4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012 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 too 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
62 from ganeti import rpc
64 import ganeti.masterd.instance # pylint: disable=W0611
67 #: Size of DRBD meta block device
71 INSTANCE_UP = [constants.ADMINST_UP]
72 INSTANCE_DOWN = [constants.ADMINST_DOWN]
73 INSTANCE_OFFLINE = [constants.ADMINST_OFFLINE]
74 INSTANCE_ONLINE = [constants.ADMINST_DOWN, constants.ADMINST_UP]
75 INSTANCE_NOT_RUNNING = [constants.ADMINST_DOWN, constants.ADMINST_OFFLINE]
79 """Data container for LU results with jobs.
81 Instances of this class returned from L{LogicalUnit.Exec} will be recognized
82 by L{mcpu.Processor._ProcessResult}. The latter will then submit the jobs
83 contained in the C{jobs} attribute and include the job IDs in the opcode
87 def __init__(self, jobs, **kwargs):
88 """Initializes this class.
90 Additional return values can be specified as keyword arguments.
92 @type jobs: list of lists of L{opcode.OpCode}
93 @param jobs: A list of lists of opcode objects
100 class LogicalUnit(object):
101 """Logical Unit base class.
103 Subclasses must follow these rules:
104 - implement ExpandNames
105 - implement CheckPrereq (except when tasklets are used)
106 - implement Exec (except when tasklets are used)
107 - implement BuildHooksEnv
108 - implement BuildHooksNodes
109 - redefine HPATH and HTYPE
110 - optionally redefine their run requirements:
111 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
113 Note that all commands require root permissions.
115 @ivar dry_run_result: the value (if any) that will be returned to the caller
116 in dry-run mode (signalled by opcode dry_run parameter)
123 def __init__(self, processor, op, context, rpc_runner):
124 """Constructor for LogicalUnit.
126 This needs to be overridden in derived classes in order to check op
130 self.proc = processor
132 self.cfg = context.cfg
133 self.glm = context.glm
135 self.owned_locks = context.glm.list_owned
136 self.context = context
137 self.rpc = rpc_runner
138 # Dicts used to declare locking needs to mcpu
139 self.needed_locks = None
140 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
142 self.remove_locks = {}
143 # Used to force good behavior when calling helper functions
144 self.recalculate_locks = {}
146 self.Log = processor.Log # pylint: disable=C0103
147 self.LogWarning = processor.LogWarning # pylint: disable=C0103
148 self.LogInfo = processor.LogInfo # pylint: disable=C0103
149 self.LogStep = processor.LogStep # pylint: disable=C0103
150 # support for dry-run
151 self.dry_run_result = None
152 # support for generic debug attribute
153 if (not hasattr(self.op, "debug_level") or
154 not isinstance(self.op.debug_level, int)):
155 self.op.debug_level = 0
160 # Validate opcode parameters and set defaults
161 self.op.Validate(True)
163 self.CheckArguments()
165 def CheckArguments(self):
166 """Check syntactic validity for the opcode arguments.
168 This method is for doing a simple syntactic check and ensure
169 validity of opcode parameters, without any cluster-related
170 checks. While the same can be accomplished in ExpandNames and/or
171 CheckPrereq, doing these separate is better because:
173 - ExpandNames is left as as purely a lock-related function
174 - CheckPrereq is run after we have acquired locks (and possible
177 The function is allowed to change the self.op attribute so that
178 later methods can no longer worry about missing parameters.
183 def ExpandNames(self):
184 """Expand names for this LU.
186 This method is called before starting to execute the opcode, and it should
187 update all the parameters of the opcode to their canonical form (e.g. a
188 short node name must be fully expanded after this method has successfully
189 completed). This way locking, hooks, logging, etc. can work correctly.
191 LUs which implement this method must also populate the self.needed_locks
192 member, as a dict with lock levels as keys, and a list of needed lock names
195 - use an empty dict if you don't need any lock
196 - if you don't need any lock at a particular level omit that level
197 - don't put anything for the BGL level
198 - if you want all locks at a level use locking.ALL_SET as a value
200 If you need to share locks (rather than acquire them exclusively) at one
201 level you can modify self.share_locks, setting a true value (usually 1) for
202 that level. By default locks are not shared.
204 This function can also define a list of tasklets, which then will be
205 executed in order instead of the usual LU-level CheckPrereq and Exec
206 functions, if those are not defined by the LU.
210 # Acquire all nodes and one instance
211 self.needed_locks = {
212 locking.LEVEL_NODE: locking.ALL_SET,
213 locking.LEVEL_INSTANCE: ['instance1.example.com'],
215 # Acquire just two nodes
216 self.needed_locks = {
217 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
220 self.needed_locks = {} # No, you can't leave it to the default value None
223 # The implementation of this method is mandatory only if the new LU is
224 # concurrent, so that old LUs don't need to be changed all at the same
227 self.needed_locks = {} # Exclusive LUs don't need locks.
229 raise NotImplementedError
231 def DeclareLocks(self, level):
232 """Declare LU locking needs for a level
234 While most LUs can just declare their locking needs at ExpandNames time,
235 sometimes there's the need to calculate some locks after having acquired
236 the ones before. This function is called just before acquiring locks at a
237 particular level, but after acquiring the ones at lower levels, and permits
238 such calculations. It can be used to modify self.needed_locks, and by
239 default it does nothing.
241 This function is only called if you have something already set in
242 self.needed_locks for the level.
244 @param level: Locking level which is going to be locked
245 @type level: member of ganeti.locking.LEVELS
249 def CheckPrereq(self):
250 """Check prerequisites for this LU.
252 This method should check that the prerequisites for the execution
253 of this LU are fulfilled. It can do internode communication, but
254 it should be idempotent - no cluster or system changes are
257 The method should raise errors.OpPrereqError in case something is
258 not fulfilled. Its return value is ignored.
260 This method should also update all the parameters of the opcode to
261 their canonical form if it hasn't been done by ExpandNames before.
264 if self.tasklets is not None:
265 for (idx, tl) in enumerate(self.tasklets):
266 logging.debug("Checking prerequisites for tasklet %s/%s",
267 idx + 1, len(self.tasklets))
272 def Exec(self, feedback_fn):
275 This method should implement the actual work. It should raise
276 errors.OpExecError for failures that are somewhat dealt with in
280 if self.tasklets is not None:
281 for (idx, tl) in enumerate(self.tasklets):
282 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
285 raise NotImplementedError
287 def BuildHooksEnv(self):
288 """Build hooks environment for this LU.
291 @return: Dictionary containing the environment that will be used for
292 running the hooks for this LU. The keys of the dict must not be prefixed
293 with "GANETI_"--that'll be added by the hooks runner. The hooks runner
294 will extend the environment with additional variables. If no environment
295 should be defined, an empty dictionary should be returned (not C{None}).
296 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
300 raise NotImplementedError
302 def BuildHooksNodes(self):
303 """Build list of nodes to run LU's hooks.
305 @rtype: tuple; (list, list)
306 @return: Tuple containing a list of node names on which the hook
307 should run before the execution and a list of node names on which the
308 hook should run after the execution. No nodes should be returned as an
309 empty list (and not None).
310 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
314 raise NotImplementedError
316 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
317 """Notify the LU about the results of its hooks.
319 This method is called every time a hooks phase is executed, and notifies
320 the Logical Unit about the hooks' result. The LU can then use it to alter
321 its result based on the hooks. By default the method does nothing and the
322 previous result is passed back unchanged but any LU can define it if it
323 wants to use the local cluster hook-scripts somehow.
325 @param phase: one of L{constants.HOOKS_PHASE_POST} or
326 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
327 @param hook_results: the results of the multi-node hooks rpc call
328 @param feedback_fn: function used send feedback back to the caller
329 @param lu_result: the previous Exec result this LU had, or None
331 @return: the new Exec result, based on the previous result
335 # API must be kept, thus we ignore the unused argument and could
336 # be a function warnings
337 # pylint: disable=W0613,R0201
340 def _ExpandAndLockInstance(self):
341 """Helper function to expand and lock an instance.
343 Many LUs that work on an instance take its name in self.op.instance_name
344 and need to expand it and then declare the expanded name for locking. This
345 function does it, and then updates self.op.instance_name to the expanded
346 name. It also initializes needed_locks as a dict, if this hasn't been done
350 if self.needed_locks is None:
351 self.needed_locks = {}
353 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
354 "_ExpandAndLockInstance called with instance-level locks set"
355 self.op.instance_name = _ExpandInstanceName(self.cfg,
356 self.op.instance_name)
357 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
359 def _LockInstancesNodes(self, primary_only=False,
360 level=locking.LEVEL_NODE):
361 """Helper function to declare instances' nodes for locking.
363 This function should be called after locking one or more instances to lock
364 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
365 with all primary or secondary nodes for instances already locked and
366 present in self.needed_locks[locking.LEVEL_INSTANCE].
368 It should be called from DeclareLocks, and for safety only works if
369 self.recalculate_locks[locking.LEVEL_NODE] is set.
371 In the future it may grow parameters to just lock some instance's nodes, or
372 to just lock primaries or secondary nodes, if needed.
374 If should be called in DeclareLocks in a way similar to::
376 if level == locking.LEVEL_NODE:
377 self._LockInstancesNodes()
379 @type primary_only: boolean
380 @param primary_only: only lock primary nodes of locked instances
381 @param level: Which lock level to use for locking nodes
384 assert level in self.recalculate_locks, \
385 "_LockInstancesNodes helper function called with no nodes to recalculate"
387 # TODO: check if we're really been called with the instance locks held
389 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
390 # future we might want to have different behaviors depending on the value
391 # of self.recalculate_locks[locking.LEVEL_NODE]
393 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
394 for _, instance in self.cfg.GetMultiInstanceInfo(locked_i):
395 wanted_nodes.append(instance.primary_node)
397 wanted_nodes.extend(instance.secondary_nodes)
399 if self.recalculate_locks[level] == constants.LOCKS_REPLACE:
400 self.needed_locks[level] = wanted_nodes
401 elif self.recalculate_locks[level] == constants.LOCKS_APPEND:
402 self.needed_locks[level].extend(wanted_nodes)
404 raise errors.ProgrammerError("Unknown recalculation mode")
406 del self.recalculate_locks[level]
409 class NoHooksLU(LogicalUnit): # pylint: disable=W0223
410 """Simple LU which runs no hooks.
412 This LU is intended as a parent for other LogicalUnits which will
413 run no hooks, in order to reduce duplicate code.
419 def BuildHooksEnv(self):
420 """Empty BuildHooksEnv for NoHooksLu.
422 This just raises an error.
425 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
427 def BuildHooksNodes(self):
428 """Empty BuildHooksNodes for NoHooksLU.
431 raise AssertionError("BuildHooksNodes called for NoHooksLU")
435 """Tasklet base class.
437 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
438 they can mix legacy code with tasklets. Locking needs to be done in the LU,
439 tasklets know nothing about locks.
441 Subclasses must follow these rules:
442 - Implement CheckPrereq
446 def __init__(self, lu):
453 def CheckPrereq(self):
454 """Check prerequisites for this tasklets.
456 This method should check whether the prerequisites for the execution of
457 this tasklet are fulfilled. It can do internode communication, but it
458 should be idempotent - no cluster or system changes are allowed.
460 The method should raise errors.OpPrereqError in case something is not
461 fulfilled. Its return value is ignored.
463 This method should also update all parameters to their canonical form if it
464 hasn't been done before.
469 def Exec(self, feedback_fn):
470 """Execute the tasklet.
472 This method should implement the actual work. It should raise
473 errors.OpExecError for failures that are somewhat dealt with in code, or
477 raise NotImplementedError
481 """Base for query utility classes.
484 #: Attribute holding field definitions
487 def __init__(self, qfilter, fields, use_locking):
488 """Initializes this class.
491 self.use_locking = use_locking
493 self.query = query.Query(self.FIELDS, fields, qfilter=qfilter,
495 self.requested_data = self.query.RequestedData()
496 self.names = self.query.RequestedNames()
498 # Sort only if no names were requested
499 self.sort_by_name = not self.names
501 self.do_locking = None
504 def _GetNames(self, lu, all_names, lock_level):
505 """Helper function to determine names asked for in the query.
509 names = lu.owned_locks(lock_level)
513 if self.wanted == locking.ALL_SET:
514 assert not self.names
515 # caller didn't specify names, so ordering is not important
516 return utils.NiceSort(names)
518 # caller specified names and we must keep the same order
520 assert not self.do_locking or lu.glm.is_owned(lock_level)
522 missing = set(self.wanted).difference(names)
524 raise errors.OpExecError("Some items were removed before retrieving"
525 " their data: %s" % missing)
527 # Return expanded names
530 def ExpandNames(self, lu):
531 """Expand names for this query.
533 See L{LogicalUnit.ExpandNames}.
536 raise NotImplementedError()
538 def DeclareLocks(self, lu, level):
539 """Declare locks for this query.
541 See L{LogicalUnit.DeclareLocks}.
544 raise NotImplementedError()
546 def _GetQueryData(self, lu):
547 """Collects all data for this query.
549 @return: Query data object
552 raise NotImplementedError()
554 def NewStyleQuery(self, lu):
555 """Collect data and execute query.
558 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
559 sort_by_name=self.sort_by_name)
561 def OldStyleQuery(self, lu):
562 """Collect data and execute query.
565 return self.query.OldStyleQuery(self._GetQueryData(lu),
566 sort_by_name=self.sort_by_name)
570 """Returns a dict declaring all lock levels shared.
573 return dict.fromkeys(locking.LEVELS, 1)
576 def _MakeLegacyNodeInfo(data):
577 """Formats the data returned by L{rpc.RpcRunner.call_node_info}.
579 Converts the data into a single dictionary. This is fine for most use cases,
580 but some require information from more than one volume group or hypervisor.
583 (bootid, (vg_info, ), (hv_info, )) = data
585 return utils.JoinDisjointDicts(utils.JoinDisjointDicts(vg_info, hv_info), {
590 def _CheckInstanceNodeGroups(cfg, instance_name, owned_groups):
591 """Checks if the owned node groups are still correct for an instance.
593 @type cfg: L{config.ConfigWriter}
594 @param cfg: The cluster configuration
595 @type instance_name: string
596 @param instance_name: Instance name
597 @type owned_groups: set or frozenset
598 @param owned_groups: List of currently owned node groups
601 inst_groups = cfg.GetInstanceNodeGroups(instance_name)
603 if not owned_groups.issuperset(inst_groups):
604 raise errors.OpPrereqError("Instance %s's node groups changed since"
605 " locks were acquired, current groups are"
606 " are '%s', owning groups '%s'; retry the"
609 utils.CommaJoin(inst_groups),
610 utils.CommaJoin(owned_groups)),
616 def _CheckNodeGroupInstances(cfg, group_uuid, owned_instances):
617 """Checks if the instances in a node group are still correct.
619 @type cfg: L{config.ConfigWriter}
620 @param cfg: The cluster configuration
621 @type group_uuid: string
622 @param group_uuid: Node group UUID
623 @type owned_instances: set or frozenset
624 @param owned_instances: List of currently owned instances
627 wanted_instances = cfg.GetNodeGroupInstances(group_uuid)
628 if owned_instances != wanted_instances:
629 raise errors.OpPrereqError("Instances in node group '%s' changed since"
630 " locks were acquired, wanted '%s', have '%s';"
631 " retry the operation" %
633 utils.CommaJoin(wanted_instances),
634 utils.CommaJoin(owned_instances)),
637 return wanted_instances
640 def _SupportsOob(cfg, node):
641 """Tells if node supports OOB.
643 @type cfg: L{config.ConfigWriter}
644 @param cfg: The cluster configuration
645 @type node: L{objects.Node}
646 @param node: The node
647 @return: The OOB script if supported or an empty string otherwise
650 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
653 def _GetWantedNodes(lu, nodes):
654 """Returns list of checked and expanded node names.
656 @type lu: L{LogicalUnit}
657 @param lu: the logical unit on whose behalf we execute
659 @param nodes: list of node names or None for all nodes
661 @return: the list of nodes, sorted
662 @raise errors.ProgrammerError: if the nodes parameter is wrong type
666 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
668 return utils.NiceSort(lu.cfg.GetNodeList())
671 def _GetWantedInstances(lu, instances):
672 """Returns list of checked and expanded instance names.
674 @type lu: L{LogicalUnit}
675 @param lu: the logical unit on whose behalf we execute
676 @type instances: list
677 @param instances: list of instance names or None for all instances
679 @return: the list of instances, sorted
680 @raise errors.OpPrereqError: if the instances parameter is wrong type
681 @raise errors.OpPrereqError: if any of the passed instances is not found
685 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
687 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
691 def _GetUpdatedParams(old_params, update_dict,
692 use_default=True, use_none=False):
693 """Return the new version of a parameter dictionary.
695 @type old_params: dict
696 @param old_params: old parameters
697 @type update_dict: dict
698 @param update_dict: dict containing new parameter values, or
699 constants.VALUE_DEFAULT to reset the parameter to its default
701 @param use_default: boolean
702 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
703 values as 'to be deleted' values
704 @param use_none: boolean
705 @type use_none: whether to recognise C{None} values as 'to be
708 @return: the new parameter dictionary
711 params_copy = copy.deepcopy(old_params)
712 for key, val in update_dict.iteritems():
713 if ((use_default and val == constants.VALUE_DEFAULT) or
714 (use_none and val is None)):
720 params_copy[key] = val
724 def _GetUpdatedIPolicy(old_ipolicy, new_ipolicy, group_policy=False):
725 """Return the new version of a instance policy.
727 @param group_policy: whether this policy applies to a group and thus
728 we should support removal of policy entries
731 use_none = use_default = group_policy
732 ipolicy = copy.deepcopy(old_ipolicy)
733 for key, value in new_ipolicy.items():
734 if key not in constants.IPOLICY_ALL_KEYS:
735 raise errors.OpPrereqError("Invalid key in new ipolicy: %s" % key,
737 if key in constants.IPOLICY_PARAMETERS:
738 utils.ForceDictType(value, constants.ISPECS_PARAMETER_TYPES)
739 ipolicy[key] = _GetUpdatedParams(old_ipolicy.get(key, {}), value,
741 use_default=use_default)
743 # FIXME: we assume all others are lists; this should be redone
745 if not value or value == [constants.VALUE_DEFAULT]:
749 raise errors.OpPrereqError("Can't unset ipolicy attribute '%s'"
750 " on the cluster'" % key,
753 ipolicy[key] = list(value)
755 objects.InstancePolicy.CheckParameterSyntax(ipolicy)
756 except errors.ConfigurationError, err:
757 raise errors.OpPrereqError("Invalid instance policy: %s" % err,
762 def _UpdateAndVerifySubDict(base, updates, type_check):
763 """Updates and verifies a dict with sub dicts of the same type.
765 @param base: The dict with the old data
766 @param updates: The dict with the new data
767 @param type_check: Dict suitable to ForceDictType to verify correct types
768 @returns: A new dict with updated and verified values
772 new = _GetUpdatedParams(old, value)
773 utils.ForceDictType(new, type_check)
776 ret = copy.deepcopy(base)
777 ret.update(dict((key, fn(base.get(key, {}), value))
778 for key, value in updates.items()))
782 def _MergeAndVerifyHvState(op_input, obj_input):
783 """Combines the hv state from an opcode with the one of the object
785 @param op_input: The input dict from the opcode
786 @param obj_input: The input dict from the objects
787 @return: The verified and updated dict
791 invalid_hvs = set(op_input) - constants.HYPER_TYPES
793 raise errors.OpPrereqError("Invalid hypervisor(s) in hypervisor state:"
794 " %s" % utils.CommaJoin(invalid_hvs),
796 if obj_input is None:
798 type_check = constants.HVSTS_PARAMETER_TYPES
799 return _UpdateAndVerifySubDict(obj_input, op_input, type_check)
804 def _MergeAndVerifyDiskState(op_input, obj_input):
805 """Combines the disk state from an opcode with the one of the object
807 @param op_input: The input dict from the opcode
808 @param obj_input: The input dict from the objects
809 @return: The verified and updated dict
812 invalid_dst = set(op_input) - constants.DS_VALID_TYPES
814 raise errors.OpPrereqError("Invalid storage type(s) in disk state: %s" %
815 utils.CommaJoin(invalid_dst),
817 type_check = constants.DSS_PARAMETER_TYPES
818 if obj_input is None:
820 return dict((key, _UpdateAndVerifySubDict(obj_input.get(key, {}), value,
822 for key, value in op_input.items())
827 def _ReleaseLocks(lu, level, names=None, keep=None):
828 """Releases locks owned by an LU.
830 @type lu: L{LogicalUnit}
831 @param level: Lock level
832 @type names: list or None
833 @param names: Names of locks to release
834 @type keep: list or None
835 @param keep: Names of locks to retain
838 assert not (keep is not None and names is not None), \
839 "Only one of the 'names' and the 'keep' parameters can be given"
841 if names is not None:
842 should_release = names.__contains__
844 should_release = lambda name: name not in keep
846 should_release = None
848 owned = lu.owned_locks(level)
850 # Not owning any lock at this level, do nothing
857 # Determine which locks to release
859 if should_release(name):
864 assert len(lu.owned_locks(level)) == (len(retain) + len(release))
866 # Release just some locks
867 lu.glm.release(level, names=release)
869 assert frozenset(lu.owned_locks(level)) == frozenset(retain)
872 lu.glm.release(level)
874 assert not lu.glm.is_owned(level), "No locks should be owned"
877 def _MapInstanceDisksToNodes(instances):
878 """Creates a map from (node, volume) to instance name.
880 @type instances: list of L{objects.Instance}
881 @rtype: dict; tuple of (node name, volume name) as key, instance name as value
884 return dict(((node, vol), inst.name)
885 for inst in instances
886 for (node, vols) in inst.MapLVsByNode().items()
890 def _RunPostHook(lu, node_name):
891 """Runs the post-hook for an opcode on a single node.
894 hm = lu.proc.BuildHooksManager(lu)
896 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
898 # pylint: disable=W0702
899 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
902 def _CheckOutputFields(static, dynamic, selected):
903 """Checks whether all selected fields are valid.
905 @type static: L{utils.FieldSet}
906 @param static: static fields set
907 @type dynamic: L{utils.FieldSet}
908 @param dynamic: dynamic fields set
915 delta = f.NonMatching(selected)
917 raise errors.OpPrereqError("Unknown output fields selected: %s"
918 % ",".join(delta), errors.ECODE_INVAL)
921 def _CheckGlobalHvParams(params):
922 """Validates that given hypervisor params are not global ones.
924 This will ensure that instances don't get customised versions of
928 used_globals = constants.HVC_GLOBALS.intersection(params)
930 msg = ("The following hypervisor parameters are global and cannot"
931 " be customized at instance level, please modify them at"
932 " cluster level: %s" % utils.CommaJoin(used_globals))
933 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
936 def _CheckNodeOnline(lu, node, msg=None):
937 """Ensure that a given node is online.
939 @param lu: the LU on behalf of which we make the check
940 @param node: the node to check
941 @param msg: if passed, should be a message to replace the default one
942 @raise errors.OpPrereqError: if the node is offline
946 msg = "Can't use offline node"
947 if lu.cfg.GetNodeInfo(node).offline:
948 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
951 def _CheckNodeNotDrained(lu, node):
952 """Ensure that a given node is not drained.
954 @param lu: the LU on behalf of which we make the check
955 @param node: the node to check
956 @raise errors.OpPrereqError: if the node is drained
959 if lu.cfg.GetNodeInfo(node).drained:
960 raise errors.OpPrereqError("Can't use drained node %s" % node,
964 def _CheckNodeVmCapable(lu, node):
965 """Ensure that a given node is vm capable.
967 @param lu: the LU on behalf of which we make the check
968 @param node: the node to check
969 @raise errors.OpPrereqError: if the node is not vm capable
972 if not lu.cfg.GetNodeInfo(node).vm_capable:
973 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
977 def _CheckNodeHasOS(lu, node, os_name, force_variant):
978 """Ensure that a node supports a given OS.
980 @param lu: the LU on behalf of which we make the check
981 @param node: the node to check
982 @param os_name: the OS to query about
983 @param force_variant: whether to ignore variant errors
984 @raise errors.OpPrereqError: if the node is not supporting the OS
987 result = lu.rpc.call_os_get(node, os_name)
988 result.Raise("OS '%s' not in supported OS list for node %s" %
990 prereq=True, ecode=errors.ECODE_INVAL)
991 if not force_variant:
992 _CheckOSVariant(result.payload, os_name)
995 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
996 """Ensure that a node has the given secondary ip.
998 @type lu: L{LogicalUnit}
999 @param lu: the LU on behalf of which we make the check
1001 @param node: the node to check
1002 @type secondary_ip: string
1003 @param secondary_ip: the ip to check
1004 @type prereq: boolean
1005 @param prereq: whether to throw a prerequisite or an execute error
1006 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
1007 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
1010 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
1011 result.Raise("Failure checking secondary ip on node %s" % node,
1012 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1013 if not result.payload:
1014 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
1015 " please fix and re-run this command" % secondary_ip)
1017 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
1019 raise errors.OpExecError(msg)
1022 def _GetClusterDomainSecret():
1023 """Reads the cluster domain secret.
1026 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
1030 def _CheckInstanceState(lu, instance, req_states, msg=None):
1031 """Ensure that an instance is in one of the required states.
1033 @param lu: the LU on behalf of which we make the check
1034 @param instance: the instance to check
1035 @param msg: if passed, should be a message to replace the default one
1036 @raise errors.OpPrereqError: if the instance is not in the required state
1040 msg = "can't use instance from outside %s states" % ", ".join(req_states)
1041 if instance.admin_state not in req_states:
1042 raise errors.OpPrereqError("Instance '%s' is marked to be %s, %s" %
1043 (instance.name, instance.admin_state, msg),
1046 if constants.ADMINST_UP not in req_states:
1047 pnode = instance.primary_node
1048 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
1049 ins_l.Raise("Can't contact node %s for instance information" % pnode,
1050 prereq=True, ecode=errors.ECODE_ENVIRON)
1052 if instance.name in ins_l.payload:
1053 raise errors.OpPrereqError("Instance %s is running, %s" %
1054 (instance.name, msg), errors.ECODE_STATE)
1057 def _ComputeMinMaxSpec(name, ipolicy, value):
1058 """Computes if value is in the desired range.
1060 @param name: name of the parameter for which we perform the check
1061 @param ipolicy: dictionary containing min, max and std values
1062 @param value: actual value that we want to use
1063 @return: None or element not meeting the criteria
1067 if value in [None, constants.VALUE_AUTO]:
1069 max_v = ipolicy[constants.ISPECS_MAX].get(name, value)
1070 min_v = ipolicy[constants.ISPECS_MIN].get(name, value)
1071 if value > max_v or min_v > value:
1072 return ("%s value %s is not in range [%s, %s]" %
1073 (name, value, min_v, max_v))
1077 def _ComputeIPolicySpecViolation(ipolicy, mem_size, cpu_count, disk_count,
1078 nic_count, disk_sizes,
1079 _compute_fn=_ComputeMinMaxSpec):
1080 """Verifies ipolicy against provided specs.
1083 @param ipolicy: The ipolicy
1085 @param mem_size: The memory size
1086 @type cpu_count: int
1087 @param cpu_count: Used cpu cores
1088 @type disk_count: int
1089 @param disk_count: Number of disks used
1090 @type nic_count: int
1091 @param nic_count: Number of nics used
1092 @type disk_sizes: list of ints
1093 @param disk_sizes: Disk sizes of used disk (len must match C{disk_count})
1094 @param _compute_fn: The compute function (unittest only)
1095 @return: A list of violations, or an empty list of no violations are found
1098 assert disk_count == len(disk_sizes)
1101 (constants.ISPEC_MEM_SIZE, mem_size),
1102 (constants.ISPEC_CPU_COUNT, cpu_count),
1103 (constants.ISPEC_DISK_COUNT, disk_count),
1104 (constants.ISPEC_NIC_COUNT, nic_count),
1105 ] + map((lambda d: (constants.ISPEC_DISK_SIZE, d)), disk_sizes)
1108 (_compute_fn(name, ipolicy, value)
1109 for (name, value) in test_settings))
1112 def _ComputeIPolicyInstanceViolation(ipolicy, instance,
1113 _compute_fn=_ComputeIPolicySpecViolation):
1114 """Compute if instance meets the specs of ipolicy.
1117 @param ipolicy: The ipolicy to verify against
1118 @type instance: L{objects.Instance}
1119 @param instance: The instance to verify
1120 @param _compute_fn: The function to verify ipolicy (unittest only)
1121 @see: L{_ComputeIPolicySpecViolation}
1124 mem_size = instance.beparams.get(constants.BE_MAXMEM, None)
1125 cpu_count = instance.beparams.get(constants.BE_VCPUS, None)
1126 disk_count = len(instance.disks)
1127 disk_sizes = [disk.size for disk in instance.disks]
1128 nic_count = len(instance.nics)
1130 return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
1134 def _ComputeIPolicyInstanceSpecViolation(ipolicy, instance_spec,
1135 _compute_fn=_ComputeIPolicySpecViolation):
1136 """Compute if instance specs meets the specs of ipolicy.
1139 @param ipolicy: The ipolicy to verify against
1140 @param instance_spec: dict
1141 @param instance_spec: The instance spec to verify
1142 @param _compute_fn: The function to verify ipolicy (unittest only)
1143 @see: L{_ComputeIPolicySpecViolation}
1146 mem_size = instance_spec.get(constants.ISPEC_MEM_SIZE, None)
1147 cpu_count = instance_spec.get(constants.ISPEC_CPU_COUNT, None)
1148 disk_count = instance_spec.get(constants.ISPEC_DISK_COUNT, 0)
1149 disk_sizes = instance_spec.get(constants.ISPEC_DISK_SIZE, [])
1150 nic_count = instance_spec.get(constants.ISPEC_NIC_COUNT, 0)
1152 return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
1156 def _ComputeIPolicyNodeViolation(ipolicy, instance, current_group,
1158 _compute_fn=_ComputeIPolicyInstanceViolation):
1159 """Compute if instance meets the specs of the new target group.
1161 @param ipolicy: The ipolicy to verify
1162 @param instance: The instance object to verify
1163 @param current_group: The current group of the instance
1164 @param target_group: The new group of the instance
1165 @param _compute_fn: The function to verify ipolicy (unittest only)
1166 @see: L{_ComputeIPolicySpecViolation}
1169 if current_group == target_group:
1172 return _compute_fn(ipolicy, instance)
1175 def _CheckTargetNodeIPolicy(lu, ipolicy, instance, node, ignore=False,
1176 _compute_fn=_ComputeIPolicyNodeViolation):
1177 """Checks that the target node is correct in terms of instance policy.
1179 @param ipolicy: The ipolicy to verify
1180 @param instance: The instance object to verify
1181 @param node: The new node to relocate
1182 @param ignore: Ignore violations of the ipolicy
1183 @param _compute_fn: The function to verify ipolicy (unittest only)
1184 @see: L{_ComputeIPolicySpecViolation}
1187 primary_node = lu.cfg.GetNodeInfo(instance.primary_node)
1188 res = _compute_fn(ipolicy, instance, primary_node.group, node.group)
1191 msg = ("Instance does not meet target node group's (%s) instance"
1192 " policy: %s") % (node.group, utils.CommaJoin(res))
1196 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
1199 def _ExpandItemName(fn, name, kind):
1200 """Expand an item name.
1202 @param fn: the function to use for expansion
1203 @param name: requested item name
1204 @param kind: text description ('Node' or 'Instance')
1205 @return: the resolved (full) name
1206 @raise errors.OpPrereqError: if the item is not found
1209 full_name = fn(name)
1210 if full_name is None:
1211 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
1216 def _ExpandNodeName(cfg, name):
1217 """Wrapper over L{_ExpandItemName} for nodes."""
1218 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
1221 def _ExpandInstanceName(cfg, name):
1222 """Wrapper over L{_ExpandItemName} for instance."""
1223 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
1226 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
1227 minmem, maxmem, vcpus, nics, disk_template, disks,
1228 bep, hvp, hypervisor_name, tags):
1229 """Builds instance related env variables for hooks
1231 This builds the hook environment from individual variables.
1234 @param name: the name of the instance
1235 @type primary_node: string
1236 @param primary_node: the name of the instance's primary node
1237 @type secondary_nodes: list
1238 @param secondary_nodes: list of secondary nodes as strings
1239 @type os_type: string
1240 @param os_type: the name of the instance's OS
1241 @type status: string
1242 @param status: the desired status of the instance
1243 @type minmem: string
1244 @param minmem: the minimum memory size of the instance
1245 @type maxmem: string
1246 @param maxmem: the maximum memory size of the instance
1248 @param vcpus: the count of VCPUs the instance has
1250 @param nics: list of tuples (ip, mac, mode, link) representing
1251 the NICs the instance has
1252 @type disk_template: string
1253 @param disk_template: the disk template of the instance
1255 @param disks: the list of (size, mode) pairs
1257 @param bep: the backend parameters for the instance
1259 @param hvp: the hypervisor parameters for the instance
1260 @type hypervisor_name: string
1261 @param hypervisor_name: the hypervisor for the instance
1263 @param tags: list of instance tags as strings
1265 @return: the hook environment for this instance
1270 "INSTANCE_NAME": name,
1271 "INSTANCE_PRIMARY": primary_node,
1272 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
1273 "INSTANCE_OS_TYPE": os_type,
1274 "INSTANCE_STATUS": status,
1275 "INSTANCE_MINMEM": minmem,
1276 "INSTANCE_MAXMEM": maxmem,
1277 # TODO(2.7) remove deprecated "memory" value
1278 "INSTANCE_MEMORY": maxmem,
1279 "INSTANCE_VCPUS": vcpus,
1280 "INSTANCE_DISK_TEMPLATE": disk_template,
1281 "INSTANCE_HYPERVISOR": hypervisor_name,
1284 nic_count = len(nics)
1285 for idx, (ip, mac, mode, link) in enumerate(nics):
1288 env["INSTANCE_NIC%d_IP" % idx] = ip
1289 env["INSTANCE_NIC%d_MAC" % idx] = mac
1290 env["INSTANCE_NIC%d_MODE" % idx] = mode
1291 env["INSTANCE_NIC%d_LINK" % idx] = link
1292 if mode == constants.NIC_MODE_BRIDGED:
1293 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
1297 env["INSTANCE_NIC_COUNT"] = nic_count
1300 disk_count = len(disks)
1301 for idx, (size, mode) in enumerate(disks):
1302 env["INSTANCE_DISK%d_SIZE" % idx] = size
1303 env["INSTANCE_DISK%d_MODE" % idx] = mode
1307 env["INSTANCE_DISK_COUNT"] = disk_count
1312 env["INSTANCE_TAGS"] = " ".join(tags)
1314 for source, kind in [(bep, "BE"), (hvp, "HV")]:
1315 for key, value in source.items():
1316 env["INSTANCE_%s_%s" % (kind, key)] = value
1321 def _NICListToTuple(lu, nics):
1322 """Build a list of nic information tuples.
1324 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
1325 value in LUInstanceQueryData.
1327 @type lu: L{LogicalUnit}
1328 @param lu: the logical unit on whose behalf we execute
1329 @type nics: list of L{objects.NIC}
1330 @param nics: list of nics to convert to hooks tuples
1334 cluster = lu.cfg.GetClusterInfo()
1338 filled_params = cluster.SimpleFillNIC(nic.nicparams)
1339 mode = filled_params[constants.NIC_MODE]
1340 link = filled_params[constants.NIC_LINK]
1341 hooks_nics.append((ip, mac, mode, link))
1345 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1346 """Builds instance related env variables for hooks from an object.
1348 @type lu: L{LogicalUnit}
1349 @param lu: the logical unit on whose behalf we execute
1350 @type instance: L{objects.Instance}
1351 @param instance: the instance for which we should build the
1353 @type override: dict
1354 @param override: dictionary with key/values that will override
1357 @return: the hook environment dictionary
1360 cluster = lu.cfg.GetClusterInfo()
1361 bep = cluster.FillBE(instance)
1362 hvp = cluster.FillHV(instance)
1364 "name": instance.name,
1365 "primary_node": instance.primary_node,
1366 "secondary_nodes": instance.secondary_nodes,
1367 "os_type": instance.os,
1368 "status": instance.admin_state,
1369 "maxmem": bep[constants.BE_MAXMEM],
1370 "minmem": bep[constants.BE_MINMEM],
1371 "vcpus": bep[constants.BE_VCPUS],
1372 "nics": _NICListToTuple(lu, instance.nics),
1373 "disk_template": instance.disk_template,
1374 "disks": [(disk.size, disk.mode) for disk in instance.disks],
1377 "hypervisor_name": instance.hypervisor,
1378 "tags": instance.tags,
1381 args.update(override)
1382 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
1385 def _AdjustCandidatePool(lu, exceptions):
1386 """Adjust the candidate pool after node operations.
1389 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1391 lu.LogInfo("Promoted nodes to master candidate role: %s",
1392 utils.CommaJoin(node.name for node in mod_list))
1393 for name in mod_list:
1394 lu.context.ReaddNode(name)
1395 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1397 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1401 def _DecideSelfPromotion(lu, exceptions=None):
1402 """Decide whether I should promote myself as a master candidate.
1405 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1406 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1407 # the new node will increase mc_max with one, so:
1408 mc_should = min(mc_should + 1, cp_size)
1409 return mc_now < mc_should
1412 def _CalculateGroupIPolicy(cluster, group):
1413 """Calculate instance policy for group.
1416 return cluster.SimpleFillIPolicy(group.ipolicy)
1419 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1420 """Check that the brigdes needed by a list of nics exist.
1423 cluster = lu.cfg.GetClusterInfo()
1424 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1425 brlist = [params[constants.NIC_LINK] for params in paramslist
1426 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1428 result = lu.rpc.call_bridges_exist(target_node, brlist)
1429 result.Raise("Error checking bridges on destination node '%s'" %
1430 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1433 def _CheckInstanceBridgesExist(lu, instance, node=None):
1434 """Check that the brigdes needed by an instance exist.
1438 node = instance.primary_node
1439 _CheckNicsBridgesExist(lu, instance.nics, node)
1442 def _CheckOSVariant(os_obj, name):
1443 """Check whether an OS name conforms to the os variants specification.
1445 @type os_obj: L{objects.OS}
1446 @param os_obj: OS object to check
1448 @param name: OS name passed by the user, to check for validity
1451 variant = objects.OS.GetVariant(name)
1452 if not os_obj.supported_variants:
1454 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'"
1455 " passed)" % (os_obj.name, variant),
1459 raise errors.OpPrereqError("OS name must include a variant",
1462 if variant not in os_obj.supported_variants:
1463 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1466 def _GetNodeInstancesInner(cfg, fn):
1467 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1470 def _GetNodeInstances(cfg, node_name):
1471 """Returns a list of all primary and secondary instances on a node.
1475 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1478 def _GetNodePrimaryInstances(cfg, node_name):
1479 """Returns primary instances on a node.
1482 return _GetNodeInstancesInner(cfg,
1483 lambda inst: node_name == inst.primary_node)
1486 def _GetNodeSecondaryInstances(cfg, node_name):
1487 """Returns secondary instances on a node.
1490 return _GetNodeInstancesInner(cfg,
1491 lambda inst: node_name in inst.secondary_nodes)
1494 def _GetStorageTypeArgs(cfg, storage_type):
1495 """Returns the arguments for a storage type.
1498 # Special case for file storage
1499 if storage_type == constants.ST_FILE:
1500 # storage.FileStorage wants a list of storage directories
1501 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1506 def _FindFaultyInstanceDisks(cfg, rpc_runner, instance, node_name, prereq):
1509 for dev in instance.disks:
1510 cfg.SetDiskID(dev, node_name)
1512 result = rpc_runner.call_blockdev_getmirrorstatus(node_name, instance.disks)
1513 result.Raise("Failed to get disk status from node %s" % node_name,
1514 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1516 for idx, bdev_status in enumerate(result.payload):
1517 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1523 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1524 """Check the sanity of iallocator and node arguments and use the
1525 cluster-wide iallocator if appropriate.
1527 Check that at most one of (iallocator, node) is specified. If none is
1528 specified, then the LU's opcode's iallocator slot is filled with the
1529 cluster-wide default iallocator.
1531 @type iallocator_slot: string
1532 @param iallocator_slot: the name of the opcode iallocator slot
1533 @type node_slot: string
1534 @param node_slot: the name of the opcode target node slot
1537 node = getattr(lu.op, node_slot, None)
1538 iallocator = getattr(lu.op, iallocator_slot, None)
1540 if node is not None and iallocator is not None:
1541 raise errors.OpPrereqError("Do not specify both, iallocator and node",
1543 elif node is None and iallocator is None:
1544 default_iallocator = lu.cfg.GetDefaultIAllocator()
1545 if default_iallocator:
1546 setattr(lu.op, iallocator_slot, default_iallocator)
1548 raise errors.OpPrereqError("No iallocator or node given and no"
1549 " cluster-wide default iallocator found;"
1550 " please specify either an iallocator or a"
1551 " node, or set a cluster-wide default"
1555 def _GetDefaultIAllocator(cfg, iallocator):
1556 """Decides on which iallocator to use.
1558 @type cfg: L{config.ConfigWriter}
1559 @param cfg: Cluster configuration object
1560 @type iallocator: string or None
1561 @param iallocator: Iallocator specified in opcode
1563 @return: Iallocator name
1567 # Use default iallocator
1568 iallocator = cfg.GetDefaultIAllocator()
1571 raise errors.OpPrereqError("No iallocator was specified, neither in the"
1572 " opcode nor as a cluster-wide default",
1578 class LUClusterPostInit(LogicalUnit):
1579 """Logical unit for running hooks after cluster initialization.
1582 HPATH = "cluster-init"
1583 HTYPE = constants.HTYPE_CLUSTER
1585 def BuildHooksEnv(self):
1590 "OP_TARGET": self.cfg.GetClusterName(),
1593 def BuildHooksNodes(self):
1594 """Build hooks nodes.
1597 return ([], [self.cfg.GetMasterNode()])
1599 def Exec(self, feedback_fn):
1606 class LUClusterDestroy(LogicalUnit):
1607 """Logical unit for destroying the cluster.
1610 HPATH = "cluster-destroy"
1611 HTYPE = constants.HTYPE_CLUSTER
1613 def BuildHooksEnv(self):
1618 "OP_TARGET": self.cfg.GetClusterName(),
1621 def BuildHooksNodes(self):
1622 """Build hooks nodes.
1627 def CheckPrereq(self):
1628 """Check prerequisites.
1630 This checks whether the cluster is empty.
1632 Any errors are signaled by raising errors.OpPrereqError.
1635 master = self.cfg.GetMasterNode()
1637 nodelist = self.cfg.GetNodeList()
1638 if len(nodelist) != 1 or nodelist[0] != master:
1639 raise errors.OpPrereqError("There are still %d node(s) in"
1640 " this cluster." % (len(nodelist) - 1),
1642 instancelist = self.cfg.GetInstanceList()
1644 raise errors.OpPrereqError("There are still %d instance(s) in"
1645 " this cluster." % len(instancelist),
1648 def Exec(self, feedback_fn):
1649 """Destroys the cluster.
1652 master_params = self.cfg.GetMasterNetworkParameters()
1654 # Run post hooks on master node before it's removed
1655 _RunPostHook(self, master_params.name)
1657 ems = self.cfg.GetUseExternalMipScript()
1658 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
1661 self.LogWarning("Error disabling the master IP address: %s",
1664 return master_params.name
1667 def _VerifyCertificate(filename):
1668 """Verifies a certificate for L{LUClusterVerifyConfig}.
1670 @type filename: string
1671 @param filename: Path to PEM file
1675 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1676 utils.ReadFile(filename))
1677 except Exception, err: # pylint: disable=W0703
1678 return (LUClusterVerifyConfig.ETYPE_ERROR,
1679 "Failed to load X509 certificate %s: %s" % (filename, err))
1682 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1683 constants.SSL_CERT_EXPIRATION_ERROR)
1686 fnamemsg = "While verifying %s: %s" % (filename, msg)
1691 return (None, fnamemsg)
1692 elif errcode == utils.CERT_WARNING:
1693 return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg)
1694 elif errcode == utils.CERT_ERROR:
1695 return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg)
1697 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1700 def _GetAllHypervisorParameters(cluster, instances):
1701 """Compute the set of all hypervisor parameters.
1703 @type cluster: L{objects.Cluster}
1704 @param cluster: the cluster object
1705 @param instances: list of L{objects.Instance}
1706 @param instances: additional instances from which to obtain parameters
1707 @rtype: list of (origin, hypervisor, parameters)
1708 @return: a list with all parameters found, indicating the hypervisor they
1709 apply to, and the origin (can be "cluster", "os X", or "instance Y")
1714 for hv_name in cluster.enabled_hypervisors:
1715 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
1717 for os_name, os_hvp in cluster.os_hvp.items():
1718 for hv_name, hv_params in os_hvp.items():
1720 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
1721 hvp_data.append(("os %s" % os_name, hv_name, full_params))
1723 # TODO: collapse identical parameter values in a single one
1724 for instance in instances:
1725 if instance.hvparams:
1726 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
1727 cluster.FillHV(instance)))
1732 class _VerifyErrors(object):
1733 """Mix-in for cluster/group verify LUs.
1735 It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects
1736 self.op and self._feedback_fn to be available.)
1740 ETYPE_FIELD = "code"
1741 ETYPE_ERROR = "ERROR"
1742 ETYPE_WARNING = "WARNING"
1744 def _Error(self, ecode, item, msg, *args, **kwargs):
1745 """Format an error message.
1747 Based on the opcode's error_codes parameter, either format a
1748 parseable error code, or a simpler error string.
1750 This must be called only from Exec and functions called from Exec.
1753 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1754 itype, etxt, _ = ecode
1755 # first complete the msg
1758 # then format the whole message
1759 if self.op.error_codes: # This is a mix-in. pylint: disable=E1101
1760 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1766 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1767 # and finally report it via the feedback_fn
1768 self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable=E1101
1770 def _ErrorIf(self, cond, ecode, *args, **kwargs):
1771 """Log an error message if the passed condition is True.
1775 or self.op.debug_simulate_errors) # pylint: disable=E1101
1777 # If the error code is in the list of ignored errors, demote the error to a
1779 (_, etxt, _) = ecode
1780 if etxt in self.op.ignore_errors: # pylint: disable=E1101
1781 kwargs[self.ETYPE_FIELD] = self.ETYPE_WARNING
1784 self._Error(ecode, *args, **kwargs)
1786 # do not mark the operation as failed for WARN cases only
1787 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1788 self.bad = self.bad or cond
1791 class LUClusterVerify(NoHooksLU):
1792 """Submits all jobs necessary to verify the cluster.
1797 def ExpandNames(self):
1798 self.needed_locks = {}
1800 def Exec(self, feedback_fn):
1803 if self.op.group_name:
1804 groups = [self.op.group_name]
1805 depends_fn = lambda: None
1807 groups = self.cfg.GetNodeGroupList()
1809 # Verify global configuration
1811 opcodes.OpClusterVerifyConfig(ignore_errors=self.op.ignore_errors)
1814 # Always depend on global verification
1815 depends_fn = lambda: [(-len(jobs), [])]
1817 jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,
1818 ignore_errors=self.op.ignore_errors,
1819 depends=depends_fn())]
1820 for group in groups)
1822 # Fix up all parameters
1823 for op in itertools.chain(*jobs): # pylint: disable=W0142
1824 op.debug_simulate_errors = self.op.debug_simulate_errors
1825 op.verbose = self.op.verbose
1826 op.error_codes = self.op.error_codes
1828 op.skip_checks = self.op.skip_checks
1829 except AttributeError:
1830 assert not isinstance(op, opcodes.OpClusterVerifyGroup)
1832 return ResultWithJobs(jobs)
1835 class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
1836 """Verifies the cluster config.
1841 def _VerifyHVP(self, hvp_data):
1842 """Verifies locally the syntax of the hypervisor parameters.
1845 for item, hv_name, hv_params in hvp_data:
1846 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
1849 hv_class = hypervisor.GetHypervisor(hv_name)
1850 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1851 hv_class.CheckParameterSyntax(hv_params)
1852 except errors.GenericError, err:
1853 self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg % str(err))
1855 def ExpandNames(self):
1856 # Information can be safely retrieved as the BGL is acquired in exclusive
1858 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER)
1859 self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
1860 self.all_node_info = self.cfg.GetAllNodesInfo()
1861 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1862 self.needed_locks = {}
1864 def Exec(self, feedback_fn):
1865 """Verify integrity of cluster, performing various test on nodes.
1869 self._feedback_fn = feedback_fn
1871 feedback_fn("* Verifying cluster config")
1873 for msg in self.cfg.VerifyConfig():
1874 self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg)
1876 feedback_fn("* Verifying cluster certificate files")
1878 for cert_filename in constants.ALL_CERT_FILES:
1879 (errcode, msg) = _VerifyCertificate(cert_filename)
1880 self._ErrorIf(errcode, constants.CV_ECLUSTERCERT, None, msg, code=errcode)
1882 feedback_fn("* Verifying hypervisor parameters")
1884 self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(),
1885 self.all_inst_info.values()))
1887 feedback_fn("* Verifying all nodes belong to an existing group")
1889 # We do this verification here because, should this bogus circumstance
1890 # occur, it would never be caught by VerifyGroup, which only acts on
1891 # nodes/instances reachable from existing node groups.
1893 dangling_nodes = set(node.name for node in self.all_node_info.values()
1894 if node.group not in self.all_group_info)
1896 dangling_instances = {}
1897 no_node_instances = []
1899 for inst in self.all_inst_info.values():
1900 if inst.primary_node in dangling_nodes:
1901 dangling_instances.setdefault(inst.primary_node, []).append(inst.name)
1902 elif inst.primary_node not in self.all_node_info:
1903 no_node_instances.append(inst.name)
1908 utils.CommaJoin(dangling_instances.get(node.name,
1910 for node in dangling_nodes]
1912 self._ErrorIf(bool(dangling_nodes), constants.CV_ECLUSTERDANGLINGNODES,
1914 "the following nodes (and their instances) belong to a non"
1915 " existing group: %s", utils.CommaJoin(pretty_dangling))
1917 self._ErrorIf(bool(no_node_instances), constants.CV_ECLUSTERDANGLINGINST,
1919 "the following instances have a non-existing primary-node:"
1920 " %s", utils.CommaJoin(no_node_instances))
1925 class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
1926 """Verifies the status of a node group.
1929 HPATH = "cluster-verify"
1930 HTYPE = constants.HTYPE_CLUSTER
1933 _HOOKS_INDENT_RE = re.compile("^", re.M)
1935 class NodeImage(object):
1936 """A class representing the logical and physical status of a node.
1939 @ivar name: the node name to which this object refers
1940 @ivar volumes: a structure as returned from
1941 L{ganeti.backend.GetVolumeList} (runtime)
1942 @ivar instances: a list of running instances (runtime)
1943 @ivar pinst: list of configured primary instances (config)
1944 @ivar sinst: list of configured secondary instances (config)
1945 @ivar sbp: dictionary of {primary-node: list of instances} for all
1946 instances for which this node is secondary (config)
1947 @ivar mfree: free memory, as reported by hypervisor (runtime)
1948 @ivar dfree: free disk, as reported by the node (runtime)
1949 @ivar offline: the offline status (config)
1950 @type rpc_fail: boolean
1951 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1952 not whether the individual keys were correct) (runtime)
1953 @type lvm_fail: boolean
1954 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1955 @type hyp_fail: boolean
1956 @ivar hyp_fail: whether the RPC call didn't return the instance list
1957 @type ghost: boolean
1958 @ivar ghost: whether this is a known node or not (config)
1959 @type os_fail: boolean
1960 @ivar os_fail: whether the RPC call didn't return valid OS data
1962 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1963 @type vm_capable: boolean
1964 @ivar vm_capable: whether the node can host instances
1967 def __init__(self, offline=False, name=None, vm_capable=True):
1976 self.offline = offline
1977 self.vm_capable = vm_capable
1978 self.rpc_fail = False
1979 self.lvm_fail = False
1980 self.hyp_fail = False
1982 self.os_fail = False
1985 def ExpandNames(self):
1986 # This raises errors.OpPrereqError on its own:
1987 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
1989 # Get instances in node group; this is unsafe and needs verification later
1990 inst_names = self.cfg.GetNodeGroupInstances(self.group_uuid)
1992 self.needed_locks = {
1993 locking.LEVEL_INSTANCE: inst_names,
1994 locking.LEVEL_NODEGROUP: [self.group_uuid],
1995 locking.LEVEL_NODE: [],
1998 self.share_locks = _ShareAll()
2000 def DeclareLocks(self, level):
2001 if level == locking.LEVEL_NODE:
2002 # Get members of node group; this is unsafe and needs verification later
2003 nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members)
2005 all_inst_info = self.cfg.GetAllInstancesInfo()
2007 # In Exec(), we warn about mirrored instances that have primary and
2008 # secondary living in separate node groups. To fully verify that
2009 # volumes for these instances are healthy, we will need to do an
2010 # extra call to their secondaries. We ensure here those nodes will
2012 for inst in self.owned_locks(locking.LEVEL_INSTANCE):
2013 # Important: access only the instances whose lock is owned
2014 if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR:
2015 nodes.update(all_inst_info[inst].secondary_nodes)
2017 self.needed_locks[locking.LEVEL_NODE] = nodes
2019 def CheckPrereq(self):
2020 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
2021 self.group_info = self.cfg.GetNodeGroup(self.group_uuid)
2023 group_nodes = set(self.group_info.members)
2024 group_instances = self.cfg.GetNodeGroupInstances(self.group_uuid)
2027 group_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
2029 unlocked_instances = \
2030 group_instances.difference(self.owned_locks(locking.LEVEL_INSTANCE))
2033 raise errors.OpPrereqError("Missing lock for nodes: %s" %
2034 utils.CommaJoin(unlocked_nodes))
2036 if unlocked_instances:
2037 raise errors.OpPrereqError("Missing lock for instances: %s" %
2038 utils.CommaJoin(unlocked_instances))
2040 self.all_node_info = self.cfg.GetAllNodesInfo()
2041 self.all_inst_info = self.cfg.GetAllInstancesInfo()
2043 self.my_node_names = utils.NiceSort(group_nodes)
2044 self.my_inst_names = utils.NiceSort(group_instances)
2046 self.my_node_info = dict((name, self.all_node_info[name])
2047 for name in self.my_node_names)
2049 self.my_inst_info = dict((name, self.all_inst_info[name])
2050 for name in self.my_inst_names)
2052 # We detect here the nodes that will need the extra RPC calls for verifying
2053 # split LV volumes; they should be locked.
