 # to ReserveLV uses the same syntax
8769 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
8770 except errors.ReservationError:
8771 raise errors.OpPrereqError("LV named %s used by another instance" %
8772 lv_name, errors.ECODE_NOTUNIQUE)
8774 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
8775 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
8777 node_lvs = self.rpc.call_lv_list([pnode.name],
8778 vg_names.payload.keys())[pnode.name]
8779 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
8780 node_lvs = node_lvs.payload
8782 delta = all_lvs.difference(node_lvs.keys())
8784 raise errors.OpPrereqError("Missing logical volume(s): %s" %
8785 utils.CommaJoin(delta),
8787 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
8789 raise errors.OpPrereqError("Online logical volumes found, cannot"
8790 " adopt: %s" % utils.CommaJoin(online_lvs),
8792 # update the size of disk based on what is found
8793 for dsk in self.disks:
8794 dsk[constants.IDISK_SIZE] = \
8795 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
8796 dsk[constants.IDISK_ADOPT])][0]))
8798 elif self.op.disk_template == constants.DT_BLOCK:
8799 # Normalize and de-duplicate device paths
8800 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
8801 for disk in self.disks])
8802 if len(all_disks) != len(self.disks):
8803 raise errors.OpPrereqError("Duplicate disk names given for adoption",
8805 baddisks = [d for d in all_disks
8806 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
8808 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
8809 " cannot be adopted" %
8810 (", ".join(baddisks),
8811 constants.ADOPTABLE_BLOCKDEV_ROOT),
8814 node_disks = self.rpc.call_bdev_sizes([pnode.name],
8815 list(all_disks))[pnode.name]
8816 node_disks.Raise("Cannot get block device information from node %s" %
8818 node_disks = node_disks.payload
8819 delta = all_disks.difference(node_disks.keys())
8821 raise errors.OpPrereqError("Missing block device(s): %s" %
8822 utils.CommaJoin(delta),
8824 for dsk in self.disks:
8825 dsk[constants.IDISK_SIZE] = \
8826 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
8828 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
8830 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
8831 # check OS parameters (remotely)
8832 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
8834 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
8836 # memory check on primary node
8838 _CheckNodeFreeMemory(self, self.pnode.name,
8839 "creating instance %s" % self.op.instance_name,
8840 self.be_full[constants.BE_MEMORY],
8843 self.dry_run_result = list(nodenames)
8845 def Exec(self, feedback_fn):
8846 """Create and add the instance to the cluster.
8849 instance = self.op.instance_name
8850 pnode_name = self.pnode.name
8852 ht_kind = self.op.hypervisor
8853 if ht_kind in constants.HTS_REQ_PORT:
8854 network_port = self.cfg.AllocatePort()
8858 disks = _GenerateDiskTemplate(self,
8859 self.op.disk_template,
8860 instance, pnode_name,
8863 self.instance_file_storage_dir,
8864 self.op.file_driver,
8868 iobj = objects.Instance(name=instance, os=self.op.os_type,
8869 primary_node=pnode_name,
8870 nics=self.nics, disks=disks,
8871 disk_template=self.op.disk_template,
8873 network_port=network_port,
8874 beparams=self.op.beparams,
8875 hvparams=self.op.hvparams,
8876 hypervisor=self.op.hypervisor,
8877 osparams=self.op.osparams,
8881 for tag in self.op.tags:
8884 if self.adopt_disks:
8885 if self.op.disk_template == constants.DT_PLAIN:
8886 # rename LVs to the newly-generated names; we need to construct
8887 # 'fake' LV disks with the old data, plus the new unique_id
8888 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
8890 for t_dsk, a_dsk in zip(tmp_disks, self.disks):
8891 rename_to.append(t_dsk.logical_id)
8892 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
8893 self.cfg.SetDiskID(t_dsk, pnode_name)
8894 result = self.rpc.call_blockdev_rename(pnode_name,
8895 zip(tmp_disks, rename_to))
8896 result.Raise("Failed to rename adoped LVs")
8898 feedback_fn("* creating instance disks...")
8900 _CreateDisks(self, iobj)
8901 except errors.OpExecError:
8902 self.LogWarning("Device creation failed, reverting...")
8904 _RemoveDisks(self, iobj)
8906 self.cfg.ReleaseDRBDMinors(instance)
8909 feedback_fn("adding instance %s to cluster config" % instance)
8911 self.cfg.AddInstance(iobj, self.proc.GetECId())
8913 # Declare that we don't want to remove the instance lock anymore, as we've
8914 # added the instance to the config
8915 del self.remove_locks[locking.LEVEL_INSTANCE]
8917 if self.op.mode == constants.INSTANCE_IMPORT:
8918 # Release unused nodes
8919 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
8922 _ReleaseLocks(self, locking.LEVEL_NODE)
8925 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
8926 feedback_fn("* wiping instance disks...")
8928 _WipeDisks(self, iobj)
8929 except errors.OpExecError, err:
8930 logging.exception("Wiping disks failed")
8931 self.LogWarning("Wiping instance disks failed (%s)", err)
8935 # Something is already wrong with the disks, don't do anything else
8937 elif self.op.wait_for_sync:
8938 disk_abort = not _WaitForSync(self, iobj)
8939 elif iobj.disk_template in constants.DTS_INT_MIRROR:
8940 # make sure the disks are not degraded (still sync-ing is ok)
8941 feedback_fn("* checking mirrors status")
8942 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
8947 _RemoveDisks(self, iobj)
8948 self.cfg.RemoveInstance(iobj.name)
8949 # Make sure the instance lock gets removed
8950 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
8951 raise errors.OpExecError("There are some degraded disks for"
8954 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
8955 if self.op.mode == constants.INSTANCE_CREATE:
8956 if not self.op.no_install:
8957 pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and
8958 not self.op.wait_for_sync)
8960 feedback_fn("* pausing disk sync to install instance OS")
8961 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
8963 for idx, success in enumerate(result.payload):
8965 logging.warn("pause-sync of instance %s for disk %d failed",
8968 feedback_fn("* running the instance OS create scripts...")
8969 # FIXME: pass debug option from opcode to backend
8970 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
8971 self.op.debug_level)
8973 feedback_fn("* resuming disk sync")
8974 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
8976 for idx, success in enumerate(result.payload):
8978 logging.warn("resume-sync of instance %s for disk %d failed",
8981 result.Raise("Could not add os for instance %s"
8982 " on node %s" % (instance, pnode_name))
8984 elif self.op.mode == constants.INSTANCE_IMPORT:
8985 feedback_fn("* running the instance OS import scripts...")
8989 for idx, image in enumerate(self.src_images):
8993 # FIXME: pass debug option from opcode to backend
8994 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
8995 constants.IEIO_FILE, (image, ),
8996 constants.IEIO_SCRIPT,
8997 (iobj.disks[idx], idx),
8999 transfers.append(dt)
9002 masterd.instance.TransferInstanceData(self, feedback_fn,
9003 self.op.src_node, pnode_name,
9004 self.pnode.secondary_ip,
9006 if not compat.all(import_result):
9007 self.LogWarning("Some disks for instance %s on node %s were not"
9008 " imported successfully" % (instance, pnode_name))
9010 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
9011 feedback_fn("* preparing remote import...")
9012 # The source cluster will stop the instance before attempting to make a
9013 # connection. In some cases stopping an instance can take a long time,
9014 # hence the shutdown timeout is added to the connection timeout.
9015 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
9016 self.op.source_shutdown_timeout)
9017 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9019 assert iobj.primary_node == self.pnode.name
9021 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
9022 self.source_x509_ca,
9023 self._cds, timeouts)
9024 if not compat.all(disk_results):
9025 # TODO: Should the instance still be started, even if some disks
9026 # failed to import (valid for local imports, too)?
9027 self.LogWarning("Some disks for instance %s on node %s were not"
9028 " imported successfully" % (instance, pnode_name))
9030 # Run rename script on newly imported instance
9031 assert iobj.name == instance
9032 feedback_fn("Running rename script for %s" % instance)
9033 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
9034 self.source_instance_name,
9035 self.op.debug_level)
9037 self.LogWarning("Failed to run rename script for %s on node"
9038 " %s: %s" % (instance, pnode_name, result.fail_msg))
9041 # also checked in the prereq part
9042 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
9046 iobj.admin_up = True
9047 self.cfg.Update(iobj, feedback_fn)
9048 logging.info("Starting instance %s on node %s", instance, pnode_name)
9049 feedback_fn("* starting instance...")
9050 result = self.rpc.call_instance_start(pnode_name, iobj,
9052 result.Raise("Could not start instance")
9054 return list(iobj.all_nodes)
9057 class LUInstanceConsole(NoHooksLU):
9058 """Connect to an instance's console.
9060 This is somewhat special in that it returns the command line that
9061 you need to run on the master node in order to connect to the
9067 def ExpandNames(self):
9068 self._ExpandAndLockInstance()
9070 def CheckPrereq(self):
9071 """Check prerequisites.
9073 This checks that the instance is in the cluster.
9076 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9077 assert self.instance is not None, \
9078 "Cannot retrieve locked instance %s" % self.op.instance_name
9079 _CheckNodeOnline(self, self.instance.primary_node)
9081 def Exec(self, feedback_fn):
9082 """Connect to the console of an instance
9085 instance = self.instance
9086 node = instance.primary_node
9088 node_insts = self.rpc.call_instance_list([node],
9089 [instance.hypervisor])[node]
9090 node_insts.Raise("Can't get node information from %s" % node)
9092 if instance.name not in node_insts.payload:
9093 if instance.admin_up:
9094 state = constants.INSTST_ERRORDOWN
9096 state = constants.INSTST_ADMINDOWN
9097 raise errors.OpExecError("Instance %s is not running (state %s)" %
9098 (instance.name, state))
9100 logging.debug("Connecting to console of %s on %s", instance.name, node)
9102 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
9105 def _GetInstanceConsole(cluster, instance):
9106 """Returns console information for an instance.
9108 @type cluster: L{objects.Cluster}
9109 @type instance: L{objects.Instance}
9113 hyper = hypervisor.GetHypervisor(instance.hypervisor)
9114 # beparams and hvparams are passed separately, to avoid editing the
9115 # instance and then saving the defaults in the instance itself.
9116 hvparams = cluster.FillHV(instance)
9117 beparams = cluster.FillBE(instance)
9118 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
9120 assert console.instance == instance.name
9121 assert console.Validate()
9123 return console.ToDict()
9126 class LUInstanceReplaceDisks(LogicalUnit):
9127 """Replace the disks of an instance.
9130 HPATH = "mirrors-replace"
9131 HTYPE = constants.HTYPE_INSTANCE
9134 def CheckArguments(self):
9135 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
9138 def ExpandNames(self):
9139 self._ExpandAndLockInstance()
9141 assert locking.LEVEL_NODE not in self.needed_locks
9142 assert locking.LEVEL_NODEGROUP not in self.needed_locks
9144 assert self.op.iallocator is None or self.op.remote_node is None, \
9145 "Conflicting options"
9147 if self.op.remote_node is not None:
9148 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9150 # Warning: do not remove the locking of the new secondary here
9151 # unless DRBD8.AddChildren is changed to work in parallel;
9152 # currently it doesn't since parallel invocations of
9153 # FindUnusedMinor will conflict
9154 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
9155 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
9157 self.needed_locks[locking.LEVEL_NODE] = []
9158 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9160 if self.op.iallocator is not None:
9161 # iallocator will select a new node in the same group
9162 self.needed_locks[locking.LEVEL_NODEGROUP] = []
9164 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
9165 self.op.iallocator, self.op.remote_node,
9166 self.op.disks, False, self.op.early_release)
9168 self.tasklets = [self.replacer]
9170 def DeclareLocks(self, level):
9171 if level == locking.LEVEL_NODEGROUP:
9172 assert self.op.remote_node is None
9173 assert self.op.iallocator is not None
9174 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
9176 self.share_locks[locking.LEVEL_NODEGROUP] = 1
9177 self.needed_locks[locking.LEVEL_NODEGROUP] = \
9178 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
9180 elif level == locking.LEVEL_NODE:
9181 if self.op.iallocator is not None:
9182 assert self.op.remote_node is None
9183 assert not self.needed_locks[locking.LEVEL_NODE]
9185 # Lock member nodes of all locked groups
9186 self.needed_locks[locking.LEVEL_NODE] = [node_name
9187 for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
9188 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
9190 self._LockInstancesNodes()
9192 def BuildHooksEnv(self):
9195 This runs on the master, the primary and all the secondaries.
9198 instance = self.replacer.instance
9200 "MODE": self.op.mode,
9201 "NEW_SECONDARY": self.op.remote_node,
9202 "OLD_SECONDARY": instance.secondary_nodes[0],
9204 env.update(_BuildInstanceHookEnvByObject(self, instance))
9207 def BuildHooksNodes(self):
9208 """Build hooks nodes.
9211 instance = self.replacer.instance
9213 self.cfg.GetMasterNode(),
9214 instance.primary_node,
9216 if self.op.remote_node is not None:
9217 nl.append(self.op.remote_node)
9220 def CheckPrereq(self):
9221 """Check prerequisites.
9224 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
9225 self.op.iallocator is None)
9227 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
9229 _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
9231 return LogicalUnit.CheckPrereq(self)
9234 class TLReplaceDisks(Tasklet):
9235 """Replaces disks for an instance.
9237 Note: Locking is not within the scope of this class.
9240 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
9241 disks, delay_iallocator, early_release):
9242 """Initializes this class.
9245 Tasklet.__init__(self, lu)
9248 self.instance_name = instance_name
9250 self.iallocator_name = iallocator_name
9251 self.remote_node = remote_node
9253 self.delay_iallocator = delay_iallocator
9254 self.early_release = early_release
9257 self.instance = None
9258 self.new_node = None
9259 self.target_node = None
9260 self.other_node = None
9261 self.remote_node_info = None
9262 self.node_secondary_ip = None
9265 def CheckArguments(mode, remote_node, iallocator):
9266 """Helper function for users of this class.
9269 # check for valid parameter combination
9270 if mode == constants.REPLACE_DISK_CHG:
9271 if remote_node is None and iallocator is None:
9272 raise errors.OpPrereqError("When changing the secondary either an"
9273 " iallocator script must be used or the"
9274 " new node given", errors.ECODE_INVAL)
9276 if remote_node is not None and iallocator is not None:
9277 raise errors.OpPrereqError("Give either the iallocator or the new"
9278 " secondary, not both", errors.ECODE_INVAL)
9280 elif remote_node is not None or iallocator is not None:
9281 # Not replacing the secondary
9282 raise errors.OpPrereqError("The iallocator and new node options can"
9283 " only be used when changing the"
9284 " secondary node", errors.ECODE_INVAL)
9287 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
9288 """Compute a new secondary node using an IAllocator.
9291 ial = IAllocator(lu.cfg, lu.rpc,
9292 mode=constants.IALLOCATOR_MODE_RELOC,
9294 relocate_from=list(relocate_from))
9296 ial.Run(iallocator_name)
9299 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
9300 " %s" % (iallocator_name, ial.info),
9303 if len(ial.result) != ial.required_nodes:
9304 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
9305 " of nodes (%s), required %s" %
9307 len(ial.result), ial.required_nodes),
9310 remote_node_name = ial.result[0]
9312 lu.LogInfo("Selected new secondary for instance '%s': %s",
9313 instance_name, remote_node_name)
9315 return remote_node_name
9317 def _FindFaultyDisks(self, node_name):
9318 """Wrapper for L{_FindFaultyInstanceDisks}.
9321 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
9324 def _CheckDisksActivated(self, instance):
9325 """Checks if the instance disks are activated.
9327 @param instance: The instance to check disks
9328 @return: True if they are activated, False otherwise
9331 nodes = instance.all_nodes
9333 for idx, dev in enumerate(instance.disks):
9335 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
9336 self.cfg.SetDiskID(dev, node)
9338 result = self.rpc.call_blockdev_find(node, dev)
9342 elif result.fail_msg or not result.payload:
9347 def CheckPrereq(self):
9348 """Check prerequisites.
9350 This checks that the instance is in the cluster.
9353 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
9354 assert instance is not None, \
9355 "Cannot retrieve locked instance %s" % self.instance_name
9357 if instance.disk_template != constants.DT_DRBD8:
9358 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
9359 " instances", errors.ECODE_INVAL)
9361 if len(instance.secondary_nodes) != 1:
9362 raise errors.OpPrereqError("The instance has a strange layout,"
9363 " expected one secondary but found %d" %
9364 len(instance.secondary_nodes),
9367 if not self.delay_iallocator:
9368 self._CheckPrereq2()
9370 def _CheckPrereq2(self):
9371 """Check prerequisites, second part.
