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_DOWN = [constants.ADMINST_DOWN]
72 INSTANCE_ONLINE = [constants.ADMINST_DOWN, constants.ADMINST_UP]
73 INSTANCE_NOT_RUNNING = [constants.ADMINST_DOWN, constants.ADMINST_OFFLINE]
75 #: Instance status in which an instance can be marked as offline/online
76 CAN_CHANGE_INSTANCE_OFFLINE = (frozenset(INSTANCE_DOWN) | frozenset([
77 constants.ADMINST_OFFLINE,
82 """Data container for LU results with jobs.
84 Instances of this class returned from L{LogicalUnit.Exec} will be recognized
85 by L{mcpu.Processor._ProcessResult}. The latter will then submit the jobs
86 contained in the C{jobs} attribute and include the job IDs in the opcode
90 def __init__(self, jobs, **kwargs):
91 """Initializes this class.
93 Additional return values can be specified as keyword arguments.
95 @type jobs: list of lists of L{opcode.OpCode}
96 @param jobs: A list of lists of opcode objects
103 class LogicalUnit(object):
104 """Logical Unit base class.
106 Subclasses must follow these rules:
107 - implement ExpandNames
108 - implement CheckPrereq (except when tasklets are used)
109 - implement Exec (except when tasklets are used)
110 - implement BuildHooksEnv
111 - implement BuildHooksNodes
112 - redefine HPATH and HTYPE
113 - optionally redefine their run requirements:
114 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
116 Note that all commands require root permissions.
118 @ivar dry_run_result: the value (if any) that will be returned to the caller
119 in dry-run mode (signalled by opcode dry_run parameter)
126 def __init__(self, processor, op, context, rpc_runner):
127 """Constructor for LogicalUnit.
129 This needs to be overridden in derived classes in order to check op
133 self.proc = processor
135 self.cfg = context.cfg
136 self.glm = context.glm
138 self.owned_locks = context.glm.list_owned
139 self.context = context
140 self.rpc = rpc_runner
141 # Dicts used to declare locking needs to mcpu
142 self.needed_locks = None
143 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
145 self.remove_locks = {}
146 # Used to force good behavior when calling helper functions
147 self.recalculate_locks = {}
149 self.Log = processor.Log # pylint: disable=C0103
150 self.LogWarning = processor.LogWarning # pylint: disable=C0103
151 self.LogInfo = processor.LogInfo # pylint: disable=C0103
152 self.LogStep = processor.LogStep # pylint: disable=C0103
153 # support for dry-run
154 self.dry_run_result = None
155 # support for generic debug attribute
156 if (not hasattr(self.op, "debug_level") or
157 not isinstance(self.op.debug_level, int)):
158 self.op.debug_level = 0
163 # Validate opcode parameters and set defaults
164 self.op.Validate(True)
166 self.CheckArguments()
168 def CheckArguments(self):
169 """Check syntactic validity for the opcode arguments.
171 This method is for doing a simple syntactic check and ensure
172 validity of opcode parameters, without any cluster-related
173 checks. While the same can be accomplished in ExpandNames and/or
174 CheckPrereq, doing these separate is better because:
176 - ExpandNames is left as as purely a lock-related function
177 - CheckPrereq is run after we have acquired locks (and possible
180 The function is allowed to change the self.op attribute so that
181 later methods can no longer worry about missing parameters.
186 def ExpandNames(self):
187 """Expand names for this LU.
189 This method is called before starting to execute the opcode, and it should
190 update all the parameters of the opcode to their canonical form (e.g. a
191 short node name must be fully expanded after this method has successfully
192 completed). This way locking, hooks, logging, etc. can work correctly.
194 LUs which implement this method must also populate the self.needed_locks
195 member, as a dict with lock levels as keys, and a list of needed lock names
198 - use an empty dict if you don't need any lock
199 - if you don't need any lock at a particular level omit that
200 level (note that in this case C{DeclareLocks} won't be called
201 at all for that level)
202 - if you need locks at a level, but you can't calculate it in
203 this function, initialise that level with an empty list and do
204 further processing in L{LogicalUnit.DeclareLocks} (see that
205 function's docstring)
206 - don't put anything for the BGL level
207 - if you want all locks at a level use L{locking.ALL_SET} as a value
209 If you need to share locks (rather than acquire them exclusively) at one
210 level you can modify self.share_locks, setting a true value (usually 1) for
211 that level. By default locks are not shared.
213 This function can also define a list of tasklets, which then will be
214 executed in order instead of the usual LU-level CheckPrereq and Exec
215 functions, if those are not defined by the LU.
219 # Acquire all nodes and one instance
220 self.needed_locks = {
221 locking.LEVEL_NODE: locking.ALL_SET,
222 locking.LEVEL_INSTANCE: ['instance1.example.com'],
224 # Acquire just two nodes
225 self.needed_locks = {
226 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
229 self.needed_locks = {} # No, you can't leave it to the default value None
232 # The implementation of this method is mandatory only if the new LU is
233 # concurrent, so that old LUs don't need to be changed all at the same
236 self.needed_locks = {} # Exclusive LUs don't need locks.
238 raise NotImplementedError
240 def DeclareLocks(self, level):
241 """Declare LU locking needs for a level
243 While most LUs can just declare their locking needs at ExpandNames time,
244 sometimes there's the need to calculate some locks after having acquired
245 the ones before. This function is called just before acquiring locks at a
246 particular level, but after acquiring the ones at lower levels, and permits
247 such calculations. It can be used to modify self.needed_locks, and by
248 default it does nothing.
250 This function is only called if you have something already set in
251 self.needed_locks for the level.
253 @param level: Locking level which is going to be locked
254 @type level: member of L{ganeti.locking.LEVELS}
258 def CheckPrereq(self):
259 """Check prerequisites for this LU.
261 This method should check that the prerequisites for the execution
262 of this LU are fulfilled. It can do internode communication, but
263 it should be idempotent - no cluster or system changes are
266 The method should raise errors.OpPrereqError in case something is
267 not fulfilled. Its return value is ignored.
269 This method should also update all the parameters of the opcode to
270 their canonical form if it hasn't been done by ExpandNames before.
273 if self.tasklets is not None:
274 for (idx, tl) in enumerate(self.tasklets):
275 logging.debug("Checking prerequisites for tasklet %s/%s",
276 idx + 1, len(self.tasklets))
281 def Exec(self, feedback_fn):
284 This method should implement the actual work. It should raise
285 errors.OpExecError for failures that are somewhat dealt with in
289 if self.tasklets is not None:
290 for (idx, tl) in enumerate(self.tasklets):
291 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
294 raise NotImplementedError
296 def BuildHooksEnv(self):
297 """Build hooks environment for this LU.
300 @return: Dictionary containing the environment that will be used for
301 running the hooks for this LU. The keys of the dict must not be prefixed
302 with "GANETI_"--that'll be added by the hooks runner. The hooks runner
303 will extend the environment with additional variables. If no environment
304 should be defined, an empty dictionary should be returned (not C{None}).
305 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
309 raise NotImplementedError
311 def BuildHooksNodes(self):
312 """Build list of nodes to run LU's hooks.
314 @rtype: tuple; (list, list)
315 @return: Tuple containing a list of node names on which the hook
316 should run before the execution and a list of node names on which the
317 hook should run after the execution. No nodes should be returned as an
318 empty list (and not None).
319 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
323 raise NotImplementedError
325 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
326 """Notify the LU about the results of its hooks.
328 This method is called every time a hooks phase is executed, and notifies
329 the Logical Unit about the hooks' result. The LU can then use it to alter
330 its result based on the hooks. By default the method does nothing and the
331 previous result is passed back unchanged but any LU can define it if it
332 wants to use the local cluster hook-scripts somehow.
334 @param phase: one of L{constants.HOOKS_PHASE_POST} or
335 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
336 @param hook_results: the results of the multi-node hooks rpc call
337 @param feedback_fn: function used send feedback back to the caller
338 @param lu_result: the previous Exec result this LU had, or None
340 @return: the new Exec result, based on the previous result
344 # API must be kept, thus we ignore the unused argument and could
345 # be a function warnings
346 # pylint: disable=W0613,R0201
349 def _ExpandAndLockInstance(self):
350 """Helper function to expand and lock an instance.
352 Many LUs that work on an instance take its name in self.op.instance_name
353 and need to expand it and then declare the expanded name for locking. This
354 function does it, and then updates self.op.instance_name to the expanded
355 name. It also initializes needed_locks as a dict, if this hasn't been done
359 if self.needed_locks is None:
360 self.needed_locks = {}
362 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
363 "_ExpandAndLockInstance called with instance-level locks set"
364 self.op.instance_name = _ExpandInstanceName(self.cfg,
365 self.op.instance_name)
366 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
368 def _LockInstancesNodes(self, primary_only=False,
369 level=locking.LEVEL_NODE):
370 """Helper function to declare instances' nodes for locking.
372 This function should be called after locking one or more instances to lock
373 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
374 with all primary or secondary nodes for instances already locked and
375 present in self.needed_locks[locking.LEVEL_INSTANCE].
377 It should be called from DeclareLocks, and for safety only works if
378 self.recalculate_locks[locking.LEVEL_NODE] is set.
380 In the future it may grow parameters to just lock some instance's nodes, or
381 to just lock primaries or secondary nodes, if needed.
383 If should be called in DeclareLocks in a way similar to::
385 if level == locking.LEVEL_NODE:
386 self._LockInstancesNodes()
388 @type primary_only: boolean
389 @param primary_only: only lock primary nodes of locked instances
390 @param level: Which lock level to use for locking nodes
393 assert level in self.recalculate_locks, \
394 "_LockInstancesNodes helper function called with no nodes to recalculate"
396 # TODO: check if we're really been called with the instance locks held
398 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
399 # future we might want to have different behaviors depending on the value
400 # of self.recalculate_locks[locking.LEVEL_NODE]
402 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
403 for _, instance in self.cfg.GetMultiInstanceInfo(locked_i):
404 wanted_nodes.append(instance.primary_node)
406 wanted_nodes.extend(instance.secondary_nodes)
408 if self.recalculate_locks[level] == constants.LOCKS_REPLACE:
409 self.needed_locks[level] = wanted_nodes
410 elif self.recalculate_locks[level] == constants.LOCKS_APPEND:
411 self.needed_locks[level].extend(wanted_nodes)
413 raise errors.ProgrammerError("Unknown recalculation mode")
415 del self.recalculate_locks[level]
418 class NoHooksLU(LogicalUnit): # pylint: disable=W0223
419 """Simple LU which runs no hooks.
421 This LU is intended as a parent for other LogicalUnits which will
422 run no hooks, in order to reduce duplicate code.
428 def BuildHooksEnv(self):
429 """Empty BuildHooksEnv for NoHooksLu.
431 This just raises an error.
434 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
436 def BuildHooksNodes(self):
437 """Empty BuildHooksNodes for NoHooksLU.
440 raise AssertionError("BuildHooksNodes called for NoHooksLU")
444 """Tasklet base class.
446 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
447 they can mix legacy code with tasklets. Locking needs to be done in the LU,
448 tasklets know nothing about locks.
450 Subclasses must follow these rules:
451 - Implement CheckPrereq
455 def __init__(self, lu):
462 def CheckPrereq(self):
463 """Check prerequisites for this tasklets.
465 This method should check whether the prerequisites for the execution of
466 this tasklet are fulfilled. It can do internode communication, but it
467 should be idempotent - no cluster or system changes are allowed.
469 The method should raise errors.OpPrereqError in case something is not
470 fulfilled. Its return value is ignored.
472 This method should also update all parameters to their canonical form if it
473 hasn't been done before.
478 def Exec(self, feedback_fn):
479 """Execute the tasklet.
481 This method should implement the actual work. It should raise
482 errors.OpExecError for failures that are somewhat dealt with in code, or
486 raise NotImplementedError
490 """Base for query utility classes.
493 #: Attribute holding field definitions
496 def __init__(self, qfilter, fields, use_locking):
497 """Initializes this class.
500 self.use_locking = use_locking
502 self.query = query.Query(self.FIELDS, fields, qfilter=qfilter,
504 self.requested_data = self.query.RequestedData()
505 self.names = self.query.RequestedNames()
507 # Sort only if no names were requested
508 self.sort_by_name = not self.names
510 self.do_locking = None
513 def _GetNames(self, lu, all_names, lock_level):
514 """Helper function to determine names asked for in the query.
518 names = lu.owned_locks(lock_level)
522 if self.wanted == locking.ALL_SET:
523 assert not self.names
524 # caller didn't specify names, so ordering is not important
525 return utils.NiceSort(names)
527 # caller specified names and we must keep the same order
529 assert not self.do_locking or lu.glm.is_owned(lock_level)
531 missing = set(self.wanted).difference(names)
533 raise errors.OpExecError("Some items were removed before retrieving"
534 " their data: %s" % missing)
536 # Return expanded names
539 def ExpandNames(self, lu):
540 """Expand names for this query.
542 See L{LogicalUnit.ExpandNames}.
545 raise NotImplementedError()
547 def DeclareLocks(self, lu, level):
548 """Declare locks for this query.
550 See L{LogicalUnit.DeclareLocks}.
553 raise NotImplementedError()
555 def _GetQueryData(self, lu):
556 """Collects all data for this query.
558 @return: Query data object
561 raise NotImplementedError()
563 def NewStyleQuery(self, lu):
564 """Collect data and execute query.
567 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
568 sort_by_name=self.sort_by_name)
570 def OldStyleQuery(self, lu):
571 """Collect data and execute query.
574 return self.query.OldStyleQuery(self._GetQueryData(lu),
575 sort_by_name=self.sort_by_name)
579 """Returns a dict declaring all lock levels shared.
582 return dict.fromkeys(locking.LEVELS, 1)
585 def _MakeLegacyNodeInfo(data):
586 """Formats the data returned by L{rpc.RpcRunner.call_node_info}.
588 Converts the data into a single dictionary. This is fine for most use cases,
589 but some require information from more than one volume group or hypervisor.
592 (bootid, (vg_info, ), (hv_info, )) = data
594 return utils.JoinDisjointDicts(utils.JoinDisjointDicts(vg_info, hv_info), {
599 def _CheckInstanceNodeGroups(cfg, instance_name, owned_groups):
600 """Checks if the owned node groups are still correct for an instance.
602 @type cfg: L{config.ConfigWriter}
603 @param cfg: The cluster configuration
604 @type instance_name: string
605 @param instance_name: Instance name
606 @type owned_groups: set or frozenset
607 @param owned_groups: List of currently owned node groups
610 inst_groups = cfg.GetInstanceNodeGroups(instance_name)
612 if not owned_groups.issuperset(inst_groups):
613 raise errors.OpPrereqError("Instance %s's node groups changed since"
614 " locks were acquired, current groups are"
615 " are '%s', owning groups '%s'; retry the"
618 utils.CommaJoin(inst_groups),
619 utils.CommaJoin(owned_groups)),
625 def _CheckNodeGroupInstances(cfg, group_uuid, owned_instances):
626 """Checks if the instances in a node group are still correct.
628 @type cfg: L{config.ConfigWriter}
629 @param cfg: The cluster configuration
630 @type group_uuid: string
631 @param group_uuid: Node group UUID
632 @type owned_instances: set or frozenset
633 @param owned_instances: List of currently owned instances
636 wanted_instances = cfg.GetNodeGroupInstances(group_uuid)
637 if owned_instances != wanted_instances:
638 raise errors.OpPrereqError("Instances in node group '%s' changed since"
639 " locks were acquired, wanted '%s', have '%s';"
640 " retry the operation" %
642 utils.CommaJoin(wanted_instances),
643 utils.CommaJoin(owned_instances)),
646 return wanted_instances
649 def _SupportsOob(cfg, node):
650 """Tells if node supports OOB.
652 @type cfg: L{config.ConfigWriter}
653 @param cfg: The cluster configuration
654 @type node: L{objects.Node}
655 @param node: The node
656 @return: The OOB script if supported or an empty string otherwise
659 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
662 def _GetWantedNodes(lu, nodes):
663 """Returns list of checked and expanded node names.
665 @type lu: L{LogicalUnit}
666 @param lu: the logical unit on whose behalf we execute
668 @param nodes: list of node names or None for all nodes
670 @return: the list of nodes, sorted
671 @raise errors.ProgrammerError: if the nodes parameter is wrong type
675 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
677 return utils.NiceSort(lu.cfg.GetNodeList())
680 def _GetWantedInstances(lu, instances):
681 """Returns list of checked and expanded instance names.
683 @type lu: L{LogicalUnit}
684 @param lu: the logical unit on whose behalf we execute
685 @type instances: list
686 @param instances: list of instance names or None for all instances
688 @return: the list of instances, sorted
689 @raise errors.OpPrereqError: if the instances parameter is wrong type
690 @raise errors.OpPrereqError: if any of the passed instances is not found
694 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
696 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
700 def _GetUpdatedParams(old_params, update_dict,
701 use_default=True, use_none=False):
702 """Return the new version of a parameter dictionary.
704 @type old_params: dict
705 @param old_params: old parameters
706 @type update_dict: dict
707 @param update_dict: dict containing new parameter values, or
708 constants.VALUE_DEFAULT to reset the parameter to its default
710 @param use_default: boolean
711 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
712 values as 'to be deleted' values
713 @param use_none: boolean
714 @type use_none: whether to recognise C{None} values as 'to be
717 @return: the new parameter dictionary
720 params_copy = copy.deepcopy(old_params)
721 for key, val in update_dict.iteritems():
722 if ((use_default and val == constants.VALUE_DEFAULT) or
723 (use_none and val is None)):
729 params_copy[key] = val
733 def _GetUpdatedIPolicy(old_ipolicy, new_ipolicy, group_policy=False):
734 """Return the new version of a instance policy.
736 @param group_policy: whether this policy applies to a group and thus
737 we should support removal of policy entries
740 use_none = use_default = group_policy
741 ipolicy = copy.deepcopy(old_ipolicy)
742 for key, value in new_ipolicy.items():
743 if key not in constants.IPOLICY_ALL_KEYS:
744 raise errors.OpPrereqError("Invalid key in new ipolicy: %s" % key,
746 if key in constants.IPOLICY_ISPECS:
747 utils.ForceDictType(value, constants.ISPECS_PARAMETER_TYPES)
748 ipolicy[key] = _GetUpdatedParams(old_ipolicy.get(key, {}), value,
750 use_default=use_default)
752 if not value or value == [constants.VALUE_DEFAULT]:
756 raise errors.OpPrereqError("Can't unset ipolicy attribute '%s'"
757 " on the cluster'" % key,
760 if key in constants.IPOLICY_PARAMETERS:
761 # FIXME: we assume all such values are float
763 ipolicy[key] = float(value)
764 except (TypeError, ValueError), err:
765 raise errors.OpPrereqError("Invalid value for attribute"
766 " '%s': '%s', error: %s" %
767 (key, value, err), errors.ECODE_INVAL)
769 # FIXME: we assume all others are lists; this should be redone
771 ipolicy[key] = list(value)
773 objects.InstancePolicy.CheckParameterSyntax(ipolicy)
774 except errors.ConfigurationError, err:
775 raise errors.OpPrereqError("Invalid instance policy: %s" % err,
780 def _UpdateAndVerifySubDict(base, updates, type_check):
781 """Updates and verifies a dict with sub dicts of the same type.
783 @param base: The dict with the old data
784 @param updates: The dict with the new data
785 @param type_check: Dict suitable to ForceDictType to verify correct types
786 @returns: A new dict with updated and verified values
790 new = _GetUpdatedParams(old, value)
791 utils.ForceDictType(new, type_check)
794 ret = copy.deepcopy(base)
795 ret.update(dict((key, fn(base.get(key, {}), value))
796 for key, value in updates.items()))
800 def _MergeAndVerifyHvState(op_input, obj_input):
801 """Combines the hv state from an opcode with the one of the object
803 @param op_input: The input dict from the opcode
804 @param obj_input: The input dict from the objects
805 @return: The verified and updated dict
809 invalid_hvs = set(op_input) - constants.HYPER_TYPES
811 raise errors.OpPrereqError("Invalid hypervisor(s) in hypervisor state:"
812 " %s" % utils.CommaJoin(invalid_hvs),
814 if obj_input is None:
816 type_check = constants.HVSTS_PARAMETER_TYPES
817 return _UpdateAndVerifySubDict(obj_input, op_input, type_check)
822 def _MergeAndVerifyDiskState(op_input, obj_input):
823 """Combines the disk state from an opcode with the one of the object
825 @param op_input: The input dict from the opcode
826 @param obj_input: The input dict from the objects
827 @return: The verified and updated dict
830 invalid_dst = set(op_input) - constants.DS_VALID_TYPES
832 raise errors.OpPrereqError("Invalid storage type(s) in disk state: %s" %
833 utils.CommaJoin(invalid_dst),
835 type_check = constants.DSS_PARAMETER_TYPES
836 if obj_input is None:
838 return dict((key, _UpdateAndVerifySubDict(obj_input.get(key, {}), value,
840 for key, value in op_input.items())
845 def _ReleaseLocks(lu, level, names=None, keep=None):
846 """Releases locks owned by an LU.
848 @type lu: L{LogicalUnit}
849 @param level: Lock level
850 @type names: list or None
851 @param names: Names of locks to release
852 @type keep: list or None
853 @param keep: Names of locks to retain
856 assert not (keep is not None and names is not None), \
857 "Only one of the 'names' and the 'keep' parameters can be given"
859 if names is not None:
860 should_release = names.__contains__
862 should_release = lambda name: name not in keep
864 should_release = None
866 owned = lu.owned_locks(level)
868 # Not owning any lock at this level, do nothing
875 # Determine which locks to release
877 if should_release(name):
882 assert len(lu.owned_locks(level)) == (len(retain) + len(release))
884 # Release just some locks
885 lu.glm.release(level, names=release)
887 assert frozenset(lu.owned_locks(level)) == frozenset(retain)
890 lu.glm.release(level)
892 assert not lu.glm.is_owned(level), "No locks should be owned"
895 def _MapInstanceDisksToNodes(instances):
896 """Creates a map from (node, volume) to instance name.
898 @type instances: list of L{objects.Instance}
899 @rtype: dict; tuple of (node name, volume name) as key, instance name as value
902 return dict(((node, vol), inst.name)
903 for inst in instances
904 for (node, vols) in inst.MapLVsByNode().items()
908 def _RunPostHook(lu, node_name):
909 """Runs the post-hook for an opcode on a single node.
912 hm = lu.proc.BuildHooksManager(lu)
914 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
916 # pylint: disable=W0702
917 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
920 def _CheckOutputFields(static, dynamic, selected):
921 """Checks whether all selected fields are valid.
923 @type static: L{utils.FieldSet}
924 @param static: static fields set
925 @type dynamic: L{utils.FieldSet}
926 @param dynamic: dynamic fields set
933 delta = f.NonMatching(selected)
935 raise errors.OpPrereqError("Unknown output fields selected: %s"
936 % ",".join(delta), errors.ECODE_INVAL)
939 def _CheckGlobalHvParams(params):
940 """Validates that given hypervisor params are not global ones.
942 This will ensure that instances don't get customised versions of
946 used_globals = constants.HVC_GLOBALS.intersection(params)
948 msg = ("The following hypervisor parameters are global and cannot"
949 " be customized at instance level, please modify them at"
950 " cluster level: %s" % utils.CommaJoin(used_globals))
951 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
954 def _CheckNodeOnline(lu, node, msg=None):
955 """Ensure that a given node is online.
957 @param lu: the LU on behalf of which we make the check
958 @param node: the node to check
959 @param msg: if passed, should be a message to replace the default one
960 @raise errors.OpPrereqError: if the node is offline
964 msg = "Can't use offline node"
965 if lu.cfg.GetNodeInfo(node).offline:
966 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
969 def _CheckNodeNotDrained(lu, node):
970 """Ensure that a given node is not drained.
972 @param lu: the LU on behalf of which we make the check
973 @param node: the node to check
974 @raise errors.OpPrereqError: if the node is drained
977 if lu.cfg.GetNodeInfo(node).drained:
978 raise errors.OpPrereqError("Can't use drained node %s" % node,
982 def _CheckNodeVmCapable(lu, node):
983 """Ensure that a given node is vm capable.
985 @param lu: the LU on behalf of which we make the check
986 @param node: the node to check
987 @raise errors.OpPrereqError: if the node is not vm capable
990 if not lu.cfg.GetNodeInfo(node).vm_capable:
991 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
995 def _CheckNodeHasOS(lu, node, os_name, force_variant):
996 """Ensure that a node supports a given OS.
998 @param lu: the LU on behalf of which we make the check
999 @param node: the node to check
1000 @param os_name: the OS to query about
1001 @param force_variant: whether to ignore variant errors
1002 @raise errors.OpPrereqError: if the node is not supporting the OS
1005 result = lu.rpc.call_os_get(node, os_name)
1006 result.Raise("OS '%s' not in supported OS list for node %s" %
1008 prereq=True, ecode=errors.ECODE_INVAL)
1009 if not force_variant:
1010 _CheckOSVariant(result.payload, os_name)
1013 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
1014 """Ensure that a node has the given secondary ip.
1016 @type lu: L{LogicalUnit}
1017 @param lu: the LU on behalf of which we make the check
1019 @param node: the node to check
1020 @type secondary_ip: string
1021 @param secondary_ip: the ip to check
1022 @type prereq: boolean
1023 @param prereq: whether to throw a prerequisite or an execute error
1024 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
1025 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
1028 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
1029 result.Raise("Failure checking secondary ip on node %s" % node,
1030 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1031 if not result.payload:
1032 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
1033 " please fix and re-run this command" % secondary_ip)
1035 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
1037 raise errors.OpExecError(msg)
1040 def _GetClusterDomainSecret():
1041 """Reads the cluster domain secret.
1044 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
1048 def _CheckInstanceState(lu, instance, req_states, msg=None):
1049 """Ensure that an instance is in one of the required states.
1051 @param lu: the LU on behalf of which we make the check
1052 @param instance: the instance to check
1053 @param msg: if passed, should be a message to replace the default one
1054 @raise errors.OpPrereqError: if the instance is not in the required state
1058 msg = "can't use instance from outside %s states" % ", ".join(req_states)
1059 if instance.admin_state not in req_states:
1060 raise errors.OpPrereqError("Instance '%s' is marked to be %s, %s" %
1061 (instance.name, instance.admin_state, msg),
1064 if constants.ADMINST_UP not in req_states:
1065 pnode = instance.primary_node
1066 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
1067 ins_l.Raise("Can't contact node %s for instance information" % pnode,
1068 prereq=True, ecode=errors.ECODE_ENVIRON)
1070 if instance.name in ins_l.payload:
1071 raise errors.OpPrereqError("Instance %s is running, %s" %
1072 (instance.name, msg), errors.ECODE_STATE)
1075 def _ComputeMinMaxSpec(name, ipolicy, value):
1076 """Computes if value is in the desired range.
1078 @param name: name of the parameter for which we perform the check
1079 @param ipolicy: dictionary containing min, max and std values
1080 @param value: actual value that we want to use
1081 @return: None or element not meeting the criteria
1085 if value in [None, constants.VALUE_AUTO]:
1087 max_v = ipolicy[constants.ISPECS_MAX].get(name, value)
1088 min_v = ipolicy[constants.ISPECS_MIN].get(name, value)
1089 if value > max_v or min_v > value:
1090 return ("%s value %s is not in range [%s, %s]" %
1091 (name, value, min_v, max_v))
1095 def _ComputeIPolicySpecViolation(ipolicy, mem_size, cpu_count, disk_count,
1096 nic_count, disk_sizes,
1097 _compute_fn=_ComputeMinMaxSpec):
1098 """Verifies ipolicy against provided specs.
1101 @param ipolicy: The ipolicy
1103 @param mem_size: The memory size
1104 @type cpu_count: int
1105 @param cpu_count: Used cpu cores
1106 @type disk_count: int
1107 @param disk_count: Number of disks used
1108 @type nic_count: int
1109 @param nic_count: Number of nics used
1110 @type disk_sizes: list of ints
1111 @param disk_sizes: Disk sizes of used disk (len must match C{disk_count})
1112 @param _compute_fn: The compute function (unittest only)
1113 @return: A list of violations, or an empty list of no violations are found
1116 assert disk_count == len(disk_sizes)
1119 (constants.ISPEC_MEM_SIZE, mem_size),
1120 (constants.ISPEC_CPU_COUNT, cpu_count),
1121 (constants.ISPEC_DISK_COUNT, disk_count),
1122 (constants.ISPEC_NIC_COUNT, nic_count),
1123 ] + map((lambda d: (constants.ISPEC_DISK_SIZE, d)), disk_sizes)
1126 (_compute_fn(name, ipolicy, value)
1127 for (name, value) in test_settings))
1130 def _ComputeIPolicyInstanceViolation(ipolicy, instance,
1131 _compute_fn=_ComputeIPolicySpecViolation):
1132 """Compute if instance meets the specs of ipolicy.
1135 @param ipolicy: The ipolicy to verify against
1136 @type instance: L{objects.Instance}
1137 @param instance: The instance to verify
1138 @param _compute_fn: The function to verify ipolicy (unittest only)
1139 @see: L{_ComputeIPolicySpecViolation}
1142 mem_size = instance.beparams.get(constants.BE_MAXMEM, None)
1143 cpu_count = instance.beparams.get(constants.BE_VCPUS, None)
1144 disk_count = len(instance.disks)
1145 disk_sizes = [disk.size for disk in instance.disks]
1146 nic_count = len(instance.nics)
1148 return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
1152 def _ComputeIPolicyInstanceSpecViolation(ipolicy, instance_spec,
1153 _compute_fn=_ComputeIPolicySpecViolation):
1154 """Compute if instance specs meets the specs of ipolicy.
1157 @param ipolicy: The ipolicy to verify against
1158 @param instance_spec: dict
1159 @param instance_spec: The instance spec to verify
1160 @param _compute_fn: The function to verify ipolicy (unittest only)
1161 @see: L{_ComputeIPolicySpecViolation}
1164 mem_size = instance_spec.get(constants.ISPEC_MEM_SIZE, None)
1165 cpu_count = instance_spec.get(constants.ISPEC_CPU_COUNT, None)
1166 disk_count = instance_spec.get(constants.ISPEC_DISK_COUNT, 0)
1167 disk_sizes = instance_spec.get(constants.ISPEC_DISK_SIZE, [])
1168 nic_count = instance_spec.get(constants.ISPEC_NIC_COUNT, 0)
1170 return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
1174 def _ComputeIPolicyNodeViolation(ipolicy, instance, current_group,
1176 _compute_fn=_ComputeIPolicyInstanceViolation):
1177 """Compute if instance meets the specs of the new target group.
1179 @param ipolicy: The ipolicy to verify
1180 @param instance: The instance object to verify
1181 @param current_group: The current group of the instance
1182 @param target_group: The new group of the instance
1183 @param _compute_fn: The function to verify ipolicy (unittest only)
1184 @see: L{_ComputeIPolicySpecViolation}
1187 if current_group == target_group:
1190 return _compute_fn(ipolicy, instance)
1193 def _CheckTargetNodeIPolicy(lu, ipolicy, instance, node, ignore=False,
1194 _compute_fn=_ComputeIPolicyNodeViolation):
1195 """Checks that the target node is correct in terms of instance policy.
1197 @param ipolicy: The ipolicy to verify
1198 @param instance: The instance object to verify
1199 @param node: The new node to relocate
1200 @param ignore: Ignore violations of the ipolicy
1201 @param _compute_fn: The function to verify ipolicy (unittest only)
1202 @see: L{_ComputeIPolicySpecViolation}
1205 primary_node = lu.cfg.GetNodeInfo(instance.primary_node)
1206 res = _compute_fn(ipolicy, instance, primary_node.group, node.group)
1209 msg = ("Instance does not meet target node group's (%s) instance"
1210 " policy: %s") % (node.group, utils.CommaJoin(res))
1214 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
1217 def _ComputeNewInstanceViolations(old_ipolicy, new_ipolicy, instances):
1218 """Computes a set of any instances that would violate the new ipolicy.
1220 @param old_ipolicy: The current (still in-place) ipolicy
1221 @param new_ipolicy: The new (to become) ipolicy
1222 @param instances: List of instances to verify
1223 @return: A list of instances which violates the new ipolicy but did not before
1226 return (_ComputeViolatingInstances(old_ipolicy, instances) -
1227 _ComputeViolatingInstances(new_ipolicy, instances))
1230 def _ExpandItemName(fn, name, kind):
1231 """Expand an item name.
1233 @param fn: the function to use for expansion
1234 @param name: requested item name
1235 @param kind: text description ('Node' or 'Instance')
1236 @return: the resolved (full) name
1237 @raise errors.OpPrereqError: if the item is not found
1240 full_name = fn(name)
1241 if full_name is None:
1242 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
1247 def _ExpandNodeName(cfg, name):
1248 """Wrapper over L{_ExpandItemName} for nodes."""
1249 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
1252 def _ExpandInstanceName(cfg, name):
1253 """Wrapper over L{_ExpandItemName} for instance."""
1254 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
1257 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
1258 minmem, maxmem, vcpus, nics, disk_template, disks,
1259 bep, hvp, hypervisor_name, tags):
1260 """Builds instance related env variables for hooks
1262 This builds the hook environment from individual variables.
1265 @param name: the name of the instance
1266 @type primary_node: string
1267 @param primary_node: the name of the instance's primary node
1268 @type secondary_nodes: list
1269 @param secondary_nodes: list of secondary nodes as strings
1270 @type os_type: string
1271 @param os_type: the name of the instance's OS
1272 @type status: string
1273 @param status: the desired status of the instance
1274 @type minmem: string
1275 @param minmem: the minimum memory size of the instance
1276 @type maxmem: string
1277 @param maxmem: the maximum memory size of the instance
1279 @param vcpus: the count of VCPUs the instance has
1281 @param nics: list of tuples (ip, mac, mode, link) representing
1282 the NICs the instance has
1283 @type disk_template: string
1284 @param disk_template: the disk template of the instance
1286 @param disks: the list of (size, mode) pairs
1288 @param bep: the backend parameters for the instance
1290 @param hvp: the hypervisor parameters for the instance
1291 @type hypervisor_name: string
1292 @param hypervisor_name: the hypervisor for the instance
1294 @param tags: list of instance tags as strings
1296 @return: the hook environment for this instance
1301 "INSTANCE_NAME": name,
1302 "INSTANCE_PRIMARY": primary_node,
1303 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
1304 "INSTANCE_OS_TYPE": os_type,
1305 "INSTANCE_STATUS": status,
1306 "INSTANCE_MINMEM": minmem,
1307 "INSTANCE_MAXMEM": maxmem,
1308 # TODO(2.7) remove deprecated "memory" value
1309 "INSTANCE_MEMORY": maxmem,
1310 "INSTANCE_VCPUS": vcpus,
1311 "INSTANCE_DISK_TEMPLATE": disk_template,
1312 "INSTANCE_HYPERVISOR": hypervisor_name,
1315 nic_count = len(nics)
1316 for idx, (ip, mac, mode, link) in enumerate(nics):
1319 env["INSTANCE_NIC%d_IP" % idx] = ip
1320 env["INSTANCE_NIC%d_MAC" % idx] = mac
1321 env["INSTANCE_NIC%d_MODE" % idx] = mode
1322 env["INSTANCE_NIC%d_LINK" % idx] = link
1323 if mode == constants.NIC_MODE_BRIDGED:
1324 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
1328 env["INSTANCE_NIC_COUNT"] = nic_count
1331 disk_count = len(disks)
1332 for idx, (size, mode) in enumerate(disks):
1333 env["INSTANCE_DISK%d_SIZE" % idx] = size
1334 env["INSTANCE_DISK%d_MODE" % idx] = mode
1338 env["INSTANCE_DISK_COUNT"] = disk_count
1343 env["INSTANCE_TAGS"] = " ".join(tags)
1345 for source, kind in [(bep, "BE"), (hvp, "HV")]:
1346 for key, value in source.items():
1347 env["INSTANCE_%s_%s" % (kind, key)] = value
1352 def _NICListToTuple(lu, nics):
1353 """Build a list of nic information tuples.
1355 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
1356 value in LUInstanceQueryData.
1358 @type lu: L{LogicalUnit}
1359 @param lu: the logical unit on whose behalf we execute
1360 @type nics: list of L{objects.NIC}
1361 @param nics: list of nics to convert to hooks tuples
1365 cluster = lu.cfg.GetClusterInfo()
1369 filled_params = cluster.SimpleFillNIC(nic.nicparams)
1370 mode = filled_params[constants.NIC_MODE]
1371 link = filled_params[constants.NIC_LINK]
1372 hooks_nics.append((ip, mac, mode, link))
1376 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1377 """Builds instance related env variables for hooks from an object.