2054 extra_lv_nodes = set()
2056 for inst in self.my_inst_info.values():
2057 if inst.disk_template in constants.DTS_INT_MIRROR:
2058 group = self.my_node_info[inst.primary_node].group
2059 for nname in inst.secondary_nodes:
2060 if self.all_node_info[nname].group != group:
2061 extra_lv_nodes.add(nname)
2063 unlocked_lv_nodes = \
2064 extra_lv_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
2066 if unlocked_lv_nodes:
2067 raise errors.OpPrereqError("these nodes could be locked: %s" %
2068 utils.CommaJoin(unlocked_lv_nodes))
2069 self.extra_lv_nodes = list(extra_lv_nodes)
2071 def _VerifyNode(self, ninfo, nresult):
2072 """Perform some basic validation on data returned from a node.
2074 - check the result data structure is well formed and has all the
2076 - check ganeti version
2078 @type ninfo: L{objects.Node}
2079 @param ninfo: the node to check
2080 @param nresult: the results from the node
2082 @return: whether overall this call was successful (and we can expect
2083 reasonable values in the respose)
2087 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2089 # main result, nresult should be a non-empty dict
2090 test = not nresult or not isinstance(nresult, dict)
2091 _ErrorIf(test, constants.CV_ENODERPC, node,
2092 "unable to verify node: no data returned")
2096 # compares ganeti version
2097 local_version = constants.PROTOCOL_VERSION
2098 remote_version = nresult.get("version", None)
2099 test = not (remote_version and
2100 isinstance(remote_version, (list, tuple)) and
2101 len(remote_version) == 2)
2102 _ErrorIf(test, constants.CV_ENODERPC, node,
2103 "connection to node returned invalid data")
2107 test = local_version != remote_version[0]
2108 _ErrorIf(test, constants.CV_ENODEVERSION, node,
2109 "incompatible protocol versions: master %s,"
2110 " node %s", local_version, remote_version[0])
2114 # node seems compatible, we can actually try to look into its results
2116 # full package version
2117 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
2118 constants.CV_ENODEVERSION, node,
2119 "software version mismatch: master %s, node %s",
2120 constants.RELEASE_VERSION, remote_version[1],
2121 code=self.ETYPE_WARNING)
2123 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
2124 if ninfo.vm_capable and isinstance(hyp_result, dict):
2125 for hv_name, hv_result in hyp_result.iteritems():
2126 test = hv_result is not None
2127 _ErrorIf(test, constants.CV_ENODEHV, node,
2128 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
2130 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
2131 if ninfo.vm_capable and isinstance(hvp_result, list):
2132 for item, hv_name, hv_result in hvp_result:
2133 _ErrorIf(True, constants.CV_ENODEHV, node,
2134 "hypervisor %s parameter verify failure (source %s): %s",
2135 hv_name, item, hv_result)
2137 test = nresult.get(constants.NV_NODESETUP,
2138 ["Missing NODESETUP results"])
2139 _ErrorIf(test, constants.CV_ENODESETUP, node, "node setup error: %s",
2144 def _VerifyNodeTime(self, ninfo, nresult,
2145 nvinfo_starttime, nvinfo_endtime):
2146 """Check the node time.
2148 @type ninfo: L{objects.Node}
2149 @param ninfo: the node to check
2150 @param nresult: the remote results for the node
2151 @param nvinfo_starttime: the start time of the RPC call
2152 @param nvinfo_endtime: the end time of the RPC call
2156 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2158 ntime = nresult.get(constants.NV_TIME, None)
2160 ntime_merged = utils.MergeTime(ntime)
2161 except (ValueError, TypeError):
2162 _ErrorIf(True, constants.CV_ENODETIME, node, "Node returned invalid time")
2165 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
2166 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
2167 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
2168 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
2172 _ErrorIf(ntime_diff is not None, constants.CV_ENODETIME, node,
2173 "Node time diverges by at least %s from master node time",
2176 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
2177 """Check the node LVM results.
2179 @type ninfo: L{objects.Node}
2180 @param ninfo: the node to check
2181 @param nresult: the remote results for the node
2182 @param vg_name: the configured VG name
2189 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2191 # checks vg existence and size > 20G
2192 vglist = nresult.get(constants.NV_VGLIST, None)
2194 _ErrorIf(test, constants.CV_ENODELVM, node, "unable to check volume groups")
2196 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
2197 constants.MIN_VG_SIZE)
2198 _ErrorIf(vgstatus, constants.CV_ENODELVM, node, vgstatus)
2201 pvlist = nresult.get(constants.NV_PVLIST, None)
2202 test = pvlist is None
2203 _ErrorIf(test, constants.CV_ENODELVM, node, "Can't get PV list from node")
2205 # check that ':' is not present in PV names, since it's a
2206 # special character for lvcreate (denotes the range of PEs to
2208 for _, pvname, owner_vg in pvlist:
2209 test = ":" in pvname
2210 _ErrorIf(test, constants.CV_ENODELVM, node,
2211 "Invalid character ':' in PV '%s' of VG '%s'",
2214 def _VerifyNodeBridges(self, ninfo, nresult, bridges):
2215 """Check the node bridges.
2217 @type ninfo: L{objects.Node}
2218 @param ninfo: the node to check
2219 @param nresult: the remote results for the node
2220 @param bridges: the expected list of bridges
2227 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2229 missing = nresult.get(constants.NV_BRIDGES, None)
2230 test = not isinstance(missing, list)
2231 _ErrorIf(test, constants.CV_ENODENET, node,
2232 "did not return valid bridge information")
2234 _ErrorIf(bool(missing), constants.CV_ENODENET, node,
2235 "missing bridges: %s" % utils.CommaJoin(sorted(missing)))
2237 def _VerifyNodeUserScripts(self, ninfo, nresult):
2238 """Check the results of user scripts presence and executability on the node
2240 @type ninfo: L{objects.Node}
2241 @param ninfo: the node to check
2242 @param nresult: the remote results for the node
2247 test = not constants.NV_USERSCRIPTS in nresult
2248 self._ErrorIf(test, constants.CV_ENODEUSERSCRIPTS, node,
2249 "did not return user scripts information")
2251 broken_scripts = nresult.get(constants.NV_USERSCRIPTS, None)
2253 self._ErrorIf(broken_scripts, constants.CV_ENODEUSERSCRIPTS, node,
2254 "user scripts not present or not executable: %s" %
2255 utils.CommaJoin(sorted(broken_scripts)))
2257 def _VerifyNodeNetwork(self, ninfo, nresult):
2258 """Check the node network connectivity results.
2260 @type ninfo: L{objects.Node}
2261 @param ninfo: the node to check
2262 @param nresult: the remote results for the node
2266 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2268 test = constants.NV_NODELIST not in nresult
2269 _ErrorIf(test, constants.CV_ENODESSH, node,
2270 "node hasn't returned node ssh connectivity data")
2272 if nresult[constants.NV_NODELIST]:
2273 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
2274 _ErrorIf(True, constants.CV_ENODESSH, node,
2275 "ssh communication with node '%s': %s", a_node, a_msg)
2277 test = constants.NV_NODENETTEST not in nresult
2278 _ErrorIf(test, constants.CV_ENODENET, node,
2279 "node hasn't returned node tcp connectivity data")
2281 if nresult[constants.NV_NODENETTEST]:
2282 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
2284 _ErrorIf(True, constants.CV_ENODENET, node,
2285 "tcp communication with node '%s': %s",
2286 anode, nresult[constants.NV_NODENETTEST][anode])
2288 test = constants.NV_MASTERIP not in nresult
2289 _ErrorIf(test, constants.CV_ENODENET, node,
2290 "node hasn't returned node master IP reachability data")
2292 if not nresult[constants.NV_MASTERIP]:
2293 if node == self.master_node:
2294 msg = "the master node cannot reach the master IP (not configured?)"
2296 msg = "cannot reach the master IP"
2297 _ErrorIf(True, constants.CV_ENODENET, node, msg)
2299 def _VerifyInstance(self, instance, instanceconfig, node_image,
2301 """Verify an instance.
2303 This function checks to see if the required block devices are
2304 available on the instance's node.
2307 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2308 node_current = instanceconfig.primary_node
2310 node_vol_should = {}
2311 instanceconfig.MapLVsByNode(node_vol_should)
2313 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(), self.group_info)
2314 err = _ComputeIPolicyInstanceViolation(ipolicy, instanceconfig)
2315 _ErrorIf(err, constants.CV_EINSTANCEPOLICY, instance, err)
2317 for node in node_vol_should:
2318 n_img = node_image[node]
2319 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2320 # ignore missing volumes on offline or broken nodes
2322 for volume in node_vol_should[node]:
2323 test = volume not in n_img.volumes
2324 _ErrorIf(test, constants.CV_EINSTANCEMISSINGDISK, instance,
2325 "volume %s missing on node %s", volume, node)
2327 if instanceconfig.admin_state == constants.ADMINST_UP:
2328 pri_img = node_image[node_current]
2329 test = instance not in pri_img.instances and not pri_img.offline
2330 _ErrorIf(test, constants.CV_EINSTANCEDOWN, instance,
2331 "instance not running on its primary node %s",
2334 diskdata = [(nname, success, status, idx)
2335 for (nname, disks) in diskstatus.items()
2336 for idx, (success, status) in enumerate(disks)]
2338 for nname, success, bdev_status, idx in diskdata:
2339 # the 'ghost node' construction in Exec() ensures that we have a
2341 snode = node_image[nname]
2342 bad_snode = snode.ghost or snode.offline
2343 _ErrorIf(instanceconfig.admin_state == constants.ADMINST_UP and
2344 not success and not bad_snode,
2345 constants.CV_EINSTANCEFAULTYDISK, instance,
2346 "couldn't retrieve status for disk/%s on %s: %s",
2347 idx, nname, bdev_status)
2348 _ErrorIf((instanceconfig.admin_state == constants.ADMINST_UP and
2349 success and bdev_status.ldisk_status == constants.LDS_FAULTY),
2350 constants.CV_EINSTANCEFAULTYDISK, instance,
2351 "disk/%s on %s is faulty", idx, nname)
2353 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
2354 """Verify if there are any unknown volumes in the cluster.
2356 The .os, .swap and backup volumes are ignored. All other volumes are
2357 reported as unknown.
2359 @type reserved: L{ganeti.utils.FieldSet}
2360 @param reserved: a FieldSet of reserved volume names
2363 for node, n_img in node_image.items():
2364 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2365 # skip non-healthy nodes
2367 for volume in n_img.volumes:
2368 test = ((node not in node_vol_should or
2369 volume not in node_vol_should[node]) and
2370 not reserved.Matches(volume))
2371 self._ErrorIf(test, constants.CV_ENODEORPHANLV, node,
2372 "volume %s is unknown", volume)
2374 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
2375 """Verify N+1 Memory Resilience.
2377 Check that if one single node dies we can still start all the
2378 instances it was primary for.
2381 cluster_info = self.cfg.GetClusterInfo()
2382 for node, n_img in node_image.items():
2383 # This code checks that every node which is now listed as
2384 # secondary has enough memory to host all instances it is
2385 # supposed to should a single other node in the cluster fail.
2386 # FIXME: not ready for failover to an arbitrary node
2387 # FIXME: does not support file-backed instances
2388 # WARNING: we currently take into account down instances as well
2389 # as up ones, considering that even if they're down someone
2390 # might want to start them even in the event of a node failure.
2392 # we're skipping offline nodes from the N+1 warning, since
2393 # most likely we don't have good memory infromation from them;
2394 # we already list instances living on such nodes, and that's
2397 #TODO(dynmem): use MINMEM for checking
2398 #TODO(dynmem): also consider ballooning out other instances
2399 for prinode, instances in n_img.sbp.items():
2401 for instance in instances:
2402 bep = cluster_info.FillBE(instance_cfg[instance])
2403 if bep[constants.BE_AUTO_BALANCE]:
2404 needed_mem += bep[constants.BE_MAXMEM]
2405 test = n_img.mfree < needed_mem
2406 self._ErrorIf(test, constants.CV_ENODEN1, node,
2407 "not enough memory to accomodate instance failovers"
2408 " should node %s fail (%dMiB needed, %dMiB available)",
2409 prinode, needed_mem, n_img.mfree)
2412 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
2413 (files_all, files_opt, files_mc, files_vm)):
2414 """Verifies file checksums collected from all nodes.
2416 @param errorif: Callback for reporting errors
2417 @param nodeinfo: List of L{objects.Node} objects
2418 @param master_node: Name of master node
2419 @param all_nvinfo: RPC results
2422 # Define functions determining which nodes to consider for a file
2425 (files_mc, lambda node: (node.master_candidate or
2426 node.name == master_node)),
2427 (files_vm, lambda node: node.vm_capable),
2430 # Build mapping from filename to list of nodes which should have the file
2432 for (files, fn) in files2nodefn:
2434 filenodes = nodeinfo
2436 filenodes = filter(fn, nodeinfo)
2437 nodefiles.update((filename,
2438 frozenset(map(operator.attrgetter("name"), filenodes)))
2439 for filename in files)
2441 assert set(nodefiles) == (files_all | files_mc | files_vm)
2443 fileinfo = dict((filename, {}) for filename in nodefiles)
2444 ignore_nodes = set()
2446 for node in nodeinfo:
2448 ignore_nodes.add(node.name)
2451 nresult = all_nvinfo[node.name]
2453 if nresult.fail_msg or not nresult.payload:
2456 node_files = nresult.payload.get(constants.NV_FILELIST, None)
2458 test = not (node_files and isinstance(node_files, dict))
2459 errorif(test, constants.CV_ENODEFILECHECK, node.name,
2460 "Node did not return file checksum data")
2462 ignore_nodes.add(node.name)
2465 # Build per-checksum mapping from filename to nodes having it
2466 for (filename, checksum) in node_files.items():
2467 assert filename in nodefiles
2468 fileinfo[filename].setdefault(checksum, set()).add(node.name)
2470 for (filename, checksums) in fileinfo.items():
2471 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
2473 # Nodes having the file
2474 with_file = frozenset(node_name
2475 for nodes in fileinfo[filename].values()
2476 for node_name in nodes) - ignore_nodes
2478 expected_nodes = nodefiles[filename] - ignore_nodes
2480 # Nodes missing file
2481 missing_file = expected_nodes - with_file
2483 if filename in files_opt:
2485 errorif(missing_file and missing_file != expected_nodes,
2486 constants.CV_ECLUSTERFILECHECK, None,
2487 "File %s is optional, but it must exist on all or no"
2488 " nodes (not found on %s)",
2489 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
2491 errorif(missing_file, constants.CV_ECLUSTERFILECHECK, None,
2492 "File %s is missing from node(s) %s", filename,
2493 utils.CommaJoin(utils.NiceSort(missing_file)))
2495 # Warn if a node has a file it shouldn't
2496 unexpected = with_file - expected_nodes
2498 constants.CV_ECLUSTERFILECHECK, None,
2499 "File %s should not exist on node(s) %s",
2500 filename, utils.CommaJoin(utils.NiceSort(unexpected)))
2502 # See if there are multiple versions of the file
2503 test = len(checksums) > 1
2505 variants = ["variant %s on %s" %
2506 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
2507 for (idx, (checksum, nodes)) in
2508 enumerate(sorted(checksums.items()))]
2512 errorif(test, constants.CV_ECLUSTERFILECHECK, None,
2513 "File %s found with %s different checksums (%s)",
2514 filename, len(checksums), "; ".join(variants))
2516 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
2518 """Verifies and the node DRBD status.
2520 @type ninfo: L{objects.Node}
2521 @param ninfo: the node to check
2522 @param nresult: the remote results for the node
2523 @param instanceinfo: the dict of instances
2524 @param drbd_helper: the configured DRBD usermode helper
2525 @param drbd_map: the DRBD map as returned by
2526 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
2530 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2533 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
2534 test = (helper_result == None)
2535 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2536 "no drbd usermode helper returned")
2538 status, payload = helper_result
2540 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2541 "drbd usermode helper check unsuccessful: %s", payload)
2542 test = status and (payload != drbd_helper)
2543 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2544 "wrong drbd usermode helper: %s", payload)
2546 # compute the DRBD minors
2548 for minor, instance in drbd_map[node].items():
2549 test = instance not in instanceinfo
2550 _ErrorIf(test, constants.CV_ECLUSTERCFG, None,
2551 "ghost instance '%s' in temporary DRBD map", instance)
2552 # ghost instance should not be running, but otherwise we
2553 # don't give double warnings (both ghost instance and
2554 # unallocated minor in use)
2556 node_drbd[minor] = (instance, False)
2558 instance = instanceinfo[instance]
2559 node_drbd[minor] = (instance.name,
2560 instance.admin_state == constants.ADMINST_UP)
2562 # and now check them
2563 used_minors = nresult.get(constants.NV_DRBDLIST, [])
2564 test = not isinstance(used_minors, (tuple, list))
2565 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2566 "cannot parse drbd status file: %s", str(used_minors))
2568 # we cannot check drbd status
2571 for minor, (iname, must_exist) in node_drbd.items():
2572 test = minor not in used_minors and must_exist
2573 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2574 "drbd minor %d of instance %s is not active", minor, iname)
2575 for minor in used_minors:
2576 test = minor not in node_drbd
2577 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2578 "unallocated drbd minor %d is in use", minor)
2580 def _UpdateNodeOS(self, ninfo, nresult, nimg):
2581 """Builds the node OS structures.
2583 @type ninfo: L{objects.Node}
2584 @param ninfo: the node to check
2585 @param nresult: the remote results for the node
2586 @param nimg: the node image object
2590 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2592 remote_os = nresult.get(constants.NV_OSLIST, None)
2593 test = (not isinstance(remote_os, list) or
2594 not compat.all(isinstance(v, list) and len(v) == 7
2595 for v in remote_os))
2597 _ErrorIf(test, constants.CV_ENODEOS, node,
2598 "node hasn't returned valid OS data")
2607 for (name, os_path, status, diagnose,
2608 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
2610 if name not in os_dict:
2613 # parameters is a list of lists instead of list of tuples due to
2614 # JSON lacking a real tuple type, fix it:
2615 parameters = [tuple(v) for v in parameters]
2616 os_dict[name].append((os_path, status, diagnose,
2617 set(variants), set(parameters), set(api_ver)))
2619 nimg.oslist = os_dict
2621 def _VerifyNodeOS(self, ninfo, nimg, base):
2622 """Verifies the node OS list.
2624 @type ninfo: L{objects.Node}
2625 @param ninfo: the node to check
2626 @param nimg: the node image object
2627 @param base: the 'template' node we match against (e.g. from the master)
2631 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2633 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
2635 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
2636 for os_name, os_data in nimg.oslist.items():
2637 assert os_data, "Empty OS status for OS %s?!" % os_name
2638 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
2639 _ErrorIf(not f_status, constants.CV_ENODEOS, node,
2640 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
2641 _ErrorIf(len(os_data) > 1, constants.CV_ENODEOS, node,
2642 "OS '%s' has multiple entries (first one shadows the rest): %s",
2643 os_name, utils.CommaJoin([v[0] for v in os_data]))
2644 # comparisons with the 'base' image
2645 test = os_name not in base.oslist
2646 _ErrorIf(test, constants.CV_ENODEOS, node,
2647 "Extra OS %s not present on reference node (%s)",
2651 assert base.oslist[os_name], "Base node has empty OS status?"
2652 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
2654 # base OS is invalid, skipping
2656 for kind, a, b in [("API version", f_api, b_api),
2657 ("variants list", f_var, b_var),
2658 ("parameters", beautify_params(f_param),
2659 beautify_params(b_param))]:
2660 _ErrorIf(a != b, constants.CV_ENODEOS, node,
2661 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
2662 kind, os_name, base.name,
2663 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
2665 # check any missing OSes
2666 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
2667 _ErrorIf(missing, constants.CV_ENODEOS, node,
2668 "OSes present on reference node %s but missing on this node: %s",
2669 base.name, utils.CommaJoin(missing))
2671 def _VerifyOob(self, ninfo, nresult):
2672 """Verifies out of band functionality of a node.
2674 @type ninfo: L{objects.Node}
2675 @param ninfo: the node to check
2676 @param nresult: the remote results for the node
2680 # We just have to verify the paths on master and/or master candidates
2681 # as the oob helper is invoked on the master
2682 if ((ninfo.master_candidate or ninfo.master_capable) and
2683 constants.NV_OOB_PATHS in nresult):
2684 for path_result in nresult[constants.NV_OOB_PATHS]:
2685 self._ErrorIf(path_result, constants.CV_ENODEOOBPATH, node, path_result)
2687 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
2688 """Verifies and updates the node volume data.
2690 This function will update a L{NodeImage}'s internal structures
2691 with data from the remote call.
2693 @type ninfo: L{objects.Node}
2694 @param ninfo: the node to check
2695 @param nresult: the remote results for the node
2696 @param nimg: the node image object
2697 @param vg_name: the configured VG name
2701 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2703 nimg.lvm_fail = True
2704 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2707 elif isinstance(lvdata, basestring):
2708 _ErrorIf(True, constants.CV_ENODELVM, node, "LVM problem on node: %s",
2709 utils.SafeEncode(lvdata))
2710 elif not isinstance(lvdata, dict):
2711 _ErrorIf(True, constants.CV_ENODELVM, node,
2712 "rpc call to node failed (lvlist)")
2714 nimg.volumes = lvdata
2715 nimg.lvm_fail = False
2717 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2718 """Verifies and updates the node instance list.
2720 If the listing was successful, then updates this node's instance
2721 list. Otherwise, it marks the RPC call as failed for the instance
2724 @type ninfo: L{objects.Node}
2725 @param ninfo: the node to check
2726 @param nresult: the remote results for the node
2727 @param nimg: the node image object
2730 idata = nresult.get(constants.NV_INSTANCELIST, None)
2731 test = not isinstance(idata, list)
2732 self._ErrorIf(test, constants.CV_ENODEHV, ninfo.name,
2733 "rpc call to node failed (instancelist): %s",
2734 utils.SafeEncode(str(idata)))
2736 nimg.hyp_fail = True
2738 nimg.instances = idata
2740 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2741 """Verifies and computes a node information map
2743 @type ninfo: L{objects.Node}
2744 @param ninfo: the node to check
2745 @param nresult: the remote results for the node
2746 @param nimg: the node image object
2747 @param vg_name: the configured VG name
2751 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2753 # try to read free memory (from the hypervisor)
2754 hv_info = nresult.get(constants.NV_HVINFO, None)
2755 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2756 _ErrorIf(test, constants.CV_ENODEHV, node,
2757 "rpc call to node failed (hvinfo)")
2760 nimg.mfree = int(hv_info["memory_free"])
2761 except (ValueError, TypeError):
2762 _ErrorIf(True, constants.CV_ENODERPC, node,
2763 "node returned invalid nodeinfo, check hypervisor")
2765 # FIXME: devise a free space model for file based instances as well
2766 if vg_name is not None:
2767 test = (constants.NV_VGLIST not in nresult or
2768 vg_name not in nresult[constants.NV_VGLIST])
2769 _ErrorIf(test, constants.CV_ENODELVM, node,
2770 "node didn't return data for the volume group '%s'"
2771 " - it is either missing or broken", vg_name)
2774 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2775 except (ValueError, TypeError):
2776 _ErrorIf(True, constants.CV_ENODERPC, node,
2777 "node returned invalid LVM info, check LVM status")
2779 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2780 """Gets per-disk status information for all instances.
2782 @type nodelist: list of strings
2783 @param nodelist: Node names
2784 @type node_image: dict of (name, L{objects.Node})
2785 @param node_image: Node objects
2786 @type instanceinfo: dict of (name, L{objects.Instance})
2787 @param instanceinfo: Instance objects
2788 @rtype: {instance: {node: [(succes, payload)]}}
2789 @return: a dictionary of per-instance dictionaries with nodes as
2790 keys and disk information as values; the disk information is a
2791 list of tuples (success, payload)
2794 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2797 node_disks_devonly = {}
2798 diskless_instances = set()
2799 diskless = constants.DT_DISKLESS
2801 for nname in nodelist:
2802 node_instances = list(itertools.chain(node_image[nname].pinst,
2803 node_image[nname].sinst))
2804 diskless_instances.update(inst for inst in node_instances
2805 if instanceinfo[inst].disk_template == diskless)
2806 disks = [(inst, disk)
2807 for inst in node_instances
2808 for disk in instanceinfo[inst].disks]
2811 # No need to collect data
2814 node_disks[nname] = disks
2816 # Creating copies as SetDiskID below will modify the objects and that can
2817 # lead to incorrect data returned from nodes
2818 devonly = [dev.Copy() for (_, dev) in disks]
2821 self.cfg.SetDiskID(dev, nname)
2823 node_disks_devonly[nname] = devonly
2825 assert len(node_disks) == len(node_disks_devonly)
2827 # Collect data from all nodes with disks
2828 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2831 assert len(result) == len(node_disks)
2835 for (nname, nres) in result.items():
2836 disks = node_disks[nname]
2839 # No data from this node
2840 data = len(disks) * [(False, "node offline")]
2843 _ErrorIf(msg, constants.CV_ENODERPC, nname,
2844 "while getting disk information: %s", msg)
2846 # No data from this node
2847 data = len(disks) * [(False, msg)]
2850 for idx, i in enumerate(nres.payload):
2851 if isinstance(i, (tuple, list)) and len(i) == 2:
2854 logging.warning("Invalid result from node %s, entry %d: %s",
2856 data.append((False, "Invalid result from the remote node"))
2858 for ((inst, _), status) in zip(disks, data):
2859 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2861 # Add empty entries for diskless instances.
2862 for inst in diskless_instances:
2863 assert inst not in instdisk
2866 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2867 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2868 compat.all(isinstance(s, (tuple, list)) and
2869 len(s) == 2 for s in statuses)
2870 for inst, nnames in instdisk.items()
2871 for nname, statuses in nnames.items())
2872 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2877 def _SshNodeSelector(group_uuid, all_nodes):
2878 """Create endless iterators for all potential SSH check hosts.
2881 nodes = [node for node in all_nodes
2882 if (node.group != group_uuid and
2884 keyfunc = operator.attrgetter("group")
2886 return map(itertools.cycle,
2887 [sorted(map(operator.attrgetter("name"), names))
2888 for _, names in itertools.groupby(sorted(nodes, key=keyfunc),
2892 def _SelectSshCheckNodes(cls, group_nodes, group_uuid, all_nodes):
2893 """Choose which nodes should talk to which other nodes.
2895 We will make nodes contact all nodes in their group, and one node from
2898 @warning: This algorithm has a known issue if one node group is much
2899 smaller than others (e.g. just one node). In such a case all other
2900 nodes will talk to the single node.
2903 online_nodes = sorted(node.name for node in group_nodes if not node.offline)
2904 sel = cls._SshNodeSelector(group_uuid, all_nodes)
2906 return (online_nodes,
2907 dict((name, sorted([i.next() for i in sel]))
2908 for name in online_nodes))
2910 def BuildHooksEnv(self):
2913 Cluster-Verify hooks just ran in the post phase and their failure makes
2914 the output be logged in the verify output and the verification to fail.
2918 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2921 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
2922 for node in self.my_node_info.values())
2926 def BuildHooksNodes(self):
2927 """Build hooks nodes.
2930 return ([], self.my_node_names)
2932 def Exec(self, feedback_fn):
2933 """Verify integrity of the node group, performing various test on nodes.
2936 # This method has too many local variables. pylint: disable=R0914
2937 feedback_fn("* Verifying group '%s'" % self.group_info.name)
2939 if not self.my_node_names:
2941 feedback_fn("* Empty node group, skipping verification")
2945 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2946 verbose = self.op.verbose
2947 self._feedback_fn = feedback_fn
2949 vg_name = self.cfg.GetVGName()
2950 drbd_helper = self.cfg.GetDRBDHelper()
2951 cluster = self.cfg.GetClusterInfo()
2952 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2953 hypervisors = cluster.enabled_hypervisors
2954 node_data_list = [self.my_node_info[name] for name in self.my_node_names]
2956 i_non_redundant = [] # Non redundant instances
2957 i_non_a_balanced = [] # Non auto-balanced instances
2958 i_offline = 0 # Count of offline instances
2959 n_offline = 0 # Count of offline nodes
2960 n_drained = 0 # Count of nodes being drained
2961 node_vol_should = {}
2963 # FIXME: verify OS list
2966 filemap = _ComputeAncillaryFiles(cluster, False)
2968 # do local checksums
2969 master_node = self.master_node = self.cfg.GetMasterNode()
2970 master_ip = self.cfg.GetMasterIP()
2972 feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
2975 if self.cfg.GetUseExternalMipScript():
2976 user_scripts.append(constants.EXTERNAL_MASTER_SETUP_SCRIPT)
2978 node_verify_param = {
2979 constants.NV_FILELIST:
2980 utils.UniqueSequence(filename
2981 for files in filemap
2982 for filename in files),
2983 constants.NV_NODELIST:
2984 self._SelectSshCheckNodes(node_data_list, self.group_uuid,
2985 self.all_node_info.values()),
2986 constants.NV_HYPERVISOR: hypervisors,
2987 constants.NV_HVPARAMS:
2988 _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
2989 constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip)
2990 for node in node_data_list
2991 if not node.offline],
2992 constants.NV_INSTANCELIST: hypervisors,
2993 constants.NV_VERSION: None,
2994 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2995 constants.NV_NODESETUP: None,
2996 constants.NV_TIME: None,
2997 constants.NV_MASTERIP: (master_node, master_ip),
2998 constants.NV_OSLIST: None,
2999 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
3000 constants.NV_USERSCRIPTS: user_scripts,
3003 if vg_name is not None:
3004 node_verify_param[constants.NV_VGLIST] = None
3005 node_verify_param[constants.NV_LVLIST] = vg_name
3006 node_verify_param[constants.NV_PVLIST] = [vg_name]
3007 node_verify_param[constants.NV_DRBDLIST] = None
3010 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
3013 # FIXME: this needs to be changed per node-group, not cluster-wide
3015 default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
3016 if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
3017 bridges.add(default_nicpp[constants.NIC_LINK])
3018 for instance in self.my_inst_info.values():
3019 for nic in instance.nics:
3020 full_nic = cluster.SimpleFillNIC(nic.nicparams)
3021 if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
3022 bridges.add(full_nic[constants.NIC_LINK])
3025 node_verify_param[constants.NV_BRIDGES] = list(bridges)
3027 # Build our expected cluster state
3028 node_image = dict((node.name, self.NodeImage(offline=node.offline,
3030 vm_capable=node.vm_capable))
3031 for node in node_data_list)
3035 for node in self.all_node_info.values():
3036 path = _SupportsOob(self.cfg, node)
3037 if path and path not in oob_paths:
3038 oob_paths.append(path)
3041 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
3043 for instance in self.my_inst_names:
3044 inst_config = self.my_inst_info[instance]
3046 for nname in inst_config.all_nodes:
3047 if nname not in node_image:
3048 gnode = self.NodeImage(name=nname)
3049 gnode.ghost = (nname not in self.all_node_info)
3050 node_image[nname] = gnode
3052 inst_config.MapLVsByNode(node_vol_should)
3054 pnode = inst_config.primary_node
3055 node_image[pnode].pinst.append(instance)
3057 for snode in inst_config.secondary_nodes:
3058 nimg = node_image[snode]
3059 nimg.sinst.append(instance)
3060 if pnode not in nimg.sbp:
3061 nimg.sbp[pnode] = []
3062 nimg.sbp[pnode].append(instance)
3064 # At this point, we have the in-memory data structures complete,
3065 # except for the runtime information, which we'll gather next
3067 # Due to the way our RPC system works, exact response times cannot be
3068 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
3069 # time before and after executing the request, we can at least have a time
3071 nvinfo_starttime = time.time()
3072 all_nvinfo = self.rpc.call_node_verify(self.my_node_names,
3074 self.cfg.GetClusterName())
3075 nvinfo_endtime = time.time()
3077 if self.extra_lv_nodes and vg_name is not None:
3079 self.rpc.call_node_verify(self.extra_lv_nodes,
3080 {constants.NV_LVLIST: vg_name},
3081 self.cfg.GetClusterName())
3083 extra_lv_nvinfo = {}
3085 all_drbd_map = self.cfg.ComputeDRBDMap()
3087 feedback_fn("* Gathering disk information (%s nodes)" %
3088 len(self.my_node_names))
3089 instdisk = self._CollectDiskInfo(self.my_node_names, node_image,
3092 feedback_fn("* Verifying configuration file consistency")
3094 # If not all nodes are being checked, we need to make sure the master node
3095 # and a non-checked vm_capable node are in the list.
3096 absent_nodes = set(self.all_node_info).difference(self.my_node_info)
3098 vf_nvinfo = all_nvinfo.copy()
3099 vf_node_info = list(self.my_node_info.values())
3100 additional_nodes = []
3101 if master_node not in self.my_node_info:
3102 additional_nodes.append(master_node)
3103 vf_node_info.append(self.all_node_info[master_node])
3104 # Add the first vm_capable node we find which is not included
3105 for node in absent_nodes:
3106 nodeinfo = self.all_node_info[node]
3107 if nodeinfo.vm_capable and not nodeinfo.offline:
3108 additional_nodes.append(node)
3109 vf_node_info.append(self.all_node_info[node])
3111 key = constants.NV_FILELIST
3112 vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes,
3113 {key: node_verify_param[key]},
3114 self.cfg.GetClusterName()))
3116 vf_nvinfo = all_nvinfo
3117 vf_node_info = self.my_node_info.values()
3119 self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap)
3121 feedback_fn("* Verifying node status")
3125 for node_i in node_data_list:
3127 nimg = node_image[node]
3131 feedback_fn("* Skipping offline node %s" % (node,))
3135 if node == master_node:
3137 elif node_i.master_candidate:
3138 ntype = "master candidate"
3139 elif node_i.drained:
3145 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
3147 msg = all_nvinfo[node].fail_msg
3148 _ErrorIf(msg, constants.CV_ENODERPC, node, "while contacting node: %s",
3151 nimg.rpc_fail = True
3154 nresult = all_nvinfo[node].payload
3156 nimg.call_ok = self._VerifyNode(node_i, nresult)
3157 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
3158 self._VerifyNodeNetwork(node_i, nresult)
3159 self._VerifyNodeUserScripts(node_i, nresult)
3160 self._VerifyOob(node_i, nresult)
3163 self._VerifyNodeLVM(node_i, nresult, vg_name)
3164 self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper,
3167 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
3168 self._UpdateNodeInstances(node_i, nresult, nimg)
3169 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
3170 self._UpdateNodeOS(node_i, nresult, nimg)
3172 if not nimg.os_fail:
3173 if refos_img is None:
3175 self._VerifyNodeOS(node_i, nimg, refos_img)
3176 self._VerifyNodeBridges(node_i, nresult, bridges)
3178 # Check whether all running instancies are primary for the node. (This
3179 # can no longer be done from _VerifyInstance below, since some of the
3180 # wrong instances could be from other node groups.)
3181 non_primary_inst = set(nimg.instances).difference(nimg.pinst)
3183 for inst in non_primary_inst:
3184 # FIXME: investigate best way to handle offline insts
3185 if inst.admin_state == constants.ADMINST_OFFLINE:
3187 feedback_fn("* Skipping offline instance %s" % inst.name)
3190 test = inst in self.all_inst_info
3191 _ErrorIf(test, constants.CV_EINSTANCEWRONGNODE, inst,
3192 "instance should not run on node %s", node_i.name)
3193 _ErrorIf(not test, constants.CV_ENODEORPHANINSTANCE, node_i.name,
3194 "node is running unknown instance %s", inst)
3196 for node, result in extra_lv_nvinfo.items():
3197 self._UpdateNodeVolumes(self.all_node_info[node], result.payload,
3198 node_image[node], vg_name)
3200 feedback_fn("* Verifying instance status")
3201 for instance in self.my_inst_names:
3203 feedback_fn("* Verifying instance %s" % instance)
3204 inst_config = self.my_inst_info[instance]
3205 self._VerifyInstance(instance, inst_config, node_image,
3207 inst_nodes_offline = []
3209 pnode = inst_config.primary_node
3210 pnode_img = node_image[pnode]
3211 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
3212 constants.CV_ENODERPC, pnode, "instance %s, connection to"
3213 " primary node failed", instance)
3215 _ErrorIf(inst_config.admin_state == constants.ADMINST_UP and
3217 constants.CV_EINSTANCEBADNODE, instance,
3218 "instance is marked as running and lives on offline node %s",
3219 inst_config.primary_node)
3221 # If the instance is non-redundant we cannot survive losing its primary
3222 # node, so we are not N+1 compliant. On the other hand we have no disk
3223 # templates with more than one secondary so that situation is not well
3225 # FIXME: does not support file-backed instances
3226 if not inst_config.secondary_nodes:
3227 i_non_redundant.append(instance)
3229 _ErrorIf(len(inst_config.secondary_nodes) > 1,
3230 constants.CV_EINSTANCELAYOUT,
3231 instance, "instance has multiple secondary nodes: %s",
3232 utils.CommaJoin(inst_config.secondary_nodes),
3233 code=self.ETYPE_WARNING)
3235 if inst_config.disk_template in constants.DTS_INT_MIRROR:
3236 pnode = inst_config.primary_node
3237 instance_nodes = utils.NiceSort(inst_config.all_nodes)
3238 instance_groups = {}
3240 for node in instance_nodes:
3241 instance_groups.setdefault(self.all_node_info[node].group,
3245 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
3246 # Sort so that we always list the primary node first.
3247 for group, nodes in sorted(instance_groups.items(),
3248 key=lambda (_, nodes): pnode in nodes,
3251 self._ErrorIf(len(instance_groups) > 1,
3252 constants.CV_EINSTANCESPLITGROUPS,
3253 instance, "instance has primary and secondary nodes in"
3254 " different groups: %s", utils.CommaJoin(pretty_list),
3255 code=self.ETYPE_WARNING)
3257 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
3258 i_non_a_balanced.append(instance)
3260 for snode in inst_config.secondary_nodes:
3261 s_img = node_image[snode]
3262 _ErrorIf(s_img.rpc_fail and not s_img.offline, constants.CV_ENODERPC,
3263 snode, "instance %s, connection to secondary node failed",
3267 inst_nodes_offline.append(snode)
3269 # warn that the instance lives on offline nodes
3270 _ErrorIf(inst_nodes_offline, constants.CV_EINSTANCEBADNODE, instance,
3271 "instance has offline secondary node(s) %s",
3272 utils.CommaJoin(inst_nodes_offline))
3273 # ... or ghost/non-vm_capable nodes
3274 for node in inst_config.all_nodes:
3275 _ErrorIf(node_image[node].ghost, constants.CV_EINSTANCEBADNODE,
3276 instance, "instance lives on ghost node %s", node)
3277 _ErrorIf(not node_image[node].vm_capable, constants.CV_EINSTANCEBADNODE,
3278 instance, "instance lives on non-vm_capable node %s", node)
3280 feedback_fn("* Verifying orphan volumes")
3281 reserved = utils.FieldSet(*cluster.reserved_lvs)
3283 # We will get spurious "unknown volume" warnings if any node of this group
3284 # is secondary for an instance whose primary is in another group. To avoid
3285 # them, we find these instances and add their volumes to node_vol_should.
3286 for inst in self.all_inst_info.values():
3287 for secondary in inst.secondary_nodes:
3288 if (secondary in self.my_node_info
3289 and inst.name not in self.my_inst_info):
3290 inst.MapLVsByNode(node_vol_should)
3293 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
3295 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
3296 feedback_fn("* Verifying N+1 Memory redundancy")
3297 self._VerifyNPlusOneMemory(node_image, self.my_inst_info)
3299 feedback_fn("* Other Notes")
3301 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
3302 % len(i_non_redundant))
3304 if i_non_a_balanced:
3305 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
3306 % len(i_non_a_balanced))
3309 feedback_fn(" - NOTICE: %d offline instance(s) found." % i_offline)
3312 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
3315 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
3319 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
3320 """Analyze the post-hooks' result
3322 This method analyses the hook result, handles it, and sends some
3323 nicely-formatted feedback back to the user.
3325 @param phase: one of L{constants.HOOKS_PHASE_POST} or
3326 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
3327 @param hooks_results: the results of the multi-node hooks rpc call
3328 @param feedback_fn: function used send feedback back to the caller
3329 @param lu_result: previous Exec result
3330 @return: the new Exec result, based on the previous result
3334 # We only really run POST phase hooks, only for non-empty groups,
3335 # and are only interested in their results
3336 if not self.my_node_names:
3339 elif phase == constants.HOOKS_PHASE_POST:
3340 # Used to change hooks' output to proper indentation
3341 feedback_fn("* Hooks Results")
3342 assert hooks_results, "invalid result from hooks"
3344 for node_name in hooks_results:
3345 res = hooks_results[node_name]
3347 test = msg and not res.offline
3348 self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,
3349 "Communication failure in hooks execution: %s", msg)
3350 if res.offline or msg:
3351 # No need to investigate payload if node is offline or gave
3354 for script, hkr, output in res.payload:
3355 test = hkr == constants.HKR_FAIL
3356 self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,
3357 "Script %s failed, output:", script)
3359 output = self._HOOKS_INDENT_RE.sub(" ", output)
3360 feedback_fn("%s" % output)
3366 class LUClusterVerifyDisks(NoHooksLU):
3367 """Verifies the cluster disks status.
3372 def ExpandNames(self):
3373 self.share_locks = _ShareAll()
3374 self.needed_locks = {
3375 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3378 def Exec(self, feedback_fn):
3379 group_names = self.owned_locks(locking.LEVEL_NODEGROUP)
3381 # Submit one instance of L{opcodes.OpGroupVerifyDisks} per node group
3382 return ResultWithJobs([[opcodes.OpGroupVerifyDisks(group_name=group)]
3383 for group in group_names])
3386 class LUGroupVerifyDisks(NoHooksLU):
3387 """Verifies the status of all disks in a node group.
3392 def ExpandNames(self):
3393 # Raises errors.OpPrereqError on its own if group can't be found
3394 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
3396 self.share_locks = _ShareAll()
3397 self.needed_locks = {
3398 locking.LEVEL_INSTANCE: [],
3399 locking.LEVEL_NODEGROUP: [],
3400 locking.LEVEL_NODE: [],
3403 def DeclareLocks(self, level):
3404 if level == locking.LEVEL_INSTANCE:
3405 assert not self.needed_locks[locking.LEVEL_INSTANCE]
3407 # Lock instances optimistically, needs verification once node and group
3408 # locks have been acquired
3409 self.needed_locks[locking.LEVEL_INSTANCE] = \
3410 self.cfg.GetNodeGroupInstances(self.group_uuid)
3412 elif level == locking.LEVEL_NODEGROUP:
3413 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
3415 self.needed_locks[locking.LEVEL_NODEGROUP] = \
3416 set([self.group_uuid] +
3417 # Lock all groups used by instances optimistically; this requires
3418 # going via the node before it's locked, requiring verification
3421 for instance_name in self.owned_locks(locking.LEVEL_INSTANCE)
3422 for group_uuid in self.cfg.GetInstanceNodeGroups(instance_name)])
3424 elif level == locking.LEVEL_NODE:
3425 # This will only lock the nodes in the group to be verified which contain
3427 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
3428 self._LockInstancesNodes()
3430 # Lock all nodes in group to be verified
3431 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
3432 member_nodes = self.cfg.GetNodeGroup(self.group_uuid).members
3433 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
3435 def CheckPrereq(self):
3436 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
3437 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
3438 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
3440 assert self.group_uuid in owned_groups
3442 # Check if locked instances are still correct
3443 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
3445 # Get instance information
3446 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
3448 # Check if node groups for locked instances are still correct
3449 for (instance_name, inst) in self.instances.items():
3450 assert owned_nodes.issuperset(inst.all_nodes), \
3451 "Instance %s's nodes changed while we kept the lock" % instance_name
3453 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
3456 assert self.group_uuid in inst_groups, \
3457 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
3459 def Exec(self, feedback_fn):
3460 """Verify integrity of cluster disks.
3462 @rtype: tuple of three items
3463 @return: a tuple of (dict of node-to-node_error, list of instances
3464 which need activate-disks, dict of instance: (node, volume) for
3469 res_instances = set()
3472 nv_dict = _MapInstanceDisksToNodes([inst
3473 for inst in self.instances.values()
3474 if inst.admin_state == constants.ADMINST_UP])
3477 nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &
3478 set(self.cfg.GetVmCapableNodeList()))
3480 node_lvs = self.rpc.call_lv_list(nodes, [])
3482 for (node, node_res) in node_lvs.items():
3483 if node_res.offline:
3486 msg = node_res.fail_msg
3488 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
3489 res_nodes[node] = msg
3492 for lv_name, (_, _, lv_online) in node_res.payload.items():
3493 inst = nv_dict.pop((node, lv_name), None)
3494 if not (lv_online or inst is None):
3495 res_instances.add(inst)
3497 # any leftover items in nv_dict are missing LVs, let's arrange the data
3499 for key, inst in nv_dict.iteritems():
3500 res_missing.setdefault(inst, []).append(list(key))
3502 return (res_nodes, list(res_instances), res_missing)
3505 class LUClusterRepairDiskSizes(NoHooksLU):
3506 """Verifies the cluster disks sizes.
3511 def ExpandNames(self):
3512 if self.op.instances:
3513 self.wanted_names = _GetWantedInstances(self, self.op.instances)
3514 self.needed_locks = {
3515 locking.LEVEL_NODE_RES: [],
3516 locking.LEVEL_INSTANCE: self.wanted_names,
3518 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
3520 self.wanted_names = None
3521 self.needed_locks = {
3522 locking.LEVEL_NODE_RES: locking.ALL_SET,
3523 locking.LEVEL_INSTANCE: locking.ALL_SET,
3525 self.share_locks = {
3526 locking.LEVEL_NODE_RES: 1,
3527 locking.LEVEL_INSTANCE: 0,
3530 def DeclareLocks(self, level):
3531 if level == locking.LEVEL_NODE_RES and self.wanted_names is not None:
3532 self._LockInstancesNodes(primary_only=True, level=level)
3534 def CheckPrereq(self):
3535 """Check prerequisites.
3537 This only checks the optional instance list against the existing names.
3540 if self.wanted_names is None:
3541 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
3543 self.wanted_instances = \
3544 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
3546 def _EnsureChildSizes(self, disk):
3547 """Ensure children of the disk have the needed disk size.
3549 This is valid mainly for DRBD8 and fixes an issue where the
3550 children have smaller disk size.
3552 @param disk: an L{ganeti.objects.Disk} object
3555 if disk.dev_type == constants.LD_DRBD8:
3556 assert disk.children, "Empty children for DRBD8?"
3557 fchild = disk.children[0]
3558 mismatch = fchild.size < disk.size
3560 self.LogInfo("Child disk has size %d, parent %d, fixing",
3561 fchild.size, disk.size)
3562 fchild.size = disk.size
3564 # and we recurse on this child only, not on the metadev
3565 return self._EnsureChildSizes(fchild) or mismatch
3569 def Exec(self, feedback_fn):
3570 """Verify the size of cluster disks.
3573 # TODO: check child disks too
3574 # TODO: check differences in size between primary/secondary nodes
3576 for instance in self.wanted_instances:
3577 pnode = instance.primary_node
3578 if pnode not in per_node_disks:
3579 per_node_disks[pnode] = []
3580 for idx, disk in enumerate(instance.disks):
3581 per_node_disks[pnode].append((instance, idx, disk))
3583 assert not (frozenset(per_node_disks.keys()) -
3584 self.owned_locks(locking.LEVEL_NODE_RES)), \
3585 "Not owning correct locks"
3586 assert not self.owned_locks(locking.LEVEL_NODE)
3589 for node, dskl in per_node_disks.items():
3590 newl = [v[2].Copy() for v in dskl]
3592 self.cfg.SetDiskID(dsk, node)
3593 result = self.rpc.call_blockdev_getsize(node, newl)
3595 self.LogWarning("Failure in blockdev_getsize call to node"
3596 " %s, ignoring", node)
3598 if len(result.payload) != len(dskl):
3599 logging.warning("Invalid result from node %s: len(dksl)=%d,"
3600 " result.payload=%s", node, len(dskl), result.payload)
3601 self.LogWarning("Invalid result from node %s, ignoring node results",
3604 for ((instance, idx, disk), size) in zip(dskl, result.payload):
3606 self.LogWarning("Disk %d of instance %s did not return size"
3607 " information, ignoring", idx, instance.name)
3609 if not isinstance(size, (int, long)):
3610 self.LogWarning("Disk %d of instance %s did not return valid"
3611 " size information, ignoring", idx, instance.name)
3614 if size != disk.size:
3615 self.LogInfo("Disk %d of instance %s has mismatched size,"
3616 " correcting: recorded %d, actual %d", idx,
3617 instance.name, disk.size, size)
3619 self.cfg.Update(instance, feedback_fn)
3620 changed.append((instance.name, idx, size))
3621 if self._EnsureChildSizes(disk):
3622 self.cfg.Update(instance, feedback_fn)
3623 changed.append((instance.name, idx, disk.size))
3627 class LUClusterRename(LogicalUnit):
3628 """Rename the cluster.
3631 HPATH = "cluster-rename"
3632 HTYPE = constants.HTYPE_CLUSTER
3634 def BuildHooksEnv(self):
3639 "OP_TARGET": self.cfg.GetClusterName(),
3640 "NEW_NAME": self.op.name,
3643 def BuildHooksNodes(self):
3644 """Build hooks nodes.
3647 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
3649 def CheckPrereq(self):
3650 """Verify that the passed name is a valid one.
3653 hostname = netutils.GetHostname(name=self.op.name,
3654 family=self.cfg.GetPrimaryIPFamily())
3656 new_name = hostname.name
3657 self.ip = new_ip = hostname.ip
3658 old_name = self.cfg.GetClusterName()
3659 old_ip = self.cfg.GetMasterIP()
3660 if new_name == old_name and new_ip == old_ip:
3661 raise errors.OpPrereqError("Neither the name nor the IP address of the"
3662 " cluster has changed",
3664 if new_ip != old_ip:
3665 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
3666 raise errors.OpPrereqError("The given cluster IP address (%s) is"
3667 " reachable on the network" %
3668 new_ip, errors.ECODE_NOTUNIQUE)
3670 self.op.name = new_name
3672 def Exec(self, feedback_fn):
3673 """Rename the cluster.
3676 clustername = self.op.name
3679 # shutdown the master IP
3680 master_params = self.cfg.GetMasterNetworkParameters()
3681 ems = self.cfg.GetUseExternalMipScript()
3682 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
3684 result.Raise("Could not disable the master role")
3687 cluster = self.cfg.GetClusterInfo()
3688 cluster.cluster_name = clustername
3689 cluster.master_ip = new_ip
3690 self.cfg.Update(cluster, feedback_fn)
3692 # update the known hosts file
3693 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
3694 node_list = self.cfg.GetOnlineNodeList()
3696 node_list.remove(master_params.name)
3699 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
3701 master_params.ip = new_ip
3702 result = self.rpc.call_node_activate_master_ip(master_params.name,
3704 msg = result.fail_msg
3706 self.LogWarning("Could not re-enable the master role on"
3707 " the master, please restart manually: %s", msg)
3712 def _ValidateNetmask(cfg, netmask):
3713 """Checks if a netmask is valid.
3715 @type cfg: L{config.ConfigWriter}
3716 @param cfg: The cluster configuration
3718 @param netmask: the netmask to be verified
3719 @raise errors.OpPrereqError: if the validation fails
3722 ip_family = cfg.GetPrimaryIPFamily()
3724 ipcls = netutils.IPAddress.GetClassFromIpFamily(ip_family)
3725 except errors.ProgrammerError:
3726 raise errors.OpPrereqError("Invalid primary ip family: %s." %
3728 if not ipcls.ValidateNetmask(netmask):
3729 raise errors.OpPrereqError("CIDR netmask (%s) not valid" %
3733 class LUClusterSetParams(LogicalUnit):
3734 """Change the parameters of the cluster.
3737 HPATH = "cluster-modify"
3738 HTYPE = constants.HTYPE_CLUSTER
3741 def CheckArguments(self):
3745 if self.op.uid_pool:
3746 uidpool.CheckUidPool(self.op.uid_pool)
3748 if self.op.add_uids:
3749 uidpool.CheckUidPool(self.op.add_uids)
3751 if self.op.remove_uids:
3752 uidpool.CheckUidPool(self.op.remove_uids)
3754 if self.op.master_netmask is not None:
3755 _ValidateNetmask(self.cfg, self.op.master_netmask)
3757 if self.op.diskparams:
3758 for dt_params in self.op.diskparams.values():
3759 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
3761 def ExpandNames(self):
3762 # FIXME: in the future maybe other cluster params won't require checking on
3763 # all nodes to be modified.
3764 self.needed_locks = {
3765 locking.LEVEL_NODE: locking.ALL_SET,
3767 self.share_locks[locking.LEVEL_NODE] = 1
3769 def BuildHooksEnv(self):
3774 "OP_TARGET": self.cfg.GetClusterName(),
3775 "NEW_VG_NAME": self.op.vg_name,
3778 def BuildHooksNodes(self):
3779 """Build hooks nodes.
3782 mn = self.cfg.GetMasterNode()
3785 def CheckPrereq(self):
3786 """Check prerequisites.
3788 This checks whether the given params don't conflict and
3789 if the given volume group is valid.