9373 This function should always be part of CheckPrereq. It was separated and is
9374 now called from Exec because during node evacuation iallocator was only
9375 called with an unmodified cluster model, not taking planned changes into
9379 instance = self.instance
9380 secondary_node = instance.secondary_nodes[0]
9382 if self.iallocator_name is None:
9383 remote_node = self.remote_node
9385 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
9386 instance.name, instance.secondary_nodes)
9388 if remote_node is None:
9389 self.remote_node_info = None
9391 assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \
9392 "Remote node '%s' is not locked" % remote_node
9394 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
9395 assert self.remote_node_info is not None, \
9396 "Cannot retrieve locked node %s" % remote_node
9398 if remote_node == self.instance.primary_node:
9399 raise errors.OpPrereqError("The specified node is the primary node of"
9400 " the instance", errors.ECODE_INVAL)
9402 if remote_node == secondary_node:
9403 raise errors.OpPrereqError("The specified node is already the"
9404 " secondary node of the instance",
9407 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
9408 constants.REPLACE_DISK_CHG):
9409 raise errors.OpPrereqError("Cannot specify disks to be replaced",
9412 if self.mode == constants.REPLACE_DISK_AUTO:
9413 if not self._CheckDisksActivated(instance):
9414 raise errors.OpPrereqError("Please run activate-disks on instance %s"
9415 " first" % self.instance_name,
9417 faulty_primary = self._FindFaultyDisks(instance.primary_node)
9418 faulty_secondary = self._FindFaultyDisks(secondary_node)
9420 if faulty_primary and faulty_secondary:
9421 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
9422 " one node and can not be repaired"
9423 " automatically" % self.instance_name,
9427 self.disks = faulty_primary
9428 self.target_node = instance.primary_node
9429 self.other_node = secondary_node
9430 check_nodes = [self.target_node, self.other_node]
9431 elif faulty_secondary:
9432 self.disks = faulty_secondary
9433 self.target_node = secondary_node
9434 self.other_node = instance.primary_node
9435 check_nodes = [self.target_node, self.other_node]
9441 # Non-automatic modes
9442 if self.mode == constants.REPLACE_DISK_PRI:
9443 self.target_node = instance.primary_node
9444 self.other_node = secondary_node
9445 check_nodes = [self.target_node, self.other_node]
9447 elif self.mode == constants.REPLACE_DISK_SEC:
9448 self.target_node = secondary_node
9449 self.other_node = instance.primary_node
9450 check_nodes = [self.target_node, self.other_node]
9452 elif self.mode == constants.REPLACE_DISK_CHG:
9453 self.new_node = remote_node
9454 self.other_node = instance.primary_node
9455 self.target_node = secondary_node
9456 check_nodes = [self.new_node, self.other_node]
9458 _CheckNodeNotDrained(self.lu, remote_node)
9459 _CheckNodeVmCapable(self.lu, remote_node)
9461 old_node_info = self.cfg.GetNodeInfo(secondary_node)
9462 assert old_node_info is not None
9463 if old_node_info.offline and not self.early_release:
9464 # doesn't make sense to delay the release
9465 self.early_release = True
9466 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
9467 " early-release mode", secondary_node)
9470 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
9473 # If not specified all disks should be replaced
9475 self.disks = range(len(self.instance.disks))
9477 for node in check_nodes:
9478 _CheckNodeOnline(self.lu, node)
9480 touched_nodes = frozenset(node_name for node_name in [self.new_node,
9483 if node_name is not None)
9485 # Release unneeded node locks
9486 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
9488 # Release any owned node group
9489 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
9490 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
9492 # Check whether disks are valid
9493 for disk_idx in self.disks:
9494 instance.FindDisk(disk_idx)
9496 # Get secondary node IP addresses
9497 self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node)
9498 in self.cfg.GetMultiNodeInfo(touched_nodes))
9500 def Exec(self, feedback_fn):
9501 """Execute disk replacement.
9503 This dispatches the disk replacement to the appropriate handler.
9506 if self.delay_iallocator:
9507 self._CheckPrereq2()
9510 # Verify owned locks before starting operation
9511 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
9512 assert set(owned_nodes) == set(self.node_secondary_ip), \
9513 ("Incorrect node locks, owning %s, expected %s" %
9514 (owned_nodes, self.node_secondary_ip.keys()))
9516 owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
9517 assert list(owned_instances) == [self.instance_name], \
9518 "Instance '%s' not locked" % self.instance_name
9520 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
9521 "Should not own any node group lock at this point"
9524 feedback_fn("No disks need replacement")
9527 feedback_fn("Replacing disk(s) %s for %s" %
9528 (utils.CommaJoin(self.disks), self.instance.name))
9530 activate_disks = (not self.instance.admin_up)
9532 # Activate the instance disks if we're replacing them on a down instance
9534 _StartInstanceDisks(self.lu, self.instance, True)
9537 # Should we replace the secondary node?
9538 if self.new_node is not None:
9539 fn = self._ExecDrbd8Secondary
9541 fn = self._ExecDrbd8DiskOnly
9543 result = fn(feedback_fn)
9545 # Deactivate the instance disks if we're replacing them on a
9548 _SafeShutdownInstanceDisks(self.lu, self.instance)
9551 # Verify owned locks
9552 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
9553 nodes = frozenset(self.node_secondary_ip)
9554 assert ((self.early_release and not owned_nodes) or
9555 (not self.early_release and not (set(owned_nodes) - nodes))), \
9556 ("Not owning the correct locks, early_release=%s, owned=%r,"
9557 " nodes=%r" % (self.early_release, owned_nodes, nodes))
9561 def _CheckVolumeGroup(self, nodes):
9562 self.lu.LogInfo("Checking volume groups")
9564 vgname = self.cfg.GetVGName()
9566 # Make sure volume group exists on all involved nodes
9567 results = self.rpc.call_vg_list(nodes)
9569 raise errors.OpExecError("Can't list volume groups on the nodes")
9573 res.Raise("Error checking node %s" % node)
9574 if vgname not in res.payload:
9575 raise errors.OpExecError("Volume group '%s' not found on node %s" %
9578 def _CheckDisksExistence(self, nodes):
9579 # Check disk existence
9580 for idx, dev in enumerate(self.instance.disks):
9581 if idx not in self.disks:
9585 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
9586 self.cfg.SetDiskID(dev, node)
9588 result = self.rpc.call_blockdev_find(node, dev)
9590 msg = result.fail_msg
9591 if msg or not result.payload:
9593 msg = "disk not found"
9594 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
9597 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
9598 for idx, dev in enumerate(self.instance.disks):
9599 if idx not in self.disks:
9602 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
9605 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
9607 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
9608 " replace disks for instance %s" %
9609 (node_name, self.instance.name))
9611 def _CreateNewStorage(self, node_name):
9612 """Create new storage on the primary or secondary node.
9614 This is only used for same-node replaces, not for changing the
9615 secondary node, hence we don't want to modify the existing disk.
9620 for idx, dev in enumerate(self.instance.disks):
9621 if idx not in self.disks:
9624 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
9626 self.cfg.SetDiskID(dev, node_name)
9628 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
9629 names = _GenerateUniqueNames(self.lu, lv_names)
9631 vg_data = dev.children[0].logical_id[0]
9632 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
9633 logical_id=(vg_data, names[0]))
9634 vg_meta = dev.children[1].logical_id[0]
9635 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
9636 logical_id=(vg_meta, names[1]))
9638 new_lvs = [lv_data, lv_meta]
9639 old_lvs = [child.Copy() for child in dev.children]
9640 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
9642 # we pass force_create=True to force the LVM creation
9643 for new_lv in new_lvs:
9644 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
9645 _GetInstanceInfoText(self.instance), False)
9649 def _CheckDevices(self, node_name, iv_names):
9650 for name, (dev, _, _) in iv_names.iteritems():
9651 self.cfg.SetDiskID(dev, node_name)
9653 result = self.rpc.call_blockdev_find(node_name, dev)
9655 msg = result.fail_msg
9656 if msg or not result.payload:
9658 msg = "disk not found"
9659 raise errors.OpExecError("Can't find DRBD device %s: %s" %
9662 if result.payload.is_degraded:
9663 raise errors.OpExecError("DRBD device %s is degraded!" % name)
9665 def _RemoveOldStorage(self, node_name, iv_names):
9666 for name, (_, old_lvs, _) in iv_names.iteritems():
9667 self.lu.LogInfo("Remove logical volumes for %s" % name)
9670 self.cfg.SetDiskID(lv, node_name)
9672 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
9674 self.lu.LogWarning("Can't remove old LV: %s" % msg,
9675 hint="remove unused LVs manually")
9677 def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
9678 """Replace a disk on the primary or secondary for DRBD 8.
9680 The algorithm for replace is quite complicated:
9682 1. for each disk to be replaced:
9684 1. create new LVs on the target node with unique names
9685 1. detach old LVs from the drbd device
9686 1. rename old LVs to name_replaced.<time_t>
9687 1. rename new LVs to old LVs
9688 1. attach the new LVs (with the old names now) to the drbd device
9690 1. wait for sync across all devices
9692 1. for each modified disk:
9694 1. remove old LVs (which have the name name_replaces.<time_t>)
9696 Failures are not very well handled.
9701 # Step: check device activation
9702 self.lu.LogStep(1, steps_total, "Check device existence")
9703 self._CheckDisksExistence([self.other_node, self.target_node])
9704 self._CheckVolumeGroup([self.target_node, self.other_node])
9706 # Step: check other node consistency
9707 self.lu.LogStep(2, steps_total, "Check peer consistency")
9708 self._CheckDisksConsistency(self.other_node,
9709 self.other_node == self.instance.primary_node,
9712 # Step: create new storage
9713 self.lu.LogStep(3, steps_total, "Allocate new storage")
9714 iv_names = self._CreateNewStorage(self.target_node)
9716 # Step: for each lv, detach+rename*2+attach
9717 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9718 for dev, old_lvs, new_lvs in iv_names.itervalues():
9719 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
9721 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
9723 result.Raise("Can't detach drbd from local storage on node"
9724 " %s for device %s" % (self.target_node, dev.iv_name))
9726 #cfg.Update(instance)
9728 # ok, we created the new LVs, so now we know we have the needed
9729 # storage; as such, we proceed on the target node to rename
9730 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
9731 # using the assumption that logical_id == physical_id (which in
9732 # turn is the unique_id on that node)
9734 # FIXME(iustin): use a better name for the replaced LVs
9735 temp_suffix = int(time.time())
9736 ren_fn = lambda d, suff: (d.physical_id[0],
9737 d.physical_id[1] + "_replaced-%s" % suff)
9739 # Build the rename list based on what LVs exist on the node
9740 rename_old_to_new = []
9741 for to_ren in old_lvs:
9742 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
9743 if not result.fail_msg and result.payload:
9745 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
9747 self.lu.LogInfo("Renaming the old LVs on the target node")
9748 result = self.rpc.call_blockdev_rename(self.target_node,
9750 result.Raise("Can't rename old LVs on node %s" % self.target_node)
9752 # Now we rename the new LVs to the old LVs
9753 self.lu.LogInfo("Renaming the new LVs on the target node")
9754 rename_new_to_old = [(new, old.physical_id)
9755 for old, new in zip(old_lvs, new_lvs)]
9756 result = self.rpc.call_blockdev_rename(self.target_node,
9758 result.Raise("Can't rename new LVs on node %s" % self.target_node)
9760 # Intermediate steps of in memory modifications
9761 for old, new in zip(old_lvs, new_lvs):
9762 new.logical_id = old.logical_id
9763 self.cfg.SetDiskID(new, self.target_node)
9765 # We need to modify old_lvs so that removal later removes the
9766 # right LVs, not the newly added ones; note that old_lvs is a
9768 for disk in old_lvs:
9769 disk.logical_id = ren_fn(disk, temp_suffix)
9770 self.cfg.SetDiskID(disk, self.target_node)
9772 # Now that the new lvs have the old name, we can add them to the device
9773 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
9774 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
9776 msg = result.fail_msg
9778 for new_lv in new_lvs:
9779 msg2 = self.rpc.call_blockdev_remove(self.target_node,
9782 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
9783 hint=("cleanup manually the unused logical"
9785 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
9788 if self.early_release:
9789 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9791 self._RemoveOldStorage(self.target_node, iv_names)
9792 # WARNING: we release both node locks here, do not do other RPCs
9793 # than WaitForSync to the primary node
9794 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9795 names=[self.target_node, self.other_node])
9798 # This can fail as the old devices are degraded and _WaitForSync
9799 # does a combined result over all disks, so we don't check its return value
9800 self.lu.LogStep(cstep, steps_total, "Sync devices")
9802 _WaitForSync(self.lu, self.instance)
9804 # Check all devices manually
9805 self._CheckDevices(self.instance.primary_node, iv_names)
9807 # Step: remove old storage
9808 if not self.early_release:
9809 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9811 self._RemoveOldStorage(self.target_node, iv_names)
9813 def _ExecDrbd8Secondary(self, feedback_fn):
9814 """Replace the secondary node for DRBD 8.
9816 The algorithm for replace is quite complicated:
9817 - for all disks of the instance:
9818 - create new LVs on the new node with same names
9819 - shutdown the drbd device on the old secondary
9820 - disconnect the drbd network on the primary
9821 - create the drbd device on the new secondary
9822 - network attach the drbd on the primary, using an artifice:
9823 the drbd code for Attach() will connect to the network if it
9824 finds a device which is connected to the good local disks but
9826 - wait for sync across all devices
9827 - remove all disks from the old secondary
9829 Failures are not very well handled.
9834 pnode = self.instance.primary_node
9836 # Step: check device activation
9837 self.lu.LogStep(1, steps_total, "Check device existence")
9838 self._CheckDisksExistence([self.instance.primary_node])
9839 self._CheckVolumeGroup([self.instance.primary_node])
9841 # Step: check other node consistency
9842 self.lu.LogStep(2, steps_total, "Check peer consistency")
9843 self._CheckDisksConsistency(self.instance.primary_node, True, True)
9845 # Step: create new storage
9846 self.lu.LogStep(3, steps_total, "Allocate new storage")
9847 for idx, dev in enumerate(self.instance.disks):
9848 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
9849 (self.new_node, idx))
9850 # we pass force_create=True to force LVM creation
9851 for new_lv in dev.children:
9852 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
9853 _GetInstanceInfoText(self.instance), False)
9855 # Step 4: dbrd minors and drbd setups changes
9856 # after this, we must manually remove the drbd minors on both the
9857 # error and the success paths
9858 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9859 minors = self.cfg.AllocateDRBDMinor([self.new_node
9860 for dev in self.instance.disks],
9862 logging.debug("Allocated minors %r", minors)
9865 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
9866 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
9867 (self.new_node, idx))
9868 # create new devices on new_node; note that we create two IDs:
9869 # one without port, so the drbd will be activated without
9870 # networking information on the new node at this stage, and one
9871 # with network, for the latter activation in step 4
9872 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
9873 if self.instance.primary_node == o_node1:
9876 assert self.instance.primary_node == o_node2, "Three-node instance?"
9879 new_alone_id = (self.instance.primary_node, self.new_node, None,
9880 p_minor, new_minor, o_secret)
9881 new_net_id = (self.instance.primary_node, self.new_node, o_port,
9882 p_minor, new_minor, o_secret)
9884 iv_names[idx] = (dev, dev.children, new_net_id)
9885 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
9887 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
9888 logical_id=new_alone_id,
9889 children=dev.children,
9892 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
9893 _GetInstanceInfoText(self.instance), False)
9894 except errors.GenericError:
9895 self.cfg.ReleaseDRBDMinors(self.instance.name)
9898 # We have new devices, shutdown the drbd on the old secondary
9899 for idx, dev in enumerate(self.instance.disks):
9900 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
9901 self.cfg.SetDiskID(dev, self.target_node)
9902 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
9904 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
9905 "node: %s" % (idx, msg),
9906 hint=("Please cleanup this device manually as"
9907 " soon as possible"))
9909 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
9910 result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
9911 self.instance.disks)[pnode]
9913 msg = result.fail_msg
9915 # detaches didn't succeed (unlikely)
9916 self.cfg.ReleaseDRBDMinors(self.instance.name)
9917 raise errors.OpExecError("Can't detach the disks from the network on"
9918 " old node: %s" % (msg,))
9920 # if we managed to detach at least one, we update all the disks of
9921 # the instance to point to the new secondary
9922 self.lu.LogInfo("Updating instance configuration")
9923 for dev, _, new_logical_id in iv_names.itervalues():
9924 dev.logical_id = new_logical_id
9925 self.cfg.SetDiskID(dev, self.instance.primary_node)
9927 self.cfg.Update(self.instance, feedback_fn)
9929 # and now perform the drbd attach
9930 self.lu.LogInfo("Attaching primary drbds to new secondary"
9931 " (standalone => connected)")
9932 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
9934 self.node_secondary_ip,
9935 self.instance.disks,
9938 for to_node, to_result in result.items():
9939 msg = to_result.fail_msg
9941 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
9943 hint=("please do a gnt-instance info to see the"
9944 " status of disks"))
9946 if self.early_release:
9947 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9949 self._RemoveOldStorage(self.target_node, iv_names)
9950 # WARNING: we release all node locks here, do not do other RPCs
9951 # than WaitForSync to the primary node
9952 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9953 names=[self.instance.primary_node,
9958 # This can fail as the old devices are degraded and _WaitForSync
9959 # does a combined result over all disks, so we don't check its return value
9960 self.lu.LogStep(cstep, steps_total, "Sync devices")
9962 _WaitForSync(self.lu, self.instance)
9964 # Check all devices manually
9965 self._CheckDevices(self.instance.primary_node, iv_names)
9967 # Step: remove old storage
9968 if not self.early_release:
9969 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9970 self._RemoveOldStorage(self.target_node, iv_names)
9973 class LURepairNodeStorage(NoHooksLU):
9974 """Repairs the volume group on a node.