1379 @type lu: L{LogicalUnit}
1380 @param lu: the logical unit on whose behalf we execute
1381 @type instance: L{objects.Instance}
1382 @param instance: the instance for which we should build the
1384 @type override: dict
1385 @param override: dictionary with key/values that will override
1388 @return: the hook environment dictionary
1391 cluster = lu.cfg.GetClusterInfo()
1392 bep = cluster.FillBE(instance)
1393 hvp = cluster.FillHV(instance)
1395 "name": instance.name,
1396 "primary_node": instance.primary_node,
1397 "secondary_nodes": instance.secondary_nodes,
1398 "os_type": instance.os,
1399 "status": instance.admin_state,
1400 "maxmem": bep[constants.BE_MAXMEM],
1401 "minmem": bep[constants.BE_MINMEM],
1402 "vcpus": bep[constants.BE_VCPUS],
1403 "nics": _NICListToTuple(lu, instance.nics),
1404 "disk_template": instance.disk_template,
1405 "disks": [(disk.size, disk.mode) for disk in instance.disks],
1408 "hypervisor_name": instance.hypervisor,
1409 "tags": instance.tags,
1412 args.update(override)
1413 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
1416 def _AdjustCandidatePool(lu, exceptions):
1417 """Adjust the candidate pool after node operations.
1420 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1422 lu.LogInfo("Promoted nodes to master candidate role: %s",
1423 utils.CommaJoin(node.name for node in mod_list))
1424 for name in mod_list:
1425 lu.context.ReaddNode(name)
1426 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1428 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1432 def _DecideSelfPromotion(lu, exceptions=None):
1433 """Decide whether I should promote myself as a master candidate.
1436 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1437 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1438 # the new node will increase mc_max with one, so:
1439 mc_should = min(mc_should + 1, cp_size)
1440 return mc_now < mc_should
1443 def _CalculateGroupIPolicy(cluster, group):
1444 """Calculate instance policy for group.
1447 return cluster.SimpleFillIPolicy(group.ipolicy)
1450 def _ComputeViolatingInstances(ipolicy, instances):
1451 """Computes a set of instances who violates given ipolicy.
1453 @param ipolicy: The ipolicy to verify
1454 @type instances: object.Instance
1455 @param instances: List of instances to verify
1456 @return: A frozenset of instance names violating the ipolicy
1459 return frozenset([inst.name for inst in instances
1460 if _ComputeIPolicyInstanceViolation(ipolicy, inst)])
1463 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1464 """Check that the brigdes needed by a list of nics exist.
1467 cluster = lu.cfg.GetClusterInfo()
1468 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1469 brlist = [params[constants.NIC_LINK] for params in paramslist
1470 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1472 result = lu.rpc.call_bridges_exist(target_node, brlist)
1473 result.Raise("Error checking bridges on destination node '%s'" %
1474 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1477 def _CheckInstanceBridgesExist(lu, instance, node=None):
1478 """Check that the brigdes needed by an instance exist.
1482 node = instance.primary_node
1483 _CheckNicsBridgesExist(lu, instance.nics, node)
1486 def _CheckOSVariant(os_obj, name):
1487 """Check whether an OS name conforms to the os variants specification.
1489 @type os_obj: L{objects.OS}
1490 @param os_obj: OS object to check
1492 @param name: OS name passed by the user, to check for validity
1495 variant = objects.OS.GetVariant(name)
1496 if not os_obj.supported_variants:
1498 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'"
1499 " passed)" % (os_obj.name, variant),
1503 raise errors.OpPrereqError("OS name must include a variant",
1506 if variant not in os_obj.supported_variants:
1507 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1510 def _GetNodeInstancesInner(cfg, fn):
1511 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1514 def _GetNodeInstances(cfg, node_name):
1515 """Returns a list of all primary and secondary instances on a node.
1519 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1522 def _GetNodePrimaryInstances(cfg, node_name):
1523 """Returns primary instances on a node.
1526 return _GetNodeInstancesInner(cfg,
1527 lambda inst: node_name == inst.primary_node)
1530 def _GetNodeSecondaryInstances(cfg, node_name):
1531 """Returns secondary instances on a node.
1534 return _GetNodeInstancesInner(cfg,
1535 lambda inst: node_name in inst.secondary_nodes)
1538 def _GetStorageTypeArgs(cfg, storage_type):
1539 """Returns the arguments for a storage type.
1542 # Special case for file storage
1543 if storage_type == constants.ST_FILE:
1544 # storage.FileStorage wants a list of storage directories
1545 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1550 def _FindFaultyInstanceDisks(cfg, rpc_runner, instance, node_name, prereq):
1553 for dev in instance.disks:
1554 cfg.SetDiskID(dev, node_name)
1556 result = rpc_runner.call_blockdev_getmirrorstatus(node_name, instance.disks)
1557 result.Raise("Failed to get disk status from node %s" % node_name,
1558 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1560 for idx, bdev_status in enumerate(result.payload):
1561 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1567 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1568 """Check the sanity of iallocator and node arguments and use the
1569 cluster-wide iallocator if appropriate.
1571 Check that at most one of (iallocator, node) is specified. If none is
1572 specified, then the LU's opcode's iallocator slot is filled with the
1573 cluster-wide default iallocator.
1575 @type iallocator_slot: string
1576 @param iallocator_slot: the name of the opcode iallocator slot
1577 @type node_slot: string
1578 @param node_slot: the name of the opcode target node slot
1581 node = getattr(lu.op, node_slot, None)
1582 iallocator = getattr(lu.op, iallocator_slot, None)
1584 if node is not None and iallocator is not None:
1585 raise errors.OpPrereqError("Do not specify both, iallocator and node",
1587 elif node is None and iallocator is None:
1588 default_iallocator = lu.cfg.GetDefaultIAllocator()
1589 if default_iallocator:
1590 setattr(lu.op, iallocator_slot, default_iallocator)
1592 raise errors.OpPrereqError("No iallocator or node given and no"
1593 " cluster-wide default iallocator found;"
1594 " please specify either an iallocator or a"
1595 " node, or set a cluster-wide default"
1599 def _GetDefaultIAllocator(cfg, iallocator):
1600 """Decides on which iallocator to use.
1602 @type cfg: L{config.ConfigWriter}
1603 @param cfg: Cluster configuration object
1604 @type iallocator: string or None
1605 @param iallocator: Iallocator specified in opcode
1607 @return: Iallocator name
1611 # Use default iallocator
1612 iallocator = cfg.GetDefaultIAllocator()
1615 raise errors.OpPrereqError("No iallocator was specified, neither in the"
1616 " opcode nor as a cluster-wide default",
1622 class LUClusterPostInit(LogicalUnit):
1623 """Logical unit for running hooks after cluster initialization.
1626 HPATH = "cluster-init"
1627 HTYPE = constants.HTYPE_CLUSTER
1629 def BuildHooksEnv(self):
1634 "OP_TARGET": self.cfg.GetClusterName(),
1637 def BuildHooksNodes(self):
1638 """Build hooks nodes.
1641 return ([], [self.cfg.GetMasterNode()])
1643 def Exec(self, feedback_fn):
1650 class LUClusterDestroy(LogicalUnit):
1651 """Logical unit for destroying the cluster.
1654 HPATH = "cluster-destroy"
1655 HTYPE = constants.HTYPE_CLUSTER
1657 def BuildHooksEnv(self):
1662 "OP_TARGET": self.cfg.GetClusterName(),
1665 def BuildHooksNodes(self):
1666 """Build hooks nodes.
1671 def CheckPrereq(self):
1672 """Check prerequisites.
1674 This checks whether the cluster is empty.
1676 Any errors are signaled by raising errors.OpPrereqError.
1679 master = self.cfg.GetMasterNode()
1681 nodelist = self.cfg.GetNodeList()
1682 if len(nodelist) != 1 or nodelist[0] != master:
1683 raise errors.OpPrereqError("There are still %d node(s) in"
1684 " this cluster." % (len(nodelist) - 1),
1686 instancelist = self.cfg.GetInstanceList()
1688 raise errors.OpPrereqError("There are still %d instance(s) in"
1689 " this cluster." % len(instancelist),
1692 def Exec(self, feedback_fn):
1693 """Destroys the cluster.
1696 master_params = self.cfg.GetMasterNetworkParameters()
1698 # Run post hooks on master node before it's removed
1699 _RunPostHook(self, master_params.name)
1701 ems = self.cfg.GetUseExternalMipScript()
1702 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
1705 self.LogWarning("Error disabling the master IP address: %s",
1708 return master_params.name
1711 def _VerifyCertificate(filename):
1712 """Verifies a certificate for L{LUClusterVerifyConfig}.
1714 @type filename: string
1715 @param filename: Path to PEM file
1719 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1720 utils.ReadFile(filename))
1721 except Exception, err: # pylint: disable=W0703
1722 return (LUClusterVerifyConfig.ETYPE_ERROR,
1723 "Failed to load X509 certificate %s: %s" % (filename, err))
1726 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1727 constants.SSL_CERT_EXPIRATION_ERROR)
1730 fnamemsg = "While verifying %s: %s" % (filename, msg)
1735 return (None, fnamemsg)
1736 elif errcode == utils.CERT_WARNING:
1737 return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg)
1738 elif errcode == utils.CERT_ERROR:
1739 return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg)
1741 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1744 def _GetAllHypervisorParameters(cluster, instances):
1745 """Compute the set of all hypervisor parameters.
1747 @type cluster: L{objects.Cluster}
1748 @param cluster: the cluster object
1749 @param instances: list of L{objects.Instance}
1750 @param instances: additional instances from which to obtain parameters
1751 @rtype: list of (origin, hypervisor, parameters)
1752 @return: a list with all parameters found, indicating the hypervisor they
1753 apply to, and the origin (can be "cluster", "os X", or "instance Y")
1758 for hv_name in cluster.enabled_hypervisors:
1759 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
1761 for os_name, os_hvp in cluster.os_hvp.items():
1762 for hv_name, hv_params in os_hvp.items():
1764 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
1765 hvp_data.append(("os %s" % os_name, hv_name, full_params))
1767 # TODO: collapse identical parameter values in a single one
1768 for instance in instances:
1769 if instance.hvparams:
1770 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
1771 cluster.FillHV(instance)))
1776 class _VerifyErrors(object):
1777 """Mix-in for cluster/group verify LUs.
1779 It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects
1780 self.op and self._feedback_fn to be available.)
1784 ETYPE_FIELD = "code"
1785 ETYPE_ERROR = "ERROR"
1786 ETYPE_WARNING = "WARNING"
1788 def _Error(self, ecode, item, msg, *args, **kwargs):
1789 """Format an error message.
1791 Based on the opcode's error_codes parameter, either format a
1792 parseable error code, or a simpler error string.
1794 This must be called only from Exec and functions called from Exec.
1797 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1798 itype, etxt, _ = ecode
1799 # first complete the msg
1802 # then format the whole message
1803 if self.op.error_codes: # This is a mix-in. pylint: disable=E1101
1804 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1810 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1811 # and finally report it via the feedback_fn
1812 self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable=E1101
1814 def _ErrorIf(self, cond, ecode, *args, **kwargs):
1815 """Log an error message if the passed condition is True.
1819 or self.op.debug_simulate_errors) # pylint: disable=E1101
1821 # If the error code is in the list of ignored errors, demote the error to a
1823 (_, etxt, _) = ecode
1824 if etxt in self.op.ignore_errors: # pylint: disable=E1101
1825 kwargs[self.ETYPE_FIELD] = self.ETYPE_WARNING
1828 self._Error(ecode, *args, **kwargs)
1830 # do not mark the operation as failed for WARN cases only
1831 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1832 self.bad = self.bad or cond
1835 class LUClusterVerify(NoHooksLU):
1836 """Submits all jobs necessary to verify the cluster.
1841 def ExpandNames(self):
1842 self.needed_locks = {}
1844 def Exec(self, feedback_fn):
1847 if self.op.group_name:
1848 groups = [self.op.group_name]
1849 depends_fn = lambda: None
1851 groups = self.cfg.GetNodeGroupList()
1853 # Verify global configuration
1855 opcodes.OpClusterVerifyConfig(ignore_errors=self.op.ignore_errors)
1858 # Always depend on global verification
1859 depends_fn = lambda: [(-len(jobs), [])]
1861 jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,
1862 ignore_errors=self.op.ignore_errors,
1863 depends=depends_fn())]
1864 for group in groups)
1866 # Fix up all parameters
1867 for op in itertools.chain(*jobs): # pylint: disable=W0142
1868 op.debug_simulate_errors = self.op.debug_simulate_errors
1869 op.verbose = self.op.verbose
1870 op.error_codes = self.op.error_codes
1872 op.skip_checks = self.op.skip_checks
1873 except AttributeError:
1874 assert not isinstance(op, opcodes.OpClusterVerifyGroup)
1876 return ResultWithJobs(jobs)
1879 class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
1880 """Verifies the cluster config.
1885 def _VerifyHVP(self, hvp_data):
1886 """Verifies locally the syntax of the hypervisor parameters.
1889 for item, hv_name, hv_params in hvp_data:
1890 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
1893 hv_class = hypervisor.GetHypervisor(hv_name)
1894 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1895 hv_class.CheckParameterSyntax(hv_params)
1896 except errors.GenericError, err:
1897 self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg % str(err))
1899 def ExpandNames(self):
1900 # Information can be safely retrieved as the BGL is acquired in exclusive
1902 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER)
1903 self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
1904 self.all_node_info = self.cfg.GetAllNodesInfo()
1905 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1906 self.needed_locks = {}
1908 def Exec(self, feedback_fn):
1909 """Verify integrity of cluster, performing various test on nodes.
1913 self._feedback_fn = feedback_fn
1915 feedback_fn("* Verifying cluster config")
1917 for msg in self.cfg.VerifyConfig():
1918 self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg)
1920 feedback_fn("* Verifying cluster certificate files")
1922 for cert_filename in constants.ALL_CERT_FILES:
1923 (errcode, msg) = _VerifyCertificate(cert_filename)
1924 self._ErrorIf(errcode, constants.CV_ECLUSTERCERT, None, msg, code=errcode)
1926 feedback_fn("* Verifying hypervisor parameters")
1928 self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(),
1929 self.all_inst_info.values()))
1931 feedback_fn("* Verifying all nodes belong to an existing group")
1933 # We do this verification here because, should this bogus circumstance
1934 # occur, it would never be caught by VerifyGroup, which only acts on
1935 # nodes/instances reachable from existing node groups.
1937 dangling_nodes = set(node.name for node in self.all_node_info.values()
1938 if node.group not in self.all_group_info)
1940 dangling_instances = {}
1941 no_node_instances = []
1943 for inst in self.all_inst_info.values():
1944 if inst.primary_node in dangling_nodes:
1945 dangling_instances.setdefault(inst.primary_node, []).append(inst.name)
1946 elif inst.primary_node not in self.all_node_info:
1947 no_node_instances.append(inst.name)
1952 utils.CommaJoin(dangling_instances.get(node.name,
1954 for node in dangling_nodes]
1956 self._ErrorIf(bool(dangling_nodes), constants.CV_ECLUSTERDANGLINGNODES,
1958 "the following nodes (and their instances) belong to a non"
1959 " existing group: %s", utils.CommaJoin(pretty_dangling))
1961 self._ErrorIf(bool(no_node_instances), constants.CV_ECLUSTERDANGLINGINST,
1963 "the following instances have a non-existing primary-node:"
1964 " %s", utils.CommaJoin(no_node_instances))
1969 class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
1970 """Verifies the status of a node group.
1973 HPATH = "cluster-verify"
1974 HTYPE = constants.HTYPE_CLUSTER
1977 _HOOKS_INDENT_RE = re.compile("^", re.M)
1979 class NodeImage(object):
1980 """A class representing the logical and physical status of a node.
1983 @ivar name: the node name to which this object refers
1984 @ivar volumes: a structure as returned from
1985 L{ganeti.backend.GetVolumeList} (runtime)
1986 @ivar instances: a list of running instances (runtime)
1987 @ivar pinst: list of configured primary instances (config)
1988 @ivar sinst: list of configured secondary instances (config)
1989 @ivar sbp: dictionary of {primary-node: list of instances} for all
1990 instances for which this node is secondary (config)
1991 @ivar mfree: free memory, as reported by hypervisor (runtime)
1992 @ivar dfree: free disk, as reported by the node (runtime)
1993 @ivar offline: the offline status (config)
1994 @type rpc_fail: boolean
1995 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1996 not whether the individual keys were correct) (runtime)
1997 @type lvm_fail: boolean
1998 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1999 @type hyp_fail: boolean
2000 @ivar hyp_fail: whether the RPC call didn't return the instance list
2001 @type ghost: boolean
2002 @ivar ghost: whether this is a known node or not (config)
2003 @type os_fail: boolean
2004 @ivar os_fail: whether the RPC call didn't return valid OS data
2006 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
2007 @type vm_capable: boolean
2008 @ivar vm_capable: whether the node can host instances
2011 def __init__(self, offline=False, name=None, vm_capable=True):
2020 self.offline = offline
2021 self.vm_capable = vm_capable
2022 self.rpc_fail = False
2023 self.lvm_fail = False
2024 self.hyp_fail = False
2026 self.os_fail = False
2029 def ExpandNames(self):
2030 # This raises errors.OpPrereqError on its own:
2031 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
2033 # Get instances in node group; this is unsafe and needs verification later
2034 inst_names = self.cfg.GetNodeGroupInstances(self.group_uuid)
2036 self.needed_locks = {
2037 locking.LEVEL_INSTANCE: inst_names,
2038 locking.LEVEL_NODEGROUP: [self.group_uuid],
2039 locking.LEVEL_NODE: [],
2042 self.share_locks = _ShareAll()
2044 def DeclareLocks(self, level):
2045 if level == locking.LEVEL_NODE:
2046 # Get members of node group; this is unsafe and needs verification later
2047 nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members)
2049 all_inst_info = self.cfg.GetAllInstancesInfo()
2051 # In Exec(), we warn about mirrored instances that have primary and
2052 # secondary living in separate node groups. To fully verify that
2053 # volumes for these instances are healthy, we will need to do an
2054 # extra call to their secondaries. We ensure here those nodes will
2056 for inst in self.owned_locks(locking.LEVEL_INSTANCE):
2057 # Important: access only the instances whose lock is owned
2058 if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR:
2059 nodes.update(all_inst_info[inst].secondary_nodes)
2061 self.needed_locks[locking.LEVEL_NODE] = nodes
2063 def CheckPrereq(self):
2064 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
2065 self.group_info = self.cfg.GetNodeGroup(self.group_uuid)
2067 group_nodes = set(self.group_info.members)
2068 group_instances = self.cfg.GetNodeGroupInstances(self.group_uuid)
2071 group_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
2073 unlocked_instances = \
2074 group_instances.difference(self.owned_locks(locking.LEVEL_INSTANCE))
2077 raise errors.OpPrereqError("Missing lock for nodes: %s" %
2078 utils.CommaJoin(unlocked_nodes))
2080 if unlocked_instances:
2081 raise errors.OpPrereqError("Missing lock for instances: %s" %
2082 utils.CommaJoin(unlocked_instances))
2084 self.all_node_info = self.cfg.GetAllNodesInfo()
2085 self.all_inst_info = self.cfg.GetAllInstancesInfo()
2087 self.my_node_names = utils.NiceSort(group_nodes)
2088 self.my_inst_names = utils.NiceSort(group_instances)
2090 self.my_node_info = dict((name, self.all_node_info[name])
2091 for name in self.my_node_names)
2093 self.my_inst_info = dict((name, self.all_inst_info[name])
2094 for name in self.my_inst_names)
2096 # We detect here the nodes that will need the extra RPC calls for verifying
2097 # split LV volumes; they should be locked.
2098 extra_lv_nodes = set()
2100 for inst in self.my_inst_info.values():
2101 if inst.disk_template in constants.DTS_INT_MIRROR:
2102 group = self.my_node_info[inst.primary_node].group
2103 for nname in inst.secondary_nodes:
2104 if self.all_node_info[nname].group != group:
2105 extra_lv_nodes.add(nname)
2107 unlocked_lv_nodes = \
2108 extra_lv_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
2110 if unlocked_lv_nodes:
2111 raise errors.OpPrereqError("these nodes could be locked: %s" %
2112 utils.CommaJoin(unlocked_lv_nodes))
2113 self.extra_lv_nodes = list(extra_lv_nodes)
2115 def _VerifyNode(self, ninfo, nresult):
2116 """Perform some basic validation on data returned from a node.
2118 - check the result data structure is well formed and has all the
2120 - check ganeti version
2122 @type ninfo: L{objects.Node}
2123 @param ninfo: the node to check
2124 @param nresult: the results from the node
2126 @return: whether overall this call was successful (and we can expect
2127 reasonable values in the respose)
2131 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2133 # main result, nresult should be a non-empty dict
2134 test = not nresult or not isinstance(nresult, dict)
2135 _ErrorIf(test, constants.CV_ENODERPC, node,
2136 "unable to verify node: no data returned")
2140 # compares ganeti version
2141 local_version = constants.PROTOCOL_VERSION
2142 remote_version = nresult.get("version", None)
2143 test = not (remote_version and
2144 isinstance(remote_version, (list, tuple)) and
2145 len(remote_version) == 2)
2146 _ErrorIf(test, constants.CV_ENODERPC, node,
2147 "connection to node returned invalid data")
2151 test = local_version != remote_version[0]
2152 _ErrorIf(test, constants.CV_ENODEVERSION, node,
2153 "incompatible protocol versions: master %s,"
2154 " node %s", local_version, remote_version[0])
2158 # node seems compatible, we can actually try to look into its results
2160 # full package version
2161 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
2162 constants.CV_ENODEVERSION, node,
2163 "software version mismatch: master %s, node %s",
2164 constants.RELEASE_VERSION, remote_version[1],
2165 code=self.ETYPE_WARNING)
2167 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
2168 if ninfo.vm_capable and isinstance(hyp_result, dict):
2169 for hv_name, hv_result in hyp_result.iteritems():
2170 test = hv_result is not None
2171 _ErrorIf(test, constants.CV_ENODEHV, node,
2172 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
2174 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
2175 if ninfo.vm_capable and isinstance(hvp_result, list):
2176 for item, hv_name, hv_result in hvp_result:
2177 _ErrorIf(True, constants.CV_ENODEHV, node,
2178 "hypervisor %s parameter verify failure (source %s): %s",
2179 hv_name, item, hv_result)
2181 test = nresult.get(constants.NV_NODESETUP,
2182 ["Missing NODESETUP results"])
2183 _ErrorIf(test, constants.CV_ENODESETUP, node, "node setup error: %s",
2188 def _VerifyNodeTime(self, ninfo, nresult,
2189 nvinfo_starttime, nvinfo_endtime):
2190 """Check the node time.
2192 @type ninfo: L{objects.Node}
2193 @param ninfo: the node to check
2194 @param nresult: the remote results for the node
2195 @param nvinfo_starttime: the start time of the RPC call
2196 @param nvinfo_endtime: the end time of the RPC call
2200 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2202 ntime = nresult.get(constants.NV_TIME, None)
2204 ntime_merged = utils.MergeTime(ntime)
2205 except (ValueError, TypeError):
2206 _ErrorIf(True, constants.CV_ENODETIME, node, "Node returned invalid time")
2209 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
2210 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
2211 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
2212 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
2216 _ErrorIf(ntime_diff is not None, constants.CV_ENODETIME, node,
2217 "Node time diverges by at least %s from master node time",
2220 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
2221 """Check the node LVM results.
2223 @type ninfo: L{objects.Node}
2224 @param ninfo: the node to check
2225 @param nresult: the remote results for the node
2226 @param vg_name: the configured VG name
2233 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2235 # checks vg existence and size > 20G
2236 vglist = nresult.get(constants.NV_VGLIST, None)
2238 _ErrorIf(test, constants.CV_ENODELVM, node, "unable to check volume groups")
2240 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
2241 constants.MIN_VG_SIZE)
2242 _ErrorIf(vgstatus, constants.CV_ENODELVM, node, vgstatus)
2245 pvlist = nresult.get(constants.NV_PVLIST, None)
2246 test = pvlist is None
2247 _ErrorIf(test, constants.CV_ENODELVM, node, "Can't get PV list from node")
2249 # check that ':' is not present in PV names, since it's a
2250 # special character for lvcreate (denotes the range of PEs to
2252 for _, pvname, owner_vg in pvlist:
2253 test = ":" in pvname
2254 _ErrorIf(test, constants.CV_ENODELVM, node,
2255 "Invalid character ':' in PV '%s' of VG '%s'",
2258 def _VerifyNodeBridges(self, ninfo, nresult, bridges):
2259 """Check the node bridges.
2261 @type ninfo: L{objects.Node}
2262 @param ninfo: the node to check
2263 @param nresult: the remote results for the node
2264 @param bridges: the expected list of bridges
2271 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2273 missing = nresult.get(constants.NV_BRIDGES, None)
2274 test = not isinstance(missing, list)
2275 _ErrorIf(test, constants.CV_ENODENET, node,
2276 "did not return valid bridge information")
2278 _ErrorIf(bool(missing), constants.CV_ENODENET, node,
2279 "missing bridges: %s" % utils.CommaJoin(sorted(missing)))
2281 def _VerifyNodeUserScripts(self, ninfo, nresult):
2282 """Check the results of user scripts presence and executability on the node
2284 @type ninfo: L{objects.Node}
2285 @param ninfo: the node to check
2286 @param nresult: the remote results for the node
2291 test = not constants.NV_USERSCRIPTS in nresult
2292 self._ErrorIf(test, constants.CV_ENODEUSERSCRIPTS, node,
2293 "did not return user scripts information")
2295 broken_scripts = nresult.get(constants.NV_USERSCRIPTS, None)
2297 self._ErrorIf(broken_scripts, constants.CV_ENODEUSERSCRIPTS, node,
2298 "user scripts not present or not executable: %s" %
2299 utils.CommaJoin(sorted(broken_scripts)))
2301 def _VerifyNodeNetwork(self, ninfo, nresult):
2302 """Check the node network connectivity results.
2304 @type ninfo: L{objects.Node}
2305 @param ninfo: the node to check
2306 @param nresult: the remote results for the node
2310 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2312 test = constants.NV_NODELIST not in nresult
2313 _ErrorIf(test, constants.CV_ENODESSH, node,
2314 "node hasn't returned node ssh connectivity data")
2316 if nresult[constants.NV_NODELIST]:
2317 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
2318 _ErrorIf(True, constants.CV_ENODESSH, node,
2319 "ssh communication with node '%s': %s", a_node, a_msg)
2321 test = constants.NV_NODENETTEST not in nresult
2322 _ErrorIf(test, constants.CV_ENODENET, node,
2323 "node hasn't returned node tcp connectivity data")
2325 if nresult[constants.NV_NODENETTEST]:
2326 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
2328 _ErrorIf(True, constants.CV_ENODENET, node,
2329 "tcp communication with node '%s': %s",
2330 anode, nresult[constants.NV_NODENETTEST][anode])
2332 test = constants.NV_MASTERIP not in nresult
2333 _ErrorIf(test, constants.CV_ENODENET, node,
2334 "node hasn't returned node master IP reachability data")
2336 if not nresult[constants.NV_MASTERIP]:
2337 if node == self.master_node:
2338 msg = "the master node cannot reach the master IP (not configured?)"
2340 msg = "cannot reach the master IP"
2341 _ErrorIf(True, constants.CV_ENODENET, node, msg)
2343 def _VerifyInstance(self, instance, instanceconfig, node_image,
2345 """Verify an instance.
2347 This function checks to see if the required block devices are
2348 available on the instance's node.
2351 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2352 node_current = instanceconfig.primary_node
2354 node_vol_should = {}
2355 instanceconfig.MapLVsByNode(node_vol_should)
2357 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(), self.group_info)
2358 err = _ComputeIPolicyInstanceViolation(ipolicy, instanceconfig)
2359 _ErrorIf(err, constants.CV_EINSTANCEPOLICY, instance, err)
2361 for node in node_vol_should:
2362 n_img = node_image[node]
2363 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2364 # ignore missing volumes on offline or broken nodes
2366 for volume in node_vol_should[node]:
2367 test = volume not in n_img.volumes
2368 _ErrorIf(test, constants.CV_EINSTANCEMISSINGDISK, instance,
2369 "volume %s missing on node %s", volume, node)
2371 if instanceconfig.admin_state == constants.ADMINST_UP:
2372 pri_img = node_image[node_current]
2373 test = instance not in pri_img.instances and not pri_img.offline
2374 _ErrorIf(test, constants.CV_EINSTANCEDOWN, instance,
2375 "instance not running on its primary node %s",
2378 diskdata = [(nname, success, status, idx)
2379 for (nname, disks) in diskstatus.items()
2380 for idx, (success, status) in enumerate(disks)]
2382 for nname, success, bdev_status, idx in diskdata:
2383 # the 'ghost node' construction in Exec() ensures that we have a
2385 snode = node_image[nname]
2386 bad_snode = snode.ghost or snode.offline
2387 _ErrorIf(instanceconfig.admin_state == constants.ADMINST_UP and
2388 not success and not bad_snode,
2389 constants.CV_EINSTANCEFAULTYDISK, instance,
2390 "couldn't retrieve status for disk/%s on %s: %s",
2391 idx, nname, bdev_status)
2392 _ErrorIf((instanceconfig.admin_state == constants.ADMINST_UP and
2393 success and bdev_status.ldisk_status == constants.LDS_FAULTY),
2394 constants.CV_EINSTANCEFAULTYDISK, instance,
2395 "disk/%s on %s is faulty", idx, nname)
2397 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
2398 """Verify if there are any unknown volumes in the cluster.
2400 The .os, .swap and backup volumes are ignored. All other volumes are
2401 reported as unknown.
2403 @type reserved: L{ganeti.utils.FieldSet}
2404 @param reserved: a FieldSet of reserved volume names
2407 for node, n_img in node_image.items():
2408 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2409 # skip non-healthy nodes
2411 for volume in n_img.volumes:
2412 test = ((node not in node_vol_should or
2413 volume not in node_vol_should[node]) and
2414 not reserved.Matches(volume))
2415 self._ErrorIf(test, constants.CV_ENODEORPHANLV, node,
2416 "volume %s is unknown", volume)
2418 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
2419 """Verify N+1 Memory Resilience.
2421 Check that if one single node dies we can still start all the
2422 instances it was primary for.
2425 cluster_info = self.cfg.GetClusterInfo()
2426 for node, n_img in node_image.items():
2427 # This code checks that every node which is now listed as
2428 # secondary has enough memory to host all instances it is
2429 # supposed to should a single other node in the cluster fail.
2430 # FIXME: not ready for failover to an arbitrary node
2431 # FIXME: does not support file-backed instances
2432 # WARNING: we currently take into account down instances as well
2433 # as up ones, considering that even if they're down someone
2434 # might want to start them even in the event of a node failure.
2436 # we're skipping offline nodes from the N+1 warning, since
2437 # most likely we don't have good memory infromation from them;
2438 # we already list instances living on such nodes, and that's
2441 #TODO(dynmem): also consider ballooning out other instances
2442 for prinode, instances in n_img.sbp.items():
2444 for instance in instances:
2445 bep = cluster_info.FillBE(instance_cfg[instance])
2446 if bep[constants.BE_AUTO_BALANCE]:
2447 needed_mem += bep[constants.BE_MINMEM]
2448 test = n_img.mfree < needed_mem
2449 self._ErrorIf(test, constants.CV_ENODEN1, node,
2450 "not enough memory to accomodate instance failovers"
2451 " should node %s fail (%dMiB needed, %dMiB available)",
2452 prinode, needed_mem, n_img.mfree)
2455 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
2456 (files_all, files_opt, files_mc, files_vm)):
2457 """Verifies file checksums collected from all nodes.
2459 @param errorif: Callback for reporting errors
2460 @param nodeinfo: List of L{objects.Node} objects
2461 @param master_node: Name of master node
2462 @param all_nvinfo: RPC results
2465 # Define functions determining which nodes to consider for a file
2468 (files_mc, lambda node: (node.master_candidate or
2469 node.name == master_node)),
2470 (files_vm, lambda node: node.vm_capable),
2473 # Build mapping from filename to list of nodes which should have the file
2475 for (files, fn) in files2nodefn:
2477 filenodes = nodeinfo
2479 filenodes = filter(fn, nodeinfo)
2480 nodefiles.update((filename,
2481 frozenset(map(operator.attrgetter("name"), filenodes)))
2482 for filename in files)
2484 assert set(nodefiles) == (files_all | files_mc | files_vm)
2486 fileinfo = dict((filename, {}) for filename in nodefiles)
2487 ignore_nodes = set()
2489 for node in nodeinfo:
2491 ignore_nodes.add(node.name)
2494 nresult = all_nvinfo[node.name]
2496 if nresult.fail_msg or not nresult.payload:
2499 node_files = nresult.payload.get(constants.NV_FILELIST, None)
2501 test = not (node_files and isinstance(node_files, dict))
2502 errorif(test, constants.CV_ENODEFILECHECK, node.name,
2503 "Node did not return file checksum data")
2505 ignore_nodes.add(node.name)
2508 # Build per-checksum mapping from filename to nodes having it
2509 for (filename, checksum) in node_files.items():
2510 assert filename in nodefiles
2511 fileinfo[filename].setdefault(checksum, set()).add(node.name)
2513 for (filename, checksums) in fileinfo.items():
2514 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
2516 # Nodes having the file
2517 with_file = frozenset(node_name
2518 for nodes in fileinfo[filename].values()
2519 for node_name in nodes) - ignore_nodes
2521 expected_nodes = nodefiles[filename] - ignore_nodes
2523 # Nodes missing file
2524 missing_file = expected_nodes - with_file
2526 if filename in files_opt:
2528 errorif(missing_file and missing_file != expected_nodes,
2529 constants.CV_ECLUSTERFILECHECK, None,
2530 "File %s is optional, but it must exist on all or no"
2531 " nodes (not found on %s)",
2532 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
2534 errorif(missing_file, constants.CV_ECLUSTERFILECHECK, None,
2535 "File %s is missing from node(s) %s", filename,
2536 utils.CommaJoin(utils.NiceSort(missing_file)))
2538 # Warn if a node has a file it shouldn't
2539 unexpected = with_file - expected_nodes
2541 constants.CV_ECLUSTERFILECHECK, None,
2542 "File %s should not exist on node(s) %s",
2543 filename, utils.CommaJoin(utils.NiceSort(unexpected)))
2545 # See if there are multiple versions of the file
2546 test = len(checksums) > 1
2548 variants = ["variant %s on %s" %
2549 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
2550 for (idx, (checksum, nodes)) in
2551 enumerate(sorted(checksums.items()))]
2555 errorif(test, constants.CV_ECLUSTERFILECHECK, None,
2556 "File %s found with %s different checksums (%s)",
2557 filename, len(checksums), "; ".join(variants))
2559 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
2561 """Verifies and the node DRBD status.
2563 @type ninfo: L{objects.Node}
2564 @param ninfo: the node to check
2565 @param nresult: the remote results for the node
2566 @param instanceinfo: the dict of instances
2567 @param drbd_helper: the configured DRBD usermode helper
2568 @param drbd_map: the DRBD map as returned by
2569 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
2573 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2576 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
2577 test = (helper_result == None)
2578 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2579 "no drbd usermode helper returned")
2581 status, payload = helper_result
2583 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2584 "drbd usermode helper check unsuccessful: %s", payload)
2585 test = status and (payload != drbd_helper)
2586 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2587 "wrong drbd usermode helper: %s", payload)
2589 # compute the DRBD minors
2591 for minor, instance in drbd_map[node].items():
2592 test = instance not in instanceinfo
2593 _ErrorIf(test, constants.CV_ECLUSTERCFG, None,
2594 "ghost instance '%s' in temporary DRBD map", instance)
2595 # ghost instance should not be running, but otherwise we
2596 # don't give double warnings (both ghost instance and
2597 # unallocated minor in use)
2599 node_drbd[minor] = (instance, False)
2601 instance = instanceinfo[instance]
2602 node_drbd[minor] = (instance.name,
2603 instance.admin_state == constants.ADMINST_UP)
2605 # and now check them
2606 used_minors = nresult.get(constants.NV_DRBDLIST, [])
2607 test = not isinstance(used_minors, (tuple, list))
2608 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2609 "cannot parse drbd status file: %s", str(used_minors))
2611 # we cannot check drbd status
2614 for minor, (iname, must_exist) in node_drbd.items():
2615 test = minor not in used_minors and must_exist
2616 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2617 "drbd minor %d of instance %s is not active", minor, iname)
2618 for minor in used_minors:
2619 test = minor not in node_drbd
2620 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2621 "unallocated drbd minor %d is in use", minor)
2623 def _UpdateNodeOS(self, ninfo, nresult, nimg):
2624 """Builds the node OS structures.
2626 @type ninfo: L{objects.Node}
2627 @param ninfo: the node to check
2628 @param nresult: the remote results for the node
2629 @param nimg: the node image object
2633 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2635 remote_os = nresult.get(constants.NV_OSLIST, None)
2636 test = (not isinstance(remote_os, list) or
2637 not compat.all(isinstance(v, list) and len(v) == 7
2638 for v in remote_os))
2640 _ErrorIf(test, constants.CV_ENODEOS, node,
2641 "node hasn't returned valid OS data")
2650 for (name, os_path, status, diagnose,
2651 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
2653 if name not in os_dict:
2656 # parameters is a list of lists instead of list of tuples due to
2657 # JSON lacking a real tuple type, fix it:
2658 parameters = [tuple(v) for v in parameters]
2659 os_dict[name].append((os_path, status, diagnose,
2660 set(variants), set(parameters), set(api_ver)))
2662 nimg.oslist = os_dict
2664 def _VerifyNodeOS(self, ninfo, nimg, base):
2665 """Verifies the node OS list.