3792 if self.op.vg_name is not None and not self.op.vg_name:
3793 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
3794 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
3795 " instances exist", errors.ECODE_INVAL)
3797 if self.op.drbd_helper is not None and not self.op.drbd_helper:
3798 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
3799 raise errors.OpPrereqError("Cannot disable drbd helper while"
3800 " drbd-based instances exist",
3803 node_list = self.owned_locks(locking.LEVEL_NODE)
3805 # if vg_name not None, checks given volume group on all nodes
3807 vglist = self.rpc.call_vg_list(node_list)
3808 for node in node_list:
3809 msg = vglist[node].fail_msg
3811 # ignoring down node
3812 self.LogWarning("Error while gathering data on node %s"
3813 " (ignoring node): %s", node, msg)
3815 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
3817 constants.MIN_VG_SIZE)
3819 raise errors.OpPrereqError("Error on node '%s': %s" %
3820 (node, vgstatus), errors.ECODE_ENVIRON)
3822 if self.op.drbd_helper:
3823 # checks given drbd helper on all nodes
3824 helpers = self.rpc.call_drbd_helper(node_list)
3825 for (node, ninfo) in self.cfg.GetMultiNodeInfo(node_list):
3827 self.LogInfo("Not checking drbd helper on offline node %s", node)
3829 msg = helpers[node].fail_msg
3831 raise errors.OpPrereqError("Error checking drbd helper on node"
3832 " '%s': %s" % (node, msg),
3833 errors.ECODE_ENVIRON)
3834 node_helper = helpers[node].payload
3835 if node_helper != self.op.drbd_helper:
3836 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
3837 (node, node_helper), errors.ECODE_ENVIRON)
3839 self.cluster = cluster = self.cfg.GetClusterInfo()
3840 # validate params changes
3841 if self.op.beparams:
3842 objects.UpgradeBeParams(self.op.beparams)
3843 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
3844 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
3846 if self.op.ndparams:
3847 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
3848 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
3850 # TODO: we need a more general way to handle resetting
3851 # cluster-level parameters to default values
3852 if self.new_ndparams["oob_program"] == "":
3853 self.new_ndparams["oob_program"] = \
3854 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
3856 if self.op.hv_state:
3857 new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
3858 self.cluster.hv_state_static)
3859 self.new_hv_state = dict((hv, cluster.SimpleFillHvState(values))
3860 for hv, values in new_hv_state.items())
3862 if self.op.disk_state:
3863 new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state,
3864 self.cluster.disk_state_static)
3865 self.new_disk_state = \
3866 dict((storage, dict((name, cluster.SimpleFillDiskState(values))
3867 for name, values in svalues.items()))
3868 for storage, svalues in new_disk_state.items())
3871 self.new_ipolicy = _GetUpdatedIPolicy(cluster.ipolicy, self.op.ipolicy,
3874 if self.op.nicparams:
3875 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
3876 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
3877 objects.NIC.CheckParameterSyntax(self.new_nicparams)
3880 # check all instances for consistency
3881 for instance in self.cfg.GetAllInstancesInfo().values():
3882 for nic_idx, nic in enumerate(instance.nics):
3883 params_copy = copy.deepcopy(nic.nicparams)
3884 params_filled = objects.FillDict(self.new_nicparams, params_copy)
3886 # check parameter syntax
3888 objects.NIC.CheckParameterSyntax(params_filled)
3889 except errors.ConfigurationError, err:
3890 nic_errors.append("Instance %s, nic/%d: %s" %
3891 (instance.name, nic_idx, err))
3893 # if we're moving instances to routed, check that they have an ip
3894 target_mode = params_filled[constants.NIC_MODE]
3895 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
3896 nic_errors.append("Instance %s, nic/%d: routed NIC with no ip"
3897 " address" % (instance.name, nic_idx))
3899 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
3900 "\n".join(nic_errors))
3902 # hypervisor list/parameters
3903 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
3904 if self.op.hvparams:
3905 for hv_name, hv_dict in self.op.hvparams.items():
3906 if hv_name not in self.new_hvparams:
3907 self.new_hvparams[hv_name] = hv_dict
3909 self.new_hvparams[hv_name].update(hv_dict)
3911 # disk template parameters
3912 self.new_diskparams = objects.FillDict(cluster.diskparams, {})
3913 if self.op.diskparams:
3914 for dt_name, dt_params in self.op.diskparams.items():
3915 if dt_name not in self.op.diskparams:
3916 self.new_diskparams[dt_name] = dt_params
3918 self.new_diskparams[dt_name].update(dt_params)
3920 # os hypervisor parameters
3921 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
3923 for os_name, hvs in self.op.os_hvp.items():
3924 if os_name not in self.new_os_hvp:
3925 self.new_os_hvp[os_name] = hvs
3927 for hv_name, hv_dict in hvs.items():
3928 if hv_name not in self.new_os_hvp[os_name]:
3929 self.new_os_hvp[os_name][hv_name] = hv_dict
3931 self.new_os_hvp[os_name][hv_name].update(hv_dict)
3934 self.new_osp = objects.FillDict(cluster.osparams, {})
3935 if self.op.osparams:
3936 for os_name, osp in self.op.osparams.items():
3937 if os_name not in self.new_osp:
3938 self.new_osp[os_name] = {}
3940 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
3943 if not self.new_osp[os_name]:
3944 # we removed all parameters
3945 del self.new_osp[os_name]
3947 # check the parameter validity (remote check)
3948 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
3949 os_name, self.new_osp[os_name])
3951 # changes to the hypervisor list
3952 if self.op.enabled_hypervisors is not None:
3953 self.hv_list = self.op.enabled_hypervisors
3954 for hv in self.hv_list:
3955 # if the hypervisor doesn't already exist in the cluster
3956 # hvparams, we initialize it to empty, and then (in both
3957 # cases) we make sure to fill the defaults, as we might not
3958 # have a complete defaults list if the hypervisor wasn't
3960 if hv not in new_hvp:
3962 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
3963 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
3965 self.hv_list = cluster.enabled_hypervisors
3967 if self.op.hvparams or self.op.enabled_hypervisors is not None:
3968 # either the enabled list has changed, or the parameters have, validate
3969 for hv_name, hv_params in self.new_hvparams.items():
3970 if ((self.op.hvparams and hv_name in self.op.hvparams) or
3971 (self.op.enabled_hypervisors and
3972 hv_name in self.op.enabled_hypervisors)):
3973 # either this is a new hypervisor, or its parameters have changed
3974 hv_class = hypervisor.GetHypervisor(hv_name)
3975 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3976 hv_class.CheckParameterSyntax(hv_params)
3977 _CheckHVParams(self, node_list, hv_name, hv_params)
3980 # no need to check any newly-enabled hypervisors, since the
3981 # defaults have already been checked in the above code-block
3982 for os_name, os_hvp in self.new_os_hvp.items():
3983 for hv_name, hv_params in os_hvp.items():
3984 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3985 # we need to fill in the new os_hvp on top of the actual hv_p
3986 cluster_defaults = self.new_hvparams.get(hv_name, {})
3987 new_osp = objects.FillDict(cluster_defaults, hv_params)
3988 hv_class = hypervisor.GetHypervisor(hv_name)
3989 hv_class.CheckParameterSyntax(new_osp)
3990 _CheckHVParams(self, node_list, hv_name, new_osp)
3992 if self.op.default_iallocator:
3993 alloc_script = utils.FindFile(self.op.default_iallocator,
3994 constants.IALLOCATOR_SEARCH_PATH,
3996 if alloc_script is None:
3997 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
3998 " specified" % self.op.default_iallocator,
4001 def Exec(self, feedback_fn):
4002 """Change the parameters of the cluster.
4005 if self.op.vg_name is not None:
4006 new_volume = self.op.vg_name
4009 if new_volume != self.cfg.GetVGName():
4010 self.cfg.SetVGName(new_volume)
4012 feedback_fn("Cluster LVM configuration already in desired"
4013 " state, not changing")
4014 if self.op.drbd_helper is not None:
4015 new_helper = self.op.drbd_helper
4018 if new_helper != self.cfg.GetDRBDHelper():
4019 self.cfg.SetDRBDHelper(new_helper)
4021 feedback_fn("Cluster DRBD helper already in desired state,"
4023 if self.op.hvparams:
4024 self.cluster.hvparams = self.new_hvparams
4026 self.cluster.os_hvp = self.new_os_hvp
4027 if self.op.enabled_hypervisors is not None:
4028 self.cluster.hvparams = self.new_hvparams
4029 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
4030 if self.op.beparams:
4031 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
4032 if self.op.nicparams:
4033 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
4035 self.cluster.ipolicy = self.new_ipolicy
4036 if self.op.osparams:
4037 self.cluster.osparams = self.new_osp
4038 if self.op.ndparams:
4039 self.cluster.ndparams = self.new_ndparams
4040 if self.op.diskparams:
4041 self.cluster.diskparams = self.new_diskparams
4042 if self.op.hv_state:
4043 self.cluster.hv_state_static = self.new_hv_state
4044 if self.op.disk_state:
4045 self.cluster.disk_state_static = self.new_disk_state
4047 if self.op.candidate_pool_size is not None:
4048 self.cluster.candidate_pool_size = self.op.candidate_pool_size
4049 # we need to update the pool size here, otherwise the save will fail
4050 _AdjustCandidatePool(self, [])
4052 if self.op.maintain_node_health is not None:
4053 if self.op.maintain_node_health and not constants.ENABLE_CONFD:
4054 feedback_fn("Note: CONFD was disabled at build time, node health"
4055 " maintenance is not useful (still enabling it)")
4056 self.cluster.maintain_node_health = self.op.maintain_node_health
4058 if self.op.prealloc_wipe_disks is not None:
4059 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
4061 if self.op.add_uids is not None:
4062 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
4064 if self.op.remove_uids is not None:
4065 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
4067 if self.op.uid_pool is not None:
4068 self.cluster.uid_pool = self.op.uid_pool
4070 if self.op.default_iallocator is not None:
4071 self.cluster.default_iallocator = self.op.default_iallocator
4073 if self.op.reserved_lvs is not None:
4074 self.cluster.reserved_lvs = self.op.reserved_lvs
4076 if self.op.use_external_mip_script is not None:
4077 self.cluster.use_external_mip_script = self.op.use_external_mip_script
4079 def helper_os(aname, mods, desc):
4081 lst = getattr(self.cluster, aname)
4082 for key, val in mods:
4083 if key == constants.DDM_ADD:
4085 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
4088 elif key == constants.DDM_REMOVE:
4092 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
4094 raise errors.ProgrammerError("Invalid modification '%s'" % key)
4096 if self.op.hidden_os:
4097 helper_os("hidden_os", self.op.hidden_os, "hidden")
4099 if self.op.blacklisted_os:
4100 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
4102 if self.op.master_netdev:
4103 master_params = self.cfg.GetMasterNetworkParameters()
4104 ems = self.cfg.GetUseExternalMipScript()
4105 feedback_fn("Shutting down master ip on the current netdev (%s)" %
4106 self.cluster.master_netdev)
4107 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
4109 result.Raise("Could not disable the master ip")
4110 feedback_fn("Changing master_netdev from %s to %s" %
4111 (master_params.netdev, self.op.master_netdev))
4112 self.cluster.master_netdev = self.op.master_netdev
4114 if self.op.master_netmask:
4115 master_params = self.cfg.GetMasterNetworkParameters()
4116 feedback_fn("Changing master IP netmask to %s" % self.op.master_netmask)
4117 result = self.rpc.call_node_change_master_netmask(master_params.name,
4118 master_params.netmask,
4119 self.op.master_netmask,
4121 master_params.netdev)
4123 msg = "Could not change the master IP netmask: %s" % result.fail_msg
4126 self.cluster.master_netmask = self.op.master_netmask
4128 self.cfg.Update(self.cluster, feedback_fn)
4130 if self.op.master_netdev:
4131 master_params = self.cfg.GetMasterNetworkParameters()
4132 feedback_fn("Starting the master ip on the new master netdev (%s)" %
4133 self.op.master_netdev)
4134 ems = self.cfg.GetUseExternalMipScript()
4135 result = self.rpc.call_node_activate_master_ip(master_params.name,
4138 self.LogWarning("Could not re-enable the master ip on"
4139 " the master, please restart manually: %s",
4143 def _UploadHelper(lu, nodes, fname):
4144 """Helper for uploading a file and showing warnings.
4147 if os.path.exists(fname):
4148 result = lu.rpc.call_upload_file(nodes, fname)
4149 for to_node, to_result in result.items():
4150 msg = to_result.fail_msg
4152 msg = ("Copy of file %s to node %s failed: %s" %
4153 (fname, to_node, msg))
4154 lu.proc.LogWarning(msg)
4157 def _ComputeAncillaryFiles(cluster, redist):
4158 """Compute files external to Ganeti which need to be consistent.
4160 @type redist: boolean
4161 @param redist: Whether to include files which need to be redistributed
4164 # Compute files for all nodes
4166 constants.SSH_KNOWN_HOSTS_FILE,
4167 constants.CONFD_HMAC_KEY,
4168 constants.CLUSTER_DOMAIN_SECRET_FILE,
4169 constants.SPICE_CERT_FILE,
4170 constants.SPICE_CACERT_FILE,
4171 constants.RAPI_USERS_FILE,
4175 files_all.update(constants.ALL_CERT_FILES)
4176 files_all.update(ssconf.SimpleStore().GetFileList())
4178 # we need to ship at least the RAPI certificate
4179 files_all.add(constants.RAPI_CERT_FILE)
4181 if cluster.modify_etc_hosts:
4182 files_all.add(constants.ETC_HOSTS)
4184 # Files which are optional, these must:
4185 # - be present in one other category as well
4186 # - either exist or not exist on all nodes of that category (mc, vm all)
4188 constants.RAPI_USERS_FILE,
4191 # Files which should only be on master candidates
4195 files_mc.add(constants.CLUSTER_CONF_FILE)
4197 # FIXME: this should also be replicated but Ganeti doesn't support files_mc
4199 files_mc.add(constants.DEFAULT_MASTER_SETUP_SCRIPT)
4201 # Files which should only be on VM-capable nodes
4202 files_vm = set(filename
4203 for hv_name in cluster.enabled_hypervisors
4204 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[0])
4206 files_opt |= set(filename
4207 for hv_name in cluster.enabled_hypervisors
4208 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[1])
4210 # Filenames in each category must be unique
4211 all_files_set = files_all | files_mc | files_vm
4212 assert (len(all_files_set) ==
4213 sum(map(len, [files_all, files_mc, files_vm]))), \
4214 "Found file listed in more than one file list"
4216 # Optional files must be present in one other category
4217 assert all_files_set.issuperset(files_opt), \
4218 "Optional file not in a different required list"
4220 return (files_all, files_opt, files_mc, files_vm)
4223 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
4224 """Distribute additional files which are part of the cluster configuration.
4226 ConfigWriter takes care of distributing the config and ssconf files, but
4227 there are more files which should be distributed to all nodes. This function
4228 makes sure those are copied.
4230 @param lu: calling logical unit
4231 @param additional_nodes: list of nodes not in the config to distribute to
4232 @type additional_vm: boolean
4233 @param additional_vm: whether the additional nodes are vm-capable or not
4236 # Gather target nodes
4237 cluster = lu.cfg.GetClusterInfo()
4238 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
4240 online_nodes = lu.cfg.GetOnlineNodeList()
4241 vm_nodes = lu.cfg.GetVmCapableNodeList()
4243 if additional_nodes is not None:
4244 online_nodes.extend(additional_nodes)
4246 vm_nodes.extend(additional_nodes)
4248 # Never distribute to master node
4249 for nodelist in [online_nodes, vm_nodes]:
4250 if master_info.name in nodelist:
4251 nodelist.remove(master_info.name)
4254 (files_all, _, files_mc, files_vm) = \
4255 _ComputeAncillaryFiles(cluster, True)
4257 # Never re-distribute configuration file from here
4258 assert not (constants.CLUSTER_CONF_FILE in files_all or
4259 constants.CLUSTER_CONF_FILE in files_vm)
4260 assert not files_mc, "Master candidates not handled in this function"
4263 (online_nodes, files_all),
4264 (vm_nodes, files_vm),
4268 for (node_list, files) in filemap:
4270 _UploadHelper(lu, node_list, fname)
4273 class LUClusterRedistConf(NoHooksLU):
4274 """Force the redistribution of cluster configuration.
4276 This is a very simple LU.
4281 def ExpandNames(self):
4282 self.needed_locks = {
4283 locking.LEVEL_NODE: locking.ALL_SET,
4285 self.share_locks[locking.LEVEL_NODE] = 1
4287 def Exec(self, feedback_fn):
4288 """Redistribute the configuration.
4291 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
4292 _RedistributeAncillaryFiles(self)
4295 class LUClusterActivateMasterIp(NoHooksLU):
4296 """Activate the master IP on the master node.
4299 def Exec(self, feedback_fn):
4300 """Activate the master IP.
4303 master_params = self.cfg.GetMasterNetworkParameters()
4304 ems = self.cfg.GetUseExternalMipScript()
4305 result = self.rpc.call_node_activate_master_ip(master_params.name,
4307 result.Raise("Could not activate the master IP")
4310 class LUClusterDeactivateMasterIp(NoHooksLU):
4311 """Deactivate the master IP on the master node.
4314 def Exec(self, feedback_fn):
4315 """Deactivate the master IP.
4318 master_params = self.cfg.GetMasterNetworkParameters()
4319 ems = self.cfg.GetUseExternalMipScript()
4320 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
4322 result.Raise("Could not deactivate the master IP")
4325 def _WaitForSync(lu, instance, disks=None, oneshot=False):
4326 """Sleep and poll for an instance's disk to sync.
4329 if not instance.disks or disks is not None and not disks:
4332 disks = _ExpandCheckDisks(instance, disks)
4335 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
4337 node = instance.primary_node
4340 lu.cfg.SetDiskID(dev, node)
4342 # TODO: Convert to utils.Retry
4345 degr_retries = 10 # in seconds, as we sleep 1 second each time
4349 cumul_degraded = False
4350 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
4351 msg = rstats.fail_msg
4353 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
4356 raise errors.RemoteError("Can't contact node %s for mirror data,"
4357 " aborting." % node)
4360 rstats = rstats.payload
4362 for i, mstat in enumerate(rstats):
4364 lu.LogWarning("Can't compute data for node %s/%s",
4365 node, disks[i].iv_name)
4368 cumul_degraded = (cumul_degraded or
4369 (mstat.is_degraded and mstat.sync_percent is None))
4370 if mstat.sync_percent is not None:
4372 if mstat.estimated_time is not None:
4373 rem_time = ("%s remaining (estimated)" %
4374 utils.FormatSeconds(mstat.estimated_time))
4375 max_time = mstat.estimated_time
4377 rem_time = "no time estimate"
4378 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
4379 (disks[i].iv_name, mstat.sync_percent, rem_time))
4381 # if we're done but degraded, let's do a few small retries, to
4382 # make sure we see a stable and not transient situation; therefore
4383 # we force restart of the loop
4384 if (done or oneshot) and cumul_degraded and degr_retries > 0:
4385 logging.info("Degraded disks found, %d retries left", degr_retries)
4393 time.sleep(min(60, max_time))
4396 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
4397 return not cumul_degraded
4400 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
4401 """Check that mirrors are not degraded.
4403 The ldisk parameter, if True, will change the test from the
4404 is_degraded attribute (which represents overall non-ok status for
4405 the device(s)) to the ldisk (representing the local storage status).
4408 lu.cfg.SetDiskID(dev, node)
4412 if on_primary or dev.AssembleOnSecondary():
4413 rstats = lu.rpc.call_blockdev_find(node, dev)
4414 msg = rstats.fail_msg
4416 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
4418 elif not rstats.payload:
4419 lu.LogWarning("Can't find disk on node %s", node)
4423 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
4425 result = result and not rstats.payload.is_degraded
4428 for child in dev.children:
4429 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
4434 class LUOobCommand(NoHooksLU):
4435 """Logical unit for OOB handling.
4439 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
4441 def ExpandNames(self):
4442 """Gather locks we need.
4445 if self.op.node_names:
4446 self.op.node_names = _GetWantedNodes(self, self.op.node_names)
4447 lock_names = self.op.node_names
4449 lock_names = locking.ALL_SET
4451 self.needed_locks = {
4452 locking.LEVEL_NODE: lock_names,
4455 def CheckPrereq(self):
4456 """Check prerequisites.
4459 - the node exists in the configuration
4462 Any errors are signaled by raising errors.OpPrereqError.
4466 self.master_node = self.cfg.GetMasterNode()
4468 assert self.op.power_delay >= 0.0
4470 if self.op.node_names:
4471 if (self.op.command in self._SKIP_MASTER and
4472 self.master_node in self.op.node_names):
4473 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
4474 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
4476 if master_oob_handler:
4477 additional_text = ("run '%s %s %s' if you want to operate on the"
4478 " master regardless") % (master_oob_handler,
4482 additional_text = "it does not support out-of-band operations"
4484 raise errors.OpPrereqError(("Operating on the master node %s is not"
4485 " allowed for %s; %s") %
4486 (self.master_node, self.op.command,
4487 additional_text), errors.ECODE_INVAL)
4489 self.op.node_names = self.cfg.GetNodeList()
4490 if self.op.command in self._SKIP_MASTER:
4491 self.op.node_names.remove(self.master_node)
4493 if self.op.command in self._SKIP_MASTER:
4494 assert self.master_node not in self.op.node_names
4496 for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names):
4498 raise errors.OpPrereqError("Node %s not found" % node_name,
4501 self.nodes.append(node)
4503 if (not self.op.ignore_status and
4504 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
4505 raise errors.OpPrereqError(("Cannot power off node %s because it is"
4506 " not marked offline") % node_name,
4509 def Exec(self, feedback_fn):
4510 """Execute OOB and return result if we expect any.
4513 master_node = self.master_node
4516 for idx, node in enumerate(utils.NiceSort(self.nodes,
4517 key=lambda node: node.name)):
4518 node_entry = [(constants.RS_NORMAL, node.name)]
4519 ret.append(node_entry)
4521 oob_program = _SupportsOob(self.cfg, node)
4524 node_entry.append((constants.RS_UNAVAIL, None))
4527 logging.info("Executing out-of-band command '%s' using '%s' on %s",
4528 self.op.command, oob_program, node.name)
4529 result = self.rpc.call_run_oob(master_node, oob_program,
4530 self.op.command, node.name,
4534 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
4535 node.name, result.fail_msg)
4536 node_entry.append((constants.RS_NODATA, None))
4539 self._CheckPayload(result)
4540 except errors.OpExecError, err:
4541 self.LogWarning("Payload returned by node '%s' is not valid: %s",
4543 node_entry.append((constants.RS_NODATA, None))
4545 if self.op.command == constants.OOB_HEALTH:
4546 # For health we should log important events
4547 for item, status in result.payload:
4548 if status in [constants.OOB_STATUS_WARNING,
4549 constants.OOB_STATUS_CRITICAL]:
4550 self.LogWarning("Item '%s' on node '%s' has status '%s'",
4551 item, node.name, status)
4553 if self.op.command == constants.OOB_POWER_ON:
4555 elif self.op.command == constants.OOB_POWER_OFF:
4556 node.powered = False
4557 elif self.op.command == constants.OOB_POWER_STATUS:
4558 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
4559 if powered != node.powered:
4560 logging.warning(("Recorded power state (%s) of node '%s' does not"
4561 " match actual power state (%s)"), node.powered,
4564 # For configuration changing commands we should update the node
4565 if self.op.command in (constants.OOB_POWER_ON,
4566 constants.OOB_POWER_OFF):
4567 self.cfg.Update(node, feedback_fn)
4569 node_entry.append((constants.RS_NORMAL, result.payload))
4571 if (self.op.command == constants.OOB_POWER_ON and
4572 idx < len(self.nodes) - 1):
4573 time.sleep(self.op.power_delay)
4577 def _CheckPayload(self, result):
4578 """Checks if the payload is valid.
4580 @param result: RPC result
4581 @raises errors.OpExecError: If payload is not valid
4585 if self.op.command == constants.OOB_HEALTH:
4586 if not isinstance(result.payload, list):
4587 errs.append("command 'health' is expected to return a list but got %s" %
4588 type(result.payload))
4590 for item, status in result.payload:
4591 if status not in constants.OOB_STATUSES:
4592 errs.append("health item '%s' has invalid status '%s'" %
4595 if self.op.command == constants.OOB_POWER_STATUS:
4596 if not isinstance(result.payload, dict):
4597 errs.append("power-status is expected to return a dict but got %s" %
4598 type(result.payload))
4600 if self.op.command in [
4601 constants.OOB_POWER_ON,
4602 constants.OOB_POWER_OFF,
4603 constants.OOB_POWER_CYCLE,
4605 if result.payload is not None:
4606 errs.append("%s is expected to not return payload but got '%s'" %
4607 (self.op.command, result.payload))
4610 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
4611 utils.CommaJoin(errs))
4614 class _OsQuery(_QueryBase):
4615 FIELDS = query.OS_FIELDS
4617 def ExpandNames(self, lu):
4618 # Lock all nodes in shared mode
4619 # Temporary removal of locks, should be reverted later
4620 # TODO: reintroduce locks when they are lighter-weight
4621 lu.needed_locks = {}
4622 #self.share_locks[locking.LEVEL_NODE] = 1
4623 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4625 # The following variables interact with _QueryBase._GetNames
4627 self.wanted = self.names
4629 self.wanted = locking.ALL_SET
4631 self.do_locking = self.use_locking
4633 def DeclareLocks(self, lu, level):
4637 def _DiagnoseByOS(rlist):
4638 """Remaps a per-node return list into an a per-os per-node dictionary
4640 @param rlist: a map with node names as keys and OS objects as values
4643 @return: a dictionary with osnames as keys and as value another
4644 map, with nodes as keys and tuples of (path, status, diagnose,
4645 variants, parameters, api_versions) as values, eg::
4647 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
4648 (/srv/..., False, "invalid api")],
4649 "node2": [(/srv/..., True, "", [], [])]}
4654 # we build here the list of nodes that didn't fail the RPC (at RPC
4655 # level), so that nodes with a non-responding node daemon don't
4656 # make all OSes invalid
4657 good_nodes = [node_name for node_name in rlist
4658 if not rlist[node_name].fail_msg]
4659 for node_name, nr in rlist.items():
4660 if nr.fail_msg or not nr.payload:
4662 for (name, path, status, diagnose, variants,
4663 params, api_versions) in nr.payload:
4664 if name not in all_os:
4665 # build a list of nodes for this os containing empty lists
4666 # for each node in node_list
4668 for nname in good_nodes:
4669 all_os[name][nname] = []
4670 # convert params from [name, help] to (name, help)
4671 params = [tuple(v) for v in params]
4672 all_os[name][node_name].append((path, status, diagnose,
4673 variants, params, api_versions))
4676 def _GetQueryData(self, lu):
4677 """Computes the list of nodes and their attributes.
4680 # Locking is not used
4681 assert not (compat.any(lu.glm.is_owned(level)
4682 for level in locking.LEVELS
4683 if level != locking.LEVEL_CLUSTER) or
4684 self.do_locking or self.use_locking)
4686 valid_nodes = [node.name
4687 for node in lu.cfg.GetAllNodesInfo().values()
4688 if not node.offline and node.vm_capable]
4689 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
4690 cluster = lu.cfg.GetClusterInfo()
4694 for (os_name, os_data) in pol.items():
4695 info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
4696 hidden=(os_name in cluster.hidden_os),
4697 blacklisted=(os_name in cluster.blacklisted_os))
4701 api_versions = set()
4703 for idx, osl in enumerate(os_data.values()):
4704 info.valid = bool(info.valid and osl and osl[0][1])
4708 (node_variants, node_params, node_api) = osl[0][3:6]
4711 variants.update(node_variants)
4712 parameters.update(node_params)
4713 api_versions.update(node_api)
4715 # Filter out inconsistent values
4716 variants.intersection_update(node_variants)
4717 parameters.intersection_update(node_params)
4718 api_versions.intersection_update(node_api)
4720 info.variants = list(variants)
4721 info.parameters = list(parameters)
4722 info.api_versions = list(api_versions)
4724 data[os_name] = info
4726 # Prepare data in requested order
4727 return [data[name] for name in self._GetNames(lu, pol.keys(), None)
4731 class LUOsDiagnose(NoHooksLU):
4732 """Logical unit for OS diagnose/query.
4738 def _BuildFilter(fields, names):
4739 """Builds a filter for querying OSes.
4742 name_filter = qlang.MakeSimpleFilter("name", names)
4744 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
4745 # respective field is not requested
4746 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]]
4747 for fname in ["hidden", "blacklisted"]
4748 if fname not in fields]
4749 if "valid" not in fields:
4750 status_filter.append([qlang.OP_TRUE, "valid"])
4753 status_filter.insert(0, qlang.OP_AND)
4755 status_filter = None
4757 if name_filter and status_filter:
4758 return [qlang.OP_AND, name_filter, status_filter]
4762 return status_filter
4764 def CheckArguments(self):
4765 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names),
4766 self.op.output_fields, False)
4768 def ExpandNames(self):
4769 self.oq.ExpandNames(self)
4771 def Exec(self, feedback_fn):
4772 return self.oq.OldStyleQuery(self)
4775 class LUNodeRemove(LogicalUnit):
4776 """Logical unit for removing a node.
4779 HPATH = "node-remove"
4780 HTYPE = constants.HTYPE_NODE
4782 def BuildHooksEnv(self):
4785 This doesn't run on the target node in the pre phase as a failed
4786 node would then be impossible to remove.
4790 "OP_TARGET": self.op.node_name,
4791 "NODE_NAME": self.op.node_name,
4794 def BuildHooksNodes(self):
4795 """Build hooks nodes.
4798 all_nodes = self.cfg.GetNodeList()
4800 all_nodes.remove(self.op.node_name)
4802 logging.warning("Node '%s', which is about to be removed, was not found"
4803 " in the list of all nodes", self.op.node_name)
4804 return (all_nodes, all_nodes)
4806 def CheckPrereq(self):
4807 """Check prerequisites.
4810 - the node exists in the configuration
4811 - it does not have primary or secondary instances
4812 - it's not the master
4814 Any errors are signaled by raising errors.OpPrereqError.
4817 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4818 node = self.cfg.GetNodeInfo(self.op.node_name)
4819 assert node is not None
4821 masternode = self.cfg.GetMasterNode()
4822 if node.name == masternode:
4823 raise errors.OpPrereqError("Node is the master node, failover to another"
4824 " node is required", errors.ECODE_INVAL)
4826 for instance_name, instance in self.cfg.GetAllInstancesInfo().items():
4827 if node.name in instance.all_nodes:
4828 raise errors.OpPrereqError("Instance %s is still running on the node,"
4829 " please remove first" % instance_name,
4831 self.op.node_name = node.name
4834 def Exec(self, feedback_fn):
4835 """Removes the node from the cluster.
4839 logging.info("Stopping the node daemon and removing configs from node %s",
4842 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
4844 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
4847 # Promote nodes to master candidate as needed
4848 _AdjustCandidatePool(self, exceptions=[node.name])
4849 self.context.RemoveNode(node.name)
4851 # Run post hooks on the node before it's removed
4852 _RunPostHook(self, node.name)
4854 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
4855 msg = result.fail_msg
4857 self.LogWarning("Errors encountered on the remote node while leaving"
4858 " the cluster: %s", msg)
4860 # Remove node from our /etc/hosts
4861 if self.cfg.GetClusterInfo().modify_etc_hosts:
4862 master_node = self.cfg.GetMasterNode()
4863 result = self.rpc.call_etc_hosts_modify(master_node,
4864 constants.ETC_HOSTS_REMOVE,
4866 result.Raise("Can't update hosts file with new host data")
4867 _RedistributeAncillaryFiles(self)
4870 class _NodeQuery(_QueryBase):
4871 FIELDS = query.NODE_FIELDS
4873 def ExpandNames(self, lu):
4874 lu.needed_locks = {}
4875 lu.share_locks = _ShareAll()
4878 self.wanted = _GetWantedNodes(lu, self.names)
4880 self.wanted = locking.ALL_SET
4882 self.do_locking = (self.use_locking and
4883 query.NQ_LIVE in self.requested_data)
4886 # If any non-static field is requested we need to lock the nodes
4887 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
4889 def DeclareLocks(self, lu, level):
4892 def _GetQueryData(self, lu):
4893 """Computes the list of nodes and their attributes.
4896 all_info = lu.cfg.GetAllNodesInfo()
4898 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
4900 # Gather data as requested
4901 if query.NQ_LIVE in self.requested_data:
4902 # filter out non-vm_capable nodes
4903 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
4905 node_data = lu.rpc.call_node_info(toquery_nodes, [lu.cfg.GetVGName()],
4906 [lu.cfg.GetHypervisorType()])
4907 live_data = dict((name, _MakeLegacyNodeInfo(nresult.payload))
4908 for (name, nresult) in node_data.items()
4909 if not nresult.fail_msg and nresult.payload)
4913 if query.NQ_INST in self.requested_data:
4914 node_to_primary = dict([(name, set()) for name in nodenames])
4915 node_to_secondary = dict([(name, set()) for name in nodenames])
4917 inst_data = lu.cfg.GetAllInstancesInfo()
4919 for inst in inst_data.values():
4920 if inst.primary_node in node_to_primary:
4921 node_to_primary[inst.primary_node].add(inst.name)
4922 for secnode in inst.secondary_nodes:
4923 if secnode in node_to_secondary:
4924 node_to_secondary[secnode].add(inst.name)
4926 node_to_primary = None
4927 node_to_secondary = None
4929 if query.NQ_OOB in self.requested_data:
4930 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
4931 for name, node in all_info.iteritems())
4935 if query.NQ_GROUP in self.requested_data:
4936 groups = lu.cfg.GetAllNodeGroupsInfo()
4940 return query.NodeQueryData([all_info[name] for name in nodenames],
4941 live_data, lu.cfg.GetMasterNode(),
4942 node_to_primary, node_to_secondary, groups,
4943 oob_support, lu.cfg.GetClusterInfo())
4946 class LUNodeQuery(NoHooksLU):
4947 """Logical unit for querying nodes.
4950 # pylint: disable=W0142
4953 def CheckArguments(self):
4954 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
4955 self.op.output_fields, self.op.use_locking)
4957 def ExpandNames(self):
4958 self.nq.ExpandNames(self)
4960 def DeclareLocks(self, level):
4961 self.nq.DeclareLocks(self, level)
4963 def Exec(self, feedback_fn):
4964 return self.nq.OldStyleQuery(self)
4967 class LUNodeQueryvols(NoHooksLU):
4968 """Logical unit for getting volumes on node(s).
4972 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
4973 _FIELDS_STATIC = utils.FieldSet("node")
4975 def CheckArguments(self):
4976 _CheckOutputFields(static=self._FIELDS_STATIC,
4977 dynamic=self._FIELDS_DYNAMIC,
4978 selected=self.op.output_fields)
4980 def ExpandNames(self):
4981 self.share_locks = _ShareAll()
4982 self.needed_locks = {}
4984 if not self.op.nodes:
4985 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4987 self.needed_locks[locking.LEVEL_NODE] = \
4988 _GetWantedNodes(self, self.op.nodes)
4990 def Exec(self, feedback_fn):
4991 """Computes the list of nodes and their attributes.
4994 nodenames = self.owned_locks(locking.LEVEL_NODE)
4995 volumes = self.rpc.call_node_volumes(nodenames)
4997 ilist = self.cfg.GetAllInstancesInfo()
4998 vol2inst = _MapInstanceDisksToNodes(ilist.values())
5001 for node in nodenames:
5002 nresult = volumes[node]
5005 msg = nresult.fail_msg
5007 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
5010 node_vols = sorted(nresult.payload,
5011 key=operator.itemgetter("dev"))
5013 for vol in node_vols:
5015 for field in self.op.output_fields:
5018 elif field == "phys":
5022 elif field == "name":
5024 elif field == "size":
5025 val = int(float(vol["size"]))
5026 elif field == "instance":
5027 val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-")
5029 raise errors.ParameterError(field)
5030 node_output.append(str(val))
5032 output.append(node_output)
5037 class LUNodeQueryStorage(NoHooksLU):
5038 """Logical unit for getting information on storage units on node(s).
5041 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
5044 def CheckArguments(self):
5045 _CheckOutputFields(static=self._FIELDS_STATIC,
5046 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
5047 selected=self.op.output_fields)
5049 def ExpandNames(self):
5050 self.share_locks = _ShareAll()
5051 self.needed_locks = {}
5054 self.needed_locks[locking.LEVEL_NODE] = \
5055 _GetWantedNodes(self, self.op.nodes)
5057 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5059 def Exec(self, feedback_fn):
5060 """Computes the list of nodes and their attributes.
5063 self.nodes = self.owned_locks(locking.LEVEL_NODE)
5065 # Always get name to sort by
5066 if constants.SF_NAME in self.op.output_fields:
5067 fields = self.op.output_fields[:]
5069 fields = [constants.SF_NAME] + self.op.output_fields
5071 # Never ask for node or type as it's only known to the LU
5072 for extra in [constants.SF_NODE, constants.SF_TYPE]:
5073 while extra in fields:
5074 fields.remove(extra)
5076 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
5077 name_idx = field_idx[constants.SF_NAME]
5079 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
5080 data = self.rpc.call_storage_list(self.nodes,
5081 self.op.storage_type, st_args,
5082 self.op.name, fields)
5086 for node in utils.NiceSort(self.nodes):
5087 nresult = data[node]
5091 msg = nresult.fail_msg
5093 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
5096 rows = dict([(row[name_idx], row) for row in nresult.payload])
5098 for name in utils.NiceSort(rows.keys()):
5103 for field in self.op.output_fields:
5104 if field == constants.SF_NODE:
5106 elif field == constants.SF_TYPE:
5107 val = self.op.storage_type
5108 elif field in field_idx:
5109 val = row[field_idx[field]]
5111 raise errors.ParameterError(field)
5120 class _InstanceQuery(_QueryBase):
5121 FIELDS = query.INSTANCE_FIELDS
5123 def ExpandNames(self, lu):
5124 lu.needed_locks = {}
5125 lu.share_locks = _ShareAll()
5128 self.wanted = _GetWantedInstances(lu, self.names)
5130 self.wanted = locking.ALL_SET
5132 self.do_locking = (self.use_locking and
5133 query.IQ_LIVE in self.requested_data)
5135 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5136 lu.needed_locks[locking.LEVEL_NODEGROUP] = []
5137 lu.needed_locks[locking.LEVEL_NODE] = []
5138 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5140 self.do_grouplocks = (self.do_locking and
5141 query.IQ_NODES in self.requested_data)
5143 def DeclareLocks(self, lu, level):
5145 if level == locking.LEVEL_NODEGROUP and self.do_grouplocks:
5146 assert not lu.needed_locks[locking.LEVEL_NODEGROUP]
5148 # Lock all groups used by instances optimistically; this requires going
5149 # via the node before it's locked, requiring verification later on
5150 lu.needed_locks[locking.LEVEL_NODEGROUP] = \
5152 for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
5153 for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
5154 elif level == locking.LEVEL_NODE:
5155 lu._LockInstancesNodes() # pylint: disable=W0212
5158 def _CheckGroupLocks(lu):
5159 owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE))
5160 owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP))
5162 # Check if node groups for locked instances are still correct
5163 for instance_name in owned_instances:
5164 _CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups)
5166 def _GetQueryData(self, lu):
5167 """Computes the list of instances and their attributes.
5170 if self.do_grouplocks:
5171 self._CheckGroupLocks(lu)
5173 cluster = lu.cfg.GetClusterInfo()
5174 all_info = lu.cfg.GetAllInstancesInfo()
5176 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
5178 instance_list = [all_info[name] for name in instance_names]
5179 nodes = frozenset(itertools.chain(*(inst.all_nodes
5180 for inst in instance_list)))
5181 hv_list = list(set([inst.hypervisor for inst in instance_list]))
5184 wrongnode_inst = set()
5186 # Gather data as requested
5187 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
5189 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
5191 result = node_data[name]
5193 # offline nodes will be in both lists
5194 assert result.fail_msg
5195 offline_nodes.append(name)
5197 bad_nodes.append(name)
5198 elif result.payload:
5199 for inst in result.payload:
5200 if inst in all_info:
5201 if all_info[inst].primary_node == name:
5202 live_data.update(result.payload)
5204 wrongnode_inst.add(inst)
5206 # orphan instance; we don't list it here as we don't
5207 # handle this case yet in the output of instance listing
5208 logging.warning("Orphan instance '%s' found on node %s",
5210 # else no instance is alive
5214 if query.IQ_DISKUSAGE in self.requested_data:
5215 disk_usage = dict((inst.name,
5216 _ComputeDiskSize(inst.disk_template,
5217 [{constants.IDISK_SIZE: disk.size}
5218 for disk in inst.disks]))
5219 for inst in instance_list)
5223 if query.IQ_CONSOLE in self.requested_data:
5225 for inst in instance_list:
5226 if inst.name in live_data:
5227 # Instance is running
5228 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
5230 consinfo[inst.name] = None
5231 assert set(consinfo.keys()) == set(instance_names)
5235 if query.IQ_NODES in self.requested_data:
5236 node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"),
5238 nodes = dict(lu.cfg.GetMultiNodeInfo(node_names))
5239 groups = dict((uuid, lu.cfg.GetNodeGroup(uuid))
5240 for uuid in set(map(operator.attrgetter("group"),
5246 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
5247 disk_usage, offline_nodes, bad_nodes,
5248 live_data, wrongnode_inst, consinfo,
5252 class LUQuery(NoHooksLU):
5253 """Query for resources/items of a certain kind.
5256 # pylint: disable=W0142
5259 def CheckArguments(self):
5260 qcls = _GetQueryImplementation(self.op.what)
5262 self.impl = qcls(self.op.qfilter, self.op.fields, self.op.use_locking)
5264 def ExpandNames(self):
5265 self.impl.ExpandNames(self)
5267 def DeclareLocks(self, level):
5268 self.impl.DeclareLocks(self, level)
5270 def Exec(self, feedback_fn):
5271 return self.impl.NewStyleQuery(self)
5274 class LUQueryFields(NoHooksLU):
5275 """Query for resources/items of a certain kind.
5278 # pylint: disable=W0142
5281 def CheckArguments(self):
5282 self.qcls = _GetQueryImplementation(self.op.what)
5284 def ExpandNames(self):
5285 self.needed_locks = {}
5287 def Exec(self, feedback_fn):
5288 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
5291 class LUNodeModifyStorage(NoHooksLU):
5292 """Logical unit for modifying a storage volume on a node.
5297 def CheckArguments(self):
5298 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5300 storage_type = self.op.storage_type
5303 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
5305 raise errors.OpPrereqError("Storage units of type '%s' can not be"
5306 " modified" % storage_type,
5309 diff = set(self.op.changes.keys()) - modifiable
5311 raise errors.OpPrereqError("The following fields can not be modified for"
5312 " storage units of type '%s': %r" %
5313 (storage_type, list(diff)),
5316 def ExpandNames(self):
5317 self.needed_locks = {
5318 locking.LEVEL_NODE: self.op.node_name,
5321 def Exec(self, feedback_fn):
5322 """Computes the list of nodes and their attributes.
5325 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
5326 result = self.rpc.call_storage_modify(self.op.node_name,
5327 self.op.storage_type, st_args,
5328 self.op.name, self.op.changes)
5329 result.Raise("Failed to modify storage unit '%s' on %s" %
5330 (self.op.name, self.op.node_name))
5333 class LUNodeAdd(LogicalUnit):
5334 """Logical unit for adding node to the cluster.
5338 HTYPE = constants.HTYPE_NODE
5339 _NFLAGS = ["master_capable", "vm_capable"]
5341 def CheckArguments(self):
5342 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
5343 # validate/normalize the node name
5344 self.hostname = netutils.GetHostname(name=self.op.node_name,
5345 family=self.primary_ip_family)
5346 self.op.node_name = self.hostname.name
5348 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
5349 raise errors.OpPrereqError("Cannot readd the master node",
5352 if self.op.readd and self.op.group:
5353 raise errors.OpPrereqError("Cannot pass a node group when a node is"
5354 " being readded", errors.ECODE_INVAL)
5356 def BuildHooksEnv(self):
5359 This will run on all nodes before, and on all nodes + the new node after.
5363 "OP_TARGET": self.op.node_name,
5364 "NODE_NAME": self.op.node_name,
5365 "NODE_PIP": self.op.primary_ip,
5366 "NODE_SIP": self.op.secondary_ip,
5367 "MASTER_CAPABLE": str(self.op.master_capable),
5368 "VM_CAPABLE": str(self.op.vm_capable),
5371 def BuildHooksNodes(self):
5372 """Build hooks nodes.
5375 # Exclude added node
5376 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name]))
5377 post_nodes = pre_nodes + [self.op.node_name, ]
5379 return (pre_nodes, post_nodes)
5381 def CheckPrereq(self):
5382 """Check prerequisites.
5385 - the new node is not already in the config
5387 - its parameters (single/dual homed) matches the cluster
5389 Any errors are signaled by raising errors.OpPrereqError.
5393 hostname = self.hostname
5394 node = hostname.name
5395 primary_ip = self.op.primary_ip = hostname.ip
5396 if self.op.secondary_ip is None:
5397 if self.primary_ip_family == netutils.IP6Address.family:
5398 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
5399 " IPv4 address must be given as secondary",
5401 self.op.secondary_ip = primary_ip
5403 secondary_ip = self.op.secondary_ip
5404 if not netutils.IP4Address.IsValid(secondary_ip):
5405 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5406 " address" % secondary_ip, errors.ECODE_INVAL)
5408 node_list = cfg.GetNodeList()
5409 if not self.op.readd and node in node_list:
5410 raise errors.OpPrereqError("Node %s is already in the configuration" %
5411 node, errors.ECODE_EXISTS)
5412 elif self.op.readd and node not in node_list:
5413 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
5416 self.changed_primary_ip = False
5418 for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list):
5419 if self.op.readd and node == existing_node_name:
5420 if existing_node.secondary_ip != secondary_ip:
5421 raise errors.OpPrereqError("Readded node doesn't have the same IP"
5422 " address configuration as before",
5424 if existing_node.primary_ip != primary_ip:
5425 self.changed_primary_ip = True
5429 if (existing_node.primary_ip == primary_ip or
5430 existing_node.secondary_ip == primary_ip or
5431 existing_node.primary_ip == secondary_ip or
5432 existing_node.secondary_ip == secondary_ip):
5433 raise errors.OpPrereqError("New node ip address(es) conflict with"
5434 " existing node %s" % existing_node.name,
5435 errors.ECODE_NOTUNIQUE)
5437 # After this 'if' block, None is no longer a valid value for the
5438 # _capable op attributes
5440 old_node = self.cfg.GetNodeInfo(node)
5441 assert old_node is not None, "Can't retrieve locked node %s" % node
5442 for attr in self._NFLAGS:
5443 if getattr(self.op, attr) is None:
5444 setattr(self.op, attr, getattr(old_node, attr))
5446 for attr in self._NFLAGS:
5447 if getattr(self.op, attr) is None:
5448 setattr(self.op, attr, True)
5450 if self.op.readd and not self.op.vm_capable:
5451 pri, sec = cfg.GetNodeInstances(node)
5453 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
5454 " flag set to false, but it already holds"
5455 " instances" % node,
5458 # check that the type of the node (single versus dual homed) is the
5459 # same as for the master
5460 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
5461 master_singlehomed = myself.secondary_ip == myself.primary_ip
5462 newbie_singlehomed = secondary_ip == primary_ip
5463 if master_singlehomed != newbie_singlehomed:
5464 if master_singlehomed:
5465 raise errors.OpPrereqError("The master has no secondary ip but the"
5466 " new node has one",
5469 raise errors.OpPrereqError("The master has a secondary ip but the"
5470 " new node doesn't have one",
5473 # checks reachability
5474 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
5475 raise errors.OpPrereqError("Node not reachable by ping",
5476 errors.ECODE_ENVIRON)
5478 if not newbie_singlehomed:
5479 # check reachability from my secondary ip to newbie's secondary ip
5480 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
5481 source=myself.secondary_ip):
5482 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5483 " based ping to node daemon port",
5484 errors.ECODE_ENVIRON)
5491 if self.op.master_capable:
5492 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
5494 self.master_candidate = False
5497 self.new_node = old_node
5499 node_group = cfg.LookupNodeGroup(self.op.group)
5500 self.new_node = objects.Node(name=node,
5501 primary_ip=primary_ip,
5502 secondary_ip=secondary_ip,
5503 master_candidate=self.master_candidate,
5504 offline=False, drained=False,
5507 if self.op.ndparams:
5508 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
5510 if self.op.hv_state:
5511 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None)
5513 if self.op.disk_state:
5514 self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None)
5516 def Exec(self, feedback_fn):
5517 """Adds the new node to the cluster.
5520 new_node = self.new_node
5521 node = new_node.name
5523 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
5526 # We adding a new node so we assume it's powered
5527 new_node.powered = True
5529 # for re-adds, reset the offline/drained/master-candidate flags;
5530 # we need to reset here, otherwise offline would prevent RPC calls
5531 # later in the procedure; this also means that if the re-add
5532 # fails, we are left with a non-offlined, broken node
5534 new_node.drained = new_node.offline = False # pylint: disable=W0201
5535 self.LogInfo("Readding a node, the offline/drained flags were reset")
5536 # if we demote the node, we do cleanup later in the procedure
5537 new_node.master_candidate = self.master_candidate
5538 if self.changed_primary_ip:
5539 new_node.primary_ip = self.op.primary_ip
5541 # copy the master/vm_capable flags
5542 for attr in self._NFLAGS:
5543 setattr(new_node, attr, getattr(self.op, attr))
5545 # notify the user about any possible mc promotion
5546 if new_node.master_candidate:
5547 self.LogInfo("Node will be a master candidate")
5549 if self.op.ndparams:
5550 new_node.ndparams = self.op.ndparams
5552 new_node.ndparams = {}
5554 if self.op.hv_state:
5555 new_node.hv_state_static = self.new_hv_state
5557 if self.op.disk_state:
5558 new_node.disk_state_static = self.new_disk_state
5560 # check connectivity
5561 result = self.rpc.call_version([node])[node]
5562 result.Raise("Can't get version information from node %s" % node)
5563 if constants.PROTOCOL_VERSION == result.payload:
5564 logging.info("Communication to node %s fine, sw version %s match",
5565 node, result.payload)
5567 raise errors.OpExecError("Version mismatch master version %s,"
5568 " node version %s" %
5569 (constants.PROTOCOL_VERSION, result.payload))
5571 # Add node to our /etc/hosts, and add key to known_hosts
5572 if self.cfg.GetClusterInfo().modify_etc_hosts:
5573 master_node = self.cfg.GetMasterNode()
5574 result = self.rpc.call_etc_hosts_modify(master_node,
5575 constants.ETC_HOSTS_ADD,
5578 result.Raise("Can't update hosts file with new host data")
5580 if new_node.secondary_ip != new_node.primary_ip:
5581 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
5584 node_verify_list = [self.cfg.GetMasterNode()]
5585 node_verify_param = {
5586 constants.NV_NODELIST: ([node], {}),
5587 # TODO: do a node-net-test as well?
5590 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
5591 self.cfg.GetClusterName())
5592 for verifier in node_verify_list:
5593 result[verifier].Raise("Cannot communicate with node %s" % verifier)
5594 nl_payload = result[verifier].payload[constants.NV_NODELIST]
5596 for failed in nl_payload:
5597 feedback_fn("ssh/hostname verification failed"
5598 " (checking from %s): %s" %
5599 (verifier, nl_payload[failed]))
5600 raise errors.OpExecError("ssh/hostname verification failed")
5603 _RedistributeAncillaryFiles(self)
5604 self.context.ReaddNode(new_node)
5605 # make sure we redistribute the config
5606 self.cfg.Update(new_node, feedback_fn)
5607 # and make sure the new node will not have old files around
5608 if not new_node.master_candidate:
5609 result = self.rpc.call_node_demote_from_mc(new_node.name)
5610 msg = result.fail_msg
5612 self.LogWarning("Node failed to demote itself from master"
5613 " candidate status: %s" % msg)
5615 _RedistributeAncillaryFiles(self, additional_nodes=[node],
5616 additional_vm=self.op.vm_capable)
5617 self.context.AddNode(new_node, self.proc.GetECId())
5620 class LUNodeSetParams(LogicalUnit):
5621 """Modifies the parameters of a node.
5623 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
5624 to the node role (as _ROLE_*)
5625 @cvar _R2F: a dictionary from node role to tuples of flags
5626 @cvar _FLAGS: a list of attribute names corresponding to the flags
5629 HPATH = "node-modify"
5630 HTYPE = constants.HTYPE_NODE
5632 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
5634 (True, False, False): _ROLE_CANDIDATE,
5635 (False, True, False): _ROLE_DRAINED,
5636 (False, False, True): _ROLE_OFFLINE,
5637 (False, False, False): _ROLE_REGULAR,
5639 _R2F = dict((v, k) for k, v in _F2R.items())
5640 _FLAGS = ["master_candidate", "drained", "offline"]
5642 def CheckArguments(self):
5643 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5644 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
5645 self.op.master_capable, self.op.vm_capable,
5646 self.op.secondary_ip, self.op.ndparams, self.op.hv_state,
5648 if all_mods.count(None) == len(all_mods):
5649 raise errors.OpPrereqError("Please pass at least one modification",
5651 if all_mods.count(True) > 1:
5652 raise errors.OpPrereqError("Can't set the node into more than one"
5653 " state at the same time",
5656 # Boolean value that tells us whether we might be demoting from MC
5657 self.might_demote = (self.op.master_candidate == False or
5658 self.op.offline == True or
5659 self.op.drained == True or
5660 self.op.master_capable == False)
5662 if self.op.secondary_ip:
5663 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
5664 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5665 " address" % self.op.secondary_ip,
5668 self.lock_all = self.op.auto_promote and self.might_demote
5669 self.lock_instances = self.op.secondary_ip is not None
5671 def _InstanceFilter(self, instance):
5672 """Filter for getting affected instances.