9979 def CheckArguments(self):
9980 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
9982 storage_type = self.op.storage_type
9984 if (constants.SO_FIX_CONSISTENCY not in
9985 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
9986 raise errors.OpPrereqError("Storage units of type '%s' can not be"
9987 " repaired" % storage_type,
9990 def ExpandNames(self):
9991 self.needed_locks = {
9992 locking.LEVEL_NODE: [self.op.node_name],
9995 def _CheckFaultyDisks(self, instance, node_name):
9996 """Ensure faulty disks abort the opcode or at least warn."""
9998 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
10000 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
10001 " node '%s'" % (instance.name, node_name),
10002 errors.ECODE_STATE)
10003 except errors.OpPrereqError, err:
10004 if self.op.ignore_consistency:
10005 self.proc.LogWarning(str(err.args[0]))
10009 def CheckPrereq(self):
10010 """Check prerequisites.
10013 # Check whether any instance on this node has faulty disks
10014 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
10015 if not inst.admin_up:
10017 check_nodes = set(inst.all_nodes)
10018 check_nodes.discard(self.op.node_name)
10019 for inst_node_name in check_nodes:
10020 self._CheckFaultyDisks(inst, inst_node_name)
10022 def Exec(self, feedback_fn):
10023 feedback_fn("Repairing storage unit '%s' on %s ..." %
10024 (self.op.name, self.op.node_name))
10026 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
10027 result = self.rpc.call_storage_execute(self.op.node_name,
10028 self.op.storage_type, st_args,
10030 constants.SO_FIX_CONSISTENCY)
10031 result.Raise("Failed to repair storage unit '%s' on %s" %
10032 (self.op.name, self.op.node_name))
10035 class LUNodeEvacuate(NoHooksLU):
10036 """Evacuates instances off a list of nodes.
10041 def CheckArguments(self):
10042 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
10044 def ExpandNames(self):
10045 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
10047 if self.op.remote_node is not None:
10048 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10049 assert self.op.remote_node
10051 if self.op.remote_node == self.op.node_name:
10052 raise errors.OpPrereqError("Can not use evacuated node as a new"
10053 " secondary node", errors.ECODE_INVAL)
10055 if self.op.mode != constants.IALLOCATOR_NEVAC_SEC:
10056 raise errors.OpPrereqError("Without the use of an iallocator only"
10057 " secondary instances can be evacuated",
10058 errors.ECODE_INVAL)
10061 self.share_locks = _ShareAll()
10062 self.needed_locks = {
10063 locking.LEVEL_INSTANCE: [],
10064 locking.LEVEL_NODEGROUP: [],
10065 locking.LEVEL_NODE: [],
10068 if self.op.remote_node is None:
10069 # Iallocator will choose any node(s) in the same group
10070 group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name])
10072 group_nodes = frozenset([self.op.remote_node])
10074 # Determine nodes to be locked
10075 self.lock_nodes = set([self.op.node_name]) | group_nodes
10077 def _DetermineInstances(self):
10078 """Builds list of instances to operate on.
10081 assert self.op.mode in constants.IALLOCATOR_NEVAC_MODES
10083 if self.op.mode == constants.IALLOCATOR_NEVAC_PRI:
10084 # Primary instances only
10085 inst_fn = _GetNodePrimaryInstances
10086 assert self.op.remote_node is None, \
10087 "Evacuating primary instances requires iallocator"
10088 elif self.op.mode == constants.IALLOCATOR_NEVAC_SEC:
10089 # Secondary instances only
10090 inst_fn = _GetNodeSecondaryInstances
10093 assert self.op.mode == constants.IALLOCATOR_NEVAC_ALL
10094 inst_fn = _GetNodeInstances
10096 return inst_fn(self.cfg, self.op.node_name)
10098 def DeclareLocks(self, level):
10099 if level == locking.LEVEL_INSTANCE:
10100 # Lock instances optimistically, needs verification once node and group
10101 # locks have been acquired
10102 self.needed_locks[locking.LEVEL_INSTANCE] = \
10103 set(i.name for i in self._DetermineInstances())
10105 elif level == locking.LEVEL_NODEGROUP:
10106 # Lock node groups optimistically, needs verification once nodes have
10108 self.needed_locks[locking.LEVEL_NODEGROUP] = \
10109 self.cfg.GetNodeGroupsFromNodes(self.lock_nodes)
10111 elif level == locking.LEVEL_NODE:
10112 self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes
10114 def CheckPrereq(self):
10116 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
10117 owned_nodes = self.owned_locks(locking.LEVEL_NODE)
10118 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
10120 assert owned_nodes == self.lock_nodes
10122 wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
10123 if owned_groups != wanted_groups:
10124 raise errors.OpExecError("Node groups changed since locks were acquired,"
10125 " current groups are '%s', used to be '%s'" %
10126 (utils.CommaJoin(wanted_groups),
10127 utils.CommaJoin(owned_groups)))
10129 # Determine affected instances
10130 self.instances = self._DetermineInstances()
10131 self.instance_names = [i.name for i in self.instances]
10133 if set(self.instance_names) != owned_instances:
10134 raise errors.OpExecError("Instances on node '%s' changed since locks"
10135 " were acquired, current instances are '%s',"
10136 " used to be '%s'" %
10137 (self.op.node_name,
10138 utils.CommaJoin(self.instance_names),
10139 utils.CommaJoin(owned_instances)))
10141 if self.instance_names:
10142 self.LogInfo("Evacuating instances from node '%s': %s",
10144 utils.CommaJoin(utils.NiceSort(self.instance_names)))
10146 self.LogInfo("No instances to evacuate from node '%s'",
10149 if self.op.remote_node is not None:
10150 for i in self.instances:
10151 if i.primary_node == self.op.remote_node:
10152 raise errors.OpPrereqError("Node %s is the primary node of"
10153 " instance %s, cannot use it as"
10155 (self.op.remote_node, i.name),
10156 errors.ECODE_INVAL)
10158 def Exec(self, feedback_fn):
10159 assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None)
10161 if not self.instance_names:
10162 # No instances to evacuate
10165 elif self.op.iallocator is not None:
10166 # TODO: Implement relocation to other group
10167 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_NODE_EVAC,
10168 evac_mode=self.op.mode,
10169 instances=list(self.instance_names))
10171 ial.Run(self.op.iallocator)
10173 if not ial.success:
10174 raise errors.OpPrereqError("Can't compute node evacuation using"
10175 " iallocator '%s': %s" %
10176 (self.op.iallocator, ial.info),
10177 errors.ECODE_NORES)
10179 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True)
10181 elif self.op.remote_node is not None:
10182 assert self.op.mode == constants.IALLOCATOR_NEVAC_SEC
10184 [opcodes.OpInstanceReplaceDisks(instance_name=instance_name,
10185 remote_node=self.op.remote_node,
10187 mode=constants.REPLACE_DISK_CHG,
10188 early_release=self.op.early_release)]
10189 for instance_name in self.instance_names
10193 raise errors.ProgrammerError("No iallocator or remote node")
10195 return ResultWithJobs(jobs)
10198 def _SetOpEarlyRelease(early_release, op):
10199 """Sets C{early_release} flag on opcodes if available.
10203 op.early_release = early_release
10204 except AttributeError:
10205 assert not isinstance(op, opcodes.OpInstanceReplaceDisks)
10210 def _NodeEvacDest(use_nodes, group, nodes):
10211 """Returns group or nodes depending on caller's choice.
10215 return utils.CommaJoin(nodes)
10220 def _LoadNodeEvacResult(lu, alloc_result, early_release, use_nodes):
10221 """Unpacks the result of change-group and node-evacuate iallocator requests.
10223 Iallocator modes L{constants.IALLOCATOR_MODE_NODE_EVAC} and
10224 L{constants.IALLOCATOR_MODE_CHG_GROUP}.
10226 @type lu: L{LogicalUnit}
10227 @param lu: Logical unit instance
10228 @type alloc_result: tuple/list
10229 @param alloc_result: Result from iallocator
10230 @type early_release: bool
10231 @param early_release: Whether to release locks early if possible
10232 @type use_nodes: bool
10233 @param use_nodes: Whether to display node names instead of groups
10236 (moved, failed, jobs) = alloc_result
10239 lu.LogWarning("Unable to evacuate instances %s",
10240 utils.CommaJoin("%s (%s)" % (name, reason)
10241 for (name, reason) in failed))
10244 lu.LogInfo("Instances to be moved: %s",
10245 utils.CommaJoin("%s (to %s)" %
10246 (name, _NodeEvacDest(use_nodes, group, nodes))
10247 for (name, group, nodes) in moved))
10249 return [map(compat.partial(_SetOpEarlyRelease, early_release),
10250 map(opcodes.OpCode.LoadOpCode, ops))
10254 class LUInstanceGrowDisk(LogicalUnit):
10255 """Grow a disk of an instance.
10258 HPATH = "disk-grow"
10259 HTYPE = constants.HTYPE_INSTANCE
10262 def ExpandNames(self):
10263 self._ExpandAndLockInstance()
10264 self.needed_locks[locking.LEVEL_NODE] = []
10265 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10267 def DeclareLocks(self, level):
10268 if level == locking.LEVEL_NODE:
10269 self._LockInstancesNodes()
10271 def BuildHooksEnv(self):
10272 """Build hooks env.
10274 This runs on the master, the primary and all the secondaries.
10278 "DISK": self.op.disk,
10279 "AMOUNT": self.op.amount,
10281 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
10284 def BuildHooksNodes(self):
10285 """Build hooks nodes.
10288 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
10291 def CheckPrereq(self):
10292 """Check prerequisites.
10294 This checks that the instance is in the cluster.
10297 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10298 assert instance is not None, \
10299 "Cannot retrieve locked instance %s" % self.op.instance_name
10300 nodenames = list(instance.all_nodes)
10301 for node in nodenames:
10302 _CheckNodeOnline(self, node)
10304 self.instance = instance
10306 if instance.disk_template not in constants.DTS_GROWABLE:
10307 raise errors.OpPrereqError("Instance's disk layout does not support"
10308 " growing", errors.ECODE_INVAL)
10310 self.disk = instance.FindDisk(self.op.disk)
10312 if instance.disk_template not in (constants.DT_FILE,
10313 constants.DT_SHARED_FILE):
10314 # TODO: check the free disk space for file, when that feature will be
10316 _CheckNodesFreeDiskPerVG(self, nodenames,
10317 self.disk.ComputeGrowth(self.op.amount))
10319 def Exec(self, feedback_fn):
10320 """Execute disk grow.
10323 instance = self.instance
10326 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
10328 raise errors.OpExecError("Cannot activate block device to grow")
10330 # First run all grow ops in dry-run mode
10331 for node in instance.all_nodes:
10332 self.cfg.SetDiskID(disk, node)
10333 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, True)
10334 result.Raise("Grow request failed to node %s" % node)
10336 # We know that (as far as we can test) operations across different
10337 # nodes will succeed, time to run it for real
10338 for node in instance.all_nodes:
10339 self.cfg.SetDiskID(disk, node)
10340 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, False)
10341 result.Raise("Grow request failed to node %s" % node)
10343 # TODO: Rewrite code to work properly
10344 # DRBD goes into sync mode for a short amount of time after executing the
10345 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
10346 # calling "resize" in sync mode fails. Sleeping for a short amount of
10347 # time is a work-around.
10350 disk.RecordGrow(self.op.amount)
10351 self.cfg.Update(instance, feedback_fn)
10352 if self.op.wait_for_sync:
10353 disk_abort = not _WaitForSync(self, instance, disks=[disk])
10355 self.proc.LogWarning("Disk sync-ing has not returned a good"
10356 " status; please check the instance")
10357 if not instance.admin_up:
10358 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
10359 elif not instance.admin_up:
10360 self.proc.LogWarning("Not shutting down the disk even if the instance is"
10361 " not supposed to be running because no wait for"
10362 " sync mode was requested")
10365 class LUInstanceQueryData(NoHooksLU):
10366 """Query runtime instance data.
10371 def ExpandNames(self):
10372 self.needed_locks = {}
10374 # Use locking if requested or when non-static information is wanted
10375 if not (self.op.static or self.op.use_locking):
10376 self.LogWarning("Non-static data requested, locks need to be acquired")
10377 self.op.use_locking = True
10379 if self.op.instances or not self.op.use_locking:
10380 # Expand instance names right here
10381 self.wanted_names = _GetWantedInstances(self, self.op.instances)
10383 # Will use acquired locks
10384 self.wanted_names = None
10386 if self.op.use_locking:
10387 self.share_locks = _ShareAll()
10389 if self.wanted_names is None:
10390 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
10392 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
10394 self.needed_locks[locking.LEVEL_NODE] = []
10395 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10397 def DeclareLocks(self, level):
10398 if self.op.use_locking and level == locking.LEVEL_NODE:
10399 self._LockInstancesNodes()
10401 def CheckPrereq(self):
10402 """Check prerequisites.
10404 This only checks the optional instance list against the existing names.
10407 if self.wanted_names is None:
10408 assert self.op.use_locking, "Locking was not used"
10409 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
10411 self.wanted_instances = \
10412 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
10414 def _ComputeBlockdevStatus(self, node, instance_name, dev):
10415 """Returns the status of a block device
10418 if self.op.static or not node:
10421 self.cfg.SetDiskID(dev, node)
10423 result = self.rpc.call_blockdev_find(node, dev)
10427 result.Raise("Can't compute disk status for %s" % instance_name)
10429 status = result.payload
10433 return (status.dev_path, status.major, status.minor,
10434 status.sync_percent, status.estimated_time,
10435 status.is_degraded, status.ldisk_status)
10437 def _ComputeDiskStatus(self, instance, snode, dev):
10438 """Compute block device status.
10441 if dev.dev_type in constants.LDS_DRBD:
10442 # we change the snode then (otherwise we use the one passed in)
10443 if dev.logical_id[0] == instance.primary_node:
10444 snode = dev.logical_id[1]
10446 snode = dev.logical_id[0]
10448 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
10449 instance.name, dev)
10450 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
10453 dev_children = map(compat.partial(self._ComputeDiskStatus,
10460 "iv_name": dev.iv_name,
10461 "dev_type": dev.dev_type,
10462 "logical_id": dev.logical_id,
10463 "physical_id": dev.physical_id,
10464 "pstatus": dev_pstatus,
10465 "sstatus": dev_sstatus,
10466 "children": dev_children,
10471 def Exec(self, feedback_fn):
10472 """Gather and return data"""
10475 cluster = self.cfg.GetClusterInfo()
10477 pri_nodes = self.cfg.GetMultiNodeInfo(i.primary_node
10478 for i in self.wanted_instances)
10479 for instance, (_, pnode) in zip(self.wanted_instances, pri_nodes):
10480 if self.op.static or pnode.offline:
10481 remote_state = None
10483 self.LogWarning("Primary node %s is marked offline, returning static"
10484 " information only for instance %s" %
10485 (pnode.name, instance.name))
10487 remote_info = self.rpc.call_instance_info(instance.primary_node,
10489 instance.hypervisor)
10490 remote_info.Raise("Error checking node %s" % instance.primary_node)
10491 remote_info = remote_info.payload
10492 if remote_info and "state" in remote_info:
10493 remote_state = "up"
10495 remote_state = "down"
10497 if instance.admin_up:
10498 config_state = "up"
10500 config_state = "down"
10502 disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
10505 result[instance.name] = {
10506 "name": instance.name,
10507 "config_state": config_state,
10508 "run_state": remote_state,
10509 "pnode": instance.primary_node,
10510 "snodes": instance.secondary_nodes,
10512 # this happens to be the same format used for hooks
10513 "nics": _NICListToTuple(self, instance.nics),
10514 "disk_template": instance.disk_template,
10516 "hypervisor": instance.hypervisor,
10517 "network_port": instance.network_port,
10518 "hv_instance": instance.hvparams,
10519 "hv_actual": cluster.FillHV(instance, skip_globals=True),
10520 "be_instance": instance.beparams,
10521 "be_actual": cluster.FillBE(instance),
10522 "os_instance": instance.osparams,
10523 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
10524 "serial_no": instance.serial_no,
10525 "mtime": instance.mtime,
10526 "ctime": instance.ctime,
10527 "uuid": instance.uuid,
10533 class LUInstanceSetParams(LogicalUnit):
10534 """Modifies an instances's parameters.