2667 @type ninfo: L{objects.Node}
2668 @param ninfo: the node to check
2669 @param nimg: the node image object
2670 @param base: the 'template' node we match against (e.g. from the master)
2674 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2676 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
2678 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
2679 for os_name, os_data in nimg.oslist.items():
2680 assert os_data, "Empty OS status for OS %s?!" % os_name
2681 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
2682 _ErrorIf(not f_status, constants.CV_ENODEOS, node,
2683 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
2684 _ErrorIf(len(os_data) > 1, constants.CV_ENODEOS, node,
2685 "OS '%s' has multiple entries (first one shadows the rest): %s",
2686 os_name, utils.CommaJoin([v[0] for v in os_data]))
2687 # comparisons with the 'base' image
2688 test = os_name not in base.oslist
2689 _ErrorIf(test, constants.CV_ENODEOS, node,
2690 "Extra OS %s not present on reference node (%s)",
2694 assert base.oslist[os_name], "Base node has empty OS status?"
2695 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
2697 # base OS is invalid, skipping
2699 for kind, a, b in [("API version", f_api, b_api),
2700 ("variants list", f_var, b_var),
2701 ("parameters", beautify_params(f_param),
2702 beautify_params(b_param))]:
2703 _ErrorIf(a != b, constants.CV_ENODEOS, node,
2704 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
2705 kind, os_name, base.name,
2706 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
2708 # check any missing OSes
2709 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
2710 _ErrorIf(missing, constants.CV_ENODEOS, node,
2711 "OSes present on reference node %s but missing on this node: %s",
2712 base.name, utils.CommaJoin(missing))
2714 def _VerifyOob(self, ninfo, nresult):
2715 """Verifies out of band functionality of a node.
2717 @type ninfo: L{objects.Node}
2718 @param ninfo: the node to check
2719 @param nresult: the remote results for the node
2723 # We just have to verify the paths on master and/or master candidates
2724 # as the oob helper is invoked on the master
2725 if ((ninfo.master_candidate or ninfo.master_capable) and
2726 constants.NV_OOB_PATHS in nresult):
2727 for path_result in nresult[constants.NV_OOB_PATHS]:
2728 self._ErrorIf(path_result, constants.CV_ENODEOOBPATH, node, path_result)
2730 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
2731 """Verifies and updates the node volume data.
2733 This function will update a L{NodeImage}'s internal structures
2734 with data from the remote call.
2736 @type ninfo: L{objects.Node}
2737 @param ninfo: the node to check
2738 @param nresult: the remote results for the node
2739 @param nimg: the node image object
2740 @param vg_name: the configured VG name
2744 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2746 nimg.lvm_fail = True
2747 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2750 elif isinstance(lvdata, basestring):
2751 _ErrorIf(True, constants.CV_ENODELVM, node, "LVM problem on node: %s",
2752 utils.SafeEncode(lvdata))
2753 elif not isinstance(lvdata, dict):
2754 _ErrorIf(True, constants.CV_ENODELVM, node,
2755 "rpc call to node failed (lvlist)")
2757 nimg.volumes = lvdata
2758 nimg.lvm_fail = False
2760 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2761 """Verifies and updates the node instance list.
2763 If the listing was successful, then updates this node's instance
2764 list. Otherwise, it marks the RPC call as failed for the instance
2767 @type ninfo: L{objects.Node}
2768 @param ninfo: the node to check
2769 @param nresult: the remote results for the node
2770 @param nimg: the node image object
2773 idata = nresult.get(constants.NV_INSTANCELIST, None)
2774 test = not isinstance(idata, list)
2775 self._ErrorIf(test, constants.CV_ENODEHV, ninfo.name,
2776 "rpc call to node failed (instancelist): %s",
2777 utils.SafeEncode(str(idata)))
2779 nimg.hyp_fail = True
2781 nimg.instances = idata
2783 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2784 """Verifies and computes a node information map
2786 @type ninfo: L{objects.Node}
2787 @param ninfo: the node to check
2788 @param nresult: the remote results for the node
2789 @param nimg: the node image object
2790 @param vg_name: the configured VG name
2794 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2796 # try to read free memory (from the hypervisor)
2797 hv_info = nresult.get(constants.NV_HVINFO, None)
2798 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2799 _ErrorIf(test, constants.CV_ENODEHV, node,
2800 "rpc call to node failed (hvinfo)")
2803 nimg.mfree = int(hv_info["memory_free"])
2804 except (ValueError, TypeError):
2805 _ErrorIf(True, constants.CV_ENODERPC, node,
2806 "node returned invalid nodeinfo, check hypervisor")
2808 # FIXME: devise a free space model for file based instances as well
2809 if vg_name is not None:
2810 test = (constants.NV_VGLIST not in nresult or
2811 vg_name not in nresult[constants.NV_VGLIST])
2812 _ErrorIf(test, constants.CV_ENODELVM, node,
2813 "node didn't return data for the volume group '%s'"
2814 " - it is either missing or broken", vg_name)
2817 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2818 except (ValueError, TypeError):
2819 _ErrorIf(True, constants.CV_ENODERPC, node,
2820 "node returned invalid LVM info, check LVM status")
2822 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2823 """Gets per-disk status information for all instances.
2825 @type nodelist: list of strings
2826 @param nodelist: Node names
2827 @type node_image: dict of (name, L{objects.Node})
2828 @param node_image: Node objects
2829 @type instanceinfo: dict of (name, L{objects.Instance})
2830 @param instanceinfo: Instance objects
2831 @rtype: {instance: {node: [(succes, payload)]}}
2832 @return: a dictionary of per-instance dictionaries with nodes as
2833 keys and disk information as values; the disk information is a
2834 list of tuples (success, payload)
2837 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2840 node_disks_devonly = {}
2841 diskless_instances = set()
2842 diskless = constants.DT_DISKLESS
2844 for nname in nodelist:
2845 node_instances = list(itertools.chain(node_image[nname].pinst,
2846 node_image[nname].sinst))
2847 diskless_instances.update(inst for inst in node_instances
2848 if instanceinfo[inst].disk_template == diskless)
2849 disks = [(inst, disk)
2850 for inst in node_instances
2851 for disk in instanceinfo[inst].disks]
2854 # No need to collect data
2857 node_disks[nname] = disks
2859 # Creating copies as SetDiskID below will modify the objects and that can
2860 # lead to incorrect data returned from nodes
2861 devonly = [dev.Copy() for (_, dev) in disks]
2864 self.cfg.SetDiskID(dev, nname)
2866 node_disks_devonly[nname] = devonly
2868 assert len(node_disks) == len(node_disks_devonly)
2870 # Collect data from all nodes with disks
2871 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2874 assert len(result) == len(node_disks)
2878 for (nname, nres) in result.items():
2879 disks = node_disks[nname]
2882 # No data from this node
2883 data = len(disks) * [(False, "node offline")]
2886 _ErrorIf(msg, constants.CV_ENODERPC, nname,
2887 "while getting disk information: %s", msg)
2889 # No data from this node
2890 data = len(disks) * [(False, msg)]
2893 for idx, i in enumerate(nres.payload):
2894 if isinstance(i, (tuple, list)) and len(i) == 2:
2897 logging.warning("Invalid result from node %s, entry %d: %s",
2899 data.append((False, "Invalid result from the remote node"))
2901 for ((inst, _), status) in zip(disks, data):
2902 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2904 # Add empty entries for diskless instances.
2905 for inst in diskless_instances:
2906 assert inst not in instdisk
2909 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2910 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2911 compat.all(isinstance(s, (tuple, list)) and
2912 len(s) == 2 for s in statuses)
2913 for inst, nnames in instdisk.items()
2914 for nname, statuses in nnames.items())
2915 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2920 def _SshNodeSelector(group_uuid, all_nodes):
2921 """Create endless iterators for all potential SSH check hosts.
2924 nodes = [node for node in all_nodes
2925 if (node.group != group_uuid and
2927 keyfunc = operator.attrgetter("group")
2929 return map(itertools.cycle,
2930 [sorted(map(operator.attrgetter("name"), names))
2931 for _, names in itertools.groupby(sorted(nodes, key=keyfunc),
2935 def _SelectSshCheckNodes(cls, group_nodes, group_uuid, all_nodes):
2936 """Choose which nodes should talk to which other nodes.
2938 We will make nodes contact all nodes in their group, and one node from
2941 @warning: This algorithm has a known issue if one node group is much
2942 smaller than others (e.g. just one node). In such a case all other
2943 nodes will talk to the single node.
2946 online_nodes = sorted(node.name for node in group_nodes if not node.offline)
2947 sel = cls._SshNodeSelector(group_uuid, all_nodes)
2949 return (online_nodes,
2950 dict((name, sorted([i.next() for i in sel]))
2951 for name in online_nodes))
2953 def BuildHooksEnv(self):
2956 Cluster-Verify hooks just ran in the post phase and their failure makes
2957 the output be logged in the verify output and the verification to fail.
2961 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2964 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
2965 for node in self.my_node_info.values())
2969 def BuildHooksNodes(self):
2970 """Build hooks nodes.
2973 return ([], self.my_node_names)
2975 def Exec(self, feedback_fn):
2976 """Verify integrity of the node group, performing various test on nodes.
2979 # This method has too many local variables. pylint: disable=R0914
2980 feedback_fn("* Verifying group '%s'" % self.group_info.name)
2982 if not self.my_node_names:
2984 feedback_fn("* Empty node group, skipping verification")
2988 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2989 verbose = self.op.verbose
2990 self._feedback_fn = feedback_fn
2992 vg_name = self.cfg.GetVGName()
2993 drbd_helper = self.cfg.GetDRBDHelper()
2994 cluster = self.cfg.GetClusterInfo()
2995 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2996 hypervisors = cluster.enabled_hypervisors
2997 node_data_list = [self.my_node_info[name] for name in self.my_node_names]
2999 i_non_redundant = [] # Non redundant instances
3000 i_non_a_balanced = [] # Non auto-balanced instances
3001 i_offline = 0 # Count of offline instances
3002 n_offline = 0 # Count of offline nodes
3003 n_drained = 0 # Count of nodes being drained
3004 node_vol_should = {}
3006 # FIXME: verify OS list
3009 filemap = _ComputeAncillaryFiles(cluster, False)
3011 # do local checksums
3012 master_node = self.master_node = self.cfg.GetMasterNode()
3013 master_ip = self.cfg.GetMasterIP()
3015 feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
3018 if self.cfg.GetUseExternalMipScript():
3019 user_scripts.append(constants.EXTERNAL_MASTER_SETUP_SCRIPT)
3021 node_verify_param = {
3022 constants.NV_FILELIST:
3023 utils.UniqueSequence(filename
3024 for files in filemap
3025 for filename in files),
3026 constants.NV_NODELIST:
3027 self._SelectSshCheckNodes(node_data_list, self.group_uuid,
3028 self.all_node_info.values()),
3029 constants.NV_HYPERVISOR: hypervisors,
3030 constants.NV_HVPARAMS:
3031 _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
3032 constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip)
3033 for node in node_data_list
3034 if not node.offline],
3035 constants.NV_INSTANCELIST: hypervisors,
3036 constants.NV_VERSION: None,
3037 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
3038 constants.NV_NODESETUP: None,
3039 constants.NV_TIME: None,
3040 constants.NV_MASTERIP: (master_node, master_ip),
3041 constants.NV_OSLIST: None,
3042 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
3043 constants.NV_USERSCRIPTS: user_scripts,
3046 if vg_name is not None:
3047 node_verify_param[constants.NV_VGLIST] = None
3048 node_verify_param[constants.NV_LVLIST] = vg_name
3049 node_verify_param[constants.NV_PVLIST] = [vg_name]
3050 node_verify_param[constants.NV_DRBDLIST] = None
3053 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
3056 # FIXME: this needs to be changed per node-group, not cluster-wide
3058 default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
3059 if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
3060 bridges.add(default_nicpp[constants.NIC_LINK])
3061 for instance in self.my_inst_info.values():
3062 for nic in instance.nics:
3063 full_nic = cluster.SimpleFillNIC(nic.nicparams)
3064 if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
3065 bridges.add(full_nic[constants.NIC_LINK])
3068 node_verify_param[constants.NV_BRIDGES] = list(bridges)
3070 # Build our expected cluster state
3071 node_image = dict((node.name, self.NodeImage(offline=node.offline,
3073 vm_capable=node.vm_capable))
3074 for node in node_data_list)
3078 for node in self.all_node_info.values():
3079 path = _SupportsOob(self.cfg, node)
3080 if path and path not in oob_paths:
3081 oob_paths.append(path)
3084 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
3086 for instance in self.my_inst_names:
3087 inst_config = self.my_inst_info[instance]
3089 for nname in inst_config.all_nodes:
3090 if nname not in node_image:
3091 gnode = self.NodeImage(name=nname)
3092 gnode.ghost = (nname not in self.all_node_info)
3093 node_image[nname] = gnode
3095 inst_config.MapLVsByNode(node_vol_should)
3097 pnode = inst_config.primary_node
3098 node_image[pnode].pinst.append(instance)
3100 for snode in inst_config.secondary_nodes:
3101 nimg = node_image[snode]
3102 nimg.sinst.append(instance)
3103 if pnode not in nimg.sbp:
3104 nimg.sbp[pnode] = []
3105 nimg.sbp[pnode].append(instance)
3107 # At this point, we have the in-memory data structures complete,
3108 # except for the runtime information, which we'll gather next
3110 # Due to the way our RPC system works, exact response times cannot be
3111 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
3112 # time before and after executing the request, we can at least have a time
3114 nvinfo_starttime = time.time()
3115 all_nvinfo = self.rpc.call_node_verify(self.my_node_names,
3117 self.cfg.GetClusterName())
3118 nvinfo_endtime = time.time()
3120 if self.extra_lv_nodes and vg_name is not None:
3122 self.rpc.call_node_verify(self.extra_lv_nodes,
3123 {constants.NV_LVLIST: vg_name},
3124 self.cfg.GetClusterName())
3126 extra_lv_nvinfo = {}
3128 all_drbd_map = self.cfg.ComputeDRBDMap()
3130 feedback_fn("* Gathering disk information (%s nodes)" %
3131 len(self.my_node_names))
3132 instdisk = self._CollectDiskInfo(self.my_node_names, node_image,
3135 feedback_fn("* Verifying configuration file consistency")
3137 # If not all nodes are being checked, we need to make sure the master node
3138 # and a non-checked vm_capable node are in the list.
3139 absent_nodes = set(self.all_node_info).difference(self.my_node_info)
3141 vf_nvinfo = all_nvinfo.copy()
3142 vf_node_info = list(self.my_node_info.values())
3143 additional_nodes = []
3144 if master_node not in self.my_node_info:
3145 additional_nodes.append(master_node)
3146 vf_node_info.append(self.all_node_info[master_node])
3147 # Add the first vm_capable node we find which is not included
3148 for node in absent_nodes:
3149 nodeinfo = self.all_node_info[node]
3150 if nodeinfo.vm_capable and not nodeinfo.offline:
3151 additional_nodes.append(node)
3152 vf_node_info.append(self.all_node_info[node])
3154 key = constants.NV_FILELIST
3155 vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes,
3156 {key: node_verify_param[key]},
3157 self.cfg.GetClusterName()))
3159 vf_nvinfo = all_nvinfo
3160 vf_node_info = self.my_node_info.values()
3162 self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap)
3164 feedback_fn("* Verifying node status")
3168 for node_i in node_data_list:
3170 nimg = node_image[node]
3174 feedback_fn("* Skipping offline node %s" % (node,))
3178 if node == master_node:
3180 elif node_i.master_candidate:
3181 ntype = "master candidate"
3182 elif node_i.drained:
3188 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
3190 msg = all_nvinfo[node].fail_msg
3191 _ErrorIf(msg, constants.CV_ENODERPC, node, "while contacting node: %s",
3194 nimg.rpc_fail = True
3197 nresult = all_nvinfo[node].payload
3199 nimg.call_ok = self._VerifyNode(node_i, nresult)
3200 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
3201 self._VerifyNodeNetwork(node_i, nresult)
3202 self._VerifyNodeUserScripts(node_i, nresult)
3203 self._VerifyOob(node_i, nresult)
3206 self._VerifyNodeLVM(node_i, nresult, vg_name)
3207 self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper,
3210 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
3211 self._UpdateNodeInstances(node_i, nresult, nimg)
3212 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
3213 self._UpdateNodeOS(node_i, nresult, nimg)
3215 if not nimg.os_fail:
3216 if refos_img is None:
3218 self._VerifyNodeOS(node_i, nimg, refos_img)
3219 self._VerifyNodeBridges(node_i, nresult, bridges)
3221 # Check whether all running instancies are primary for the node. (This
3222 # can no longer be done from _VerifyInstance below, since some of the
3223 # wrong instances could be from other node groups.)
3224 non_primary_inst = set(nimg.instances).difference(nimg.pinst)
3226 for inst in non_primary_inst:
3227 # FIXME: investigate best way to handle offline insts
3228 if inst.admin_state == constants.ADMINST_OFFLINE:
3230 feedback_fn("* Skipping offline instance %s" % inst.name)
3233 test = inst in self.all_inst_info
3234 _ErrorIf(test, constants.CV_EINSTANCEWRONGNODE, inst,
3235 "instance should not run on node %s", node_i.name)
3236 _ErrorIf(not test, constants.CV_ENODEORPHANINSTANCE, node_i.name,
3237 "node is running unknown instance %s", inst)
3239 for node, result in extra_lv_nvinfo.items():
3240 self._UpdateNodeVolumes(self.all_node_info[node], result.payload,
3241 node_image[node], vg_name)
3243 feedback_fn("* Verifying instance status")
3244 for instance in self.my_inst_names:
3246 feedback_fn("* Verifying instance %s" % instance)
3247 inst_config = self.my_inst_info[instance]
3248 self._VerifyInstance(instance, inst_config, node_image,
3250 inst_nodes_offline = []
3252 pnode = inst_config.primary_node
3253 pnode_img = node_image[pnode]
3254 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
3255 constants.CV_ENODERPC, pnode, "instance %s, connection to"
3256 " primary node failed", instance)
3258 _ErrorIf(inst_config.admin_state == constants.ADMINST_UP and
3260 constants.CV_EINSTANCEBADNODE, instance,
3261 "instance is marked as running and lives on offline node %s",
3262 inst_config.primary_node)
3264 # If the instance is non-redundant we cannot survive losing its primary
3265 # node, so we are not N+1 compliant. On the other hand we have no disk
3266 # templates with more than one secondary so that situation is not well
3268 # FIXME: does not support file-backed instances
3269 if not inst_config.secondary_nodes:
3270 i_non_redundant.append(instance)
3272 _ErrorIf(len(inst_config.secondary_nodes) > 1,
3273 constants.CV_EINSTANCELAYOUT,
3274 instance, "instance has multiple secondary nodes: %s",
3275 utils.CommaJoin(inst_config.secondary_nodes),
3276 code=self.ETYPE_WARNING)
3278 if inst_config.disk_template in constants.DTS_INT_MIRROR:
3279 pnode = inst_config.primary_node
3280 instance_nodes = utils.NiceSort(inst_config.all_nodes)
3281 instance_groups = {}
3283 for node in instance_nodes:
3284 instance_groups.setdefault(self.all_node_info[node].group,
3288 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
3289 # Sort so that we always list the primary node first.
3290 for group, nodes in sorted(instance_groups.items(),
3291 key=lambda (_, nodes): pnode in nodes,
3294 self._ErrorIf(len(instance_groups) > 1,
3295 constants.CV_EINSTANCESPLITGROUPS,
3296 instance, "instance has primary and secondary nodes in"
3297 " different groups: %s", utils.CommaJoin(pretty_list),
3298 code=self.ETYPE_WARNING)
3300 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
3301 i_non_a_balanced.append(instance)
3303 for snode in inst_config.secondary_nodes:
3304 s_img = node_image[snode]
3305 _ErrorIf(s_img.rpc_fail and not s_img.offline, constants.CV_ENODERPC,
3306 snode, "instance %s, connection to secondary node failed",
3310 inst_nodes_offline.append(snode)
3312 # warn that the instance lives on offline nodes
3313 _ErrorIf(inst_nodes_offline, constants.CV_EINSTANCEBADNODE, instance,
3314 "instance has offline secondary node(s) %s",
3315 utils.CommaJoin(inst_nodes_offline))
3316 # ... or ghost/non-vm_capable nodes
3317 for node in inst_config.all_nodes:
3318 _ErrorIf(node_image[node].ghost, constants.CV_EINSTANCEBADNODE,
3319 instance, "instance lives on ghost node %s", node)
3320 _ErrorIf(not node_image[node].vm_capable, constants.CV_EINSTANCEBADNODE,
3321 instance, "instance lives on non-vm_capable node %s", node)
3323 feedback_fn("* Verifying orphan volumes")
3324 reserved = utils.FieldSet(*cluster.reserved_lvs)
3326 # We will get spurious "unknown volume" warnings if any node of this group
3327 # is secondary for an instance whose primary is in another group. To avoid
3328 # them, we find these instances and add their volumes to node_vol_should.
3329 for inst in self.all_inst_info.values():
3330 for secondary in inst.secondary_nodes:
3331 if (secondary in self.my_node_info
3332 and inst.name not in self.my_inst_info):
3333 inst.MapLVsByNode(node_vol_should)
3336 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
3338 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
3339 feedback_fn("* Verifying N+1 Memory redundancy")
3340 self._VerifyNPlusOneMemory(node_image, self.my_inst_info)
3342 feedback_fn("* Other Notes")
3344 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
3345 % len(i_non_redundant))
3347 if i_non_a_balanced:
3348 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
3349 % len(i_non_a_balanced))
3352 feedback_fn(" - NOTICE: %d offline instance(s) found." % i_offline)
3355 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
3358 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
3362 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
3363 """Analyze the post-hooks' result
3365 This method analyses the hook result, handles it, and sends some
3366 nicely-formatted feedback back to the user.
3368 @param phase: one of L{constants.HOOKS_PHASE_POST} or
3369 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
3370 @param hooks_results: the results of the multi-node hooks rpc call
3371 @param feedback_fn: function used send feedback back to the caller
3372 @param lu_result: previous Exec result
3373 @return: the new Exec result, based on the previous result
3377 # We only really run POST phase hooks, only for non-empty groups,
3378 # and are only interested in their results
3379 if not self.my_node_names:
3382 elif phase == constants.HOOKS_PHASE_POST:
3383 # Used to change hooks' output to proper indentation
3384 feedback_fn("* Hooks Results")
3385 assert hooks_results, "invalid result from hooks"
3387 for node_name in hooks_results:
3388 res = hooks_results[node_name]
3390 test = msg and not res.offline
3391 self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,
3392 "Communication failure in hooks execution: %s", msg)
3393 if res.offline or msg:
3394 # No need to investigate payload if node is offline or gave
3397 for script, hkr, output in res.payload:
3398 test = hkr == constants.HKR_FAIL
3399 self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,
3400 "Script %s failed, output:", script)
3402 output = self._HOOKS_INDENT_RE.sub(" ", output)
3403 feedback_fn("%s" % output)
3409 class LUClusterVerifyDisks(NoHooksLU):
3410 """Verifies the cluster disks status.
3415 def ExpandNames(self):
3416 self.share_locks = _ShareAll()
3417 self.needed_locks = {
3418 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3421 def Exec(self, feedback_fn):
3422 group_names = self.owned_locks(locking.LEVEL_NODEGROUP)
3424 # Submit one instance of L{opcodes.OpGroupVerifyDisks} per node group
3425 return ResultWithJobs([[opcodes.OpGroupVerifyDisks(group_name=group)]
3426 for group in group_names])
3429 class LUGroupVerifyDisks(NoHooksLU):
3430 """Verifies the status of all disks in a node group.
3435 def ExpandNames(self):
3436 # Raises errors.OpPrereqError on its own if group can't be found
3437 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
3439 self.share_locks = _ShareAll()
3440 self.needed_locks = {
3441 locking.LEVEL_INSTANCE: [],
3442 locking.LEVEL_NODEGROUP: [],
3443 locking.LEVEL_NODE: [],
3446 def DeclareLocks(self, level):
3447 if level == locking.LEVEL_INSTANCE:
3448 assert not self.needed_locks[locking.LEVEL_INSTANCE]
3450 # Lock instances optimistically, needs verification once node and group
3451 # locks have been acquired
3452 self.needed_locks[locking.LEVEL_INSTANCE] = \
3453 self.cfg.GetNodeGroupInstances(self.group_uuid)
3455 elif level == locking.LEVEL_NODEGROUP:
3456 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
3458 self.needed_locks[locking.LEVEL_NODEGROUP] = \
3459 set([self.group_uuid] +
3460 # Lock all groups used by instances optimistically; this requires
3461 # going via the node before it's locked, requiring verification
3464 for instance_name in self.owned_locks(locking.LEVEL_INSTANCE)
3465 for group_uuid in self.cfg.GetInstanceNodeGroups(instance_name)])
3467 elif level == locking.LEVEL_NODE:
3468 # This will only lock the nodes in the group to be verified which contain
3470 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
3471 self._LockInstancesNodes()
3473 # Lock all nodes in group to be verified
3474 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
3475 member_nodes = self.cfg.GetNodeGroup(self.group_uuid).members
3476 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
3478 def CheckPrereq(self):
3479 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
3480 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
3481 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
3483 assert self.group_uuid in owned_groups
3485 # Check if locked instances are still correct
3486 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
3488 # Get instance information
3489 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
3491 # Check if node groups for locked instances are still correct
3492 for (instance_name, inst) in self.instances.items():
3493 assert owned_nodes.issuperset(inst.all_nodes), \
3494 "Instance %s's nodes changed while we kept the lock" % instance_name
3496 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
3499 assert self.group_uuid in inst_groups, \
3500 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
3502 def Exec(self, feedback_fn):
3503 """Verify integrity of cluster disks.
3505 @rtype: tuple of three items
3506 @return: a tuple of (dict of node-to-node_error, list of instances
3507 which need activate-disks, dict of instance: (node, volume) for
3512 res_instances = set()
3515 nv_dict = _MapInstanceDisksToNodes([inst
3516 for inst in self.instances.values()
3517 if inst.admin_state == constants.ADMINST_UP])
3520 nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &
3521 set(self.cfg.GetVmCapableNodeList()))
3523 node_lvs = self.rpc.call_lv_list(nodes, [])
3525 for (node, node_res) in node_lvs.items():
3526 if node_res.offline:
3529 msg = node_res.fail_msg
3531 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
3532 res_nodes[node] = msg
3535 for lv_name, (_, _, lv_online) in node_res.payload.items():
3536 inst = nv_dict.pop((node, lv_name), None)
3537 if not (lv_online or inst is None):
3538 res_instances.add(inst)
3540 # any leftover items in nv_dict are missing LVs, let's arrange the data
3542 for key, inst in nv_dict.iteritems():
3543 res_missing.setdefault(inst, []).append(list(key))
3545 return (res_nodes, list(res_instances), res_missing)
3548 class LUClusterRepairDiskSizes(NoHooksLU):
3549 """Verifies the cluster disks sizes.
3554 def ExpandNames(self):
3555 if self.op.instances:
3556 self.wanted_names = _GetWantedInstances(self, self.op.instances)
3557 self.needed_locks = {
3558 locking.LEVEL_NODE_RES: [],
3559 locking.LEVEL_INSTANCE: self.wanted_names,
3561 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
3563 self.wanted_names = None
3564 self.needed_locks = {
3565 locking.LEVEL_NODE_RES: locking.ALL_SET,
3566 locking.LEVEL_INSTANCE: locking.ALL_SET,
3568 self.share_locks = {
3569 locking.LEVEL_NODE_RES: 1,
3570 locking.LEVEL_INSTANCE: 0,
3573 def DeclareLocks(self, level):
3574 if level == locking.LEVEL_NODE_RES and self.wanted_names is not None:
3575 self._LockInstancesNodes(primary_only=True, level=level)
3577 def CheckPrereq(self):
3578 """Check prerequisites.
3580 This only checks the optional instance list against the existing names.
3583 if self.wanted_names is None:
3584 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
3586 self.wanted_instances = \
3587 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
3589 def _EnsureChildSizes(self, disk):
3590 """Ensure children of the disk have the needed disk size.
3592 This is valid mainly for DRBD8 and fixes an issue where the
3593 children have smaller disk size.
3595 @param disk: an L{ganeti.objects.Disk} object
3598 if disk.dev_type == constants.LD_DRBD8:
3599 assert disk.children, "Empty children for DRBD8?"
3600 fchild = disk.children[0]
3601 mismatch = fchild.size < disk.size
3603 self.LogInfo("Child disk has size %d, parent %d, fixing",
3604 fchild.size, disk.size)
3605 fchild.size = disk.size
3607 # and we recurse on this child only, not on the metadev
3608 return self._EnsureChildSizes(fchild) or mismatch
3612 def Exec(self, feedback_fn):
3613 """Verify the size of cluster disks.
3616 # TODO: check child disks too
3617 # TODO: check differences in size between primary/secondary nodes
3619 for instance in self.wanted_instances:
3620 pnode = instance.primary_node
3621 if pnode not in per_node_disks:
3622 per_node_disks[pnode] = []
3623 for idx, disk in enumerate(instance.disks):
3624 per_node_disks[pnode].append((instance, idx, disk))
3626 assert not (frozenset(per_node_disks.keys()) -
3627 self.owned_locks(locking.LEVEL_NODE_RES)), \
3628 "Not owning correct locks"
3629 assert not self.owned_locks(locking.LEVEL_NODE)
3632 for node, dskl in per_node_disks.items():
3633 newl = [v[2].Copy() for v in dskl]
3635 self.cfg.SetDiskID(dsk, node)
3636 result = self.rpc.call_blockdev_getsize(node, newl)
3638 self.LogWarning("Failure in blockdev_getsize call to node"
3639 " %s, ignoring", node)
3641 if len(result.payload) != len(dskl):
3642 logging.warning("Invalid result from node %s: len(dksl)=%d,"
3643 " result.payload=%s", node, len(dskl), result.payload)
3644 self.LogWarning("Invalid result from node %s, ignoring node results",
3647 for ((instance, idx, disk), size) in zip(dskl, result.payload):
3649 self.LogWarning("Disk %d of instance %s did not return size"
3650 " information, ignoring", idx, instance.name)
3652 if not isinstance(size, (int, long)):
3653 self.LogWarning("Disk %d of instance %s did not return valid"
3654 " size information, ignoring", idx, instance.name)
3657 if size != disk.size:
3658 self.LogInfo("Disk %d of instance %s has mismatched size,"
3659 " correcting: recorded %d, actual %d", idx,
3660 instance.name, disk.size, size)
3662 self.cfg.Update(instance, feedback_fn)
3663 changed.append((instance.name, idx, size))
3664 if self._EnsureChildSizes(disk):
3665 self.cfg.Update(instance, feedback_fn)
3666 changed.append((instance.name, idx, disk.size))
3670 class LUClusterRename(LogicalUnit):
3671 """Rename the cluster.
3674 HPATH = "cluster-rename"
3675 HTYPE = constants.HTYPE_CLUSTER
3677 def BuildHooksEnv(self):
3682 "OP_TARGET": self.cfg.GetClusterName(),
3683 "NEW_NAME": self.op.name,
3686 def BuildHooksNodes(self):
3687 """Build hooks nodes.
3690 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
3692 def CheckPrereq(self):
3693 """Verify that the passed name is a valid one.
3696 hostname = netutils.GetHostname(name=self.op.name,
3697 family=self.cfg.GetPrimaryIPFamily())
3699 new_name = hostname.name
3700 self.ip = new_ip = hostname.ip
3701 old_name = self.cfg.GetClusterName()
3702 old_ip = self.cfg.GetMasterIP()
3703 if new_name == old_name and new_ip == old_ip:
3704 raise errors.OpPrereqError("Neither the name nor the IP address of the"
3705 " cluster has changed",
3707 if new_ip != old_ip:
3708 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
3709 raise errors.OpPrereqError("The given cluster IP address (%s) is"
3710 " reachable on the network" %
3711 new_ip, errors.ECODE_NOTUNIQUE)
3713 self.op.name = new_name
3715 def Exec(self, feedback_fn):
3716 """Rename the cluster.
3719 clustername = self.op.name
3722 # shutdown the master IP
3723 master_params = self.cfg.GetMasterNetworkParameters()
3724 ems = self.cfg.GetUseExternalMipScript()
3725 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
3727 result.Raise("Could not disable the master role")
3730 cluster = self.cfg.GetClusterInfo()
3731 cluster.cluster_name = clustername
3732 cluster.master_ip = new_ip
3733 self.cfg.Update(cluster, feedback_fn)
3735 # update the known hosts file
3736 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
3737 node_list = self.cfg.GetOnlineNodeList()
3739 node_list.remove(master_params.name)
3742 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
3744 master_params.ip = new_ip
3745 result = self.rpc.call_node_activate_master_ip(master_params.name,
3747 msg = result.fail_msg
3749 self.LogWarning("Could not re-enable the master role on"
3750 " the master, please restart manually: %s", msg)
3755 def _ValidateNetmask(cfg, netmask):
3756 """Checks if a netmask is valid.
3758 @type cfg: L{config.ConfigWriter}
3759 @param cfg: The cluster configuration
3761 @param netmask: the netmask to be verified
3762 @raise errors.OpPrereqError: if the validation fails
3765 ip_family = cfg.GetPrimaryIPFamily()
3767 ipcls = netutils.IPAddress.GetClassFromIpFamily(ip_family)
3768 except errors.ProgrammerError:
3769 raise errors.OpPrereqError("Invalid primary ip family: %s." %
3771 if not ipcls.ValidateNetmask(netmask):
3772 raise errors.OpPrereqError("CIDR netmask (%s) not valid" %
3776 class LUClusterSetParams(LogicalUnit):
3777 """Change the parameters of the cluster.
3780 HPATH = "cluster-modify"
3781 HTYPE = constants.HTYPE_CLUSTER
3784 def CheckArguments(self):
3788 if self.op.uid_pool:
3789 uidpool.CheckUidPool(self.op.uid_pool)
3791 if self.op.add_uids:
3792 uidpool.CheckUidPool(self.op.add_uids)
3794 if self.op.remove_uids:
3795 uidpool.CheckUidPool(self.op.remove_uids)
3797 if self.op.master_netmask is not None:
3798 _ValidateNetmask(self.cfg, self.op.master_netmask)
3800 if self.op.diskparams:
3801 for dt_params in self.op.diskparams.values():
3802 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
3804 def ExpandNames(self):
3805 # FIXME: in the future maybe other cluster params won't require checking on
3806 # all nodes to be modified.
3807 self.needed_locks = {
3808 locking.LEVEL_NODE: locking.ALL_SET,
3809 locking.LEVEL_INSTANCE: locking.ALL_SET,
3810 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3812 self.share_locks = {
3813 locking.LEVEL_NODE: 1,
3814 locking.LEVEL_INSTANCE: 1,
3815 locking.LEVEL_NODEGROUP: 1,
3818 def BuildHooksEnv(self):
3823 "OP_TARGET": self.cfg.GetClusterName(),
3824 "NEW_VG_NAME": self.op.vg_name,
3827 def BuildHooksNodes(self):
3828 """Build hooks nodes.
3831 mn = self.cfg.GetMasterNode()
3834 def CheckPrereq(self):
3835 """Check prerequisites.
3837 This checks whether the given params don't conflict and
3838 if the given volume group is valid.