5675 return (instance.disk_template in constants.DTS_INT_MIRROR and
5676 self.op.node_name in instance.all_nodes)
5678 def ExpandNames(self):
5680 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
5682 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
5684 # Since modifying a node can have severe effects on currently running
5685 # operations the resource lock is at least acquired in shared mode
5686 self.needed_locks[locking.LEVEL_NODE_RES] = \
5687 self.needed_locks[locking.LEVEL_NODE]
5689 # Get node resource and instance locks in shared mode; they are not used
5690 # for anything but read-only access
5691 self.share_locks[locking.LEVEL_NODE_RES] = 1
5692 self.share_locks[locking.LEVEL_INSTANCE] = 1
5694 if self.lock_instances:
5695 self.needed_locks[locking.LEVEL_INSTANCE] = \
5696 frozenset(self.cfg.GetInstancesInfoByFilter(self._InstanceFilter))
5698 def BuildHooksEnv(self):
5701 This runs on the master node.
5705 "OP_TARGET": self.op.node_name,
5706 "MASTER_CANDIDATE": str(self.op.master_candidate),
5707 "OFFLINE": str(self.op.offline),
5708 "DRAINED": str(self.op.drained),
5709 "MASTER_CAPABLE": str(self.op.master_capable),
5710 "VM_CAPABLE": str(self.op.vm_capable),
5713 def BuildHooksNodes(self):
5714 """Build hooks nodes.
5717 nl = [self.cfg.GetMasterNode(), self.op.node_name]
5720 def CheckPrereq(self):
5721 """Check prerequisites.
5723 This only checks the instance list against the existing names.
5726 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
5728 if self.lock_instances:
5729 affected_instances = \
5730 self.cfg.GetInstancesInfoByFilter(self._InstanceFilter)
5732 # Verify instance locks
5733 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
5734 wanted_instances = frozenset(affected_instances.keys())
5735 if wanted_instances - owned_instances:
5736 raise errors.OpPrereqError("Instances affected by changing node %s's"
5737 " secondary IP address have changed since"
5738 " locks were acquired, wanted '%s', have"
5739 " '%s'; retry the operation" %
5741 utils.CommaJoin(wanted_instances),
5742 utils.CommaJoin(owned_instances)),
5745 affected_instances = None
5747 if (self.op.master_candidate is not None or
5748 self.op.drained is not None or
5749 self.op.offline is not None):
5750 # we can't change the master's node flags
5751 if self.op.node_name == self.cfg.GetMasterNode():
5752 raise errors.OpPrereqError("The master role can be changed"
5753 " only via master-failover",
5756 if self.op.master_candidate and not node.master_capable:
5757 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
5758 " it a master candidate" % node.name,
5761 if self.op.vm_capable == False:
5762 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
5764 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
5765 " the vm_capable flag" % node.name,
5768 if node.master_candidate and self.might_demote and not self.lock_all:
5769 assert not self.op.auto_promote, "auto_promote set but lock_all not"
5770 # check if after removing the current node, we're missing master
5772 (mc_remaining, mc_should, _) = \
5773 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
5774 if mc_remaining < mc_should:
5775 raise errors.OpPrereqError("Not enough master candidates, please"
5776 " pass auto promote option to allow"
5777 " promotion", errors.ECODE_STATE)
5779 self.old_flags = old_flags = (node.master_candidate,
5780 node.drained, node.offline)
5781 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
5782 self.old_role = old_role = self._F2R[old_flags]
5784 # Check for ineffective changes
5785 for attr in self._FLAGS:
5786 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
5787 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
5788 setattr(self.op, attr, None)
5790 # Past this point, any flag change to False means a transition
5791 # away from the respective state, as only real changes are kept
5793 # TODO: We might query the real power state if it supports OOB
5794 if _SupportsOob(self.cfg, node):
5795 if self.op.offline is False and not (node.powered or
5796 self.op.powered == True):
5797 raise errors.OpPrereqError(("Node %s needs to be turned on before its"
5798 " offline status can be reset") %
5800 elif self.op.powered is not None:
5801 raise errors.OpPrereqError(("Unable to change powered state for node %s"
5802 " as it does not support out-of-band"
5803 " handling") % self.op.node_name)
5805 # If we're being deofflined/drained, we'll MC ourself if needed
5806 if (self.op.drained == False or self.op.offline == False or
5807 (self.op.master_capable and not node.master_capable)):
5808 if _DecideSelfPromotion(self):
5809 self.op.master_candidate = True
5810 self.LogInfo("Auto-promoting node to master candidate")
5812 # If we're no longer master capable, we'll demote ourselves from MC
5813 if self.op.master_capable == False and node.master_candidate:
5814 self.LogInfo("Demoting from master candidate")
5815 self.op.master_candidate = False
5818 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
5819 if self.op.master_candidate:
5820 new_role = self._ROLE_CANDIDATE
5821 elif self.op.drained:
5822 new_role = self._ROLE_DRAINED
5823 elif self.op.offline:
5824 new_role = self._ROLE_OFFLINE
5825 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
5826 # False is still in new flags, which means we're un-setting (the
5828 new_role = self._ROLE_REGULAR
5829 else: # no new flags, nothing, keep old role
5832 self.new_role = new_role
5834 if old_role == self._ROLE_OFFLINE and new_role != old_role:
5835 # Trying to transition out of offline status
5836 # TODO: Use standard RPC runner, but make sure it works when the node is
5837 # still marked offline
5838 result = rpc.BootstrapRunner().call_version([node.name])[node.name]
5840 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
5841 " to report its version: %s" %
5842 (node.name, result.fail_msg),
5845 self.LogWarning("Transitioning node from offline to online state"
5846 " without using re-add. Please make sure the node"
5849 if self.op.secondary_ip:
5850 # Ok even without locking, because this can't be changed by any LU
5851 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
5852 master_singlehomed = master.secondary_ip == master.primary_ip
5853 if master_singlehomed and self.op.secondary_ip:
5854 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
5855 " homed cluster", errors.ECODE_INVAL)
5857 assert not (frozenset(affected_instances) -
5858 self.owned_locks(locking.LEVEL_INSTANCE))
5861 if affected_instances:
5862 raise errors.OpPrereqError("Cannot change secondary IP address:"
5863 " offline node has instances (%s)"
5864 " configured to use it" %
5865 utils.CommaJoin(affected_instances.keys()))
5867 # On online nodes, check that no instances are running, and that
5868 # the node has the new ip and we can reach it.
5869 for instance in affected_instances.values():
5870 _CheckInstanceState(self, instance, INSTANCE_DOWN,
5871 msg="cannot change secondary ip")
5873 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
5874 if master.name != node.name:
5875 # check reachability from master secondary ip to new secondary ip
5876 if not netutils.TcpPing(self.op.secondary_ip,
5877 constants.DEFAULT_NODED_PORT,
5878 source=master.secondary_ip):
5879 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5880 " based ping to node daemon port",
5881 errors.ECODE_ENVIRON)
5883 if self.op.ndparams:
5884 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
5885 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
5886 self.new_ndparams = new_ndparams
5888 if self.op.hv_state:
5889 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
5890 self.node.hv_state_static)
5892 if self.op.disk_state:
5893 self.new_disk_state = \
5894 _MergeAndVerifyDiskState(self.op.disk_state,
5895 self.node.disk_state_static)
5897 def Exec(self, feedback_fn):
5902 old_role = self.old_role
5903 new_role = self.new_role
5907 if self.op.ndparams:
5908 node.ndparams = self.new_ndparams
5910 if self.op.powered is not None:
5911 node.powered = self.op.powered
5913 if self.op.hv_state:
5914 node.hv_state_static = self.new_hv_state
5916 if self.op.disk_state:
5917 node.disk_state_static = self.new_disk_state
5919 for attr in ["master_capable", "vm_capable"]:
5920 val = getattr(self.op, attr)
5922 setattr(node, attr, val)
5923 result.append((attr, str(val)))
5925 if new_role != old_role:
5926 # Tell the node to demote itself, if no longer MC and not offline
5927 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
5928 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
5930 self.LogWarning("Node failed to demote itself: %s", msg)
5932 new_flags = self._R2F[new_role]
5933 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
5935 result.append((desc, str(nf)))
5936 (node.master_candidate, node.drained, node.offline) = new_flags
5938 # we locked all nodes, we adjust the CP before updating this node
5940 _AdjustCandidatePool(self, [node.name])
5942 if self.op.secondary_ip:
5943 node.secondary_ip = self.op.secondary_ip
5944 result.append(("secondary_ip", self.op.secondary_ip))
5946 # this will trigger configuration file update, if needed
5947 self.cfg.Update(node, feedback_fn)
5949 # this will trigger job queue propagation or cleanup if the mc
5951 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
5952 self.context.ReaddNode(node)
5957 class LUNodePowercycle(NoHooksLU):
5958 """Powercycles a node.
5963 def CheckArguments(self):
5964 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5965 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
5966 raise errors.OpPrereqError("The node is the master and the force"
5967 " parameter was not set",
5970 def ExpandNames(self):
5971 """Locking for PowercycleNode.
5973 This is a last-resort option and shouldn't block on other
5974 jobs. Therefore, we grab no locks.
5977 self.needed_locks = {}
5979 def Exec(self, feedback_fn):
5983 result = self.rpc.call_node_powercycle(self.op.node_name,
5984 self.cfg.GetHypervisorType())
5985 result.Raise("Failed to schedule the reboot")
5986 return result.payload
5989 class LUClusterQuery(NoHooksLU):
5990 """Query cluster configuration.
5995 def ExpandNames(self):
5996 self.needed_locks = {}
5998 def Exec(self, feedback_fn):
5999 """Return cluster config.
6002 cluster = self.cfg.GetClusterInfo()
6005 # Filter just for enabled hypervisors
6006 for os_name, hv_dict in cluster.os_hvp.items():
6007 os_hvp[os_name] = {}
6008 for hv_name, hv_params in hv_dict.items():
6009 if hv_name in cluster.enabled_hypervisors:
6010 os_hvp[os_name][hv_name] = hv_params
6012 # Convert ip_family to ip_version
6013 primary_ip_version = constants.IP4_VERSION
6014 if cluster.primary_ip_family == netutils.IP6Address.family:
6015 primary_ip_version = constants.IP6_VERSION
6018 "software_version": constants.RELEASE_VERSION,
6019 "protocol_version": constants.PROTOCOL_VERSION,
6020 "config_version": constants.CONFIG_VERSION,
6021 "os_api_version": max(constants.OS_API_VERSIONS),
6022 "export_version": constants.EXPORT_VERSION,
6023 "architecture": (platform.architecture()[0], platform.machine()),
6024 "name": cluster.cluster_name,
6025 "master": cluster.master_node,
6026 "default_hypervisor": cluster.primary_hypervisor,
6027 "enabled_hypervisors": cluster.enabled_hypervisors,
6028 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
6029 for hypervisor_name in cluster.enabled_hypervisors]),
6031 "beparams": cluster.beparams,
6032 "osparams": cluster.osparams,
6033 "ipolicy": cluster.ipolicy,
6034 "nicparams": cluster.nicparams,
6035 "ndparams": cluster.ndparams,
6036 "candidate_pool_size": cluster.candidate_pool_size,
6037 "master_netdev": cluster.master_netdev,
6038 "master_netmask": cluster.master_netmask,
6039 "use_external_mip_script": cluster.use_external_mip_script,
6040 "volume_group_name": cluster.volume_group_name,
6041 "drbd_usermode_helper": cluster.drbd_usermode_helper,
6042 "file_storage_dir": cluster.file_storage_dir,
6043 "shared_file_storage_dir": cluster.shared_file_storage_dir,
6044 "maintain_node_health": cluster.maintain_node_health,
6045 "ctime": cluster.ctime,
6046 "mtime": cluster.mtime,
6047 "uuid": cluster.uuid,
6048 "tags": list(cluster.GetTags()),
6049 "uid_pool": cluster.uid_pool,
6050 "default_iallocator": cluster.default_iallocator,
6051 "reserved_lvs": cluster.reserved_lvs,
6052 "primary_ip_version": primary_ip_version,
6053 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
6054 "hidden_os": cluster.hidden_os,
6055 "blacklisted_os": cluster.blacklisted_os,
6061 class LUClusterConfigQuery(NoHooksLU):
6062 """Return configuration values.
6066 _FIELDS_DYNAMIC = utils.FieldSet()
6067 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
6068 "watcher_pause", "volume_group_name")
6070 def CheckArguments(self):
6071 _CheckOutputFields(static=self._FIELDS_STATIC,
6072 dynamic=self._FIELDS_DYNAMIC,
6073 selected=self.op.output_fields)
6075 def ExpandNames(self):
6076 self.needed_locks = {}
6078 def Exec(self, feedback_fn):
6079 """Dump a representation of the cluster config to the standard output.
6083 for field in self.op.output_fields:
6084 if field == "cluster_name":
6085 entry = self.cfg.GetClusterName()
6086 elif field == "master_node":
6087 entry = self.cfg.GetMasterNode()
6088 elif field == "drain_flag":
6089 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
6090 elif field == "watcher_pause":
6091 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
6092 elif field == "volume_group_name":
6093 entry = self.cfg.GetVGName()
6095 raise errors.ParameterError(field)
6096 values.append(entry)
6100 class LUInstanceActivateDisks(NoHooksLU):
6101 """Bring up an instance's disks.
6106 def ExpandNames(self):
6107 self._ExpandAndLockInstance()
6108 self.needed_locks[locking.LEVEL_NODE] = []
6109 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6111 def DeclareLocks(self, level):
6112 if level == locking.LEVEL_NODE:
6113 self._LockInstancesNodes()
6115 def CheckPrereq(self):
6116 """Check prerequisites.
6118 This checks that the instance is in the cluster.
6121 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6122 assert self.instance is not None, \
6123 "Cannot retrieve locked instance %s" % self.op.instance_name
6124 _CheckNodeOnline(self, self.instance.primary_node)
6126 def Exec(self, feedback_fn):
6127 """Activate the disks.
6130 disks_ok, disks_info = \
6131 _AssembleInstanceDisks(self, self.instance,
6132 ignore_size=self.op.ignore_size)
6134 raise errors.OpExecError("Cannot activate block devices")
6139 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
6141 """Prepare the block devices for an instance.
6143 This sets up the block devices on all nodes.
6145 @type lu: L{LogicalUnit}
6146 @param lu: the logical unit on whose behalf we execute
6147 @type instance: L{objects.Instance}
6148 @param instance: the instance for whose disks we assemble
6149 @type disks: list of L{objects.Disk} or None
6150 @param disks: which disks to assemble (or all, if None)
6151 @type ignore_secondaries: boolean
6152 @param ignore_secondaries: if true, errors on secondary nodes
6153 won't result in an error return from the function
6154 @type ignore_size: boolean
6155 @param ignore_size: if true, the current known size of the disk
6156 will not be used during the disk activation, useful for cases
6157 when the size is wrong
6158 @return: False if the operation failed, otherwise a list of
6159 (host, instance_visible_name, node_visible_name)
6160 with the mapping from node devices to instance devices
6165 iname = instance.name
6166 disks = _ExpandCheckDisks(instance, disks)
6168 # With the two passes mechanism we try to reduce the window of
6169 # opportunity for the race condition of switching DRBD to primary
6170 # before handshaking occured, but we do not eliminate it
6172 # The proper fix would be to wait (with some limits) until the
6173 # connection has been made and drbd transitions from WFConnection
6174 # into any other network-connected state (Connected, SyncTarget,
6177 # 1st pass, assemble on all nodes in secondary mode
6178 for idx, inst_disk in enumerate(disks):
6179 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
6181 node_disk = node_disk.Copy()
6182 node_disk.UnsetSize()
6183 lu.cfg.SetDiskID(node_disk, node)
6184 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
6185 msg = result.fail_msg
6187 lu.proc.LogWarning("Could not prepare block device %s on node %s"
6188 " (is_primary=False, pass=1): %s",
6189 inst_disk.iv_name, node, msg)
6190 if not ignore_secondaries:
6193 # FIXME: race condition on drbd migration to primary
6195 # 2nd pass, do only the primary node
6196 for idx, inst_disk in enumerate(disks):
6199 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
6200 if node != instance.primary_node:
6203 node_disk = node_disk.Copy()
6204 node_disk.UnsetSize()
6205 lu.cfg.SetDiskID(node_disk, node)
6206 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
6207 msg = result.fail_msg
6209 lu.proc.LogWarning("Could not prepare block device %s on node %s"
6210 " (is_primary=True, pass=2): %s",
6211 inst_disk.iv_name, node, msg)
6214 dev_path = result.payload
6216 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
6218 # leave the disks configured for the primary node
6219 # this is a workaround that would be fixed better by
6220 # improving the logical/physical id handling
6222 lu.cfg.SetDiskID(disk, instance.primary_node)
6224 return disks_ok, device_info
6227 def _StartInstanceDisks(lu, instance, force):
6228 """Start the disks of an instance.
6231 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
6232 ignore_secondaries=force)
6234 _ShutdownInstanceDisks(lu, instance)
6235 if force is not None and not force:
6236 lu.proc.LogWarning("", hint="If the message above refers to a"
6238 " you can retry the operation using '--force'.")
6239 raise errors.OpExecError("Disk consistency error")
6242 class LUInstanceDeactivateDisks(NoHooksLU):
6243 """Shutdown an instance's disks.
6248 def ExpandNames(self):
6249 self._ExpandAndLockInstance()
6250 self.needed_locks[locking.LEVEL_NODE] = []
6251 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6253 def DeclareLocks(self, level):
6254 if level == locking.LEVEL_NODE:
6255 self._LockInstancesNodes()
6257 def CheckPrereq(self):
6258 """Check prerequisites.
6260 This checks that the instance is in the cluster.
6263 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6264 assert self.instance is not None, \
6265 "Cannot retrieve locked instance %s" % self.op.instance_name
6267 def Exec(self, feedback_fn):
6268 """Deactivate the disks
6271 instance = self.instance
6273 _ShutdownInstanceDisks(self, instance)
6275 _SafeShutdownInstanceDisks(self, instance)
6278 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
6279 """Shutdown block devices of an instance.
6281 This function checks if an instance is running, before calling
6282 _ShutdownInstanceDisks.
6285 _CheckInstanceState(lu, instance, INSTANCE_DOWN, msg="cannot shutdown disks")
6286 _ShutdownInstanceDisks(lu, instance, disks=disks)
6289 def _ExpandCheckDisks(instance, disks):
6290 """Return the instance disks selected by the disks list
6292 @type disks: list of L{objects.Disk} or None
6293 @param disks: selected disks
6294 @rtype: list of L{objects.Disk}
6295 @return: selected instance disks to act on
6299 return instance.disks
6301 if not set(disks).issubset(instance.disks):
6302 raise errors.ProgrammerError("Can only act on disks belonging to the"
6307 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
6308 """Shutdown block devices of an instance.
6310 This does the shutdown on all nodes of the instance.
6312 If the ignore_primary is false, errors on the primary node are
6317 disks = _ExpandCheckDisks(instance, disks)
6320 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
6321 lu.cfg.SetDiskID(top_disk, node)
6322 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
6323 msg = result.fail_msg
6325 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
6326 disk.iv_name, node, msg)
6327 if ((node == instance.primary_node and not ignore_primary) or
6328 (node != instance.primary_node and not result.offline)):
6333 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
6334 """Checks if a node has enough free memory.
6336 This function check if a given node has the needed amount of free
6337 memory. In case the node has less memory or we cannot get the
6338 information from the node, this function raise an OpPrereqError
6341 @type lu: C{LogicalUnit}
6342 @param lu: a logical unit from which we get configuration data
6344 @param node: the node to check
6345 @type reason: C{str}
6346 @param reason: string to use in the error message
6347 @type requested: C{int}
6348 @param requested: the amount of memory in MiB to check for
6349 @type hypervisor_name: C{str}
6350 @param hypervisor_name: the hypervisor to ask for memory stats
6351 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
6352 we cannot check the node
6355 nodeinfo = lu.rpc.call_node_info([node], None, [hypervisor_name])
6356 nodeinfo[node].Raise("Can't get data from node %s" % node,
6357 prereq=True, ecode=errors.ECODE_ENVIRON)
6358 (_, _, (hv_info, )) = nodeinfo[node].payload
6360 free_mem = hv_info.get("memory_free", None)
6361 if not isinstance(free_mem, int):
6362 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
6363 " was '%s'" % (node, free_mem),
6364 errors.ECODE_ENVIRON)
6365 if requested > free_mem:
6366 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
6367 " needed %s MiB, available %s MiB" %
6368 (node, reason, requested, free_mem),
6372 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
6373 """Checks if nodes have enough free disk space in the all VGs.
6375 This function check if all given nodes have the needed amount of
6376 free disk. In case any node has less disk or we cannot get the
6377 information from the node, this function raise an OpPrereqError
6380 @type lu: C{LogicalUnit}
6381 @param lu: a logical unit from which we get configuration data
6382 @type nodenames: C{list}
6383 @param nodenames: the list of node names to check
6384 @type req_sizes: C{dict}
6385 @param req_sizes: the hash of vg and corresponding amount of disk in
6387 @raise errors.OpPrereqError: if the node doesn't have enough disk,
6388 or we cannot check the node
6391 for vg, req_size in req_sizes.items():
6392 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
6395 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
6396 """Checks if nodes have enough free disk space in the specified VG.
6398 This function check if all given nodes have the needed amount of
6399 free disk. In case any node has less disk or we cannot get the
6400 information from the node, this function raise an OpPrereqError
6403 @type lu: C{LogicalUnit}
6404 @param lu: a logical unit from which we get configuration data
6405 @type nodenames: C{list}
6406 @param nodenames: the list of node names to check
6408 @param vg: the volume group to check
6409 @type requested: C{int}
6410 @param requested: the amount of disk in MiB to check for
6411 @raise errors.OpPrereqError: if the node doesn't have enough disk,
6412 or we cannot check the node
6415 nodeinfo = lu.rpc.call_node_info(nodenames, [vg], None)
6416 for node in nodenames:
6417 info = nodeinfo[node]
6418 info.Raise("Cannot get current information from node %s" % node,
6419 prereq=True, ecode=errors.ECODE_ENVIRON)
6420 (_, (vg_info, ), _) = info.payload
6421 vg_free = vg_info.get("vg_free", None)
6422 if not isinstance(vg_free, int):
6423 raise errors.OpPrereqError("Can't compute free disk space on node"
6424 " %s for vg %s, result was '%s'" %
6425 (node, vg, vg_free), errors.ECODE_ENVIRON)
6426 if requested > vg_free:
6427 raise errors.OpPrereqError("Not enough disk space on target node %s"
6428 " vg %s: required %d MiB, available %d MiB" %
6429 (node, vg, requested, vg_free),
6433 def _CheckNodesPhysicalCPUs(lu, nodenames, requested, hypervisor_name):
6434 """Checks if nodes have enough physical CPUs
6436 This function checks if all given nodes have the needed number of
6437 physical CPUs. In case any node has less CPUs or we cannot get the
6438 information from the node, this function raises an OpPrereqError
6441 @type lu: C{LogicalUnit}
6442 @param lu: a logical unit from which we get configuration data
6443 @type nodenames: C{list}
6444 @param nodenames: the list of node names to check
6445 @type requested: C{int}
6446 @param requested: the minimum acceptable number of physical CPUs
6447 @raise errors.OpPrereqError: if the node doesn't have enough CPUs,
6448 or we cannot check the node
6451 nodeinfo = lu.rpc.call_node_info(nodenames, None, [hypervisor_name])
6452 for node in nodenames:
6453 info = nodeinfo[node]
6454 info.Raise("Cannot get current information from node %s" % node,
6455 prereq=True, ecode=errors.ECODE_ENVIRON)
6456 (_, _, (hv_info, )) = info.payload
6457 num_cpus = hv_info.get("cpu_total", None)
6458 if not isinstance(num_cpus, int):
6459 raise errors.OpPrereqError("Can't compute the number of physical CPUs"
6460 " on node %s, result was '%s'" %
6461 (node, num_cpus), errors.ECODE_ENVIRON)
6462 if requested > num_cpus:
6463 raise errors.OpPrereqError("Node %s has %s physical CPUs, but %s are "
6464 "required" % (node, num_cpus, requested),
6468 class LUInstanceStartup(LogicalUnit):
6469 """Starts an instance.
6472 HPATH = "instance-start"
6473 HTYPE = constants.HTYPE_INSTANCE
6476 def CheckArguments(self):
6478 if self.op.beparams:
6479 # fill the beparams dict
6480 objects.UpgradeBeParams(self.op.beparams)
6481 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6483 def ExpandNames(self):
6484 self._ExpandAndLockInstance()
6485 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
6487 def DeclareLocks(self, level):
6488 if level == locking.LEVEL_NODE_RES:
6489 self._LockInstancesNodes(primary_only=True, level=locking.LEVEL_NODE_RES)
6491 def BuildHooksEnv(self):
6494 This runs on master, primary and secondary nodes of the instance.
6498 "FORCE": self.op.force,
6501 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6505 def BuildHooksNodes(self):
6506 """Build hooks nodes.
6509 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6512 def CheckPrereq(self):
6513 """Check prerequisites.
6515 This checks that the instance is in the cluster.
6518 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6519 assert self.instance is not None, \
6520 "Cannot retrieve locked instance %s" % self.op.instance_name
6523 if self.op.hvparams:
6524 # check hypervisor parameter syntax (locally)
6525 cluster = self.cfg.GetClusterInfo()
6526 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6527 filled_hvp = cluster.FillHV(instance)
6528 filled_hvp.update(self.op.hvparams)
6529 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
6530 hv_type.CheckParameterSyntax(filled_hvp)
6531 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
6533 _CheckInstanceState(self, instance, INSTANCE_ONLINE)
6535 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
6537 if self.primary_offline and self.op.ignore_offline_nodes:
6538 self.proc.LogWarning("Ignoring offline primary node")
6540 if self.op.hvparams or self.op.beparams:
6541 self.proc.LogWarning("Overridden parameters are ignored")
6543 _CheckNodeOnline(self, instance.primary_node)
6545 bep = self.cfg.GetClusterInfo().FillBE(instance)
6546 bep.update(self.op.beparams)
6548 # check bridges existence
6549 _CheckInstanceBridgesExist(self, instance)
6551 remote_info = self.rpc.call_instance_info(instance.primary_node,
6553 instance.hypervisor)
6554 remote_info.Raise("Error checking node %s" % instance.primary_node,
6555 prereq=True, ecode=errors.ECODE_ENVIRON)
6556 if not remote_info.payload: # not running already
6557 _CheckNodeFreeMemory(self, instance.primary_node,
6558 "starting instance %s" % instance.name,
6559 bep[constants.BE_MINMEM], instance.hypervisor)
6561 def Exec(self, feedback_fn):
6562 """Start the instance.
6565 instance = self.instance
6566 force = self.op.force
6568 if not self.op.no_remember:
6569 self.cfg.MarkInstanceUp(instance.name)
6571 if self.primary_offline:
6572 assert self.op.ignore_offline_nodes
6573 self.proc.LogInfo("Primary node offline, marked instance as started")
6575 node_current = instance.primary_node
6577 _StartInstanceDisks(self, instance, force)
6580 self.rpc.call_instance_start(node_current,
6581 (instance, self.op.hvparams,
6583 self.op.startup_paused)
6584 msg = result.fail_msg
6586 _ShutdownInstanceDisks(self, instance)
6587 raise errors.OpExecError("Could not start instance: %s" % msg)
6590 class LUInstanceReboot(LogicalUnit):
6591 """Reboot an instance.
6594 HPATH = "instance-reboot"
6595 HTYPE = constants.HTYPE_INSTANCE
6598 def ExpandNames(self):
6599 self._ExpandAndLockInstance()
6601 def BuildHooksEnv(self):
6604 This runs on master, primary and secondary nodes of the instance.
6608 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
6609 "REBOOT_TYPE": self.op.reboot_type,
6610 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6613 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6617 def BuildHooksNodes(self):
6618 """Build hooks nodes.
6621 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6624 def CheckPrereq(self):
6625 """Check prerequisites.
6627 This checks that the instance is in the cluster.
6630 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6631 assert self.instance is not None, \
6632 "Cannot retrieve locked instance %s" % self.op.instance_name
6633 _CheckInstanceState(self, instance, INSTANCE_ONLINE)
6634 _CheckNodeOnline(self, instance.primary_node)
6636 # check bridges existence
6637 _CheckInstanceBridgesExist(self, instance)
6639 def Exec(self, feedback_fn):
6640 """Reboot the instance.
6643 instance = self.instance
6644 ignore_secondaries = self.op.ignore_secondaries
6645 reboot_type = self.op.reboot_type
6647 remote_info = self.rpc.call_instance_info(instance.primary_node,
6649 instance.hypervisor)
6650 remote_info.Raise("Error checking node %s" % instance.primary_node)
6651 instance_running = bool(remote_info.payload)
6653 node_current = instance.primary_node
6655 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
6656 constants.INSTANCE_REBOOT_HARD]:
6657 for disk in instance.disks:
6658 self.cfg.SetDiskID(disk, node_current)
6659 result = self.rpc.call_instance_reboot(node_current, instance,
6661 self.op.shutdown_timeout)
6662 result.Raise("Could not reboot instance")
6664 if instance_running:
6665 result = self.rpc.call_instance_shutdown(node_current, instance,
6666 self.op.shutdown_timeout)
6667 result.Raise("Could not shutdown instance for full reboot")
6668 _ShutdownInstanceDisks(self, instance)
6670 self.LogInfo("Instance %s was already stopped, starting now",
6672 _StartInstanceDisks(self, instance, ignore_secondaries)
6673 result = self.rpc.call_instance_start(node_current,
6674 (instance, None, None), False)
6675 msg = result.fail_msg
6677 _ShutdownInstanceDisks(self, instance)
6678 raise errors.OpExecError("Could not start instance for"
6679 " full reboot: %s" % msg)
6681 self.cfg.MarkInstanceUp(instance.name)
6684 class LUInstanceShutdown(LogicalUnit):
6685 """Shutdown an instance.
6688 HPATH = "instance-stop"
6689 HTYPE = constants.HTYPE_INSTANCE
6692 def ExpandNames(self):
6693 self._ExpandAndLockInstance()
6695 def BuildHooksEnv(self):
6698 This runs on master, primary and secondary nodes of the instance.
6701 env = _BuildInstanceHookEnvByObject(self, self.instance)
6702 env["TIMEOUT"] = self.op.timeout
6705 def BuildHooksNodes(self):
6706 """Build hooks nodes.
6709 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6712 def CheckPrereq(self):
6713 """Check prerequisites.
6715 This checks that the instance is in the cluster.
6718 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6719 assert self.instance is not None, \
6720 "Cannot retrieve locked instance %s" % self.op.instance_name
6722 _CheckInstanceState(self, self.instance, INSTANCE_ONLINE)
6724 self.primary_offline = \
6725 self.cfg.GetNodeInfo(self.instance.primary_node).offline
6727 if self.primary_offline and self.op.ignore_offline_nodes:
6728 self.proc.LogWarning("Ignoring offline primary node")
6730 _CheckNodeOnline(self, self.instance.primary_node)
6732 def Exec(self, feedback_fn):
6733 """Shutdown the instance.
6736 instance = self.instance
6737 node_current = instance.primary_node
6738 timeout = self.op.timeout
6740 if not self.op.no_remember:
6741 self.cfg.MarkInstanceDown(instance.name)
6743 if self.primary_offline:
6744 assert self.op.ignore_offline_nodes
6745 self.proc.LogInfo("Primary node offline, marked instance as stopped")
6747 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
6748 msg = result.fail_msg
6750 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
6752 _ShutdownInstanceDisks(self, instance)
6755 class LUInstanceReinstall(LogicalUnit):
6756 """Reinstall an instance.
6759 HPATH = "instance-reinstall"
6760 HTYPE = constants.HTYPE_INSTANCE
6763 def ExpandNames(self):
6764 self._ExpandAndLockInstance()
6766 def BuildHooksEnv(self):
6769 This runs on master, primary and secondary nodes of the instance.
6772 return _BuildInstanceHookEnvByObject(self, self.instance)
6774 def BuildHooksNodes(self):
6775 """Build hooks nodes.
6778 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6781 def CheckPrereq(self):
6782 """Check prerequisites.
6784 This checks that the instance is in the cluster and is not running.
6787 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6788 assert instance is not None, \
6789 "Cannot retrieve locked instance %s" % self.op.instance_name
6790 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
6791 " offline, cannot reinstall")
6792 for node in instance.secondary_nodes:
6793 _CheckNodeOnline(self, node, "Instance secondary node offline,"
6794 " cannot reinstall")
6796 if instance.disk_template == constants.DT_DISKLESS:
6797 raise errors.OpPrereqError("Instance '%s' has no disks" %
6798 self.op.instance_name,
6800 _CheckInstanceState(self, instance, INSTANCE_DOWN, msg="cannot reinstall")
6802 if self.op.os_type is not None:
6804 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
6805 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
6806 instance_os = self.op.os_type
6808 instance_os = instance.os
6810 nodelist = list(instance.all_nodes)
6812 if self.op.osparams:
6813 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
6814 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
6815 self.os_inst = i_osdict # the new dict (without defaults)
6819 self.instance = instance
6821 def Exec(self, feedback_fn):
6822 """Reinstall the instance.
6825 inst = self.instance
6827 if self.op.os_type is not None:
6828 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
6829 inst.os = self.op.os_type
6830 # Write to configuration
6831 self.cfg.Update(inst, feedback_fn)
6833 _StartInstanceDisks(self, inst, None)
6835 feedback_fn("Running the instance OS create scripts...")
6836 # FIXME: pass debug option from opcode to backend
6837 result = self.rpc.call_instance_os_add(inst.primary_node,
6838 (inst, self.os_inst), True,
6839 self.op.debug_level)
6840 result.Raise("Could not install OS for instance %s on node %s" %
6841 (inst.name, inst.primary_node))
6843 _ShutdownInstanceDisks(self, inst)
6846 class LUInstanceRecreateDisks(LogicalUnit):
6847 """Recreate an instance's missing disks.
6850 HPATH = "instance-recreate-disks"
6851 HTYPE = constants.HTYPE_INSTANCE
6854 def CheckArguments(self):
6855 # normalise the disk list
6856 self.op.disks = sorted(frozenset(self.op.disks))
6858 def ExpandNames(self):
6859 self._ExpandAndLockInstance()
6860 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6862 self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes]
6863 self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes)
6865 self.needed_locks[locking.LEVEL_NODE] = []
6867 def DeclareLocks(self, level):
6868 if level == locking.LEVEL_NODE:
6869 # if we replace the nodes, we only need to lock the old primary,
6870 # otherwise we need to lock all nodes for disk re-creation
6871 primary_only = bool(self.op.nodes)
6872 self._LockInstancesNodes(primary_only=primary_only)
6873 elif level == locking.LEVEL_NODE_RES:
6875 self.needed_locks[locking.LEVEL_NODE_RES] = \
6876 self.needed_locks[locking.LEVEL_NODE][:]
6878 def BuildHooksEnv(self):
6881 This runs on master, primary and secondary nodes of the instance.
6884 return _BuildInstanceHookEnvByObject(self, self.instance)
6886 def BuildHooksNodes(self):
6887 """Build hooks nodes.
6890 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6893 def CheckPrereq(self):
6894 """Check prerequisites.
6896 This checks that the instance is in the cluster and is not running.
6899 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6900 assert instance is not None, \
6901 "Cannot retrieve locked instance %s" % self.op.instance_name
6903 if len(self.op.nodes) != len(instance.all_nodes):
6904 raise errors.OpPrereqError("Instance %s currently has %d nodes, but"
6905 " %d replacement nodes were specified" %
6906 (instance.name, len(instance.all_nodes),
6907 len(self.op.nodes)),
6909 assert instance.disk_template != constants.DT_DRBD8 or \
6910 len(self.op.nodes) == 2
6911 assert instance.disk_template != constants.DT_PLAIN or \
6912 len(self.op.nodes) == 1
6913 primary_node = self.op.nodes[0]
6915 primary_node = instance.primary_node
6916 _CheckNodeOnline(self, primary_node)
6918 if instance.disk_template == constants.DT_DISKLESS:
6919 raise errors.OpPrereqError("Instance '%s' has no disks" %
6920 self.op.instance_name, errors.ECODE_INVAL)
6921 # if we replace nodes *and* the old primary is offline, we don't
6923 assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE)
6924 assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE_RES)
6925 old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
6926 if not (self.op.nodes and old_pnode.offline):
6927 _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
6928 msg="cannot recreate disks")
6930 if not self.op.disks:
6931 self.op.disks = range(len(instance.disks))
6933 for idx in self.op.disks:
6934 if idx >= len(instance.disks):
6935 raise errors.OpPrereqError("Invalid disk index '%s'" % idx,
6937 if self.op.disks != range(len(instance.disks)) and self.op.nodes:
6938 raise errors.OpPrereqError("Can't recreate disks partially and"
6939 " change the nodes at the same time",
6941 self.instance = instance
6943 def Exec(self, feedback_fn):
6944 """Recreate the disks.
6947 instance = self.instance
6949 assert (self.owned_locks(locking.LEVEL_NODE) ==
6950 self.owned_locks(locking.LEVEL_NODE_RES))
6953 mods = [] # keeps track of needed logical_id changes
6955 for idx, disk in enumerate(instance.disks):
6956 if idx not in self.op.disks: # disk idx has not been passed in
6959 # update secondaries for disks, if needed
6961 if disk.dev_type == constants.LD_DRBD8:
6962 # need to update the nodes and minors
6963 assert len(self.op.nodes) == 2
6964 assert len(disk.logical_id) == 6 # otherwise disk internals
6966 (_, _, old_port, _, _, old_secret) = disk.logical_id
6967 new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name)
6968 new_id = (self.op.nodes[0], self.op.nodes[1], old_port,
6969 new_minors[0], new_minors[1], old_secret)
6970 assert len(disk.logical_id) == len(new_id)
6971 mods.append((idx, new_id))
6973 # now that we have passed all asserts above, we can apply the mods
6974 # in a single run (to avoid partial changes)
6975 for idx, new_id in mods:
6976 instance.disks[idx].logical_id = new_id
6978 # change primary node, if needed
6980 instance.primary_node = self.op.nodes[0]
6981 self.LogWarning("Changing the instance's nodes, you will have to"
6982 " remove any disks left on the older nodes manually")
6985 self.cfg.Update(instance, feedback_fn)
6987 _CreateDisks(self, instance, to_skip=to_skip)
6990 class LUInstanceRename(LogicalUnit):
6991 """Rename an instance.
6994 HPATH = "instance-rename"
6995 HTYPE = constants.HTYPE_INSTANCE
6997 def CheckArguments(self):
7001 if self.op.ip_check and not self.op.name_check:
7002 # TODO: make the ip check more flexible and not depend on the name check
7003 raise errors.OpPrereqError("IP address check requires a name check",
7006 def BuildHooksEnv(self):
7009 This runs on master, primary and secondary nodes of the instance.
7012 env = _BuildInstanceHookEnvByObject(self, self.instance)
7013 env["INSTANCE_NEW_NAME"] = self.op.new_name
7016 def BuildHooksNodes(self):
7017 """Build hooks nodes.
7020 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7023 def CheckPrereq(self):
7024 """Check prerequisites.
7026 This checks that the instance is in the cluster and is not running.
7029 self.op.instance_name = _ExpandInstanceName(self.cfg,
7030 self.op.instance_name)
7031 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7032 assert instance is not None
7033 _CheckNodeOnline(self, instance.primary_node)
7034 _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
7035 msg="cannot rename")
7036 self.instance = instance
7038 new_name = self.op.new_name
7039 if self.op.name_check:
7040 hostname = netutils.GetHostname(name=new_name)
7041 if hostname.name != new_name:
7042 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
7044 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
7045 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
7046 " same as given hostname '%s'") %
7047 (hostname.name, self.op.new_name),
7049 new_name = self.op.new_name = hostname.name
7050 if (self.op.ip_check and
7051 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
7052 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7053 (hostname.ip, new_name),
7054 errors.ECODE_NOTUNIQUE)
7056 instance_list = self.cfg.GetInstanceList()
7057 if new_name in instance_list and new_name != instance.name:
7058 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7059 new_name, errors.ECODE_EXISTS)
7061 def Exec(self, feedback_fn):
7062 """Rename the instance.
7065 inst = self.instance
7066 old_name = inst.name
7068 rename_file_storage = False
7069 if (inst.disk_template in constants.DTS_FILEBASED and
7070 self.op.new_name != inst.name):
7071 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
7072 rename_file_storage = True
7074 self.cfg.RenameInstance(inst.name, self.op.new_name)
7075 # Change the instance lock. This is definitely safe while we hold the BGL.
7076 # Otherwise the new lock would have to be added in acquired mode.
7078 self.glm.remove(locking.LEVEL_INSTANCE, old_name)
7079 self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
7081 # re-read the instance from the configuration after rename
7082 inst = self.cfg.GetInstanceInfo(self.op.new_name)
7084 if rename_file_storage:
7085 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
7086 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
7087 old_file_storage_dir,
7088 new_file_storage_dir)
7089 result.Raise("Could not rename on node %s directory '%s' to '%s'"
7090 " (but the instance has been renamed in Ganeti)" %
7091 (inst.primary_node, old_file_storage_dir,
7092 new_file_storage_dir))
7094 _StartInstanceDisks(self, inst, None)
7096 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
7097 old_name, self.op.debug_level)
7098 msg = result.fail_msg
7100 msg = ("Could not run OS rename script for instance %s on node %s"
7101 " (but the instance has been renamed in Ganeti): %s" %
7102 (inst.name, inst.primary_node, msg))
7103 self.proc.LogWarning(msg)
7105 _ShutdownInstanceDisks(self, inst)
7110 class LUInstanceRemove(LogicalUnit):
7111 """Remove an instance.
7114 HPATH = "instance-remove"
7115 HTYPE = constants.HTYPE_INSTANCE
7118 def ExpandNames(self):
7119 self._ExpandAndLockInstance()
7120 self.needed_locks[locking.LEVEL_NODE] = []
7121 self.needed_locks[locking.LEVEL_NODE_RES] = []
7122 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7124 def DeclareLocks(self, level):
7125 if level == locking.LEVEL_NODE:
7126 self._LockInstancesNodes()
7127 elif level == locking.LEVEL_NODE_RES:
7129 self.needed_locks[locking.LEVEL_NODE_RES] = \
7130 self.needed_locks[locking.LEVEL_NODE][:]
7132 def BuildHooksEnv(self):
7135 This runs on master, primary and secondary nodes of the instance.
7138 env = _BuildInstanceHookEnvByObject(self, self.instance)
7139 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
7142 def BuildHooksNodes(self):
7143 """Build hooks nodes.
7146 nl = [self.cfg.GetMasterNode()]
7147 nl_post = list(self.instance.all_nodes) + nl
7148 return (nl, nl_post)
7150 def CheckPrereq(self):
7151 """Check prerequisites.
7153 This checks that the instance is in the cluster.
7156 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7157 assert self.instance is not None, \
7158 "Cannot retrieve locked instance %s" % self.op.instance_name
7160 def Exec(self, feedback_fn):
7161 """Remove the instance.
7164 instance = self.instance
7165 logging.info("Shutting down instance %s on node %s",
7166 instance.name, instance.primary_node)
7168 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
7169 self.op.shutdown_timeout)
7170 msg = result.fail_msg
7172 if self.op.ignore_failures:
7173 feedback_fn("Warning: can't shutdown instance: %s" % msg)
7175 raise errors.OpExecError("Could not shutdown instance %s on"
7177 (instance.name, instance.primary_node, msg))
7179 assert (self.owned_locks(locking.LEVEL_NODE) ==
7180 self.owned_locks(locking.LEVEL_NODE_RES))
7181 assert not (set(instance.all_nodes) -
7182 self.owned_locks(locking.LEVEL_NODE)), \
7183 "Not owning correct locks"
7185 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
7188 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
7189 """Utility function to remove an instance.
7192 logging.info("Removing block devices for instance %s", instance.name)
7194 if not _RemoveDisks(lu, instance):
7195 if not ignore_failures:
7196 raise errors.OpExecError("Can't remove instance's disks")
7197 feedback_fn("Warning: can't remove instance's disks")
7199 logging.info("Removing instance %s out of cluster config", instance.name)
7201 lu.cfg.RemoveInstance(instance.name)
7203 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
7204 "Instance lock removal conflict"
7206 # Remove lock for the instance
7207 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
7210 class LUInstanceQuery(NoHooksLU):
7211 """Logical unit for querying instances.
7214 # pylint: disable=W0142
7217 def CheckArguments(self):
7218 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
7219 self.op.output_fields, self.op.use_locking)
7221 def ExpandNames(self):
7222 self.iq.ExpandNames(self)
7224 def DeclareLocks(self, level):
7225 self.iq.DeclareLocks(self, level)
7227 def Exec(self, feedback_fn):
7228 return self.iq.OldStyleQuery(self)
7231 class LUInstanceFailover(LogicalUnit):
7232 """Failover an instance.
7235 HPATH = "instance-failover"
7236 HTYPE = constants.HTYPE_INSTANCE
7239 def CheckArguments(self):
7240 """Check the arguments.
7243 self.iallocator = getattr(self.op, "iallocator", None)
7244 self.target_node = getattr(self.op, "target_node", None)
7246 def ExpandNames(self):
7247 self._ExpandAndLockInstance()
7249 if self.op.target_node is not None:
7250 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7252 self.needed_locks[locking.LEVEL_NODE] = []
7253 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7255 ignore_consistency = self.op.ignore_consistency
7256 shutdown_timeout = self.op.shutdown_timeout
7257 self._migrater = TLMigrateInstance(self, self.op.instance_name,
7260 ignore_consistency=ignore_consistency,
7261 shutdown_timeout=shutdown_timeout,
7262 ignore_ipolicy=self.op.ignore_ipolicy)
7263 self.tasklets = [self._migrater]
7265 def DeclareLocks(self, level):
7266 if level == locking.LEVEL_NODE:
7267 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
7268 if instance.disk_template in constants.DTS_EXT_MIRROR:
7269 if self.op.target_node is None:
7270 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7272 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
7273 self.op.target_node]
7274 del self.recalculate_locks[locking.LEVEL_NODE]
7276 self._LockInstancesNodes()
7278 def BuildHooksEnv(self):
7281 This runs on master, primary and secondary nodes of the instance.
7284 instance = self._migrater.instance
7285 source_node = instance.primary_node
7286 target_node = self.op.target_node
7288 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
7289 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
7290 "OLD_PRIMARY": source_node,
7291 "NEW_PRIMARY": target_node,
7294 if instance.disk_template in constants.DTS_INT_MIRROR:
7295 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
7296 env["NEW_SECONDARY"] = source_node
7298 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
7300 env.update(_BuildInstanceHookEnvByObject(self, instance))
7304 def BuildHooksNodes(self):
7305 """Build hooks nodes.
7308 instance = self._migrater.instance
7309 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
7310 return (nl, nl + [instance.primary_node])
7313 class LUInstanceMigrate(LogicalUnit):
7314 """Migrate an instance.
7316 This is migration without shutting down, compared to the failover,
7317 which is done with shutdown.
7320 HPATH = "instance-migrate"
7321 HTYPE = constants.HTYPE_INSTANCE
7324 def ExpandNames(self):
7325 self._ExpandAndLockInstance()
7327 if self.op.target_node is not None:
7328 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7330 self.needed_locks[locking.LEVEL_NODE] = []
7331 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7333 self._migrater = TLMigrateInstance(self, self.op.instance_name,
7334 cleanup=self.op.cleanup,
7336 fallback=self.op.allow_failover,
7337 ignore_ipolicy=self.op.ignore_ipolicy)
7338 self.tasklets = [self._migrater]
7340 def DeclareLocks(self, level):
7341 if level == locking.LEVEL_NODE:
7342 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
7343 if instance.disk_template in constants.DTS_EXT_MIRROR:
7344 if self.op.target_node is None:
7345 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7347 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
7348 self.op.target_node]
7349 del self.recalculate_locks[locking.LEVEL_NODE]
7351 self._LockInstancesNodes()
7353 def BuildHooksEnv(self):
7356 This runs on master, primary and secondary nodes of the instance.
7359 instance = self._migrater.instance
7360 source_node = instance.primary_node
7361 target_node = self.op.target_node
7362 env = _BuildInstanceHookEnvByObject(self, instance)
7364 "MIGRATE_LIVE": self._migrater.live,
7365 "MIGRATE_CLEANUP": self.op.cleanup,
7366 "OLD_PRIMARY": source_node,
7367 "NEW_PRIMARY": target_node,
7370 if instance.disk_template in constants.DTS_INT_MIRROR:
7371 env["OLD_SECONDARY"] = target_node
7372 env["NEW_SECONDARY"] = source_node
7374 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
7378 def BuildHooksNodes(self):
7379 """Build hooks nodes.
7382 instance = self._migrater.instance
7383 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
7384 return (nl, nl + [instance.primary_node])
7387 class LUInstanceMove(LogicalUnit):
7388 """Move an instance by data-copying.
7391 HPATH = "instance-move"
7392 HTYPE = constants.HTYPE_INSTANCE
7395 def ExpandNames(self):
7396 self._ExpandAndLockInstance()
7397 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7398 self.op.target_node = target_node
7399 self.needed_locks[locking.LEVEL_NODE] = [target_node]
7400 self.needed_locks[locking.LEVEL_NODE_RES] = []
7401 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7403 def DeclareLocks(self, level):
7404 if level == locking.LEVEL_NODE:
7405 self._LockInstancesNodes(primary_only=True)
7406 elif level == locking.LEVEL_NODE_RES:
7408 self.needed_locks[locking.LEVEL_NODE_RES] = \
7409 self.needed_locks[locking.LEVEL_NODE][:]
7411 def BuildHooksEnv(self):
7414 This runs on master, primary and secondary nodes of the instance.
7418 "TARGET_NODE": self.op.target_node,
7419 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
7421 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7424 def BuildHooksNodes(self):
7425 """Build hooks nodes.
7429 self.cfg.GetMasterNode(),
7430 self.instance.primary_node,
7431 self.op.target_node,
7435 def CheckPrereq(self):
7436 """Check prerequisites.
7438 This checks that the instance is in the cluster.
7441 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7442 assert self.instance is not None, \
7443 "Cannot retrieve locked instance %s" % self.op.instance_name
7445 node = self.cfg.GetNodeInfo(self.op.target_node)
7446 assert node is not None, \
7447 "Cannot retrieve locked node %s" % self.op.target_node
7449 self.target_node = target_node = node.name
7451 if target_node == instance.primary_node:
7452 raise errors.OpPrereqError("Instance %s is already on the node %s" %
7453 (instance.name, target_node),
7456 bep = self.cfg.GetClusterInfo().FillBE(instance)
7458 for idx, dsk in enumerate(instance.disks):
7459 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
7460 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
7461 " cannot copy" % idx, errors.ECODE_STATE)
7463 _CheckNodeOnline(self, target_node)
7464 _CheckNodeNotDrained(self, target_node)
7465 _CheckNodeVmCapable(self, target_node)
7466 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(),
7467 self.cfg.GetNodeGroup(node.group))
7468 _CheckTargetNodeIPolicy(self, ipolicy, instance, node,
7469 ignore=self.op.ignore_ipolicy)
7471 if instance.admin_state == constants.ADMINST_UP:
7472 # check memory requirements on the secondary node
7473 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
7474 instance.name, bep[constants.BE_MAXMEM],
7475 instance.hypervisor)
7477 self.LogInfo("Not checking memory on the secondary node as"
7478 " instance will not be started")
7480 # check bridge existance
7481 _CheckInstanceBridgesExist(self, instance, node=target_node)
7483 def Exec(self, feedback_fn):
7484 """Move an instance.
7486 The move is done by shutting it down on its present node, copying
7487 the data over (slow) and starting it on the new node.
7490 instance = self.instance
7492 source_node = instance.primary_node
7493 target_node = self.target_node
7495 self.LogInfo("Shutting down instance %s on source node %s",
7496 instance.name, source_node)
7498 assert (self.owned_locks(locking.LEVEL_NODE) ==
7499 self.owned_locks(locking.LEVEL_NODE_RES))
7501 result = self.rpc.call_instance_shutdown(source_node, instance,
7502 self.op.shutdown_timeout)
7503 msg = result.fail_msg
7505 if self.op.ignore_consistency:
7506 self.proc.LogWarning("Could not shutdown instance %s on node %s."
7507 " Proceeding anyway. Please make sure node"
7508 " %s is down. Error details: %s",
7509 instance.name, source_node, source_node, msg)
7511 raise errors.OpExecError("Could not shutdown instance %s on"
7513 (instance.name, source_node, msg))
7515 # create the target disks
7517 _CreateDisks(self, instance, target_node=target_node)
7518 except errors.OpExecError:
7519 self.LogWarning("Device creation failed, reverting...")
7521 _RemoveDisks(self, instance, target_node=target_node)
7523 self.cfg.ReleaseDRBDMinors(instance.name)
7526 cluster_name = self.cfg.GetClusterInfo().cluster_name
7529 # activate, get path, copy the data over
7530 for idx, disk in enumerate(instance.disks):
7531 self.LogInfo("Copying data for disk %d", idx)
7532 result = self.rpc.call_blockdev_assemble(target_node, disk,
7533 instance.name, True, idx)
7535 self.LogWarning("Can't assemble newly created disk %d: %s",
7536 idx, result.fail_msg)
7537 errs.append(result.fail_msg)
7539 dev_path = result.payload
7540 result = self.rpc.call_blockdev_export(source_node, disk,
7541 target_node, dev_path,
7544 self.LogWarning("Can't copy data over for disk %d: %s",
7545 idx, result.fail_msg)
7546 errs.append(result.fail_msg)
7550 self.LogWarning("Some disks failed to copy, aborting")
7552 _RemoveDisks(self, instance, target_node=target_node)
7554 self.cfg.ReleaseDRBDMinors(instance.name)
7555 raise errors.OpExecError("Errors during disk copy: %s" %
7558 instance.primary_node = target_node
7559 self.cfg.Update(instance, feedback_fn)
7561 self.LogInfo("Removing the disks on the original node")
7562 _RemoveDisks(self, instance, target_node=source_node)
7564 # Only start the instance if it's marked as up
7565 if instance.admin_state == constants.ADMINST_UP:
7566 self.LogInfo("Starting instance %s on node %s",
7567 instance.name, target_node)
7569 disks_ok, _ = _AssembleInstanceDisks(self, instance,
7570 ignore_secondaries=True)
7572 _ShutdownInstanceDisks(self, instance)
7573 raise errors.OpExecError("Can't activate the instance's disks")
7575 result = self.rpc.call_instance_start(target_node,
7576 (instance, None, None), False)
7577 msg = result.fail_msg
7579 _ShutdownInstanceDisks(self, instance)
7580 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
7581 (instance.name, target_node, msg))
7584 class LUNodeMigrate(LogicalUnit):
7585 """Migrate all instances from a node.