10537 HPATH = "instance-modify"
10538 HTYPE = constants.HTYPE_INSTANCE
10541 def CheckArguments(self):
10542 if not (self.op.nics or self.op.disks or self.op.disk_template or
10543 self.op.hvparams or self.op.beparams or self.op.os_name):
10544 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
10546 if self.op.hvparams:
10547 _CheckGlobalHvParams(self.op.hvparams)
10551 for disk_op, disk_dict in self.op.disks:
10552 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
10553 if disk_op == constants.DDM_REMOVE:
10554 disk_addremove += 1
10556 elif disk_op == constants.DDM_ADD:
10557 disk_addremove += 1
10559 if not isinstance(disk_op, int):
10560 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
10561 if not isinstance(disk_dict, dict):
10562 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
10563 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
10565 if disk_op == constants.DDM_ADD:
10566 mode = disk_dict.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
10567 if mode not in constants.DISK_ACCESS_SET:
10568 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
10569 errors.ECODE_INVAL)
10570 size = disk_dict.get(constants.IDISK_SIZE, None)
10572 raise errors.OpPrereqError("Required disk parameter size missing",
10573 errors.ECODE_INVAL)
10576 except (TypeError, ValueError), err:
10577 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
10578 str(err), errors.ECODE_INVAL)
10579 disk_dict[constants.IDISK_SIZE] = size
10581 # modification of disk
10582 if constants.IDISK_SIZE in disk_dict:
10583 raise errors.OpPrereqError("Disk size change not possible, use"
10584 " grow-disk", errors.ECODE_INVAL)
10586 if disk_addremove > 1:
10587 raise errors.OpPrereqError("Only one disk add or remove operation"
10588 " supported at a time", errors.ECODE_INVAL)
10590 if self.op.disks and self.op.disk_template is not None:
10591 raise errors.OpPrereqError("Disk template conversion and other disk"
10592 " changes not supported at the same time",
10593 errors.ECODE_INVAL)
10595 if (self.op.disk_template and
10596 self.op.disk_template in constants.DTS_INT_MIRROR and
10597 self.op.remote_node is None):
10598 raise errors.OpPrereqError("Changing the disk template to a mirrored"
10599 " one requires specifying a secondary node",
10600 errors.ECODE_INVAL)
10604 for nic_op, nic_dict in self.op.nics:
10605 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
10606 if nic_op == constants.DDM_REMOVE:
10609 elif nic_op == constants.DDM_ADD:
10612 if not isinstance(nic_op, int):
10613 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
10614 if not isinstance(nic_dict, dict):
10615 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
10616 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
10618 # nic_dict should be a dict
10619 nic_ip = nic_dict.get(constants.INIC_IP, None)
10620 if nic_ip is not None:
10621 if nic_ip.lower() == constants.VALUE_NONE:
10622 nic_dict[constants.INIC_IP] = None
10624 if not netutils.IPAddress.IsValid(nic_ip):
10625 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
10626 errors.ECODE_INVAL)
10628 nic_bridge = nic_dict.get("bridge", None)
10629 nic_link = nic_dict.get(constants.INIC_LINK, None)
10630 if nic_bridge and nic_link:
10631 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
10632 " at the same time", errors.ECODE_INVAL)
10633 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
10634 nic_dict["bridge"] = None
10635 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
10636 nic_dict[constants.INIC_LINK] = None
10638 if nic_op == constants.DDM_ADD:
10639 nic_mac = nic_dict.get(constants.INIC_MAC, None)
10640 if nic_mac is None:
10641 nic_dict[constants.INIC_MAC] = constants.VALUE_AUTO
10643 if constants.INIC_MAC in nic_dict:
10644 nic_mac = nic_dict[constants.INIC_MAC]
10645 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
10646 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
10648 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
10649 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
10650 " modifying an existing nic",
10651 errors.ECODE_INVAL)
10653 if nic_addremove > 1:
10654 raise errors.OpPrereqError("Only one NIC add or remove operation"
10655 " supported at a time", errors.ECODE_INVAL)
10657 def ExpandNames(self):
10658 self._ExpandAndLockInstance()
10659 self.needed_locks[locking.LEVEL_NODE] = []
10660 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10662 def DeclareLocks(self, level):
10663 if level == locking.LEVEL_NODE:
10664 self._LockInstancesNodes()
10665 if self.op.disk_template and self.op.remote_node:
10666 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10667 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
10669 def BuildHooksEnv(self):
10670 """Build hooks env.
10672 This runs on the master, primary and secondaries.
10676 if constants.BE_MEMORY in self.be_new:
10677 args["memory"] = self.be_new[constants.BE_MEMORY]
10678 if constants.BE_VCPUS in self.be_new:
10679 args["vcpus"] = self.be_new[constants.BE_VCPUS]
10680 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
10681 # information at all.
10684 nic_override = dict(self.op.nics)
10685 for idx, nic in enumerate(self.instance.nics):
10686 if idx in nic_override:
10687 this_nic_override = nic_override[idx]
10689 this_nic_override = {}
10690 if constants.INIC_IP in this_nic_override:
10691 ip = this_nic_override[constants.INIC_IP]
10694 if constants.INIC_MAC in this_nic_override:
10695 mac = this_nic_override[constants.INIC_MAC]
10698 if idx in self.nic_pnew:
10699 nicparams = self.nic_pnew[idx]
10701 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
10702 mode = nicparams[constants.NIC_MODE]
10703 link = nicparams[constants.NIC_LINK]
10704 args["nics"].append((ip, mac, mode, link))
10705 if constants.DDM_ADD in nic_override:
10706 ip = nic_override[constants.DDM_ADD].get(constants.INIC_IP, None)
10707 mac = nic_override[constants.DDM_ADD][constants.INIC_MAC]
10708 nicparams = self.nic_pnew[constants.DDM_ADD]
10709 mode = nicparams[constants.NIC_MODE]
10710 link = nicparams[constants.NIC_LINK]
10711 args["nics"].append((ip, mac, mode, link))
10712 elif constants.DDM_REMOVE in nic_override:
10713 del args["nics"][-1]
10715 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
10716 if self.op.disk_template:
10717 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
10721 def BuildHooksNodes(self):
10722 """Build hooks nodes.
10725 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
10728 def CheckPrereq(self):
10729 """Check prerequisites.
10731 This only checks the instance list against the existing names.
10734 # checking the new params on the primary/secondary nodes
10736 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10737 cluster = self.cluster = self.cfg.GetClusterInfo()
10738 assert self.instance is not None, \
10739 "Cannot retrieve locked instance %s" % self.op.instance_name
10740 pnode = instance.primary_node
10741 nodelist = list(instance.all_nodes)
10744 if self.op.os_name and not self.op.force:
10745 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
10746 self.op.force_variant)
10747 instance_os = self.op.os_name
10749 instance_os = instance.os
10751 if self.op.disk_template:
10752 if instance.disk_template == self.op.disk_template:
10753 raise errors.OpPrereqError("Instance already has disk template %s" %
10754 instance.disk_template, errors.ECODE_INVAL)
10756 if (instance.disk_template,
10757 self.op.disk_template) not in self._DISK_CONVERSIONS:
10758 raise errors.OpPrereqError("Unsupported disk template conversion from"
10759 " %s to %s" % (instance.disk_template,
10760 self.op.disk_template),
10761 errors.ECODE_INVAL)
10762 _CheckInstanceDown(self, instance, "cannot change disk template")
10763 if self.op.disk_template in constants.DTS_INT_MIRROR:
10764 if self.op.remote_node == pnode:
10765 raise errors.OpPrereqError("Given new secondary node %s is the same"
10766 " as the primary node of the instance" %
10767 self.op.remote_node, errors.ECODE_STATE)
10768 _CheckNodeOnline(self, self.op.remote_node)
10769 _CheckNodeNotDrained(self, self.op.remote_node)
10770 # FIXME: here we assume that the old instance type is DT_PLAIN
10771 assert instance.disk_template == constants.DT_PLAIN
10772 disks = [{constants.IDISK_SIZE: d.size,
10773 constants.IDISK_VG: d.logical_id[0]}
10774 for d in instance.disks]
10775 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
10776 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
10778 # hvparams processing
10779 if self.op.hvparams:
10780 hv_type = instance.hypervisor
10781 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
10782 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
10783 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
10786 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
10787 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
10788 self.hv_new = hv_new # the new actual values
10789 self.hv_inst = i_hvdict # the new dict (without defaults)
10791 self.hv_new = self.hv_inst = {}
10793 # beparams processing
10794 if self.op.beparams:
10795 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
10797 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
10798 be_new = cluster.SimpleFillBE(i_bedict)
10799 self.be_new = be_new # the new actual values
10800 self.be_inst = i_bedict # the new dict (without defaults)
10802 self.be_new = self.be_inst = {}
10803 be_old = cluster.FillBE(instance)
10805 # osparams processing
10806 if self.op.osparams:
10807 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
10808 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
10809 self.os_inst = i_osdict # the new dict (without defaults)
10815 if (constants.BE_MEMORY in self.op.beparams and not self.op.force and
10816 be_new[constants.BE_MEMORY] > be_old[constants.BE_MEMORY]):
10817 mem_check_list = [pnode]
10818 if be_new[constants.BE_AUTO_BALANCE]:
10819 # either we changed auto_balance to yes or it was from before
10820 mem_check_list.extend(instance.secondary_nodes)
10821 instance_info = self.rpc.call_instance_info(pnode, instance.name,
10822 instance.hypervisor)
10823 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
10824 instance.hypervisor)
10825 pninfo = nodeinfo[pnode]
10826 msg = pninfo.fail_msg
10828 # Assume the primary node is unreachable and go ahead
10829 self.warn.append("Can't get info from primary node %s: %s" %
10831 elif not isinstance(pninfo.payload.get("memory_free", None), int):
10832 self.warn.append("Node data from primary node %s doesn't contain"
10833 " free memory information" % pnode)
10834 elif instance_info.fail_msg:
10835 self.warn.append("Can't get instance runtime information: %s" %
10836 instance_info.fail_msg)
10838 if instance_info.payload:
10839 current_mem = int(instance_info.payload["memory"])
10841 # Assume instance not running
10842 # (there is a slight race condition here, but it's not very probable,
10843 # and we have no other way to check)
10845 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
10846 pninfo.payload["memory_free"])
10848 raise errors.OpPrereqError("This change will prevent the instance"
10849 " from starting, due to %d MB of memory"
10850 " missing on its primary node" % miss_mem,
10851 errors.ECODE_NORES)
10853 if be_new[constants.BE_AUTO_BALANCE]:
10854 for node, nres in nodeinfo.items():
10855 if node not in instance.secondary_nodes:
10857 nres.Raise("Can't get info from secondary node %s" % node,
10858 prereq=True, ecode=errors.ECODE_STATE)
10859 if not isinstance(nres.payload.get("memory_free", None), int):
10860 raise errors.OpPrereqError("Secondary node %s didn't return free"
10861 " memory information" % node,
10862 errors.ECODE_STATE)
10863 elif be_new[constants.BE_MEMORY] > nres.payload["memory_free"]:
10864 raise errors.OpPrereqError("This change will prevent the instance"
10865 " from failover to its secondary node"
10866 " %s, due to not enough memory" % node,
10867 errors.ECODE_STATE)
10871 self.nic_pinst = {}
10872 for nic_op, nic_dict in self.op.nics:
10873 if nic_op == constants.DDM_REMOVE:
10874 if not instance.nics:
10875 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
10876 errors.ECODE_INVAL)
10878 if nic_op != constants.DDM_ADD:
10880 if not instance.nics:
10881 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
10882 " no NICs" % nic_op,
10883 errors.ECODE_INVAL)
10884 if nic_op < 0 or nic_op >= len(instance.nics):
10885 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
10887 (nic_op, len(instance.nics) - 1),
10888 errors.ECODE_INVAL)
10889 old_nic_params = instance.nics[nic_op].nicparams
10890 old_nic_ip = instance.nics[nic_op].ip
10892 old_nic_params = {}
10895 update_params_dict = dict([(key, nic_dict[key])
10896 for key in constants.NICS_PARAMETERS
10897 if key in nic_dict])
10899 if "bridge" in nic_dict:
10900 update_params_dict[constants.NIC_LINK] = nic_dict["bridge"]
10902 new_nic_params = _GetUpdatedParams(old_nic_params,
10903 update_params_dict)
10904 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
10905 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
10906 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
10907 self.nic_pinst[nic_op] = new_nic_params
10908 self.nic_pnew[nic_op] = new_filled_nic_params
10909 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
10911 if new_nic_mode == constants.NIC_MODE_BRIDGED:
10912 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
10913 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
10915 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
10917 self.warn.append(msg)
10919 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
10920 if new_nic_mode == constants.NIC_MODE_ROUTED:
10921 if constants.INIC_IP in nic_dict:
10922 nic_ip = nic_dict[constants.INIC_IP]
10924 nic_ip = old_nic_ip
10926 raise errors.OpPrereqError("Cannot set the nic ip to None"
10927 " on a routed nic", errors.ECODE_INVAL)
10928 if constants.INIC_MAC in nic_dict:
10929 nic_mac = nic_dict[constants.INIC_MAC]
10930 if nic_mac is None:
10931 raise errors.OpPrereqError("Cannot set the nic mac to None",
10932 errors.ECODE_INVAL)
10933 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
10934 # otherwise generate the mac
10935 nic_dict[constants.INIC_MAC] = \
10936 self.cfg.GenerateMAC(self.proc.GetECId())
10938 # or validate/reserve the current one
10940 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
10941 except errors.ReservationError:
10942 raise errors.OpPrereqError("MAC address %s already in use"
10943 " in cluster" % nic_mac,
10944 errors.ECODE_NOTUNIQUE)
10947 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
10948 raise errors.OpPrereqError("Disk operations not supported for"
10949 " diskless instances",
10950 errors.ECODE_INVAL)
10951 for disk_op, _ in self.op.disks:
10952 if disk_op == constants.DDM_REMOVE:
10953 if len(instance.disks) == 1:
10954 raise errors.OpPrereqError("Cannot remove the last disk of"
10955 " an instance", errors.ECODE_INVAL)
10956 _CheckInstanceDown(self, instance, "cannot remove disks")
10958 if (disk_op == constants.DDM_ADD and
10959 len(instance.disks) >= constants.MAX_DISKS):
10960 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
10961 " add more" % constants.MAX_DISKS,
10962 errors.ECODE_STATE)
10963 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
10965 if disk_op < 0 or disk_op >= len(instance.disks):
10966 raise errors.OpPrereqError("Invalid disk index %s, valid values"
10968 (disk_op, len(instance.disks)),
10969 errors.ECODE_INVAL)
10973 def _ConvertPlainToDrbd(self, feedback_fn):
10974 """Converts an instance from plain to drbd.
10977 feedback_fn("Converting template to drbd")
10978 instance = self.instance
10979 pnode = instance.primary_node
10980 snode = self.op.remote_node
10982 # create a fake disk info for _GenerateDiskTemplate
10983 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
10984 constants.IDISK_VG: d.logical_id[0]}
10985 for d in instance.disks]
10986 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
10987 instance.name, pnode, [snode],
10988 disk_info, None, None, 0, feedback_fn)
10989 info = _GetInstanceInfoText(instance)
10990 feedback_fn("Creating aditional volumes...")
10991 # first, create the missing data and meta devices
10992 for disk in new_disks:
10993 # unfortunately this is... not too nice
10994 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
10996 for child in disk.children:
10997 _CreateSingleBlockDev(self, snode, instance, child, info, True)
10998 # at this stage, all new LVs have been created, we can rename the
11000 feedback_fn("Renaming original volumes...")