3841 if self.op.vg_name is not None and not self.op.vg_name:
3842 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
3843 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
3844 " instances exist", errors.ECODE_INVAL)
3846 if self.op.drbd_helper is not None and not self.op.drbd_helper:
3847 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
3848 raise errors.OpPrereqError("Cannot disable drbd helper while"
3849 " drbd-based instances exist",
3852 node_list = self.owned_locks(locking.LEVEL_NODE)
3854 # if vg_name not None, checks given volume group on all nodes
3856 vglist = self.rpc.call_vg_list(node_list)
3857 for node in node_list:
3858 msg = vglist[node].fail_msg
3860 # ignoring down node
3861 self.LogWarning("Error while gathering data on node %s"
3862 " (ignoring node): %s", node, msg)
3864 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
3866 constants.MIN_VG_SIZE)
3868 raise errors.OpPrereqError("Error on node '%s': %s" %
3869 (node, vgstatus), errors.ECODE_ENVIRON)
3871 if self.op.drbd_helper:
3872 # checks given drbd helper on all nodes
3873 helpers = self.rpc.call_drbd_helper(node_list)
3874 for (node, ninfo) in self.cfg.GetMultiNodeInfo(node_list):
3876 self.LogInfo("Not checking drbd helper on offline node %s", node)
3878 msg = helpers[node].fail_msg
3880 raise errors.OpPrereqError("Error checking drbd helper on node"
3881 " '%s': %s" % (node, msg),
3882 errors.ECODE_ENVIRON)
3883 node_helper = helpers[node].payload
3884 if node_helper != self.op.drbd_helper:
3885 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
3886 (node, node_helper), errors.ECODE_ENVIRON)
3888 self.cluster = cluster = self.cfg.GetClusterInfo()
3889 # validate params changes
3890 if self.op.beparams:
3891 objects.UpgradeBeParams(self.op.beparams)
3892 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
3893 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
3895 if self.op.ndparams:
3896 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
3897 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
3899 # TODO: we need a more general way to handle resetting
3900 # cluster-level parameters to default values
3901 if self.new_ndparams["oob_program"] == "":
3902 self.new_ndparams["oob_program"] = \
3903 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
3905 if self.op.hv_state:
3906 new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
3907 self.cluster.hv_state_static)
3908 self.new_hv_state = dict((hv, cluster.SimpleFillHvState(values))
3909 for hv, values in new_hv_state.items())
3911 if self.op.disk_state:
3912 new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state,
3913 self.cluster.disk_state_static)
3914 self.new_disk_state = \
3915 dict((storage, dict((name, cluster.SimpleFillDiskState(values))
3916 for name, values in svalues.items()))
3917 for storage, svalues in new_disk_state.items())
3920 self.new_ipolicy = _GetUpdatedIPolicy(cluster.ipolicy, self.op.ipolicy,
3923 all_instances = self.cfg.GetAllInstancesInfo().values()
3925 for group in self.cfg.GetAllNodeGroupsInfo().values():
3926 instances = frozenset([inst for inst in all_instances
3927 if compat.any(node in group.members
3928 for node in inst.all_nodes)])
3929 new_ipolicy = objects.FillIPolicy(self.new_ipolicy, group.ipolicy)
3930 new = _ComputeNewInstanceViolations(_CalculateGroupIPolicy(cluster,
3932 new_ipolicy, instances)
3934 violations.update(new)
3937 self.LogWarning("After the ipolicy change the following instances"
3938 " violate them: %s",
3939 utils.CommaJoin(violations))
3941 if self.op.nicparams:
3942 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
3943 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
3944 objects.NIC.CheckParameterSyntax(self.new_nicparams)
3947 # check all instances for consistency
3948 for instance in self.cfg.GetAllInstancesInfo().values():
3949 for nic_idx, nic in enumerate(instance.nics):
3950 params_copy = copy.deepcopy(nic.nicparams)
3951 params_filled = objects.FillDict(self.new_nicparams, params_copy)
3953 # check parameter syntax
3955 objects.NIC.CheckParameterSyntax(params_filled)
3956 except errors.ConfigurationError, err:
3957 nic_errors.append("Instance %s, nic/%d: %s" %
3958 (instance.name, nic_idx, err))
3960 # if we're moving instances to routed, check that they have an ip
3961 target_mode = params_filled[constants.NIC_MODE]
3962 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
3963 nic_errors.append("Instance %s, nic/%d: routed NIC with no ip"
3964 " address" % (instance.name, nic_idx))
3966 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
3967 "\n".join(nic_errors))
3969 # hypervisor list/parameters
3970 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
3971 if self.op.hvparams:
3972 for hv_name, hv_dict in self.op.hvparams.items():
3973 if hv_name not in self.new_hvparams:
3974 self.new_hvparams[hv_name] = hv_dict
3976 self.new_hvparams[hv_name].update(hv_dict)
3978 # disk template parameters
3979 self.new_diskparams = objects.FillDict(cluster.diskparams, {})
3980 if self.op.diskparams:
3981 for dt_name, dt_params in self.op.diskparams.items():
3982 if dt_name not in self.op.diskparams:
3983 self.new_diskparams[dt_name] = dt_params
3985 self.new_diskparams[dt_name].update(dt_params)
3987 # os hypervisor parameters
3988 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
3990 for os_name, hvs in self.op.os_hvp.items():
3991 if os_name not in self.new_os_hvp:
3992 self.new_os_hvp[os_name] = hvs
3994 for hv_name, hv_dict in hvs.items():
3995 if hv_name not in self.new_os_hvp[os_name]:
3996 self.new_os_hvp[os_name][hv_name] = hv_dict
3998 self.new_os_hvp[os_name][hv_name].update(hv_dict)
4001 self.new_osp = objects.FillDict(cluster.osparams, {})
4002 if self.op.osparams:
4003 for os_name, osp in self.op.osparams.items():
4004 if os_name not in self.new_osp:
4005 self.new_osp[os_name] = {}
4007 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
4010 if not self.new_osp[os_name]:
4011 # we removed all parameters
4012 del self.new_osp[os_name]
4014 # check the parameter validity (remote check)
4015 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
4016 os_name, self.new_osp[os_name])
4018 # changes to the hypervisor list
4019 if self.op.enabled_hypervisors is not None:
4020 self.hv_list = self.op.enabled_hypervisors
4021 for hv in self.hv_list:
4022 # if the hypervisor doesn't already exist in the cluster
4023 # hvparams, we initialize it to empty, and then (in both
4024 # cases) we make sure to fill the defaults, as we might not
4025 # have a complete defaults list if the hypervisor wasn't
4027 if hv not in new_hvp:
4029 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
4030 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
4032 self.hv_list = cluster.enabled_hypervisors
4034 if self.op.hvparams or self.op.enabled_hypervisors is not None:
4035 # either the enabled list has changed, or the parameters have, validate
4036 for hv_name, hv_params in self.new_hvparams.items():
4037 if ((self.op.hvparams and hv_name in self.op.hvparams) or
4038 (self.op.enabled_hypervisors and
4039 hv_name in self.op.enabled_hypervisors)):
4040 # either this is a new hypervisor, or its parameters have changed
4041 hv_class = hypervisor.GetHypervisor(hv_name)
4042 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
4043 hv_class.CheckParameterSyntax(hv_params)
4044 _CheckHVParams(self, node_list, hv_name, hv_params)
4047 # no need to check any newly-enabled hypervisors, since the
4048 # defaults have already been checked in the above code-block
4049 for os_name, os_hvp in self.new_os_hvp.items():
4050 for hv_name, hv_params in os_hvp.items():
4051 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
4052 # we need to fill in the new os_hvp on top of the actual hv_p
4053 cluster_defaults = self.new_hvparams.get(hv_name, {})
4054 new_osp = objects.FillDict(cluster_defaults, hv_params)
4055 hv_class = hypervisor.GetHypervisor(hv_name)
4056 hv_class.CheckParameterSyntax(new_osp)
4057 _CheckHVParams(self, node_list, hv_name, new_osp)
4059 if self.op.default_iallocator:
4060 alloc_script = utils.FindFile(self.op.default_iallocator,
4061 constants.IALLOCATOR_SEARCH_PATH,
4063 if alloc_script is None:
4064 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
4065 " specified" % self.op.default_iallocator,
4068 def Exec(self, feedback_fn):
4069 """Change the parameters of the cluster.
4072 if self.op.vg_name is not None:
4073 new_volume = self.op.vg_name
4076 if new_volume != self.cfg.GetVGName():
4077 self.cfg.SetVGName(new_volume)
4079 feedback_fn("Cluster LVM configuration already in desired"
4080 " state, not changing")
4081 if self.op.drbd_helper is not None:
4082 new_helper = self.op.drbd_helper
4085 if new_helper != self.cfg.GetDRBDHelper():
4086 self.cfg.SetDRBDHelper(new_helper)
4088 feedback_fn("Cluster DRBD helper already in desired state,"
4090 if self.op.hvparams:
4091 self.cluster.hvparams = self.new_hvparams
4093 self.cluster.os_hvp = self.new_os_hvp
4094 if self.op.enabled_hypervisors is not None:
4095 self.cluster.hvparams = self.new_hvparams
4096 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
4097 if self.op.beparams:
4098 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
4099 if self.op.nicparams:
4100 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
4102 self.cluster.ipolicy = self.new_ipolicy
4103 if self.op.osparams:
4104 self.cluster.osparams = self.new_osp
4105 if self.op.ndparams:
4106 self.cluster.ndparams = self.new_ndparams
4107 if self.op.diskparams:
4108 self.cluster.diskparams = self.new_diskparams
4109 if self.op.hv_state:
4110 self.cluster.hv_state_static = self.new_hv_state
4111 if self.op.disk_state:
4112 self.cluster.disk_state_static = self.new_disk_state
4114 if self.op.candidate_pool_size is not None:
4115 self.cluster.candidate_pool_size = self.op.candidate_pool_size
4116 # we need to update the pool size here, otherwise the save will fail
4117 _AdjustCandidatePool(self, [])
4119 if self.op.maintain_node_health is not None:
4120 if self.op.maintain_node_health and not constants.ENABLE_CONFD:
4121 feedback_fn("Note: CONFD was disabled at build time, node health"
4122 " maintenance is not useful (still enabling it)")
4123 self.cluster.maintain_node_health = self.op.maintain_node_health
4125 if self.op.prealloc_wipe_disks is not None:
4126 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
4128 if self.op.add_uids is not None:
4129 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
4131 if self.op.remove_uids is not None:
4132 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
4134 if self.op.uid_pool is not None:
4135 self.cluster.uid_pool = self.op.uid_pool
4137 if self.op.default_iallocator is not None:
4138 self.cluster.default_iallocator = self.op.default_iallocator
4140 if self.op.reserved_lvs is not None:
4141 self.cluster.reserved_lvs = self.op.reserved_lvs
4143 if self.op.use_external_mip_script is not None:
4144 self.cluster.use_external_mip_script = self.op.use_external_mip_script
4146 def helper_os(aname, mods, desc):
4148 lst = getattr(self.cluster, aname)
4149 for key, val in mods:
4150 if key == constants.DDM_ADD:
4152 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
4155 elif key == constants.DDM_REMOVE:
4159 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
4161 raise errors.ProgrammerError("Invalid modification '%s'" % key)
4163 if self.op.hidden_os:
4164 helper_os("hidden_os", self.op.hidden_os, "hidden")
4166 if self.op.blacklisted_os:
4167 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
4169 if self.op.master_netdev:
4170 master_params = self.cfg.GetMasterNetworkParameters()
4171 ems = self.cfg.GetUseExternalMipScript()
4172 feedback_fn("Shutting down master ip on the current netdev (%s)" %
4173 self.cluster.master_netdev)
4174 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
4176 result.Raise("Could not disable the master ip")
4177 feedback_fn("Changing master_netdev from %s to %s" %
4178 (master_params.netdev, self.op.master_netdev))
4179 self.cluster.master_netdev = self.op.master_netdev
4181 if self.op.master_netmask:
4182 master_params = self.cfg.GetMasterNetworkParameters()
4183 feedback_fn("Changing master IP netmask to %s" % self.op.master_netmask)
4184 result = self.rpc.call_node_change_master_netmask(master_params.name,
4185 master_params.netmask,
4186 self.op.master_netmask,
4188 master_params.netdev)
4190 msg = "Could not change the master IP netmask: %s" % result.fail_msg
4193 self.cluster.master_netmask = self.op.master_netmask
4195 self.cfg.Update(self.cluster, feedback_fn)
4197 if self.op.master_netdev:
4198 master_params = self.cfg.GetMasterNetworkParameters()
4199 feedback_fn("Starting the master ip on the new master netdev (%s)" %
4200 self.op.master_netdev)
4201 ems = self.cfg.GetUseExternalMipScript()
4202 result = self.rpc.call_node_activate_master_ip(master_params.name,
4205 self.LogWarning("Could not re-enable the master ip on"
4206 " the master, please restart manually: %s",
4210 def _UploadHelper(lu, nodes, fname):
4211 """Helper for uploading a file and showing warnings.
4214 if os.path.exists(fname):
4215 result = lu.rpc.call_upload_file(nodes, fname)
4216 for to_node, to_result in result.items():
4217 msg = to_result.fail_msg
4219 msg = ("Copy of file %s to node %s failed: %s" %
4220 (fname, to_node, msg))
4221 lu.proc.LogWarning(msg)
4224 def _ComputeAncillaryFiles(cluster, redist):
4225 """Compute files external to Ganeti which need to be consistent.
4227 @type redist: boolean
4228 @param redist: Whether to include files which need to be redistributed
4231 # Compute files for all nodes
4233 constants.SSH_KNOWN_HOSTS_FILE,
4234 constants.CONFD_HMAC_KEY,
4235 constants.CLUSTER_DOMAIN_SECRET_FILE,
4236 constants.SPICE_CERT_FILE,
4237 constants.SPICE_CACERT_FILE,
4238 constants.RAPI_USERS_FILE,
4242 files_all.update(constants.ALL_CERT_FILES)
4243 files_all.update(ssconf.SimpleStore().GetFileList())
4245 # we need to ship at least the RAPI certificate
4246 files_all.add(constants.RAPI_CERT_FILE)
4248 if cluster.modify_etc_hosts:
4249 files_all.add(constants.ETC_HOSTS)
4251 # Files which are optional, these must:
4252 # - be present in one other category as well
4253 # - either exist or not exist on all nodes of that category (mc, vm all)
4255 constants.RAPI_USERS_FILE,
4258 # Files which should only be on master candidates
4262 files_mc.add(constants.CLUSTER_CONF_FILE)
4264 # FIXME: this should also be replicated but Ganeti doesn't support files_mc
4266 files_mc.add(constants.DEFAULT_MASTER_SETUP_SCRIPT)
4268 # Files which should only be on VM-capable nodes
4269 files_vm = set(filename
4270 for hv_name in cluster.enabled_hypervisors
4271 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[0])
4273 files_opt |= set(filename
4274 for hv_name in cluster.enabled_hypervisors
4275 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[1])
4277 # Filenames in each category must be unique
4278 all_files_set = files_all | files_mc | files_vm
4279 assert (len(all_files_set) ==
4280 sum(map(len, [files_all, files_mc, files_vm]))), \
4281 "Found file listed in more than one file list"
4283 # Optional files must be present in one other category
4284 assert all_files_set.issuperset(files_opt), \
4285 "Optional file not in a different required list"
4287 return (files_all, files_opt, files_mc, files_vm)
4290 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
4291 """Distribute additional files which are part of the cluster configuration.
4293 ConfigWriter takes care of distributing the config and ssconf files, but
4294 there are more files which should be distributed to all nodes. This function
4295 makes sure those are copied.
4297 @param lu: calling logical unit
4298 @param additional_nodes: list of nodes not in the config to distribute to
4299 @type additional_vm: boolean
4300 @param additional_vm: whether the additional nodes are vm-capable or not
4303 # Gather target nodes
4304 cluster = lu.cfg.GetClusterInfo()
4305 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
4307 online_nodes = lu.cfg.GetOnlineNodeList()
4308 vm_nodes = lu.cfg.GetVmCapableNodeList()
4310 if additional_nodes is not None:
4311 online_nodes.extend(additional_nodes)
4313 vm_nodes.extend(additional_nodes)
4315 # Never distribute to master node
4316 for nodelist in [online_nodes, vm_nodes]:
4317 if master_info.name in nodelist:
4318 nodelist.remove(master_info.name)
4321 (files_all, _, files_mc, files_vm) = \
4322 _ComputeAncillaryFiles(cluster, True)
4324 # Never re-distribute configuration file from here
4325 assert not (constants.CLUSTER_CONF_FILE in files_all or
4326 constants.CLUSTER_CONF_FILE in files_vm)
4327 assert not files_mc, "Master candidates not handled in this function"
4330 (online_nodes, files_all),
4331 (vm_nodes, files_vm),
4335 for (node_list, files) in filemap:
4337 _UploadHelper(lu, node_list, fname)
4340 class LUClusterRedistConf(NoHooksLU):
4341 """Force the redistribution of cluster configuration.
4343 This is a very simple LU.
4348 def ExpandNames(self):
4349 self.needed_locks = {
4350 locking.LEVEL_NODE: locking.ALL_SET,
4352 self.share_locks[locking.LEVEL_NODE] = 1
4354 def Exec(self, feedback_fn):
4355 """Redistribute the configuration.
4358 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
4359 _RedistributeAncillaryFiles(self)
4362 class LUClusterActivateMasterIp(NoHooksLU):
4363 """Activate the master IP on the master node.
4366 def Exec(self, feedback_fn):
4367 """Activate the master IP.
4370 master_params = self.cfg.GetMasterNetworkParameters()
4371 ems = self.cfg.GetUseExternalMipScript()
4372 result = self.rpc.call_node_activate_master_ip(master_params.name,
4374 result.Raise("Could not activate the master IP")
4377 class LUClusterDeactivateMasterIp(NoHooksLU):
4378 """Deactivate the master IP on the master node.
4381 def Exec(self, feedback_fn):
4382 """Deactivate the master IP.
4385 master_params = self.cfg.GetMasterNetworkParameters()
4386 ems = self.cfg.GetUseExternalMipScript()
4387 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
4389 result.Raise("Could not deactivate the master IP")
4392 def _WaitForSync(lu, instance, disks=None, oneshot=False):
4393 """Sleep and poll for an instance's disk to sync.
4396 if not instance.disks or disks is not None and not disks:
4399 disks = _ExpandCheckDisks(instance, disks)
4402 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
4404 node = instance.primary_node
4407 lu.cfg.SetDiskID(dev, node)
4409 # TODO: Convert to utils.Retry
4412 degr_retries = 10 # in seconds, as we sleep 1 second each time
4416 cumul_degraded = False
4417 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
4418 msg = rstats.fail_msg
4420 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
4423 raise errors.RemoteError("Can't contact node %s for mirror data,"
4424 " aborting." % node)
4427 rstats = rstats.payload
4429 for i, mstat in enumerate(rstats):
4431 lu.LogWarning("Can't compute data for node %s/%s",
4432 node, disks[i].iv_name)
4435 cumul_degraded = (cumul_degraded or
4436 (mstat.is_degraded and mstat.sync_percent is None))
4437 if mstat.sync_percent is not None:
4439 if mstat.estimated_time is not None:
4440 rem_time = ("%s remaining (estimated)" %
4441 utils.FormatSeconds(mstat.estimated_time))
4442 max_time = mstat.estimated_time
4444 rem_time = "no time estimate"
4445 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
4446 (disks[i].iv_name, mstat.sync_percent, rem_time))
4448 # if we're done but degraded, let's do a few small retries, to
4449 # make sure we see a stable and not transient situation; therefore
4450 # we force restart of the loop
4451 if (done or oneshot) and cumul_degraded and degr_retries > 0:
4452 logging.info("Degraded disks found, %d retries left", degr_retries)
4460 time.sleep(min(60, max_time))
4463 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
4464 return not cumul_degraded
4467 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
4468 """Check that mirrors are not degraded.
4470 The ldisk parameter, if True, will change the test from the
4471 is_degraded attribute (which represents overall non-ok status for
4472 the device(s)) to the ldisk (representing the local storage status).
4475 lu.cfg.SetDiskID(dev, node)
4479 if on_primary or dev.AssembleOnSecondary():
4480 rstats = lu.rpc.call_blockdev_find(node, dev)
4481 msg = rstats.fail_msg
4483 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
4485 elif not rstats.payload:
4486 lu.LogWarning("Can't find disk on node %s", node)
4490 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
4492 result = result and not rstats.payload.is_degraded
4495 for child in dev.children:
4496 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
4501 class LUOobCommand(NoHooksLU):
4502 """Logical unit for OOB handling.
4506 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
4508 def ExpandNames(self):
4509 """Gather locks we need.
4512 if self.op.node_names:
4513 self.op.node_names = _GetWantedNodes(self, self.op.node_names)
4514 lock_names = self.op.node_names
4516 lock_names = locking.ALL_SET
4518 self.needed_locks = {
4519 locking.LEVEL_NODE: lock_names,
4522 def CheckPrereq(self):
4523 """Check prerequisites.
4526 - the node exists in the configuration
4529 Any errors are signaled by raising errors.OpPrereqError.
4533 self.master_node = self.cfg.GetMasterNode()
4535 assert self.op.power_delay >= 0.0
4537 if self.op.node_names:
4538 if (self.op.command in self._SKIP_MASTER and
4539 self.master_node in self.op.node_names):
4540 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
4541 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
4543 if master_oob_handler:
4544 additional_text = ("run '%s %s %s' if you want to operate on the"
4545 " master regardless") % (master_oob_handler,
4549 additional_text = "it does not support out-of-band operations"
4551 raise errors.OpPrereqError(("Operating on the master node %s is not"
4552 " allowed for %s; %s") %
4553 (self.master_node, self.op.command,
4554 additional_text), errors.ECODE_INVAL)
4556 self.op.node_names = self.cfg.GetNodeList()
4557 if self.op.command in self._SKIP_MASTER:
4558 self.op.node_names.remove(self.master_node)
4560 if self.op.command in self._SKIP_MASTER:
4561 assert self.master_node not in self.op.node_names
4563 for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names):
4565 raise errors.OpPrereqError("Node %s not found" % node_name,
4568 self.nodes.append(node)
4570 if (not self.op.ignore_status and
4571 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
4572 raise errors.OpPrereqError(("Cannot power off node %s because it is"
4573 " not marked offline") % node_name,
4576 def Exec(self, feedback_fn):
4577 """Execute OOB and return result if we expect any.
4580 master_node = self.master_node
4583 for idx, node in enumerate(utils.NiceSort(self.nodes,
4584 key=lambda node: node.name)):
4585 node_entry = [(constants.RS_NORMAL, node.name)]
4586 ret.append(node_entry)
4588 oob_program = _SupportsOob(self.cfg, node)
4591 node_entry.append((constants.RS_UNAVAIL, None))
4594 logging.info("Executing out-of-band command '%s' using '%s' on %s",
4595 self.op.command, oob_program, node.name)
4596 result = self.rpc.call_run_oob(master_node, oob_program,
4597 self.op.command, node.name,
4601 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
4602 node.name, result.fail_msg)
4603 node_entry.append((constants.RS_NODATA, None))
4606 self._CheckPayload(result)
4607 except errors.OpExecError, err:
4608 self.LogWarning("Payload returned by node '%s' is not valid: %s",
4610 node_entry.append((constants.RS_NODATA, None))
4612 if self.op.command == constants.OOB_HEALTH:
4613 # For health we should log important events
4614 for item, status in result.payload:
4615 if status in [constants.OOB_STATUS_WARNING,
4616 constants.OOB_STATUS_CRITICAL]:
4617 self.LogWarning("Item '%s' on node '%s' has status '%s'",
4618 item, node.name, status)
4620 if self.op.command == constants.OOB_POWER_ON:
4622 elif self.op.command == constants.OOB_POWER_OFF:
4623 node.powered = False
4624 elif self.op.command == constants.OOB_POWER_STATUS:
4625 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
4626 if powered != node.powered:
4627 logging.warning(("Recorded power state (%s) of node '%s' does not"
4628 " match actual power state (%s)"), node.powered,
4631 # For configuration changing commands we should update the node
4632 if self.op.command in (constants.OOB_POWER_ON,
4633 constants.OOB_POWER_OFF):
4634 self.cfg.Update(node, feedback_fn)
4636 node_entry.append((constants.RS_NORMAL, result.payload))
4638 if (self.op.command == constants.OOB_POWER_ON and
4639 idx < len(self.nodes) - 1):
4640 time.sleep(self.op.power_delay)
4644 def _CheckPayload(self, result):
4645 """Checks if the payload is valid.
4647 @param result: RPC result
4648 @raises errors.OpExecError: If payload is not valid
4652 if self.op.command == constants.OOB_HEALTH:
4653 if not isinstance(result.payload, list):
4654 errs.append("command 'health' is expected to return a list but got %s" %
4655 type(result.payload))
4657 for item, status in result.payload:
4658 if status not in constants.OOB_STATUSES:
4659 errs.append("health item '%s' has invalid status '%s'" %
4662 if self.op.command == constants.OOB_POWER_STATUS:
4663 if not isinstance(result.payload, dict):
4664 errs.append("power-status is expected to return a dict but got %s" %
4665 type(result.payload))
4667 if self.op.command in [
4668 constants.OOB_POWER_ON,
4669 constants.OOB_POWER_OFF,
4670 constants.OOB_POWER_CYCLE,
4672 if result.payload is not None:
4673 errs.append("%s is expected to not return payload but got '%s'" %
4674 (self.op.command, result.payload))
4677 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
4678 utils.CommaJoin(errs))
4681 class _OsQuery(_QueryBase):
4682 FIELDS = query.OS_FIELDS
4684 def ExpandNames(self, lu):
4685 # Lock all nodes in shared mode
4686 # Temporary removal of locks, should be reverted later
4687 # TODO: reintroduce locks when they are lighter-weight
4688 lu.needed_locks = {}
4689 #self.share_locks[locking.LEVEL_NODE] = 1
4690 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4692 # The following variables interact with _QueryBase._GetNames
4694 self.wanted = self.names
4696 self.wanted = locking.ALL_SET
4698 self.do_locking = self.use_locking
4700 def DeclareLocks(self, lu, level):
4704 def _DiagnoseByOS(rlist):
4705 """Remaps a per-node return list into an a per-os per-node dictionary
4707 @param rlist: a map with node names as keys and OS objects as values
4710 @return: a dictionary with osnames as keys and as value another
4711 map, with nodes as keys and tuples of (path, status, diagnose,
4712 variants, parameters, api_versions) as values, eg::
4714 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
4715 (/srv/..., False, "invalid api")],
4716 "node2": [(/srv/..., True, "", [], [])]}
4721 # we build here the list of nodes that didn't fail the RPC (at RPC
4722 # level), so that nodes with a non-responding node daemon don't
4723 # make all OSes invalid
4724 good_nodes = [node_name for node_name in rlist
4725 if not rlist[node_name].fail_msg]
4726 for node_name, nr in rlist.items():
4727 if nr.fail_msg or not nr.payload:
4729 for (name, path, status, diagnose, variants,
4730 params, api_versions) in nr.payload:
4731 if name not in all_os:
4732 # build a list of nodes for this os containing empty lists
4733 # for each node in node_list
4735 for nname in good_nodes:
4736 all_os[name][nname] = []
4737 # convert params from [name, help] to (name, help)
4738 params = [tuple(v) for v in params]
4739 all_os[name][node_name].append((path, status, diagnose,
4740 variants, params, api_versions))
4743 def _GetQueryData(self, lu):
4744 """Computes the list of nodes and their attributes.
4747 # Locking is not used
4748 assert not (compat.any(lu.glm.is_owned(level)
4749 for level in locking.LEVELS
4750 if level != locking.LEVEL_CLUSTER) or
4751 self.do_locking or self.use_locking)
4753 valid_nodes = [node.name
4754 for node in lu.cfg.GetAllNodesInfo().values()
4755 if not node.offline and node.vm_capable]
4756 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
4757 cluster = lu.cfg.GetClusterInfo()
4761 for (os_name, os_data) in pol.items():
4762 info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
4763 hidden=(os_name in cluster.hidden_os),
4764 blacklisted=(os_name in cluster.blacklisted_os))
4768 api_versions = set()
4770 for idx, osl in enumerate(os_data.values()):
4771 info.valid = bool(info.valid and osl and osl[0][1])
4775 (node_variants, node_params, node_api) = osl[0][3:6]
4778 variants.update(node_variants)
4779 parameters.update(node_params)
4780 api_versions.update(node_api)
4782 # Filter out inconsistent values
4783 variants.intersection_update(node_variants)
4784 parameters.intersection_update(node_params)
4785 api_versions.intersection_update(node_api)
4787 info.variants = list(variants)
4788 info.parameters = list(parameters)
4789 info.api_versions = list(api_versions)
4791 data[os_name] = info
4793 # Prepare data in requested order
4794 return [data[name] for name in self._GetNames(lu, pol.keys(), None)
4798 class LUOsDiagnose(NoHooksLU):
4799 """Logical unit for OS diagnose/query.
4805 def _BuildFilter(fields, names):
4806 """Builds a filter for querying OSes.
4809 name_filter = qlang.MakeSimpleFilter("name", names)
4811 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
4812 # respective field is not requested
4813 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]]
4814 for fname in ["hidden", "blacklisted"]
4815 if fname not in fields]
4816 if "valid" not in fields:
4817 status_filter.append([qlang.OP_TRUE, "valid"])
4820 status_filter.insert(0, qlang.OP_AND)
4822 status_filter = None
4824 if name_filter and status_filter:
4825 return [qlang.OP_AND, name_filter, status_filter]
4829 return status_filter
4831 def CheckArguments(self):
4832 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names),
4833 self.op.output_fields, False)
4835 def ExpandNames(self):
4836 self.oq.ExpandNames(self)
4838 def Exec(self, feedback_fn):
4839 return self.oq.OldStyleQuery(self)
4842 class LUNodeRemove(LogicalUnit):
4843 """Logical unit for removing a node.
4846 HPATH = "node-remove"
4847 HTYPE = constants.HTYPE_NODE
4849 def BuildHooksEnv(self):
4852 This doesn't run on the target node in the pre phase as a failed
4853 node would then be impossible to remove.
4857 "OP_TARGET": self.op.node_name,
4858 "NODE_NAME": self.op.node_name,
4861 def BuildHooksNodes(self):
4862 """Build hooks nodes.
4865 all_nodes = self.cfg.GetNodeList()
4867 all_nodes.remove(self.op.node_name)
4869 logging.warning("Node '%s', which is about to be removed, was not found"
4870 " in the list of all nodes", self.op.node_name)
4871 return (all_nodes, all_nodes)
4873 def CheckPrereq(self):
4874 """Check prerequisites.
4877 - the node exists in the configuration
4878 - it does not have primary or secondary instances
4879 - it's not the master
4881 Any errors are signaled by raising errors.OpPrereqError.
4884 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4885 node = self.cfg.GetNodeInfo(self.op.node_name)
4886 assert node is not None
4888 masternode = self.cfg.GetMasterNode()
4889 if node.name == masternode:
4890 raise errors.OpPrereqError("Node is the master node, failover to another"
4891 " node is required", errors.ECODE_INVAL)
4893 for instance_name, instance in self.cfg.GetAllInstancesInfo().items():
4894 if node.name in instance.all_nodes:
4895 raise errors.OpPrereqError("Instance %s is still running on the node,"
4896 " please remove first" % instance_name,
4898 self.op.node_name = node.name
4901 def Exec(self, feedback_fn):
4902 """Removes the node from the cluster.
4906 logging.info("Stopping the node daemon and removing configs from node %s",
4909 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
4911 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
4914 # Promote nodes to master candidate as needed
4915 _AdjustCandidatePool(self, exceptions=[node.name])
4916 self.context.RemoveNode(node.name)
4918 # Run post hooks on the node before it's removed
4919 _RunPostHook(self, node.name)
4921 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
4922 msg = result.fail_msg
4924 self.LogWarning("Errors encountered on the remote node while leaving"
4925 " the cluster: %s", msg)
4927 # Remove node from our /etc/hosts
4928 if self.cfg.GetClusterInfo().modify_etc_hosts:
4929 master_node = self.cfg.GetMasterNode()
4930 result = self.rpc.call_etc_hosts_modify(master_node,
4931 constants.ETC_HOSTS_REMOVE,
4933 result.Raise("Can't update hosts file with new host data")
4934 _RedistributeAncillaryFiles(self)
4937 class _NodeQuery(_QueryBase):
4938 FIELDS = query.NODE_FIELDS
4940 def ExpandNames(self, lu):
4941 lu.needed_locks = {}
4942 lu.share_locks = _ShareAll()
4945 self.wanted = _GetWantedNodes(lu, self.names)
4947 self.wanted = locking.ALL_SET
4949 self.do_locking = (self.use_locking and
4950 query.NQ_LIVE in self.requested_data)
4953 # If any non-static field is requested we need to lock the nodes
4954 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
4956 def DeclareLocks(self, lu, level):
4959 def _GetQueryData(self, lu):
4960 """Computes the list of nodes and their attributes.
4963 all_info = lu.cfg.GetAllNodesInfo()
4965 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
4967 # Gather data as requested
4968 if query.NQ_LIVE in self.requested_data:
4969 # filter out non-vm_capable nodes
4970 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
4972 node_data = lu.rpc.call_node_info(toquery_nodes, [lu.cfg.GetVGName()],
4973 [lu.cfg.GetHypervisorType()])
4974 live_data = dict((name, _MakeLegacyNodeInfo(nresult.payload))
4975 for (name, nresult) in node_data.items()
4976 if not nresult.fail_msg and nresult.payload)
4980 if query.NQ_INST in self.requested_data:
4981 node_to_primary = dict([(name, set()) for name in nodenames])
4982 node_to_secondary = dict([(name, set()) for name in nodenames])
4984 inst_data = lu.cfg.GetAllInstancesInfo()
4986 for inst in inst_data.values():
4987 if inst.primary_node in node_to_primary:
4988 node_to_primary[inst.primary_node].add(inst.name)
4989 for secnode in inst.secondary_nodes:
4990 if secnode in node_to_secondary:
4991 node_to_secondary[secnode].add(inst.name)
4993 node_to_primary = None
4994 node_to_secondary = None
4996 if query.NQ_OOB in self.requested_data:
4997 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
4998 for name, node in all_info.iteritems())
5002 if query.NQ_GROUP in self.requested_data:
5003 groups = lu.cfg.GetAllNodeGroupsInfo()
5007 return query.NodeQueryData([all_info[name] for name in nodenames],
5008 live_data, lu.cfg.GetMasterNode(),
5009 node_to_primary, node_to_secondary, groups,
5010 oob_support, lu.cfg.GetClusterInfo())
5013 class LUNodeQuery(NoHooksLU):
5014 """Logical unit for querying nodes.
5017 # pylint: disable=W0142
5020 def CheckArguments(self):
5021 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
5022 self.op.output_fields, self.op.use_locking)
5024 def ExpandNames(self):
5025 self.nq.ExpandNames(self)
5027 def DeclareLocks(self, level):
5028 self.nq.DeclareLocks(self, level)
5030 def Exec(self, feedback_fn):
5031 return self.nq.OldStyleQuery(self)
5034 class LUNodeQueryvols(NoHooksLU):
5035 """Logical unit for getting volumes on node(s).
5039 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
5040 _FIELDS_STATIC = utils.FieldSet("node")
5042 def CheckArguments(self):
5043 _CheckOutputFields(static=self._FIELDS_STATIC,
5044 dynamic=self._FIELDS_DYNAMIC,
5045 selected=self.op.output_fields)
5047 def ExpandNames(self):
5048 self.share_locks = _ShareAll()
5049 self.needed_locks = {}
5051 if not self.op.nodes:
5052 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5054 self.needed_locks[locking.LEVEL_NODE] = \
5055 _GetWantedNodes(self, self.op.nodes)
5057 def Exec(self, feedback_fn):
5058 """Computes the list of nodes and their attributes.
5061 nodenames = self.owned_locks(locking.LEVEL_NODE)
5062 volumes = self.rpc.call_node_volumes(nodenames)
5064 ilist = self.cfg.GetAllInstancesInfo()
5065 vol2inst = _MapInstanceDisksToNodes(ilist.values())
5068 for node in nodenames:
5069 nresult = volumes[node]
5072 msg = nresult.fail_msg
5074 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
5077 node_vols = sorted(nresult.payload,
5078 key=operator.itemgetter("dev"))
5080 for vol in node_vols:
5082 for field in self.op.output_fields:
5085 elif field == "phys":
5089 elif field == "name":
5091 elif field == "size":
5092 val = int(float(vol["size"]))
5093 elif field == "instance":
5094 val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-")
5096 raise errors.ParameterError(field)
5097 node_output.append(str(val))
5099 output.append(node_output)
5104 class LUNodeQueryStorage(NoHooksLU):
5105 """Logical unit for getting information on storage units on node(s).
5108 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
5111 def CheckArguments(self):
5112 _CheckOutputFields(static=self._FIELDS_STATIC,
5113 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
5114 selected=self.op.output_fields)
5116 def ExpandNames(self):
5117 self.share_locks = _ShareAll()
5118 self.needed_locks = {}
5121 self.needed_locks[locking.LEVEL_NODE] = \
5122 _GetWantedNodes(self, self.op.nodes)
5124 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5126 def Exec(self, feedback_fn):
5127 """Computes the list of nodes and their attributes.
5130 self.nodes = self.owned_locks(locking.LEVEL_NODE)
5132 # Always get name to sort by
5133 if constants.SF_NAME in self.op.output_fields:
5134 fields = self.op.output_fields[:]
5136 fields = [constants.SF_NAME] + self.op.output_fields
5138 # Never ask for node or type as it's only known to the LU
5139 for extra in [constants.SF_NODE, constants.SF_TYPE]:
5140 while extra in fields:
5141 fields.remove(extra)
5143 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
5144 name_idx = field_idx[constants.SF_NAME]
5146 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
5147 data = self.rpc.call_storage_list(self.nodes,
5148 self.op.storage_type, st_args,
5149 self.op.name, fields)
5153 for node in utils.NiceSort(self.nodes):
5154 nresult = data[node]
5158 msg = nresult.fail_msg
5160 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
5163 rows = dict([(row[name_idx], row) for row in nresult.payload])
5165 for name in utils.NiceSort(rows.keys()):
5170 for field in self.op.output_fields:
5171 if field == constants.SF_NODE:
5173 elif field == constants.SF_TYPE:
5174 val = self.op.storage_type
5175 elif field in field_idx:
5176 val = row[field_idx[field]]
5178 raise errors.ParameterError(field)
5187 class _InstanceQuery(_QueryBase):
5188 FIELDS = query.INSTANCE_FIELDS
5190 def ExpandNames(self, lu):
5191 lu.needed_locks = {}
5192 lu.share_locks = _ShareAll()
5195 self.wanted = _GetWantedInstances(lu, self.names)
5197 self.wanted = locking.ALL_SET
5199 self.do_locking = (self.use_locking and
5200 query.IQ_LIVE in self.requested_data)
5202 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5203 lu.needed_locks[locking.LEVEL_NODEGROUP] = []
5204 lu.needed_locks[locking.LEVEL_NODE] = []
5205 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5207 self.do_grouplocks = (self.do_locking and
5208 query.IQ_NODES in self.requested_data)
5210 def DeclareLocks(self, lu, level):
5212 if level == locking.LEVEL_NODEGROUP and self.do_grouplocks:
5213 assert not lu.needed_locks[locking.LEVEL_NODEGROUP]
5215 # Lock all groups used by instances optimistically; this requires going
5216 # via the node before it's locked, requiring verification later on
5217 lu.needed_locks[locking.LEVEL_NODEGROUP] = \
5219 for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
5220 for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
5221 elif level == locking.LEVEL_NODE:
5222 lu._LockInstancesNodes() # pylint: disable=W0212
5225 def _CheckGroupLocks(lu):
5226 owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE))
5227 owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP))
5229 # Check if node groups for locked instances are still correct
5230 for instance_name in owned_instances:
5231 _CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups)
5233 def _GetQueryData(self, lu):
5234 """Computes the list of instances and their attributes.