7588 HPATH = "node-migrate"
7589 HTYPE = constants.HTYPE_NODE
7592 def CheckArguments(self):
7595 def ExpandNames(self):
7596 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
7598 self.share_locks = _ShareAll()
7599 self.needed_locks = {
7600 locking.LEVEL_NODE: [self.op.node_name],
7603 def BuildHooksEnv(self):
7606 This runs on the master, the primary and all the secondaries.
7610 "NODE_NAME": self.op.node_name,
7613 def BuildHooksNodes(self):
7614 """Build hooks nodes.
7617 nl = [self.cfg.GetMasterNode()]
7620 def CheckPrereq(self):
7623 def Exec(self, feedback_fn):
7624 # Prepare jobs for migration instances
7626 [opcodes.OpInstanceMigrate(instance_name=inst.name,
7629 iallocator=self.op.iallocator,
7630 target_node=self.op.target_node,
7631 ignore_ipolicy=self.op.ignore_ipolicy)]
7632 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name)
7635 # TODO: Run iallocator in this opcode and pass correct placement options to
7636 # OpInstanceMigrate. Since other jobs can modify the cluster between
7637 # running the iallocator and the actual migration, a good consistency model
7638 # will have to be found.
7640 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
7641 frozenset([self.op.node_name]))
7643 return ResultWithJobs(jobs)
7646 class TLMigrateInstance(Tasklet):
7647 """Tasklet class for instance migration.
7650 @ivar live: whether the migration will be done live or non-live;
7651 this variable is initalized only after CheckPrereq has run
7652 @type cleanup: boolean
7653 @ivar cleanup: Wheater we cleanup from a failed migration
7654 @type iallocator: string
7655 @ivar iallocator: The iallocator used to determine target_node
7656 @type target_node: string
7657 @ivar target_node: If given, the target_node to reallocate the instance to
7658 @type failover: boolean
7659 @ivar failover: Whether operation results in failover or migration
7660 @type fallback: boolean
7661 @ivar fallback: Whether fallback to failover is allowed if migration not
7663 @type ignore_consistency: boolean
7664 @ivar ignore_consistency: Wheter we should ignore consistency between source
7666 @type shutdown_timeout: int
7667 @ivar shutdown_timeout: In case of failover timeout of the shutdown
7668 @type ignore_ipolicy: bool
7669 @ivar ignore_ipolicy: If true, we can ignore instance policy when migrating
7674 _MIGRATION_POLL_INTERVAL = 1 # seconds
7675 _MIGRATION_FEEDBACK_INTERVAL = 10 # seconds
7677 def __init__(self, lu, instance_name, cleanup=False,
7678 failover=False, fallback=False,
7679 ignore_consistency=False,
7680 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT,
7681 ignore_ipolicy=False):
7682 """Initializes this class.
7685 Tasklet.__init__(self, lu)
7688 self.instance_name = instance_name
7689 self.cleanup = cleanup
7690 self.live = False # will be overridden later
7691 self.failover = failover
7692 self.fallback = fallback
7693 self.ignore_consistency = ignore_consistency
7694 self.shutdown_timeout = shutdown_timeout
7695 self.ignore_ipolicy = ignore_ipolicy
7697 def CheckPrereq(self):
7698 """Check prerequisites.
7700 This checks that the instance is in the cluster.
7703 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
7704 instance = self.cfg.GetInstanceInfo(instance_name)
7705 assert instance is not None
7706 self.instance = instance
7707 cluster = self.cfg.GetClusterInfo()
7709 if (not self.cleanup and
7710 not instance.admin_state == constants.ADMINST_UP and
7711 not self.failover and self.fallback):
7712 self.lu.LogInfo("Instance is marked down or offline, fallback allowed,"
7713 " switching to failover")
7714 self.failover = True
7716 if instance.disk_template not in constants.DTS_MIRRORED:
7721 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
7722 " %s" % (instance.disk_template, text),
7725 if instance.disk_template in constants.DTS_EXT_MIRROR:
7726 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
7728 if self.lu.op.iallocator:
7729 self._RunAllocator()
7731 # We set set self.target_node as it is required by
7733 self.target_node = self.lu.op.target_node
7735 # Check that the target node is correct in terms of instance policy
7736 nodeinfo = self.cfg.GetNodeInfo(self.target_node)
7737 group_info = self.cfg.GetNodeGroup(nodeinfo.group)
7738 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
7739 _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
7740 ignore=self.ignore_ipolicy)
7742 # self.target_node is already populated, either directly or by the
7744 target_node = self.target_node
7745 if self.target_node == instance.primary_node:
7746 raise errors.OpPrereqError("Cannot migrate instance %s"
7747 " to its primary (%s)" %
7748 (instance.name, instance.primary_node))
7750 if len(self.lu.tasklets) == 1:
7751 # It is safe to release locks only when we're the only tasklet
7753 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
7754 keep=[instance.primary_node, self.target_node])
7757 secondary_nodes = instance.secondary_nodes
7758 if not secondary_nodes:
7759 raise errors.ConfigurationError("No secondary node but using"
7760 " %s disk template" %
7761 instance.disk_template)
7762 target_node = secondary_nodes[0]
7763 if self.lu.op.iallocator or (self.lu.op.target_node and
7764 self.lu.op.target_node != target_node):
7766 text = "failed over"
7769 raise errors.OpPrereqError("Instances with disk template %s cannot"
7770 " be %s to arbitrary nodes"
7771 " (neither an iallocator nor a target"
7772 " node can be passed)" %
7773 (instance.disk_template, text),
7775 nodeinfo = self.cfg.GetNodeInfo(target_node)
7776 group_info = self.cfg.GetNodeGroup(nodeinfo.group)
7777 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
7778 _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
7779 ignore=self.ignore_ipolicy)
7781 i_be = cluster.FillBE(instance)
7783 # check memory requirements on the secondary node
7784 if (not self.cleanup and
7785 (not self.failover or instance.admin_state == constants.ADMINST_UP)):
7786 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
7787 instance.name, i_be[constants.BE_MAXMEM],
7788 instance.hypervisor)
7790 self.lu.LogInfo("Not checking memory on the secondary node as"
7791 " instance will not be started")
7793 # check if failover must be forced instead of migration
7794 if (not self.cleanup and not self.failover and
7795 i_be[constants.BE_ALWAYS_FAILOVER]):
7797 self.lu.LogInfo("Instance configured to always failover; fallback"
7799 self.failover = True
7801 raise errors.OpPrereqError("This instance has been configured to"
7802 " always failover, please allow failover",
7805 # check bridge existance
7806 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
7808 if not self.cleanup:
7809 _CheckNodeNotDrained(self.lu, target_node)
7810 if not self.failover:
7811 result = self.rpc.call_instance_migratable(instance.primary_node,
7813 if result.fail_msg and self.fallback:
7814 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
7816 self.failover = True
7818 result.Raise("Can't migrate, please use failover",
7819 prereq=True, ecode=errors.ECODE_STATE)
7821 assert not (self.failover and self.cleanup)
7823 if not self.failover:
7824 if self.lu.op.live is not None and self.lu.op.mode is not None:
7825 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
7826 " parameters are accepted",
7828 if self.lu.op.live is not None:
7830 self.lu.op.mode = constants.HT_MIGRATION_LIVE
7832 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
7833 # reset the 'live' parameter to None so that repeated
7834 # invocations of CheckPrereq do not raise an exception
7835 self.lu.op.live = None
7836 elif self.lu.op.mode is None:
7837 # read the default value from the hypervisor
7838 i_hv = cluster.FillHV(self.instance, skip_globals=False)
7839 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
7841 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
7843 # Failover is never live
7846 def _RunAllocator(self):
7847 """Run the allocator based on input opcode.
7850 # FIXME: add a self.ignore_ipolicy option
7851 ial = IAllocator(self.cfg, self.rpc,
7852 mode=constants.IALLOCATOR_MODE_RELOC,
7853 name=self.instance_name,
7854 # TODO See why hail breaks with a single node below
7855 relocate_from=[self.instance.primary_node,
7856 self.instance.primary_node],
7859 ial.Run(self.lu.op.iallocator)
7862 raise errors.OpPrereqError("Can't compute nodes using"
7863 " iallocator '%s': %s" %
7864 (self.lu.op.iallocator, ial.info),
7866 if len(ial.result) != ial.required_nodes:
7867 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7868 " of nodes (%s), required %s" %
7869 (self.lu.op.iallocator, len(ial.result),
7870 ial.required_nodes), errors.ECODE_FAULT)
7871 self.target_node = ial.result[0]
7872 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7873 self.instance_name, self.lu.op.iallocator,
7874 utils.CommaJoin(ial.result))
7876 def _WaitUntilSync(self):
7877 """Poll with custom rpc for disk sync.
7879 This uses our own step-based rpc call.
7882 self.feedback_fn("* wait until resync is done")
7886 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
7888 self.instance.disks)
7890 for node, nres in result.items():
7891 nres.Raise("Cannot resync disks on node %s" % node)
7892 node_done, node_percent = nres.payload
7893 all_done = all_done and node_done
7894 if node_percent is not None:
7895 min_percent = min(min_percent, node_percent)
7897 if min_percent < 100:
7898 self.feedback_fn(" - progress: %.1f%%" % min_percent)
7901 def _EnsureSecondary(self, node):
7902 """Demote a node to secondary.
7905 self.feedback_fn("* switching node %s to secondary mode" % node)
7907 for dev in self.instance.disks:
7908 self.cfg.SetDiskID(dev, node)
7910 result = self.rpc.call_blockdev_close(node, self.instance.name,
7911 self.instance.disks)
7912 result.Raise("Cannot change disk to secondary on node %s" % node)
7914 def _GoStandalone(self):
7915 """Disconnect from the network.
7918 self.feedback_fn("* changing into standalone mode")
7919 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
7920 self.instance.disks)
7921 for node, nres in result.items():
7922 nres.Raise("Cannot disconnect disks node %s" % node)
7924 def _GoReconnect(self, multimaster):
7925 """Reconnect to the network.
7931 msg = "single-master"
7932 self.feedback_fn("* changing disks into %s mode" % msg)
7933 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
7934 self.instance.disks,
7935 self.instance.name, multimaster)
7936 for node, nres in result.items():
7937 nres.Raise("Cannot change disks config on node %s" % node)
7939 def _ExecCleanup(self):
7940 """Try to cleanup after a failed migration.
7942 The cleanup is done by:
7943 - check that the instance is running only on one node
7944 (and update the config if needed)
7945 - change disks on its secondary node to secondary
7946 - wait until disks are fully synchronized
7947 - disconnect from the network
7948 - change disks into single-master mode
7949 - wait again until disks are fully synchronized
7952 instance = self.instance
7953 target_node = self.target_node
7954 source_node = self.source_node
7956 # check running on only one node
7957 self.feedback_fn("* checking where the instance actually runs"
7958 " (if this hangs, the hypervisor might be in"
7960 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
7961 for node, result in ins_l.items():
7962 result.Raise("Can't contact node %s" % node)
7964 runningon_source = instance.name in ins_l[source_node].payload
7965 runningon_target = instance.name in ins_l[target_node].payload
7967 if runningon_source and runningon_target:
7968 raise errors.OpExecError("Instance seems to be running on two nodes,"
7969 " or the hypervisor is confused; you will have"
7970 " to ensure manually that it runs only on one"
7971 " and restart this operation")
7973 if not (runningon_source or runningon_target):
7974 raise errors.OpExecError("Instance does not seem to be running at all;"
7975 " in this case it's safer to repair by"
7976 " running 'gnt-instance stop' to ensure disk"
7977 " shutdown, and then restarting it")
7979 if runningon_target:
7980 # the migration has actually succeeded, we need to update the config
7981 self.feedback_fn("* instance running on secondary node (%s),"
7982 " updating config" % target_node)
7983 instance.primary_node = target_node
7984 self.cfg.Update(instance, self.feedback_fn)
7985 demoted_node = source_node
7987 self.feedback_fn("* instance confirmed to be running on its"
7988 " primary node (%s)" % source_node)
7989 demoted_node = target_node
7991 if instance.disk_template in constants.DTS_INT_MIRROR:
7992 self._EnsureSecondary(demoted_node)
7994 self._WaitUntilSync()
7995 except errors.OpExecError:
7996 # we ignore here errors, since if the device is standalone, it
7997 # won't be able to sync
7999 self._GoStandalone()
8000 self._GoReconnect(False)
8001 self._WaitUntilSync()
8003 self.feedback_fn("* done")
8005 def _RevertDiskStatus(self):
8006 """Try to revert the disk status after a failed migration.
8009 target_node = self.target_node
8010 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
8014 self._EnsureSecondary(target_node)
8015 self._GoStandalone()
8016 self._GoReconnect(False)
8017 self._WaitUntilSync()
8018 except errors.OpExecError, err:
8019 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
8020 " please try to recover the instance manually;"
8021 " error '%s'" % str(err))
8023 def _AbortMigration(self):
8024 """Call the hypervisor code to abort a started migration.
8027 instance = self.instance
8028 target_node = self.target_node
8029 source_node = self.source_node
8030 migration_info = self.migration_info
8032 abort_result = self.rpc.call_instance_finalize_migration_dst(target_node,
8036 abort_msg = abort_result.fail_msg
8038 logging.error("Aborting migration failed on target node %s: %s",
8039 target_node, abort_msg)
8040 # Don't raise an exception here, as we stil have to try to revert the
8041 # disk status, even if this step failed.
8043 abort_result = self.rpc.call_instance_finalize_migration_src(source_node,
8044 instance, False, self.live)
8045 abort_msg = abort_result.fail_msg
8047 logging.error("Aborting migration failed on source node %s: %s",
8048 source_node, abort_msg)
8050 def _ExecMigration(self):
8051 """Migrate an instance.
8053 The migrate is done by:
8054 - change the disks into dual-master mode
8055 - wait until disks are fully synchronized again
8056 - migrate the instance
8057 - change disks on the new secondary node (the old primary) to secondary
8058 - wait until disks are fully synchronized
8059 - change disks into single-master mode
8062 instance = self.instance
8063 target_node = self.target_node
8064 source_node = self.source_node
8066 # Check for hypervisor version mismatch and warn the user.
8067 nodeinfo = self.rpc.call_node_info([source_node, target_node],
8068 None, [self.instance.hypervisor])
8069 for ninfo in nodeinfo.values():
8070 ninfo.Raise("Unable to retrieve node information from node '%s'" %
8072 (_, _, (src_info, )) = nodeinfo[source_node].payload
8073 (_, _, (dst_info, )) = nodeinfo[target_node].payload
8075 if ((constants.HV_NODEINFO_KEY_VERSION in src_info) and
8076 (constants.HV_NODEINFO_KEY_VERSION in dst_info)):
8077 src_version = src_info[constants.HV_NODEINFO_KEY_VERSION]
8078 dst_version = dst_info[constants.HV_NODEINFO_KEY_VERSION]
8079 if src_version != dst_version:
8080 self.feedback_fn("* warning: hypervisor version mismatch between"
8081 " source (%s) and target (%s) node" %
8082 (src_version, dst_version))
8084 self.feedback_fn("* checking disk consistency between source and target")
8085 for dev in instance.disks:
8086 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
8087 raise errors.OpExecError("Disk %s is degraded or not fully"
8088 " synchronized on target node,"
8089 " aborting migration" % dev.iv_name)
8091 # First get the migration information from the remote node
8092 result = self.rpc.call_migration_info(source_node, instance)
8093 msg = result.fail_msg
8095 log_err = ("Failed fetching source migration information from %s: %s" %
8097 logging.error(log_err)
8098 raise errors.OpExecError(log_err)
8100 self.migration_info = migration_info = result.payload
8102 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
8103 # Then switch the disks to master/master mode
8104 self._EnsureSecondary(target_node)
8105 self._GoStandalone()
8106 self._GoReconnect(True)
8107 self._WaitUntilSync()
8109 self.feedback_fn("* preparing %s to accept the instance" % target_node)
8110 result = self.rpc.call_accept_instance(target_node,
8113 self.nodes_ip[target_node])
8115 msg = result.fail_msg
8117 logging.error("Instance pre-migration failed, trying to revert"
8118 " disk status: %s", msg)
8119 self.feedback_fn("Pre-migration failed, aborting")
8120 self._AbortMigration()
8121 self._RevertDiskStatus()
8122 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
8123 (instance.name, msg))
8125 self.feedback_fn("* migrating instance to %s" % target_node)
8126 result = self.rpc.call_instance_migrate(source_node, instance,
8127 self.nodes_ip[target_node],
8129 msg = result.fail_msg
8131 logging.error("Instance migration failed, trying to revert"
8132 " disk status: %s", msg)
8133 self.feedback_fn("Migration failed, aborting")
8134 self._AbortMigration()
8135 self._RevertDiskStatus()
8136 raise errors.OpExecError("Could not migrate instance %s: %s" %
8137 (instance.name, msg))
8139 self.feedback_fn("* starting memory transfer")
8140 last_feedback = time.time()
8142 result = self.rpc.call_instance_get_migration_status(source_node,
8144 msg = result.fail_msg
8145 ms = result.payload # MigrationStatus instance
8146 if msg or (ms.status in constants.HV_MIGRATION_FAILED_STATUSES):
8147 logging.error("Instance migration failed, trying to revert"
8148 " disk status: %s", msg)
8149 self.feedback_fn("Migration failed, aborting")
8150 self._AbortMigration()
8151 self._RevertDiskStatus()
8152 raise errors.OpExecError("Could not migrate instance %s: %s" %
8153 (instance.name, msg))
8155 if result.payload.status != constants.HV_MIGRATION_ACTIVE:
8156 self.feedback_fn("* memory transfer complete")
8159 if (utils.TimeoutExpired(last_feedback,
8160 self._MIGRATION_FEEDBACK_INTERVAL) and
8161 ms.transferred_ram is not None):
8162 mem_progress = 100 * float(ms.transferred_ram) / float(ms.total_ram)
8163 self.feedback_fn("* memory transfer progress: %.2f %%" % mem_progress)
8164 last_feedback = time.time()
8166 time.sleep(self._MIGRATION_POLL_INTERVAL)
8168 result = self.rpc.call_instance_finalize_migration_src(source_node,
8172 msg = result.fail_msg
8174 logging.error("Instance migration succeeded, but finalization failed"
8175 " on the source node: %s", msg)
8176 raise errors.OpExecError("Could not finalize instance migration: %s" %
8179 instance.primary_node = target_node
8181 # distribute new instance config to the other nodes
8182 self.cfg.Update(instance, self.feedback_fn)
8184 result = self.rpc.call_instance_finalize_migration_dst(target_node,
8188 msg = result.fail_msg
8190 logging.error("Instance migration succeeded, but finalization failed"
8191 " on the target node: %s", msg)
8192 raise errors.OpExecError("Could not finalize instance migration: %s" %
8195 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
8196 self._EnsureSecondary(source_node)
8197 self._WaitUntilSync()
8198 self._GoStandalone()
8199 self._GoReconnect(False)
8200 self._WaitUntilSync()
8202 self.feedback_fn("* done")
8204 def _ExecFailover(self):
8205 """Failover an instance.
8207 The failover is done by shutting it down on its present node and
8208 starting it on the secondary.
8211 instance = self.instance
8212 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
8214 source_node = instance.primary_node
8215 target_node = self.target_node
8217 if instance.admin_state == constants.ADMINST_UP:
8218 self.feedback_fn("* checking disk consistency between source and target")
8219 for dev in instance.disks:
8220 # for drbd, these are drbd over lvm
8221 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
8222 if primary_node.offline:
8223 self.feedback_fn("Node %s is offline, ignoring degraded disk %s on"
8225 (primary_node.name, dev.iv_name, target_node))
8226 elif not self.ignore_consistency:
8227 raise errors.OpExecError("Disk %s is degraded on target node,"
8228 " aborting failover" % dev.iv_name)
8230 self.feedback_fn("* not checking disk consistency as instance is not"
8233 self.feedback_fn("* shutting down instance on source node")
8234 logging.info("Shutting down instance %s on node %s",
8235 instance.name, source_node)
8237 result = self.rpc.call_instance_shutdown(source_node, instance,
8238 self.shutdown_timeout)
8239 msg = result.fail_msg
8241 if self.ignore_consistency or primary_node.offline:
8242 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
8243 " proceeding anyway; please make sure node"
8244 " %s is down; error details: %s",
8245 instance.name, source_node, source_node, msg)
8247 raise errors.OpExecError("Could not shutdown instance %s on"
8249 (instance.name, source_node, msg))
8251 self.feedback_fn("* deactivating the instance's disks on source node")
8252 if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True):
8253 raise errors.OpExecError("Can't shut down the instance's disks")
8255 instance.primary_node = target_node
8256 # distribute new instance config to the other nodes
8257 self.cfg.Update(instance, self.feedback_fn)
8259 # Only start the instance if it's marked as up
8260 if instance.admin_state == constants.ADMINST_UP:
8261 self.feedback_fn("* activating the instance's disks on target node %s" %
8263 logging.info("Starting instance %s on node %s",
8264 instance.name, target_node)
8266 disks_ok, _ = _AssembleInstanceDisks(self.lu, instance,
8267 ignore_secondaries=True)
8269 _ShutdownInstanceDisks(self.lu, instance)
8270 raise errors.OpExecError("Can't activate the instance's disks")
8272 self.feedback_fn("* starting the instance on the target node %s" %
8274 result = self.rpc.call_instance_start(target_node, (instance, None, None),
8276 msg = result.fail_msg
8278 _ShutdownInstanceDisks(self.lu, instance)
8279 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
8280 (instance.name, target_node, msg))
8282 def Exec(self, feedback_fn):
8283 """Perform the migration.
8286 self.feedback_fn = feedback_fn
8287 self.source_node = self.instance.primary_node
8289 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
8290 if self.instance.disk_template in constants.DTS_INT_MIRROR:
8291 self.target_node = self.instance.secondary_nodes[0]
8292 # Otherwise self.target_node has been populated either
8293 # directly, or through an iallocator.
8295 self.all_nodes = [self.source_node, self.target_node]
8296 self.nodes_ip = dict((name, node.secondary_ip) for (name, node)
8297 in self.cfg.GetMultiNodeInfo(self.all_nodes))
8300 feedback_fn("Failover instance %s" % self.instance.name)
8301 self._ExecFailover()
8303 feedback_fn("Migrating instance %s" % self.instance.name)
8306 return self._ExecCleanup()
8308 return self._ExecMigration()
8311 def _CreateBlockDev(lu, node, instance, device, force_create,
8313 """Create a tree of block devices on a given node.
8315 If this device type has to be created on secondaries, create it and
8318 If not, just recurse to children keeping the same 'force' value.
8320 @param lu: the lu on whose behalf we execute
8321 @param node: the node on which to create the device
8322 @type instance: L{objects.Instance}
8323 @param instance: the instance which owns the device
8324 @type device: L{objects.Disk}
8325 @param device: the device to create
8326 @type force_create: boolean
8327 @param force_create: whether to force creation of this device; this
8328 will be change to True whenever we find a device which has
8329 CreateOnSecondary() attribute
8330 @param info: the extra 'metadata' we should attach to the device
8331 (this will be represented as a LVM tag)
8332 @type force_open: boolean
8333 @param force_open: this parameter will be passes to the
8334 L{backend.BlockdevCreate} function where it specifies
8335 whether we run on primary or not, and it affects both
8336 the child assembly and the device own Open() execution
8339 if device.CreateOnSecondary():
8343 for child in device.children:
8344 _CreateBlockDev(lu, node, instance, child, force_create,
8347 if not force_create:
8350 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
8353 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
8354 """Create a single block device on a given node.
8356 This will not recurse over children of the device, so they must be
8359 @param lu: the lu on whose behalf we execute
8360 @param node: the node on which to create the device
8361 @type instance: L{objects.Instance}
8362 @param instance: the instance which owns the device
8363 @type device: L{objects.Disk}
8364 @param device: the device to create
8365 @param info: the extra 'metadata' we should attach to the device
8366 (this will be represented as a LVM tag)
8367 @type force_open: boolean
8368 @param force_open: this parameter will be passes to the
8369 L{backend.BlockdevCreate} function where it specifies
8370 whether we run on primary or not, and it affects both
8371 the child assembly and the device own Open() execution
8374 lu.cfg.SetDiskID(device, node)
8375 result = lu.rpc.call_blockdev_create(node, device, device.size,
8376 instance.name, force_open, info)
8377 result.Raise("Can't create block device %s on"
8378 " node %s for instance %s" % (device, node, instance.name))
8379 if device.physical_id is None:
8380 device.physical_id = result.payload
8383 def _GenerateUniqueNames(lu, exts):
8384 """Generate a suitable LV name.
8386 This will generate a logical volume name for the given instance.
8391 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
8392 results.append("%s%s" % (new_id, val))
8396 def _ComputeLDParams(disk_template, disk_params):
8397 """Computes Logical Disk parameters from Disk Template parameters.
8399 @type disk_template: string
8400 @param disk_template: disk template, one of L{constants.DISK_TEMPLATES}
8401 @type disk_params: dict
8402 @param disk_params: disk template parameters; dict(template_name -> parameters
8404 @return: a list of dicts, one for each node of the disk hierarchy. Each dict
8405 contains the LD parameters of the node. The tree is flattened in-order.
8408 if disk_template not in constants.DISK_TEMPLATES:
8409 raise errors.ProgrammerError("Unknown disk template %s" % disk_template)
8412 dt_params = disk_params[disk_template]
8413 if disk_template == constants.DT_DRBD8:
8415 constants.LDP_RESYNC_RATE: dt_params[constants.DRBD_RESYNC_RATE],
8416 constants.LDP_BARRIERS: dt_params[constants.DRBD_DISK_BARRIERS],
8417 constants.LDP_NO_META_FLUSH: dt_params[constants.DRBD_META_BARRIERS],
8418 constants.LDP_DEFAULT_METAVG: dt_params[constants.DRBD_DEFAULT_METAVG],
8419 constants.LDP_DISK_CUSTOM: dt_params[constants.DRBD_DISK_CUSTOM],
8420 constants.LDP_NET_CUSTOM: dt_params[constants.DRBD_NET_CUSTOM],
8421 constants.LDP_DYNAMIC_RESYNC: dt_params[constants.DRBD_DYNAMIC_RESYNC],
8422 constants.LDP_PLAN_AHEAD: dt_params[constants.DRBD_PLAN_AHEAD],
8423 constants.LDP_FILL_TARGET: dt_params[constants.DRBD_FILL_TARGET],
8424 constants.LDP_DELAY_TARGET: dt_params[constants.DRBD_DELAY_TARGET],
8425 constants.LDP_MAX_RATE: dt_params[constants.DRBD_MAX_RATE],
8426 constants.LDP_MIN_RATE: dt_params[constants.DRBD_MIN_RATE],
8430 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_DRBD8],
8433 result.append(drbd_params)
8437 constants.LDP_STRIPES: dt_params[constants.DRBD_DATA_STRIPES],
8440 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_LV],
8442 result.append(data_params)
8446 constants.LDP_STRIPES: dt_params[constants.DRBD_META_STRIPES],
8449 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_LV],
8451 result.append(meta_params)
8453 elif (disk_template == constants.DT_FILE or
8454 disk_template == constants.DT_SHARED_FILE):
8455 result.append(constants.DISK_LD_DEFAULTS[constants.LD_FILE])
8457 elif disk_template == constants.DT_PLAIN:
8459 constants.LDP_STRIPES: dt_params[constants.LV_STRIPES],
8462 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_LV],
8464 result.append(params)
8466 elif disk_template == constants.DT_BLOCK:
8467 result.append(constants.DISK_LD_DEFAULTS[constants.LD_BLOCKDEV])
8472 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
8473 iv_name, p_minor, s_minor, drbd_params, data_params,
8475 """Generate a drbd8 device complete with its children.
8478 assert len(vgnames) == len(names) == 2
8479 port = lu.cfg.AllocatePort()
8480 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
8482 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
8483 logical_id=(vgnames[0], names[0]),
8485 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,
8486 logical_id=(vgnames[1], names[1]),
8488 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
8489 logical_id=(primary, secondary, port,
8492 children=[dev_data, dev_meta],
8493 iv_name=iv_name, params=drbd_params)
8497 def _GenerateDiskTemplate(lu, template_name,
8498 instance_name, primary_node,
8499 secondary_nodes, disk_info,
8500 file_storage_dir, file_driver,
8501 base_index, feedback_fn, disk_params):
8502 """Generate the entire disk layout for a given template type.
8505 #TODO: compute space requirements
8507 vgname = lu.cfg.GetVGName()
8508 disk_count = len(disk_info)
8510 ld_params = _ComputeLDParams(template_name, disk_params)
8511 if template_name == constants.DT_DISKLESS:
8513 elif template_name == constants.DT_PLAIN:
8515 raise errors.ProgrammerError("Wrong template configuration")
8517 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
8518 for i in range(disk_count)])
8519 for idx, disk in enumerate(disk_info):
8520 disk_index = idx + base_index
8521 vg = disk.get(constants.IDISK_VG, vgname)
8522 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
8523 disk_dev = objects.Disk(dev_type=constants.LD_LV,
8524 size=disk[constants.IDISK_SIZE],
8525 logical_id=(vg, names[idx]),
8526 iv_name="disk/%d" % disk_index,
8527 mode=disk[constants.IDISK_MODE],
8528 params=ld_params[0])
8529 disks.append(disk_dev)
8530 elif template_name == constants.DT_DRBD8:
8531 drbd_params, data_params, meta_params = ld_params
8532 if len(secondary_nodes) != 1:
8533 raise errors.ProgrammerError("Wrong template configuration")
8534 remote_node = secondary_nodes[0]
8535 minors = lu.cfg.AllocateDRBDMinor(
8536 [primary_node, remote_node] * len(disk_info), instance_name)
8539 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
8540 for i in range(disk_count)]):
8541 names.append(lv_prefix + "_data")
8542 names.append(lv_prefix + "_meta")
8543 for idx, disk in enumerate(disk_info):
8544 disk_index = idx + base_index
8545 drbd_default_metavg = drbd_params[constants.LDP_DEFAULT_METAVG]
8546 data_vg = disk.get(constants.IDISK_VG, vgname)
8547 meta_vg = disk.get(constants.IDISK_METAVG, drbd_default_metavg)
8548 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
8549 disk[constants.IDISK_SIZE],
8551 names[idx * 2:idx * 2 + 2],
8552 "disk/%d" % disk_index,
8553 minors[idx * 2], minors[idx * 2 + 1],
8554 drbd_params, data_params, meta_params)
8555 disk_dev.mode = disk[constants.IDISK_MODE]
8556 disks.append(disk_dev)
8557 elif template_name == constants.DT_FILE:
8559 raise errors.ProgrammerError("Wrong template configuration")
8561 opcodes.RequireFileStorage()
8563 for idx, disk in enumerate(disk_info):
8564 disk_index = idx + base_index
8565 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
8566 size=disk[constants.IDISK_SIZE],
8567 iv_name="disk/%d" % disk_index,
8568 logical_id=(file_driver,
8569 "%s/disk%d" % (file_storage_dir,
8571 mode=disk[constants.IDISK_MODE],
8572 params=ld_params[0])
8573 disks.append(disk_dev)
8574 elif template_name == constants.DT_SHARED_FILE:
8576 raise errors.ProgrammerError("Wrong template configuration")
8578 opcodes.RequireSharedFileStorage()
8580 for idx, disk in enumerate(disk_info):
8581 disk_index = idx + base_index
8582 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
8583 size=disk[constants.IDISK_SIZE],
8584 iv_name="disk/%d" % disk_index,
8585 logical_id=(file_driver,
8586 "%s/disk%d" % (file_storage_dir,
8588 mode=disk[constants.IDISK_MODE],
8589 params=ld_params[0])
8590 disks.append(disk_dev)
8591 elif template_name == constants.DT_BLOCK:
8593 raise errors.ProgrammerError("Wrong template configuration")
8595 for idx, disk in enumerate(disk_info):
8596 disk_index = idx + base_index
8597 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV,
8598 size=disk[constants.IDISK_SIZE],
8599 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL,
8600 disk[constants.IDISK_ADOPT]),
8601 iv_name="disk/%d" % disk_index,
8602 mode=disk[constants.IDISK_MODE],
8603 params=ld_params[0])
8604 disks.append(disk_dev)
8607 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
8611 def _GetInstanceInfoText(instance):
8612 """Compute that text that should be added to the disk's metadata.
8615 return "originstname+%s" % instance.name
8618 def _CalcEta(time_taken, written, total_size):
8619 """Calculates the ETA based on size written and total size.
8621 @param time_taken: The time taken so far
8622 @param written: amount written so far
8623 @param total_size: The total size of data to be written
8624 @return: The remaining time in seconds
8627 avg_time = time_taken / float(written)
8628 return (total_size - written) * avg_time
8631 def _WipeDisks(lu, instance):
8632 """Wipes instance disks.
8634 @type lu: L{LogicalUnit}
8635 @param lu: the logical unit on whose behalf we execute
8636 @type instance: L{objects.Instance}
8637 @param instance: the instance whose disks we should create
8638 @return: the success of the wipe
8641 node = instance.primary_node
8643 for device in instance.disks:
8644 lu.cfg.SetDiskID(device, node)
8646 logging.info("Pause sync of instance %s disks", instance.name)
8647 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
8649 for idx, success in enumerate(result.payload):
8651 logging.warn("pause-sync of instance %s for disks %d failed",
8655 for idx, device in enumerate(instance.disks):
8656 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
8657 # MAX_WIPE_CHUNK at max
8658 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
8659 constants.MIN_WIPE_CHUNK_PERCENT)
8660 # we _must_ make this an int, otherwise rounding errors will
8662 wipe_chunk_size = int(wipe_chunk_size)
8664 lu.LogInfo("* Wiping disk %d", idx)
8665 logging.info("Wiping disk %d for instance %s, node %s using"
8666 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
8671 start_time = time.time()
8673 while offset < size:
8674 wipe_size = min(wipe_chunk_size, size - offset)
8675 logging.debug("Wiping disk %d, offset %s, chunk %s",
8676 idx, offset, wipe_size)
8677 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
8678 result.Raise("Could not wipe disk %d at offset %d for size %d" %
8679 (idx, offset, wipe_size))
8682 if now - last_output >= 60:
8683 eta = _CalcEta(now - start_time, offset, size)
8684 lu.LogInfo(" - done: %.1f%% ETA: %s" %
8685 (offset / float(size) * 100, utils.FormatSeconds(eta)))
8688 logging.info("Resume sync of instance %s disks", instance.name)
8690 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
8692 for idx, success in enumerate(result.payload):
8694 lu.LogWarning("Resume sync of disk %d failed, please have a"
8695 " look at the status and troubleshoot the issue", idx)
8696 logging.warn("resume-sync of instance %s for disks %d failed",
8700 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
8701 """Create all disks for an instance.
8703 This abstracts away some work from AddInstance.
8705 @type lu: L{LogicalUnit}
8706 @param lu: the logical unit on whose behalf we execute
8707 @type instance: L{objects.Instance}
8708 @param instance: the instance whose disks we should create
8710 @param to_skip: list of indices to skip
8711 @type target_node: string
8712 @param target_node: if passed, overrides the target node for creation
8714 @return: the success of the creation
8717 info = _GetInstanceInfoText(instance)
8718 if target_node is None:
8719 pnode = instance.primary_node
8720 all_nodes = instance.all_nodes
8725 if instance.disk_template in constants.DTS_FILEBASED:
8726 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
8727 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
8729 result.Raise("Failed to create directory '%s' on"
8730 " node %s" % (file_storage_dir, pnode))
8732 # Note: this needs to be kept in sync with adding of disks in
8733 # LUInstanceSetParams
8734 for idx, device in enumerate(instance.disks):
8735 if to_skip and idx in to_skip:
8737 logging.info("Creating volume %s for instance %s",
8738 device.iv_name, instance.name)
8740 for node in all_nodes:
8741 f_create = node == pnode
8742 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
8745 def _RemoveDisks(lu, instance, target_node=None):
8746 """Remove all disks for an instance.
8748 This abstracts away some work from `AddInstance()` and
8749 `RemoveInstance()`. Note that in case some of the devices couldn't
8750 be removed, the removal will continue with the other ones (compare
8751 with `_CreateDisks()`).
8753 @type lu: L{LogicalUnit}
8754 @param lu: the logical unit on whose behalf we execute
8755 @type instance: L{objects.Instance}
8756 @param instance: the instance whose disks we should remove
8757 @type target_node: string
8758 @param target_node: used to override the node on which to remove the disks
8760 @return: the success of the removal
8763 logging.info("Removing block devices for instance %s", instance.name)
8766 for device in instance.disks:
8768 edata = [(target_node, device)]
8770 edata = device.ComputeNodeTree(instance.primary_node)
8771 for node, disk in edata:
8772 lu.cfg.SetDiskID(disk, node)
8773 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
8775 lu.LogWarning("Could not remove block device %s on node %s,"
8776 " continuing anyway: %s", device.iv_name, node, msg)
8779 # if this is a DRBD disk, return its port to the pool
8780 if device.dev_type in constants.LDS_DRBD:
8781 tcp_port = device.logical_id[2]
8782 lu.cfg.AddTcpUdpPort(tcp_port)
8784 if instance.disk_template == constants.DT_FILE:
8785 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
8789 tgt = instance.primary_node
8790 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
8792 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
8793 file_storage_dir, instance.primary_node, result.fail_msg)
8799 def _ComputeDiskSizePerVG(disk_template, disks):
8800 """Compute disk size requirements in the volume group
8803 def _compute(disks, payload):
8804 """Universal algorithm.
8809 vgs[disk[constants.IDISK_VG]] = \
8810 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
8814 # Required free disk space as a function of disk and swap space
8816 constants.DT_DISKLESS: {},
8817 constants.DT_PLAIN: _compute(disks, 0),
8818 # 128 MB are added for drbd metadata for each disk
8819 constants.DT_DRBD8: _compute(disks, DRBD_META_SIZE),
8820 constants.DT_FILE: {},
8821 constants.DT_SHARED_FILE: {},
8824 if disk_template not in req_size_dict:
8825 raise errors.ProgrammerError("Disk template '%s' size requirement"
8826 " is unknown" % disk_template)
8828 return req_size_dict[disk_template]
8831 def _ComputeDiskSize(disk_template, disks):
8832 """Compute disk size requirements in the volume group
8835 # Required free disk space as a function of disk and swap space
8837 constants.DT_DISKLESS: None,
8838 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
8839 # 128 MB are added for drbd metadata for each disk
8841 sum(d[constants.IDISK_SIZE] + DRBD_META_SIZE for d in disks),
8842 constants.DT_FILE: None,
8843 constants.DT_SHARED_FILE: 0,
8844 constants.DT_BLOCK: 0,
8847 if disk_template not in req_size_dict:
8848 raise errors.ProgrammerError("Disk template '%s' size requirement"
8849 " is unknown" % disk_template)
8851 return req_size_dict[disk_template]
8854 def _FilterVmNodes(lu, nodenames):
8855 """Filters out non-vm_capable nodes from a list.
8857 @type lu: L{LogicalUnit}
8858 @param lu: the logical unit for which we check
8859 @type nodenames: list
8860 @param nodenames: the list of nodes on which we should check
8862 @return: the list of vm-capable nodes
8865 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
8866 return [name for name in nodenames if name not in vm_nodes]
8869 def _CheckHVParams(lu, nodenames, hvname, hvparams):
8870 """Hypervisor parameter validation.
8872 This function abstract the hypervisor parameter validation to be
8873 used in both instance create and instance modify.
8875 @type lu: L{LogicalUnit}
8876 @param lu: the logical unit for which we check
8877 @type nodenames: list
8878 @param nodenames: the list of nodes on which we should check
8879 @type hvname: string
8880 @param hvname: the name of the hypervisor we should use
8881 @type hvparams: dict
8882 @param hvparams: the parameters which we need to check
8883 @raise errors.OpPrereqError: if the parameters are not valid
8886 nodenames = _FilterVmNodes(lu, nodenames)
8888 cluster = lu.cfg.GetClusterInfo()
8889 hvfull = objects.FillDict(cluster.hvparams.get(hvname, {}), hvparams)
8891 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames, hvname, hvfull)
8892 for node in nodenames:
8896 info.Raise("Hypervisor parameter validation failed on node %s" % node)
8899 def _CheckOSParams(lu, required, nodenames, osname, osparams):
8900 """OS parameters validation.
8902 @type lu: L{LogicalUnit}
8903 @param lu: the logical unit for which we check
8904 @type required: boolean
8905 @param required: whether the validation should fail if the OS is not
8907 @type nodenames: list
8908 @param nodenames: the list of nodes on which we should check
8909 @type osname: string
8910 @param osname: the name of the hypervisor we should use
8911 @type osparams: dict
8912 @param osparams: the parameters which we need to check
8913 @raise errors.OpPrereqError: if the parameters are not valid
8916 nodenames = _FilterVmNodes(lu, nodenames)
8917 result = lu.rpc.call_os_validate(nodenames, required, osname,
8918 [constants.OS_VALIDATE_PARAMETERS],
8920 for node, nres in result.items():
8921 # we don't check for offline cases since this should be run only
8922 # against the master node and/or an instance's nodes
8923 nres.Raise("OS Parameters validation failed on node %s" % node)
8924 if not nres.payload:
8925 lu.LogInfo("OS %s not found on node %s, validation skipped",
8929 class LUInstanceCreate(LogicalUnit):
8930 """Create an instance.
8933 HPATH = "instance-add"
8934 HTYPE = constants.HTYPE_INSTANCE
8937 def CheckArguments(self):
8941 # do not require name_check to ease forward/backward compatibility
8943 if self.op.no_install and self.op.start:
8944 self.LogInfo("No-installation mode selected, disabling startup")
8945 self.op.start = False
8946 # validate/normalize the instance name
8947 self.op.instance_name = \
8948 netutils.Hostname.GetNormalizedName(self.op.instance_name)
8950 if self.op.ip_check and not self.op.name_check:
8951 # TODO: make the ip check more flexible and not depend on the name check
8952 raise errors.OpPrereqError("Cannot do IP address check without a name"
8953 " check", errors.ECODE_INVAL)
8955 # check nics' parameter names
8956 for nic in self.op.nics:
8957 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
8959 # check disks. parameter names and consistent adopt/no-adopt strategy
8960 has_adopt = has_no_adopt = False
8961 for disk in self.op.disks:
8962 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
8963 if constants.IDISK_ADOPT in disk:
8967 if has_adopt and has_no_adopt:
8968 raise errors.OpPrereqError("Either all disks are adopted or none is",
8971 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
8972 raise errors.OpPrereqError("Disk adoption is not supported for the"
8973 " '%s' disk template" %
8974 self.op.disk_template,
8976 if self.op.iallocator is not None:
8977 raise errors.OpPrereqError("Disk adoption not allowed with an"
8978 " iallocator script", errors.ECODE_INVAL)
8979 if self.op.mode == constants.INSTANCE_IMPORT:
8980 raise errors.OpPrereqError("Disk adoption not allowed for"
8981 " instance import", errors.ECODE_INVAL)
8983 if self.op.disk_template in constants.DTS_MUST_ADOPT:
8984 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
8985 " but no 'adopt' parameter given" %
8986 self.op.disk_template,
8989 self.adopt_disks = has_adopt
8991 # instance name verification
8992 if self.op.name_check:
8993 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
8994 self.op.instance_name = self.hostname1.name
8995 # used in CheckPrereq for ip ping check
8996 self.check_ip = self.hostname1.ip
8998 self.check_ip = None
9000 # file storage checks
9001 if (self.op.file_driver and
9002 not self.op.file_driver in constants.FILE_DRIVER):
9003 raise errors.OpPrereqError("Invalid file driver name '%s'" %
9004 self.op.file_driver, errors.ECODE_INVAL)
9006 if self.op.disk_template == constants.DT_FILE:
9007 opcodes.RequireFileStorage()
9008 elif self.op.disk_template == constants.DT_SHARED_FILE:
9009 opcodes.RequireSharedFileStorage()
9011 ### Node/iallocator related checks
9012 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
9014 if self.op.pnode is not None:
9015 if self.op.disk_template in constants.DTS_INT_MIRROR:
9016 if self.op.snode is None:
9017 raise errors.OpPrereqError("The networked disk templates need"
9018 " a mirror node", errors.ECODE_INVAL)
9020 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
9022 self.op.snode = None
9024 self._cds = _GetClusterDomainSecret()
9026 if self.op.mode == constants.INSTANCE_IMPORT:
9027 # On import force_variant must be True, because if we forced it at
9028 # initial install, our only chance when importing it back is that it
9030 self.op.force_variant = True
9032 if self.op.no_install:
9033 self.LogInfo("No-installation mode has no effect during import")
9035 elif self.op.mode == constants.INSTANCE_CREATE:
9036 if self.op.os_type is None:
9037 raise errors.OpPrereqError("No guest OS specified",
9039 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
9040 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
9041 " installation" % self.op.os_type,
9043 if self.op.disk_template is None:
9044 raise errors.OpPrereqError("No disk template specified",
9047 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
9048 # Check handshake to ensure both clusters have the same domain secret
9049 src_handshake = self.op.source_handshake
9050 if not src_handshake:
9051 raise errors.OpPrereqError("Missing source handshake",
9054 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
9057 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
9060 # Load and check source CA
9061 self.source_x509_ca_pem = self.op.source_x509_ca
9062 if not self.source_x509_ca_pem:
9063 raise errors.OpPrereqError("Missing source X509 CA",
9067 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
9069 except OpenSSL.crypto.Error, err:
9070 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
9071 (err, ), errors.ECODE_INVAL)
9073 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9074 if errcode is not None:
9075 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
9078 self.source_x509_ca = cert
9080 src_instance_name = self.op.source_instance_name
9081 if not src_instance_name:
9082 raise errors.OpPrereqError("Missing source instance name",
9085 self.source_instance_name = \
9086 netutils.GetHostname(name=src_instance_name).name
9089 raise errors.OpPrereqError("Invalid instance creation mode %r" %
9090 self.op.mode, errors.ECODE_INVAL)
9092 def ExpandNames(self):
9093 """ExpandNames for CreateInstance.
9095 Figure out the right locks for instance creation.
9098 self.needed_locks = {}
9100 instance_name = self.op.instance_name
9101 # this is just a preventive check, but someone might still add this
9102 # instance in the meantime, and creation will fail at lock-add time
9103 if instance_name in self.cfg.GetInstanceList():
9104 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
9105 instance_name, errors.ECODE_EXISTS)
9107 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
9109 if self.op.iallocator:
9110 # TODO: Find a solution to not lock all nodes in the cluster, e.g. by
9111 # specifying a group on instance creation and then selecting nodes from
9113 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9114 self.needed_locks[locking.LEVEL_NODE_RES] = locking.ALL_SET
9116 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
9117 nodelist = [self.op.pnode]
9118 if self.op.snode is not None:
9119 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
9120 nodelist.append(self.op.snode)
9121 self.needed_locks[locking.LEVEL_NODE] = nodelist
9122 # Lock resources of instance's primary and secondary nodes (copy to
9123 # prevent accidential modification)
9124 self.needed_locks[locking.LEVEL_NODE_RES] = list(nodelist)
9126 # in case of import lock the source node too
9127 if self.op.mode == constants.INSTANCE_IMPORT:
9128 src_node = self.op.src_node
9129 src_path = self.op.src_path
9131 if src_path is None:
9132 self.op.src_path = src_path = self.op.instance_name
9134 if src_node is None:
9135 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9136 self.op.src_node = None
9137 if os.path.isabs(src_path):
9138 raise errors.OpPrereqError("Importing an instance from a path"
9139 " requires a source node option",
9142 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
9143 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
9144 self.needed_locks[locking.LEVEL_NODE].append(src_node)
9145 if not os.path.isabs(src_path):
9146 self.op.src_path = src_path = \
9147 utils.PathJoin(constants.EXPORT_DIR, src_path)
9149 def _RunAllocator(self):
9150 """Run the allocator based on input opcode.
9153 nics = [n.ToDict() for n in self.nics]
9154 ial = IAllocator(self.cfg, self.rpc,
9155 mode=constants.IALLOCATOR_MODE_ALLOC,
9156 name=self.op.instance_name,
9157 disk_template=self.op.disk_template,
9160 vcpus=self.be_full[constants.BE_VCPUS],
9161 memory=self.be_full[constants.BE_MAXMEM],
9164 hypervisor=self.op.hypervisor,
9167 ial.Run(self.op.iallocator)
9170 raise errors.OpPrereqError("Can't compute nodes using"
9171 " iallocator '%s': %s" %
9172 (self.op.iallocator, ial.info),
9174 if len(ial.result) != ial.required_nodes:
9175 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
9176 " of nodes (%s), required %s" %
9177 (self.op.iallocator, len(ial.result),
9178 ial.required_nodes), errors.ECODE_FAULT)
9179 self.op.pnode = ial.result[0]
9180 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
9181 self.op.instance_name, self.op.iallocator,
9182 utils.CommaJoin(ial.result))
9183 if ial.required_nodes == 2:
9184 self.op.snode = ial.result[1]
9186 def BuildHooksEnv(self):
9189 This runs on master, primary and secondary nodes of the instance.
9193 "ADD_MODE": self.op.mode,
9195 if self.op.mode == constants.INSTANCE_IMPORT:
9196 env["SRC_NODE"] = self.op.src_node
9197 env["SRC_PATH"] = self.op.src_path
9198 env["SRC_IMAGES"] = self.src_images
9200 env.update(_BuildInstanceHookEnv(
9201 name=self.op.instance_name,
9202 primary_node=self.op.pnode,
9203 secondary_nodes=self.secondaries,
9204 status=self.op.start,
9205 os_type=self.op.os_type,
9206 minmem=self.be_full[constants.BE_MINMEM],
9207 maxmem=self.be_full[constants.BE_MAXMEM],
9208 vcpus=self.be_full[constants.BE_VCPUS],
9209 nics=_NICListToTuple(self, self.nics),
9210 disk_template=self.op.disk_template,
9211 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
9212 for d in self.disks],
9215 hypervisor_name=self.op.hypervisor,
9221 def BuildHooksNodes(self):
9222 """Build hooks nodes.
9225 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
9228 def _ReadExportInfo(self):
9229 """Reads the export information from disk.
9231 It will override the opcode source node and path with the actual
9232 information, if these two were not specified before.
9234 @return: the export information
9237 assert self.op.mode == constants.INSTANCE_IMPORT
9239 src_node = self.op.src_node
9240 src_path = self.op.src_path
9242 if src_node is None:
9243 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
9244 exp_list = self.rpc.call_export_list(locked_nodes)
9246 for node in exp_list:
9247 if exp_list[node].fail_msg:
9249 if src_path in exp_list[node].payload:
9251 self.op.src_node = src_node = node
9252 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
9256 raise errors.OpPrereqError("No export found for relative path %s" %
9257 src_path, errors.ECODE_INVAL)
9259 _CheckNodeOnline(self, src_node)
9260 result = self.rpc.call_export_info(src_node, src_path)
9261 result.Raise("No export or invalid export found in dir %s" % src_path)
9263 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
9264 if not export_info.has_section(constants.INISECT_EXP):
9265 raise errors.ProgrammerError("Corrupted export config",
9266 errors.ECODE_ENVIRON)
9268 ei_version = export_info.get(constants.INISECT_EXP, "version")
9269 if (int(ei_version) != constants.EXPORT_VERSION):
9270 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
9271 (ei_version, constants.EXPORT_VERSION),
9272 errors.ECODE_ENVIRON)
9275 def _ReadExportParams(self, einfo):
9276 """Use export parameters as defaults.