11001 rename_list = [(o, n.children[0].logical_id)
11002 for (o, n) in zip(instance.disks, new_disks)]
11003 result = self.rpc.call_blockdev_rename(pnode, rename_list)
11004 result.Raise("Failed to rename original LVs")
11006 feedback_fn("Initializing DRBD devices...")
11007 # all child devices are in place, we can now create the DRBD devices
11008 for disk in new_disks:
11009 for node in [pnode, snode]:
11010 f_create = node == pnode
11011 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
11013 # at this point, the instance has been modified
11014 instance.disk_template = constants.DT_DRBD8
11015 instance.disks = new_disks
11016 self.cfg.Update(instance, feedback_fn)
11018 # disks are created, waiting for sync
11019 disk_abort = not _WaitForSync(self, instance,
11020 oneshot=not self.op.wait_for_sync)
11022 raise errors.OpExecError("There are some degraded disks for"
11023 " this instance, please cleanup manually")
11025 def _ConvertDrbdToPlain(self, feedback_fn):
11026 """Converts an instance from drbd to plain.
11029 instance = self.instance
11030 assert len(instance.secondary_nodes) == 1
11031 pnode = instance.primary_node
11032 snode = instance.secondary_nodes[0]
11033 feedback_fn("Converting template to plain")
11035 old_disks = instance.disks
11036 new_disks = [d.children[0] for d in old_disks]
11038 # copy over size and mode
11039 for parent, child in zip(old_disks, new_disks):
11040 child.size = parent.size
11041 child.mode = parent.mode
11043 # update instance structure
11044 instance.disks = new_disks
11045 instance.disk_template = constants.DT_PLAIN
11046 self.cfg.Update(instance, feedback_fn)
11048 feedback_fn("Removing volumes on the secondary node...")
11049 for disk in old_disks:
11050 self.cfg.SetDiskID(disk, snode)
11051 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
11053 self.LogWarning("Could not remove block device %s on node %s,"
11054 " continuing anyway: %s", disk.iv_name, snode, msg)
11056 feedback_fn("Removing unneeded volumes on the primary node...")
11057 for idx, disk in enumerate(old_disks):
11058 meta = disk.children[1]
11059 self.cfg.SetDiskID(meta, pnode)
11060 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
11062 self.LogWarning("Could not remove metadata for disk %d on node %s,"
11063 " continuing anyway: %s", idx, pnode, msg)
11065 def Exec(self, feedback_fn):
11066 """Modifies an instance.
11068 All parameters take effect only at the next restart of the instance.
11071 # Process here the warnings from CheckPrereq, as we don't have a
11072 # feedback_fn there.
11073 for warn in self.warn:
11074 feedback_fn("WARNING: %s" % warn)
11077 instance = self.instance
11079 for disk_op, disk_dict in self.op.disks:
11080 if disk_op == constants.DDM_REMOVE:
11081 # remove the last disk
11082 device = instance.disks.pop()
11083 device_idx = len(instance.disks)
11084 for node, disk in device.ComputeNodeTree(instance.primary_node):
11085 self.cfg.SetDiskID(disk, node)
11086 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
11088 self.LogWarning("Could not remove disk/%d on node %s: %s,"
11089 " continuing anyway", device_idx, node, msg)
11090 result.append(("disk/%d" % device_idx, "remove"))
11091 elif disk_op == constants.DDM_ADD:
11093 if instance.disk_template in (constants.DT_FILE,
11094 constants.DT_SHARED_FILE):
11095 file_driver, file_path = instance.disks[0].logical_id
11096 file_path = os.path.dirname(file_path)
11098 file_driver = file_path = None
11099 disk_idx_base = len(instance.disks)
11100 new_disk = _GenerateDiskTemplate(self,
11101 instance.disk_template,
11102 instance.name, instance.primary_node,
11103 instance.secondary_nodes,
11107 disk_idx_base, feedback_fn)[0]
11108 instance.disks.append(new_disk)
11109 info = _GetInstanceInfoText(instance)
11111 logging.info("Creating volume %s for instance %s",
11112 new_disk.iv_name, instance.name)
11113 # Note: this needs to be kept in sync with _CreateDisks
11115 for node in instance.all_nodes:
11116 f_create = node == instance.primary_node
11118 _CreateBlockDev(self, node, instance, new_disk,
11119 f_create, info, f_create)
11120 except errors.OpExecError, err:
11121 self.LogWarning("Failed to create volume %s (%s) on"
11123 new_disk.iv_name, new_disk, node, err)
11124 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
11125 (new_disk.size, new_disk.mode)))
11127 # change a given disk
11128 instance.disks[disk_op].mode = disk_dict[constants.IDISK_MODE]
11129 result.append(("disk.mode/%d" % disk_op,
11130 disk_dict[constants.IDISK_MODE]))
11132 if self.op.disk_template:
11133 r_shut = _ShutdownInstanceDisks(self, instance)
11135 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
11136 " proceed with disk template conversion")
11137 mode = (instance.disk_template, self.op.disk_template)
11139 self._DISK_CONVERSIONS[mode](self, feedback_fn)
11141 self.cfg.ReleaseDRBDMinors(instance.name)
11143 result.append(("disk_template", self.op.disk_template))
11146 for nic_op, nic_dict in self.op.nics:
11147 if nic_op == constants.DDM_REMOVE:
11148 # remove the last nic
11149 del instance.nics[-1]
11150 result.append(("nic.%d" % len(instance.nics), "remove"))
11151 elif nic_op == constants.DDM_ADD:
11152 # mac and bridge should be set, by now
11153 mac = nic_dict[constants.INIC_MAC]
11154 ip = nic_dict.get(constants.INIC_IP, None)
11155 nicparams = self.nic_pinst[constants.DDM_ADD]
11156 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
11157 instance.nics.append(new_nic)
11158 result.append(("nic.%d" % (len(instance.nics) - 1),
11159 "add:mac=%s,ip=%s,mode=%s,link=%s" %
11160 (new_nic.mac, new_nic.ip,
11161 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
11162 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
11165 for key in (constants.INIC_MAC, constants.INIC_IP):
11166 if key in nic_dict:
11167 setattr(instance.nics[nic_op], key, nic_dict[key])
11168 if nic_op in self.nic_pinst:
11169 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
11170 for key, val in nic_dict.iteritems():
11171 result.append(("nic.%s/%d" % (key, nic_op), val))
11174 if self.op.hvparams:
11175 instance.hvparams = self.hv_inst
11176 for key, val in self.op.hvparams.iteritems():
11177 result.append(("hv/%s" % key, val))
11180 if self.op.beparams:
11181 instance.beparams = self.be_inst
11182 for key, val in self.op.beparams.iteritems():
11183 result.append(("be/%s" % key, val))
11186 if self.op.os_name:
11187 instance.os = self.op.os_name
11190 if self.op.osparams:
11191 instance.osparams = self.os_inst
11192 for key, val in self.op.osparams.iteritems():
11193 result.append(("os/%s" % key, val))
11195 self.cfg.Update(instance, feedback_fn)
11199 _DISK_CONVERSIONS = {
11200 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
11201 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
11205 class LUInstanceChangeGroup(LogicalUnit):
11206 HPATH = "instance-change-group"
11207 HTYPE = constants.HTYPE_INSTANCE
11210 def ExpandNames(self):
11211 self.share_locks = _ShareAll()
11212 self.needed_locks = {
11213 locking.LEVEL_NODEGROUP: [],
11214 locking.LEVEL_NODE: [],
11217 self._ExpandAndLockInstance()
11219 if self.op.target_groups:
11220 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
11221 self.op.target_groups)
11223 self.req_target_uuids = None
11225 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
11227 def DeclareLocks(self, level):
11228 if level == locking.LEVEL_NODEGROUP:
11229 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
11231 if self.req_target_uuids:
11232 lock_groups = set(self.req_target_uuids)
11234 # Lock all groups used by instance optimistically; this requires going
11235 # via the node before it's locked, requiring verification later on
11236 instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
11237 lock_groups.update(instance_groups)
11239 # No target groups, need to lock all of them
11240 lock_groups = locking.ALL_SET
11242 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
11244 elif level == locking.LEVEL_NODE:
11245 if self.req_target_uuids:
11246 # Lock all nodes used by instances
11247 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
11248 self._LockInstancesNodes()
11250 # Lock all nodes in all potential target groups
11251 lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) -
11252 self.cfg.GetInstanceNodeGroups(self.op.instance_name))
11253 member_nodes = [node_name
11254 for group in lock_groups
11255 for node_name in self.cfg.GetNodeGroup(group).members]
11256 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
11258 # Lock all nodes as all groups are potential targets
11259 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11261 def CheckPrereq(self):
11262 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
11263 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
11264 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
11266 assert (self.req_target_uuids is None or
11267 owned_groups.issuperset(self.req_target_uuids))
11268 assert owned_instances == set([self.op.instance_name])
11270 # Get instance information
11271 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
11273 # Check if node groups for locked instance are still correct
11274 assert owned_nodes.issuperset(self.instance.all_nodes), \
11275 ("Instance %s's nodes changed while we kept the lock" %
11276 self.op.instance_name)
11278 inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name,
11281 if self.req_target_uuids:
11282 # User requested specific target groups
11283 self.target_uuids = self.req_target_uuids
11285 # All groups except those used by the instance are potential targets
11286 self.target_uuids = owned_groups - inst_groups
11288 conflicting_groups = self.target_uuids & inst_groups
11289 if conflicting_groups:
11290 raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are"
11291 " used by the instance '%s'" %
11292 (utils.CommaJoin(conflicting_groups),
11293 self.op.instance_name),
11294 errors.ECODE_INVAL)
11296 if not self.target_uuids:
11297 raise errors.OpPrereqError("There are no possible target groups",
11298 errors.ECODE_INVAL)
11300 def BuildHooksEnv(self):
11301 """Build hooks env.
11304 assert self.target_uuids
11307 "TARGET_GROUPS": " ".join(self.target_uuids),
11310 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11314 def BuildHooksNodes(self):
11315 """Build hooks nodes.
11318 mn = self.cfg.GetMasterNode()
11319 return ([mn], [mn])
11321 def Exec(self, feedback_fn):
11322 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
11324 assert instances == [self.op.instance_name], "Instance not locked"
11326 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
11327 instances=instances, target_groups=list(self.target_uuids))
11329 ial.Run(self.op.iallocator)
11331 if not ial.success:
11332 raise errors.OpPrereqError("Can't compute solution for changing group of"
11333 " instance '%s' using iallocator '%s': %s" %
11334 (self.op.instance_name, self.op.iallocator,
11336 errors.ECODE_NORES)
11338 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
11340 self.LogInfo("Iallocator returned %s job(s) for changing group of"
11341 " instance '%s'", len(jobs), self.op.instance_name)
11343 return ResultWithJobs(jobs)
11346 class LUBackupQuery(NoHooksLU):
11347 """Query the exports list
11352 def ExpandNames(self):
11353 self.needed_locks = {}
11354 self.share_locks[locking.LEVEL_NODE] = 1
11355 if not self.op.nodes:
11356 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11358 self.needed_locks[locking.LEVEL_NODE] = \
11359 _GetWantedNodes(self, self.op.nodes)
11361 def Exec(self, feedback_fn):
11362 """Compute the list of all the exported system images.
11365 @return: a dictionary with the structure node->(export-list)
11366 where export-list is a list of the instances exported on
11370 self.nodes = self.owned_locks(locking.LEVEL_NODE)
11371 rpcresult = self.rpc.call_export_list(self.nodes)
11373 for node in rpcresult:
11374 if rpcresult[node].fail_msg:
11375 result[node] = False
11377 result[node] = rpcresult[node].payload
11382 class LUBackupPrepare(NoHooksLU):
11383 """Prepares an instance for an export and returns useful information.
11388 def ExpandNames(self):
11389 self._ExpandAndLockInstance()
11391 def CheckPrereq(self):
11392 """Check prerequisites.
11395 instance_name = self.op.instance_name
11397 self.instance = self.cfg.GetInstanceInfo(instance_name)
11398 assert self.instance is not None, \
11399 "Cannot retrieve locked instance %s" % self.op.instance_name
11400 _CheckNodeOnline(self, self.instance.primary_node)
11402 self._cds = _GetClusterDomainSecret()
11404 def Exec(self, feedback_fn):
11405 """Prepares an instance for an export.
11408 instance = self.instance
11410 if self.op.mode == constants.EXPORT_MODE_REMOTE:
11411 salt = utils.GenerateSecret(8)
11413 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
11414 result = self.rpc.call_x509_cert_create(instance.primary_node,
11415 constants.RIE_CERT_VALIDITY)
11416 result.Raise("Can't create X509 key and certificate on %s" % result.node)
11418 (name, cert_pem) = result.payload
11420 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
11424 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
11425 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
11427 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
11433 class LUBackupExport(LogicalUnit):
11434 """Export an instance to an image in the cluster.
11437 HPATH = "instance-export"
11438 HTYPE = constants.HTYPE_INSTANCE
11441 def CheckArguments(self):
11442 """Check the arguments.
11445 self.x509_key_name = self.op.x509_key_name
11446 self.dest_x509_ca_pem = self.op.destination_x509_ca
11448 if self.op.mode == constants.EXPORT_MODE_REMOTE:
11449 if not self.x509_key_name:
11450 raise errors.OpPrereqError("Missing X509 key name for encryption",
11451 errors.ECODE_INVAL)
11453 if not self.dest_x509_ca_pem:
11454 raise errors.OpPrereqError("Missing destination X509 CA",
11455 errors.ECODE_INVAL)
11457 def ExpandNames(self):
11458 self._ExpandAndLockInstance()
11460 # Lock all nodes for local exports
11461 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11462 # FIXME: lock only instance primary and destination node
11464 # Sad but true, for now we have do lock all nodes, as we don't know where
11465 # the previous export might be, and in this LU we search for it and
11466 # remove it from its current node. In the future we could fix this by:
11467 # - making a tasklet to search (share-lock all), then create the
11468 # new one, then one to remove, after
11469 # - removing the removal operation altogether
11470 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11472 def DeclareLocks(self, level):
11473 """Last minute lock declaration."""
11474 # All nodes are locked anyway, so nothing to do here.
11476 def BuildHooksEnv(self):
11477 """Build hooks env.
11479 This will run on the master, primary node and target node.
11483 "EXPORT_MODE": self.op.mode,
11484 "EXPORT_NODE": self.op.target_node,
11485 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
11486 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
11487 # TODO: Generic function for boolean env variables
11488 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
11491 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11495 def BuildHooksNodes(self):
11496 """Build hooks nodes.
11499 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
11501 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11502 nl.append(self.op.target_node)
11506 def CheckPrereq(self):
11507 """Check prerequisites.
11509 This checks that the instance and node names are valid.
11512 instance_name = self.op.instance_name
11514 self.instance = self.cfg.GetInstanceInfo(instance_name)
11515 assert self.instance is not None, \
11516 "Cannot retrieve locked instance %s" % self.op.instance_name
11517 _CheckNodeOnline(self, self.instance.primary_node)
11519 if (self.op.remove_instance and self.instance.admin_up and
11520 not self.op.shutdown):
11521 raise errors.OpPrereqError("Can not remove instance without shutting it"
11524 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11525 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
11526 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
11527 assert self.dst_node is not None
11529 _CheckNodeOnline(self, self.dst_node.name)
11530 _CheckNodeNotDrained(self, self.dst_node.name)
11533 self.dest_disk_info = None
11534 self.dest_x509_ca = None
11536 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
11537 self.dst_node = None
11539 if len(self.op.target_node) != len(self.instance.disks):
11540 raise errors.OpPrereqError(("Received destination information for %s"
11541 " disks, but instance %s has %s disks") %
11542 (len(self.op.target_node), instance_name,
11543 len(self.instance.disks)),
11544 errors.ECODE_INVAL)
11546 cds = _GetClusterDomainSecret()
11548 # Check X509 key name
11550 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
11551 except (TypeError, ValueError), err:
11552 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
11554 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
11555 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
11556 errors.ECODE_INVAL)
11558 # Load and verify CA
11560 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
11561 except OpenSSL.crypto.Error, err:
11562 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
11563 (err, ), errors.ECODE_INVAL)
11565 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
11566 if errcode is not None:
11567 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
11568 (msg, ), errors.ECODE_INVAL)
11570 self.dest_x509_ca = cert
11572 # Verify target information
11574 for idx, disk_data in enumerate(self.op.target_node):
11576 (host, port, magic) = \
11577 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
11578 except errors.GenericError, err:
11579 raise errors.OpPrereqError("Target info for disk %s: %s" %
11580 (idx, err), errors.ECODE_INVAL)
11582 disk_info.append((host, port, magic))
11584 assert len(disk_info) == len(self.op.target_node)
11585 self.dest_disk_info = disk_info
11588 raise errors.ProgrammerError("Unhandled export mode %r" %
11591 # instance disk type verification
11592 # TODO: Implement export support for file-based disks
11593 for disk in self.instance.disks:
11594 if disk.dev_type == constants.LD_FILE:
11595 raise errors.OpPrereqError("Export not supported for instances with"
11596 " file-based disks", errors.ECODE_INVAL)
11598 def _CleanupExports(self, feedback_fn):
11599 """Removes exports of current instance from all other nodes.