5237 if self.do_grouplocks:
5238 self._CheckGroupLocks(lu)
5240 cluster = lu.cfg.GetClusterInfo()
5241 all_info = lu.cfg.GetAllInstancesInfo()
5243 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
5245 instance_list = [all_info[name] for name in instance_names]
5246 nodes = frozenset(itertools.chain(*(inst.all_nodes
5247 for inst in instance_list)))
5248 hv_list = list(set([inst.hypervisor for inst in instance_list]))
5251 wrongnode_inst = set()
5253 # Gather data as requested
5254 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
5256 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
5258 result = node_data[name]
5260 # offline nodes will be in both lists
5261 assert result.fail_msg
5262 offline_nodes.append(name)
5264 bad_nodes.append(name)
5265 elif result.payload:
5266 for inst in result.payload:
5267 if inst in all_info:
5268 if all_info[inst].primary_node == name:
5269 live_data.update(result.payload)
5271 wrongnode_inst.add(inst)
5273 # orphan instance; we don't list it here as we don't
5274 # handle this case yet in the output of instance listing
5275 logging.warning("Orphan instance '%s' found on node %s",
5277 # else no instance is alive
5281 if query.IQ_DISKUSAGE in self.requested_data:
5282 disk_usage = dict((inst.name,
5283 _ComputeDiskSize(inst.disk_template,
5284 [{constants.IDISK_SIZE: disk.size}
5285 for disk in inst.disks]))
5286 for inst in instance_list)
5290 if query.IQ_CONSOLE in self.requested_data:
5292 for inst in instance_list:
5293 if inst.name in live_data:
5294 # Instance is running
5295 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
5297 consinfo[inst.name] = None
5298 assert set(consinfo.keys()) == set(instance_names)
5302 if query.IQ_NODES in self.requested_data:
5303 node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"),
5305 nodes = dict(lu.cfg.GetMultiNodeInfo(node_names))
5306 groups = dict((uuid, lu.cfg.GetNodeGroup(uuid))
5307 for uuid in set(map(operator.attrgetter("group"),
5313 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
5314 disk_usage, offline_nodes, bad_nodes,
5315 live_data, wrongnode_inst, consinfo,
5319 class LUQuery(NoHooksLU):
5320 """Query for resources/items of a certain kind.
5323 # pylint: disable=W0142
5326 def CheckArguments(self):
5327 qcls = _GetQueryImplementation(self.op.what)
5329 self.impl = qcls(self.op.qfilter, self.op.fields, self.op.use_locking)
5331 def ExpandNames(self):
5332 self.impl.ExpandNames(self)
5334 def DeclareLocks(self, level):
5335 self.impl.DeclareLocks(self, level)
5337 def Exec(self, feedback_fn):
5338 return self.impl.NewStyleQuery(self)
5341 class LUQueryFields(NoHooksLU):
5342 """Query for resources/items of a certain kind.
5345 # pylint: disable=W0142
5348 def CheckArguments(self):
5349 self.qcls = _GetQueryImplementation(self.op.what)
5351 def ExpandNames(self):
5352 self.needed_locks = {}
5354 def Exec(self, feedback_fn):
5355 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
5358 class LUNodeModifyStorage(NoHooksLU):
5359 """Logical unit for modifying a storage volume on a node.
5364 def CheckArguments(self):
5365 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5367 storage_type = self.op.storage_type
5370 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
5372 raise errors.OpPrereqError("Storage units of type '%s' can not be"
5373 " modified" % storage_type,
5376 diff = set(self.op.changes.keys()) - modifiable
5378 raise errors.OpPrereqError("The following fields can not be modified for"
5379 " storage units of type '%s': %r" %
5380 (storage_type, list(diff)),
5383 def ExpandNames(self):
5384 self.needed_locks = {
5385 locking.LEVEL_NODE: self.op.node_name,
5388 def Exec(self, feedback_fn):
5389 """Computes the list of nodes and their attributes.
5392 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
5393 result = self.rpc.call_storage_modify(self.op.node_name,
5394 self.op.storage_type, st_args,
5395 self.op.name, self.op.changes)
5396 result.Raise("Failed to modify storage unit '%s' on %s" %
5397 (self.op.name, self.op.node_name))
5400 class LUNodeAdd(LogicalUnit):
5401 """Logical unit for adding node to the cluster.
5405 HTYPE = constants.HTYPE_NODE
5406 _NFLAGS = ["master_capable", "vm_capable"]
5408 def CheckArguments(self):
5409 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
5410 # validate/normalize the node name
5411 self.hostname = netutils.GetHostname(name=self.op.node_name,
5412 family=self.primary_ip_family)
5413 self.op.node_name = self.hostname.name
5415 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
5416 raise errors.OpPrereqError("Cannot readd the master node",
5419 if self.op.readd and self.op.group:
5420 raise errors.OpPrereqError("Cannot pass a node group when a node is"
5421 " being readded", errors.ECODE_INVAL)
5423 def BuildHooksEnv(self):
5426 This will run on all nodes before, and on all nodes + the new node after.
5430 "OP_TARGET": self.op.node_name,
5431 "NODE_NAME": self.op.node_name,
5432 "NODE_PIP": self.op.primary_ip,
5433 "NODE_SIP": self.op.secondary_ip,
5434 "MASTER_CAPABLE": str(self.op.master_capable),
5435 "VM_CAPABLE": str(self.op.vm_capable),
5438 def BuildHooksNodes(self):
5439 """Build hooks nodes.
5442 # Exclude added node
5443 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name]))
5444 post_nodes = pre_nodes + [self.op.node_name, ]
5446 return (pre_nodes, post_nodes)
5448 def CheckPrereq(self):
5449 """Check prerequisites.
5452 - the new node is not already in the config
5454 - its parameters (single/dual homed) matches the cluster
5456 Any errors are signaled by raising errors.OpPrereqError.
5460 hostname = self.hostname
5461 node = hostname.name
5462 primary_ip = self.op.primary_ip = hostname.ip
5463 if self.op.secondary_ip is None:
5464 if self.primary_ip_family == netutils.IP6Address.family:
5465 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
5466 " IPv4 address must be given as secondary",
5468 self.op.secondary_ip = primary_ip
5470 secondary_ip = self.op.secondary_ip
5471 if not netutils.IP4Address.IsValid(secondary_ip):
5472 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5473 " address" % secondary_ip, errors.ECODE_INVAL)
5475 node_list = cfg.GetNodeList()
5476 if not self.op.readd and node in node_list:
5477 raise errors.OpPrereqError("Node %s is already in the configuration" %
5478 node, errors.ECODE_EXISTS)
5479 elif self.op.readd and node not in node_list:
5480 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
5483 self.changed_primary_ip = False
5485 for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list):
5486 if self.op.readd and node == existing_node_name:
5487 if existing_node.secondary_ip != secondary_ip:
5488 raise errors.OpPrereqError("Readded node doesn't have the same IP"
5489 " address configuration as before",
5491 if existing_node.primary_ip != primary_ip:
5492 self.changed_primary_ip = True
5496 if (existing_node.primary_ip == primary_ip or
5497 existing_node.secondary_ip == primary_ip or
5498 existing_node.primary_ip == secondary_ip or
5499 existing_node.secondary_ip == secondary_ip):
5500 raise errors.OpPrereqError("New node ip address(es) conflict with"
5501 " existing node %s" % existing_node.name,
5502 errors.ECODE_NOTUNIQUE)
5504 # After this 'if' block, None is no longer a valid value for the
5505 # _capable op attributes
5507 old_node = self.cfg.GetNodeInfo(node)
5508 assert old_node is not None, "Can't retrieve locked node %s" % node
5509 for attr in self._NFLAGS:
5510 if getattr(self.op, attr) is None:
5511 setattr(self.op, attr, getattr(old_node, attr))
5513 for attr in self._NFLAGS:
5514 if getattr(self.op, attr) is None:
5515 setattr(self.op, attr, True)
5517 if self.op.readd and not self.op.vm_capable:
5518 pri, sec = cfg.GetNodeInstances(node)
5520 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
5521 " flag set to false, but it already holds"
5522 " instances" % node,
5525 # check that the type of the node (single versus dual homed) is the
5526 # same as for the master
5527 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
5528 master_singlehomed = myself.secondary_ip == myself.primary_ip
5529 newbie_singlehomed = secondary_ip == primary_ip
5530 if master_singlehomed != newbie_singlehomed:
5531 if master_singlehomed:
5532 raise errors.OpPrereqError("The master has no secondary ip but the"
5533 " new node has one",
5536 raise errors.OpPrereqError("The master has a secondary ip but the"
5537 " new node doesn't have one",
5540 # checks reachability
5541 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
5542 raise errors.OpPrereqError("Node not reachable by ping",
5543 errors.ECODE_ENVIRON)
5545 if not newbie_singlehomed:
5546 # check reachability from my secondary ip to newbie's secondary ip
5547 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
5548 source=myself.secondary_ip):
5549 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5550 " based ping to node daemon port",
5551 errors.ECODE_ENVIRON)
5558 if self.op.master_capable:
5559 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
5561 self.master_candidate = False
5564 self.new_node = old_node
5566 node_group = cfg.LookupNodeGroup(self.op.group)
5567 self.new_node = objects.Node(name=node,
5568 primary_ip=primary_ip,
5569 secondary_ip=secondary_ip,
5570 master_candidate=self.master_candidate,
5571 offline=False, drained=False,
5574 if self.op.ndparams:
5575 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
5577 if self.op.hv_state:
5578 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None)
5580 if self.op.disk_state:
5581 self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None)
5583 def Exec(self, feedback_fn):
5584 """Adds the new node to the cluster.
5587 new_node = self.new_node
5588 node = new_node.name
5590 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
5593 # We adding a new node so we assume it's powered
5594 new_node.powered = True
5596 # for re-adds, reset the offline/drained/master-candidate flags;
5597 # we need to reset here, otherwise offline would prevent RPC calls
5598 # later in the procedure; this also means that if the re-add
5599 # fails, we are left with a non-offlined, broken node
5601 new_node.drained = new_node.offline = False # pylint: disable=W0201
5602 self.LogInfo("Readding a node, the offline/drained flags were reset")
5603 # if we demote the node, we do cleanup later in the procedure
5604 new_node.master_candidate = self.master_candidate
5605 if self.changed_primary_ip:
5606 new_node.primary_ip = self.op.primary_ip
5608 # copy the master/vm_capable flags
5609 for attr in self._NFLAGS:
5610 setattr(new_node, attr, getattr(self.op, attr))
5612 # notify the user about any possible mc promotion
5613 if new_node.master_candidate:
5614 self.LogInfo("Node will be a master candidate")
5616 if self.op.ndparams:
5617 new_node.ndparams = self.op.ndparams
5619 new_node.ndparams = {}
5621 if self.op.hv_state:
5622 new_node.hv_state_static = self.new_hv_state
5624 if self.op.disk_state:
5625 new_node.disk_state_static = self.new_disk_state
5627 # check connectivity
5628 result = self.rpc.call_version([node])[node]
5629 result.Raise("Can't get version information from node %s" % node)
5630 if constants.PROTOCOL_VERSION == result.payload:
5631 logging.info("Communication to node %s fine, sw version %s match",
5632 node, result.payload)
5634 raise errors.OpExecError("Version mismatch master version %s,"
5635 " node version %s" %
5636 (constants.PROTOCOL_VERSION, result.payload))
5638 # Add node to our /etc/hosts, and add key to known_hosts
5639 if self.cfg.GetClusterInfo().modify_etc_hosts:
5640 master_node = self.cfg.GetMasterNode()
5641 result = self.rpc.call_etc_hosts_modify(master_node,
5642 constants.ETC_HOSTS_ADD,
5645 result.Raise("Can't update hosts file with new host data")
5647 if new_node.secondary_ip != new_node.primary_ip:
5648 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
5651 node_verify_list = [self.cfg.GetMasterNode()]
5652 node_verify_param = {
5653 constants.NV_NODELIST: ([node], {}),
5654 # TODO: do a node-net-test as well?
5657 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
5658 self.cfg.GetClusterName())
5659 for verifier in node_verify_list:
5660 result[verifier].Raise("Cannot communicate with node %s" % verifier)
5661 nl_payload = result[verifier].payload[constants.NV_NODELIST]
5663 for failed in nl_payload:
5664 feedback_fn("ssh/hostname verification failed"
5665 " (checking from %s): %s" %
5666 (verifier, nl_payload[failed]))
5667 raise errors.OpExecError("ssh/hostname verification failed")
5670 _RedistributeAncillaryFiles(self)
5671 self.context.ReaddNode(new_node)
5672 # make sure we redistribute the config
5673 self.cfg.Update(new_node, feedback_fn)
5674 # and make sure the new node will not have old files around
5675 if not new_node.master_candidate:
5676 result = self.rpc.call_node_demote_from_mc(new_node.name)
5677 msg = result.fail_msg
5679 self.LogWarning("Node failed to demote itself from master"
5680 " candidate status: %s" % msg)
5682 _RedistributeAncillaryFiles(self, additional_nodes=[node],
5683 additional_vm=self.op.vm_capable)
5684 self.context.AddNode(new_node, self.proc.GetECId())
5687 class LUNodeSetParams(LogicalUnit):
5688 """Modifies the parameters of a node.
5690 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
5691 to the node role (as _ROLE_*)
5692 @cvar _R2F: a dictionary from node role to tuples of flags
5693 @cvar _FLAGS: a list of attribute names corresponding to the flags
5696 HPATH = "node-modify"
5697 HTYPE = constants.HTYPE_NODE
5699 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
5701 (True, False, False): _ROLE_CANDIDATE,
5702 (False, True, False): _ROLE_DRAINED,
5703 (False, False, True): _ROLE_OFFLINE,
5704 (False, False, False): _ROLE_REGULAR,
5706 _R2F = dict((v, k) for k, v in _F2R.items())
5707 _FLAGS = ["master_candidate", "drained", "offline"]
5709 def CheckArguments(self):
5710 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5711 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
5712 self.op.master_capable, self.op.vm_capable,
5713 self.op.secondary_ip, self.op.ndparams, self.op.hv_state,
5715 if all_mods.count(None) == len(all_mods):
5716 raise errors.OpPrereqError("Please pass at least one modification",
5718 if all_mods.count(True) > 1:
5719 raise errors.OpPrereqError("Can't set the node into more than one"
5720 " state at the same time",
5723 # Boolean value that tells us whether we might be demoting from MC
5724 self.might_demote = (self.op.master_candidate == False or
5725 self.op.offline == True or
5726 self.op.drained == True or
5727 self.op.master_capable == False)
5729 if self.op.secondary_ip:
5730 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
5731 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5732 " address" % self.op.secondary_ip,
5735 self.lock_all = self.op.auto_promote and self.might_demote
5736 self.lock_instances = self.op.secondary_ip is not None
5738 def _InstanceFilter(self, instance):
5739 """Filter for getting affected instances.
5742 return (instance.disk_template in constants.DTS_INT_MIRROR and
5743 self.op.node_name in instance.all_nodes)
5745 def ExpandNames(self):
5747 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
5749 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
5751 # Since modifying a node can have severe effects on currently running
5752 # operations the resource lock is at least acquired in shared mode
5753 self.needed_locks[locking.LEVEL_NODE_RES] = \
5754 self.needed_locks[locking.LEVEL_NODE]
5756 # Get node resource and instance locks in shared mode; they are not used
5757 # for anything but read-only access
5758 self.share_locks[locking.LEVEL_NODE_RES] = 1
5759 self.share_locks[locking.LEVEL_INSTANCE] = 1
5761 if self.lock_instances:
5762 self.needed_locks[locking.LEVEL_INSTANCE] = \
5763 frozenset(self.cfg.GetInstancesInfoByFilter(self._InstanceFilter))
5765 def BuildHooksEnv(self):
5768 This runs on the master node.
5772 "OP_TARGET": self.op.node_name,
5773 "MASTER_CANDIDATE": str(self.op.master_candidate),
5774 "OFFLINE": str(self.op.offline),
5775 "DRAINED": str(self.op.drained),
5776 "MASTER_CAPABLE": str(self.op.master_capable),
5777 "VM_CAPABLE": str(self.op.vm_capable),
5780 def BuildHooksNodes(self):
5781 """Build hooks nodes.
5784 nl = [self.cfg.GetMasterNode(), self.op.node_name]
5787 def CheckPrereq(self):
5788 """Check prerequisites.
5790 This only checks the instance list against the existing names.
5793 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
5795 if self.lock_instances:
5796 affected_instances = \
5797 self.cfg.GetInstancesInfoByFilter(self._InstanceFilter)
5799 # Verify instance locks
5800 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
5801 wanted_instances = frozenset(affected_instances.keys())
5802 if wanted_instances - owned_instances:
5803 raise errors.OpPrereqError("Instances affected by changing node %s's"
5804 " secondary IP address have changed since"
5805 " locks were acquired, wanted '%s', have"
5806 " '%s'; retry the operation" %
5808 utils.CommaJoin(wanted_instances),
5809 utils.CommaJoin(owned_instances)),
5812 affected_instances = None
5814 if (self.op.master_candidate is not None or
5815 self.op.drained is not None or
5816 self.op.offline is not None):
5817 # we can't change the master's node flags
5818 if self.op.node_name == self.cfg.GetMasterNode():
5819 raise errors.OpPrereqError("The master role can be changed"
5820 " only via master-failover",
5823 if self.op.master_candidate and not node.master_capable:
5824 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
5825 " it a master candidate" % node.name,
5828 if self.op.vm_capable == False:
5829 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
5831 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
5832 " the vm_capable flag" % node.name,
5835 if node.master_candidate and self.might_demote and not self.lock_all:
5836 assert not self.op.auto_promote, "auto_promote set but lock_all not"
5837 # check if after removing the current node, we're missing master
5839 (mc_remaining, mc_should, _) = \
5840 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
5841 if mc_remaining < mc_should:
5842 raise errors.OpPrereqError("Not enough master candidates, please"
5843 " pass auto promote option to allow"
5844 " promotion", errors.ECODE_STATE)
5846 self.old_flags = old_flags = (node.master_candidate,
5847 node.drained, node.offline)
5848 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
5849 self.old_role = old_role = self._F2R[old_flags]
5851 # Check for ineffective changes
5852 for attr in self._FLAGS:
5853 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
5854 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
5855 setattr(self.op, attr, None)
5857 # Past this point, any flag change to False means a transition
5858 # away from the respective state, as only real changes are kept
5860 # TODO: We might query the real power state if it supports OOB
5861 if _SupportsOob(self.cfg, node):
5862 if self.op.offline is False and not (node.powered or
5863 self.op.powered == True):
5864 raise errors.OpPrereqError(("Node %s needs to be turned on before its"
5865 " offline status can be reset") %
5867 elif self.op.powered is not None:
5868 raise errors.OpPrereqError(("Unable to change powered state for node %s"
5869 " as it does not support out-of-band"
5870 " handling") % self.op.node_name)
5872 # If we're being deofflined/drained, we'll MC ourself if needed
5873 if (self.op.drained == False or self.op.offline == False or
5874 (self.op.master_capable and not node.master_capable)):
5875 if _DecideSelfPromotion(self):
5876 self.op.master_candidate = True
5877 self.LogInfo("Auto-promoting node to master candidate")
5879 # If we're no longer master capable, we'll demote ourselves from MC
5880 if self.op.master_capable == False and node.master_candidate:
5881 self.LogInfo("Demoting from master candidate")
5882 self.op.master_candidate = False
5885 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
5886 if self.op.master_candidate:
5887 new_role = self._ROLE_CANDIDATE
5888 elif self.op.drained:
5889 new_role = self._ROLE_DRAINED
5890 elif self.op.offline:
5891 new_role = self._ROLE_OFFLINE
5892 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
5893 # False is still in new flags, which means we're un-setting (the
5895 new_role = self._ROLE_REGULAR
5896 else: # no new flags, nothing, keep old role
5899 self.new_role = new_role
5901 if old_role == self._ROLE_OFFLINE and new_role != old_role:
5902 # Trying to transition out of offline status
5903 # TODO: Use standard RPC runner, but make sure it works when the node is
5904 # still marked offline
5905 result = rpc.BootstrapRunner().call_version([node.name])[node.name]
5907 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
5908 " to report its version: %s" %
5909 (node.name, result.fail_msg),
5912 self.LogWarning("Transitioning node from offline to online state"
5913 " without using re-add. Please make sure the node"
5916 if self.op.secondary_ip:
5917 # Ok even without locking, because this can't be changed by any LU
5918 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
5919 master_singlehomed = master.secondary_ip == master.primary_ip
5920 if master_singlehomed and self.op.secondary_ip:
5921 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
5922 " homed cluster", errors.ECODE_INVAL)
5924 assert not (frozenset(affected_instances) -
5925 self.owned_locks(locking.LEVEL_INSTANCE))
5928 if affected_instances:
5929 raise errors.OpPrereqError("Cannot change secondary IP address:"
5930 " offline node has instances (%s)"
5931 " configured to use it" %
5932 utils.CommaJoin(affected_instances.keys()))
5934 # On online nodes, check that no instances are running, and that
5935 # the node has the new ip and we can reach it.
5936 for instance in affected_instances.values():
5937 _CheckInstanceState(self, instance, INSTANCE_DOWN,
5938 msg="cannot change secondary ip")
5940 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
5941 if master.name != node.name:
5942 # check reachability from master secondary ip to new secondary ip
5943 if not netutils.TcpPing(self.op.secondary_ip,
5944 constants.DEFAULT_NODED_PORT,
5945 source=master.secondary_ip):
5946 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5947 " based ping to node daemon port",
5948 errors.ECODE_ENVIRON)
5950 if self.op.ndparams:
5951 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
5952 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
5953 self.new_ndparams = new_ndparams
5955 if self.op.hv_state:
5956 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
5957 self.node.hv_state_static)
5959 if self.op.disk_state:
5960 self.new_disk_state = \
5961 _MergeAndVerifyDiskState(self.op.disk_state,
5962 self.node.disk_state_static)
5964 def Exec(self, feedback_fn):
5969 old_role = self.old_role
5970 new_role = self.new_role
5974 if self.op.ndparams:
5975 node.ndparams = self.new_ndparams
5977 if self.op.powered is not None:
5978 node.powered = self.op.powered
5980 if self.op.hv_state:
5981 node.hv_state_static = self.new_hv_state
5983 if self.op.disk_state:
5984 node.disk_state_static = self.new_disk_state
5986 for attr in ["master_capable", "vm_capable"]:
5987 val = getattr(self.op, attr)
5989 setattr(node, attr, val)
5990 result.append((attr, str(val)))
5992 if new_role != old_role:
5993 # Tell the node to demote itself, if no longer MC and not offline
5994 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
5995 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
5997 self.LogWarning("Node failed to demote itself: %s", msg)
5999 new_flags = self._R2F[new_role]
6000 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
6002 result.append((desc, str(nf)))
6003 (node.master_candidate, node.drained, node.offline) = new_flags
6005 # we locked all nodes, we adjust the CP before updating this node
6007 _AdjustCandidatePool(self, [node.name])
6009 if self.op.secondary_ip:
6010 node.secondary_ip = self.op.secondary_ip
6011 result.append(("secondary_ip", self.op.secondary_ip))
6013 # this will trigger configuration file update, if needed
6014 self.cfg.Update(node, feedback_fn)
6016 # this will trigger job queue propagation or cleanup if the mc
6018 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
6019 self.context.ReaddNode(node)
6024 class LUNodePowercycle(NoHooksLU):
6025 """Powercycles a node.
6030 def CheckArguments(self):
6031 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6032 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
6033 raise errors.OpPrereqError("The node is the master and the force"
6034 " parameter was not set",
6037 def ExpandNames(self):
6038 """Locking for PowercycleNode.
6040 This is a last-resort option and shouldn't block on other
6041 jobs. Therefore, we grab no locks.
6044 self.needed_locks = {}
6046 def Exec(self, feedback_fn):
6050 result = self.rpc.call_node_powercycle(self.op.node_name,
6051 self.cfg.GetHypervisorType())
6052 result.Raise("Failed to schedule the reboot")
6053 return result.payload
6056 class LUClusterQuery(NoHooksLU):
6057 """Query cluster configuration.
6062 def ExpandNames(self):
6063 self.needed_locks = {}
6065 def Exec(self, feedback_fn):
6066 """Return cluster config.
6069 cluster = self.cfg.GetClusterInfo()
6072 # Filter just for enabled hypervisors
6073 for os_name, hv_dict in cluster.os_hvp.items():
6074 os_hvp[os_name] = {}
6075 for hv_name, hv_params in hv_dict.items():
6076 if hv_name in cluster.enabled_hypervisors:
6077 os_hvp[os_name][hv_name] = hv_params
6079 # Convert ip_family to ip_version
6080 primary_ip_version = constants.IP4_VERSION
6081 if cluster.primary_ip_family == netutils.IP6Address.family:
6082 primary_ip_version = constants.IP6_VERSION
6085 "software_version": constants.RELEASE_VERSION,
6086 "protocol_version": constants.PROTOCOL_VERSION,
6087 "config_version": constants.CONFIG_VERSION,
6088 "os_api_version": max(constants.OS_API_VERSIONS),
6089 "export_version": constants.EXPORT_VERSION,
6090 "architecture": (platform.architecture()[0], platform.machine()),
6091 "name": cluster.cluster_name,
6092 "master": cluster.master_node,
6093 "default_hypervisor": cluster.primary_hypervisor,
6094 "enabled_hypervisors": cluster.enabled_hypervisors,
6095 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
6096 for hypervisor_name in cluster.enabled_hypervisors]),
6098 "beparams": cluster.beparams,
6099 "osparams": cluster.osparams,
6100 "ipolicy": cluster.ipolicy,
6101 "nicparams": cluster.nicparams,
6102 "ndparams": cluster.ndparams,
6103 "candidate_pool_size": cluster.candidate_pool_size,
6104 "master_netdev": cluster.master_netdev,
6105 "master_netmask": cluster.master_netmask,
6106 "use_external_mip_script": cluster.use_external_mip_script,
6107 "volume_group_name": cluster.volume_group_name,
6108 "drbd_usermode_helper": cluster.drbd_usermode_helper,
6109 "file_storage_dir": cluster.file_storage_dir,
6110 "shared_file_storage_dir": cluster.shared_file_storage_dir,
6111 "maintain_node_health": cluster.maintain_node_health,
6112 "ctime": cluster.ctime,
6113 "mtime": cluster.mtime,
6114 "uuid": cluster.uuid,
6115 "tags": list(cluster.GetTags()),
6116 "uid_pool": cluster.uid_pool,
6117 "default_iallocator": cluster.default_iallocator,
6118 "reserved_lvs": cluster.reserved_lvs,
6119 "primary_ip_version": primary_ip_version,
6120 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
6121 "hidden_os": cluster.hidden_os,
6122 "blacklisted_os": cluster.blacklisted_os,
6128 class LUClusterConfigQuery(NoHooksLU):
6129 """Return configuration values.
6133 _FIELDS_DYNAMIC = utils.FieldSet()
6134 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
6135 "watcher_pause", "volume_group_name")
6137 def CheckArguments(self):
6138 _CheckOutputFields(static=self._FIELDS_STATIC,
6139 dynamic=self._FIELDS_DYNAMIC,
6140 selected=self.op.output_fields)
6142 def ExpandNames(self):
6143 self.needed_locks = {}
6145 def Exec(self, feedback_fn):
6146 """Dump a representation of the cluster config to the standard output.
6150 for field in self.op.output_fields:
6151 if field == "cluster_name":
6152 entry = self.cfg.GetClusterName()
6153 elif field == "master_node":
6154 entry = self.cfg.GetMasterNode()
6155 elif field == "drain_flag":
6156 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
6157 elif field == "watcher_pause":
6158 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
6159 elif field == "volume_group_name":
6160 entry = self.cfg.GetVGName()
6162 raise errors.ParameterError(field)
6163 values.append(entry)
6167 class LUInstanceActivateDisks(NoHooksLU):
6168 """Bring up an instance's disks.
6173 def ExpandNames(self):
6174 self._ExpandAndLockInstance()
6175 self.needed_locks[locking.LEVEL_NODE] = []
6176 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6178 def DeclareLocks(self, level):
6179 if level == locking.LEVEL_NODE:
6180 self._LockInstancesNodes()
6182 def CheckPrereq(self):
6183 """Check prerequisites.
6185 This checks that the instance is in the cluster.
6188 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6189 assert self.instance is not None, \
6190 "Cannot retrieve locked instance %s" % self.op.instance_name
6191 _CheckNodeOnline(self, self.instance.primary_node)
6193 def Exec(self, feedback_fn):
6194 """Activate the disks.
6197 disks_ok, disks_info = \
6198 _AssembleInstanceDisks(self, self.instance,
6199 ignore_size=self.op.ignore_size)
6201 raise errors.OpExecError("Cannot activate block devices")
6206 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
6208 """Prepare the block devices for an instance.
6210 This sets up the block devices on all nodes.
6212 @type lu: L{LogicalUnit}
6213 @param lu: the logical unit on whose behalf we execute
6214 @type instance: L{objects.Instance}
6215 @param instance: the instance for whose disks we assemble
6216 @type disks: list of L{objects.Disk} or None
6217 @param disks: which disks to assemble (or all, if None)
6218 @type ignore_secondaries: boolean
6219 @param ignore_secondaries: if true, errors on secondary nodes
6220 won't result in an error return from the function
6221 @type ignore_size: boolean
6222 @param ignore_size: if true, the current known size of the disk
6223 will not be used during the disk activation, useful for cases
6224 when the size is wrong
6225 @return: False if the operation failed, otherwise a list of
6226 (host, instance_visible_name, node_visible_name)
6227 with the mapping from node devices to instance devices
6232 iname = instance.name
6233 disks = _ExpandCheckDisks(instance, disks)
6235 # With the two passes mechanism we try to reduce the window of
6236 # opportunity for the race condition of switching DRBD to primary
6237 # before handshaking occured, but we do not eliminate it
6239 # The proper fix would be to wait (with some limits) until the
6240 # connection has been made and drbd transitions from WFConnection
6241 # into any other network-connected state (Connected, SyncTarget,
6244 # 1st pass, assemble on all nodes in secondary mode
6245 for idx, inst_disk in enumerate(disks):
6246 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
6248 node_disk = node_disk.Copy()
6249 node_disk.UnsetSize()
6250 lu.cfg.SetDiskID(node_disk, node)
6251 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
6252 msg = result.fail_msg
6254 lu.proc.LogWarning("Could not prepare block device %s on node %s"
6255 " (is_primary=False, pass=1): %s",
6256 inst_disk.iv_name, node, msg)
6257 if not ignore_secondaries:
6260 # FIXME: race condition on drbd migration to primary
6262 # 2nd pass, do only the primary node
6263 for idx, inst_disk in enumerate(disks):
6266 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
6267 if node != instance.primary_node:
6270 node_disk = node_disk.Copy()
6271 node_disk.UnsetSize()
6272 lu.cfg.SetDiskID(node_disk, node)
6273 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
6274 msg = result.fail_msg
6276 lu.proc.LogWarning("Could not prepare block device %s on node %s"
6277 " (is_primary=True, pass=2): %s",
6278 inst_disk.iv_name, node, msg)
6281 dev_path = result.payload
6283 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
6285 # leave the disks configured for the primary node
6286 # this is a workaround that would be fixed better by
6287 # improving the logical/physical id handling
6289 lu.cfg.SetDiskID(disk, instance.primary_node)
6291 return disks_ok, device_info
6294 def _StartInstanceDisks(lu, instance, force):
6295 """Start the disks of an instance.
6298 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
6299 ignore_secondaries=force)
6301 _ShutdownInstanceDisks(lu, instance)
6302 if force is not None and not force:
6303 lu.proc.LogWarning("", hint="If the message above refers to a"
6305 " you can retry the operation using '--force'.")
6306 raise errors.OpExecError("Disk consistency error")
6309 class LUInstanceDeactivateDisks(NoHooksLU):
6310 """Shutdown an instance's disks.
6315 def ExpandNames(self):
6316 self._ExpandAndLockInstance()
6317 self.needed_locks[locking.LEVEL_NODE] = []
6318 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6320 def DeclareLocks(self, level):
6321 if level == locking.LEVEL_NODE:
6322 self._LockInstancesNodes()
6324 def CheckPrereq(self):
6325 """Check prerequisites.
6327 This checks that the instance is in the cluster.
6330 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6331 assert self.instance is not None, \
6332 "Cannot retrieve locked instance %s" % self.op.instance_name
6334 def Exec(self, feedback_fn):
6335 """Deactivate the disks
6338 instance = self.instance
6340 _ShutdownInstanceDisks(self, instance)
6342 _SafeShutdownInstanceDisks(self, instance)
6345 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
6346 """Shutdown block devices of an instance.
6348 This function checks if an instance is running, before calling
6349 _ShutdownInstanceDisks.
6352 _CheckInstanceState(lu, instance, INSTANCE_DOWN, msg="cannot shutdown disks")
6353 _ShutdownInstanceDisks(lu, instance, disks=disks)
6356 def _ExpandCheckDisks(instance, disks):
6357 """Return the instance disks selected by the disks list
6359 @type disks: list of L{objects.Disk} or None
6360 @param disks: selected disks
6361 @rtype: list of L{objects.Disk}
6362 @return: selected instance disks to act on
6366 return instance.disks
6368 if not set(disks).issubset(instance.disks):
6369 raise errors.ProgrammerError("Can only act on disks belonging to the"
6374 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
6375 """Shutdown block devices of an instance.
6377 This does the shutdown on all nodes of the instance.
6379 If the ignore_primary is false, errors on the primary node are
6384 disks = _ExpandCheckDisks(instance, disks)
6387 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
6388 lu.cfg.SetDiskID(top_disk, node)
6389 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
6390 msg = result.fail_msg
6392 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
6393 disk.iv_name, node, msg)
6394 if ((node == instance.primary_node and not ignore_primary) or
6395 (node != instance.primary_node and not result.offline)):
6400 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
6401 """Checks if a node has enough free memory.
6403 This function check if a given node has the needed amount of free
6404 memory. In case the node has less memory or we cannot get the
6405 information from the node, this function raise an OpPrereqError
6408 @type lu: C{LogicalUnit}
6409 @param lu: a logical unit from which we get configuration data
6411 @param node: the node to check
6412 @type reason: C{str}
6413 @param reason: string to use in the error message
6414 @type requested: C{int}
6415 @param requested: the amount of memory in MiB to check for
6416 @type hypervisor_name: C{str}
6417 @param hypervisor_name: the hypervisor to ask for memory stats
6419 @return: node current free memory
6420 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
6421 we cannot check the node
6424 nodeinfo = lu.rpc.call_node_info([node], None, [hypervisor_name])
6425 nodeinfo[node].Raise("Can't get data from node %s" % node,
6426 prereq=True, ecode=errors.ECODE_ENVIRON)
6427 (_, _, (hv_info, )) = nodeinfo[node].payload
6429 free_mem = hv_info.get("memory_free", None)
6430 if not isinstance(free_mem, int):
6431 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
6432 " was '%s'" % (node, free_mem),
6433 errors.ECODE_ENVIRON)
6434 if requested > free_mem:
6435 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
6436 " needed %s MiB, available %s MiB" %
6437 (node, reason, requested, free_mem),
6442 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
6443 """Checks if nodes have enough free disk space in the all VGs.
6445 This function check if all given nodes have the needed amount of
6446 free disk. In case any node has less disk or we cannot get the
6447 information from the node, this function raise an OpPrereqError
6450 @type lu: C{LogicalUnit}
6451 @param lu: a logical unit from which we get configuration data
6452 @type nodenames: C{list}
6453 @param nodenames: the list of node names to check
6454 @type req_sizes: C{dict}
6455 @param req_sizes: the hash of vg and corresponding amount of disk in
6457 @raise errors.OpPrereqError: if the node doesn't have enough disk,
6458 or we cannot check the node
6461 for vg, req_size in req_sizes.items():
6462 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
6465 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
6466 """Checks if nodes have enough free disk space in the specified VG.