9278 In case the opcode doesn't specify (as in override) some instance
9279 parameters, then try to use them from the export information, if
9283 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
9285 if self.op.disk_template is None:
9286 if einfo.has_option(constants.INISECT_INS, "disk_template"):
9287 self.op.disk_template = einfo.get(constants.INISECT_INS,
9289 if self.op.disk_template not in constants.DISK_TEMPLATES:
9290 raise errors.OpPrereqError("Disk template specified in configuration"
9291 " file is not one of the allowed values:"
9292 " %s" % " ".join(constants.DISK_TEMPLATES))
9294 raise errors.OpPrereqError("No disk template specified and the export"
9295 " is missing the disk_template information",
9298 if not self.op.disks:
9300 # TODO: import the disk iv_name too
9301 for idx in range(constants.MAX_DISKS):
9302 if einfo.has_option(constants.INISECT_INS, "disk%d_size" % idx):
9303 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
9304 disks.append({constants.IDISK_SIZE: disk_sz})
9305 self.op.disks = disks
9306 if not disks and self.op.disk_template != constants.DT_DISKLESS:
9307 raise errors.OpPrereqError("No disk info specified and the export"
9308 " is missing the disk information",
9311 if not self.op.nics:
9313 for idx in range(constants.MAX_NICS):
9314 if einfo.has_option(constants.INISECT_INS, "nic%d_mac" % idx):
9316 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
9317 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
9324 if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):
9325 self.op.tags = einfo.get(constants.INISECT_INS, "tags").split()
9327 if (self.op.hypervisor is None and
9328 einfo.has_option(constants.INISECT_INS, "hypervisor")):
9329 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
9331 if einfo.has_section(constants.INISECT_HYP):
9332 # use the export parameters but do not override the ones
9333 # specified by the user
9334 for name, value in einfo.items(constants.INISECT_HYP):
9335 if name not in self.op.hvparams:
9336 self.op.hvparams[name] = value
9338 if einfo.has_section(constants.INISECT_BEP):
9339 # use the parameters, without overriding
9340 for name, value in einfo.items(constants.INISECT_BEP):
9341 if name not in self.op.beparams:
9342 self.op.beparams[name] = value
9343 # Compatibility for the old "memory" be param
9344 if name == constants.BE_MEMORY:
9345 if constants.BE_MAXMEM not in self.op.beparams:
9346 self.op.beparams[constants.BE_MAXMEM] = value
9347 if constants.BE_MINMEM not in self.op.beparams:
9348 self.op.beparams[constants.BE_MINMEM] = value
9350 # try to read the parameters old style, from the main section
9351 for name in constants.BES_PARAMETERS:
9352 if (name not in self.op.beparams and
9353 einfo.has_option(constants.INISECT_INS, name)):
9354 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
9356 if einfo.has_section(constants.INISECT_OSP):
9357 # use the parameters, without overriding
9358 for name, value in einfo.items(constants.INISECT_OSP):
9359 if name not in self.op.osparams:
9360 self.op.osparams[name] = value
9362 def _RevertToDefaults(self, cluster):
9363 """Revert the instance parameters to the default values.
9367 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
9368 for name in self.op.hvparams.keys():
9369 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
9370 del self.op.hvparams[name]
9372 be_defs = cluster.SimpleFillBE({})
9373 for name in self.op.beparams.keys():
9374 if name in be_defs and be_defs[name] == self.op.beparams[name]:
9375 del self.op.beparams[name]
9377 nic_defs = cluster.SimpleFillNIC({})
9378 for nic in self.op.nics:
9379 for name in constants.NICS_PARAMETERS:
9380 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
9383 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
9384 for name in self.op.osparams.keys():
9385 if name in os_defs and os_defs[name] == self.op.osparams[name]:
9386 del self.op.osparams[name]
9388 def _CalculateFileStorageDir(self):
9389 """Calculate final instance file storage dir.
9392 # file storage dir calculation/check
9393 self.instance_file_storage_dir = None
9394 if self.op.disk_template in constants.DTS_FILEBASED:
9395 # build the full file storage dir path
9398 if self.op.disk_template == constants.DT_SHARED_FILE:
9399 get_fsd_fn = self.cfg.GetSharedFileStorageDir
9401 get_fsd_fn = self.cfg.GetFileStorageDir
9403 cfg_storagedir = get_fsd_fn()
9404 if not cfg_storagedir:
9405 raise errors.OpPrereqError("Cluster file storage dir not defined")
9406 joinargs.append(cfg_storagedir)
9408 if self.op.file_storage_dir is not None:
9409 joinargs.append(self.op.file_storage_dir)
9411 joinargs.append(self.op.instance_name)
9413 # pylint: disable=W0142
9414 self.instance_file_storage_dir = utils.PathJoin(*joinargs)
9416 def CheckPrereq(self): # pylint: disable=R0914
9417 """Check prerequisites.
9420 self._CalculateFileStorageDir()
9422 if self.op.mode == constants.INSTANCE_IMPORT:
9423 export_info = self._ReadExportInfo()
9424 self._ReadExportParams(export_info)
9426 if (not self.cfg.GetVGName() and
9427 self.op.disk_template not in constants.DTS_NOT_LVM):
9428 raise errors.OpPrereqError("Cluster does not support lvm-based"
9429 " instances", errors.ECODE_STATE)
9431 if (self.op.hypervisor is None or
9432 self.op.hypervisor == constants.VALUE_AUTO):
9433 self.op.hypervisor = self.cfg.GetHypervisorType()
9435 cluster = self.cfg.GetClusterInfo()
9436 enabled_hvs = cluster.enabled_hypervisors
9437 if self.op.hypervisor not in enabled_hvs:
9438 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
9439 " cluster (%s)" % (self.op.hypervisor,
9440 ",".join(enabled_hvs)),
9443 # Check tag validity
9444 for tag in self.op.tags:
9445 objects.TaggableObject.ValidateTag(tag)
9447 # check hypervisor parameter syntax (locally)
9448 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
9449 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
9451 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
9452 hv_type.CheckParameterSyntax(filled_hvp)
9453 self.hv_full = filled_hvp
9454 # check that we don't specify global parameters on an instance
9455 _CheckGlobalHvParams(self.op.hvparams)
9457 # fill and remember the beparams dict
9458 default_beparams = cluster.beparams[constants.PP_DEFAULT]
9459 for param, value in self.op.beparams.iteritems():
9460 if value == constants.VALUE_AUTO:
9461 self.op.beparams[param] = default_beparams[param]
9462 objects.UpgradeBeParams(self.op.beparams)
9463 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
9464 self.be_full = cluster.SimpleFillBE(self.op.beparams)
9466 # build os parameters
9467 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
9469 # now that hvp/bep are in final format, let's reset to defaults,
9471 if self.op.identify_defaults:
9472 self._RevertToDefaults(cluster)
9476 for idx, nic in enumerate(self.op.nics):
9477 nic_mode_req = nic.get(constants.INIC_MODE, None)
9478 nic_mode = nic_mode_req
9479 if nic_mode is None or nic_mode == constants.VALUE_AUTO:
9480 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
9482 # in routed mode, for the first nic, the default ip is 'auto'
9483 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
9484 default_ip_mode = constants.VALUE_AUTO
9486 default_ip_mode = constants.VALUE_NONE
9488 # ip validity checks
9489 ip = nic.get(constants.INIC_IP, default_ip_mode)
9490 if ip is None or ip.lower() == constants.VALUE_NONE:
9492 elif ip.lower() == constants.VALUE_AUTO:
9493 if not self.op.name_check:
9494 raise errors.OpPrereqError("IP address set to auto but name checks"
9495 " have been skipped",
9497 nic_ip = self.hostname1.ip
9499 if not netutils.IPAddress.IsValid(ip):
9500 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
9504 # TODO: check the ip address for uniqueness
9505 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
9506 raise errors.OpPrereqError("Routed nic mode requires an ip address",
9509 # MAC address verification
9510 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
9511 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9512 mac = utils.NormalizeAndValidateMac(mac)
9515 self.cfg.ReserveMAC(mac, self.proc.GetECId())
9516 except errors.ReservationError:
9517 raise errors.OpPrereqError("MAC address %s already in use"
9518 " in cluster" % mac,
9519 errors.ECODE_NOTUNIQUE)
9521 # Build nic parameters
9522 link = nic.get(constants.INIC_LINK, None)
9523 if link == constants.VALUE_AUTO:
9524 link = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_LINK]
9527 nicparams[constants.NIC_MODE] = nic_mode
9529 nicparams[constants.NIC_LINK] = link
9531 check_params = cluster.SimpleFillNIC(nicparams)
9532 objects.NIC.CheckParameterSyntax(check_params)
9533 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
9535 # disk checks/pre-build
9536 default_vg = self.cfg.GetVGName()
9538 for disk in self.op.disks:
9539 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
9540 if mode not in constants.DISK_ACCESS_SET:
9541 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
9542 mode, errors.ECODE_INVAL)
9543 size = disk.get(constants.IDISK_SIZE, None)
9545 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
9548 except (TypeError, ValueError):
9549 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
9552 data_vg = disk.get(constants.IDISK_VG, default_vg)
9554 constants.IDISK_SIZE: size,
9555 constants.IDISK_MODE: mode,
9556 constants.IDISK_VG: data_vg,
9558 if constants.IDISK_METAVG in disk:
9559 new_disk[constants.IDISK_METAVG] = disk[constants.IDISK_METAVG]
9560 if constants.IDISK_ADOPT in disk:
9561 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
9562 self.disks.append(new_disk)
9564 if self.op.mode == constants.INSTANCE_IMPORT:
9566 for idx in range(len(self.disks)):
9567 option = "disk%d_dump" % idx
9568 if export_info.has_option(constants.INISECT_INS, option):
9569 # FIXME: are the old os-es, disk sizes, etc. useful?
9570 export_name = export_info.get(constants.INISECT_INS, option)
9571 image = utils.PathJoin(self.op.src_path, export_name)
9572 disk_images.append(image)
9574 disk_images.append(False)
9576 self.src_images = disk_images
9578 old_name = export_info.get(constants.INISECT_INS, "name")
9579 if self.op.instance_name == old_name:
9580 for idx, nic in enumerate(self.nics):
9581 if nic.mac == constants.VALUE_AUTO:
9582 nic_mac_ini = "nic%d_mac" % idx
9583 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
9585 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
9587 # ip ping checks (we use the same ip that was resolved in ExpandNames)
9588 if self.op.ip_check:
9589 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
9590 raise errors.OpPrereqError("IP %s of instance %s already in use" %
9591 (self.check_ip, self.op.instance_name),
9592 errors.ECODE_NOTUNIQUE)
9594 #### mac address generation
9595 # By generating here the mac address both the allocator and the hooks get
9596 # the real final mac address rather than the 'auto' or 'generate' value.
9597 # There is a race condition between the generation and the instance object
9598 # creation, which means that we know the mac is valid now, but we're not
9599 # sure it will be when we actually add the instance. If things go bad
9600 # adding the instance will abort because of a duplicate mac, and the
9601 # creation job will fail.
9602 for nic in self.nics:
9603 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9604 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
9608 if self.op.iallocator is not None:
9609 self._RunAllocator()
9611 # Release all unneeded node locks
9612 _ReleaseLocks(self, locking.LEVEL_NODE,
9613 keep=filter(None, [self.op.pnode, self.op.snode,
9615 _ReleaseLocks(self, locking.LEVEL_NODE_RES,
9616 keep=filter(None, [self.op.pnode, self.op.snode,
9619 #### node related checks
9621 # check primary node
9622 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
9623 assert self.pnode is not None, \
9624 "Cannot retrieve locked node %s" % self.op.pnode
9626 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
9627 pnode.name, errors.ECODE_STATE)
9629 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
9630 pnode.name, errors.ECODE_STATE)
9631 if not pnode.vm_capable:
9632 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
9633 " '%s'" % pnode.name, errors.ECODE_STATE)
9635 self.secondaries = []
9637 # mirror node verification
9638 if self.op.disk_template in constants.DTS_INT_MIRROR:
9639 if self.op.snode == pnode.name:
9640 raise errors.OpPrereqError("The secondary node cannot be the"
9641 " primary node", errors.ECODE_INVAL)
9642 _CheckNodeOnline(self, self.op.snode)
9643 _CheckNodeNotDrained(self, self.op.snode)
9644 _CheckNodeVmCapable(self, self.op.snode)
9645 self.secondaries.append(self.op.snode)
9647 snode = self.cfg.GetNodeInfo(self.op.snode)
9648 if pnode.group != snode.group:
9649 self.LogWarning("The primary and secondary nodes are in two"
9650 " different node groups; the disk parameters"
9651 " from the first disk's node group will be"
9654 nodenames = [pnode.name] + self.secondaries
9656 # Verify instance specs
9658 constants.ISPEC_MEM_SIZE: self.be_full.get(constants.BE_MAXMEM, None),
9659 constants.ISPEC_CPU_COUNT: self.be_full.get(constants.BE_VCPUS, None),
9660 constants.ISPEC_DISK_COUNT: len(self.disks),
9661 constants.ISPEC_DISK_SIZE: [disk["size"] for disk in self.disks],
9662 constants.ISPEC_NIC_COUNT: len(self.nics),
9665 group_info = self.cfg.GetNodeGroup(pnode.group)
9666 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
9667 res = _ComputeIPolicyInstanceSpecViolation(ipolicy, ispec)
9668 if not self.op.ignore_ipolicy and res:
9669 raise errors.OpPrereqError(("Instance allocation to group %s violates"
9670 " policy: %s") % (pnode.group,
9671 utils.CommaJoin(res)),
9674 # disk parameters (not customizable at instance or node level)
9675 # just use the primary node parameters, ignoring the secondary.
9676 self.diskparams = group_info.diskparams
9678 if not self.adopt_disks:
9679 # Check lv size requirements, if not adopting
9680 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
9681 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
9683 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
9684 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
9685 disk[constants.IDISK_ADOPT])
9686 for disk in self.disks])
9687 if len(all_lvs) != len(self.disks):
9688 raise errors.OpPrereqError("Duplicate volume names given for adoption",
9690 for lv_name in all_lvs:
9692 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
9693 # to ReserveLV uses the same syntax
9694 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
9695 except errors.ReservationError:
9696 raise errors.OpPrereqError("LV named %s used by another instance" %
9697 lv_name, errors.ECODE_NOTUNIQUE)
9699 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
9700 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
9702 node_lvs = self.rpc.call_lv_list([pnode.name],
9703 vg_names.payload.keys())[pnode.name]
9704 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
9705 node_lvs = node_lvs.payload
9707 delta = all_lvs.difference(node_lvs.keys())
9709 raise errors.OpPrereqError("Missing logical volume(s): %s" %
9710 utils.CommaJoin(delta),
9712 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
9714 raise errors.OpPrereqError("Online logical volumes found, cannot"
9715 " adopt: %s" % utils.CommaJoin(online_lvs),
9717 # update the size of disk based on what is found
9718 for dsk in self.disks:
9719 dsk[constants.IDISK_SIZE] = \
9720 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
9721 dsk[constants.IDISK_ADOPT])][0]))
9723 elif self.op.disk_template == constants.DT_BLOCK:
9724 # Normalize and de-duplicate device paths
9725 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
9726 for disk in self.disks])
9727 if len(all_disks) != len(self.disks):
9728 raise errors.OpPrereqError("Duplicate disk names given for adoption",
9730 baddisks = [d for d in all_disks
9731 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
9733 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
9734 " cannot be adopted" %
9735 (", ".join(baddisks),
9736 constants.ADOPTABLE_BLOCKDEV_ROOT),
9739 node_disks = self.rpc.call_bdev_sizes([pnode.name],
9740 list(all_disks))[pnode.name]
9741 node_disks.Raise("Cannot get block device information from node %s" %
9743 node_disks = node_disks.payload
9744 delta = all_disks.difference(node_disks.keys())
9746 raise errors.OpPrereqError("Missing block device(s): %s" %
9747 utils.CommaJoin(delta),
9749 for dsk in self.disks:
9750 dsk[constants.IDISK_SIZE] = \
9751 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
9753 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
9755 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
9756 # check OS parameters (remotely)
9757 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
9759 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
9761 # memory check on primary node
9762 #TODO(dynmem): use MINMEM for checking
9764 _CheckNodeFreeMemory(self, self.pnode.name,
9765 "creating instance %s" % self.op.instance_name,
9766 self.be_full[constants.BE_MAXMEM],
9769 self.dry_run_result = list(nodenames)
9771 def Exec(self, feedback_fn):
9772 """Create and add the instance to the cluster.
9775 instance = self.op.instance_name
9776 pnode_name = self.pnode.name
9778 assert not (self.owned_locks(locking.LEVEL_NODE_RES) -
9779 self.owned_locks(locking.LEVEL_NODE)), \
9780 "Node locks differ from node resource locks"
9782 ht_kind = self.op.hypervisor
9783 if ht_kind in constants.HTS_REQ_PORT:
9784 network_port = self.cfg.AllocatePort()
9788 disks = _GenerateDiskTemplate(self,
9789 self.op.disk_template,
9790 instance, pnode_name,
9793 self.instance_file_storage_dir,
9794 self.op.file_driver,
9799 iobj = objects.Instance(name=instance, os=self.op.os_type,
9800 primary_node=pnode_name,
9801 nics=self.nics, disks=disks,
9802 disk_template=self.op.disk_template,
9803 admin_state=constants.ADMINST_DOWN,
9804 network_port=network_port,
9805 beparams=self.op.beparams,
9806 hvparams=self.op.hvparams,
9807 hypervisor=self.op.hypervisor,
9808 osparams=self.op.osparams,
9812 for tag in self.op.tags:
9815 if self.adopt_disks:
9816 if self.op.disk_template == constants.DT_PLAIN:
9817 # rename LVs to the newly-generated names; we need to construct
9818 # 'fake' LV disks with the old data, plus the new unique_id
9819 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
9821 for t_dsk, a_dsk in zip(tmp_disks, self.disks):
9822 rename_to.append(t_dsk.logical_id)
9823 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
9824 self.cfg.SetDiskID(t_dsk, pnode_name)
9825 result = self.rpc.call_blockdev_rename(pnode_name,
9826 zip(tmp_disks, rename_to))
9827 result.Raise("Failed to rename adoped LVs")
9829 feedback_fn("* creating instance disks...")
9831 _CreateDisks(self, iobj)
9832 except errors.OpExecError:
9833 self.LogWarning("Device creation failed, reverting...")
9835 _RemoveDisks(self, iobj)
9837 self.cfg.ReleaseDRBDMinors(instance)
9840 feedback_fn("adding instance %s to cluster config" % instance)
9842 self.cfg.AddInstance(iobj, self.proc.GetECId())
9844 # Declare that we don't want to remove the instance lock anymore, as we've
9845 # added the instance to the config
9846 del self.remove_locks[locking.LEVEL_INSTANCE]
9848 if self.op.mode == constants.INSTANCE_IMPORT:
9849 # Release unused nodes
9850 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
9853 _ReleaseLocks(self, locking.LEVEL_NODE)
9856 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
9857 feedback_fn("* wiping instance disks...")
9859 _WipeDisks(self, iobj)
9860 except errors.OpExecError, err:
9861 logging.exception("Wiping disks failed")
9862 self.LogWarning("Wiping instance disks failed (%s)", err)
9866 # Something is already wrong with the disks, don't do anything else
9868 elif self.op.wait_for_sync:
9869 disk_abort = not _WaitForSync(self, iobj)
9870 elif iobj.disk_template in constants.DTS_INT_MIRROR:
9871 # make sure the disks are not degraded (still sync-ing is ok)
9872 feedback_fn("* checking mirrors status")
9873 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
9878 _RemoveDisks(self, iobj)
9879 self.cfg.RemoveInstance(iobj.name)
9880 # Make sure the instance lock gets removed
9881 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
9882 raise errors.OpExecError("There are some degraded disks for"
9885 # Release all node resource locks
9886 _ReleaseLocks(self, locking.LEVEL_NODE_RES)
9888 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
9889 if self.op.mode == constants.INSTANCE_CREATE:
9890 if not self.op.no_install:
9891 pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and
9892 not self.op.wait_for_sync)
9894 feedback_fn("* pausing disk sync to install instance OS")
9895 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
9897 for idx, success in enumerate(result.payload):
9899 logging.warn("pause-sync of instance %s for disk %d failed",
9902 feedback_fn("* running the instance OS create scripts...")
9903 # FIXME: pass debug option from opcode to backend
9905 self.rpc.call_instance_os_add(pnode_name, (iobj, None), False,
9906 self.op.debug_level)
9908 feedback_fn("* resuming disk sync")
9909 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
9911 for idx, success in enumerate(result.payload):
9913 logging.warn("resume-sync of instance %s for disk %d failed",
9916 os_add_result.Raise("Could not add os for instance %s"
9917 " on node %s" % (instance, pnode_name))
9919 elif self.op.mode == constants.INSTANCE_IMPORT:
9920 feedback_fn("* running the instance OS import scripts...")
9924 for idx, image in enumerate(self.src_images):
9928 # FIXME: pass debug option from opcode to backend
9929 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
9930 constants.IEIO_FILE, (image, ),
9931 constants.IEIO_SCRIPT,
9932 (iobj.disks[idx], idx),
9934 transfers.append(dt)
9937 masterd.instance.TransferInstanceData(self, feedback_fn,
9938 self.op.src_node, pnode_name,
9939 self.pnode.secondary_ip,
9941 if not compat.all(import_result):
9942 self.LogWarning("Some disks for instance %s on node %s were not"
9943 " imported successfully" % (instance, pnode_name))
9945 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
9946 feedback_fn("* preparing remote import...")
9947 # The source cluster will stop the instance before attempting to make a
9948 # connection. In some cases stopping an instance can take a long time,
9949 # hence the shutdown timeout is added to the connection timeout.
9950 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
9951 self.op.source_shutdown_timeout)
9952 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9954 assert iobj.primary_node == self.pnode.name
9956 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
9957 self.source_x509_ca,
9958 self._cds, timeouts)
9959 if not compat.all(disk_results):
9960 # TODO: Should the instance still be started, even if some disks
9961 # failed to import (valid for local imports, too)?
9962 self.LogWarning("Some disks for instance %s on node %s were not"
9963 " imported successfully" % (instance, pnode_name))
9965 # Run rename script on newly imported instance
9966 assert iobj.name == instance
9967 feedback_fn("Running rename script for %s" % instance)
9968 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
9969 self.source_instance_name,
9970 self.op.debug_level)
9972 self.LogWarning("Failed to run rename script for %s on node"
9973 " %s: %s" % (instance, pnode_name, result.fail_msg))
9976 # also checked in the prereq part
9977 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
9980 assert not self.owned_locks(locking.LEVEL_NODE_RES)
9983 iobj.admin_state = constants.ADMINST_UP
9984 self.cfg.Update(iobj, feedback_fn)
9985 logging.info("Starting instance %s on node %s", instance, pnode_name)
9986 feedback_fn("* starting instance...")
9987 result = self.rpc.call_instance_start(pnode_name, (iobj, None, None),
9989 result.Raise("Could not start instance")
9991 return list(iobj.all_nodes)
9994 class LUInstanceConsole(NoHooksLU):
9995 """Connect to an instance's console.
9997 This is somewhat special in that it returns the command line that
9998 you need to run on the master node in order to connect to the
10004 def ExpandNames(self):
10005 self.share_locks = _ShareAll()
10006 self._ExpandAndLockInstance()
10008 def CheckPrereq(self):
10009 """Check prerequisites.
10011 This checks that the instance is in the cluster.
10014 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10015 assert self.instance is not None, \
10016 "Cannot retrieve locked instance %s" % self.op.instance_name
10017 _CheckNodeOnline(self, self.instance.primary_node)
10019 def Exec(self, feedback_fn):
10020 """Connect to the console of an instance
10023 instance = self.instance
10024 node = instance.primary_node
10026 node_insts = self.rpc.call_instance_list([node],
10027 [instance.hypervisor])[node]
10028 node_insts.Raise("Can't get node information from %s" % node)
10030 if instance.name not in node_insts.payload:
10031 if instance.admin_state == constants.ADMINST_UP:
10032 state = constants.INSTST_ERRORDOWN
10033 elif instance.admin_state == constants.ADMINST_DOWN:
10034 state = constants.INSTST_ADMINDOWN
10036 state = constants.INSTST_ADMINOFFLINE
10037 raise errors.OpExecError("Instance %s is not running (state %s)" %
10038 (instance.name, state))
10040 logging.debug("Connecting to console of %s on %s", instance.name, node)
10042 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
10045 def _GetInstanceConsole(cluster, instance):
10046 """Returns console information for an instance.
10048 @type cluster: L{objects.Cluster}
10049 @type instance: L{objects.Instance}
10053 hyper = hypervisor.GetHypervisor(instance.hypervisor)
10054 # beparams and hvparams are passed separately, to avoid editing the
10055 # instance and then saving the defaults in the instance itself.
10056 hvparams = cluster.FillHV(instance)
10057 beparams = cluster.FillBE(instance)
10058 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
10060 assert console.instance == instance.name
10061 assert console.Validate()
10063 return console.ToDict()
10066 class LUInstanceReplaceDisks(LogicalUnit):
10067 """Replace the disks of an instance.
10070 HPATH = "mirrors-replace"
10071 HTYPE = constants.HTYPE_INSTANCE
10074 def CheckArguments(self):
10075 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
10076 self.op.iallocator)
10078 def ExpandNames(self):
10079 self._ExpandAndLockInstance()
10081 assert locking.LEVEL_NODE not in self.needed_locks
10082 assert locking.LEVEL_NODE_RES not in self.needed_locks
10083 assert locking.LEVEL_NODEGROUP not in self.needed_locks
10085 assert self.op.iallocator is None or self.op.remote_node is None, \
10086 "Conflicting options"
10088 if self.op.remote_node is not None:
10089 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10091 # Warning: do not remove the locking of the new secondary here
10092 # unless DRBD8.AddChildren is changed to work in parallel;
10093 # currently it doesn't since parallel invocations of
10094 # FindUnusedMinor will conflict
10095 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
10096 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
10098 self.needed_locks[locking.LEVEL_NODE] = []
10099 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10101 if self.op.iallocator is not None:
10102 # iallocator will select a new node in the same group
10103 self.needed_locks[locking.LEVEL_NODEGROUP] = []
10105 self.needed_locks[locking.LEVEL_NODE_RES] = []
10107 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
10108 self.op.iallocator, self.op.remote_node,
10109 self.op.disks, False, self.op.early_release,
10110 self.op.ignore_ipolicy)
10112 self.tasklets = [self.replacer]
10114 def DeclareLocks(self, level):
10115 if level == locking.LEVEL_NODEGROUP:
10116 assert self.op.remote_node is None
10117 assert self.op.iallocator is not None
10118 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
10120 self.share_locks[locking.LEVEL_NODEGROUP] = 1
10121 # Lock all groups used by instance optimistically; this requires going
10122 # via the node before it's locked, requiring verification later on
10123 self.needed_locks[locking.LEVEL_NODEGROUP] = \
10124 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
10126 elif level == locking.LEVEL_NODE:
10127 if self.op.iallocator is not None:
10128 assert self.op.remote_node is None
10129 assert not self.needed_locks[locking.LEVEL_NODE]
10131 # Lock member nodes of all locked groups
10132 self.needed_locks[locking.LEVEL_NODE] = [node_name
10133 for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
10134 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
10136 self._LockInstancesNodes()
10137 elif level == locking.LEVEL_NODE_RES:
10139 self.needed_locks[locking.LEVEL_NODE_RES] = \
10140 self.needed_locks[locking.LEVEL_NODE]
10142 def BuildHooksEnv(self):
10143 """Build hooks env.
10145 This runs on the master, the primary and all the secondaries.
10148 instance = self.replacer.instance
10150 "MODE": self.op.mode,
10151 "NEW_SECONDARY": self.op.remote_node,
10152 "OLD_SECONDARY": instance.secondary_nodes[0],
10154 env.update(_BuildInstanceHookEnvByObject(self, instance))
10157 def BuildHooksNodes(self):
10158 """Build hooks nodes.
10161 instance = self.replacer.instance
10163 self.cfg.GetMasterNode(),
10164 instance.primary_node,
10166 if self.op.remote_node is not None:
10167 nl.append(self.op.remote_node)
10170 def CheckPrereq(self):
10171 """Check prerequisites.
10174 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
10175 self.op.iallocator is None)
10177 # Verify if node group locks are still correct
10178 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
10180 _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
10182 return LogicalUnit.CheckPrereq(self)
10185 class TLReplaceDisks(Tasklet):
10186 """Replaces disks for an instance.
10188 Note: Locking is not within the scope of this class.
10191 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
10192 disks, delay_iallocator, early_release, ignore_ipolicy):
10193 """Initializes this class.
10196 Tasklet.__init__(self, lu)
10199 self.instance_name = instance_name
10201 self.iallocator_name = iallocator_name
10202 self.remote_node = remote_node
10204 self.delay_iallocator = delay_iallocator
10205 self.early_release = early_release
10206 self.ignore_ipolicy = ignore_ipolicy
10209 self.instance = None
10210 self.new_node = None
10211 self.target_node = None
10212 self.other_node = None
10213 self.remote_node_info = None
10214 self.node_secondary_ip = None
10217 def CheckArguments(mode, remote_node, iallocator):
10218 """Helper function for users of this class.
10221 # check for valid parameter combination
10222 if mode == constants.REPLACE_DISK_CHG:
10223 if remote_node is None and iallocator is None:
10224 raise errors.OpPrereqError("When changing the secondary either an"
10225 " iallocator script must be used or the"
10226 " new node given", errors.ECODE_INVAL)
10228 if remote_node is not None and iallocator is not None:
10229 raise errors.OpPrereqError("Give either the iallocator or the new"
10230 " secondary, not both", errors.ECODE_INVAL)
10232 elif remote_node is not None or iallocator is not None:
10233 # Not replacing the secondary
10234 raise errors.OpPrereqError("The iallocator and new node options can"
10235 " only be used when changing the"
10236 " secondary node", errors.ECODE_INVAL)
10239 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
10240 """Compute a new secondary node using an IAllocator.
10243 ial = IAllocator(lu.cfg, lu.rpc,
10244 mode=constants.IALLOCATOR_MODE_RELOC,
10245 name=instance_name,
10246 relocate_from=list(relocate_from))
10248 ial.Run(iallocator_name)
10250 if not ial.success:
10251 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
10252 " %s" % (iallocator_name, ial.info),
10253 errors.ECODE_NORES)
10255 if len(ial.result) != ial.required_nodes:
10256 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
10257 " of nodes (%s), required %s" %
10259 len(ial.result), ial.required_nodes),
10260 errors.ECODE_FAULT)
10262 remote_node_name = ial.result[0]
10264 lu.LogInfo("Selected new secondary for instance '%s': %s",
10265 instance_name, remote_node_name)
10267 return remote_node_name
10269 def _FindFaultyDisks(self, node_name):
10270 """Wrapper for L{_FindFaultyInstanceDisks}.
10273 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
10276 def _CheckDisksActivated(self, instance):
10277 """Checks if the instance disks are activated.
10279 @param instance: The instance to check disks
10280 @return: True if they are activated, False otherwise
10283 nodes = instance.all_nodes
10285 for idx, dev in enumerate(instance.disks):
10287 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
10288 self.cfg.SetDiskID(dev, node)
10290 result = self.rpc.call_blockdev_find(node, dev)
10294 elif result.fail_msg or not result.payload:
10299 def CheckPrereq(self):
10300 """Check prerequisites.
10302 This checks that the instance is in the cluster.
10305 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
10306 assert instance is not None, \
10307 "Cannot retrieve locked instance %s" % self.instance_name
10309 if instance.disk_template != constants.DT_DRBD8:
10310 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
10311 " instances", errors.ECODE_INVAL)
10313 if len(instance.secondary_nodes) != 1:
10314 raise errors.OpPrereqError("The instance has a strange layout,"
10315 " expected one secondary but found %d" %
10316 len(instance.secondary_nodes),
10317 errors.ECODE_FAULT)
10319 if not self.delay_iallocator:
10320 self._CheckPrereq2()
10322 def _CheckPrereq2(self):
10323 """Check prerequisites, second part.
10325 This function should always be part of CheckPrereq. It was separated and is
10326 now called from Exec because during node evacuation iallocator was only
10327 called with an unmodified cluster model, not taking planned changes into
10331 instance = self.instance
10332 secondary_node = instance.secondary_nodes[0]
10334 if self.iallocator_name is None:
10335 remote_node = self.remote_node
10337 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
10338 instance.name, instance.secondary_nodes)
10340 if remote_node is None:
10341 self.remote_node_info = None
10343 assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \
10344 "Remote node '%s' is not locked" % remote_node
10346 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
10347 assert self.remote_node_info is not None, \
10348 "Cannot retrieve locked node %s" % remote_node
10350 if remote_node == self.instance.primary_node:
10351 raise errors.OpPrereqError("The specified node is the primary node of"
10352 " the instance", errors.ECODE_INVAL)
10354 if remote_node == secondary_node:
10355 raise errors.OpPrereqError("The specified node is already the"
10356 " secondary node of the instance",
10357 errors.ECODE_INVAL)
10359 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
10360 constants.REPLACE_DISK_CHG):
10361 raise errors.OpPrereqError("Cannot specify disks to be replaced",
10362 errors.ECODE_INVAL)
10364 if self.mode == constants.REPLACE_DISK_AUTO:
10365 if not self._CheckDisksActivated(instance):
10366 raise errors.OpPrereqError("Please run activate-disks on instance %s"
10367 " first" % self.instance_name,
10368 errors.ECODE_STATE)
10369 faulty_primary = self._FindFaultyDisks(instance.primary_node)
10370 faulty_secondary = self._FindFaultyDisks(secondary_node)
10372 if faulty_primary and faulty_secondary:
10373 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
10374 " one node and can not be repaired"
10375 " automatically" % self.instance_name,
10376 errors.ECODE_STATE)
10379 self.disks = faulty_primary
10380 self.target_node = instance.primary_node
10381 self.other_node = secondary_node
10382 check_nodes = [self.target_node, self.other_node]
10383 elif faulty_secondary:
10384 self.disks = faulty_secondary
10385 self.target_node = secondary_node
10386 self.other_node = instance.primary_node
10387 check_nodes = [self.target_node, self.other_node]
10393 # Non-automatic modes
10394 if self.mode == constants.REPLACE_DISK_PRI:
10395 self.target_node = instance.primary_node
10396 self.other_node = secondary_node
10397 check_nodes = [self.target_node, self.other_node]
10399 elif self.mode == constants.REPLACE_DISK_SEC:
10400 self.target_node = secondary_node
10401 self.other_node = instance.primary_node
10402 check_nodes = [self.target_node, self.other_node]
10404 elif self.mode == constants.REPLACE_DISK_CHG:
10405 self.new_node = remote_node
10406 self.other_node = instance.primary_node
10407 self.target_node = secondary_node
10408 check_nodes = [self.new_node, self.other_node]
10410 _CheckNodeNotDrained(self.lu, remote_node)
10411 _CheckNodeVmCapable(self.lu, remote_node)
10413 old_node_info = self.cfg.GetNodeInfo(secondary_node)
10414 assert old_node_info is not None
10415 if old_node_info.offline and not self.early_release:
10416 # doesn't make sense to delay the release
10417 self.early_release = True
10418 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
10419 " early-release mode", secondary_node)
10422 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
10425 # If not specified all disks should be replaced
10427 self.disks = range(len(self.instance.disks))
10429 # TODO: This is ugly, but right now we can't distinguish between internal
10430 # submitted opcode and external one. We should fix that.
10431 if self.remote_node_info:
10432 # We change the node, lets verify it still meets instance policy
10433 new_group_info = self.cfg.GetNodeGroup(self.remote_node_info.group)
10434 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(),
10436 _CheckTargetNodeIPolicy(self, ipolicy, instance, self.remote_node_info,
10437 ignore=self.ignore_ipolicy)
10439 # TODO: compute disk parameters
10440 primary_node_info = self.cfg.GetNodeInfo(instance.primary_node)
10441 secondary_node_info = self.cfg.GetNodeInfo(secondary_node)
10442 if primary_node_info.group != secondary_node_info.group:
10443 self.lu.LogInfo("The instance primary and secondary nodes are in two"
10444 " different node groups; the disk parameters of the"
10445 " primary node's group will be applied.")
10447 self.diskparams = self.cfg.GetNodeGroup(primary_node_info.group).diskparams
10449 for node in check_nodes:
10450 _CheckNodeOnline(self.lu, node)
10452 touched_nodes = frozenset(node_name for node_name in [self.new_node,
10455 if node_name is not None)
10457 # Release unneeded node and node resource locks
10458 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
10459 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES, keep=touched_nodes)
10461 # Release any owned node group
10462 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
10463 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
10465 # Check whether disks are valid
10466 for disk_idx in self.disks:
10467 instance.FindDisk(disk_idx)
10469 # Get secondary node IP addresses
10470 self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node)
10471 in self.cfg.GetMultiNodeInfo(touched_nodes))
10473 def Exec(self, feedback_fn):
10474 """Execute disk replacement.
10476 This dispatches the disk replacement to the appropriate handler.
10479 if self.delay_iallocator:
10480 self._CheckPrereq2()
10483 # Verify owned locks before starting operation
10484 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
10485 assert set(owned_nodes) == set(self.node_secondary_ip), \
10486 ("Incorrect node locks, owning %s, expected %s" %
10487 (owned_nodes, self.node_secondary_ip.keys()))
10488 assert (self.lu.owned_locks(locking.LEVEL_NODE) ==
10489 self.lu.owned_locks(locking.LEVEL_NODE_RES))
10491 owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
10492 assert list(owned_instances) == [self.instance_name], \
10493 "Instance '%s' not locked" % self.instance_name
10495 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
10496 "Should not own any node group lock at this point"
10499 feedback_fn("No disks need replacement")
10502 feedback_fn("Replacing disk(s) %s for %s" %
10503 (utils.CommaJoin(self.disks), self.instance.name))
10505 activate_disks = (self.instance.admin_state != constants.ADMINST_UP)
10507 # Activate the instance disks if we're replacing them on a down instance
10509 _StartInstanceDisks(self.lu, self.instance, True)
10512 # Should we replace the secondary node?
10513 if self.new_node is not None:
10514 fn = self._ExecDrbd8Secondary
10516 fn = self._ExecDrbd8DiskOnly
10518 result = fn(feedback_fn)
10520 # Deactivate the instance disks if we're replacing them on a
10523 _SafeShutdownInstanceDisks(self.lu, self.instance)
10525 assert not self.lu.owned_locks(locking.LEVEL_NODE)
10528 # Verify owned locks
10529 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE_RES)
10530 nodes = frozenset(self.node_secondary_ip)
10531 assert ((self.early_release and not owned_nodes) or
10532 (not self.early_release and not (set(owned_nodes) - nodes))), \
10533 ("Not owning the correct locks, early_release=%s, owned=%r,"
10534 " nodes=%r" % (self.early_release, owned_nodes, nodes))
10538 def _CheckVolumeGroup(self, nodes):
10539 self.lu.LogInfo("Checking volume groups")
10541 vgname = self.cfg.GetVGName()
10543 # Make sure volume group exists on all involved nodes
10544 results = self.rpc.call_vg_list(nodes)
10546 raise errors.OpExecError("Can't list volume groups on the nodes")
10549 res = results[node]
10550 res.Raise("Error checking node %s" % node)
10551 if vgname not in res.payload:
10552 raise errors.OpExecError("Volume group '%s' not found on node %s" %
10555 def _CheckDisksExistence(self, nodes):
10556 # Check disk existence
10557 for idx, dev in enumerate(self.instance.disks):
10558 if idx not in self.disks:
10562 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
10563 self.cfg.SetDiskID(dev, node)
10565 result = self.rpc.call_blockdev_find(node, dev)
10567 msg = result.fail_msg
10568 if msg or not result.payload:
10570 msg = "disk not found"
10571 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
10574 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
10575 for idx, dev in enumerate(self.instance.disks):
10576 if idx not in self.disks:
10579 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
10582 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
10584 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
10585 " replace disks for instance %s" %
10586 (node_name, self.instance.name))
10588 def _CreateNewStorage(self, node_name):
10589 """Create new storage on the primary or secondary node.
10591 This is only used for same-node replaces, not for changing the
10592 secondary node, hence we don't want to modify the existing disk.
10597 for idx, dev in enumerate(self.instance.disks):
10598 if idx not in self.disks:
10601 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
10603 self.cfg.SetDiskID(dev, node_name)
10605 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
10606 names = _GenerateUniqueNames(self.lu, lv_names)
10608 _, data_p, meta_p = _ComputeLDParams(constants.DT_DRBD8, self.diskparams)
10610 vg_data = dev.children[0].logical_id[0]
10611 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
10612 logical_id=(vg_data, names[0]), params=data_p)
10613 vg_meta = dev.children[1].logical_id[0]
10614 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,
10615 logical_id=(vg_meta, names[1]), params=meta_p)
10617 new_lvs = [lv_data, lv_meta]
10618 old_lvs = [child.Copy() for child in dev.children]
10619 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
10621 # we pass force_create=True to force the LVM creation
10622 for new_lv in new_lvs:
10623 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
10624 _GetInstanceInfoText(self.instance), False)
10628 def _CheckDevices(self, node_name, iv_names):
10629 for name, (dev, _, _) in iv_names.iteritems():
10630 self.cfg.SetDiskID(dev, node_name)
10632 result = self.rpc.call_blockdev_find(node_name, dev)
10634 msg = result.fail_msg
10635 if msg or not result.payload:
10637 msg = "disk not found"
10638 raise errors.OpExecError("Can't find DRBD device %s: %s" %
10641 if result.payload.is_degraded:
10642 raise errors.OpExecError("DRBD device %s is degraded!" % name)
10644 def _RemoveOldStorage(self, node_name, iv_names):
10645 for name, (_, old_lvs, _) in iv_names.iteritems():
10646 self.lu.LogInfo("Remove logical volumes for %s" % name)
10649 self.cfg.SetDiskID(lv, node_name)
10651 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
10653 self.lu.LogWarning("Can't remove old LV: %s" % msg,
10654 hint="remove unused LVs manually")
10656 def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
10657 """Replace a disk on the primary or secondary for DRBD 8.
10659 The algorithm for replace is quite complicated:
10661 1. for each disk to be replaced:
10663 1. create new LVs on the target node with unique names
10664 1. detach old LVs from the drbd device
10665 1. rename old LVs to name_replaced.<time_t>
10666 1. rename new LVs to old LVs
10667 1. attach the new LVs (with the old names now) to the drbd device
10669 1. wait for sync across all devices
10671 1. for each modified disk:
10673 1. remove old LVs (which have the name name_replaces.<time_t>)
10675 Failures are not very well handled.
10680 # Step: check device activation
10681 self.lu.LogStep(1, steps_total, "Check device existence")
10682 self._CheckDisksExistence([self.other_node, self.target_node])
10683 self._CheckVolumeGroup([self.target_node, self.other_node])
10685 # Step: check other node consistency
10686 self.lu.LogStep(2, steps_total, "Check peer consistency")
10687 self._CheckDisksConsistency(self.other_node,
10688 self.other_node == self.instance.primary_node,
10691 # Step: create new storage
10692 self.lu.LogStep(3, steps_total, "Allocate new storage")
10693 iv_names = self._CreateNewStorage(self.target_node)
10695 # Step: for each lv, detach+rename*2+attach
10696 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
10697 for dev, old_lvs, new_lvs in iv_names.itervalues():
10698 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
10700 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
10702 result.Raise("Can't detach drbd from local storage on node"
10703 " %s for device %s" % (self.target_node, dev.iv_name))
10705 #cfg.Update(instance)
10707 # ok, we created the new LVs, so now we know we have the needed
10708 # storage; as such, we proceed on the target node to rename
10709 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
10710 # using the assumption that logical_id == physical_id (which in
10711 # turn is the unique_id on that node)
10713 # FIXME(iustin): use a better name for the replaced LVs
10714 temp_suffix = int(time.time())
10715 ren_fn = lambda d, suff: (d.physical_id[0],
10716 d.physical_id[1] + "_replaced-%s" % suff)
10718 # Build the rename list based on what LVs exist on the node
10719 rename_old_to_new = []
10720 for to_ren in old_lvs:
10721 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
10722 if not result.fail_msg and result.payload:
10724 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
10726 self.lu.LogInfo("Renaming the old LVs on the target node")
10727 result = self.rpc.call_blockdev_rename(self.target_node,
10729 result.Raise("Can't rename old LVs on node %s" % self.target_node)
10731 # Now we rename the new LVs to the old LVs
10732 self.lu.LogInfo("Renaming the new LVs on the target node")
10733 rename_new_to_old = [(new, old.physical_id)
10734 for old, new in zip(old_lvs, new_lvs)]
10735 result = self.rpc.call_blockdev_rename(self.target_node,
10737 result.Raise("Can't rename new LVs on node %s" % self.target_node)
10739 # Intermediate steps of in memory modifications
10740 for old, new in zip(old_lvs, new_lvs):
10741 new.logical_id = old.logical_id
10742 self.cfg.SetDiskID(new, self.target_node)
10744 # We need to modify old_lvs so that removal later removes the
10745 # right LVs, not the newly added ones; note that old_lvs is a
10747 for disk in old_lvs:
10748 disk.logical_id = ren_fn(disk, temp_suffix)
10749 self.cfg.SetDiskID(disk, self.target_node)
10751 # Now that the new lvs have the old name, we can add them to the device
10752 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
10753 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
10755 msg = result.fail_msg
10757 for new_lv in new_lvs:
10758 msg2 = self.rpc.call_blockdev_remove(self.target_node,
10761 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
10762 hint=("cleanup manually the unused logical"
10764 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
10766 cstep = itertools.count(5)
10768 if self.early_release:
10769 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10770 self._RemoveOldStorage(self.target_node, iv_names)
10771 # TODO: Check if releasing locks early still makes sense
10772 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
10774 # Release all resource locks except those used by the instance
10775 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
10776 keep=self.node_secondary_ip.keys())
10778 # Release all node locks while waiting for sync
10779 _ReleaseLocks(self.lu, locking.LEVEL_NODE)
10781 # TODO: Can the instance lock be downgraded here? Take the optional disk
10782 # shutdown in the caller into consideration.
10785 # This can fail as the old devices are degraded and _WaitForSync
10786 # does a combined result over all disks, so we don't check its return value
10787 self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
10788 _WaitForSync(self.lu, self.instance)
10790 # Check all devices manually
10791 self._CheckDevices(self.instance.primary_node, iv_names)
10793 # Step: remove old storage
10794 if not self.early_release:
10795 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10796 self._RemoveOldStorage(self.target_node, iv_names)
10798 def _ExecDrbd8Secondary(self, feedback_fn):
10799 """Replace the secondary node for DRBD 8.
10801 The algorithm for replace is quite complicated:
10802 - for all disks of the instance:
10803 - create new LVs on the new node with same names
10804 - shutdown the drbd device on the old secondary
10805 - disconnect the drbd network on the primary
10806 - create the drbd device on the new secondary
10807 - network attach the drbd on the primary, using an artifice:
10808 the drbd code for Attach() will connect to the network if it
10809 finds a device which is connected to the good local disks but
10810 not network enabled
10811 - wait for sync across all devices
10812 - remove all disks from the old secondary
10814 Failures are not very well handled.
10819 pnode = self.instance.primary_node
10821 # Step: check device activation
10822 self.lu.LogStep(1, steps_total, "Check device existence")
10823 self._CheckDisksExistence([self.instance.primary_node])
10824 self._CheckVolumeGroup([self.instance.primary_node])
10826 # Step: check other node consistency
10827 self.lu.LogStep(2, steps_total, "Check peer consistency")
10828 self._CheckDisksConsistency(self.instance.primary_node, True, True)
10830 # Step: create new storage
10831 self.lu.LogStep(3, steps_total, "Allocate new storage")
10832 for idx, dev in enumerate(self.instance.disks):
10833 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
10834 (self.new_node, idx))
10835 # we pass force_create=True to force LVM creation
10836 for new_lv in dev.children:
10837 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
10838 _GetInstanceInfoText(self.instance), False)
10840 # Step 4: dbrd minors and drbd setups changes
10841 # after this, we must manually remove the drbd minors on both the
10842 # error and the success paths
10843 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
10844 minors = self.cfg.AllocateDRBDMinor([self.new_node
10845 for dev in self.instance.disks],
10846 self.instance.name)
10847 logging.debug("Allocated minors %r", minors)
10850 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
10851 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
10852 (self.new_node, idx))
10853 # create new devices on new_node; note that we create two IDs:
10854 # one without port, so the drbd will be activated without
10855 # networking information on the new node at this stage, and one
10856 # with network, for the latter activation in step 4
10857 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
10858 if self.instance.primary_node == o_node1:
10861 assert self.instance.primary_node == o_node2, "Three-node instance?"
10864 new_alone_id = (self.instance.primary_node, self.new_node, None,
10865 p_minor, new_minor, o_secret)
10866 new_net_id = (self.instance.primary_node, self.new_node, o_port,
10867 p_minor, new_minor, o_secret)
10869 iv_names[idx] = (dev, dev.children, new_net_id)
10870 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
10872 drbd_params, _, _ = _ComputeLDParams(constants.DT_DRBD8, self.diskparams)
10873 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
10874 logical_id=new_alone_id,
10875 children=dev.children,
10877 params=drbd_params)
10879 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
10880 _GetInstanceInfoText(self.instance), False)
10881 except errors.GenericError:
10882 self.cfg.ReleaseDRBDMinors(self.instance.name)
10885 # We have new devices, shutdown the drbd on the old secondary
10886 for idx, dev in enumerate(self.instance.disks):
10887 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
10888 self.cfg.SetDiskID(dev, self.target_node)
10889 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
10891 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
10892 "node: %s" % (idx, msg),
10893 hint=("Please cleanup this device manually as"
10894 " soon as possible"))
10896 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
10897 result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
10898 self.instance.disks)[pnode]
10900 msg = result.fail_msg
10902 # detaches didn't succeed (unlikely)
10903 self.cfg.ReleaseDRBDMinors(self.instance.name)
10904 raise errors.OpExecError("Can't detach the disks from the network on"
10905 " old node: %s" % (msg,))
10907 # if we managed to detach at least one, we update all the disks of
10908 # the instance to point to the new secondary
10909 self.lu.LogInfo("Updating instance configuration")
10910 for dev, _, new_logical_id in iv_names.itervalues():
10911 dev.logical_id = new_logical_id
10912 self.cfg.SetDiskID(dev, self.instance.primary_node)
10914 self.cfg.Update(self.instance, feedback_fn)
10916 # Release all node locks (the configuration has been updated)
10917 _ReleaseLocks(self.lu, locking.LEVEL_NODE)
10919 # and now perform the drbd attach
10920 self.lu.LogInfo("Attaching primary drbds to new secondary"
10921 " (standalone => connected)")
10922 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
10924 self.node_secondary_ip,
10925 self.instance.disks,
10926 self.instance.name,
10928 for to_node, to_result in result.items():
10929 msg = to_result.fail_msg
10931 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
10933 hint=("please do a gnt-instance info to see the"
10934 " status of disks"))
10936 cstep = itertools.count(5)
10938 if self.early_release:
10939 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10940 self._RemoveOldStorage(self.target_node, iv_names)
10941 # TODO: Check if releasing locks early still makes sense
10942 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
10944 # Release all resource locks except those used by the instance
10945 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
10946 keep=self.node_secondary_ip.keys())
10948 # TODO: Can the instance lock be downgraded here? Take the optional disk
10949 # shutdown in the caller into consideration.
10952 # This can fail as the old devices are degraded and _WaitForSync
10953 # does a combined result over all disks, so we don't check its return value
10954 self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
10955 _WaitForSync(self.lu, self.instance)
10957 # Check all devices manually
10958 self._CheckDevices(self.instance.primary_node, iv_names)
10960 # Step: remove old storage
10961 if not self.early_release:
10962 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10963 self._RemoveOldStorage(self.target_node, iv_names)
10966 class LURepairNodeStorage(NoHooksLU):
10967 """Repairs the volume group on a node.
10972 def CheckArguments(self):
10973 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
10975 storage_type = self.op.storage_type
10977 if (constants.SO_FIX_CONSISTENCY not in
10978 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
10979 raise errors.OpPrereqError("Storage units of type '%s' can not be"
10980 " repaired" % storage_type,
10981 errors.ECODE_INVAL)
10983 def ExpandNames(self):
10984 self.needed_locks = {
10985 locking.LEVEL_NODE: [self.op.node_name],
10988 def _CheckFaultyDisks(self, instance, node_name):
10989 """Ensure faulty disks abort the opcode or at least warn."""
10991 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
10993 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
10994 " node '%s'" % (instance.name, node_name),
10995 errors.ECODE_STATE)
10996 except errors.OpPrereqError, err:
10997 if self.op.ignore_consistency:
10998 self.proc.LogWarning(str(err.args[0]))
11002 def CheckPrereq(self):
11003 """Check prerequisites.
11006 # Check whether any instance on this node has faulty disks
11007 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
11008 if inst.admin_state != constants.ADMINST_UP:
11010 check_nodes = set(inst.all_nodes)
11011 check_nodes.discard(self.op.node_name)
11012 for inst_node_name in check_nodes:
11013 self._CheckFaultyDisks(inst, inst_node_name)
11015 def Exec(self, feedback_fn):
11016 feedback_fn("Repairing storage unit '%s' on %s ..." %
11017 (self.op.name, self.op.node_name))
11019 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
11020 result = self.rpc.call_storage_execute(self.op.node_name,
11021 self.op.storage_type, st_args,
11023 constants.SO_FIX_CONSISTENCY)
11024 result.Raise("Failed to repair storage unit '%s' on %s" %
11025 (self.op.name, self.op.node_name))
11028 class LUNodeEvacuate(NoHooksLU):
11029 """Evacuates instances off a list of nodes.
11034 _MODE2IALLOCATOR = {
11035 constants.NODE_EVAC_PRI: constants.IALLOCATOR_NEVAC_PRI,
11036 constants.NODE_EVAC_SEC: constants.IALLOCATOR_NEVAC_SEC,
11037 constants.NODE_EVAC_ALL: constants.IALLOCATOR_NEVAC_ALL,
11039 assert frozenset(_MODE2IALLOCATOR.keys()) == constants.NODE_EVAC_MODES
11040 assert (frozenset(_MODE2IALLOCATOR.values()) ==
11041 constants.IALLOCATOR_NEVAC_MODES)
11043 def CheckArguments(self):
11044 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
11046 def ExpandNames(self):
11047 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
11049 if self.op.remote_node is not None:
11050 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
11051 assert self.op.remote_node
11053 if self.op.remote_node == self.op.node_name:
11054 raise errors.OpPrereqError("Can not use evacuated node as a new"
11055 " secondary node", errors.ECODE_INVAL)
11057 if self.op.mode != constants.NODE_EVAC_SEC:
11058 raise errors.OpPrereqError("Without the use of an iallocator only"
11059 " secondary instances can be evacuated",
11060 errors.ECODE_INVAL)
11063 self.share_locks = _ShareAll()
11064 self.needed_locks = {
11065 locking.LEVEL_INSTANCE: [],
11066 locking.LEVEL_NODEGROUP: [],
11067 locking.LEVEL_NODE: [],
11070 # Determine nodes (via group) optimistically, needs verification once locks
11071 # have been acquired
11072 self.lock_nodes = self._DetermineNodes()
11074 def _DetermineNodes(self):
11075 """Gets the list of nodes to operate on.