11601 If an instance in a cluster with nodes A..D was exported to node C, its
11602 exports will be removed from the nodes A, B and D.
11605 assert self.op.mode != constants.EXPORT_MODE_REMOTE
11607 nodelist = self.cfg.GetNodeList()
11608 nodelist.remove(self.dst_node.name)
11610 # on one-node clusters nodelist will be empty after the removal
11611 # if we proceed the backup would be removed because OpBackupQuery
11612 # substitutes an empty list with the full cluster node list.
11613 iname = self.instance.name
11615 feedback_fn("Removing old exports for instance %s" % iname)
11616 exportlist = self.rpc.call_export_list(nodelist)
11617 for node in exportlist:
11618 if exportlist[node].fail_msg:
11620 if iname in exportlist[node].payload:
11621 msg = self.rpc.call_export_remove(node, iname).fail_msg
11623 self.LogWarning("Could not remove older export for instance %s"
11624 " on node %s: %s", iname, node, msg)
11626 def Exec(self, feedback_fn):
11627 """Export an instance to an image in the cluster.
11630 assert self.op.mode in constants.EXPORT_MODES
11632 instance = self.instance
11633 src_node = instance.primary_node
11635 if self.op.shutdown:
11636 # shutdown the instance, but not the disks
11637 feedback_fn("Shutting down instance %s" % instance.name)
11638 result = self.rpc.call_instance_shutdown(src_node, instance,
11639 self.op.shutdown_timeout)
11640 # TODO: Maybe ignore failures if ignore_remove_failures is set
11641 result.Raise("Could not shutdown instance %s on"
11642 " node %s" % (instance.name, src_node))
11644 # set the disks ID correctly since call_instance_start needs the
11645 # correct drbd minor to create the symlinks
11646 for disk in instance.disks:
11647 self.cfg.SetDiskID(disk, src_node)
11649 activate_disks = (not instance.admin_up)
11652 # Activate the instance disks if we'exporting a stopped instance
11653 feedback_fn("Activating disks for %s" % instance.name)
11654 _StartInstanceDisks(self, instance, None)
11657 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
11660 helper.CreateSnapshots()
11662 if (self.op.shutdown and instance.admin_up and
11663 not self.op.remove_instance):
11664 assert not activate_disks
11665 feedback_fn("Starting instance %s" % instance.name)
11666 result = self.rpc.call_instance_start(src_node, instance,
11668 msg = result.fail_msg
11670 feedback_fn("Failed to start instance: %s" % msg)
11671 _ShutdownInstanceDisks(self, instance)
11672 raise errors.OpExecError("Could not start instance: %s" % msg)
11674 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11675 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
11676 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
11677 connect_timeout = constants.RIE_CONNECT_TIMEOUT
11678 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
11680 (key_name, _, _) = self.x509_key_name
11683 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
11686 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
11687 key_name, dest_ca_pem,
11692 # Check for backwards compatibility
11693 assert len(dresults) == len(instance.disks)
11694 assert compat.all(isinstance(i, bool) for i in dresults), \
11695 "Not all results are boolean: %r" % dresults
11699 feedback_fn("Deactivating disks for %s" % instance.name)
11700 _ShutdownInstanceDisks(self, instance)
11702 if not (compat.all(dresults) and fin_resu):
11705 failures.append("export finalization")
11706 if not compat.all(dresults):
11707 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
11709 failures.append("disk export: disk(s) %s" % fdsk)
11711 raise errors.OpExecError("Export failed, errors in %s" %
11712 utils.CommaJoin(failures))
11714 # At this point, the export was successful, we can cleanup/finish
11716 # Remove instance if requested
11717 if self.op.remove_instance:
11718 feedback_fn("Removing instance %s" % instance.name)
11719 _RemoveInstance(self, feedback_fn, instance,
11720 self.op.ignore_remove_failures)
11722 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11723 self._CleanupExports(feedback_fn)
11725 return fin_resu, dresults
11728 class LUBackupRemove(NoHooksLU):
11729 """Remove exports related to the named instance.
11734 def ExpandNames(self):
11735 self.needed_locks = {}
11736 # We need all nodes to be locked in order for RemoveExport to work, but we
11737 # don't need to lock the instance itself, as nothing will happen to it (and
11738 # we can remove exports also for a removed instance)
11739 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11741 def Exec(self, feedback_fn):
11742 """Remove any export.
11745 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
11746 # If the instance was not found we'll try with the name that was passed in.
11747 # This will only work if it was an FQDN, though.
11749 if not instance_name:
11751 instance_name = self.op.instance_name
11753 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
11754 exportlist = self.rpc.call_export_list(locked_nodes)
11756 for node in exportlist:
11757 msg = exportlist[node].fail_msg
11759 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
11761 if instance_name in exportlist[node].payload:
11763 result = self.rpc.call_export_remove(node, instance_name)
11764 msg = result.fail_msg
11766 logging.error("Could not remove export for instance %s"
11767 " on node %s: %s", instance_name, node, msg)
11769 if fqdn_warn and not found:
11770 feedback_fn("Export not found. If trying to remove an export belonging"
11771 " to a deleted instance please use its Fully Qualified"
11775 class LUGroupAdd(LogicalUnit):
11776 """Logical unit for creating node groups.
11779 HPATH = "group-add"
11780 HTYPE = constants.HTYPE_GROUP
11783 def ExpandNames(self):
11784 # We need the new group's UUID here so that we can create and acquire the
11785 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
11786 # that it should not check whether the UUID exists in the configuration.
11787 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
11788 self.needed_locks = {}
11789 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
11791 def CheckPrereq(self):
11792 """Check prerequisites.
11794 This checks that the given group name is not an existing node group
11799 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11800 except errors.OpPrereqError:
11803 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
11804 " node group (UUID: %s)" %
11805 (self.op.group_name, existing_uuid),
11806 errors.ECODE_EXISTS)
11808 if self.op.ndparams:
11809 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
11811 def BuildHooksEnv(self):
11812 """Build hooks env.
11816 "GROUP_NAME": self.op.group_name,
11819 def BuildHooksNodes(self):
11820 """Build hooks nodes.
11823 mn = self.cfg.GetMasterNode()
11824 return ([mn], [mn])
11826 def Exec(self, feedback_fn):
11827 """Add the node group to the cluster.
11830 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
11831 uuid=self.group_uuid,
11832 alloc_policy=self.op.alloc_policy,
11833 ndparams=self.op.ndparams)
11835 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
11836 del self.remove_locks[locking.LEVEL_NODEGROUP]
11839 class LUGroupAssignNodes(NoHooksLU):
11840 """Logical unit for assigning nodes to groups.
11845 def ExpandNames(self):
11846 # These raise errors.OpPrereqError on their own:
11847 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11848 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
11850 # We want to lock all the affected nodes and groups. We have readily
11851 # available the list of nodes, and the *destination* group. To gather the
11852 # list of "source" groups, we need to fetch node information later on.
11853 self.needed_locks = {
11854 locking.LEVEL_NODEGROUP: set([self.group_uuid]),
11855 locking.LEVEL_NODE: self.op.nodes,
11858 def DeclareLocks(self, level):
11859 if level == locking.LEVEL_NODEGROUP:
11860 assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1
11862 # Try to get all affected nodes' groups without having the group or node
11863 # lock yet. Needs verification later in the code flow.
11864 groups = self.cfg.GetNodeGroupsFromNodes(self.op.nodes)
11866 self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)
11868 def CheckPrereq(self):
11869 """Check prerequisites.
11872 assert self.needed_locks[locking.LEVEL_NODEGROUP]
11873 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
11874 frozenset(self.op.nodes))
11876 expected_locks = (set([self.group_uuid]) |
11877 self.cfg.GetNodeGroupsFromNodes(self.op.nodes))
11878 actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP)
11879 if actual_locks != expected_locks:
11880 raise errors.OpExecError("Nodes changed groups since locks were acquired,"
11881 " current groups are '%s', used to be '%s'" %
11882 (utils.CommaJoin(expected_locks),
11883 utils.CommaJoin(actual_locks)))
11885 self.node_data = self.cfg.GetAllNodesInfo()
11886 self.group = self.cfg.GetNodeGroup(self.group_uuid)
11887 instance_data = self.cfg.GetAllInstancesInfo()
11889 if self.group is None:
11890 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
11891 (self.op.group_name, self.group_uuid))
11893 (new_splits, previous_splits) = \
11894 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
11895 for node in self.op.nodes],
11896 self.node_data, instance_data)
11899 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
11901 if not self.op.force:
11902 raise errors.OpExecError("The following instances get split by this"
11903 " change and --force was not given: %s" %
11906 self.LogWarning("This operation will split the following instances: %s",
11909 if previous_splits:
11910 self.LogWarning("In addition, these already-split instances continue"
11911 " to be split across groups: %s",
11912 utils.CommaJoin(utils.NiceSort(previous_splits)))
11914 def Exec(self, feedback_fn):
11915 """Assign nodes to a new group.
11918 for node in self.op.nodes:
11919 self.node_data[node].group = self.group_uuid
11921 # FIXME: Depends on side-effects of modifying the result of
11922 # C{cfg.GetAllNodesInfo}
11924 self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
11927 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
11928 """Check for split instances after a node assignment.
11930 This method considers a series of node assignments as an atomic operation,
11931 and returns information about split instances after applying the set of
11934 In particular, it returns information about newly split instances, and
11935 instances that were already split, and remain so after the change.
11937 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
11940 @type changes: list of (node_name, new_group_uuid) pairs.
11941 @param changes: list of node assignments to consider.
11942 @param node_data: a dict with data for all nodes
11943 @param instance_data: a dict with all instances to consider
11944 @rtype: a two-tuple
11945 @return: a list of instances that were previously okay and result split as a
11946 consequence of this change, and a list of instances that were previously
11947 split and this change does not fix.
11950 changed_nodes = dict((node, group) for node, group in changes
11951 if node_data[node].group != group)
11953 all_split_instances = set()
11954 previously_split_instances = set()
11956 def InstanceNodes(instance):
11957 return [instance.primary_node] + list(instance.secondary_nodes)
11959 for inst in instance_data.values():
11960 if inst.disk_template not in constants.DTS_INT_MIRROR:
11963 instance_nodes = InstanceNodes(inst)
11965 if len(set(node_data[node].group for node in instance_nodes)) > 1:
11966 previously_split_instances.add(inst.name)
11968 if len(set(changed_nodes.get(node, node_data[node].group)
11969 for node in instance_nodes)) > 1:
11970 all_split_instances.add(inst.name)
11972 return (list(all_split_instances - previously_split_instances),
11973 list(previously_split_instances & all_split_instances))
11976 class _GroupQuery(_QueryBase):
11977 FIELDS = query.GROUP_FIELDS
11979 def ExpandNames(self, lu):
11980 lu.needed_locks = {}
11982 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
11983 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
11986 self.wanted = [name_to_uuid[name]
11987 for name in utils.NiceSort(name_to_uuid.keys())]
11989 # Accept names to be either names or UUIDs.
11992 all_uuid = frozenset(self._all_groups.keys())
11994 for name in self.names:
11995 if name in all_uuid:
11996 self.wanted.append(name)
11997 elif name in name_to_uuid:
11998 self.wanted.append(name_to_uuid[name])
12000 missing.append(name)
12003 raise errors.OpPrereqError("Some groups do not exist: %s" %
12004 utils.CommaJoin(missing),
12005 errors.ECODE_NOENT)
12007 def DeclareLocks(self, lu, level):
12010 def _GetQueryData(self, lu):
12011 """Computes the list of node groups and their attributes.
12014 do_nodes = query.GQ_NODE in self.requested_data
12015 do_instances = query.GQ_INST in self.requested_data
12017 group_to_nodes = None
12018 group_to_instances = None
12020 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
12021 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
12022 # latter GetAllInstancesInfo() is not enough, for we have to go through
12023 # instance->node. Hence, we will need to process nodes even if we only need
12024 # instance information.
12025 if do_nodes or do_instances:
12026 all_nodes = lu.cfg.GetAllNodesInfo()
12027 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
12030 for node in all_nodes.values():
12031 if node.group in group_to_nodes:
12032 group_to_nodes[node.group].append(node.name)
12033 node_to_group[node.name] = node.group
12036 all_instances = lu.cfg.GetAllInstancesInfo()
12037 group_to_instances = dict((uuid, []) for uuid in self.wanted)
12039 for instance in all_instances.values():
12040 node = instance.primary_node
12041 if node in node_to_group:
12042 group_to_instances[node_to_group[node]].append(instance.name)
12045 # Do not pass on node information if it was not requested.
12046 group_to_nodes = None
12048 return query.GroupQueryData([self._all_groups[uuid]
12049 for uuid in self.wanted],
12050 group_to_nodes, group_to_instances)
12053 class LUGroupQuery(NoHooksLU):
12054 """Logical unit for querying node groups.
12059 def CheckArguments(self):
12060 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
12061 self.op.output_fields, False)
12063 def ExpandNames(self):
12064 self.gq.ExpandNames(self)
12066 def DeclareLocks(self, level):
12067 self.gq.DeclareLocks(self, level)
12069 def Exec(self, feedback_fn):
12070 return self.gq.OldStyleQuery(self)
12073 class LUGroupSetParams(LogicalUnit):
12074 """Modifies the parameters of a node group.
12077 HPATH = "group-modify"
12078 HTYPE = constants.HTYPE_GROUP
12081 def CheckArguments(self):
12084 self.op.alloc_policy,
12087 if all_changes.count(None) == len(all_changes):
12088 raise errors.OpPrereqError("Please pass at least one modification",
12089 errors.ECODE_INVAL)
12091 def ExpandNames(self):
12092 # This raises errors.OpPrereqError on its own:
12093 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12095 self.needed_locks = {
12096 locking.LEVEL_NODEGROUP: [self.group_uuid],
12099 def CheckPrereq(self):
12100 """Check prerequisites.
12103 self.group = self.cfg.GetNodeGroup(self.group_uuid)
12105 if self.group is None:
12106 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
12107 (self.op.group_name, self.group_uuid))
12109 if self.op.ndparams:
12110 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
12111 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
12112 self.new_ndparams = new_ndparams
12114 def BuildHooksEnv(self):
12115 """Build hooks env.
12119 "GROUP_NAME": self.op.group_name,
12120 "NEW_ALLOC_POLICY": self.op.alloc_policy,
12123 def BuildHooksNodes(self):
12124 """Build hooks nodes.
12127 mn = self.cfg.GetMasterNode()
12128 return ([mn], [mn])
12130 def Exec(self, feedback_fn):
12131 """Modifies the node group.
12136 if self.op.ndparams:
12137 self.group.ndparams = self.new_ndparams
12138 result.append(("ndparams", str(self.group.ndparams)))
12140 if self.op.alloc_policy:
12141 self.group.alloc_policy = self.op.alloc_policy
12143 self.cfg.Update(self.group, feedback_fn)
12147 class LUGroupRemove(LogicalUnit):
12148 HPATH = "group-remove"
12149 HTYPE = constants.HTYPE_GROUP
12152 def ExpandNames(self):
12153 # This will raises errors.OpPrereqError on its own:
12154 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12155 self.needed_locks = {
12156 locking.LEVEL_NODEGROUP: [self.group_uuid],
12159 def CheckPrereq(self):
12160 """Check prerequisites.
12162 This checks that the given group name exists as a node group, that is
12163 empty (i.e., contains no nodes), and that is not the last group of the
12167 # Verify that the group is empty.
12168 group_nodes = [node.name
12169 for node in self.cfg.GetAllNodesInfo().values()
12170 if node.group == self.group_uuid]
12173 raise errors.OpPrereqError("Group '%s' not empty, has the following"
12175 (self.op.group_name,
12176 utils.CommaJoin(utils.NiceSort(group_nodes))),
12177 errors.ECODE_STATE)
12179 # Verify the cluster would not be left group-less.
12180 if len(self.cfg.GetNodeGroupList()) == 1:
12181 raise errors.OpPrereqError("Group '%s' is the only group,"
12182 " cannot be removed" %
12183 self.op.group_name,
12184 errors.ECODE_STATE)
12186 def BuildHooksEnv(self):
12187 """Build hooks env.