6468 This function check if all given nodes have the needed amount of
6469 free disk. In case any node has less disk or we cannot get the
6470 information from the node, this function raise an OpPrereqError
6473 @type lu: C{LogicalUnit}
6474 @param lu: a logical unit from which we get configuration data
6475 @type nodenames: C{list}
6476 @param nodenames: the list of node names to check
6478 @param vg: the volume group to check
6479 @type requested: C{int}
6480 @param requested: the amount of disk in MiB to check for
6481 @raise errors.OpPrereqError: if the node doesn't have enough disk,
6482 or we cannot check the node
6485 nodeinfo = lu.rpc.call_node_info(nodenames, [vg], None)
6486 for node in nodenames:
6487 info = nodeinfo[node]
6488 info.Raise("Cannot get current information from node %s" % node,
6489 prereq=True, ecode=errors.ECODE_ENVIRON)
6490 (_, (vg_info, ), _) = info.payload
6491 vg_free = vg_info.get("vg_free", None)
6492 if not isinstance(vg_free, int):
6493 raise errors.OpPrereqError("Can't compute free disk space on node"
6494 " %s for vg %s, result was '%s'" %
6495 (node, vg, vg_free), errors.ECODE_ENVIRON)
6496 if requested > vg_free:
6497 raise errors.OpPrereqError("Not enough disk space on target node %s"
6498 " vg %s: required %d MiB, available %d MiB" %
6499 (node, vg, requested, vg_free),
6503 def _CheckNodesPhysicalCPUs(lu, nodenames, requested, hypervisor_name):
6504 """Checks if nodes have enough physical CPUs
6506 This function checks if all given nodes have the needed number of
6507 physical CPUs. In case any node has less CPUs or we cannot get the
6508 information from the node, this function raises an OpPrereqError
6511 @type lu: C{LogicalUnit}
6512 @param lu: a logical unit from which we get configuration data
6513 @type nodenames: C{list}
6514 @param nodenames: the list of node names to check
6515 @type requested: C{int}
6516 @param requested: the minimum acceptable number of physical CPUs
6517 @raise errors.OpPrereqError: if the node doesn't have enough CPUs,
6518 or we cannot check the node
6521 nodeinfo = lu.rpc.call_node_info(nodenames, None, [hypervisor_name])
6522 for node in nodenames:
6523 info = nodeinfo[node]
6524 info.Raise("Cannot get current information from node %s" % node,
6525 prereq=True, ecode=errors.ECODE_ENVIRON)
6526 (_, _, (hv_info, )) = info.payload
6527 num_cpus = hv_info.get("cpu_total", None)
6528 if not isinstance(num_cpus, int):
6529 raise errors.OpPrereqError("Can't compute the number of physical CPUs"
6530 " on node %s, result was '%s'" %
6531 (node, num_cpus), errors.ECODE_ENVIRON)
6532 if requested > num_cpus:
6533 raise errors.OpPrereqError("Node %s has %s physical CPUs, but %s are "
6534 "required" % (node, num_cpus, requested),
6538 class LUInstanceStartup(LogicalUnit):
6539 """Starts an instance.
6542 HPATH = "instance-start"
6543 HTYPE = constants.HTYPE_INSTANCE
6546 def CheckArguments(self):
6548 if self.op.beparams:
6549 # fill the beparams dict
6550 objects.UpgradeBeParams(self.op.beparams)
6551 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6553 def ExpandNames(self):
6554 self._ExpandAndLockInstance()
6555 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
6557 def DeclareLocks(self, level):
6558 if level == locking.LEVEL_NODE_RES:
6559 self._LockInstancesNodes(primary_only=True, level=locking.LEVEL_NODE_RES)
6561 def BuildHooksEnv(self):
6564 This runs on master, primary and secondary nodes of the instance.
6568 "FORCE": self.op.force,
6571 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6575 def BuildHooksNodes(self):
6576 """Build hooks nodes.
6579 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6582 def CheckPrereq(self):
6583 """Check prerequisites.
6585 This checks that the instance is in the cluster.
6588 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6589 assert self.instance is not None, \
6590 "Cannot retrieve locked instance %s" % self.op.instance_name
6593 if self.op.hvparams:
6594 # check hypervisor parameter syntax (locally)
6595 cluster = self.cfg.GetClusterInfo()
6596 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6597 filled_hvp = cluster.FillHV(instance)
6598 filled_hvp.update(self.op.hvparams)
6599 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
6600 hv_type.CheckParameterSyntax(filled_hvp)
6601 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
6603 _CheckInstanceState(self, instance, INSTANCE_ONLINE)
6605 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
6607 if self.primary_offline and self.op.ignore_offline_nodes:
6608 self.proc.LogWarning("Ignoring offline primary node")
6610 if self.op.hvparams or self.op.beparams:
6611 self.proc.LogWarning("Overridden parameters are ignored")
6613 _CheckNodeOnline(self, instance.primary_node)
6615 bep = self.cfg.GetClusterInfo().FillBE(instance)
6616 bep.update(self.op.beparams)
6618 # check bridges existence
6619 _CheckInstanceBridgesExist(self, instance)
6621 remote_info = self.rpc.call_instance_info(instance.primary_node,
6623 instance.hypervisor)
6624 remote_info.Raise("Error checking node %s" % instance.primary_node,
6625 prereq=True, ecode=errors.ECODE_ENVIRON)
6626 if not remote_info.payload: # not running already
6627 _CheckNodeFreeMemory(self, instance.primary_node,
6628 "starting instance %s" % instance.name,
6629 bep[constants.BE_MINMEM], instance.hypervisor)
6631 def Exec(self, feedback_fn):
6632 """Start the instance.
6635 instance = self.instance
6636 force = self.op.force
6638 if not self.op.no_remember:
6639 self.cfg.MarkInstanceUp(instance.name)
6641 if self.primary_offline:
6642 assert self.op.ignore_offline_nodes
6643 self.proc.LogInfo("Primary node offline, marked instance as started")
6645 node_current = instance.primary_node
6647 _StartInstanceDisks(self, instance, force)
6650 self.rpc.call_instance_start(node_current,
6651 (instance, self.op.hvparams,
6653 self.op.startup_paused)
6654 msg = result.fail_msg
6656 _ShutdownInstanceDisks(self, instance)
6657 raise errors.OpExecError("Could not start instance: %s" % msg)
6660 class LUInstanceReboot(LogicalUnit):
6661 """Reboot an instance.
6664 HPATH = "instance-reboot"
6665 HTYPE = constants.HTYPE_INSTANCE
6668 def ExpandNames(self):
6669 self._ExpandAndLockInstance()
6671 def BuildHooksEnv(self):
6674 This runs on master, primary and secondary nodes of the instance.
6678 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
6679 "REBOOT_TYPE": self.op.reboot_type,
6680 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6683 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6687 def BuildHooksNodes(self):
6688 """Build hooks nodes.
6691 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6694 def CheckPrereq(self):
6695 """Check prerequisites.
6697 This checks that the instance is in the cluster.
6700 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6701 assert self.instance is not None, \
6702 "Cannot retrieve locked instance %s" % self.op.instance_name
6703 _CheckInstanceState(self, instance, INSTANCE_ONLINE)
6704 _CheckNodeOnline(self, instance.primary_node)
6706 # check bridges existence
6707 _CheckInstanceBridgesExist(self, instance)
6709 def Exec(self, feedback_fn):
6710 """Reboot the instance.
6713 instance = self.instance
6714 ignore_secondaries = self.op.ignore_secondaries
6715 reboot_type = self.op.reboot_type
6717 remote_info = self.rpc.call_instance_info(instance.primary_node,
6719 instance.hypervisor)
6720 remote_info.Raise("Error checking node %s" % instance.primary_node)
6721 instance_running = bool(remote_info.payload)
6723 node_current = instance.primary_node
6725 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
6726 constants.INSTANCE_REBOOT_HARD]:
6727 for disk in instance.disks:
6728 self.cfg.SetDiskID(disk, node_current)
6729 result = self.rpc.call_instance_reboot(node_current, instance,
6731 self.op.shutdown_timeout)
6732 result.Raise("Could not reboot instance")
6734 if instance_running:
6735 result = self.rpc.call_instance_shutdown(node_current, instance,
6736 self.op.shutdown_timeout)
6737 result.Raise("Could not shutdown instance for full reboot")
6738 _ShutdownInstanceDisks(self, instance)
6740 self.LogInfo("Instance %s was already stopped, starting now",
6742 _StartInstanceDisks(self, instance, ignore_secondaries)
6743 result = self.rpc.call_instance_start(node_current,
6744 (instance, None, None), False)
6745 msg = result.fail_msg
6747 _ShutdownInstanceDisks(self, instance)
6748 raise errors.OpExecError("Could not start instance for"
6749 " full reboot: %s" % msg)
6751 self.cfg.MarkInstanceUp(instance.name)
6754 class LUInstanceShutdown(LogicalUnit):
6755 """Shutdown an instance.
6758 HPATH = "instance-stop"
6759 HTYPE = constants.HTYPE_INSTANCE
6762 def ExpandNames(self):
6763 self._ExpandAndLockInstance()
6765 def BuildHooksEnv(self):
6768 This runs on master, primary and secondary nodes of the instance.
6771 env = _BuildInstanceHookEnvByObject(self, self.instance)
6772 env["TIMEOUT"] = self.op.timeout
6775 def BuildHooksNodes(self):
6776 """Build hooks nodes.
6779 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6782 def CheckPrereq(self):
6783 """Check prerequisites.
6785 This checks that the instance is in the cluster.
6788 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6789 assert self.instance is not None, \
6790 "Cannot retrieve locked instance %s" % self.op.instance_name
6792 _CheckInstanceState(self, self.instance, INSTANCE_ONLINE)
6794 self.primary_offline = \
6795 self.cfg.GetNodeInfo(self.instance.primary_node).offline
6797 if self.primary_offline and self.op.ignore_offline_nodes:
6798 self.proc.LogWarning("Ignoring offline primary node")
6800 _CheckNodeOnline(self, self.instance.primary_node)
6802 def Exec(self, feedback_fn):
6803 """Shutdown the instance.
6806 instance = self.instance
6807 node_current = instance.primary_node
6808 timeout = self.op.timeout
6810 if not self.op.no_remember:
6811 self.cfg.MarkInstanceDown(instance.name)
6813 if self.primary_offline:
6814 assert self.op.ignore_offline_nodes
6815 self.proc.LogInfo("Primary node offline, marked instance as stopped")
6817 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
6818 msg = result.fail_msg
6820 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
6822 _ShutdownInstanceDisks(self, instance)
6825 class LUInstanceReinstall(LogicalUnit):
6826 """Reinstall an instance.
6829 HPATH = "instance-reinstall"
6830 HTYPE = constants.HTYPE_INSTANCE
6833 def ExpandNames(self):
6834 self._ExpandAndLockInstance()
6836 def BuildHooksEnv(self):
6839 This runs on master, primary and secondary nodes of the instance.
6842 return _BuildInstanceHookEnvByObject(self, self.instance)
6844 def BuildHooksNodes(self):
6845 """Build hooks nodes.
6848 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6851 def CheckPrereq(self):
6852 """Check prerequisites.
6854 This checks that the instance is in the cluster and is not running.
6857 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6858 assert instance is not None, \
6859 "Cannot retrieve locked instance %s" % self.op.instance_name
6860 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
6861 " offline, cannot reinstall")
6862 for node in instance.secondary_nodes:
6863 _CheckNodeOnline(self, node, "Instance secondary node offline,"
6864 " cannot reinstall")
6866 if instance.disk_template == constants.DT_DISKLESS:
6867 raise errors.OpPrereqError("Instance '%s' has no disks" %
6868 self.op.instance_name,
6870 _CheckInstanceState(self, instance, INSTANCE_DOWN, msg="cannot reinstall")
6872 if self.op.os_type is not None:
6874 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
6875 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
6876 instance_os = self.op.os_type
6878 instance_os = instance.os
6880 nodelist = list(instance.all_nodes)
6882 if self.op.osparams:
6883 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
6884 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
6885 self.os_inst = i_osdict # the new dict (without defaults)
6889 self.instance = instance
6891 def Exec(self, feedback_fn):
6892 """Reinstall the instance.
6895 inst = self.instance
6897 if self.op.os_type is not None:
6898 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
6899 inst.os = self.op.os_type
6900 # Write to configuration
6901 self.cfg.Update(inst, feedback_fn)
6903 _StartInstanceDisks(self, inst, None)
6905 feedback_fn("Running the instance OS create scripts...")
6906 # FIXME: pass debug option from opcode to backend
6907 result = self.rpc.call_instance_os_add(inst.primary_node,
6908 (inst, self.os_inst), True,
6909 self.op.debug_level)
6910 result.Raise("Could not install OS for instance %s on node %s" %
6911 (inst.name, inst.primary_node))
6913 _ShutdownInstanceDisks(self, inst)
6916 class LUInstanceRecreateDisks(LogicalUnit):
6917 """Recreate an instance's missing disks.
6920 HPATH = "instance-recreate-disks"
6921 HTYPE = constants.HTYPE_INSTANCE
6924 _MODIFYABLE = frozenset([
6925 constants.IDISK_SIZE,
6926 constants.IDISK_MODE,
6929 # New or changed disk parameters may have different semantics
6930 assert constants.IDISK_PARAMS == (_MODIFYABLE | frozenset([
6931 constants.IDISK_ADOPT,
6933 # TODO: Implement support changing VG while recreating
6935 constants.IDISK_METAVG,
6938 def CheckArguments(self):
6939 if self.op.disks and ht.TPositiveInt(self.op.disks[0]):
6940 # Normalize and convert deprecated list of disk indices
6941 self.op.disks = [(idx, {}) for idx in sorted(frozenset(self.op.disks))]
6943 duplicates = utils.FindDuplicates(map(compat.fst, self.op.disks))
6945 raise errors.OpPrereqError("Some disks have been specified more than"
6946 " once: %s" % utils.CommaJoin(duplicates),
6949 for (idx, params) in self.op.disks:
6950 utils.ForceDictType(params, constants.IDISK_PARAMS_TYPES)
6951 unsupported = frozenset(params.keys()) - self._MODIFYABLE
6953 raise errors.OpPrereqError("Parameters for disk %s try to change"
6954 " unmodifyable parameter(s): %s" %
6955 (idx, utils.CommaJoin(unsupported)),
6958 def ExpandNames(self):
6959 self._ExpandAndLockInstance()
6960 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6962 self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes]
6963 self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes)
6965 self.needed_locks[locking.LEVEL_NODE] = []
6966 self.needed_locks[locking.LEVEL_NODE_RES] = []
6968 def DeclareLocks(self, level):
6969 if level == locking.LEVEL_NODE:
6970 # if we replace the nodes, we only need to lock the old primary,
6971 # otherwise we need to lock all nodes for disk re-creation
6972 primary_only = bool(self.op.nodes)
6973 self._LockInstancesNodes(primary_only=primary_only)
6974 elif level == locking.LEVEL_NODE_RES:
6976 self.needed_locks[locking.LEVEL_NODE_RES] = \
6977 self.needed_locks[locking.LEVEL_NODE][:]
6979 def BuildHooksEnv(self):
6982 This runs on master, primary and secondary nodes of the instance.
6985 return _BuildInstanceHookEnvByObject(self, self.instance)
6987 def BuildHooksNodes(self):
6988 """Build hooks nodes.
6991 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6994 def CheckPrereq(self):
6995 """Check prerequisites.
6997 This checks that the instance is in the cluster and is not running.
7000 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7001 assert instance is not None, \
7002 "Cannot retrieve locked instance %s" % self.op.instance_name
7004 if len(self.op.nodes) != len(instance.all_nodes):
7005 raise errors.OpPrereqError("Instance %s currently has %d nodes, but"
7006 " %d replacement nodes were specified" %
7007 (instance.name, len(instance.all_nodes),
7008 len(self.op.nodes)),
7010 assert instance.disk_template != constants.DT_DRBD8 or \
7011 len(self.op.nodes) == 2
7012 assert instance.disk_template != constants.DT_PLAIN or \
7013 len(self.op.nodes) == 1
7014 primary_node = self.op.nodes[0]
7016 primary_node = instance.primary_node
7017 _CheckNodeOnline(self, primary_node)
7019 if instance.disk_template == constants.DT_DISKLESS:
7020 raise errors.OpPrereqError("Instance '%s' has no disks" %
7021 self.op.instance_name, errors.ECODE_INVAL)
7023 # if we replace nodes *and* the old primary is offline, we don't
7025 assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE)
7026 assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE_RES)
7027 old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
7028 if not (self.op.nodes and old_pnode.offline):
7029 _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
7030 msg="cannot recreate disks")
7033 self.disks = dict(self.op.disks)
7035 self.disks = dict((idx, {}) for idx in range(len(instance.disks)))
7037 maxidx = max(self.disks.keys())
7038 if maxidx >= len(instance.disks):
7039 raise errors.OpPrereqError("Invalid disk index '%s'" % maxidx,
7042 if (self.op.nodes and
7043 sorted(self.disks.keys()) != range(len(instance.disks))):
7044 raise errors.OpPrereqError("Can't recreate disks partially and"
7045 " change the nodes at the same time",
7048 self.instance = instance
7050 def Exec(self, feedback_fn):
7051 """Recreate the disks.
7054 instance = self.instance
7056 assert (self.owned_locks(locking.LEVEL_NODE) ==
7057 self.owned_locks(locking.LEVEL_NODE_RES))
7060 mods = [] # keeps track of needed changes
7062 for idx, disk in enumerate(instance.disks):
7064 changes = self.disks[idx]
7066 # Disk should not be recreated
7070 # update secondaries for disks, if needed
7071 if self.op.nodes and disk.dev_type == constants.LD_DRBD8:
7072 # need to update the nodes and minors
7073 assert len(self.op.nodes) == 2
7074 assert len(disk.logical_id) == 6 # otherwise disk internals
7076 (_, _, old_port, _, _, old_secret) = disk.logical_id
7077 new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name)
7078 new_id = (self.op.nodes[0], self.op.nodes[1], old_port,
7079 new_minors[0], new_minors[1], old_secret)
7080 assert len(disk.logical_id) == len(new_id)
7084 mods.append((idx, new_id, changes))
7086 # now that we have passed all asserts above, we can apply the mods
7087 # in a single run (to avoid partial changes)
7088 for idx, new_id, changes in mods:
7089 disk = instance.disks[idx]
7090 if new_id is not None:
7091 assert disk.dev_type == constants.LD_DRBD8
7092 disk.logical_id = new_id
7094 disk.Update(size=changes.get(constants.IDISK_SIZE, None),
7095 mode=changes.get(constants.IDISK_MODE, None))
7097 # change primary node, if needed
7099 instance.primary_node = self.op.nodes[0]
7100 self.LogWarning("Changing the instance's nodes, you will have to"
7101 " remove any disks left on the older nodes manually")
7104 self.cfg.Update(instance, feedback_fn)
7106 _CreateDisks(self, instance, to_skip=to_skip)
7109 class LUInstanceRename(LogicalUnit):
7110 """Rename an instance.
7113 HPATH = "instance-rename"
7114 HTYPE = constants.HTYPE_INSTANCE
7116 def CheckArguments(self):
7120 if self.op.ip_check and not self.op.name_check:
7121 # TODO: make the ip check more flexible and not depend on the name check
7122 raise errors.OpPrereqError("IP address check requires a name check",
7125 def BuildHooksEnv(self):
7128 This runs on master, primary and secondary nodes of the instance.
7131 env = _BuildInstanceHookEnvByObject(self, self.instance)
7132 env["INSTANCE_NEW_NAME"] = self.op.new_name
7135 def BuildHooksNodes(self):
7136 """Build hooks nodes.
7139 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7142 def CheckPrereq(self):
7143 """Check prerequisites.
7145 This checks that the instance is in the cluster and is not running.
7148 self.op.instance_name = _ExpandInstanceName(self.cfg,
7149 self.op.instance_name)
7150 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7151 assert instance is not None
7152 _CheckNodeOnline(self, instance.primary_node)
7153 _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
7154 msg="cannot rename")
7155 self.instance = instance
7157 new_name = self.op.new_name
7158 if self.op.name_check:
7159 hostname = netutils.GetHostname(name=new_name)
7160 if hostname.name != new_name:
7161 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
7163 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
7164 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
7165 " same as given hostname '%s'") %
7166 (hostname.name, self.op.new_name),
7168 new_name = self.op.new_name = hostname.name
7169 if (self.op.ip_check and
7170 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
7171 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7172 (hostname.ip, new_name),
7173 errors.ECODE_NOTUNIQUE)
7175 instance_list = self.cfg.GetInstanceList()
7176 if new_name in instance_list and new_name != instance.name:
7177 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7178 new_name, errors.ECODE_EXISTS)
7180 def Exec(self, feedback_fn):
7181 """Rename the instance.
7184 inst = self.instance
7185 old_name = inst.name
7187 rename_file_storage = False
7188 if (inst.disk_template in constants.DTS_FILEBASED and
7189 self.op.new_name != inst.name):
7190 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
7191 rename_file_storage = True
7193 self.cfg.RenameInstance(inst.name, self.op.new_name)
7194 # Change the instance lock. This is definitely safe while we hold the BGL.
7195 # Otherwise the new lock would have to be added in acquired mode.
7197 self.glm.remove(locking.LEVEL_INSTANCE, old_name)
7198 self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
7200 # re-read the instance from the configuration after rename
7201 inst = self.cfg.GetInstanceInfo(self.op.new_name)
7203 if rename_file_storage:
7204 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
7205 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
7206 old_file_storage_dir,
7207 new_file_storage_dir)
7208 result.Raise("Could not rename on node %s directory '%s' to '%s'"
7209 " (but the instance has been renamed in Ganeti)" %
7210 (inst.primary_node, old_file_storage_dir,
7211 new_file_storage_dir))
7213 _StartInstanceDisks(self, inst, None)
7215 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
7216 old_name, self.op.debug_level)
7217 msg = result.fail_msg
7219 msg = ("Could not run OS rename script for instance %s on node %s"
7220 " (but the instance has been renamed in Ganeti): %s" %
7221 (inst.name, inst.primary_node, msg))
7222 self.proc.LogWarning(msg)
7224 _ShutdownInstanceDisks(self, inst)
7229 class LUInstanceRemove(LogicalUnit):
7230 """Remove an instance.
7233 HPATH = "instance-remove"
7234 HTYPE = constants.HTYPE_INSTANCE
7237 def ExpandNames(self):
7238 self._ExpandAndLockInstance()
7239 self.needed_locks[locking.LEVEL_NODE] = []
7240 self.needed_locks[locking.LEVEL_NODE_RES] = []
7241 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7243 def DeclareLocks(self, level):
7244 if level == locking.LEVEL_NODE:
7245 self._LockInstancesNodes()
7246 elif level == locking.LEVEL_NODE_RES:
7248 self.needed_locks[locking.LEVEL_NODE_RES] = \
7249 self.needed_locks[locking.LEVEL_NODE][:]
7251 def BuildHooksEnv(self):
7254 This runs on master, primary and secondary nodes of the instance.
7257 env = _BuildInstanceHookEnvByObject(self, self.instance)
7258 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
7261 def BuildHooksNodes(self):
7262 """Build hooks nodes.
7265 nl = [self.cfg.GetMasterNode()]
7266 nl_post = list(self.instance.all_nodes) + nl
7267 return (nl, nl_post)
7269 def CheckPrereq(self):
7270 """Check prerequisites.
7272 This checks that the instance is in the cluster.
7275 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7276 assert self.instance is not None, \
7277 "Cannot retrieve locked instance %s" % self.op.instance_name
7279 def Exec(self, feedback_fn):
7280 """Remove the instance.
7283 instance = self.instance
7284 logging.info("Shutting down instance %s on node %s",
7285 instance.name, instance.primary_node)
7287 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
7288 self.op.shutdown_timeout)
7289 msg = result.fail_msg
7291 if self.op.ignore_failures:
7292 feedback_fn("Warning: can't shutdown instance: %s" % msg)
7294 raise errors.OpExecError("Could not shutdown instance %s on"
7296 (instance.name, instance.primary_node, msg))
7298 assert (self.owned_locks(locking.LEVEL_NODE) ==
7299 self.owned_locks(locking.LEVEL_NODE_RES))
7300 assert not (set(instance.all_nodes) -
7301 self.owned_locks(locking.LEVEL_NODE)), \
7302 "Not owning correct locks"
7304 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
7307 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
7308 """Utility function to remove an instance.
7311 logging.info("Removing block devices for instance %s", instance.name)
7313 if not _RemoveDisks(lu, instance):
7314 if not ignore_failures:
7315 raise errors.OpExecError("Can't remove instance's disks")
7316 feedback_fn("Warning: can't remove instance's disks")
7318 logging.info("Removing instance %s out of cluster config", instance.name)
7320 lu.cfg.RemoveInstance(instance.name)
7322 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
7323 "Instance lock removal conflict"
7325 # Remove lock for the instance
7326 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
7329 class LUInstanceQuery(NoHooksLU):
7330 """Logical unit for querying instances.
7333 # pylint: disable=W0142
7336 def CheckArguments(self):
7337 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
7338 self.op.output_fields, self.op.use_locking)
7340 def ExpandNames(self):
7341 self.iq.ExpandNames(self)
7343 def DeclareLocks(self, level):
7344 self.iq.DeclareLocks(self, level)
7346 def Exec(self, feedback_fn):
7347 return self.iq.OldStyleQuery(self)
7350 class LUInstanceFailover(LogicalUnit):
7351 """Failover an instance.
7354 HPATH = "instance-failover"
7355 HTYPE = constants.HTYPE_INSTANCE
7358 def CheckArguments(self):
7359 """Check the arguments.
7362 self.iallocator = getattr(self.op, "iallocator", None)
7363 self.target_node = getattr(self.op, "target_node", None)
7365 def ExpandNames(self):
7366 self._ExpandAndLockInstance()
7368 if self.op.target_node is not None:
7369 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7371 self.needed_locks[locking.LEVEL_NODE] = []
7372 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7374 self.needed_locks[locking.LEVEL_NODE_RES] = []
7375 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
7377 ignore_consistency = self.op.ignore_consistency
7378 shutdown_timeout = self.op.shutdown_timeout
7379 self._migrater = TLMigrateInstance(self, self.op.instance_name,
7382 ignore_consistency=ignore_consistency,
7383 shutdown_timeout=shutdown_timeout,
7384 ignore_ipolicy=self.op.ignore_ipolicy)
7385 self.tasklets = [self._migrater]
7387 def DeclareLocks(self, level):
7388 if level == locking.LEVEL_NODE:
7389 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
7390 if instance.disk_template in constants.DTS_EXT_MIRROR:
7391 if self.op.target_node is None:
7392 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7394 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
7395 self.op.target_node]
7396 del self.recalculate_locks[locking.LEVEL_NODE]
7398 self._LockInstancesNodes()
7399 elif level == locking.LEVEL_NODE_RES:
7401 self.needed_locks[locking.LEVEL_NODE_RES] = \
7402 self.needed_locks[locking.LEVEL_NODE][:]
7404 def BuildHooksEnv(self):
7407 This runs on master, primary and secondary nodes of the instance.
7410 instance = self._migrater.instance
7411 source_node = instance.primary_node
7412 target_node = self.op.target_node
7414 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
7415 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
7416 "OLD_PRIMARY": source_node,
7417 "NEW_PRIMARY": target_node,
7420 if instance.disk_template in constants.DTS_INT_MIRROR:
7421 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
7422 env["NEW_SECONDARY"] = source_node
7424 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
7426 env.update(_BuildInstanceHookEnvByObject(self, instance))
7430 def BuildHooksNodes(self):
7431 """Build hooks nodes.
7434 instance = self._migrater.instance
7435 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
7436 return (nl, nl + [instance.primary_node])
7439 class LUInstanceMigrate(LogicalUnit):
7440 """Migrate an instance.
7442 This is migration without shutting down, compared to the failover,
7443 which is done with shutdown.
7446 HPATH = "instance-migrate"
7447 HTYPE = constants.HTYPE_INSTANCE
7450 def ExpandNames(self):
7451 self._ExpandAndLockInstance()
7453 if self.op.target_node is not None:
7454 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7456 self.needed_locks[locking.LEVEL_NODE] = []
7457 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7459 self.needed_locks[locking.LEVEL_NODE] = []
7460 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7463 TLMigrateInstance(self, self.op.instance_name,
7464 cleanup=self.op.cleanup,
7466 fallback=self.op.allow_failover,
7467 allow_runtime_changes=self.op.allow_runtime_changes,
7468 ignore_ipolicy=self.op.ignore_ipolicy)
7469 self.tasklets = [self._migrater]
7471 def DeclareLocks(self, level):
7472 if level == locking.LEVEL_NODE:
7473 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
7474 if instance.disk_template in constants.DTS_EXT_MIRROR:
7475 if self.op.target_node is None:
7476 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7478 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
7479 self.op.target_node]
7480 del self.recalculate_locks[locking.LEVEL_NODE]
7482 self._LockInstancesNodes()
7483 elif level == locking.LEVEL_NODE_RES:
7485 self.needed_locks[locking.LEVEL_NODE_RES] = \
7486 self.needed_locks[locking.LEVEL_NODE][:]
7488 def BuildHooksEnv(self):
7491 This runs on master, primary and secondary nodes of the instance.
7494 instance = self._migrater.instance
7495 source_node = instance.primary_node
7496 target_node = self.op.target_node
7497 env = _BuildInstanceHookEnvByObject(self, instance)
7499 "MIGRATE_LIVE": self._migrater.live,
7500 "MIGRATE_CLEANUP": self.op.cleanup,
7501 "OLD_PRIMARY": source_node,
7502 "NEW_PRIMARY": target_node,
7503 "ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes,
7506 if instance.disk_template in constants.DTS_INT_MIRROR:
7507 env["OLD_SECONDARY"] = target_node
7508 env["NEW_SECONDARY"] = source_node
7510 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
7514 def BuildHooksNodes(self):
7515 """Build hooks nodes.
7518 instance = self._migrater.instance
7519 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
7520 return (nl, nl + [instance.primary_node])
7523 class LUInstanceMove(LogicalUnit):
7524 """Move an instance by data-copying.
7527 HPATH = "instance-move"
7528 HTYPE = constants.HTYPE_INSTANCE
7531 def ExpandNames(self):
7532 self._ExpandAndLockInstance()
7533 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7534 self.op.target_node = target_node
7535 self.needed_locks[locking.LEVEL_NODE] = [target_node]
7536 self.needed_locks[locking.LEVEL_NODE_RES] = []
7537 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7539 def DeclareLocks(self, level):
7540 if level == locking.LEVEL_NODE:
7541 self._LockInstancesNodes(primary_only=True)
7542 elif level == locking.LEVEL_NODE_RES:
7544 self.needed_locks[locking.LEVEL_NODE_RES] = \
7545 self.needed_locks[locking.LEVEL_NODE][:]
7547 def BuildHooksEnv(self):
7550 This runs on master, primary and secondary nodes of the instance.
7554 "TARGET_NODE": self.op.target_node,
7555 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
7557 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7560 def BuildHooksNodes(self):
7561 """Build hooks nodes.
7565 self.cfg.GetMasterNode(),
7566 self.instance.primary_node,
7567 self.op.target_node,
7571 def CheckPrereq(self):
7572 """Check prerequisites.
7574 This checks that the instance is in the cluster.
7577 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7578 assert self.instance is not None, \
7579 "Cannot retrieve locked instance %s" % self.op.instance_name
7581 node = self.cfg.GetNodeInfo(self.op.target_node)
7582 assert node is not None, \
7583 "Cannot retrieve locked node %s" % self.op.target_node
7585 self.target_node = target_node = node.name
7587 if target_node == instance.primary_node:
7588 raise errors.OpPrereqError("Instance %s is already on the node %s" %
7589 (instance.name, target_node),
7592 bep = self.cfg.GetClusterInfo().FillBE(instance)
7594 for idx, dsk in enumerate(instance.disks):
7595 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
7596 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
7597 " cannot copy" % idx, errors.ECODE_STATE)
7599 _CheckNodeOnline(self, target_node)
7600 _CheckNodeNotDrained(self, target_node)
7601 _CheckNodeVmCapable(self, target_node)
7602 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(),
7603 self.cfg.GetNodeGroup(node.group))
7604 _CheckTargetNodeIPolicy(self, ipolicy, instance, node,
7605 ignore=self.op.ignore_ipolicy)
7607 if instance.admin_state == constants.ADMINST_UP:
7608 # check memory requirements on the secondary node
7609 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
7610 instance.name, bep[constants.BE_MAXMEM],
7611 instance.hypervisor)
7613 self.LogInfo("Not checking memory on the secondary node as"
7614 " instance will not be started")
7616 # check bridge existance
7617 _CheckInstanceBridgesExist(self, instance, node=target_node)
7619 def Exec(self, feedback_fn):
7620 """Move an instance.
7622 The move is done by shutting it down on its present node, copying
7623 the data over (slow) and starting it on the new node.
7626 instance = self.instance
7628 source_node = instance.primary_node
7629 target_node = self.target_node
7631 self.LogInfo("Shutting down instance %s on source node %s",
7632 instance.name, source_node)
7634 assert (self.owned_locks(locking.LEVEL_NODE) ==
7635 self.owned_locks(locking.LEVEL_NODE_RES))
7637 result = self.rpc.call_instance_shutdown(source_node, instance,
7638 self.op.shutdown_timeout)
7639 msg = result.fail_msg
7641 if self.op.ignore_consistency:
7642 self.proc.LogWarning("Could not shutdown instance %s on node %s."
7643 " Proceeding anyway. Please make sure node"
7644 " %s is down. Error details: %s",
7645 instance.name, source_node, source_node, msg)
7647 raise errors.OpExecError("Could not shutdown instance %s on"
7649 (instance.name, source_node, msg))
7651 # create the target disks
7653 _CreateDisks(self, instance, target_node=target_node)
7654 except errors.OpExecError:
7655 self.LogWarning("Device creation failed, reverting...")
7657 _RemoveDisks(self, instance, target_node=target_node)
7659 self.cfg.ReleaseDRBDMinors(instance.name)
7662 cluster_name = self.cfg.GetClusterInfo().cluster_name
7665 # activate, get path, copy the data over
7666 for idx, disk in enumerate(instance.disks):
7667 self.LogInfo("Copying data for disk %d", idx)
7668 result = self.rpc.call_blockdev_assemble(target_node, disk,
7669 instance.name, True, idx)
7671 self.LogWarning("Can't assemble newly created disk %d: %s",
7672 idx, result.fail_msg)
7673 errs.append(result.fail_msg)
7675 dev_path = result.payload
7676 result = self.rpc.call_blockdev_export(source_node, disk,
7677 target_node, dev_path,
7680 self.LogWarning("Can't copy data over for disk %d: %s",
7681 idx, result.fail_msg)
7682 errs.append(result.fail_msg)
7686 self.LogWarning("Some disks failed to copy, aborting")
7688 _RemoveDisks(self, instance, target_node=target_node)
7690 self.cfg.ReleaseDRBDMinors(instance.name)
7691 raise errors.OpExecError("Errors during disk copy: %s" %
7694 instance.primary_node = target_node
7695 self.cfg.Update(instance, feedback_fn)
7697 self.LogInfo("Removing the disks on the original node")
7698 _RemoveDisks(self, instance, target_node=source_node)
7700 # Only start the instance if it's marked as up
7701 if instance.admin_state == constants.ADMINST_UP:
7702 self.LogInfo("Starting instance %s on node %s",
7703 instance.name, target_node)
7705 disks_ok, _ = _AssembleInstanceDisks(self, instance,
7706 ignore_secondaries=True)
7708 _ShutdownInstanceDisks(self, instance)
7709 raise errors.OpExecError("Can't activate the instance's disks")
7711 result = self.rpc.call_instance_start(target_node,
7712 (instance, None, None), False)
7713 msg = result.fail_msg
7715 _ShutdownInstanceDisks(self, instance)
7716 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
7717 (instance.name, target_node, msg))
7720 class LUNodeMigrate(LogicalUnit):
7721 """Migrate all instances from a node.
7724 HPATH = "node-migrate"
7725 HTYPE = constants.HTYPE_NODE
7728 def CheckArguments(self):
7731 def ExpandNames(self):
7732 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
7734 self.share_locks = _ShareAll()
7735 self.needed_locks = {
7736 locking.LEVEL_NODE: [self.op.node_name],
7739 def BuildHooksEnv(self):
7742 This runs on the master, the primary and all the secondaries.
7746 "NODE_NAME": self.op.node_name,
7747 "ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes,
7750 def BuildHooksNodes(self):
7751 """Build hooks nodes.
7754 nl = [self.cfg.GetMasterNode()]
7757 def CheckPrereq(self):
7760 def Exec(self, feedback_fn):
7761 # Prepare jobs for migration instances
7762 allow_runtime_changes = self.op.allow_runtime_changes
7764 [opcodes.OpInstanceMigrate(instance_name=inst.name,
7767 iallocator=self.op.iallocator,
7768 target_node=self.op.target_node,
7769 allow_runtime_changes=allow_runtime_changes,
7770 ignore_ipolicy=self.op.ignore_ipolicy)]
7771 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name)
7774 # TODO: Run iallocator in this opcode and pass correct placement options to
7775 # OpInstanceMigrate. Since other jobs can modify the cluster between
7776 # running the iallocator and the actual migration, a good consistency model
7777 # will have to be found.