11078 if self.op.remote_node is None:
11079 # Iallocator will choose any node(s) in the same group
11080 group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name])
11082 group_nodes = frozenset([self.op.remote_node])
11084 # Determine nodes to be locked
11085 return set([self.op.node_name]) | group_nodes
11087 def _DetermineInstances(self):
11088 """Builds list of instances to operate on.
11091 assert self.op.mode in constants.NODE_EVAC_MODES
11093 if self.op.mode == constants.NODE_EVAC_PRI:
11094 # Primary instances only
11095 inst_fn = _GetNodePrimaryInstances
11096 assert self.op.remote_node is None, \
11097 "Evacuating primary instances requires iallocator"
11098 elif self.op.mode == constants.NODE_EVAC_SEC:
11099 # Secondary instances only
11100 inst_fn = _GetNodeSecondaryInstances
11103 assert self.op.mode == constants.NODE_EVAC_ALL
11104 inst_fn = _GetNodeInstances
11105 # TODO: In 2.6, change the iallocator interface to take an evacuation mode
11107 raise errors.OpPrereqError("Due to an issue with the iallocator"
11108 " interface it is not possible to evacuate"
11109 " all instances at once; specify explicitly"
11110 " whether to evacuate primary or secondary"
11112 errors.ECODE_INVAL)
11114 return inst_fn(self.cfg, self.op.node_name)
11116 def DeclareLocks(self, level):
11117 if level == locking.LEVEL_INSTANCE:
11118 # Lock instances optimistically, needs verification once node and group
11119 # locks have been acquired
11120 self.needed_locks[locking.LEVEL_INSTANCE] = \
11121 set(i.name for i in self._DetermineInstances())
11123 elif level == locking.LEVEL_NODEGROUP:
11124 # Lock node groups for all potential target nodes optimistically, needs
11125 # verification once nodes have been acquired
11126 self.needed_locks[locking.LEVEL_NODEGROUP] = \
11127 self.cfg.GetNodeGroupsFromNodes(self.lock_nodes)
11129 elif level == locking.LEVEL_NODE:
11130 self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes
11132 def CheckPrereq(self):
11134 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
11135 owned_nodes = self.owned_locks(locking.LEVEL_NODE)
11136 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
11138 need_nodes = self._DetermineNodes()
11140 if not owned_nodes.issuperset(need_nodes):
11141 raise errors.OpPrereqError("Nodes in same group as '%s' changed since"
11142 " locks were acquired, current nodes are"
11143 " are '%s', used to be '%s'; retry the"
11145 (self.op.node_name,
11146 utils.CommaJoin(need_nodes),
11147 utils.CommaJoin(owned_nodes)),
11148 errors.ECODE_STATE)
11150 wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
11151 if owned_groups != wanted_groups:
11152 raise errors.OpExecError("Node groups changed since locks were acquired,"
11153 " current groups are '%s', used to be '%s';"
11154 " retry the operation" %
11155 (utils.CommaJoin(wanted_groups),
11156 utils.CommaJoin(owned_groups)))
11158 # Determine affected instances
11159 self.instances = self._DetermineInstances()
11160 self.instance_names = [i.name for i in self.instances]
11162 if set(self.instance_names) != owned_instances:
11163 raise errors.OpExecError("Instances on node '%s' changed since locks"
11164 " were acquired, current instances are '%s',"
11165 " used to be '%s'; retry the operation" %
11166 (self.op.node_name,
11167 utils.CommaJoin(self.instance_names),
11168 utils.CommaJoin(owned_instances)))
11170 if self.instance_names:
11171 self.LogInfo("Evacuating instances from node '%s': %s",
11173 utils.CommaJoin(utils.NiceSort(self.instance_names)))
11175 self.LogInfo("No instances to evacuate from node '%s'",
11178 if self.op.remote_node is not None:
11179 for i in self.instances:
11180 if i.primary_node == self.op.remote_node:
11181 raise errors.OpPrereqError("Node %s is the primary node of"
11182 " instance %s, cannot use it as"
11184 (self.op.remote_node, i.name),
11185 errors.ECODE_INVAL)
11187 def Exec(self, feedback_fn):
11188 assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None)
11190 if not self.instance_names:
11191 # No instances to evacuate
11194 elif self.op.iallocator is not None:
11195 # TODO: Implement relocation to other group
11196 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_NODE_EVAC,
11197 evac_mode=self._MODE2IALLOCATOR[self.op.mode],
11198 instances=list(self.instance_names))
11200 ial.Run(self.op.iallocator)
11202 if not ial.success:
11203 raise errors.OpPrereqError("Can't compute node evacuation using"
11204 " iallocator '%s': %s" %
11205 (self.op.iallocator, ial.info),
11206 errors.ECODE_NORES)
11208 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True)
11210 elif self.op.remote_node is not None:
11211 assert self.op.mode == constants.NODE_EVAC_SEC
11213 [opcodes.OpInstanceReplaceDisks(instance_name=instance_name,
11214 remote_node=self.op.remote_node,
11216 mode=constants.REPLACE_DISK_CHG,
11217 early_release=self.op.early_release)]
11218 for instance_name in self.instance_names
11222 raise errors.ProgrammerError("No iallocator or remote node")
11224 return ResultWithJobs(jobs)
11227 def _SetOpEarlyRelease(early_release, op):
11228 """Sets C{early_release} flag on opcodes if available.
11232 op.early_release = early_release
11233 except AttributeError:
11234 assert not isinstance(op, opcodes.OpInstanceReplaceDisks)
11239 def _NodeEvacDest(use_nodes, group, nodes):
11240 """Returns group or nodes depending on caller's choice.
11244 return utils.CommaJoin(nodes)
11249 def _LoadNodeEvacResult(lu, alloc_result, early_release, use_nodes):
11250 """Unpacks the result of change-group and node-evacuate iallocator requests.
11252 Iallocator modes L{constants.IALLOCATOR_MODE_NODE_EVAC} and
11253 L{constants.IALLOCATOR_MODE_CHG_GROUP}.
11255 @type lu: L{LogicalUnit}
11256 @param lu: Logical unit instance
11257 @type alloc_result: tuple/list
11258 @param alloc_result: Result from iallocator
11259 @type early_release: bool
11260 @param early_release: Whether to release locks early if possible
11261 @type use_nodes: bool
11262 @param use_nodes: Whether to display node names instead of groups
11265 (moved, failed, jobs) = alloc_result
11268 failreason = utils.CommaJoin("%s (%s)" % (name, reason)
11269 for (name, reason) in failed)
11270 lu.LogWarning("Unable to evacuate instances %s", failreason)
11271 raise errors.OpExecError("Unable to evacuate instances %s" % failreason)
11274 lu.LogInfo("Instances to be moved: %s",
11275 utils.CommaJoin("%s (to %s)" %
11276 (name, _NodeEvacDest(use_nodes, group, nodes))
11277 for (name, group, nodes) in moved))
11279 return [map(compat.partial(_SetOpEarlyRelease, early_release),
11280 map(opcodes.OpCode.LoadOpCode, ops))
11284 class LUInstanceGrowDisk(LogicalUnit):
11285 """Grow a disk of an instance.
11288 HPATH = "disk-grow"
11289 HTYPE = constants.HTYPE_INSTANCE
11292 def ExpandNames(self):
11293 self._ExpandAndLockInstance()
11294 self.needed_locks[locking.LEVEL_NODE] = []
11295 self.needed_locks[locking.LEVEL_NODE_RES] = []
11296 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11297 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
11299 def DeclareLocks(self, level):
11300 if level == locking.LEVEL_NODE:
11301 self._LockInstancesNodes()
11302 elif level == locking.LEVEL_NODE_RES:
11304 self.needed_locks[locking.LEVEL_NODE_RES] = \
11305 self.needed_locks[locking.LEVEL_NODE][:]
11307 def BuildHooksEnv(self):
11308 """Build hooks env.
11310 This runs on the master, the primary and all the secondaries.
11314 "DISK": self.op.disk,
11315 "AMOUNT": self.op.amount,
11317 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11320 def BuildHooksNodes(self):
11321 """Build hooks nodes.
11324 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
11327 def CheckPrereq(self):
11328 """Check prerequisites.
11330 This checks that the instance is in the cluster.
11333 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
11334 assert instance is not None, \
11335 "Cannot retrieve locked instance %s" % self.op.instance_name
11336 nodenames = list(instance.all_nodes)
11337 for node in nodenames:
11338 _CheckNodeOnline(self, node)
11340 self.instance = instance
11342 if instance.disk_template not in constants.DTS_GROWABLE:
11343 raise errors.OpPrereqError("Instance's disk layout does not support"
11344 " growing", errors.ECODE_INVAL)
11346 self.disk = instance.FindDisk(self.op.disk)
11348 if instance.disk_template not in (constants.DT_FILE,
11349 constants.DT_SHARED_FILE):
11350 # TODO: check the free disk space for file, when that feature will be
11352 _CheckNodesFreeDiskPerVG(self, nodenames,
11353 self.disk.ComputeGrowth(self.op.amount))
11355 def Exec(self, feedback_fn):
11356 """Execute disk grow.
11359 instance = self.instance
11362 assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
11363 assert (self.owned_locks(locking.LEVEL_NODE) ==
11364 self.owned_locks(locking.LEVEL_NODE_RES))
11366 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
11368 raise errors.OpExecError("Cannot activate block device to grow")
11370 feedback_fn("Growing disk %s of instance '%s' by %s" %
11371 (self.op.disk, instance.name,
11372 utils.FormatUnit(self.op.amount, "h")))
11374 # First run all grow ops in dry-run mode
11375 for node in instance.all_nodes:
11376 self.cfg.SetDiskID(disk, node)
11377 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, True)
11378 result.Raise("Grow request failed to node %s" % node)
11380 # We know that (as far as we can test) operations across different
11381 # nodes will succeed, time to run it for real
11382 for node in instance.all_nodes:
11383 self.cfg.SetDiskID(disk, node)
11384 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, False)
11385 result.Raise("Grow request failed to node %s" % node)
11387 # TODO: Rewrite code to work properly
11388 # DRBD goes into sync mode for a short amount of time after executing the
11389 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
11390 # calling "resize" in sync mode fails. Sleeping for a short amount of
11391 # time is a work-around.
11394 disk.RecordGrow(self.op.amount)
11395 self.cfg.Update(instance, feedback_fn)
11397 # Changes have been recorded, release node lock
11398 _ReleaseLocks(self, locking.LEVEL_NODE)
11400 # Downgrade lock while waiting for sync
11401 self.glm.downgrade(locking.LEVEL_INSTANCE)
11403 if self.op.wait_for_sync:
11404 disk_abort = not _WaitForSync(self, instance, disks=[disk])
11406 self.proc.LogWarning("Disk sync-ing has not returned a good"
11407 " status; please check the instance")
11408 if instance.admin_state != constants.ADMINST_UP:
11409 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
11410 elif instance.admin_state != constants.ADMINST_UP:
11411 self.proc.LogWarning("Not shutting down the disk even if the instance is"
11412 " not supposed to be running because no wait for"
11413 " sync mode was requested")
11415 assert self.owned_locks(locking.LEVEL_NODE_RES)
11416 assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
11419 class LUInstanceQueryData(NoHooksLU):
11420 """Query runtime instance data.
11425 def ExpandNames(self):
11426 self.needed_locks = {}
11428 # Use locking if requested or when non-static information is wanted
11429 if not (self.op.static or self.op.use_locking):
11430 self.LogWarning("Non-static data requested, locks need to be acquired")
11431 self.op.use_locking = True
11433 if self.op.instances or not self.op.use_locking:
11434 # Expand instance names right here
11435 self.wanted_names = _GetWantedInstances(self, self.op.instances)
11437 # Will use acquired locks
11438 self.wanted_names = None
11440 if self.op.use_locking:
11441 self.share_locks = _ShareAll()
11443 if self.wanted_names is None:
11444 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
11446 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
11448 self.needed_locks[locking.LEVEL_NODE] = []
11449 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11451 def DeclareLocks(self, level):
11452 if self.op.use_locking and level == locking.LEVEL_NODE:
11453 self._LockInstancesNodes()
11455 def CheckPrereq(self):
11456 """Check prerequisites.
11458 This only checks the optional instance list against the existing names.
11461 if self.wanted_names is None:
11462 assert self.op.use_locking, "Locking was not used"
11463 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
11465 self.wanted_instances = \
11466 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
11468 def _ComputeBlockdevStatus(self, node, instance_name, dev):
11469 """Returns the status of a block device
11472 if self.op.static or not node:
11475 self.cfg.SetDiskID(dev, node)
11477 result = self.rpc.call_blockdev_find(node, dev)
11481 result.Raise("Can't compute disk status for %s" % instance_name)
11483 status = result.payload
11487 return (status.dev_path, status.major, status.minor,
11488 status.sync_percent, status.estimated_time,
11489 status.is_degraded, status.ldisk_status)
11491 def _ComputeDiskStatus(self, instance, snode, dev):
11492 """Compute block device status.
11495 if dev.dev_type in constants.LDS_DRBD:
11496 # we change the snode then (otherwise we use the one passed in)
11497 if dev.logical_id[0] == instance.primary_node:
11498 snode = dev.logical_id[1]
11500 snode = dev.logical_id[0]
11502 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
11503 instance.name, dev)
11504 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
11507 dev_children = map(compat.partial(self._ComputeDiskStatus,
11514 "iv_name": dev.iv_name,
11515 "dev_type": dev.dev_type,
11516 "logical_id": dev.logical_id,
11517 "physical_id": dev.physical_id,
11518 "pstatus": dev_pstatus,
11519 "sstatus": dev_sstatus,
11520 "children": dev_children,
11525 def Exec(self, feedback_fn):
11526 """Gather and return data"""
11529 cluster = self.cfg.GetClusterInfo()
11531 pri_nodes = self.cfg.GetMultiNodeInfo(i.primary_node
11532 for i in self.wanted_instances)
11533 for instance, (_, pnode) in zip(self.wanted_instances, pri_nodes):
11534 if self.op.static or pnode.offline:
11535 remote_state = None
11537 self.LogWarning("Primary node %s is marked offline, returning static"
11538 " information only for instance %s" %
11539 (pnode.name, instance.name))
11541 remote_info = self.rpc.call_instance_info(instance.primary_node,
11543 instance.hypervisor)
11544 remote_info.Raise("Error checking node %s" % instance.primary_node)
11545 remote_info = remote_info.payload
11546 if remote_info and "state" in remote_info:
11547 remote_state = "up"
11549 if instance.admin_state == constants.ADMINST_UP:
11550 remote_state = "down"
11552 remote_state = instance.admin_state
11554 disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
11557 result[instance.name] = {
11558 "name": instance.name,
11559 "config_state": instance.admin_state,
11560 "run_state": remote_state,
11561 "pnode": instance.primary_node,
11562 "snodes": instance.secondary_nodes,
11564 # this happens to be the same format used for hooks
11565 "nics": _NICListToTuple(self, instance.nics),
11566 "disk_template": instance.disk_template,
11568 "hypervisor": instance.hypervisor,
11569 "network_port": instance.network_port,
11570 "hv_instance": instance.hvparams,
11571 "hv_actual": cluster.FillHV(instance, skip_globals=True),
11572 "be_instance": instance.beparams,
11573 "be_actual": cluster.FillBE(instance),
11574 "os_instance": instance.osparams,
11575 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
11576 "serial_no": instance.serial_no,
11577 "mtime": instance.mtime,
11578 "ctime": instance.ctime,
11579 "uuid": instance.uuid,
11585 class LUInstanceSetParams(LogicalUnit):
11586 """Modifies an instances's parameters.
11589 HPATH = "instance-modify"
11590 HTYPE = constants.HTYPE_INSTANCE
11593 def CheckArguments(self):
11594 if not (self.op.nics or self.op.disks or self.op.disk_template or
11595 self.op.hvparams or self.op.beparams or self.op.os_name or
11596 self.op.online_inst or self.op.offline_inst or
11597 self.op.runtime_mem):
11598 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
11600 if self.op.hvparams:
11601 _CheckGlobalHvParams(self.op.hvparams)
11605 for disk_op, disk_dict in self.op.disks:
11606 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
11607 if disk_op == constants.DDM_REMOVE:
11608 disk_addremove += 1
11610 elif disk_op == constants.DDM_ADD:
11611 disk_addremove += 1
11613 if not isinstance(disk_op, int):
11614 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
11615 if not isinstance(disk_dict, dict):
11616 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
11617 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
11619 if disk_op == constants.DDM_ADD:
11620 mode = disk_dict.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
11621 if mode not in constants.DISK_ACCESS_SET:
11622 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
11623 errors.ECODE_INVAL)
11624 size = disk_dict.get(constants.IDISK_SIZE, None)
11626 raise errors.OpPrereqError("Required disk parameter size missing",
11627 errors.ECODE_INVAL)
11630 except (TypeError, ValueError), err:
11631 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
11632 str(err), errors.ECODE_INVAL)
11633 disk_dict[constants.IDISK_SIZE] = size
11635 # modification of disk
11636 if constants.IDISK_SIZE in disk_dict:
11637 raise errors.OpPrereqError("Disk size change not possible, use"
11638 " grow-disk", errors.ECODE_INVAL)
11640 if disk_addremove > 1:
11641 raise errors.OpPrereqError("Only one disk add or remove operation"
11642 " supported at a time", errors.ECODE_INVAL)
11644 if self.op.disks and self.op.disk_template is not None:
11645 raise errors.OpPrereqError("Disk template conversion and other disk"
11646 " changes not supported at the same time",
11647 errors.ECODE_INVAL)
11649 if (self.op.disk_template and
11650 self.op.disk_template in constants.DTS_INT_MIRROR and
11651 self.op.remote_node is None):
11652 raise errors.OpPrereqError("Changing the disk template to a mirrored"
11653 " one requires specifying a secondary node",
11654 errors.ECODE_INVAL)
11658 for nic_op, nic_dict in self.op.nics:
11659 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
11660 if nic_op == constants.DDM_REMOVE:
11663 elif nic_op == constants.DDM_ADD:
11666 if not isinstance(nic_op, int):
11667 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
11668 if not isinstance(nic_dict, dict):
11669 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
11670 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
11672 # nic_dict should be a dict
11673 nic_ip = nic_dict.get(constants.INIC_IP, None)
11674 if nic_ip is not None:
11675 if nic_ip.lower() == constants.VALUE_NONE:
11676 nic_dict[constants.INIC_IP] = None
11678 if not netutils.IPAddress.IsValid(nic_ip):
11679 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
11680 errors.ECODE_INVAL)
11682 nic_bridge = nic_dict.get("bridge", None)
11683 nic_link = nic_dict.get(constants.INIC_LINK, None)
11684 if nic_bridge and nic_link:
11685 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
11686 " at the same time", errors.ECODE_INVAL)
11687 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
11688 nic_dict["bridge"] = None
11689 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
11690 nic_dict[constants.INIC_LINK] = None
11692 if nic_op == constants.DDM_ADD:
11693 nic_mac = nic_dict.get(constants.INIC_MAC, None)
11694 if nic_mac is None:
11695 nic_dict[constants.INIC_MAC] = constants.VALUE_AUTO
11697 if constants.INIC_MAC in nic_dict:
11698 nic_mac = nic_dict[constants.INIC_MAC]
11699 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
11700 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
11702 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
11703 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
11704 " modifying an existing nic",
11705 errors.ECODE_INVAL)
11707 if nic_addremove > 1:
11708 raise errors.OpPrereqError("Only one NIC add or remove operation"
11709 " supported at a time", errors.ECODE_INVAL)
11711 def ExpandNames(self):
11712 self._ExpandAndLockInstance()
11713 # Can't even acquire node locks in shared mode as upcoming changes in
11714 # Ganeti 2.6 will start to modify the node object on disk conversion
11715 self.needed_locks[locking.LEVEL_NODE] = []
11716 self.needed_locks[locking.LEVEL_NODE_RES] = []
11717 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11719 def DeclareLocks(self, level):
11720 if level == locking.LEVEL_NODE:
11721 self._LockInstancesNodes()
11722 if self.op.disk_template and self.op.remote_node:
11723 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
11724 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
11725 elif level == locking.LEVEL_NODE_RES and self.op.disk_template:
11727 self.needed_locks[locking.LEVEL_NODE_RES] = \
11728 self.needed_locks[locking.LEVEL_NODE][:]
11730 def BuildHooksEnv(self):
11731 """Build hooks env.
11733 This runs on the master, primary and secondaries.
11737 if constants.BE_MINMEM in self.be_new:
11738 args["minmem"] = self.be_new[constants.BE_MINMEM]
11739 if constants.BE_MAXMEM in self.be_new:
11740 args["maxmem"] = self.be_new[constants.BE_MAXMEM]
11741 if constants.BE_VCPUS in self.be_new:
11742 args["vcpus"] = self.be_new[constants.BE_VCPUS]
11743 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
11744 # information at all.
11747 nic_override = dict(self.op.nics)
11748 for idx, nic in enumerate(self.instance.nics):
11749 if idx in nic_override:
11750 this_nic_override = nic_override[idx]
11752 this_nic_override = {}
11753 if constants.INIC_IP in this_nic_override:
11754 ip = this_nic_override[constants.INIC_IP]
11757 if constants.INIC_MAC in this_nic_override:
11758 mac = this_nic_override[constants.INIC_MAC]
11761 if idx in self.nic_pnew:
11762 nicparams = self.nic_pnew[idx]
11764 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
11765 mode = nicparams[constants.NIC_MODE]
11766 link = nicparams[constants.NIC_LINK]
11767 args["nics"].append((ip, mac, mode, link))
11768 if constants.DDM_ADD in nic_override:
11769 ip = nic_override[constants.DDM_ADD].get(constants.INIC_IP, None)
11770 mac = nic_override[constants.DDM_ADD][constants.INIC_MAC]
11771 nicparams = self.nic_pnew[constants.DDM_ADD]
11772 mode = nicparams[constants.NIC_MODE]
11773 link = nicparams[constants.NIC_LINK]
11774 args["nics"].append((ip, mac, mode, link))
11775 elif constants.DDM_REMOVE in nic_override:
11776 del args["nics"][-1]
11778 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
11779 if self.op.disk_template:
11780 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
11781 if self.op.runtime_mem:
11782 env["RUNTIME_MEMORY"] = self.op.runtime_mem
11786 def BuildHooksNodes(self):
11787 """Build hooks nodes.
11790 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
11793 def CheckPrereq(self):
11794 """Check prerequisites.
11796 This only checks the instance list against the existing names.
11799 # checking the new params on the primary/secondary nodes
11801 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
11802 cluster = self.cluster = self.cfg.GetClusterInfo()
11803 assert self.instance is not None, \
11804 "Cannot retrieve locked instance %s" % self.op.instance_name
11805 pnode = instance.primary_node
11806 nodelist = list(instance.all_nodes)
11807 pnode_info = self.cfg.GetNodeInfo(pnode)
11808 self.diskparams = self.cfg.GetNodeGroup(pnode_info.group).diskparams
11811 if self.op.os_name and not self.op.force:
11812 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
11813 self.op.force_variant)
11814 instance_os = self.op.os_name
11816 instance_os = instance.os
11818 if self.op.disk_template:
11819 if instance.disk_template == self.op.disk_template:
11820 raise errors.OpPrereqError("Instance already has disk template %s" %
11821 instance.disk_template, errors.ECODE_INVAL)
11823 if (instance.disk_template,
11824 self.op.disk_template) not in self._DISK_CONVERSIONS:
11825 raise errors.OpPrereqError("Unsupported disk template conversion from"
11826 " %s to %s" % (instance.disk_template,
11827 self.op.disk_template),
11828 errors.ECODE_INVAL)
11829 _CheckInstanceState(self, instance, INSTANCE_DOWN,
11830 msg="cannot change disk template")
11831 if self.op.disk_template in constants.DTS_INT_MIRROR:
11832 if self.op.remote_node == pnode:
11833 raise errors.OpPrereqError("Given new secondary node %s is the same"
11834 " as the primary node of the instance" %
11835 self.op.remote_node, errors.ECODE_STATE)
11836 _CheckNodeOnline(self, self.op.remote_node)
11837 _CheckNodeNotDrained(self, self.op.remote_node)
11838 # FIXME: here we assume that the old instance type is DT_PLAIN
11839 assert instance.disk_template == constants.DT_PLAIN
11840 disks = [{constants.IDISK_SIZE: d.size,
11841 constants.IDISK_VG: d.logical_id[0]}
11842 for d in instance.disks]
11843 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
11844 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
11846 snode_info = self.cfg.GetNodeInfo(self.op.remote_node)
11847 snode_group = self.cfg.GetNodeGroup(snode_info.group)
11848 ipolicy = _CalculateGroupIPolicy(cluster, snode_group)
11849 _CheckTargetNodeIPolicy(self, ipolicy, instance, snode_info,
11850 ignore=self.op.ignore_ipolicy)
11851 if pnode_info.group != snode_info.group:
11852 self.LogWarning("The primary and secondary nodes are in two"
11853 " different node groups; the disk parameters"
11854 " from the first disk's node group will be"
11857 # hvparams processing
11858 if self.op.hvparams:
11859 hv_type = instance.hypervisor
11860 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
11861 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
11862 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
11865 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
11866 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
11867 self.hv_proposed = self.hv_new = hv_new # the new actual values
11868 self.hv_inst = i_hvdict # the new dict (without defaults)
11870 self.hv_proposed = cluster.SimpleFillHV(instance.hypervisor, instance.os,
11872 self.hv_new = self.hv_inst = {}
11874 # beparams processing
11875 if self.op.beparams:
11876 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
11878 objects.UpgradeBeParams(i_bedict)
11879 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
11880 be_new = cluster.SimpleFillBE(i_bedict)
11881 self.be_proposed = self.be_new = be_new # the new actual values
11882 self.be_inst = i_bedict # the new dict (without defaults)
11884 self.be_new = self.be_inst = {}
11885 self.be_proposed = cluster.SimpleFillBE(instance.beparams)
11886 be_old = cluster.FillBE(instance)
11888 # CPU param validation -- checking every time a paramtere is
11889 # changed to cover all cases where either CPU mask or vcpus have
11891 if (constants.BE_VCPUS in self.be_proposed and
11892 constants.HV_CPU_MASK in self.hv_proposed):
11894 utils.ParseMultiCpuMask(self.hv_proposed[constants.HV_CPU_MASK])
11895 # Verify mask is consistent with number of vCPUs. Can skip this
11896 # test if only 1 entry in the CPU mask, which means same mask
11897 # is applied to all vCPUs.
11898 if (len(cpu_list) > 1 and
11899 len(cpu_list) != self.be_proposed[constants.BE_VCPUS]):
11900 raise errors.OpPrereqError("Number of vCPUs [%d] does not match the"
11902 (self.be_proposed[constants.BE_VCPUS],
11903 self.hv_proposed[constants.HV_CPU_MASK]),
11904 errors.ECODE_INVAL)
11906 # Only perform this test if a new CPU mask is given
11907 if constants.HV_CPU_MASK in self.hv_new:
11908 # Calculate the largest CPU number requested
11909 max_requested_cpu = max(map(max, cpu_list))
11910 # Check that all of the instance's nodes have enough physical CPUs to
11911 # satisfy the requested CPU mask
11912 _CheckNodesPhysicalCPUs(self, instance.all_nodes,
11913 max_requested_cpu + 1, instance.hypervisor)
11915 # osparams processing
11916 if self.op.osparams:
11917 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
11918 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
11919 self.os_inst = i_osdict # the new dict (without defaults)
11925 #TODO(dynmem): do the appropriate check involving MINMEM
11926 if (constants.BE_MAXMEM in self.op.beparams and not self.op.force and
11927 be_new[constants.BE_MAXMEM] > be_old[constants.BE_MAXMEM]):
11928 mem_check_list = [pnode]
11929 if be_new[constants.BE_AUTO_BALANCE]:
11930 # either we changed auto_balance to yes or it was from before
11931 mem_check_list.extend(instance.secondary_nodes)
11932 instance_info = self.rpc.call_instance_info(pnode, instance.name,
11933 instance.hypervisor)
11934 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
11935 [instance.hypervisor])
11936 pninfo = nodeinfo[pnode]
11937 msg = pninfo.fail_msg
11939 # Assume the primary node is unreachable and go ahead
11940 self.warn.append("Can't get info from primary node %s: %s" %
11943 (_, _, (pnhvinfo, )) = pninfo.payload
11944 if not isinstance(pnhvinfo.get("memory_free", None), int):
11945 self.warn.append("Node data from primary node %s doesn't contain"
11946 " free memory information" % pnode)
11947 elif instance_info.fail_msg:
11948 self.warn.append("Can't get instance runtime information: %s" %
11949 instance_info.fail_msg)
11951 if instance_info.payload:
11952 current_mem = int(instance_info.payload["memory"])
11954 # Assume instance not running
11955 # (there is a slight race condition here, but it's not very
11956 # probable, and we have no other way to check)
11957 # TODO: Describe race condition
11959 #TODO(dynmem): do the appropriate check involving MINMEM
11960 miss_mem = (be_new[constants.BE_MAXMEM] - current_mem -
11961 pnhvinfo["memory_free"])
11963 raise errors.OpPrereqError("This change will prevent the instance"
11964 " from starting, due to %d MB of memory"
11965 " missing on its primary node" %
11967 errors.ECODE_NORES)
11969 if be_new[constants.BE_AUTO_BALANCE]:
11970 for node, nres in nodeinfo.items():
11971 if node not in instance.secondary_nodes:
11973 nres.Raise("Can't get info from secondary node %s" % node,
11974 prereq=True, ecode=errors.ECODE_STATE)
11975 (_, _, (nhvinfo, )) = nres.payload
11976 if not isinstance(nhvinfo.get("memory_free", None), int):
11977 raise errors.OpPrereqError("Secondary node %s didn't return free"
11978 " memory information" % node,
11979 errors.ECODE_STATE)
11980 #TODO(dynmem): do the appropriate check involving MINMEM
11981 elif be_new[constants.BE_MAXMEM] > nhvinfo["memory_free"]:
11982 raise errors.OpPrereqError("This change will prevent the instance"
11983 " from failover to its secondary node"
11984 " %s, due to not enough memory" % node,
11985 errors.ECODE_STATE)
11987 if self.op.runtime_mem:
11988 remote_info = self.rpc.call_instance_info(instance.primary_node,
11990 instance.hypervisor)
11991 remote_info.Raise("Error checking node %s" % instance.primary_node)
11992 if not remote_info.payload: # not running already
11993 raise errors.OpPrereqError("Instance %s is not running" % instance.name,
11994 errors.ECODE_STATE)
11996 current_memory = remote_info.payload["memory"]
11997 if (not self.op.force and
11998 (self.op.runtime_mem > self.be_proposed[constants.BE_MAXMEM] or
11999 self.op.runtime_mem < self.be_proposed[constants.BE_MINMEM])):
12000 raise errors.OpPrereqError("Instance %s must have memory between %d"
12001 " and %d MB of memory unless --force is"
12002 " given" % (instance.name,
12003 self.be_proposed[constants.BE_MINMEM],
12004 self.be_proposed[constants.BE_MAXMEM]),
12005 errors.ECODE_INVAL)
12007 if self.op.runtime_mem > current_memory:
12008 _CheckNodeFreeMemory(self, instance.primary_node,
12009 "ballooning memory for instance %s" %
12011 self.op.memory - current_memory,
12012 instance.hypervisor)
12016 self.nic_pinst = {}
12017 for nic_op, nic_dict in self.op.nics:
12018 if nic_op == constants.DDM_REMOVE:
12019 if not instance.nics:
12020 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
12021 errors.ECODE_INVAL)
12023 if nic_op != constants.DDM_ADD:
12025 if not instance.nics:
12026 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
12027 " no NICs" % nic_op,
12028 errors.ECODE_INVAL)
12029 if nic_op < 0 or nic_op >= len(instance.nics):
12030 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
12032 (nic_op, len(instance.nics) - 1),
12033 errors.ECODE_INVAL)
12034 old_nic_params = instance.nics[nic_op].nicparams
12035 old_nic_ip = instance.nics[nic_op].ip
12037 old_nic_params = {}
12040 update_params_dict = dict([(key, nic_dict[key])
12041 for key in constants.NICS_PARAMETERS
12042 if key in nic_dict])
12044 if "bridge" in nic_dict:
12045 update_params_dict[constants.NIC_LINK] = nic_dict["bridge"]
12047 new_nic_params = _GetUpdatedParams(old_nic_params,
12048 update_params_dict)
12049 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
12050 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
12051 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
12052 self.nic_pinst[nic_op] = new_nic_params
12053 self.nic_pnew[nic_op] = new_filled_nic_params
12054 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
12056 if new_nic_mode == constants.NIC_MODE_BRIDGED:
12057 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
12058 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
12060 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
12062 self.warn.append(msg)
12064 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
12065 if new_nic_mode == constants.NIC_MODE_ROUTED:
12066 if constants.INIC_IP in nic_dict:
12067 nic_ip = nic_dict[constants.INIC_IP]
12069 nic_ip = old_nic_ip
12071 raise errors.OpPrereqError("Cannot set the nic ip to None"
12072 " on a routed nic", errors.ECODE_INVAL)
12073 if constants.INIC_MAC in nic_dict:
12074 nic_mac = nic_dict[constants.INIC_MAC]
12075 if nic_mac is None:
12076 raise errors.OpPrereqError("Cannot set the nic mac to None",
12077 errors.ECODE_INVAL)
12078 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
12079 # otherwise generate the mac
12080 nic_dict[constants.INIC_MAC] = \
12081 self.cfg.GenerateMAC(self.proc.GetECId())
12083 # or validate/reserve the current one
12085 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
12086 except errors.ReservationError:
12087 raise errors.OpPrereqError("MAC address %s already in use"
12088 " in cluster" % nic_mac,
12089 errors.ECODE_NOTUNIQUE)
12092 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
12093 raise errors.OpPrereqError("Disk operations not supported for"
12094 " diskless instances",
12095 errors.ECODE_INVAL)
12096 for disk_op, _ in self.op.disks:
12097 if disk_op == constants.DDM_REMOVE:
12098 if len(instance.disks) == 1:
12099 raise errors.OpPrereqError("Cannot remove the last disk of"
12100 " an instance", errors.ECODE_INVAL)
12101 _CheckInstanceState(self, instance, INSTANCE_DOWN,
12102 msg="cannot remove disks")
12104 if (disk_op == constants.DDM_ADD and
12105 len(instance.disks) >= constants.MAX_DISKS):
12106 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
12107 " add more" % constants.MAX_DISKS,
12108 errors.ECODE_STATE)
12109 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
12111 if disk_op < 0 or disk_op >= len(instance.disks):
12112 raise errors.OpPrereqError("Invalid disk index %s, valid values"
12114 (disk_op, len(instance.disks)),
12115 errors.ECODE_INVAL)
12117 # disabling the instance
12118 if self.op.offline_inst:
12119 _CheckInstanceState(self, instance, INSTANCE_DOWN,
12120 msg="cannot change instance state to offline")
12122 # enabling the instance
12123 if self.op.online_inst:
12124 _CheckInstanceState(self, instance, INSTANCE_OFFLINE,
12125 msg="cannot make instance go online")
12127 def _ConvertPlainToDrbd(self, feedback_fn):
12128 """Converts an instance from plain to drbd.
12131 feedback_fn("Converting template to drbd")
12132 instance = self.instance
12133 pnode = instance.primary_node
12134 snode = self.op.remote_node
12136 assert instance.disk_template == constants.DT_PLAIN
12138 # create a fake disk info for _GenerateDiskTemplate
12139 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
12140 constants.IDISK_VG: d.logical_id[0]}
12141 for d in instance.disks]
12142 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
12143 instance.name, pnode, [snode],
12144 disk_info, None, None, 0, feedback_fn,
12146 info = _GetInstanceInfoText(instance)
12147 feedback_fn("Creating aditional volumes...")
12148 # first, create the missing data and meta devices
12149 for disk in new_disks:
12150 # unfortunately this is... not too nice
12151 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
12153 for child in disk.children:
12154 _CreateSingleBlockDev(self, snode, instance, child, info, True)
12155 # at this stage, all new LVs have been created, we can rename the
12157 feedback_fn("Renaming original volumes...")
12158 rename_list = [(o, n.children[0].logical_id)
12159 for (o, n) in zip(instance.disks, new_disks)]
12160 result = self.rpc.call_blockdev_rename(pnode, rename_list)
12161 result.Raise("Failed to rename original LVs")
12163 feedback_fn("Initializing DRBD devices...")
12164 # all child devices are in place, we can now create the DRBD devices
12165 for disk in new_disks:
12166 for node in [pnode, snode]:
12167 f_create = node == pnode
12168 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
12170 # at this point, the instance has been modified
12171 instance.disk_template = constants.DT_DRBD8
12172 instance.disks = new_disks
12173 self.cfg.Update(instance, feedback_fn)
12175 # Release node locks while waiting for sync
12176 _ReleaseLocks(self, locking.LEVEL_NODE)
12178 # disks are created, waiting for sync
12179 disk_abort = not _WaitForSync(self, instance,
12180 oneshot=not self.op.wait_for_sync)
12182 raise errors.OpExecError("There are some degraded disks for"
12183 " this instance, please cleanup manually")
12185 # Node resource locks will be released by caller
12187 def _ConvertDrbdToPlain(self, feedback_fn):
12188 """Converts an instance from drbd to plain.
12191 instance = self.instance
12193 assert len(instance.secondary_nodes) == 1
12194 assert instance.disk_template == constants.DT_DRBD8
12196 pnode = instance.primary_node
12197 snode = instance.secondary_nodes[0]
12198 feedback_fn("Converting template to plain")
12200 old_disks = instance.disks
12201 new_disks = [d.children[0] for d in old_disks]
12203 # copy over size and mode
12204 for parent, child in zip(old_disks, new_disks):
12205 child.size = parent.size
12206 child.mode = parent.mode
12208 # update instance structure
12209 instance.disks = new_disks
12210 instance.disk_template = constants.DT_PLAIN
12211 self.cfg.Update(instance, feedback_fn)
12213 # Release locks in case removing disks takes a while
12214 _ReleaseLocks(self, locking.LEVEL_NODE)
12216 feedback_fn("Removing volumes on the secondary node...")
12217 for disk in old_disks:
12218 self.cfg.SetDiskID(disk, snode)
12219 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
12221 self.LogWarning("Could not remove block device %s on node %s,"
12222 " continuing anyway: %s", disk.iv_name, snode, msg)
12224 feedback_fn("Removing unneeded volumes on the primary node...")
12225 for idx, disk in enumerate(old_disks):
12226 meta = disk.children[1]
12227 self.cfg.SetDiskID(meta, pnode)
12228 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
12230 self.LogWarning("Could not remove metadata for disk %d on node %s,"
12231 " continuing anyway: %s", idx, pnode, msg)
12233 # this is a DRBD disk, return its port to the pool
12234 for disk in old_disks:
12235 tcp_port = disk.logical_id[2]
12236 self.cfg.AddTcpUdpPort(tcp_port)
12238 # Node resource locks will be released by caller
12240 def Exec(self, feedback_fn):
12241 """Modifies an instance.
12243 All parameters take effect only at the next restart of the instance.
12246 # Process here the warnings from CheckPrereq, as we don't have a
12247 # feedback_fn there.
12248 for warn in self.warn:
12249 feedback_fn("WARNING: %s" % warn)
12251 assert ((self.op.disk_template is None) ^
12252 bool(self.owned_locks(locking.LEVEL_NODE_RES))), \
12253 "Not owning any node resource locks"
12256 instance = self.instance
12259 if self.op.runtime_mem:
12260 rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
12262 self.op.runtime_mem)
12263 rpcres.Raise("Cannot modify instance runtime memory")
12264 result.append(("runtime_memory", self.op.runtime_mem))
12267 for disk_op, disk_dict in self.op.disks:
12268 if disk_op == constants.DDM_REMOVE:
12269 # remove the last disk
12270 device = instance.disks.pop()
12271 device_idx = len(instance.disks)
12272 for node, disk in device.ComputeNodeTree(instance.primary_node):
12273 self.cfg.SetDiskID(disk, node)
12274 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
12276 self.LogWarning("Could not remove disk/%d on node %s: %s,"
12277 " continuing anyway", device_idx, node, msg)
12278 result.append(("disk/%d" % device_idx, "remove"))
12280 # if this is a DRBD disk, return its port to the pool
12281 if device.dev_type in constants.LDS_DRBD:
12282 tcp_port = device.logical_id[2]
12283 self.cfg.AddTcpUdpPort(tcp_port)
12284 elif disk_op == constants.DDM_ADD:
12286 if instance.disk_template in (constants.DT_FILE,
12287 constants.DT_SHARED_FILE):
12288 file_driver, file_path = instance.disks[0].logical_id
12289 file_path = os.path.dirname(file_path)
12291 file_driver = file_path = None
12292 disk_idx_base = len(instance.disks)
12293 new_disk = _GenerateDiskTemplate(self,
12294 instance.disk_template,
12295 instance.name, instance.primary_node,
12296 instance.secondary_nodes,
12302 self.diskparams)[0]
12303 instance.disks.append(new_disk)
12304 info = _GetInstanceInfoText(instance)
12306 logging.info("Creating volume %s for instance %s",
12307 new_disk.iv_name, instance.name)
12308 # Note: this needs to be kept in sync with _CreateDisks
12310 for node in instance.all_nodes:
12311 f_create = node == instance.primary_node
12313 _CreateBlockDev(self, node, instance, new_disk,
12314 f_create, info, f_create)
12315 except errors.OpExecError, err:
12316 self.LogWarning("Failed to create volume %s (%s) on"
12318 new_disk.iv_name, new_disk, node, err)
12319 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
12320 (new_disk.size, new_disk.mode)))
12322 # change a given disk
12323 instance.disks[disk_op].mode = disk_dict[constants.IDISK_MODE]
12324 result.append(("disk.mode/%d" % disk_op,
12325 disk_dict[constants.IDISK_MODE]))
12327 if self.op.disk_template:
12329 check_nodes = set(instance.all_nodes)
12330 if self.op.remote_node:
12331 check_nodes.add(self.op.remote_node)
12332 for level in [locking.LEVEL_NODE, locking.LEVEL_NODE_RES]:
12333 owned = self.owned_locks(level)
12334 assert not (check_nodes - owned), \
12335 ("Not owning the correct locks, owning %r, expected at least %r" %
12336 (owned, check_nodes))
12338 r_shut = _ShutdownInstanceDisks(self, instance)
12340 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
12341 " proceed with disk template conversion")
12342 mode = (instance.disk_template, self.op.disk_template)
12344 self._DISK_CONVERSIONS[mode](self, feedback_fn)
12346 self.cfg.ReleaseDRBDMinors(instance.name)
12348 result.append(("disk_template", self.op.disk_template))
12350 assert instance.disk_template == self.op.disk_template, \
12351 ("Expected disk template '%s', found '%s'" %
12352 (self.op.disk_template, instance.disk_template))
12354 # Release node and resource locks if there are any (they might already have
12355 # been released during disk conversion)
12356 _ReleaseLocks(self, locking.LEVEL_NODE)
12357 _ReleaseLocks(self, locking.LEVEL_NODE_RES)
12360 for nic_op, nic_dict in self.op.nics:
12361 if nic_op == constants.DDM_REMOVE:
12362 # remove the last nic
12363 del instance.nics[-1]
12364 result.append(("nic.%d" % len(instance.nics), "remove"))
12365 elif nic_op == constants.DDM_ADD:
12366 # mac and bridge should be set, by now
12367 mac = nic_dict[constants.INIC_MAC]
12368 ip = nic_dict.get(constants.INIC_IP, None)
12369 nicparams = self.nic_pinst[constants.DDM_ADD]
12370 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
12371 instance.nics.append(new_nic)
12372 result.append(("nic.%d" % (len(instance.nics) - 1),
12373 "add:mac=%s,ip=%s,mode=%s,link=%s" %
12374 (new_nic.mac, new_nic.ip,
12375 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
12376 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
12379 for key in (constants.INIC_MAC, constants.INIC_IP):
12380 if key in nic_dict:
12381 setattr(instance.nics[nic_op], key, nic_dict[key])
12382 if nic_op in self.nic_pinst:
12383 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
12384 for key, val in nic_dict.iteritems():
12385 result.append(("nic.%s/%d" % (key, nic_op), val))
12388 if self.op.hvparams:
12389 instance.hvparams = self.hv_inst
12390 for key, val in self.op.hvparams.iteritems():
12391 result.append(("hv/%s" % key, val))
12394 if self.op.beparams:
12395 instance.beparams = self.be_inst
12396 for key, val in self.op.beparams.iteritems():
12397 result.append(("be/%s" % key, val))
12400 if self.op.os_name:
12401 instance.os = self.op.os_name
12404 if self.op.osparams:
12405 instance.osparams = self.os_inst
12406 for key, val in self.op.osparams.iteritems():
12407 result.append(("os/%s" % key, val))
12409 # online/offline instance
12410 if self.op.online_inst:
12411 self.cfg.MarkInstanceDown(instance.name)
12412 result.append(("admin_state", constants.ADMINST_DOWN))
12413 if self.op.offline_inst:
12414 self.cfg.MarkInstanceOffline(instance.name)
12415 result.append(("admin_state", constants.ADMINST_OFFLINE))
12417 self.cfg.Update(instance, feedback_fn)
12419 assert not (self.owned_locks(locking.LEVEL_NODE_RES) or
12420 self.owned_locks(locking.LEVEL_NODE)), \
12421 "All node locks should have been released by now"
12425 _DISK_CONVERSIONS = {
12426 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
12427 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
12431 class LUInstanceChangeGroup(LogicalUnit):
12432 HPATH = "instance-change-group"
12433 HTYPE = constants.HTYPE_INSTANCE
12436 def ExpandNames(self):
12437 self.share_locks = _ShareAll()
12438 self.needed_locks = {
12439 locking.LEVEL_NODEGROUP: [],
12440 locking.LEVEL_NODE: [],
12443 self._ExpandAndLockInstance()
12445 if self.op.target_groups:
12446 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
12447 self.op.target_groups)
12449 self.req_target_uuids = None
12451 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
12453 def DeclareLocks(self, level):
12454 if level == locking.LEVEL_NODEGROUP:
12455 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
12457 if self.req_target_uuids:
12458 lock_groups = set(self.req_target_uuids)
12460 # Lock all groups used by instance optimistically; this requires going
12461 # via the node before it's locked, requiring verification later on
12462 instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
12463 lock_groups.update(instance_groups)
12465 # No target groups, need to lock all of them
12466 lock_groups = locking.ALL_SET
12468 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
12470 elif level == locking.LEVEL_NODE:
12471 if self.req_target_uuids:
12472 # Lock all nodes used by instances
12473 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
12474 self._LockInstancesNodes()
12476 # Lock all nodes in all potential target groups
12477 lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) -
12478 self.cfg.GetInstanceNodeGroups(self.op.instance_name))
12479 member_nodes = [node_name
12480 for group in lock_groups
12481 for node_name in self.cfg.GetNodeGroup(group).members]
12482 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
12484 # Lock all nodes as all groups are potential targets
12485 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12487 def CheckPrereq(self):
12488 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
12489 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12490 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
12492 assert (self.req_target_uuids is None or
12493 owned_groups.issuperset(self.req_target_uuids))
12494 assert owned_instances == set([self.op.instance_name])
12496 # Get instance information
12497 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
12499 # Check if node groups for locked instance are still correct
12500 assert owned_nodes.issuperset(self.instance.all_nodes), \
12501 ("Instance %s's nodes changed while we kept the lock" %
12502 self.op.instance_name)
12504 inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name,
12507 if self.req_target_uuids:
12508 # User requested specific target groups
12509 self.target_uuids = self.req_target_uuids
12511 # All groups except those used by the instance are potential targets
12512 self.target_uuids = owned_groups - inst_groups
12514 conflicting_groups = self.target_uuids & inst_groups
12515 if conflicting_groups:
12516 raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are"
12517 " used by the instance '%s'" %
12518 (utils.CommaJoin(conflicting_groups),
12519 self.op.instance_name),
12520 errors.ECODE_INVAL)
12522 if not self.target_uuids:
12523 raise errors.OpPrereqError("There are no possible target groups",
12524 errors.ECODE_INVAL)
12526 def BuildHooksEnv(self):
12527 """Build hooks env.
12530 assert self.target_uuids
12533 "TARGET_GROUPS": " ".join(self.target_uuids),
12536 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
12540 def BuildHooksNodes(self):
12541 """Build hooks nodes.
12544 mn = self.cfg.GetMasterNode()
12545 return ([mn], [mn])
12547 def Exec(self, feedback_fn):
12548 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
12550 assert instances == [self.op.instance_name], "Instance not locked"
12552 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
12553 instances=instances, target_groups=list(self.target_uuids))
12555 ial.Run(self.op.iallocator)
12557 if not ial.success:
12558 raise errors.OpPrereqError("Can't compute solution for changing group of"
12559 " instance '%s' using iallocator '%s': %s" %
12560 (self.op.instance_name, self.op.iallocator,
12562 errors.ECODE_NORES)
12564 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
12566 self.LogInfo("Iallocator returned %s job(s) for changing group of"
12567 " instance '%s'", len(jobs), self.op.instance_name)
12569 return ResultWithJobs(jobs)
12572 class LUBackupQuery(NoHooksLU):
12573 """Query the exports list
12578 def ExpandNames(self):
12579 self.needed_locks = {}
12580 self.share_locks[locking.LEVEL_NODE] = 1
12581 if not self.op.nodes:
12582 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12584 self.needed_locks[locking.LEVEL_NODE] = \
12585 _GetWantedNodes(self, self.op.nodes)
12587 def Exec(self, feedback_fn):
12588 """Compute the list of all the exported system images.
12591 @return: a dictionary with the structure node->(export-list)
12592 where export-list is a list of the instances exported on
12596 self.nodes = self.owned_locks(locking.LEVEL_NODE)
12597 rpcresult = self.rpc.call_export_list(self.nodes)
12599 for node in rpcresult:
12600 if rpcresult[node].fail_msg:
12601 result[node] = False
12603 result[node] = rpcresult[node].payload
12608 class LUBackupPrepare(NoHooksLU):
12609 """Prepares an instance for an export and returns useful information.
12614 def ExpandNames(self):
12615 self._ExpandAndLockInstance()
12617 def CheckPrereq(self):
12618 """Check prerequisites.
12621 instance_name = self.op.instance_name
12623 self.instance = self.cfg.GetInstanceInfo(instance_name)
12624 assert self.instance is not None, \
12625 "Cannot retrieve locked instance %s" % self.op.instance_name
12626 _CheckNodeOnline(self, self.instance.primary_node)
12628 self._cds = _GetClusterDomainSecret()
12630 def Exec(self, feedback_fn):
12631 """Prepares an instance for an export.
12634 instance = self.instance
12636 if self.op.mode == constants.EXPORT_MODE_REMOTE:
12637 salt = utils.GenerateSecret(8)
12639 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
12640 result = self.rpc.call_x509_cert_create(instance.primary_node,
12641 constants.RIE_CERT_VALIDITY)
12642 result.Raise("Can't create X509 key and certificate on %s" % result.node)
12644 (name, cert_pem) = result.payload
12646 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
12650 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
12651 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
12653 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
12659 class LUBackupExport(LogicalUnit):
12660 """Export an instance to an image in the cluster.
12663 HPATH = "instance-export"
12664 HTYPE = constants.HTYPE_INSTANCE
12667 def CheckArguments(self):
12668 """Check the arguments.
12671 self.x509_key_name = self.op.x509_key_name
12672 self.dest_x509_ca_pem = self.op.destination_x509_ca
12674 if self.op.mode == constants.EXPORT_MODE_REMOTE:
12675 if not self.x509_key_name:
12676 raise errors.OpPrereqError("Missing X509 key name for encryption",
12677 errors.ECODE_INVAL)
12679 if not self.dest_x509_ca_pem:
12680 raise errors.OpPrereqError("Missing destination X509 CA",
12681 errors.ECODE_INVAL)
12683 def ExpandNames(self):
12684 self._ExpandAndLockInstance()
12686 # Lock all nodes for local exports
12687 if self.op.mode == constants.EXPORT_MODE_LOCAL:
12688 # FIXME: lock only instance primary and destination node
12690 # Sad but true, for now we have do lock all nodes, as we don't know where
12691 # the previous export might be, and in this LU we search for it and
12692 # remove it from its current node. In the future we could fix this by:
12693 # - making a tasklet to search (share-lock all), then create the
12694 # new one, then one to remove, after
12695 # - removing the removal operation altogether
12696 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12698 def DeclareLocks(self, level):
12699 """Last minute lock declaration."""
12700 # All nodes are locked anyway, so nothing to do here.
12702 def BuildHooksEnv(self):
12703 """Build hooks env.
12705 This will run on the master, primary node and target node.
12709 "EXPORT_MODE": self.op.mode,
12710 "EXPORT_NODE": self.op.target_node,
12711 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
12712 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
12713 # TODO: Generic function for boolean env variables
12714 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
12717 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
12721 def BuildHooksNodes(self):
12722 """Build hooks nodes.
12725 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
12727 if self.op.mode == constants.EXPORT_MODE_LOCAL:
12728 nl.append(self.op.target_node)
12732 def CheckPrereq(self):
12733 """Check prerequisites.