12191 "GROUP_NAME": self.op.group_name,
12194 def BuildHooksNodes(self):
12195 """Build hooks nodes.
12198 mn = self.cfg.GetMasterNode()
12199 return ([mn], [mn])
12201 def Exec(self, feedback_fn):
12202 """Remove the node group.
12206 self.cfg.RemoveNodeGroup(self.group_uuid)
12207 except errors.ConfigurationError:
12208 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
12209 (self.op.group_name, self.group_uuid))
12211 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
12214 class LUGroupRename(LogicalUnit):
12215 HPATH = "group-rename"
12216 HTYPE = constants.HTYPE_GROUP
12219 def ExpandNames(self):
12220 # This raises errors.OpPrereqError on its own:
12221 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12223 self.needed_locks = {
12224 locking.LEVEL_NODEGROUP: [self.group_uuid],
12227 def CheckPrereq(self):
12228 """Check prerequisites.
12230 Ensures requested new name is not yet used.
12234 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
12235 except errors.OpPrereqError:
12238 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
12239 " node group (UUID: %s)" %
12240 (self.op.new_name, new_name_uuid),
12241 errors.ECODE_EXISTS)
12243 def BuildHooksEnv(self):
12244 """Build hooks env.
12248 "OLD_NAME": self.op.group_name,
12249 "NEW_NAME": self.op.new_name,
12252 def BuildHooksNodes(self):
12253 """Build hooks nodes.
12256 mn = self.cfg.GetMasterNode()
12258 all_nodes = self.cfg.GetAllNodesInfo()
12259 all_nodes.pop(mn, None)
12262 run_nodes.extend(node.name for node in all_nodes.values()
12263 if node.group == self.group_uuid)
12265 return (run_nodes, run_nodes)
12267 def Exec(self, feedback_fn):
12268 """Rename the node group.
12271 group = self.cfg.GetNodeGroup(self.group_uuid)
12274 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
12275 (self.op.group_name, self.group_uuid))
12277 group.name = self.op.new_name
12278 self.cfg.Update(group, feedback_fn)
12280 return self.op.new_name
12283 class LUGroupEvacuate(LogicalUnit):
12284 HPATH = "group-evacuate"
12285 HTYPE = constants.HTYPE_GROUP
12288 def ExpandNames(self):
12289 # This raises errors.OpPrereqError on its own:
12290 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12292 if self.op.target_groups:
12293 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
12294 self.op.target_groups)
12296 self.req_target_uuids = []
12298 if self.group_uuid in self.req_target_uuids:
12299 raise errors.OpPrereqError("Group to be evacuated (%s) can not be used"
12300 " as a target group (targets are %s)" %
12302 utils.CommaJoin(self.req_target_uuids)),
12303 errors.ECODE_INVAL)
12305 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
12307 self.share_locks = _ShareAll()
12308 self.needed_locks = {
12309 locking.LEVEL_INSTANCE: [],
12310 locking.LEVEL_NODEGROUP: [],
12311 locking.LEVEL_NODE: [],
12314 def DeclareLocks(self, level):
12315 if level == locking.LEVEL_INSTANCE:
12316 assert not self.needed_locks[locking.LEVEL_INSTANCE]
12318 # Lock instances optimistically, needs verification once node and group
12319 # locks have been acquired
12320 self.needed_locks[locking.LEVEL_INSTANCE] = \
12321 self.cfg.GetNodeGroupInstances(self.group_uuid)
12323 elif level == locking.LEVEL_NODEGROUP:
12324 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
12326 if self.req_target_uuids:
12327 lock_groups = set([self.group_uuid] + self.req_target_uuids)
12329 # Lock all groups used by instances optimistically; this requires going
12330 # via the node before it's locked, requiring verification later on
12331 lock_groups.update(group_uuid
12332 for instance_name in
12333 self.owned_locks(locking.LEVEL_INSTANCE)
12335 self.cfg.GetInstanceNodeGroups(instance_name))
12337 # No target groups, need to lock all of them
12338 lock_groups = locking.ALL_SET
12340 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
12342 elif level == locking.LEVEL_NODE:
12343 # This will only lock the nodes in the group to be evacuated which
12344 # contain actual instances
12345 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
12346 self._LockInstancesNodes()
12348 # Lock all nodes in group to be evacuated and target groups
12349 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12350 assert self.group_uuid in owned_groups
12351 member_nodes = [node_name
12352 for group in owned_groups
12353 for node_name in self.cfg.GetNodeGroup(group).members]
12354 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
12356 def CheckPrereq(self):
12357 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
12358 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12359 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
12361 assert owned_groups.issuperset(self.req_target_uuids)
12362 assert self.group_uuid in owned_groups
12364 # Check if locked instances are still correct
12365 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
12367 # Get instance information
12368 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
12370 # Check if node groups for locked instances are still correct
12371 for instance_name in owned_instances:
12372 inst = self.instances[instance_name]
12373 assert owned_nodes.issuperset(inst.all_nodes), \
12374 "Instance %s's nodes changed while we kept the lock" % instance_name
12376 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
12379 assert self.group_uuid in inst_groups, \
12380 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
12382 if self.req_target_uuids:
12383 # User requested specific target groups
12384 self.target_uuids = self.req_target_uuids
12386 # All groups except the one to be evacuated are potential targets
12387 self.target_uuids = [group_uuid for group_uuid in owned_groups
12388 if group_uuid != self.group_uuid]
12390 if not self.target_uuids:
12391 raise errors.OpPrereqError("There are no possible target groups",
12392 errors.ECODE_INVAL)
12394 def BuildHooksEnv(self):
12395 """Build hooks env.
12399 "GROUP_NAME": self.op.group_name,
12400 "TARGET_GROUPS": " ".join(self.target_uuids),
12403 def BuildHooksNodes(self):
12404 """Build hooks nodes.
12407 mn = self.cfg.GetMasterNode()
12409 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
12411 run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members
12413 return (run_nodes, run_nodes)
12415 def Exec(self, feedback_fn):
12416 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
12418 assert self.group_uuid not in self.target_uuids
12420 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
12421 instances=instances, target_groups=self.target_uuids)
12423 ial.Run(self.op.iallocator)
12425 if not ial.success:
12426 raise errors.OpPrereqError("Can't compute group evacuation using"
12427 " iallocator '%s': %s" %
12428 (self.op.iallocator, ial.info),
12429 errors.ECODE_NORES)
12431 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
12433 self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s",
12434 len(jobs), self.op.group_name)
12436 return ResultWithJobs(jobs)
12439 class TagsLU(NoHooksLU): # pylint: disable=W0223
12440 """Generic tags LU.
12442 This is an abstract class which is the parent of all the other tags LUs.
12445 def ExpandNames(self):
12446 self.group_uuid = None
12447 self.needed_locks = {}
12448 if self.op.kind == constants.TAG_NODE:
12449 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
12450 self.needed_locks[locking.LEVEL_NODE] = self.op.name
12451 elif self.op.kind == constants.TAG_INSTANCE:
12452 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
12453 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
12454 elif self.op.kind == constants.TAG_NODEGROUP:
12455 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
12457 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
12458 # not possible to acquire the BGL based on opcode parameters)
12460 def CheckPrereq(self):
12461 """Check prerequisites.
12464 if self.op.kind == constants.TAG_CLUSTER:
12465 self.target = self.cfg.GetClusterInfo()
12466 elif self.op.kind == constants.TAG_NODE:
12467 self.target = self.cfg.GetNodeInfo(self.op.name)
12468 elif self.op.kind == constants.TAG_INSTANCE:
12469 self.target = self.cfg.GetInstanceInfo(self.op.name)
12470 elif self.op.kind == constants.TAG_NODEGROUP:
12471 self.target = self.cfg.GetNodeGroup(self.group_uuid)
12473 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
12474 str(self.op.kind), errors.ECODE_INVAL)
12477 class LUTagsGet(TagsLU):
12478 """Returns the tags of a given object.
12483 def ExpandNames(self):
12484 TagsLU.ExpandNames(self)
12486 # Share locks as this is only a read operation
12487 self.share_locks = _ShareAll()
12489 def Exec(self, feedback_fn):
12490 """Returns the tag list.
12493 return list(self.target.GetTags())
12496 class LUTagsSearch(NoHooksLU):
12497 """Searches the tags for a given pattern.
12502 def ExpandNames(self):
12503 self.needed_locks = {}
12505 def CheckPrereq(self):
12506 """Check prerequisites.
12508 This checks the pattern passed for validity by compiling it.
12512 self.re = re.compile(self.op.pattern)
12513 except re.error, err:
12514 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
12515 (self.op.pattern, err), errors.ECODE_INVAL)
12517 def Exec(self, feedback_fn):
12518 """Returns the tag list.
12522 tgts = [("/cluster", cfg.GetClusterInfo())]
12523 ilist = cfg.GetAllInstancesInfo().values()
12524 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
12525 nlist = cfg.GetAllNodesInfo().values()
12526 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
12527 tgts.extend(("/nodegroup/%s" % n.name, n)
12528 for n in cfg.GetAllNodeGroupsInfo().values())
12530 for path, target in tgts:
12531 for tag in target.GetTags():
12532 if self.re.search(tag):
12533 results.append((path, tag))
12537 class LUTagsSet(TagsLU):
12538 """Sets a tag on a given object.
12543 def CheckPrereq(self):
12544 """Check prerequisites.
12546 This checks the type and length of the tag name and value.
12549 TagsLU.CheckPrereq(self)
12550 for tag in self.op.tags:
12551 objects.TaggableObject.ValidateTag(tag)
12553 def Exec(self, feedback_fn):
12558 for tag in self.op.tags:
12559 self.target.AddTag(tag)
12560 except errors.TagError, err:
12561 raise errors.OpExecError("Error while setting tag: %s" % str(err))
12562 self.cfg.Update(self.target, feedback_fn)
12565 class LUTagsDel(TagsLU):
12566 """Delete a list of tags from a given object.
12571 def CheckPrereq(self):
12572 """Check prerequisites.
12574 This checks that we have the given tag.
12577 TagsLU.CheckPrereq(self)
12578 for tag in self.op.tags:
12579 objects.TaggableObject.ValidateTag(tag)
12580 del_tags = frozenset(self.op.tags)
12581 cur_tags = self.target.GetTags()
12583 diff_tags = del_tags - cur_tags
12585 diff_names = ("'%s'" % i for i in sorted(diff_tags))
12586 raise errors.OpPrereqError("Tag(s) %s not found" %
12587 (utils.CommaJoin(diff_names), ),
12588 errors.ECODE_NOENT)
12590 def Exec(self, feedback_fn):
12591 """Remove the tag from the object.
12594 for tag in self.op.tags:
12595 self.target.RemoveTag(tag)
12596 self.cfg.Update(self.target, feedback_fn)
12599 class LUTestDelay(NoHooksLU):
12600 """Sleep for a specified amount of time.
12602 This LU sleeps on the master and/or nodes for a specified amount of
12608 def ExpandNames(self):
12609 """Expand names and set required locks.
12611 This expands the node list, if any.
12614 self.needed_locks = {}
12615 if self.op.on_nodes:
12616 # _GetWantedNodes can be used here, but is not always appropriate to use
12617 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
12618 # more information.
12619 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
12620 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
12622 def _TestDelay(self):
12623 """Do the actual sleep.
12626 if self.op.on_master:
12627 if not utils.TestDelay(self.op.duration):
12628 raise errors.OpExecError("Error during master delay test")
12629 if self.op.on_nodes:
12630 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
12631 for node, node_result in result.items():
12632 node_result.Raise("Failure during rpc call to node %s" % node)
12634 def Exec(self, feedback_fn):
12635 """Execute the test delay opcode, with the wanted repetitions.
12638 if self.op.repeat == 0:
12641 top_value = self.op.repeat - 1
12642 for i in range(self.op.repeat):
12643 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
12647 class LUTestJqueue(NoHooksLU):
12648 """Utility LU to test some aspects of the job queue.
12653 # Must be lower than default timeout for WaitForJobChange to see whether it
12654 # notices changed jobs
12655 _CLIENT_CONNECT_TIMEOUT = 20.0
12656 _CLIENT_CONFIRM_TIMEOUT = 60.0
12659 def _NotifyUsingSocket(cls, cb, errcls):
12660 """Opens a Unix socket and waits for another program to connect.
12663 @param cb: Callback to send socket name to client
12664 @type errcls: class
12665 @param errcls: Exception class to use for errors
12668 # Using a temporary directory as there's no easy way to create temporary
12669 # sockets without writing a custom loop around tempfile.mktemp and
12671 tmpdir = tempfile.mkdtemp()
12673 tmpsock = utils.PathJoin(tmpdir, "sock")
12675 logging.debug("Creating temporary socket at %s", tmpsock)
12676 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
12681 # Send details to client
12684 # Wait for client to connect before continuing
12685 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
12687 (conn, _) = sock.accept()
12688 except socket.error, err:
12689 raise errcls("Client didn't connect in time (%s)" % err)
12693 # Remove as soon as client is connected
12694 shutil.rmtree(tmpdir)
12696 # Wait for client to close
12699 # pylint: disable=E1101
12700 # Instance of '_socketobject' has no ... member
12701 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
12703 except socket.error, err:
12704 raise errcls("Client failed to confirm notification (%s)" % err)
12708 def _SendNotification(self, test, arg, sockname):
12709 """Sends a notification to the client.
12712 @param test: Test name
12713 @param arg: Test argument (depends on test)
12714 @type sockname: string
12715 @param sockname: Socket path
12718 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
12720 def _Notify(self, prereq, test, arg):
12721 """Notifies the client of a test.
12724 @param prereq: Whether this is a prereq-phase test
12726 @param test: Test name
12727 @param arg: Test argument (depends on test)
12731 errcls = errors.OpPrereqError
12733 errcls = errors.OpExecError
12735 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
12739 def CheckArguments(self):
12740 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
12741 self.expandnames_calls = 0
12743 def ExpandNames(self):
12744 checkargs_calls = getattr(self, "checkargs_calls", 0)
12745 if checkargs_calls < 1:
12746 raise errors.ProgrammerError("CheckArguments was not called")
12748 self.expandnames_calls += 1
12750 if self.op.notify_waitlock:
12751 self._Notify(True, constants.JQT_EXPANDNAMES, None)
12753 self.LogInfo("Expanding names")
12755 # Get lock on master node (just to get a lock, not for a particular reason)
12756 self.needed_locks = {
12757 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
12760 def Exec(self, feedback_fn):
12761 if self.expandnames_calls < 1:
12762 raise errors.ProgrammerError("ExpandNames was not called")
12764 if self.op.notify_exec:
12765 self._Notify(False, constants.JQT_EXEC, None)
12767 self.LogInfo("Executing")
12769 if self.op.log_messages:
12770 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
12771 for idx, msg in enumerate(self.op.log_messages):
12772 self.LogInfo("Sending log message %s", idx + 1)
12773 feedback_fn(constants.JQT_MSGPREFIX + msg)
12774 # Report how many test messages have been sent
12775 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
12778 raise errors.OpExecError("Opcode failure was requested")
12783 class IAllocator(object):
12784 """IAllocator framework.
12786 An IAllocator instance has three sets of attributes:
12787 - cfg that is needed to query the cluster
12788 - input data (all members of the _KEYS class attribute are required)
12789 - four buffer attributes (in|out_data|text), that represent the
12790 input (to the external script) in text and data structure format,
12791 and the output from it, again in two formats
12792 - the result variables from the script (success, info, nodes) for
12796 # pylint: disable=R0902
12797 # lots of instance attributes
12799 def __init__(self, cfg, rpc, mode, **kwargs):
12802 # init buffer variables
12803 self.in_text = self.out_text = self.in_data = self.out_data = None
12804 # init all input fields so that pylint is happy
12806 self.memory = self.disks = self.disk_template = None
12807 self.os = self.tags = self.nics = self.vcpus = None
12808 self.hypervisor = None
12809 self.relocate_from = None
12811 self.instances = None
12812 self.evac_mode = None
12813 self.target_groups = []
12815 self.required_nodes = None
12816 # init result fields
12817 self.success = self.info = self.result = None
12820 (fn, keydata, self._result_check) = self._MODE_DATA[self.mode]
12822 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
12823 " IAllocator" % self.mode)
12825 keyset = [n for (n, _) in keydata]
12828 if key not in keyset:
12829 raise errors.ProgrammerError("Invalid input parameter '%s' to"
12830 " IAllocator" % key)
12831 setattr(self, key, kwargs[key])
12834 if key not in kwargs:
12835 raise errors.ProgrammerError("Missing input parameter '%s' to"
12836 " IAllocator" % key)
12837 self._BuildInputData(compat.partial(fn, self), keydata)
12839 def _ComputeClusterData(self):
12840 """Compute the generic allocator input data.