7779 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
7780 frozenset([self.op.node_name]))
7782 return ResultWithJobs(jobs)
7785 class TLMigrateInstance(Tasklet):
7786 """Tasklet class for instance migration.
7789 @ivar live: whether the migration will be done live or non-live;
7790 this variable is initalized only after CheckPrereq has run
7791 @type cleanup: boolean
7792 @ivar cleanup: Wheater we cleanup from a failed migration
7793 @type iallocator: string
7794 @ivar iallocator: The iallocator used to determine target_node
7795 @type target_node: string
7796 @ivar target_node: If given, the target_node to reallocate the instance to
7797 @type failover: boolean
7798 @ivar failover: Whether operation results in failover or migration
7799 @type fallback: boolean
7800 @ivar fallback: Whether fallback to failover is allowed if migration not
7802 @type ignore_consistency: boolean
7803 @ivar ignore_consistency: Wheter we should ignore consistency between source
7805 @type shutdown_timeout: int
7806 @ivar shutdown_timeout: In case of failover timeout of the shutdown
7807 @type ignore_ipolicy: bool
7808 @ivar ignore_ipolicy: If true, we can ignore instance policy when migrating
7813 _MIGRATION_POLL_INTERVAL = 1 # seconds
7814 _MIGRATION_FEEDBACK_INTERVAL = 10 # seconds
7816 def __init__(self, lu, instance_name, cleanup=False,
7817 failover=False, fallback=False,
7818 ignore_consistency=False,
7819 allow_runtime_changes=True,
7820 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT,
7821 ignore_ipolicy=False):
7822 """Initializes this class.
7825 Tasklet.__init__(self, lu)
7828 self.instance_name = instance_name
7829 self.cleanup = cleanup
7830 self.live = False # will be overridden later
7831 self.failover = failover
7832 self.fallback = fallback
7833 self.ignore_consistency = ignore_consistency
7834 self.shutdown_timeout = shutdown_timeout
7835 self.ignore_ipolicy = ignore_ipolicy
7836 self.allow_runtime_changes = allow_runtime_changes
7838 def CheckPrereq(self):
7839 """Check prerequisites.
7841 This checks that the instance is in the cluster.
7844 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
7845 instance = self.cfg.GetInstanceInfo(instance_name)
7846 assert instance is not None
7847 self.instance = instance
7848 cluster = self.cfg.GetClusterInfo()
7850 if (not self.cleanup and
7851 not instance.admin_state == constants.ADMINST_UP and
7852 not self.failover and self.fallback):
7853 self.lu.LogInfo("Instance is marked down or offline, fallback allowed,"
7854 " switching to failover")
7855 self.failover = True
7857 if instance.disk_template not in constants.DTS_MIRRORED:
7862 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
7863 " %s" % (instance.disk_template, text),
7866 if instance.disk_template in constants.DTS_EXT_MIRROR:
7867 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
7869 if self.lu.op.iallocator:
7870 self._RunAllocator()
7872 # We set set self.target_node as it is required by
7874 self.target_node = self.lu.op.target_node
7876 # Check that the target node is correct in terms of instance policy
7877 nodeinfo = self.cfg.GetNodeInfo(self.target_node)
7878 group_info = self.cfg.GetNodeGroup(nodeinfo.group)
7879 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
7880 _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
7881 ignore=self.ignore_ipolicy)
7883 # self.target_node is already populated, either directly or by the
7885 target_node = self.target_node
7886 if self.target_node == instance.primary_node:
7887 raise errors.OpPrereqError("Cannot migrate instance %s"
7888 " to its primary (%s)" %
7889 (instance.name, instance.primary_node))
7891 if len(self.lu.tasklets) == 1:
7892 # It is safe to release locks only when we're the only tasklet
7894 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
7895 keep=[instance.primary_node, self.target_node])
7898 secondary_nodes = instance.secondary_nodes
7899 if not secondary_nodes:
7900 raise errors.ConfigurationError("No secondary node but using"
7901 " %s disk template" %
7902 instance.disk_template)
7903 target_node = secondary_nodes[0]
7904 if self.lu.op.iallocator or (self.lu.op.target_node and
7905 self.lu.op.target_node != target_node):
7907 text = "failed over"
7910 raise errors.OpPrereqError("Instances with disk template %s cannot"
7911 " be %s to arbitrary nodes"
7912 " (neither an iallocator nor a target"
7913 " node can be passed)" %
7914 (instance.disk_template, text),
7916 nodeinfo = self.cfg.GetNodeInfo(target_node)
7917 group_info = self.cfg.GetNodeGroup(nodeinfo.group)
7918 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
7919 _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
7920 ignore=self.ignore_ipolicy)
7922 i_be = cluster.FillBE(instance)
7924 # check memory requirements on the secondary node
7925 if (not self.cleanup and
7926 (not self.failover or instance.admin_state == constants.ADMINST_UP)):
7927 self.tgt_free_mem = _CheckNodeFreeMemory(self.lu, target_node,
7928 "migrating instance %s" %
7930 i_be[constants.BE_MINMEM],
7931 instance.hypervisor)
7933 self.lu.LogInfo("Not checking memory on the secondary node as"
7934 " instance will not be started")
7936 # check if failover must be forced instead of migration
7937 if (not self.cleanup and not self.failover and
7938 i_be[constants.BE_ALWAYS_FAILOVER]):
7940 self.lu.LogInfo("Instance configured to always failover; fallback"
7942 self.failover = True
7944 raise errors.OpPrereqError("This instance has been configured to"
7945 " always failover, please allow failover",
7948 # check bridge existance
7949 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
7951 if not self.cleanup:
7952 _CheckNodeNotDrained(self.lu, target_node)
7953 if not self.failover:
7954 result = self.rpc.call_instance_migratable(instance.primary_node,
7956 if result.fail_msg and self.fallback:
7957 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
7959 self.failover = True
7961 result.Raise("Can't migrate, please use failover",
7962 prereq=True, ecode=errors.ECODE_STATE)
7964 assert not (self.failover and self.cleanup)
7966 if not self.failover:
7967 if self.lu.op.live is not None and self.lu.op.mode is not None:
7968 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
7969 " parameters are accepted",
7971 if self.lu.op.live is not None:
7973 self.lu.op.mode = constants.HT_MIGRATION_LIVE
7975 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
7976 # reset the 'live' parameter to None so that repeated
7977 # invocations of CheckPrereq do not raise an exception
7978 self.lu.op.live = None
7979 elif self.lu.op.mode is None:
7980 # read the default value from the hypervisor
7981 i_hv = cluster.FillHV(self.instance, skip_globals=False)
7982 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
7984 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
7986 # Failover is never live
7989 if not (self.failover or self.cleanup):
7990 remote_info = self.rpc.call_instance_info(instance.primary_node,
7992 instance.hypervisor)
7993 remote_info.Raise("Error checking instance on node %s" %
7994 instance.primary_node)
7995 instance_running = bool(remote_info.payload)
7996 if instance_running:
7997 self.current_mem = int(remote_info.payload["memory"])
7999 def _RunAllocator(self):
8000 """Run the allocator based on input opcode.
8003 # FIXME: add a self.ignore_ipolicy option
8004 ial = IAllocator(self.cfg, self.rpc,
8005 mode=constants.IALLOCATOR_MODE_RELOC,
8006 name=self.instance_name,
8007 # TODO See why hail breaks with a single node below
8008 relocate_from=[self.instance.primary_node,
8009 self.instance.primary_node],
8012 ial.Run(self.lu.op.iallocator)
8015 raise errors.OpPrereqError("Can't compute nodes using"
8016 " iallocator '%s': %s" %
8017 (self.lu.op.iallocator, ial.info),
8019 if len(ial.result) != ial.required_nodes:
8020 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8021 " of nodes (%s), required %s" %
8022 (self.lu.op.iallocator, len(ial.result),
8023 ial.required_nodes), errors.ECODE_FAULT)
8024 self.target_node = ial.result[0]
8025 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
8026 self.instance_name, self.lu.op.iallocator,
8027 utils.CommaJoin(ial.result))
8029 def _WaitUntilSync(self):
8030 """Poll with custom rpc for disk sync.
8032 This uses our own step-based rpc call.
8035 self.feedback_fn("* wait until resync is done")
8039 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
8041 self.instance.disks)
8043 for node, nres in result.items():
8044 nres.Raise("Cannot resync disks on node %s" % node)
8045 node_done, node_percent = nres.payload
8046 all_done = all_done and node_done
8047 if node_percent is not None:
8048 min_percent = min(min_percent, node_percent)
8050 if min_percent < 100:
8051 self.feedback_fn(" - progress: %.1f%%" % min_percent)
8054 def _EnsureSecondary(self, node):
8055 """Demote a node to secondary.
8058 self.feedback_fn("* switching node %s to secondary mode" % node)
8060 for dev in self.instance.disks:
8061 self.cfg.SetDiskID(dev, node)
8063 result = self.rpc.call_blockdev_close(node, self.instance.name,
8064 self.instance.disks)
8065 result.Raise("Cannot change disk to secondary on node %s" % node)
8067 def _GoStandalone(self):
8068 """Disconnect from the network.
8071 self.feedback_fn("* changing into standalone mode")
8072 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
8073 self.instance.disks)
8074 for node, nres in result.items():
8075 nres.Raise("Cannot disconnect disks node %s" % node)
8077 def _GoReconnect(self, multimaster):
8078 """Reconnect to the network.
8084 msg = "single-master"
8085 self.feedback_fn("* changing disks into %s mode" % msg)
8086 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
8087 self.instance.disks,
8088 self.instance.name, multimaster)
8089 for node, nres in result.items():
8090 nres.Raise("Cannot change disks config on node %s" % node)
8092 def _ExecCleanup(self):
8093 """Try to cleanup after a failed migration.
8095 The cleanup is done by:
8096 - check that the instance is running only on one node
8097 (and update the config if needed)
8098 - change disks on its secondary node to secondary
8099 - wait until disks are fully synchronized
8100 - disconnect from the network
8101 - change disks into single-master mode
8102 - wait again until disks are fully synchronized
8105 instance = self.instance
8106 target_node = self.target_node
8107 source_node = self.source_node
8109 # check running on only one node
8110 self.feedback_fn("* checking where the instance actually runs"
8111 " (if this hangs, the hypervisor might be in"
8113 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
8114 for node, result in ins_l.items():
8115 result.Raise("Can't contact node %s" % node)
8117 runningon_source = instance.name in ins_l[source_node].payload
8118 runningon_target = instance.name in ins_l[target_node].payload
8120 if runningon_source and runningon_target:
8121 raise errors.OpExecError("Instance seems to be running on two nodes,"
8122 " or the hypervisor is confused; you will have"
8123 " to ensure manually that it runs only on one"
8124 " and restart this operation")
8126 if not (runningon_source or runningon_target):
8127 raise errors.OpExecError("Instance does not seem to be running at all;"
8128 " in this case it's safer to repair by"
8129 " running 'gnt-instance stop' to ensure disk"
8130 " shutdown, and then restarting it")
8132 if runningon_target:
8133 # the migration has actually succeeded, we need to update the config
8134 self.feedback_fn("* instance running on secondary node (%s),"
8135 " updating config" % target_node)
8136 instance.primary_node = target_node
8137 self.cfg.Update(instance, self.feedback_fn)
8138 demoted_node = source_node
8140 self.feedback_fn("* instance confirmed to be running on its"
8141 " primary node (%s)" % source_node)
8142 demoted_node = target_node
8144 if instance.disk_template in constants.DTS_INT_MIRROR:
8145 self._EnsureSecondary(demoted_node)
8147 self._WaitUntilSync()
8148 except errors.OpExecError:
8149 # we ignore here errors, since if the device is standalone, it
8150 # won't be able to sync
8152 self._GoStandalone()
8153 self._GoReconnect(False)
8154 self._WaitUntilSync()
8156 self.feedback_fn("* done")
8158 def _RevertDiskStatus(self):
8159 """Try to revert the disk status after a failed migration.
8162 target_node = self.target_node
8163 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
8167 self._EnsureSecondary(target_node)
8168 self._GoStandalone()
8169 self._GoReconnect(False)
8170 self._WaitUntilSync()
8171 except errors.OpExecError, err:
8172 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
8173 " please try to recover the instance manually;"
8174 " error '%s'" % str(err))
8176 def _AbortMigration(self):
8177 """Call the hypervisor code to abort a started migration.
8180 instance = self.instance
8181 target_node = self.target_node
8182 source_node = self.source_node
8183 migration_info = self.migration_info
8185 abort_result = self.rpc.call_instance_finalize_migration_dst(target_node,
8189 abort_msg = abort_result.fail_msg
8191 logging.error("Aborting migration failed on target node %s: %s",
8192 target_node, abort_msg)
8193 # Don't raise an exception here, as we stil have to try to revert the
8194 # disk status, even if this step failed.
8196 abort_result = self.rpc.call_instance_finalize_migration_src(source_node,
8197 instance, False, self.live)
8198 abort_msg = abort_result.fail_msg
8200 logging.error("Aborting migration failed on source node %s: %s",
8201 source_node, abort_msg)
8203 def _ExecMigration(self):
8204 """Migrate an instance.
8206 The migrate is done by:
8207 - change the disks into dual-master mode
8208 - wait until disks are fully synchronized again
8209 - migrate the instance
8210 - change disks on the new secondary node (the old primary) to secondary
8211 - wait until disks are fully synchronized
8212 - change disks into single-master mode
8215 instance = self.instance
8216 target_node = self.target_node
8217 source_node = self.source_node
8219 # Check for hypervisor version mismatch and warn the user.
8220 nodeinfo = self.rpc.call_node_info([source_node, target_node],
8221 None, [self.instance.hypervisor])
8222 for ninfo in nodeinfo.values():
8223 ninfo.Raise("Unable to retrieve node information from node '%s'" %
8225 (_, _, (src_info, )) = nodeinfo[source_node].payload
8226 (_, _, (dst_info, )) = nodeinfo[target_node].payload
8228 if ((constants.HV_NODEINFO_KEY_VERSION in src_info) and
8229 (constants.HV_NODEINFO_KEY_VERSION in dst_info)):
8230 src_version = src_info[constants.HV_NODEINFO_KEY_VERSION]
8231 dst_version = dst_info[constants.HV_NODEINFO_KEY_VERSION]
8232 if src_version != dst_version:
8233 self.feedback_fn("* warning: hypervisor version mismatch between"
8234 " source (%s) and target (%s) node" %
8235 (src_version, dst_version))
8237 self.feedback_fn("* checking disk consistency between source and target")
8238 for (idx, dev) in enumerate(instance.disks):
8239 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
8240 raise errors.OpExecError("Disk %s is degraded or not fully"
8241 " synchronized on target node,"
8242 " aborting migration" % idx)
8244 if self.current_mem > self.tgt_free_mem:
8245 if not self.allow_runtime_changes:
8246 raise errors.OpExecError("Memory ballooning not allowed and not enough"
8247 " free memory to fit instance %s on target"
8248 " node %s (have %dMB, need %dMB)" %
8249 (instance.name, target_node,
8250 self.tgt_free_mem, self.current_mem))
8251 self.feedback_fn("* setting instance memory to %s" % self.tgt_free_mem)
8252 rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
8255 rpcres.Raise("Cannot modify instance runtime memory")
8257 # First get the migration information from the remote node
8258 result = self.rpc.call_migration_info(source_node, instance)
8259 msg = result.fail_msg
8261 log_err = ("Failed fetching source migration information from %s: %s" %
8263 logging.error(log_err)
8264 raise errors.OpExecError(log_err)
8266 self.migration_info = migration_info = result.payload
8268 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
8269 # Then switch the disks to master/master mode
8270 self._EnsureSecondary(target_node)
8271 self._GoStandalone()
8272 self._GoReconnect(True)
8273 self._WaitUntilSync()
8275 self.feedback_fn("* preparing %s to accept the instance" % target_node)
8276 result = self.rpc.call_accept_instance(target_node,
8279 self.nodes_ip[target_node])
8281 msg = result.fail_msg
8283 logging.error("Instance pre-migration failed, trying to revert"
8284 " disk status: %s", msg)
8285 self.feedback_fn("Pre-migration failed, aborting")
8286 self._AbortMigration()
8287 self._RevertDiskStatus()
8288 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
8289 (instance.name, msg))
8291 self.feedback_fn("* migrating instance to %s" % target_node)
8292 result = self.rpc.call_instance_migrate(source_node, instance,
8293 self.nodes_ip[target_node],
8295 msg = result.fail_msg
8297 logging.error("Instance migration failed, trying to revert"
8298 " disk status: %s", msg)
8299 self.feedback_fn("Migration failed, aborting")
8300 self._AbortMigration()
8301 self._RevertDiskStatus()
8302 raise errors.OpExecError("Could not migrate instance %s: %s" %
8303 (instance.name, msg))
8305 self.feedback_fn("* starting memory transfer")
8306 last_feedback = time.time()
8308 result = self.rpc.call_instance_get_migration_status(source_node,
8310 msg = result.fail_msg
8311 ms = result.payload # MigrationStatus instance
8312 if msg or (ms.status in constants.HV_MIGRATION_FAILED_STATUSES):
8313 logging.error("Instance migration failed, trying to revert"
8314 " disk status: %s", msg)
8315 self.feedback_fn("Migration failed, aborting")
8316 self._AbortMigration()
8317 self._RevertDiskStatus()
8318 raise errors.OpExecError("Could not migrate instance %s: %s" %
8319 (instance.name, msg))
8321 if result.payload.status != constants.HV_MIGRATION_ACTIVE:
8322 self.feedback_fn("* memory transfer complete")
8325 if (utils.TimeoutExpired(last_feedback,
8326 self._MIGRATION_FEEDBACK_INTERVAL) and
8327 ms.transferred_ram is not None):
8328 mem_progress = 100 * float(ms.transferred_ram) / float(ms.total_ram)
8329 self.feedback_fn("* memory transfer progress: %.2f %%" % mem_progress)
8330 last_feedback = time.time()
8332 time.sleep(self._MIGRATION_POLL_INTERVAL)
8334 result = self.rpc.call_instance_finalize_migration_src(source_node,
8338 msg = result.fail_msg
8340 logging.error("Instance migration succeeded, but finalization failed"
8341 " on the source node: %s", msg)
8342 raise errors.OpExecError("Could not finalize instance migration: %s" %
8345 instance.primary_node = target_node
8347 # distribute new instance config to the other nodes
8348 self.cfg.Update(instance, self.feedback_fn)
8350 result = self.rpc.call_instance_finalize_migration_dst(target_node,
8354 msg = result.fail_msg
8356 logging.error("Instance migration succeeded, but finalization failed"
8357 " on the target node: %s", msg)
8358 raise errors.OpExecError("Could not finalize instance migration: %s" %
8361 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
8362 self._EnsureSecondary(source_node)
8363 self._WaitUntilSync()
8364 self._GoStandalone()
8365 self._GoReconnect(False)
8366 self._WaitUntilSync()
8368 # If the instance's disk template is `rbd' and there was a successful
8369 # migration, unmap the device from the source node.
8370 if self.instance.disk_template == constants.DT_RBD:
8371 disks = _ExpandCheckDisks(instance, instance.disks)
8372 self.feedback_fn("* unmapping instance's disks from %s" % source_node)
8374 result = self.rpc.call_blockdev_shutdown(source_node, disk)
8375 msg = result.fail_msg
8377 logging.error("Migration was successful, but couldn't unmap the"
8378 " block device %s on source node %s: %s",
8379 disk.iv_name, source_node, msg)
8380 logging.error("You need to unmap the device %s manually on %s",
8381 disk.iv_name, source_node)
8383 self.feedback_fn("* done")
8385 def _ExecFailover(self):
8386 """Failover an instance.
8388 The failover is done by shutting it down on its present node and
8389 starting it on the secondary.
8392 instance = self.instance
8393 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
8395 source_node = instance.primary_node
8396 target_node = self.target_node
8398 if instance.admin_state == constants.ADMINST_UP:
8399 self.feedback_fn("* checking disk consistency between source and target")
8400 for (idx, dev) in enumerate(instance.disks):
8401 # for drbd, these are drbd over lvm
8402 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
8403 if primary_node.offline:
8404 self.feedback_fn("Node %s is offline, ignoring degraded disk %s on"
8406 (primary_node.name, idx, target_node))
8407 elif not self.ignore_consistency:
8408 raise errors.OpExecError("Disk %s is degraded on target node,"
8409 " aborting failover" % idx)
8411 self.feedback_fn("* not checking disk consistency as instance is not"
8414 self.feedback_fn("* shutting down instance on source node")
8415 logging.info("Shutting down instance %s on node %s",
8416 instance.name, source_node)
8418 result = self.rpc.call_instance_shutdown(source_node, instance,
8419 self.shutdown_timeout)
8420 msg = result.fail_msg
8422 if self.ignore_consistency or primary_node.offline:
8423 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
8424 " proceeding anyway; please make sure node"
8425 " %s is down; error details: %s",
8426 instance.name, source_node, source_node, msg)
8428 raise errors.OpExecError("Could not shutdown instance %s on"
8430 (instance.name, source_node, msg))
8432 self.feedback_fn("* deactivating the instance's disks on source node")
8433 if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True):
8434 raise errors.OpExecError("Can't shut down the instance's disks")
8436 instance.primary_node = target_node
8437 # distribute new instance config to the other nodes
8438 self.cfg.Update(instance, self.feedback_fn)
8440 # Only start the instance if it's marked as up
8441 if instance.admin_state == constants.ADMINST_UP:
8442 self.feedback_fn("* activating the instance's disks on target node %s" %
8444 logging.info("Starting instance %s on node %s",
8445 instance.name, target_node)
8447 disks_ok, _ = _AssembleInstanceDisks(self.lu, instance,
8448 ignore_secondaries=True)
8450 _ShutdownInstanceDisks(self.lu, instance)
8451 raise errors.OpExecError("Can't activate the instance's disks")
8453 self.feedback_fn("* starting the instance on the target node %s" %
8455 result = self.rpc.call_instance_start(target_node, (instance, None, None),
8457 msg = result.fail_msg
8459 _ShutdownInstanceDisks(self.lu, instance)
8460 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
8461 (instance.name, target_node, msg))
8463 def Exec(self, feedback_fn):
8464 """Perform the migration.
8467 self.feedback_fn = feedback_fn
8468 self.source_node = self.instance.primary_node
8470 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
8471 if self.instance.disk_template in constants.DTS_INT_MIRROR:
8472 self.target_node = self.instance.secondary_nodes[0]
8473 # Otherwise self.target_node has been populated either
8474 # directly, or through an iallocator.
8476 self.all_nodes = [self.source_node, self.target_node]
8477 self.nodes_ip = dict((name, node.secondary_ip) for (name, node)
8478 in self.cfg.GetMultiNodeInfo(self.all_nodes))
8481 feedback_fn("Failover instance %s" % self.instance.name)
8482 self._ExecFailover()
8484 feedback_fn("Migrating instance %s" % self.instance.name)
8487 return self._ExecCleanup()
8489 return self._ExecMigration()
8492 def _CreateBlockDev(lu, node, instance, device, force_create,
8494 """Create a tree of block devices on a given node.
8496 If this device type has to be created on secondaries, create it and
8499 If not, just recurse to children keeping the same 'force' value.
8501 @param lu: the lu on whose behalf we execute
8502 @param node: the node on which to create the device
8503 @type instance: L{objects.Instance}
8504 @param instance: the instance which owns the device
8505 @type device: L{objects.Disk}
8506 @param device: the device to create
8507 @type force_create: boolean
8508 @param force_create: whether to force creation of this device; this
8509 will be change to True whenever we find a device which has
8510 CreateOnSecondary() attribute
8511 @param info: the extra 'metadata' we should attach to the device
8512 (this will be represented as a LVM tag)
8513 @type force_open: boolean
8514 @param force_open: this parameter will be passes to the
8515 L{backend.BlockdevCreate} function where it specifies
8516 whether we run on primary or not, and it affects both
8517 the child assembly and the device own Open() execution
8520 if device.CreateOnSecondary():
8524 for child in device.children:
8525 _CreateBlockDev(lu, node, instance, child, force_create,
8528 if not force_create:
8531 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
8534 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
8535 """Create a single block device on a given node.
8537 This will not recurse over children of the device, so they must be
8540 @param lu: the lu on whose behalf we execute
8541 @param node: the node on which to create the device
8542 @type instance: L{objects.Instance}
8543 @param instance: the instance which owns the device
8544 @type device: L{objects.Disk}
8545 @param device: the device to create
8546 @param info: the extra 'metadata' we should attach to the device
8547 (this will be represented as a LVM tag)
8548 @type force_open: boolean
8549 @param force_open: this parameter will be passes to the
8550 L{backend.BlockdevCreate} function where it specifies
8551 whether we run on primary or not, and it affects both
8552 the child assembly and the device own Open() execution
8555 lu.cfg.SetDiskID(device, node)
8556 result = lu.rpc.call_blockdev_create(node, device, device.size,
8557 instance.name, force_open, info)
8558 result.Raise("Can't create block device %s on"
8559 " node %s for instance %s" % (device, node, instance.name))
8560 if device.physical_id is None:
8561 device.physical_id = result.payload
8564 def _GenerateUniqueNames(lu, exts):
8565 """Generate a suitable LV name.
8567 This will generate a logical volume name for the given instance.
8572 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
8573 results.append("%s%s" % (new_id, val))
8577 def _ComputeLDParams(disk_template, disk_params):
8578 """Computes Logical Disk parameters from Disk Template parameters.
8580 @type disk_template: string
8581 @param disk_template: disk template, one of L{constants.DISK_TEMPLATES}
8582 @type disk_params: dict
8583 @param disk_params: disk template parameters; dict(template_name -> parameters
8585 @return: a list of dicts, one for each node of the disk hierarchy. Each dict
8586 contains the LD parameters of the node. The tree is flattened in-order.
8589 if disk_template not in constants.DISK_TEMPLATES:
8590 raise errors.ProgrammerError("Unknown disk template %s" % disk_template)
8593 dt_params = disk_params[disk_template]
8594 if disk_template == constants.DT_DRBD8:
8596 constants.LDP_RESYNC_RATE: dt_params[constants.DRBD_RESYNC_RATE],
8597 constants.LDP_BARRIERS: dt_params[constants.DRBD_DISK_BARRIERS],
8598 constants.LDP_NO_META_FLUSH: dt_params[constants.DRBD_META_BARRIERS],
8599 constants.LDP_DEFAULT_METAVG: dt_params[constants.DRBD_DEFAULT_METAVG],
8600 constants.LDP_DISK_CUSTOM: dt_params[constants.DRBD_DISK_CUSTOM],
8601 constants.LDP_NET_CUSTOM: dt_params[constants.DRBD_NET_CUSTOM],
8602 constants.LDP_DYNAMIC_RESYNC: dt_params[constants.DRBD_DYNAMIC_RESYNC],
8603 constants.LDP_PLAN_AHEAD: dt_params[constants.DRBD_PLAN_AHEAD],
8604 constants.LDP_FILL_TARGET: dt_params[constants.DRBD_FILL_TARGET],
8605 constants.LDP_DELAY_TARGET: dt_params[constants.DRBD_DELAY_TARGET],
8606 constants.LDP_MAX_RATE: dt_params[constants.DRBD_MAX_RATE],
8607 constants.LDP_MIN_RATE: dt_params[constants.DRBD_MIN_RATE],
8611 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_DRBD8],
8614 result.append(drbd_params)
8618 constants.LDP_STRIPES: dt_params[constants.DRBD_DATA_STRIPES],
8621 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_LV],
8623 result.append(data_params)
8627 constants.LDP_STRIPES: dt_params[constants.DRBD_META_STRIPES],
8630 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_LV],
8632 result.append(meta_params)
8634 elif (disk_template == constants.DT_FILE or
8635 disk_template == constants.DT_SHARED_FILE):
8636 result.append(constants.DISK_LD_DEFAULTS[constants.LD_FILE])
8638 elif disk_template == constants.DT_PLAIN:
8640 constants.LDP_STRIPES: dt_params[constants.LV_STRIPES],
8643 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_LV],
8645 result.append(params)
8647 elif disk_template == constants.DT_BLOCK:
8648 result.append(constants.DISK_LD_DEFAULTS[constants.LD_BLOCKDEV])
8650 elif disk_template == constants.DT_RBD:
8652 constants.LDP_POOL: dt_params[constants.RBD_POOL]
8655 objects.FillDict(constants.DISK_LD_DEFAULTS[constants.LD_RBD],
8657 result.append(params)
8662 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
8663 iv_name, p_minor, s_minor, drbd_params, data_params,
8665 """Generate a drbd8 device complete with its children.
8668 assert len(vgnames) == len(names) == 2
8669 port = lu.cfg.AllocatePort()
8670 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
8672 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
8673 logical_id=(vgnames[0], names[0]),
8675 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,
8676 logical_id=(vgnames[1], names[1]),
8678 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
8679 logical_id=(primary, secondary, port,
8682 children=[dev_data, dev_meta],
8683 iv_name=iv_name, params=drbd_params)
8687 _DISK_TEMPLATE_NAME_PREFIX = {
8688 constants.DT_PLAIN: "",
8689 constants.DT_RBD: ".rbd",
8693 _DISK_TEMPLATE_DEVICE_TYPE = {
8694 constants.DT_PLAIN: constants.LD_LV,
8695 constants.DT_FILE: constants.LD_FILE,
8696 constants.DT_SHARED_FILE: constants.LD_FILE,
8697 constants.DT_BLOCK: constants.LD_BLOCKDEV,
8698 constants.DT_RBD: constants.LD_RBD,
8702 def _GenerateDiskTemplate(lu, template_name, instance_name, primary_node,
8703 secondary_nodes, disk_info, file_storage_dir, file_driver, base_index,
8704 feedback_fn, disk_params,
8705 _req_file_storage=opcodes.RequireFileStorage,
8706 _req_shr_file_storage=opcodes.RequireSharedFileStorage):
8707 """Generate the entire disk layout for a given template type.
8710 #TODO: compute space requirements
8712 vgname = lu.cfg.GetVGName()
8713 disk_count = len(disk_info)
8715 ld_params = _ComputeLDParams(template_name, disk_params)
8717 if template_name == constants.DT_DISKLESS:
8719 elif template_name == constants.DT_DRBD8:
8720 drbd_params, data_params, meta_params = ld_params
8721 if len(secondary_nodes) != 1:
8722 raise errors.ProgrammerError("Wrong template configuration")
8723 remote_node = secondary_nodes[0]
8724 minors = lu.cfg.AllocateDRBDMinor(
8725 [primary_node, remote_node] * len(disk_info), instance_name)
8728 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
8729 for i in range(disk_count)]):
8730 names.append(lv_prefix + "_data")
8731 names.append(lv_prefix + "_meta")
8732 for idx, disk in enumerate(disk_info):
8733 disk_index = idx + base_index
8734 drbd_default_metavg = drbd_params[constants.LDP_DEFAULT_METAVG]
8735 data_vg = disk.get(constants.IDISK_VG, vgname)
8736 meta_vg = disk.get(constants.IDISK_METAVG, drbd_default_metavg)
8737 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
8738 disk[constants.IDISK_SIZE],
8740 names[idx * 2:idx * 2 + 2],
8741 "disk/%d" % disk_index,
8742 minors[idx * 2], minors[idx * 2 + 1],
8743 drbd_params, data_params, meta_params)
8744 disk_dev.mode = disk[constants.IDISK_MODE]
8745 disks.append(disk_dev)
8748 raise errors.ProgrammerError("Wrong template configuration")
8750 if template_name == constants.DT_FILE:
8752 elif template_name == constants.DT_SHARED_FILE:
8753 _req_shr_file_storage()
8755 name_prefix = _DISK_TEMPLATE_NAME_PREFIX.get(template_name, None)
8756 if name_prefix is None:
8759 names = _GenerateUniqueNames(lu, ["%s.disk%s" %
8760 (name_prefix, base_index + i)
8761 for i in range(disk_count)])
8763 dev_type = _DISK_TEMPLATE_DEVICE_TYPE[template_name]
8765 if template_name == constants.DT_PLAIN:
8766 def logical_id_fn(idx, _, disk):
8767 vg = disk.get(constants.IDISK_VG, vgname)
8768 return (vg, names[idx])
8769 elif template_name in (constants.DT_FILE, constants.DT_SHARED_FILE):
8771 lambda _, disk_index, disk: (file_driver,
8772 "%s/disk%d" % (file_storage_dir,
8774 elif template_name == constants.DT_BLOCK:
8776 lambda idx, disk_index, disk: (constants.BLOCKDEV_DRIVER_MANUAL,
8777 disk[constants.IDISK_ADOPT])
8778 elif template_name == constants.DT_RBD:
8779 logical_id_fn = lambda idx, _, disk: ("rbd", names[idx])
8781 raise errors.ProgrammerError("Unknown disk template '%s'" % template_name)
8783 for idx, disk in enumerate(disk_info):
8784 disk_index = idx + base_index
8785 size = disk[constants.IDISK_SIZE]
8786 feedback_fn("* disk %s, size %s" %
8787 (disk_index, utils.FormatUnit(size, "h")))
8788 disks.append(objects.Disk(dev_type=dev_type, size=size,
8789 logical_id=logical_id_fn(idx, disk_index, disk),
8790 iv_name="disk/%d" % disk_index,
8791 mode=disk[constants.IDISK_MODE],
8792 params=ld_params[0]))
8797 def _GetInstanceInfoText(instance):
8798 """Compute that text that should be added to the disk's metadata.
8801 return "originstname+%s" % instance.name
8804 def _CalcEta(time_taken, written, total_size):
8805 """Calculates the ETA based on size written and total size.
8807 @param time_taken: The time taken so far
8808 @param written: amount written so far
8809 @param total_size: The total size of data to be written
8810 @return: The remaining time in seconds
8813 avg_time = time_taken / float(written)
8814 return (total_size - written) * avg_time
8817 def _WipeDisks(lu, instance):
8818 """Wipes instance disks.
8820 @type lu: L{LogicalUnit}
8821 @param lu: the logical unit on whose behalf we execute
8822 @type instance: L{objects.Instance}
8823 @param instance: the instance whose disks we should create
8824 @return: the success of the wipe
8827 node = instance.primary_node
8829 for device in instance.disks:
8830 lu.cfg.SetDiskID(device, node)
8832 logging.info("Pause sync of instance %s disks", instance.name)
8833 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
8835 for idx, success in enumerate(result.payload):
8837 logging.warn("pause-sync of instance %s for disks %d failed",
8841 for idx, device in enumerate(instance.disks):
8842 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
8843 # MAX_WIPE_CHUNK at max
8844 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
8845 constants.MIN_WIPE_CHUNK_PERCENT)
8846 # we _must_ make this an int, otherwise rounding errors will
8848 wipe_chunk_size = int(wipe_chunk_size)
8850 lu.LogInfo("* Wiping disk %d", idx)
8851 logging.info("Wiping disk %d for instance %s, node %s using"
8852 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
8857 start_time = time.time()
8859 while offset < size:
8860 wipe_size = min(wipe_chunk_size, size - offset)
8861 logging.debug("Wiping disk %d, offset %s, chunk %s",
8862 idx, offset, wipe_size)
8863 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
8864 result.Raise("Could not wipe disk %d at offset %d for size %d" %
8865 (idx, offset, wipe_size))
8868 if now - last_output >= 60:
8869 eta = _CalcEta(now - start_time, offset, size)
8870 lu.LogInfo(" - done: %.1f%% ETA: %s" %
8871 (offset / float(size) * 100, utils.FormatSeconds(eta)))
8874 logging.info("Resume sync of instance %s disks", instance.name)
8876 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
8878 for idx, success in enumerate(result.payload):
8880 lu.LogWarning("Resume sync of disk %d failed, please have a"
8881 " look at the status and troubleshoot the issue", idx)
8882 logging.warn("resume-sync of instance %s for disks %d failed",
8886 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
8887 """Create all disks for an instance.
8889 This abstracts away some work from AddInstance.
8891 @type lu: L{LogicalUnit}
8892 @param lu: the logical unit on whose behalf we execute
8893 @type instance: L{objects.Instance}
8894 @param instance: the instance whose disks we should create
8896 @param to_skip: list of indices to skip
8897 @type target_node: string
8898 @param target_node: if passed, overrides the target node for creation
8900 @return: the success of the creation
8903 info = _GetInstanceInfoText(instance)
8904 if target_node is None:
8905 pnode = instance.primary_node
8906 all_nodes = instance.all_nodes
8911 if instance.disk_template in constants.DTS_FILEBASED:
8912 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
8913 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
8915 result.Raise("Failed to create directory '%s' on"
8916 " node %s" % (file_storage_dir, pnode))
8918 # Note: this needs to be kept in sync with adding of disks in
8919 # LUInstanceSetParams
8920 for idx, device in enumerate(instance.disks):
8921 if to_skip and idx in to_skip:
8923 logging.info("Creating disk %s for instance '%s'", idx, instance.name)
8925 for node in all_nodes:
8926 f_create = node == pnode
8927 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
8930 def _RemoveDisks(lu, instance, target_node=None):
8931 """Remove all disks for an instance.