12735 This checks that the instance and node names are valid.
12738 instance_name = self.op.instance_name
12740 self.instance = self.cfg.GetInstanceInfo(instance_name)
12741 assert self.instance is not None, \
12742 "Cannot retrieve locked instance %s" % self.op.instance_name
12743 _CheckNodeOnline(self, self.instance.primary_node)
12745 if (self.op.remove_instance and
12746 self.instance.admin_state == constants.ADMINST_UP and
12747 not self.op.shutdown):
12748 raise errors.OpPrereqError("Can not remove instance without shutting it"
12751 if self.op.mode == constants.EXPORT_MODE_LOCAL:
12752 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
12753 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
12754 assert self.dst_node is not None
12756 _CheckNodeOnline(self, self.dst_node.name)
12757 _CheckNodeNotDrained(self, self.dst_node.name)
12760 self.dest_disk_info = None
12761 self.dest_x509_ca = None
12763 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
12764 self.dst_node = None
12766 if len(self.op.target_node) != len(self.instance.disks):
12767 raise errors.OpPrereqError(("Received destination information for %s"
12768 " disks, but instance %s has %s disks") %
12769 (len(self.op.target_node), instance_name,
12770 len(self.instance.disks)),
12771 errors.ECODE_INVAL)
12773 cds = _GetClusterDomainSecret()
12775 # Check X509 key name
12777 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
12778 except (TypeError, ValueError), err:
12779 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
12781 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
12782 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
12783 errors.ECODE_INVAL)
12785 # Load and verify CA
12787 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
12788 except OpenSSL.crypto.Error, err:
12789 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
12790 (err, ), errors.ECODE_INVAL)
12792 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
12793 if errcode is not None:
12794 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
12795 (msg, ), errors.ECODE_INVAL)
12797 self.dest_x509_ca = cert
12799 # Verify target information
12801 for idx, disk_data in enumerate(self.op.target_node):
12803 (host, port, magic) = \
12804 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
12805 except errors.GenericError, err:
12806 raise errors.OpPrereqError("Target info for disk %s: %s" %
12807 (idx, err), errors.ECODE_INVAL)
12809 disk_info.append((host, port, magic))
12811 assert len(disk_info) == len(self.op.target_node)
12812 self.dest_disk_info = disk_info
12815 raise errors.ProgrammerError("Unhandled export mode %r" %
12818 # instance disk type verification
12819 # TODO: Implement export support for file-based disks
12820 for disk in self.instance.disks:
12821 if disk.dev_type == constants.LD_FILE:
12822 raise errors.OpPrereqError("Export not supported for instances with"
12823 " file-based disks", errors.ECODE_INVAL)
12825 def _CleanupExports(self, feedback_fn):
12826 """Removes exports of current instance from all other nodes.
12828 If an instance in a cluster with nodes A..D was exported to node C, its
12829 exports will be removed from the nodes A, B and D.
12832 assert self.op.mode != constants.EXPORT_MODE_REMOTE
12834 nodelist = self.cfg.GetNodeList()
12835 nodelist.remove(self.dst_node.name)
12837 # on one-node clusters nodelist will be empty after the removal
12838 # if we proceed the backup would be removed because OpBackupQuery
12839 # substitutes an empty list with the full cluster node list.
12840 iname = self.instance.name
12842 feedback_fn("Removing old exports for instance %s" % iname)
12843 exportlist = self.rpc.call_export_list(nodelist)
12844 for node in exportlist:
12845 if exportlist[node].fail_msg:
12847 if iname in exportlist[node].payload:
12848 msg = self.rpc.call_export_remove(node, iname).fail_msg
12850 self.LogWarning("Could not remove older export for instance %s"
12851 " on node %s: %s", iname, node, msg)
12853 def Exec(self, feedback_fn):
12854 """Export an instance to an image in the cluster.
12857 assert self.op.mode in constants.EXPORT_MODES
12859 instance = self.instance
12860 src_node = instance.primary_node
12862 if self.op.shutdown:
12863 # shutdown the instance, but not the disks
12864 feedback_fn("Shutting down instance %s" % instance.name)
12865 result = self.rpc.call_instance_shutdown(src_node, instance,
12866 self.op.shutdown_timeout)
12867 # TODO: Maybe ignore failures if ignore_remove_failures is set
12868 result.Raise("Could not shutdown instance %s on"
12869 " node %s" % (instance.name, src_node))
12871 # set the disks ID correctly since call_instance_start needs the
12872 # correct drbd minor to create the symlinks
12873 for disk in instance.disks:
12874 self.cfg.SetDiskID(disk, src_node)
12876 activate_disks = (instance.admin_state != constants.ADMINST_UP)
12879 # Activate the instance disks if we'exporting a stopped instance
12880 feedback_fn("Activating disks for %s" % instance.name)
12881 _StartInstanceDisks(self, instance, None)
12884 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
12887 helper.CreateSnapshots()
12889 if (self.op.shutdown and
12890 instance.admin_state == constants.ADMINST_UP and
12891 not self.op.remove_instance):
12892 assert not activate_disks
12893 feedback_fn("Starting instance %s" % instance.name)
12894 result = self.rpc.call_instance_start(src_node,
12895 (instance, None, None), False)
12896 msg = result.fail_msg
12898 feedback_fn("Failed to start instance: %s" % msg)
12899 _ShutdownInstanceDisks(self, instance)
12900 raise errors.OpExecError("Could not start instance: %s" % msg)
12902 if self.op.mode == constants.EXPORT_MODE_LOCAL:
12903 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
12904 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
12905 connect_timeout = constants.RIE_CONNECT_TIMEOUT
12906 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
12908 (key_name, _, _) = self.x509_key_name
12911 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
12914 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
12915 key_name, dest_ca_pem,
12920 # Check for backwards compatibility
12921 assert len(dresults) == len(instance.disks)
12922 assert compat.all(isinstance(i, bool) for i in dresults), \
12923 "Not all results are boolean: %r" % dresults
12927 feedback_fn("Deactivating disks for %s" % instance.name)
12928 _ShutdownInstanceDisks(self, instance)
12930 if not (compat.all(dresults) and fin_resu):
12933 failures.append("export finalization")
12934 if not compat.all(dresults):
12935 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
12937 failures.append("disk export: disk(s) %s" % fdsk)
12939 raise errors.OpExecError("Export failed, errors in %s" %
12940 utils.CommaJoin(failures))
12942 # At this point, the export was successful, we can cleanup/finish
12944 # Remove instance if requested
12945 if self.op.remove_instance:
12946 feedback_fn("Removing instance %s" % instance.name)
12947 _RemoveInstance(self, feedback_fn, instance,
12948 self.op.ignore_remove_failures)
12950 if self.op.mode == constants.EXPORT_MODE_LOCAL:
12951 self._CleanupExports(feedback_fn)
12953 return fin_resu, dresults
12956 class LUBackupRemove(NoHooksLU):
12957 """Remove exports related to the named instance.
12962 def ExpandNames(self):
12963 self.needed_locks = {}
12964 # We need all nodes to be locked in order for RemoveExport to work, but we
12965 # don't need to lock the instance itself, as nothing will happen to it (and
12966 # we can remove exports also for a removed instance)
12967 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12969 def Exec(self, feedback_fn):
12970 """Remove any export.
12973 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
12974 # If the instance was not found we'll try with the name that was passed in.
12975 # This will only work if it was an FQDN, though.
12977 if not instance_name:
12979 instance_name = self.op.instance_name
12981 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
12982 exportlist = self.rpc.call_export_list(locked_nodes)
12984 for node in exportlist:
12985 msg = exportlist[node].fail_msg
12987 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
12989 if instance_name in exportlist[node].payload:
12991 result = self.rpc.call_export_remove(node, instance_name)
12992 msg = result.fail_msg
12994 logging.error("Could not remove export for instance %s"
12995 " on node %s: %s", instance_name, node, msg)
12997 if fqdn_warn and not found:
12998 feedback_fn("Export not found. If trying to remove an export belonging"
12999 " to a deleted instance please use its Fully Qualified"
13003 class LUGroupAdd(LogicalUnit):
13004 """Logical unit for creating node groups.
13007 HPATH = "group-add"
13008 HTYPE = constants.HTYPE_GROUP
13011 def ExpandNames(self):
13012 # We need the new group's UUID here so that we can create and acquire the
13013 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
13014 # that it should not check whether the UUID exists in the configuration.
13015 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
13016 self.needed_locks = {}
13017 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
13019 def CheckPrereq(self):
13020 """Check prerequisites.
13022 This checks that the given group name is not an existing node group
13027 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13028 except errors.OpPrereqError:
13031 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
13032 " node group (UUID: %s)" %
13033 (self.op.group_name, existing_uuid),
13034 errors.ECODE_EXISTS)
13036 if self.op.ndparams:
13037 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
13039 if self.op.hv_state:
13040 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None)
13042 self.new_hv_state = None
13044 if self.op.disk_state:
13045 self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None)
13047 self.new_disk_state = None
13049 if self.op.diskparams:
13050 for templ in constants.DISK_TEMPLATES:
13051 if templ not in self.op.diskparams:
13052 self.op.diskparams[templ] = {}
13053 utils.ForceDictType(self.op.diskparams[templ], constants.DISK_DT_TYPES)
13055 self.op.diskparams = self.cfg.GetClusterInfo().diskparams
13057 if self.op.ipolicy:
13058 cluster = self.cfg.GetClusterInfo()
13059 full_ipolicy = cluster.SimpleFillIPolicy(self.op.ipolicy)
13061 objects.InstancePolicy.CheckParameterSyntax(full_ipolicy)
13062 except errors.ConfigurationError, err:
13063 raise errors.OpPrereqError("Invalid instance policy: %s" % err,
13064 errors.ECODE_INVAL)
13066 def BuildHooksEnv(self):
13067 """Build hooks env.
13071 "GROUP_NAME": self.op.group_name,
13074 def BuildHooksNodes(self):
13075 """Build hooks nodes.
13078 mn = self.cfg.GetMasterNode()
13079 return ([mn], [mn])
13081 def Exec(self, feedback_fn):
13082 """Add the node group to the cluster.
13085 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
13086 uuid=self.group_uuid,
13087 alloc_policy=self.op.alloc_policy,
13088 ndparams=self.op.ndparams,
13089 diskparams=self.op.diskparams,
13090 ipolicy=self.op.ipolicy,
13091 hv_state_static=self.new_hv_state,
13092 disk_state_static=self.new_disk_state)
13094 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
13095 del self.remove_locks[locking.LEVEL_NODEGROUP]
13098 class LUGroupAssignNodes(NoHooksLU):
13099 """Logical unit for assigning nodes to groups.
13104 def ExpandNames(self):
13105 # These raise errors.OpPrereqError on their own:
13106 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13107 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
13109 # We want to lock all the affected nodes and groups. We have readily
13110 # available the list of nodes, and the *destination* group. To gather the
13111 # list of "source" groups, we need to fetch node information later on.
13112 self.needed_locks = {
13113 locking.LEVEL_NODEGROUP: set([self.group_uuid]),
13114 locking.LEVEL_NODE: self.op.nodes,
13117 def DeclareLocks(self, level):
13118 if level == locking.LEVEL_NODEGROUP:
13119 assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1
13121 # Try to get all affected nodes' groups without having the group or node
13122 # lock yet. Needs verification later in the code flow.
13123 groups = self.cfg.GetNodeGroupsFromNodes(self.op.nodes)
13125 self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)
13127 def CheckPrereq(self):
13128 """Check prerequisites.
13131 assert self.needed_locks[locking.LEVEL_NODEGROUP]
13132 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
13133 frozenset(self.op.nodes))
13135 expected_locks = (set([self.group_uuid]) |
13136 self.cfg.GetNodeGroupsFromNodes(self.op.nodes))
13137 actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP)
13138 if actual_locks != expected_locks:
13139 raise errors.OpExecError("Nodes changed groups since locks were acquired,"
13140 " current groups are '%s', used to be '%s'" %
13141 (utils.CommaJoin(expected_locks),
13142 utils.CommaJoin(actual_locks)))
13144 self.node_data = self.cfg.GetAllNodesInfo()
13145 self.group = self.cfg.GetNodeGroup(self.group_uuid)
13146 instance_data = self.cfg.GetAllInstancesInfo()
13148 if self.group is None:
13149 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13150 (self.op.group_name, self.group_uuid))
13152 (new_splits, previous_splits) = \
13153 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
13154 for node in self.op.nodes],
13155 self.node_data, instance_data)
13158 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
13160 if not self.op.force:
13161 raise errors.OpExecError("The following instances get split by this"
13162 " change and --force was not given: %s" %
13165 self.LogWarning("This operation will split the following instances: %s",
13168 if previous_splits:
13169 self.LogWarning("In addition, these already-split instances continue"
13170 " to be split across groups: %s",
13171 utils.CommaJoin(utils.NiceSort(previous_splits)))
13173 def Exec(self, feedback_fn):
13174 """Assign nodes to a new group.
13177 mods = [(node_name, self.group_uuid) for node_name in self.op.nodes]
13179 self.cfg.AssignGroupNodes(mods)
13182 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
13183 """Check for split instances after a node assignment.
13185 This method considers a series of node assignments as an atomic operation,
13186 and returns information about split instances after applying the set of
13189 In particular, it returns information about newly split instances, and
13190 instances that were already split, and remain so after the change.
13192 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
13195 @type changes: list of (node_name, new_group_uuid) pairs.
13196 @param changes: list of node assignments to consider.
13197 @param node_data: a dict with data for all nodes
13198 @param instance_data: a dict with all instances to consider
13199 @rtype: a two-tuple
13200 @return: a list of instances that were previously okay and result split as a
13201 consequence of this change, and a list of instances that were previously
13202 split and this change does not fix.
13205 changed_nodes = dict((node, group) for node, group in changes
13206 if node_data[node].group != group)
13208 all_split_instances = set()
13209 previously_split_instances = set()
13211 def InstanceNodes(instance):
13212 return [instance.primary_node] + list(instance.secondary_nodes)
13214 for inst in instance_data.values():
13215 if inst.disk_template not in constants.DTS_INT_MIRROR:
13218 instance_nodes = InstanceNodes(inst)
13220 if len(set(node_data[node].group for node in instance_nodes)) > 1:
13221 previously_split_instances.add(inst.name)
13223 if len(set(changed_nodes.get(node, node_data[node].group)
13224 for node in instance_nodes)) > 1:
13225 all_split_instances.add(inst.name)
13227 return (list(all_split_instances - previously_split_instances),
13228 list(previously_split_instances & all_split_instances))
13231 class _GroupQuery(_QueryBase):
13232 FIELDS = query.GROUP_FIELDS
13234 def ExpandNames(self, lu):
13235 lu.needed_locks = {}
13237 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
13238 self._cluster = lu.cfg.GetClusterInfo()
13239 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
13242 self.wanted = [name_to_uuid[name]
13243 for name in utils.NiceSort(name_to_uuid.keys())]
13245 # Accept names to be either names or UUIDs.
13248 all_uuid = frozenset(self._all_groups.keys())
13250 for name in self.names:
13251 if name in all_uuid:
13252 self.wanted.append(name)
13253 elif name in name_to_uuid:
13254 self.wanted.append(name_to_uuid[name])
13256 missing.append(name)
13259 raise errors.OpPrereqError("Some groups do not exist: %s" %
13260 utils.CommaJoin(missing),
13261 errors.ECODE_NOENT)
13263 def DeclareLocks(self, lu, level):
13266 def _GetQueryData(self, lu):
13267 """Computes the list of node groups and their attributes.
13270 do_nodes = query.GQ_NODE in self.requested_data
13271 do_instances = query.GQ_INST in self.requested_data
13273 group_to_nodes = None
13274 group_to_instances = None
13276 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
13277 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
13278 # latter GetAllInstancesInfo() is not enough, for we have to go through
13279 # instance->node. Hence, we will need to process nodes even if we only need
13280 # instance information.
13281 if do_nodes or do_instances:
13282 all_nodes = lu.cfg.GetAllNodesInfo()
13283 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
13286 for node in all_nodes.values():
13287 if node.group in group_to_nodes:
13288 group_to_nodes[node.group].append(node.name)
13289 node_to_group[node.name] = node.group
13292 all_instances = lu.cfg.GetAllInstancesInfo()
13293 group_to_instances = dict((uuid, []) for uuid in self.wanted)
13295 for instance in all_instances.values():
13296 node = instance.primary_node
13297 if node in node_to_group:
13298 group_to_instances[node_to_group[node]].append(instance.name)
13301 # Do not pass on node information if it was not requested.
13302 group_to_nodes = None
13304 return query.GroupQueryData(self._cluster,
13305 [self._all_groups[uuid]
13306 for uuid in self.wanted],
13307 group_to_nodes, group_to_instances)
13310 class LUGroupQuery(NoHooksLU):
13311 """Logical unit for querying node groups.
13316 def CheckArguments(self):
13317 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
13318 self.op.output_fields, False)
13320 def ExpandNames(self):
13321 self.gq.ExpandNames(self)
13323 def DeclareLocks(self, level):
13324 self.gq.DeclareLocks(self, level)
13326 def Exec(self, feedback_fn):
13327 return self.gq.OldStyleQuery(self)
13330 class LUGroupSetParams(LogicalUnit):
13331 """Modifies the parameters of a node group.
13334 HPATH = "group-modify"
13335 HTYPE = constants.HTYPE_GROUP
13338 def CheckArguments(self):
13341 self.op.diskparams,
13342 self.op.alloc_policy,
13344 self.op.disk_state,
13348 if all_changes.count(None) == len(all_changes):
13349 raise errors.OpPrereqError("Please pass at least one modification",
13350 errors.ECODE_INVAL)
13352 def ExpandNames(self):
13353 # This raises errors.OpPrereqError on its own:
13354 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13356 self.needed_locks = {
13357 locking.LEVEL_NODEGROUP: [self.group_uuid],
13360 def CheckPrereq(self):
13361 """Check prerequisites.
13364 self.group = self.cfg.GetNodeGroup(self.group_uuid)
13366 if self.group is None:
13367 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13368 (self.op.group_name, self.group_uuid))
13370 if self.op.ndparams:
13371 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
13372 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
13373 self.new_ndparams = new_ndparams
13375 if self.op.diskparams:
13376 self.new_diskparams = dict()
13377 for templ in constants.DISK_TEMPLATES:
13378 if templ not in self.op.diskparams:
13379 self.op.diskparams[templ] = {}
13380 new_templ_params = _GetUpdatedParams(self.group.diskparams[templ],
13381 self.op.diskparams[templ])
13382 utils.ForceDictType(new_templ_params, constants.DISK_DT_TYPES)
13383 self.new_diskparams[templ] = new_templ_params
13385 if self.op.hv_state:
13386 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
13387 self.group.hv_state_static)
13389 if self.op.disk_state:
13390 self.new_disk_state = \
13391 _MergeAndVerifyDiskState(self.op.disk_state,
13392 self.group.disk_state_static)
13394 if self.op.ipolicy:
13395 self.new_ipolicy = _GetUpdatedIPolicy(self.group.ipolicy,
13399 def BuildHooksEnv(self):
13400 """Build hooks env.
13404 "GROUP_NAME": self.op.group_name,
13405 "NEW_ALLOC_POLICY": self.op.alloc_policy,
13408 def BuildHooksNodes(self):
13409 """Build hooks nodes.
13412 mn = self.cfg.GetMasterNode()
13413 return ([mn], [mn])
13415 def Exec(self, feedback_fn):
13416 """Modifies the node group.
13421 if self.op.ndparams:
13422 self.group.ndparams = self.new_ndparams
13423 result.append(("ndparams", str(self.group.ndparams)))
13425 if self.op.diskparams:
13426 self.group.diskparams = self.new_diskparams
13427 result.append(("diskparams", str(self.group.diskparams)))
13429 if self.op.alloc_policy:
13430 self.group.alloc_policy = self.op.alloc_policy
13432 if self.op.hv_state:
13433 self.group.hv_state_static = self.new_hv_state
13435 if self.op.disk_state:
13436 self.group.disk_state_static = self.new_disk_state
13438 if self.op.ipolicy:
13439 self.group.ipolicy = self.new_ipolicy
13441 self.cfg.Update(self.group, feedback_fn)
13445 class LUGroupRemove(LogicalUnit):
13446 HPATH = "group-remove"
13447 HTYPE = constants.HTYPE_GROUP
13450 def ExpandNames(self):
13451 # This will raises errors.OpPrereqError on its own:
13452 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13453 self.needed_locks = {
13454 locking.LEVEL_NODEGROUP: [self.group_uuid],
13457 def CheckPrereq(self):
13458 """Check prerequisites.
13460 This checks that the given group name exists as a node group, that is
13461 empty (i.e., contains no nodes), and that is not the last group of the
13465 # Verify that the group is empty.
13466 group_nodes = [node.name
13467 for node in self.cfg.GetAllNodesInfo().values()
13468 if node.group == self.group_uuid]
13471 raise errors.OpPrereqError("Group '%s' not empty, has the following"
13473 (self.op.group_name,
13474 utils.CommaJoin(utils.NiceSort(group_nodes))),
13475 errors.ECODE_STATE)
13477 # Verify the cluster would not be left group-less.
13478 if len(self.cfg.GetNodeGroupList()) == 1:
13479 raise errors.OpPrereqError("Group '%s' is the only group,"
13480 " cannot be removed" %
13481 self.op.group_name,
13482 errors.ECODE_STATE)
13484 def BuildHooksEnv(self):
13485 """Build hooks env.
13489 "GROUP_NAME": self.op.group_name,
13492 def BuildHooksNodes(self):
13493 """Build hooks nodes.
13496 mn = self.cfg.GetMasterNode()
13497 return ([mn], [mn])
13499 def Exec(self, feedback_fn):
13500 """Remove the node group.
13504 self.cfg.RemoveNodeGroup(self.group_uuid)
13505 except errors.ConfigurationError:
13506 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
13507 (self.op.group_name, self.group_uuid))
13509 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
13512 class LUGroupRename(LogicalUnit):
13513 HPATH = "group-rename"
13514 HTYPE = constants.HTYPE_GROUP
13517 def ExpandNames(self):
13518 # This raises errors.OpPrereqError on its own:
13519 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13521 self.needed_locks = {
13522 locking.LEVEL_NODEGROUP: [self.group_uuid],
13525 def CheckPrereq(self):
13526 """Check prerequisites.
13528 Ensures requested new name is not yet used.
13532 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
13533 except errors.OpPrereqError:
13536 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
13537 " node group (UUID: %s)" %
13538 (self.op.new_name, new_name_uuid),
13539 errors.ECODE_EXISTS)
13541 def BuildHooksEnv(self):
13542 """Build hooks env.
13546 "OLD_NAME": self.op.group_name,
13547 "NEW_NAME": self.op.new_name,
13550 def BuildHooksNodes(self):
13551 """Build hooks nodes.
13554 mn = self.cfg.GetMasterNode()
13556 all_nodes = self.cfg.GetAllNodesInfo()
13557 all_nodes.pop(mn, None)
13560 run_nodes.extend(node.name for node in all_nodes.values()
13561 if node.group == self.group_uuid)
13563 return (run_nodes, run_nodes)
13565 def Exec(self, feedback_fn):
13566 """Rename the node group.
13569 group = self.cfg.GetNodeGroup(self.group_uuid)
13572 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13573 (self.op.group_name, self.group_uuid))
13575 group.name = self.op.new_name
13576 self.cfg.Update(group, feedback_fn)
13578 return self.op.new_name
13581 class LUGroupEvacuate(LogicalUnit):
13582 HPATH = "group-evacuate"
13583 HTYPE = constants.HTYPE_GROUP
13586 def ExpandNames(self):
13587 # This raises errors.OpPrereqError on its own:
13588 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13590 if self.op.target_groups:
13591 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
13592 self.op.target_groups)
13594 self.req_target_uuids = []
13596 if self.group_uuid in self.req_target_uuids:
13597 raise errors.OpPrereqError("Group to be evacuated (%s) can not be used"
13598 " as a target group (targets are %s)" %
13600 utils.CommaJoin(self.req_target_uuids)),
13601 errors.ECODE_INVAL)
13603 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
13605 self.share_locks = _ShareAll()
13606 self.needed_locks = {
13607 locking.LEVEL_INSTANCE: [],
13608 locking.LEVEL_NODEGROUP: [],
13609 locking.LEVEL_NODE: [],
13612 def DeclareLocks(self, level):
13613 if level == locking.LEVEL_INSTANCE:
13614 assert not self.needed_locks[locking.LEVEL_INSTANCE]
13616 # Lock instances optimistically, needs verification once node and group
13617 # locks have been acquired
13618 self.needed_locks[locking.LEVEL_INSTANCE] = \
13619 self.cfg.GetNodeGroupInstances(self.group_uuid)
13621 elif level == locking.LEVEL_NODEGROUP:
13622 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
13624 if self.req_target_uuids:
13625 lock_groups = set([self.group_uuid] + self.req_target_uuids)
13627 # Lock all groups used by instances optimistically; this requires going
13628 # via the node before it's locked, requiring verification later on
13629 lock_groups.update(group_uuid
13630 for instance_name in
13631 self.owned_locks(locking.LEVEL_INSTANCE)
13633 self.cfg.GetInstanceNodeGroups(instance_name))
13635 # No target groups, need to lock all of them
13636 lock_groups = locking.ALL_SET
13638 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
13640 elif level == locking.LEVEL_NODE:
13641 # This will only lock the nodes in the group to be evacuated which
13642 # contain actual instances
13643 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
13644 self._LockInstancesNodes()
13646 # Lock all nodes in group to be evacuated and target groups
13647 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
13648 assert self.group_uuid in owned_groups
13649 member_nodes = [node_name
13650 for group in owned_groups
13651 for node_name in self.cfg.GetNodeGroup(group).members]
13652 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
13654 def CheckPrereq(self):
13655 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
13656 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
13657 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
13659 assert owned_groups.issuperset(self.req_target_uuids)
13660 assert self.group_uuid in owned_groups
13662 # Check if locked instances are still correct
13663 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
13665 # Get instance information
13666 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
13668 # Check if node groups for locked instances are still correct
13669 for instance_name in owned_instances:
13670 inst = self.instances[instance_name]
13671 assert owned_nodes.issuperset(inst.all_nodes), \
13672 "Instance %s's nodes changed while we kept the lock" % instance_name
13674 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
13677 assert self.group_uuid in inst_groups, \
13678 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
13680 if self.req_target_uuids:
13681 # User requested specific target groups
13682 self.target_uuids = self.req_target_uuids
13684 # All groups except the one to be evacuated are potential targets
13685 self.target_uuids = [group_uuid for group_uuid in owned_groups
13686 if group_uuid != self.group_uuid]
13688 if not self.target_uuids:
13689 raise errors.OpPrereqError("There are no possible target groups",
13690 errors.ECODE_INVAL)
13692 def BuildHooksEnv(self):
13693 """Build hooks env.
13697 "GROUP_NAME": self.op.group_name,
13698 "TARGET_GROUPS": " ".join(self.target_uuids),
13701 def BuildHooksNodes(self):
13702 """Build hooks nodes.
13705 mn = self.cfg.GetMasterNode()
13707 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
13709 run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members
13711 return (run_nodes, run_nodes)
13713 def Exec(self, feedback_fn):
13714 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
13716 assert self.group_uuid not in self.target_uuids
13718 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
13719 instances=instances, target_groups=self.target_uuids)
13721 ial.Run(self.op.iallocator)
13723 if not ial.success:
13724 raise errors.OpPrereqError("Can't compute group evacuation using"
13725 " iallocator '%s': %s" %
13726 (self.op.iallocator, ial.info),
13727 errors.ECODE_NORES)
13729 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
13731 self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s",
13732 len(jobs), self.op.group_name)
13734 return ResultWithJobs(jobs)
13737 class TagsLU(NoHooksLU): # pylint: disable=W0223
13738 """Generic tags LU.
13740 This is an abstract class which is the parent of all the other tags LUs.
13743 def ExpandNames(self):
13744 self.group_uuid = None
13745 self.needed_locks = {}
13746 if self.op.kind == constants.TAG_NODE:
13747 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
13748 self.needed_locks[locking.LEVEL_NODE] = self.op.name
13749 elif self.op.kind == constants.TAG_INSTANCE:
13750 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
13751 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
13752 elif self.op.kind == constants.TAG_NODEGROUP:
13753 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
13755 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
13756 # not possible to acquire the BGL based on opcode parameters)
13758 def CheckPrereq(self):
13759 """Check prerequisites.
13762 if self.op.kind == constants.TAG_CLUSTER:
13763 self.target = self.cfg.GetClusterInfo()
13764 elif self.op.kind == constants.TAG_NODE:
13765 self.target = self.cfg.GetNodeInfo(self.op.name)
13766 elif self.op.kind == constants.TAG_INSTANCE:
13767 self.target = self.cfg.GetInstanceInfo(self.op.name)
13768 elif self.op.kind == constants.TAG_NODEGROUP:
13769 self.target = self.cfg.GetNodeGroup(self.group_uuid)
13771 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
13772 str(self.op.kind), errors.ECODE_INVAL)
13775 class LUTagsGet(TagsLU):
13776 """Returns the tags of a given object.
13781 def ExpandNames(self):
13782 TagsLU.ExpandNames(self)
13784 # Share locks as this is only a read operation
13785 self.share_locks = _ShareAll()
13787 def Exec(self, feedback_fn):
13788 """Returns the tag list.
13791 return list(self.target.GetTags())
13794 class LUTagsSearch(NoHooksLU):
13795 """Searches the tags for a given pattern.
13800 def ExpandNames(self):
13801 self.needed_locks = {}
13803 def CheckPrereq(self):
13804 """Check prerequisites.
13806 This checks the pattern passed for validity by compiling it.
13810 self.re = re.compile(self.op.pattern)
13811 except re.error, err:
13812 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
13813 (self.op.pattern, err), errors.ECODE_INVAL)
13815 def Exec(self, feedback_fn):
13816 """Returns the tag list.
13820 tgts = [("/cluster", cfg.GetClusterInfo())]
13821 ilist = cfg.GetAllInstancesInfo().values()
13822 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
13823 nlist = cfg.GetAllNodesInfo().values()
13824 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
13825 tgts.extend(("/nodegroup/%s" % n.name, n)
13826 for n in cfg.GetAllNodeGroupsInfo().values())
13828 for path, target in tgts:
13829 for tag in target.GetTags():
13830 if self.re.search(tag):
13831 results.append((path, tag))
13835 class LUTagsSet(TagsLU):
13836 """Sets a tag on a given object.
13841 def CheckPrereq(self):
13842 """Check prerequisites.
13844 This checks the type and length of the tag name and value.
13847 TagsLU.CheckPrereq(self)
13848 for tag in self.op.tags:
13849 objects.TaggableObject.ValidateTag(tag)
13851 def Exec(self, feedback_fn):
13856 for tag in self.op.tags:
13857 self.target.AddTag(tag)
13858 except errors.TagError, err:
13859 raise errors.OpExecError("Error while setting tag: %s" % str(err))
13860 self.cfg.Update(self.target, feedback_fn)
13863 class LUTagsDel(TagsLU):
13864 """Delete a list of tags from a given object.
13869 def CheckPrereq(self):
13870 """Check prerequisites.
13872 This checks that we have the given tag.
13875 TagsLU.CheckPrereq(self)
13876 for tag in self.op.tags:
13877 objects.TaggableObject.ValidateTag(tag)
13878 del_tags = frozenset(self.op.tags)
13879 cur_tags = self.target.GetTags()
13881 diff_tags = del_tags - cur_tags
13883 diff_names = ("'%s'" % i for i in sorted(diff_tags))
13884 raise errors.OpPrereqError("Tag(s) %s not found" %
13885 (utils.CommaJoin(diff_names), ),
13886 errors.ECODE_NOENT)
13888 def Exec(self, feedback_fn):
13889 """Remove the tag from the object.
13892 for tag in self.op.tags:
13893 self.target.RemoveTag(tag)
13894 self.cfg.Update(self.target, feedback_fn)
13897 class LUTestDelay(NoHooksLU):
13898 """Sleep for a specified amount of time.
13900 This LU sleeps on the master and/or nodes for a specified amount of
13906 def ExpandNames(self):
13907 """Expand names and set required locks.
13909 This expands the node list, if any.
13912 self.needed_locks = {}
13913 if self.op.on_nodes:
13914 # _GetWantedNodes can be used here, but is not always appropriate to use
13915 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
13916 # more information.
13917 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
13918 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
13920 def _TestDelay(self):
13921 """Do the actual sleep.
13924 if self.op.on_master:
13925 if not utils.TestDelay(self.op.duration):
13926 raise errors.OpExecError("Error during master delay test")
13927 if self.op.on_nodes:
13928 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
13929 for node, node_result in result.items():
13930 node_result.Raise("Failure during rpc call to node %s" % node)
13932 def Exec(self, feedback_fn):
13933 """Execute the test delay opcode, with the wanted repetitions.
13936 if self.op.repeat == 0:
13939 top_value = self.op.repeat - 1
13940 for i in range(self.op.repeat):
13941 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
13945 class LUTestJqueue(NoHooksLU):
13946 """Utility LU to test some aspects of the job queue.
13951 # Must be lower than default timeout for WaitForJobChange to see whether it
13952 # notices changed jobs
13953 _CLIENT_CONNECT_TIMEOUT = 20.0
13954 _CLIENT_CONFIRM_TIMEOUT = 60.0
13957 def _NotifyUsingSocket(cls, cb, errcls):
13958 """Opens a Unix socket and waits for another program to connect.
13961 @param cb: Callback to send socket name to client
13962 @type errcls: class
13963 @param errcls: Exception class to use for errors
13966 # Using a temporary directory as there's no easy way to create temporary
13967 # sockets without writing a custom loop around tempfile.mktemp and
13969 tmpdir = tempfile.mkdtemp()
13971 tmpsock = utils.PathJoin(tmpdir, "sock")
13973 logging.debug("Creating temporary socket at %s", tmpsock)
13974 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
13979 # Send details to client
13982 # Wait for client to connect before continuing
13983 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
13985 (conn, _) = sock.accept()
13986 except socket.error, err:
13987 raise errcls("Client didn't connect in time (%s)" % err)
13991 # Remove as soon as client is connected
13992 shutil.rmtree(tmpdir)
13994 # Wait for client to close
13997 # pylint: disable=E1101
13998 # Instance of '_socketobject' has no ... member
13999 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
14001 except socket.error, err:
14002 raise errcls("Client failed to confirm notification (%s)" % err)
14006 def _SendNotification(self, test, arg, sockname):
14007 """Sends a notification to the client.
14010 @param test: Test name
14011 @param arg: Test argument (depends on test)
14012 @type sockname: string
14013 @param sockname: Socket path
14016 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
14018 def _Notify(self, prereq, test, arg):
14019 """Notifies the client of a test.
14022 @param prereq: Whether this is a prereq-phase test
14024 @param test: Test name
14025 @param arg: Test argument (depends on test)
14029 errcls = errors.OpPrereqError
14031 errcls = errors.OpExecError
14033 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
14037 def CheckArguments(self):
14038 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
14039 self.expandnames_calls = 0
14041 def ExpandNames(self):
14042 checkargs_calls = getattr(self, "checkargs_calls", 0)
14043 if checkargs_calls < 1:
14044 raise errors.ProgrammerError("CheckArguments was not called")
14046 self.expandnames_calls += 1
14048 if self.op.notify_waitlock:
14049 self._Notify(True, constants.JQT_EXPANDNAMES, None)
14051 self.LogInfo("Expanding names")
14053 # Get lock on master node (just to get a lock, not for a particular reason)
14054 self.needed_locks = {
14055 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
14058 def Exec(self, feedback_fn):
14059 if self.expandnames_calls < 1:
14060 raise errors.ProgrammerError("ExpandNames was not called")
14062 if self.op.notify_exec:
14063 self._Notify(False, constants.JQT_EXEC, None)
14065 self.LogInfo("Executing")
14067 if self.op.log_messages:
14068 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
14069 for idx, msg in enumerate(self.op.log_messages):
14070 self.LogInfo("Sending log message %s", idx + 1)
14071 feedback_fn(constants.JQT_MSGPREFIX + msg)
14072 # Report how many test messages have been sent
14073 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
14076 raise errors.OpExecError("Opcode failure was requested")
14081 class IAllocator(object):
14082 """IAllocator framework.
14084 An IAllocator instance has three sets of attributes:
14085 - cfg that is needed to query the cluster
14086 - input data (all members of the _KEYS class attribute are required)
14087 - four buffer attributes (in|out_data|text), that represent the
14088 input (to the external script) in text and data structure format,
14089 and the output from it, again in two formats
14090 - the result variables from the script (success, info, nodes) for
14094 # pylint: disable=R0902
14095 # lots of instance attributes
14097 def __init__(self, cfg, rpc_runner, mode, **kwargs):
14099 self.rpc = rpc_runner
14100 # init buffer variables
14101 self.in_text = self.out_text = self.in_data = self.out_data = None
14102 # init all input fields so that pylint is happy
14104 self.memory = self.disks = self.disk_template = None
14105 self.os = self.tags = self.nics = self.vcpus = None
14106 self.hypervisor = None
14107 self.relocate_from = None
14109 self.instances = None
14110 self.evac_mode = None
14111 self.target_groups = []
14113 self.required_nodes = None
14114 # init result fields
14115 self.success = self.info = self.result = None
14118 (fn, keydata, self._result_check) = self._MODE_DATA[self.mode]
14120 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
14121 " IAllocator" % self.mode)
14123 keyset = [n for (n, _) in keydata]
14126 if key not in keyset:
14127 raise errors.ProgrammerError("Invalid input parameter '%s' to"
14128 " IAllocator" % key)
14129 setattr(self, key, kwargs[key])
14132 if key not in kwargs:
14133 raise errors.ProgrammerError("Missing input parameter '%s' to"
14134 " IAllocator" % key)
14135 self._BuildInputData(compat.partial(fn, self), keydata)
14137 def _ComputeClusterData(self):
14138 """Compute the generic allocator input data.
14140 This is the data that is independent of the actual operation.
14144 cluster_info = cfg.GetClusterInfo()
14147 "version": constants.IALLOCATOR_VERSION,
14148 "cluster_name": cfg.GetClusterName(),
14149 "cluster_tags": list(cluster_info.GetTags()),
14150 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
14151 # we don't have job IDs
14153 ninfo = cfg.GetAllNodesInfo()
14154 iinfo = cfg.GetAllInstancesInfo().values()
14155 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
14158 node_list = [n.name for n in ninfo.values() if n.vm_capable]
14160 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
14161 hypervisor_name = self.hypervisor
14162 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
14163 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
14165 hypervisor_name = cluster_info.primary_hypervisor
14167 node_data = self.rpc.call_node_info(node_list, [cfg.GetVGName()],
14170 self.rpc.call_all_instances_info(node_list,
14171 cluster_info.enabled_hypervisors)
14173 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
14175 config_ndata = self._ComputeBasicNodeData(ninfo)
14176 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
14177 i_list, config_ndata)
14178 assert len(data["nodes"]) == len(ninfo), \
14179 "Incomplete node data computed"
14181 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
14183 self.in_data = data
14186 def _ComputeNodeGroupData(cfg):
14187 """Compute node groups data.
14190 ng = dict((guuid, {
14191 "name": gdata.name,
14192 "alloc_policy": gdata.alloc_policy,
14194 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items())
14199 def _ComputeBasicNodeData(node_cfg):
14200 """Compute global node data.
14203 @returns: a dict of name: (node dict, node config)
14206 # fill in static (config-based) values
14207 node_results = dict((ninfo.name, {
14208 "tags": list(ninfo.GetTags()),
14209 "primary_ip": ninfo.primary_ip,
14210 "secondary_ip": ninfo.secondary_ip,
14211 "offline": ninfo.offline,
14212 "drained": ninfo.drained,
14213 "master_candidate": ninfo.master_candidate,
14214 "group": ninfo.group,
14215 "master_capable": ninfo.master_capable,
14216 "vm_capable": ninfo.vm_capable,
14218 for ninfo in node_cfg.values())
14220 return node_results
14223 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
14225 """Compute global node data.
14227 @param node_results: the basic node structures as filled from the config
14230 #TODO(dynmem): compute the right data on MAX and MIN memory
14231 # make a copy of the current dict
14232 node_results = dict(node_results)
14233 for nname, nresult in node_data.items():
14234 assert nname in node_results, "Missing basic data for node %s" % nname
14235 ninfo = node_cfg[nname]
14237 if not (ninfo.offline or ninfo.drained):
14238 nresult.Raise("Can't get data for node %s" % nname)
14239 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
14241 remote_info = _MakeLegacyNodeInfo(nresult.payload)
14243 for attr in ["memory_total", "memory_free", "memory_dom0",
14244 "vg_size", "vg_free", "cpu_total"]:
14245 if attr not in remote_info:
14246 raise errors.OpExecError("Node '%s' didn't return attribute"
14247 " '%s'" % (nname, attr))
14248 if not isinstance(remote_info[attr], int):
14249 raise errors.OpExecError("Node '%s' returned invalid value"
14251 (nname, attr, remote_info[attr]))
14252 # compute memory used by primary instances
14253 i_p_mem = i_p_up_mem = 0
14254 for iinfo, beinfo in i_list:
14255 if iinfo.primary_node == nname:
14256 i_p_mem += beinfo[constants.BE_MAXMEM]
14257 if iinfo.name not in node_iinfo[nname].payload:
14260 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]["memory"])
14261 i_mem_diff = beinfo[constants.BE_MAXMEM] - i_used_mem
14262 remote_info["memory_free"] -= max(0, i_mem_diff)
14264 if iinfo.admin_state == constants.ADMINST_UP:
14265 i_p_up_mem += beinfo[constants.BE_MAXMEM]
14267 # compute memory used by instances
14269 "total_memory": remote_info["memory_total"],
14270 "reserved_memory": remote_info["memory_dom0"],
14271 "free_memory": remote_info["memory_free"],
14272 "total_disk": remote_info["vg_size"],
14273 "free_disk": remote_info["vg_free"],
14274 "total_cpus": remote_info["cpu_total"],
14275 "i_pri_memory": i_p_mem,
14276 "i_pri_up_memory": i_p_up_mem,
14278 pnr_dyn.update(node_results[nname])
14279 node_results[nname] = pnr_dyn
14281 return node_results
14284 def _ComputeInstanceData(cluster_info, i_list):
14285 """Compute global instance data.
14289 for iinfo, beinfo in i_list:
14291 for nic in iinfo.nics:
14292 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
14296 "mode": filled_params[constants.NIC_MODE],
14297 "link": filled_params[constants.NIC_LINK],
14299 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
14300 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
14301 nic_data.append(nic_dict)
14303 "tags": list(iinfo.GetTags()),
14304 "admin_state": iinfo.admin_state,
14305 "vcpus": beinfo[constants.BE_VCPUS],
14306 "memory": beinfo[constants.BE_MAXMEM],
14308 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
14310 "disks": [{constants.IDISK_SIZE: dsk.size,
14311 constants.IDISK_MODE: dsk.mode}
14312 for dsk in iinfo.disks],
14313 "disk_template": iinfo.disk_template,
14314 "hypervisor": iinfo.hypervisor,
14316 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
14318 instance_data[iinfo.name] = pir
14320 return instance_data
14322 def _AddNewInstance(self):
14323 """Add new instance data to allocator structure.
14325 This in combination with _AllocatorGetClusterData will create the
14326 correct structure needed as input for the allocator.
14328 The checks for the completeness of the opcode must have already been
14332 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
14334 if self.disk_template in constants.DTS_INT_MIRROR:
14335 self.required_nodes = 2
14337 self.required_nodes = 1
14341 "disk_template": self.disk_template,
14344 "vcpus": self.vcpus,
14345 "memory": self.memory,
14346 "disks": self.disks,
14347 "disk_space_total": disk_space,
14349 "required_nodes": self.required_nodes,
14350 "hypervisor": self.hypervisor,
14355 def _AddRelocateInstance(self):
14356 """Add relocate instance data to allocator structure.
14358 This in combination with _IAllocatorGetClusterData will create the
14359 correct structure needed as input for the allocator.
14361 The checks for the completeness of the opcode must have already been
14365 instance = self.cfg.GetInstanceInfo(self.name)
14366 if instance is None:
14367 raise errors.ProgrammerError("Unknown instance '%s' passed to"
14368 " IAllocator" % self.name)
14370 if instance.disk_template not in constants.DTS_MIRRORED:
14371 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
14372 errors.ECODE_INVAL)
14374 if instance.disk_template in constants.DTS_INT_MIRROR and \
14375 len(instance.secondary_nodes) != 1:
14376 raise errors.OpPrereqError("Instance has not exactly one secondary node",
14377 errors.ECODE_STATE)
14379 self.required_nodes = 1
14380 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
14381 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
14385 "disk_space_total": disk_space,
14386 "required_nodes": self.required_nodes,
14387 "relocate_from": self.relocate_from,
14391 def _AddNodeEvacuate(self):
14392 """Get data for node-evacuate requests.
14396 "instances": self.instances,
14397 "evac_mode": self.evac_mode,
14400 def _AddChangeGroup(self):
14401 """Get data for node-evacuate requests.
14405 "instances": self.instances,
14406 "target_groups": self.target_groups,
14409 def _BuildInputData(self, fn, keydata):
14410 """Build input data structures.
14413 self._ComputeClusterData()
14416 request["type"] = self.mode
14417 for keyname, keytype in keydata:
14418 if keyname not in request:
14419 raise errors.ProgrammerError("Request parameter %s is missing" %
14421 val = request[keyname]
14422 if not keytype(val):
14423 raise errors.ProgrammerError("Request parameter %s doesn't pass"
14424 " validation, value %s, expected"
14425 " type %s" % (keyname, val, keytype))
14426 self.in_data["request"] = request
14428 self.in_text = serializer.Dump(self.in_data)
14430 _STRING_LIST = ht.TListOf(ht.TString)
14431 _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
14432 # pylint: disable=E1101
14433 # Class '...' has no 'OP_ID' member
14434 "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
14435 opcodes.OpInstanceMigrate.OP_ID,
14436 opcodes.OpInstanceReplaceDisks.OP_ID])
14440 ht.TListOf(ht.TAnd(ht.TIsLength(3),
14441 ht.TItems([ht.TNonEmptyString,
14442 ht.TNonEmptyString,
14443 ht.TListOf(ht.TNonEmptyString),
14446 ht.TListOf(ht.TAnd(ht.TIsLength(2),
14447 ht.TItems([ht.TNonEmptyString,
14450 _NEVAC_RESULT = ht.TAnd(ht.TIsLength(3),
14451 ht.TItems([_NEVAC_MOVED, _NEVAC_FAILED, _JOB_LIST]))
14454 constants.IALLOCATOR_MODE_ALLOC:
14457 ("name", ht.TString),
14458 ("memory", ht.TInt),
14459 ("disks", ht.TListOf(ht.TDict)),
14460 ("disk_template", ht.TString),
14461 ("os", ht.TString),
14462 ("tags", _STRING_LIST),
14463 ("nics", ht.TListOf(ht.TDict)),
14464 ("vcpus", ht.TInt),
14465 ("hypervisor", ht.TString),
14467 constants.IALLOCATOR_MODE_RELOC:
14468 (_AddRelocateInstance,
14469 [("name", ht.TString), ("relocate_from", _STRING_LIST)],
14471 constants.IALLOCATOR_MODE_NODE_EVAC:
14472 (_AddNodeEvacuate, [
14473 ("instances", _STRING_LIST),
14474 ("evac_mode", ht.TElemOf(constants.IALLOCATOR_NEVAC_MODES)),
14476 constants.IALLOCATOR_MODE_CHG_GROUP:
14477 (_AddChangeGroup, [
14478 ("instances", _STRING_LIST),
14479 ("target_groups", _STRING_LIST),
14483 def Run(self, name, validate=True, call_fn=None):
14484 """Run an instance allocator and return the results.
14487 if call_fn is None:
14488 call_fn = self.rpc.call_iallocator_runner
14490 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
14491 result.Raise("Failure while running the iallocator script")
14493 self.out_text = result.payload
14495 self._ValidateResult()
14497 def _ValidateResult(self):
14498 """Process the allocator results.
14500 This will process and if successful save the result in
14501 self.out_data and the other parameters.
14505 rdict = serializer.Load(self.out_text)
14506 except Exception, err:
14507 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
14509 if not isinstance(rdict, dict):
14510 raise errors.OpExecError("Can't parse iallocator results: not a dict")
14512 # TODO: remove backwards compatiblity in later versions
14513 if "nodes" in rdict and "result" not in rdict:
14514 rdict["result"] = rdict["nodes"]
14517 for key in "success", "info", "result":
14518 if key not in rdict:
14519 raise errors.OpExecError("Can't parse iallocator results:"
14520 " missing key '%s'" % key)
14521 setattr(self, key, rdict[key])
14523 if not self._result_check(self.result):
14524 raise errors.OpExecError("Iallocator returned invalid result,"
14525 " expected %s, got %s" %
14526 (self._result_check, self.result),
14527 errors.ECODE_INVAL)
14529 if self.mode == constants.IALLOCATOR_MODE_RELOC:
14530 assert self.relocate_from is not None
14531 assert self.required_nodes == 1
14533 node2group = dict((name, ndata["group"])
14534 for (name, ndata) in self.in_data["nodes"].items())
14536 fn = compat.partial(self._NodesToGroups, node2group,
14537 self.in_data["nodegroups"])
14539 instance = self.cfg.GetInstanceInfo(self.name)
14540 request_groups = fn(self.relocate_from + [instance.primary_node])
14541 result_groups = fn(rdict["result"] + [instance.primary_node])
14543 if self.success and not set(result_groups).issubset(request_groups):
14544 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
14545 " differ from original groups (%s)" %
14546 (utils.CommaJoin(result_groups),
14547 utils.CommaJoin(request_groups)))
14549 elif self.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
14550 assert self.evac_mode in constants.IALLOCATOR_NEVAC_MODES
14552 self.out_data = rdict
14555 def _NodesToGroups(node2group, groups, nodes):
14556 """Returns a list of unique group names for a list of nodes.
14558 @type node2group: dict
14559 @param node2group: Map from node name to group UUID
14561 @param groups: Group information
14563 @param nodes: Node names
14570 group_uuid = node2group[node]
14572 # Ignore unknown node
14576 group = groups[group_uuid]
14578 # Can't find group, let's use UUID
14579 group_name = group_uuid
14581 group_name = group["name"]
14583 result.add(group_name)
14585 return sorted(result)
14588 class LUTestAllocator(NoHooksLU):
14589 """Run allocator tests.
14591 This LU runs the allocator tests
14594 def CheckPrereq(self):
14595 """Check prerequisites.
14597 This checks the opcode parameters depending on the director and mode test.
14600 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
14601 for attr in ["memory", "disks", "disk_template",
14602 "os", "tags", "nics", "vcpus"]:
14603 if not hasattr(self.op, attr):
14604 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
14605 attr, errors.ECODE_INVAL)
14606 iname = self.cfg.ExpandInstanceName(self.op.name)
14607 if iname is not None:
14608 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
14609 iname, errors.ECODE_EXISTS)
14610 if not isinstance(self.op.nics, list):
14611 raise errors.OpPrereqError("Invalid parameter 'nics'",
14612 errors.ECODE_INVAL)
14613 if not isinstance(self.op.disks, list):
14614 raise errors.OpPrereqError("Invalid parameter 'disks'",
14615 errors.ECODE_INVAL)
14616 for row in self.op.disks:
14617 if (not isinstance(row, dict) or
14618 constants.IDISK_SIZE not in row or
14619 not isinstance(row[constants.IDISK_SIZE], int) or
14620 constants.IDISK_MODE not in row or
14621 row[constants.IDISK_MODE] not in constants.DISK_ACCESS_SET):
14622 raise errors.OpPrereqError("Invalid contents of the 'disks'"
14623 " parameter", errors.ECODE_INVAL)
14624 if self.op.hypervisor is None:
14625 self.op.hypervisor = self.cfg.GetHypervisorType()
14626 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
14627 fname = _ExpandInstanceName(self.cfg, self.op.name)
14628 self.op.name = fname
14629 self.relocate_from = \
14630 list(self.cfg.GetInstanceInfo(fname).secondary_nodes)
14631 elif self.op.mode in (constants.IALLOCATOR_MODE_CHG_GROUP,
14632 constants.IALLOCATOR_MODE_NODE_EVAC):
14633 if not self.op.instances:
14634 raise errors.OpPrereqError("Missing instances", errors.ECODE_INVAL)
14635 self.op.instances = _GetWantedInstances(self, self.op.instances)
14637 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
14638 self.op.mode, errors.ECODE_INVAL)
14640 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
14641 if self.op.allocator is None:
14642 raise errors.OpPrereqError("Missing allocator name",
14643 errors.ECODE_INVAL)
14644 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
14645 raise errors.OpPrereqError("Wrong allocator test '%s'" %
14646 self.op.direction, errors.ECODE_INVAL)
14648 def Exec(self, feedback_fn):
14649 """Run the allocator test.
14652 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
14653 ial = IAllocator(self.cfg, self.rpc,
14656 memory=self.op.memory,
14657 disks=self.op.disks,
14658 disk_template=self.op.disk_template,
14662 vcpus=self.op.vcpus,
14663 hypervisor=self.op.hypervisor,
14665 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
14666 ial = IAllocator(self.cfg, self.rpc,
14669 relocate_from=list(self.relocate_from),
14671 elif self.op.mode == constants.IALLOCATOR_MODE_CHG_GROUP:
14672 ial = IAllocator(self.cfg, self.rpc,
14674 instances=self.op.instances,
14675 target_groups=self.op.target_groups)
14676 elif self.op.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
14677 ial = IAllocator(self.cfg, self.rpc,
14679 instances=self.op.instances,
14680 evac_mode=self.op.evac_mode)
14682 raise errors.ProgrammerError("Uncatched mode %s in"
14683 " LUTestAllocator.Exec", self.op.mode)
14685 if self.op.direction == constants.IALLOCATOR_DIR_IN:
14686 result = ial.in_text
14688 ial.Run(self.op.allocator, validate=False)
14689 result = ial.out_text
14693 #: Query type implementations
14695 constants.QR_INSTANCE: _InstanceQuery,
14696 constants.QR_NODE: _NodeQuery,
14697 constants.QR_GROUP: _GroupQuery,
14698 constants.QR_OS: _OsQuery,
14701 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
14704 def _GetQueryImplementation(name):
14705 """Returns the implemtnation for a query type.
14707 @param name: Query type, must be one of L{constants.QR_VIA_OP}
14711 return _QUERY_IMPL[name]
14713 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
14714 errors.ECODE_INVAL)