12842 This is the data that is independent of the actual operation.
12846 cluster_info = cfg.GetClusterInfo()
12849 "version": constants.IALLOCATOR_VERSION,
12850 "cluster_name": cfg.GetClusterName(),
12851 "cluster_tags": list(cluster_info.GetTags()),
12852 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
12853 # we don't have job IDs
12855 ninfo = cfg.GetAllNodesInfo()
12856 iinfo = cfg.GetAllInstancesInfo().values()
12857 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
12860 node_list = [n.name for n in ninfo.values() if n.vm_capable]
12862 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
12863 hypervisor_name = self.hypervisor
12864 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
12865 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
12867 hypervisor_name = cluster_info.enabled_hypervisors[0]
12869 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
12872 self.rpc.call_all_instances_info(node_list,
12873 cluster_info.enabled_hypervisors)
12875 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
12877 config_ndata = self._ComputeBasicNodeData(ninfo)
12878 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
12879 i_list, config_ndata)
12880 assert len(data["nodes"]) == len(ninfo), \
12881 "Incomplete node data computed"
12883 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
12885 self.in_data = data
12888 def _ComputeNodeGroupData(cfg):
12889 """Compute node groups data.
12892 ng = dict((guuid, {
12893 "name": gdata.name,
12894 "alloc_policy": gdata.alloc_policy,
12896 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items())
12901 def _ComputeBasicNodeData(node_cfg):
12902 """Compute global node data.
12905 @returns: a dict of name: (node dict, node config)
12908 # fill in static (config-based) values
12909 node_results = dict((ninfo.name, {
12910 "tags": list(ninfo.GetTags()),
12911 "primary_ip": ninfo.primary_ip,
12912 "secondary_ip": ninfo.secondary_ip,
12913 "offline": ninfo.offline,
12914 "drained": ninfo.drained,
12915 "master_candidate": ninfo.master_candidate,
12916 "group": ninfo.group,
12917 "master_capable": ninfo.master_capable,
12918 "vm_capable": ninfo.vm_capable,
12920 for ninfo in node_cfg.values())
12922 return node_results
12925 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
12927 """Compute global node data.
12929 @param node_results: the basic node structures as filled from the config
12932 # make a copy of the current dict
12933 node_results = dict(node_results)
12934 for nname, nresult in node_data.items():
12935 assert nname in node_results, "Missing basic data for node %s" % nname
12936 ninfo = node_cfg[nname]
12938 if not (ninfo.offline or ninfo.drained):
12939 nresult.Raise("Can't get data for node %s" % nname)
12940 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
12942 remote_info = nresult.payload
12944 for attr in ["memory_total", "memory_free", "memory_dom0",
12945 "vg_size", "vg_free", "cpu_total"]:
12946 if attr not in remote_info:
12947 raise errors.OpExecError("Node '%s' didn't return attribute"
12948 " '%s'" % (nname, attr))
12949 if not isinstance(remote_info[attr], int):
12950 raise errors.OpExecError("Node '%s' returned invalid value"
12952 (nname, attr, remote_info[attr]))
12953 # compute memory used by primary instances
12954 i_p_mem = i_p_up_mem = 0
12955 for iinfo, beinfo in i_list:
12956 if iinfo.primary_node == nname:
12957 i_p_mem += beinfo[constants.BE_MEMORY]
12958 if iinfo.name not in node_iinfo[nname].payload:
12961 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]["memory"])
12962 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
12963 remote_info["memory_free"] -= max(0, i_mem_diff)
12966 i_p_up_mem += beinfo[constants.BE_MEMORY]
12968 # compute memory used by instances
12970 "total_memory": remote_info["memory_total"],
12971 "reserved_memory": remote_info["memory_dom0"],
12972 "free_memory": remote_info["memory_free"],
12973 "total_disk": remote_info["vg_size"],
12974 "free_disk": remote_info["vg_free"],
12975 "total_cpus": remote_info["cpu_total"],
12976 "i_pri_memory": i_p_mem,
12977 "i_pri_up_memory": i_p_up_mem,
12979 pnr_dyn.update(node_results[nname])
12980 node_results[nname] = pnr_dyn
12982 return node_results
12985 def _ComputeInstanceData(cluster_info, i_list):
12986 """Compute global instance data.
12990 for iinfo, beinfo in i_list:
12992 for nic in iinfo.nics:
12993 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
12997 "mode": filled_params[constants.NIC_MODE],
12998 "link": filled_params[constants.NIC_LINK],
13000 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
13001 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
13002 nic_data.append(nic_dict)
13004 "tags": list(iinfo.GetTags()),
13005 "admin_up": iinfo.admin_up,
13006 "vcpus": beinfo[constants.BE_VCPUS],
13007 "memory": beinfo[constants.BE_MEMORY],
13009 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
13011 "disks": [{constants.IDISK_SIZE: dsk.size,
13012 constants.IDISK_MODE: dsk.mode}
13013 for dsk in iinfo.disks],
13014 "disk_template": iinfo.disk_template,
13015 "hypervisor": iinfo.hypervisor,
13017 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
13019 instance_data[iinfo.name] = pir
13021 return instance_data
13023 def _AddNewInstance(self):
13024 """Add new instance data to allocator structure.
13026 This in combination with _AllocatorGetClusterData will create the
13027 correct structure needed as input for the allocator.
13029 The checks for the completeness of the opcode must have already been
13033 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
13035 if self.disk_template in constants.DTS_INT_MIRROR:
13036 self.required_nodes = 2
13038 self.required_nodes = 1
13042 "disk_template": self.disk_template,
13045 "vcpus": self.vcpus,
13046 "memory": self.memory,
13047 "disks": self.disks,
13048 "disk_space_total": disk_space,
13050 "required_nodes": self.required_nodes,
13051 "hypervisor": self.hypervisor,
13056 def _AddRelocateInstance(self):
13057 """Add relocate instance data to allocator structure.
13059 This in combination with _IAllocatorGetClusterData will create the
13060 correct structure needed as input for the allocator.
13062 The checks for the completeness of the opcode must have already been
13066 instance = self.cfg.GetInstanceInfo(self.name)
13067 if instance is None:
13068 raise errors.ProgrammerError("Unknown instance '%s' passed to"
13069 " IAllocator" % self.name)
13071 if instance.disk_template not in constants.DTS_MIRRORED:
13072 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
13073 errors.ECODE_INVAL)
13075 if instance.disk_template in constants.DTS_INT_MIRROR and \
13076 len(instance.secondary_nodes) != 1:
13077 raise errors.OpPrereqError("Instance has not exactly one secondary node",
13078 errors.ECODE_STATE)
13080 self.required_nodes = 1
13081 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
13082 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
13086 "disk_space_total": disk_space,
13087 "required_nodes": self.required_nodes,
13088 "relocate_from": self.relocate_from,
13092 def _AddNodeEvacuate(self):
13093 """Get data for node-evacuate requests.
13097 "instances": self.instances,
13098 "evac_mode": self.evac_mode,
13101 def _AddChangeGroup(self):
13102 """Get data for node-evacuate requests.
13106 "instances": self.instances,
13107 "target_groups": self.target_groups,
13110 def _BuildInputData(self, fn, keydata):
13111 """Build input data structures.
13114 self._ComputeClusterData()
13117 request["type"] = self.mode
13118 for keyname, keytype in keydata:
13119 if keyname not in request:
13120 raise errors.ProgrammerError("Request parameter %s is missing" %
13122 val = request[keyname]
13123 if not keytype(val):
13124 raise errors.ProgrammerError("Request parameter %s doesn't pass"
13125 " validation, value %s, expected"
13126 " type %s" % (keyname, val, keytype))
13127 self.in_data["request"] = request
13129 self.in_text = serializer.Dump(self.in_data)
13131 _STRING_LIST = ht.TListOf(ht.TString)
13132 _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
13133 # pylint: disable=E1101
13134 # Class '...' has no 'OP_ID' member
13135 "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
13136 opcodes.OpInstanceMigrate.OP_ID,
13137 opcodes.OpInstanceReplaceDisks.OP_ID])
13141 ht.TListOf(ht.TAnd(ht.TIsLength(3),
13142 ht.TItems([ht.TNonEmptyString,
13143 ht.TNonEmptyString,
13144 ht.TListOf(ht.TNonEmptyString),
13147 ht.TListOf(ht.TAnd(ht.TIsLength(2),
13148 ht.TItems([ht.TNonEmptyString,
13151 _NEVAC_RESULT = ht.TAnd(ht.TIsLength(3),
13152 ht.TItems([_NEVAC_MOVED, _NEVAC_FAILED, _JOB_LIST]))
13155 constants.IALLOCATOR_MODE_ALLOC:
13158 ("name", ht.TString),
13159 ("memory", ht.TInt),
13160 ("disks", ht.TListOf(ht.TDict)),
13161 ("disk_template", ht.TString),
13162 ("os", ht.TString),
13163 ("tags", _STRING_LIST),
13164 ("nics", ht.TListOf(ht.TDict)),
13165 ("vcpus", ht.TInt),
13166 ("hypervisor", ht.TString),
13168 constants.IALLOCATOR_MODE_RELOC:
13169 (_AddRelocateInstance,
13170 [("name", ht.TString), ("relocate_from", _STRING_LIST)],
13172 constants.IALLOCATOR_MODE_NODE_EVAC:
13173 (_AddNodeEvacuate, [
13174 ("instances", _STRING_LIST),
13175 ("evac_mode", ht.TElemOf(constants.IALLOCATOR_NEVAC_MODES)),
13177 constants.IALLOCATOR_MODE_CHG_GROUP:
13178 (_AddChangeGroup, [
13179 ("instances", _STRING_LIST),
13180 ("target_groups", _STRING_LIST),
13184 def Run(self, name, validate=True, call_fn=None):
13185 """Run an instance allocator and return the results.
13188 if call_fn is None:
13189 call_fn = self.rpc.call_iallocator_runner
13191 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
13192 result.Raise("Failure while running the iallocator script")
13194 self.out_text = result.payload
13196 self._ValidateResult()
13198 def _ValidateResult(self):
13199 """Process the allocator results.
13201 This will process and if successful save the result in
13202 self.out_data and the other parameters.
13206 rdict = serializer.Load(self.out_text)
13207 except Exception, err:
13208 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
13210 if not isinstance(rdict, dict):
13211 raise errors.OpExecError("Can't parse iallocator results: not a dict")
13213 # TODO: remove backwards compatiblity in later versions
13214 if "nodes" in rdict and "result" not in rdict:
13215 rdict["result"] = rdict["nodes"]
13218 for key in "success", "info", "result":
13219 if key not in rdict:
13220 raise errors.OpExecError("Can't parse iallocator results:"
13221 " missing key '%s'" % key)
13222 setattr(self, key, rdict[key])
13224 if not self._result_check(self.result):
13225 raise errors.OpExecError("Iallocator returned invalid result,"
13226 " expected %s, got %s" %
13227 (self._result_check, self.result),
13228 errors.ECODE_INVAL)
13230 if self.mode == constants.IALLOCATOR_MODE_RELOC:
13231 assert self.relocate_from is not None
13232 assert self.required_nodes == 1
13234 node2group = dict((name, ndata["group"])
13235 for (name, ndata) in self.in_data["nodes"].items())
13237 fn = compat.partial(self._NodesToGroups, node2group,
13238 self.in_data["nodegroups"])
13240 instance = self.cfg.GetInstanceInfo(self.name)
13241 request_groups = fn(self.relocate_from + [instance.primary_node])
13242 result_groups = fn(rdict["result"] + [instance.primary_node])
13244 if self.success and not set(result_groups).issubset(request_groups):
13245 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
13246 " differ from original groups (%s)" %
13247 (utils.CommaJoin(result_groups),
13248 utils.CommaJoin(request_groups)))
13250 elif self.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
13251 assert self.evac_mode in constants.IALLOCATOR_NEVAC_MODES
13253 self.out_data = rdict
13256 def _NodesToGroups(node2group, groups, nodes):
13257 """Returns a list of unique group names for a list of nodes.
13259 @type node2group: dict
13260 @param node2group: Map from node name to group UUID
13262 @param groups: Group information
13264 @param nodes: Node names
13271 group_uuid = node2group[node]
13273 # Ignore unknown node
13277 group = groups[group_uuid]
13279 # Can't find group, let's use UUID
13280 group_name = group_uuid
13282 group_name = group["name"]
13284 result.add(group_name)
13286 return sorted(result)
13289 class LUTestAllocator(NoHooksLU):
13290 """Run allocator tests.
13292 This LU runs the allocator tests
13295 def CheckPrereq(self):
13296 """Check prerequisites.
13298 This checks the opcode parameters depending on the director and mode test.
13301 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
13302 for attr in ["memory", "disks", "disk_template",
13303 "os", "tags", "nics", "vcpus"]:
13304 if not hasattr(self.op, attr):
13305 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
13306 attr, errors.ECODE_INVAL)
13307 iname = self.cfg.ExpandInstanceName(self.op.name)
13308 if iname is not None:
13309 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
13310 iname, errors.ECODE_EXISTS)
13311 if not isinstance(self.op.nics, list):
13312 raise errors.OpPrereqError("Invalid parameter 'nics'",
13313 errors.ECODE_INVAL)
13314 if not isinstance(self.op.disks, list):
13315 raise errors.OpPrereqError("Invalid parameter 'disks'",
13316 errors.ECODE_INVAL)
13317 for row in self.op.disks:
13318 if (not isinstance(row, dict) or
13319 constants.IDISK_SIZE not in row or
13320 not isinstance(row[constants.IDISK_SIZE], int) or
13321 constants.IDISK_MODE not in row or
13322 row[constants.IDISK_MODE] not in constants.DISK_ACCESS_SET):
13323 raise errors.OpPrereqError("Invalid contents of the 'disks'"
13324 " parameter", errors.ECODE_INVAL)
13325 if self.op.hypervisor is None:
13326 self.op.hypervisor = self.cfg.GetHypervisorType()
13327 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
13328 fname = _ExpandInstanceName(self.cfg, self.op.name)
13329 self.op.name = fname
13330 self.relocate_from = \
13331 list(self.cfg.GetInstanceInfo(fname).secondary_nodes)
13332 elif self.op.mode in (constants.IALLOCATOR_MODE_CHG_GROUP,
13333 constants.IALLOCATOR_MODE_NODE_EVAC):
13334 if not self.op.instances:
13335 raise errors.OpPrereqError("Missing instances", errors.ECODE_INVAL)
13336 self.op.instances = _GetWantedInstances(self, self.op.instances)
13338 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
13339 self.op.mode, errors.ECODE_INVAL)
13341 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
13342 if self.op.allocator is None:
13343 raise errors.OpPrereqError("Missing allocator name",
13344 errors.ECODE_INVAL)
13345 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
13346 raise errors.OpPrereqError("Wrong allocator test '%s'" %
13347 self.op.direction, errors.ECODE_INVAL)
13349 def Exec(self, feedback_fn):
13350 """Run the allocator test.
13353 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
13354 ial = IAllocator(self.cfg, self.rpc,
13357 memory=self.op.memory,
13358 disks=self.op.disks,
13359 disk_template=self.op.disk_template,
13363 vcpus=self.op.vcpus,
13364 hypervisor=self.op.hypervisor,
13366 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
13367 ial = IAllocator(self.cfg, self.rpc,
13370 relocate_from=list(self.relocate_from),
13372 elif self.op.mode == constants.IALLOCATOR_MODE_CHG_GROUP:
13373 ial = IAllocator(self.cfg, self.rpc,
13375 instances=self.op.instances,
13376 target_groups=self.op.target_groups)
13377 elif self.op.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
13378 ial = IAllocator(self.cfg, self.rpc,
13380 instances=self.op.instances,
13381 evac_mode=self.op.evac_mode)
13383 raise errors.ProgrammerError("Uncatched mode %s in"
13384 " LUTestAllocator.Exec", self.op.mode)
13386 if self.op.direction == constants.IALLOCATOR_DIR_IN:
13387 result = ial.in_text
13389 ial.Run(self.op.allocator, validate=False)
13390 result = ial.out_text
13394 #: Query type implementations
13396 constants.QR_INSTANCE: _InstanceQuery,
13397 constants.QR_NODE: _NodeQuery,
13398 constants.QR_GROUP: _GroupQuery,
13399 constants.QR_OS: _OsQuery,
13402 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
13405 def _GetQueryImplementation(name):
13406 """Returns the implemtnation for a query type.
13408 @param name: Query type, must be one of L{constants.QR_VIA_OP}
13412 return _QUERY_IMPL[name]
13414 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
13415 errors.ECODE_INVAL)