8933 This abstracts away some work from `AddInstance()` and
8934 `RemoveInstance()`. Note that in case some of the devices couldn't
8935 be removed, the removal will continue with the other ones (compare
8936 with `_CreateDisks()`).
8938 @type lu: L{LogicalUnit}
8939 @param lu: the logical unit on whose behalf we execute
8940 @type instance: L{objects.Instance}
8941 @param instance: the instance whose disks we should remove
8942 @type target_node: string
8943 @param target_node: used to override the node on which to remove the disks
8945 @return: the success of the removal
8948 logging.info("Removing block devices for instance %s", instance.name)
8951 for (idx, device) in enumerate(instance.disks):
8953 edata = [(target_node, device)]
8955 edata = device.ComputeNodeTree(instance.primary_node)
8956 for node, disk in edata:
8957 lu.cfg.SetDiskID(disk, node)
8958 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
8960 lu.LogWarning("Could not remove disk %s on node %s,"
8961 " continuing anyway: %s", idx, node, msg)
8964 # if this is a DRBD disk, return its port to the pool
8965 if device.dev_type in constants.LDS_DRBD:
8966 tcp_port = device.logical_id[2]
8967 lu.cfg.AddTcpUdpPort(tcp_port)
8969 if instance.disk_template == constants.DT_FILE:
8970 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
8974 tgt = instance.primary_node
8975 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
8977 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
8978 file_storage_dir, instance.primary_node, result.fail_msg)
8984 def _ComputeDiskSizePerVG(disk_template, disks):
8985 """Compute disk size requirements in the volume group
8988 def _compute(disks, payload):
8989 """Universal algorithm.
8994 vgs[disk[constants.IDISK_VG]] = \
8995 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
8999 # Required free disk space as a function of disk and swap space
9001 constants.DT_DISKLESS: {},
9002 constants.DT_PLAIN: _compute(disks, 0),
9003 # 128 MB are added for drbd metadata for each disk
9004 constants.DT_DRBD8: _compute(disks, DRBD_META_SIZE),
9005 constants.DT_FILE: {},
9006 constants.DT_SHARED_FILE: {},
9009 if disk_template not in req_size_dict:
9010 raise errors.ProgrammerError("Disk template '%s' size requirement"
9011 " is unknown" % disk_template)
9013 return req_size_dict[disk_template]
9016 def _ComputeDiskSize(disk_template, disks):
9017 """Compute disk size requirements in the volume group
9020 # Required free disk space as a function of disk and swap space
9022 constants.DT_DISKLESS: None,
9023 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
9024 # 128 MB are added for drbd metadata for each disk
9026 sum(d[constants.IDISK_SIZE] + DRBD_META_SIZE for d in disks),
9027 constants.DT_FILE: None,
9028 constants.DT_SHARED_FILE: 0,
9029 constants.DT_BLOCK: 0,
9030 constants.DT_RBD: 0,
9033 if disk_template not in req_size_dict:
9034 raise errors.ProgrammerError("Disk template '%s' size requirement"
9035 " is unknown" % disk_template)
9037 return req_size_dict[disk_template]
9040 def _FilterVmNodes(lu, nodenames):
9041 """Filters out non-vm_capable nodes from a list.
9043 @type lu: L{LogicalUnit}
9044 @param lu: the logical unit for which we check
9045 @type nodenames: list
9046 @param nodenames: the list of nodes on which we should check
9048 @return: the list of vm-capable nodes
9051 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
9052 return [name for name in nodenames if name not in vm_nodes]
9055 def _CheckHVParams(lu, nodenames, hvname, hvparams):
9056 """Hypervisor parameter validation.
9058 This function abstract the hypervisor parameter validation to be
9059 used in both instance create and instance modify.
9061 @type lu: L{LogicalUnit}
9062 @param lu: the logical unit for which we check
9063 @type nodenames: list
9064 @param nodenames: the list of nodes on which we should check
9065 @type hvname: string
9066 @param hvname: the name of the hypervisor we should use
9067 @type hvparams: dict
9068 @param hvparams: the parameters which we need to check
9069 @raise errors.OpPrereqError: if the parameters are not valid
9072 nodenames = _FilterVmNodes(lu, nodenames)
9074 cluster = lu.cfg.GetClusterInfo()
9075 hvfull = objects.FillDict(cluster.hvparams.get(hvname, {}), hvparams)
9077 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames, hvname, hvfull)
9078 for node in nodenames:
9082 info.Raise("Hypervisor parameter validation failed on node %s" % node)
9085 def _CheckOSParams(lu, required, nodenames, osname, osparams):
9086 """OS parameters validation.
9088 @type lu: L{LogicalUnit}
9089 @param lu: the logical unit for which we check
9090 @type required: boolean
9091 @param required: whether the validation should fail if the OS is not
9093 @type nodenames: list
9094 @param nodenames: the list of nodes on which we should check
9095 @type osname: string
9096 @param osname: the name of the hypervisor we should use
9097 @type osparams: dict
9098 @param osparams: the parameters which we need to check
9099 @raise errors.OpPrereqError: if the parameters are not valid
9102 nodenames = _FilterVmNodes(lu, nodenames)
9103 result = lu.rpc.call_os_validate(nodenames, required, osname,
9104 [constants.OS_VALIDATE_PARAMETERS],
9106 for node, nres in result.items():
9107 # we don't check for offline cases since this should be run only
9108 # against the master node and/or an instance's nodes
9109 nres.Raise("OS Parameters validation failed on node %s" % node)
9110 if not nres.payload:
9111 lu.LogInfo("OS %s not found on node %s, validation skipped",
9115 class LUInstanceCreate(LogicalUnit):
9116 """Create an instance.
9119 HPATH = "instance-add"
9120 HTYPE = constants.HTYPE_INSTANCE
9123 def CheckArguments(self):
9127 # do not require name_check to ease forward/backward compatibility
9129 if self.op.no_install and self.op.start:
9130 self.LogInfo("No-installation mode selected, disabling startup")
9131 self.op.start = False
9132 # validate/normalize the instance name
9133 self.op.instance_name = \
9134 netutils.Hostname.GetNormalizedName(self.op.instance_name)
9136 if self.op.ip_check and not self.op.name_check:
9137 # TODO: make the ip check more flexible and not depend on the name check
9138 raise errors.OpPrereqError("Cannot do IP address check without a name"
9139 " check", errors.ECODE_INVAL)
9141 # check nics' parameter names
9142 for nic in self.op.nics:
9143 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
9145 # check disks. parameter names and consistent adopt/no-adopt strategy
9146 has_adopt = has_no_adopt = False
9147 for disk in self.op.disks:
9148 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
9149 if constants.IDISK_ADOPT in disk:
9153 if has_adopt and has_no_adopt:
9154 raise errors.OpPrereqError("Either all disks are adopted or none is",
9157 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
9158 raise errors.OpPrereqError("Disk adoption is not supported for the"
9159 " '%s' disk template" %
9160 self.op.disk_template,
9162 if self.op.iallocator is not None:
9163 raise errors.OpPrereqError("Disk adoption not allowed with an"
9164 " iallocator script", errors.ECODE_INVAL)
9165 if self.op.mode == constants.INSTANCE_IMPORT:
9166 raise errors.OpPrereqError("Disk adoption not allowed for"
9167 " instance import", errors.ECODE_INVAL)
9169 if self.op.disk_template in constants.DTS_MUST_ADOPT:
9170 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
9171 " but no 'adopt' parameter given" %
9172 self.op.disk_template,
9175 self.adopt_disks = has_adopt
9177 # instance name verification
9178 if self.op.name_check:
9179 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
9180 self.op.instance_name = self.hostname1.name
9181 # used in CheckPrereq for ip ping check
9182 self.check_ip = self.hostname1.ip
9184 self.check_ip = None
9186 # file storage checks
9187 if (self.op.file_driver and
9188 not self.op.file_driver in constants.FILE_DRIVER):
9189 raise errors.OpPrereqError("Invalid file driver name '%s'" %
9190 self.op.file_driver, errors.ECODE_INVAL)
9192 if self.op.disk_template == constants.DT_FILE:
9193 opcodes.RequireFileStorage()
9194 elif self.op.disk_template == constants.DT_SHARED_FILE:
9195 opcodes.RequireSharedFileStorage()
9197 ### Node/iallocator related checks
9198 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
9200 if self.op.pnode is not None:
9201 if self.op.disk_template in constants.DTS_INT_MIRROR:
9202 if self.op.snode is None:
9203 raise errors.OpPrereqError("The networked disk templates need"
9204 " a mirror node", errors.ECODE_INVAL)
9206 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
9208 self.op.snode = None
9210 self._cds = _GetClusterDomainSecret()
9212 if self.op.mode == constants.INSTANCE_IMPORT:
9213 # On import force_variant must be True, because if we forced it at
9214 # initial install, our only chance when importing it back is that it
9216 self.op.force_variant = True
9218 if self.op.no_install:
9219 self.LogInfo("No-installation mode has no effect during import")
9221 elif self.op.mode == constants.INSTANCE_CREATE:
9222 if self.op.os_type is None:
9223 raise errors.OpPrereqError("No guest OS specified",
9225 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
9226 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
9227 " installation" % self.op.os_type,
9229 if self.op.disk_template is None:
9230 raise errors.OpPrereqError("No disk template specified",
9233 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
9234 # Check handshake to ensure both clusters have the same domain secret
9235 src_handshake = self.op.source_handshake
9236 if not src_handshake:
9237 raise errors.OpPrereqError("Missing source handshake",
9240 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
9243 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
9246 # Load and check source CA
9247 self.source_x509_ca_pem = self.op.source_x509_ca
9248 if not self.source_x509_ca_pem:
9249 raise errors.OpPrereqError("Missing source X509 CA",
9253 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
9255 except OpenSSL.crypto.Error, err:
9256 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
9257 (err, ), errors.ECODE_INVAL)
9259 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9260 if errcode is not None:
9261 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
9264 self.source_x509_ca = cert
9266 src_instance_name = self.op.source_instance_name
9267 if not src_instance_name:
9268 raise errors.OpPrereqError("Missing source instance name",
9271 self.source_instance_name = \
9272 netutils.GetHostname(name=src_instance_name).name
9275 raise errors.OpPrereqError("Invalid instance creation mode %r" %
9276 self.op.mode, errors.ECODE_INVAL)
9278 def ExpandNames(self):
9279 """ExpandNames for CreateInstance.
9281 Figure out the right locks for instance creation.
9284 self.needed_locks = {}
9286 instance_name = self.op.instance_name
9287 # this is just a preventive check, but someone might still add this
9288 # instance in the meantime, and creation will fail at lock-add time
9289 if instance_name in self.cfg.GetInstanceList():
9290 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
9291 instance_name, errors.ECODE_EXISTS)
9293 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
9295 if self.op.iallocator:
9296 # TODO: Find a solution to not lock all nodes in the cluster, e.g. by
9297 # specifying a group on instance creation and then selecting nodes from
9299 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9300 self.needed_locks[locking.LEVEL_NODE_RES] = locking.ALL_SET
9302 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
9303 nodelist = [self.op.pnode]
9304 if self.op.snode is not None:
9305 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
9306 nodelist.append(self.op.snode)
9307 self.needed_locks[locking.LEVEL_NODE] = nodelist
9308 # Lock resources of instance's primary and secondary nodes (copy to
9309 # prevent accidential modification)
9310 self.needed_locks[locking.LEVEL_NODE_RES] = list(nodelist)
9312 # in case of import lock the source node too
9313 if self.op.mode == constants.INSTANCE_IMPORT:
9314 src_node = self.op.src_node
9315 src_path = self.op.src_path
9317 if src_path is None:
9318 self.op.src_path = src_path = self.op.instance_name
9320 if src_node is None:
9321 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9322 self.op.src_node = None
9323 if os.path.isabs(src_path):
9324 raise errors.OpPrereqError("Importing an instance from a path"
9325 " requires a source node option",
9328 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
9329 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
9330 self.needed_locks[locking.LEVEL_NODE].append(src_node)
9331 if not os.path.isabs(src_path):
9332 self.op.src_path = src_path = \
9333 utils.PathJoin(constants.EXPORT_DIR, src_path)
9335 def _RunAllocator(self):
9336 """Run the allocator based on input opcode.
9339 nics = [n.ToDict() for n in self.nics]
9340 ial = IAllocator(self.cfg, self.rpc,
9341 mode=constants.IALLOCATOR_MODE_ALLOC,
9342 name=self.op.instance_name,
9343 disk_template=self.op.disk_template,
9346 vcpus=self.be_full[constants.BE_VCPUS],
9347 memory=self.be_full[constants.BE_MAXMEM],
9350 hypervisor=self.op.hypervisor,
9353 ial.Run(self.op.iallocator)
9356 raise errors.OpPrereqError("Can't compute nodes using"
9357 " iallocator '%s': %s" %
9358 (self.op.iallocator, ial.info),
9360 if len(ial.result) != ial.required_nodes:
9361 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
9362 " of nodes (%s), required %s" %
9363 (self.op.iallocator, len(ial.result),
9364 ial.required_nodes), errors.ECODE_FAULT)
9365 self.op.pnode = ial.result[0]
9366 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
9367 self.op.instance_name, self.op.iallocator,
9368 utils.CommaJoin(ial.result))
9369 if ial.required_nodes == 2:
9370 self.op.snode = ial.result[1]
9372 def BuildHooksEnv(self):
9375 This runs on master, primary and secondary nodes of the instance.
9379 "ADD_MODE": self.op.mode,
9381 if self.op.mode == constants.INSTANCE_IMPORT:
9382 env["SRC_NODE"] = self.op.src_node
9383 env["SRC_PATH"] = self.op.src_path
9384 env["SRC_IMAGES"] = self.src_images
9386 env.update(_BuildInstanceHookEnv(
9387 name=self.op.instance_name,
9388 primary_node=self.op.pnode,
9389 secondary_nodes=self.secondaries,
9390 status=self.op.start,
9391 os_type=self.op.os_type,
9392 minmem=self.be_full[constants.BE_MINMEM],
9393 maxmem=self.be_full[constants.BE_MAXMEM],
9394 vcpus=self.be_full[constants.BE_VCPUS],
9395 nics=_NICListToTuple(self, self.nics),
9396 disk_template=self.op.disk_template,
9397 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
9398 for d in self.disks],
9401 hypervisor_name=self.op.hypervisor,
9407 def BuildHooksNodes(self):
9408 """Build hooks nodes.
9411 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
9414 def _ReadExportInfo(self):
9415 """Reads the export information from disk.
9417 It will override the opcode source node and path with the actual
9418 information, if these two were not specified before.
9420 @return: the export information
9423 assert self.op.mode == constants.INSTANCE_IMPORT
9425 src_node = self.op.src_node
9426 src_path = self.op.src_path
9428 if src_node is None:
9429 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
9430 exp_list = self.rpc.call_export_list(locked_nodes)
9432 for node in exp_list:
9433 if exp_list[node].fail_msg:
9435 if src_path in exp_list[node].payload:
9437 self.op.src_node = src_node = node
9438 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
9442 raise errors.OpPrereqError("No export found for relative path %s" %
9443 src_path, errors.ECODE_INVAL)
9445 _CheckNodeOnline(self, src_node)
9446 result = self.rpc.call_export_info(src_node, src_path)
9447 result.Raise("No export or invalid export found in dir %s" % src_path)
9449 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
9450 if not export_info.has_section(constants.INISECT_EXP):
9451 raise errors.ProgrammerError("Corrupted export config",
9452 errors.ECODE_ENVIRON)
9454 ei_version = export_info.get(constants.INISECT_EXP, "version")
9455 if (int(ei_version) != constants.EXPORT_VERSION):
9456 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
9457 (ei_version, constants.EXPORT_VERSION),
9458 errors.ECODE_ENVIRON)
9461 def _ReadExportParams(self, einfo):
9462 """Use export parameters as defaults.
9464 In case the opcode doesn't specify (as in override) some instance
9465 parameters, then try to use them from the export information, if
9469 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
9471 if self.op.disk_template is None:
9472 if einfo.has_option(constants.INISECT_INS, "disk_template"):
9473 self.op.disk_template = einfo.get(constants.INISECT_INS,
9475 if self.op.disk_template not in constants.DISK_TEMPLATES:
9476 raise errors.OpPrereqError("Disk template specified in configuration"
9477 " file is not one of the allowed values:"
9478 " %s" % " ".join(constants.DISK_TEMPLATES))
9480 raise errors.OpPrereqError("No disk template specified and the export"
9481 " is missing the disk_template information",
9484 if not self.op.disks:
9486 # TODO: import the disk iv_name too
9487 for idx in range(constants.MAX_DISKS):
9488 if einfo.has_option(constants.INISECT_INS, "disk%d_size" % idx):
9489 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
9490 disks.append({constants.IDISK_SIZE: disk_sz})
9491 self.op.disks = disks
9492 if not disks and self.op.disk_template != constants.DT_DISKLESS:
9493 raise errors.OpPrereqError("No disk info specified and the export"
9494 " is missing the disk information",
9497 if not self.op.nics:
9499 for idx in range(constants.MAX_NICS):
9500 if einfo.has_option(constants.INISECT_INS, "nic%d_mac" % idx):
9502 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
9503 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
9510 if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):
9511 self.op.tags = einfo.get(constants.INISECT_INS, "tags").split()
9513 if (self.op.hypervisor is None and
9514 einfo.has_option(constants.INISECT_INS, "hypervisor")):
9515 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
9517 if einfo.has_section(constants.INISECT_HYP):
9518 # use the export parameters but do not override the ones
9519 # specified by the user
9520 for name, value in einfo.items(constants.INISECT_HYP):
9521 if name not in self.op.hvparams:
9522 self.op.hvparams[name] = value
9524 if einfo.has_section(constants.INISECT_BEP):
9525 # use the parameters, without overriding
9526 for name, value in einfo.items(constants.INISECT_BEP):
9527 if name not in self.op.beparams:
9528 self.op.beparams[name] = value
9529 # Compatibility for the old "memory" be param
9530 if name == constants.BE_MEMORY:
9531 if constants.BE_MAXMEM not in self.op.beparams:
9532 self.op.beparams[constants.BE_MAXMEM] = value
9533 if constants.BE_MINMEM not in self.op.beparams:
9534 self.op.beparams[constants.BE_MINMEM] = value
9536 # try to read the parameters old style, from the main section
9537 for name in constants.BES_PARAMETERS:
9538 if (name not in self.op.beparams and
9539 einfo.has_option(constants.INISECT_INS, name)):
9540 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
9542 if einfo.has_section(constants.INISECT_OSP):
9543 # use the parameters, without overriding
9544 for name, value in einfo.items(constants.INISECT_OSP):
9545 if name not in self.op.osparams:
9546 self.op.osparams[name] = value
9548 def _RevertToDefaults(self, cluster):
9549 """Revert the instance parameters to the default values.
9553 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
9554 for name in self.op.hvparams.keys():
9555 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
9556 del self.op.hvparams[name]
9558 be_defs = cluster.SimpleFillBE({})
9559 for name in self.op.beparams.keys():
9560 if name in be_defs and be_defs[name] == self.op.beparams[name]:
9561 del self.op.beparams[name]
9563 nic_defs = cluster.SimpleFillNIC({})
9564 for nic in self.op.nics:
9565 for name in constants.NICS_PARAMETERS:
9566 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
9569 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
9570 for name in self.op.osparams.keys():
9571 if name in os_defs and os_defs[name] == self.op.osparams[name]:
9572 del self.op.osparams[name]
9574 def _CalculateFileStorageDir(self):
9575 """Calculate final instance file storage dir.
9578 # file storage dir calculation/check
9579 self.instance_file_storage_dir = None
9580 if self.op.disk_template in constants.DTS_FILEBASED:
9581 # build the full file storage dir path
9584 if self.op.disk_template == constants.DT_SHARED_FILE:
9585 get_fsd_fn = self.cfg.GetSharedFileStorageDir
9587 get_fsd_fn = self.cfg.GetFileStorageDir
9589 cfg_storagedir = get_fsd_fn()
9590 if not cfg_storagedir:
9591 raise errors.OpPrereqError("Cluster file storage dir not defined")
9592 joinargs.append(cfg_storagedir)
9594 if self.op.file_storage_dir is not None:
9595 joinargs.append(self.op.file_storage_dir)
9597 joinargs.append(self.op.instance_name)
9599 # pylint: disable=W0142
9600 self.instance_file_storage_dir = utils.PathJoin(*joinargs)
9602 def CheckPrereq(self): # pylint: disable=R0914
9603 """Check prerequisites.
9606 self._CalculateFileStorageDir()
9608 if self.op.mode == constants.INSTANCE_IMPORT:
9609 export_info = self._ReadExportInfo()
9610 self._ReadExportParams(export_info)
9612 if (not self.cfg.GetVGName() and
9613 self.op.disk_template not in constants.DTS_NOT_LVM):
9614 raise errors.OpPrereqError("Cluster does not support lvm-based"
9615 " instances", errors.ECODE_STATE)
9617 if (self.op.hypervisor is None or
9618 self.op.hypervisor == constants.VALUE_AUTO):
9619 self.op.hypervisor = self.cfg.GetHypervisorType()
9621 cluster = self.cfg.GetClusterInfo()
9622 enabled_hvs = cluster.enabled_hypervisors
9623 if self.op.hypervisor not in enabled_hvs:
9624 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
9625 " cluster (%s)" % (self.op.hypervisor,
9626 ",".join(enabled_hvs)),
9629 # Check tag validity
9630 for tag in self.op.tags:
9631 objects.TaggableObject.ValidateTag(tag)
9633 # check hypervisor parameter syntax (locally)
9634 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
9635 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
9637 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
9638 hv_type.CheckParameterSyntax(filled_hvp)
9639 self.hv_full = filled_hvp
9640 # check that we don't specify global parameters on an instance
9641 _CheckGlobalHvParams(self.op.hvparams)
9643 # fill and remember the beparams dict
9644 default_beparams = cluster.beparams[constants.PP_DEFAULT]
9645 for param, value in self.op.beparams.iteritems():
9646 if value == constants.VALUE_AUTO:
9647 self.op.beparams[param] = default_beparams[param]
9648 objects.UpgradeBeParams(self.op.beparams)
9649 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
9650 self.be_full = cluster.SimpleFillBE(self.op.beparams)
9652 # build os parameters
9653 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
9655 # now that hvp/bep are in final format, let's reset to defaults,
9657 if self.op.identify_defaults:
9658 self._RevertToDefaults(cluster)
9662 for idx, nic in enumerate(self.op.nics):
9663 nic_mode_req = nic.get(constants.INIC_MODE, None)
9664 nic_mode = nic_mode_req
9665 if nic_mode is None or nic_mode == constants.VALUE_AUTO:
9666 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
9668 # in routed mode, for the first nic, the default ip is 'auto'
9669 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
9670 default_ip_mode = constants.VALUE_AUTO
9672 default_ip_mode = constants.VALUE_NONE
9674 # ip validity checks
9675 ip = nic.get(constants.INIC_IP, default_ip_mode)
9676 if ip is None or ip.lower() == constants.VALUE_NONE:
9678 elif ip.lower() == constants.VALUE_AUTO:
9679 if not self.op.name_check:
9680 raise errors.OpPrereqError("IP address set to auto but name checks"
9681 " have been skipped",
9683 nic_ip = self.hostname1.ip
9685 if not netutils.IPAddress.IsValid(ip):
9686 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
9690 # TODO: check the ip address for uniqueness
9691 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
9692 raise errors.OpPrereqError("Routed nic mode requires an ip address",
9695 # MAC address verification
9696 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
9697 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9698 mac = utils.NormalizeAndValidateMac(mac)
9701 self.cfg.ReserveMAC(mac, self.proc.GetECId())
9702 except errors.ReservationError:
9703 raise errors.OpPrereqError("MAC address %s already in use"
9704 " in cluster" % mac,
9705 errors.ECODE_NOTUNIQUE)
9707 # Build nic parameters
9708 link = nic.get(constants.INIC_LINK, None)
9709 if link == constants.VALUE_AUTO:
9710 link = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_LINK]
9713 nicparams[constants.NIC_MODE] = nic_mode
9715 nicparams[constants.NIC_LINK] = link
9717 check_params = cluster.SimpleFillNIC(nicparams)
9718 objects.NIC.CheckParameterSyntax(check_params)
9719 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
9721 # disk checks/pre-build
9722 default_vg = self.cfg.GetVGName()
9724 for disk in self.op.disks:
9725 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
9726 if mode not in constants.DISK_ACCESS_SET:
9727 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
9728 mode, errors.ECODE_INVAL)
9729 size = disk.get(constants.IDISK_SIZE, None)
9731 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
9734 except (TypeError, ValueError):
9735 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
9738 data_vg = disk.get(constants.IDISK_VG, default_vg)
9740 constants.IDISK_SIZE: size,
9741 constants.IDISK_MODE: mode,
9742 constants.IDISK_VG: data_vg,
9744 if constants.IDISK_METAVG in disk:
9745 new_disk[constants.IDISK_METAVG] = disk[constants.IDISK_METAVG]
9746 if constants.IDISK_ADOPT in disk:
9747 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
9748 self.disks.append(new_disk)
9750 if self.op.mode == constants.INSTANCE_IMPORT:
9752 for idx in range(len(self.disks)):
9753 option = "disk%d_dump" % idx
9754 if export_info.has_option(constants.INISECT_INS, option):
9755 # FIXME: are the old os-es, disk sizes, etc. useful?
9756 export_name = export_info.get(constants.INISECT_INS, option)
9757 image = utils.PathJoin(self.op.src_path, export_name)
9758 disk_images.append(image)
9760 disk_images.append(False)
9762 self.src_images = disk_images
9764 old_name = export_info.get(constants.INISECT_INS, "name")
9765 if self.op.instance_name == old_name:
9766 for idx, nic in enumerate(self.nics):
9767 if nic.mac == constants.VALUE_AUTO:
9768 nic_mac_ini = "nic%d_mac" % idx
9769 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
9771 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
9773 # ip ping checks (we use the same ip that was resolved in ExpandNames)
9774 if self.op.ip_check:
9775 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
9776 raise errors.OpPrereqError("IP %s of instance %s already in use" %
9777 (self.check_ip, self.op.instance_name),
9778 errors.ECODE_NOTUNIQUE)
9780 #### mac address generation
9781 # By generating here the mac address both the allocator and the hooks get
9782 # the real final mac address rather than the 'auto' or 'generate' value.
9783 # There is a race condition between the generation and the instance object
9784 # creation, which means that we know the mac is valid now, but we're not
9785 # sure it will be when we actually add the instance. If things go bad
9786 # adding the instance will abort because of a duplicate mac, and the
9787 # creation job will fail.
9788 for nic in self.nics:
9789 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9790 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
9794 if self.op.iallocator is not None:
9795 self._RunAllocator()
9797 # Release all unneeded node locks
9798 _ReleaseLocks(self, locking.LEVEL_NODE,
9799 keep=filter(None, [self.op.pnode, self.op.snode,
9801 _ReleaseLocks(self, locking.LEVEL_NODE_RES,
9802 keep=filter(None, [self.op.pnode, self.op.snode,
9805 #### node related checks
9807 # check primary node
9808 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
9809 assert self.pnode is not None, \
9810 "Cannot retrieve locked node %s" % self.op.pnode
9812 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
9813 pnode.name, errors.ECODE_STATE)
9815 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
9816 pnode.name, errors.ECODE_STATE)
9817 if not pnode.vm_capable:
9818 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
9819 " '%s'" % pnode.name, errors.ECODE_STATE)
9821 self.secondaries = []
9823 # mirror node verification
9824 if self.op.disk_template in constants.DTS_INT_MIRROR:
9825 if self.op.snode == pnode.name:
9826 raise errors.OpPrereqError("The secondary node cannot be the"
9827 " primary node", errors.ECODE_INVAL)
9828 _CheckNodeOnline(self, self.op.snode)
9829 _CheckNodeNotDrained(self, self.op.snode)
9830 _CheckNodeVmCapable(self, self.op.snode)
9831 self.secondaries.append(self.op.snode)
9833 snode = self.cfg.GetNodeInfo(self.op.snode)
9834 if pnode.group != snode.group:
9835 self.LogWarning("The primary and secondary nodes are in two"
9836 " different node groups; the disk parameters"
9837 " from the first disk's node group will be"
9840 nodenames = [pnode.name] + self.secondaries
9842 # Verify instance specs
9844 constants.ISPEC_MEM_SIZE: self.be_full.get(constants.BE_MAXMEM, None),
9845 constants.ISPEC_CPU_COUNT: self.be_full.get(constants.BE_VCPUS, None),
9846 constants.ISPEC_DISK_COUNT: len(self.disks),
9847 constants.ISPEC_DISK_SIZE: [disk["size"] for disk in self.disks],
9848 constants.ISPEC_NIC_COUNT: len(self.nics),
9851 group_info = self.cfg.GetNodeGroup(pnode.group)
9852 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
9853 res = _ComputeIPolicyInstanceSpecViolation(ipolicy, ispec)
9854 if not self.op.ignore_ipolicy and res:
9855 raise errors.OpPrereqError(("Instance allocation to group %s violates"
9856 " policy: %s") % (pnode.group,
9857 utils.CommaJoin(res)),
9860 # disk parameters (not customizable at instance or node level)
9861 # just use the primary node parameters, ignoring the secondary.
9862 self.diskparams = group_info.diskparams
9864 if not self.adopt_disks:
9865 if self.op.disk_template == constants.DT_RBD:
9866 # _CheckRADOSFreeSpace() is just a placeholder.
9867 # Any function that checks prerequisites can be placed here.
9868 # Check if there is enough space on the RADOS cluster.
9869 _CheckRADOSFreeSpace()
9871 # Check lv size requirements, if not adopting
9872 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
9873 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
9875 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
9876 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
9877 disk[constants.IDISK_ADOPT])
9878 for disk in self.disks])
9879 if len(all_lvs) != len(self.disks):
9880 raise errors.OpPrereqError("Duplicate volume names given for adoption",
9882 for lv_name in all_lvs:
9884 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
9885 # to ReserveLV uses the same syntax
9886 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
9887 except errors.ReservationError:
9888 raise errors.OpPrereqError("LV named %s used by another instance" %
9889 lv_name, errors.ECODE_NOTUNIQUE)
9891 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
9892 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
9894 node_lvs = self.rpc.call_lv_list([pnode.name],
9895 vg_names.payload.keys())[pnode.name]
9896 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
9897 node_lvs = node_lvs.payload
9899 delta = all_lvs.difference(node_lvs.keys())
9901 raise errors.OpPrereqError("Missing logical volume(s): %s" %
9902 utils.CommaJoin(delta),
9904 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
9906 raise errors.OpPrereqError("Online logical volumes found, cannot"
9907 " adopt: %s" % utils.CommaJoin(online_lvs),
9909 # update the size of disk based on what is found
9910 for dsk in self.disks:
9911 dsk[constants.IDISK_SIZE] = \
9912 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
9913 dsk[constants.IDISK_ADOPT])][0]))
9915 elif self.op.disk_template == constants.DT_BLOCK:
9916 # Normalize and de-duplicate device paths
9917 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
9918 for disk in self.disks])
9919 if len(all_disks) != len(self.disks):
9920 raise errors.OpPrereqError("Duplicate disk names given for adoption",
9922 baddisks = [d for d in all_disks
9923 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
9925 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
9926 " cannot be adopted" %
9927 (", ".join(baddisks),
9928 constants.ADOPTABLE_BLOCKDEV_ROOT),
9931 node_disks = self.rpc.call_bdev_sizes([pnode.name],
9932 list(all_disks))[pnode.name]
9933 node_disks.Raise("Cannot get block device information from node %s" %
9935 node_disks = node_disks.payload
9936 delta = all_disks.difference(node_disks.keys())
9938 raise errors.OpPrereqError("Missing block device(s): %s" %
9939 utils.CommaJoin(delta),
9941 for dsk in self.disks:
9942 dsk[constants.IDISK_SIZE] = \
9943 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
9945 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
9947 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
9948 # check OS parameters (remotely)
9949 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
9951 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
9953 # memory check on primary node
9954 #TODO(dynmem): use MINMEM for checking
9956 _CheckNodeFreeMemory(self, self.pnode.name,
9957 "creating instance %s" % self.op.instance_name,
9958 self.be_full[constants.BE_MAXMEM],
9961 self.dry_run_result = list(nodenames)
9963 def Exec(self, feedback_fn):
9964 """Create and add the instance to the cluster.
9967 instance = self.op.instance_name
9968 pnode_name = self.pnode.name
9970 assert not (self.owned_locks(locking.LEVEL_NODE_RES) -
9971 self.owned_locks(locking.LEVEL_NODE)), \
9972 "Node locks differ from node resource locks"
9974 ht_kind = self.op.hypervisor
9975 if ht_kind in constants.HTS_REQ_PORT:
9976 network_port = self.cfg.AllocatePort()
9980 disks = _GenerateDiskTemplate(self,
9981 self.op.disk_template,
9982 instance, pnode_name,
9985 self.instance_file_storage_dir,
9986 self.op.file_driver,
9991 iobj = objects.Instance(name=instance, os=self.op.os_type,
9992 primary_node=pnode_name,
9993 nics=self.nics, disks=disks,
9994 disk_template=self.op.disk_template,
9995 admin_state=constants.ADMINST_DOWN,
9996 network_port=network_port,
9997 beparams=self.op.beparams,
9998 hvparams=self.op.hvparams,
9999 hypervisor=self.op.hypervisor,
10000 osparams=self.op.osparams,
10004 for tag in self.op.tags:
10007 if self.adopt_disks:
10008 if self.op.disk_template == constants.DT_PLAIN:
10009 # rename LVs to the newly-generated names; we need to construct
10010 # 'fake' LV disks with the old data, plus the new unique_id
10011 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
10013 for t_dsk, a_dsk in zip(tmp_disks, self.disks):
10014 rename_to.append(t_dsk.logical_id)
10015 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
10016 self.cfg.SetDiskID(t_dsk, pnode_name)
10017 result = self.rpc.call_blockdev_rename(pnode_name,
10018 zip(tmp_disks, rename_to))
10019 result.Raise("Failed to rename adoped LVs")
10021 feedback_fn("* creating instance disks...")
10023 _CreateDisks(self, iobj)
10024 except errors.OpExecError:
10025 self.LogWarning("Device creation failed, reverting...")
10027 _RemoveDisks(self, iobj)
10029 self.cfg.ReleaseDRBDMinors(instance)
10032 feedback_fn("adding instance %s to cluster config" % instance)
10034 self.cfg.AddInstance(iobj, self.proc.GetECId())
10036 # Declare that we don't want to remove the instance lock anymore, as we've
10037 # added the instance to the config
10038 del self.remove_locks[locking.LEVEL_INSTANCE]
10040 if self.op.mode == constants.INSTANCE_IMPORT:
10041 # Release unused nodes
10042 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
10044 # Release all nodes
10045 _ReleaseLocks(self, locking.LEVEL_NODE)
10048 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
10049 feedback_fn("* wiping instance disks...")
10051 _WipeDisks(self, iobj)
10052 except errors.OpExecError, err:
10053 logging.exception("Wiping disks failed")
10054 self.LogWarning("Wiping instance disks failed (%s)", err)
10058 # Something is already wrong with the disks, don't do anything else
10060 elif self.op.wait_for_sync:
10061 disk_abort = not _WaitForSync(self, iobj)
10062 elif iobj.disk_template in constants.DTS_INT_MIRROR:
10063 # make sure the disks are not degraded (still sync-ing is ok)
10064 feedback_fn("* checking mirrors status")
10065 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
10070 _RemoveDisks(self, iobj)
10071 self.cfg.RemoveInstance(iobj.name)
10072 # Make sure the instance lock gets removed
10073 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
10074 raise errors.OpExecError("There are some degraded disks for"
10077 # Release all node resource locks
10078 _ReleaseLocks(self, locking.LEVEL_NODE_RES)
10080 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
10081 if self.op.mode == constants.INSTANCE_CREATE:
10082 if not self.op.no_install:
10083 pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and
10084 not self.op.wait_for_sync)
10086 feedback_fn("* pausing disk sync to install instance OS")
10087 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
10089 for idx, success in enumerate(result.payload):
10091 logging.warn("pause-sync of instance %s for disk %d failed",
10094 feedback_fn("* running the instance OS create scripts...")
10095 # FIXME: pass debug option from opcode to backend
10097 self.rpc.call_instance_os_add(pnode_name, (iobj, None), False,
10098 self.op.debug_level)
10100 feedback_fn("* resuming disk sync")
10101 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
10103 for idx, success in enumerate(result.payload):
10105 logging.warn("resume-sync of instance %s for disk %d failed",
10108 os_add_result.Raise("Could not add os for instance %s"
10109 " on node %s" % (instance, pnode_name))
10111 elif self.op.mode == constants.INSTANCE_IMPORT:
10112 feedback_fn("* running the instance OS import scripts...")
10116 for idx, image in enumerate(self.src_images):
10120 # FIXME: pass debug option from opcode to backend
10121 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
10122 constants.IEIO_FILE, (image, ),
10123 constants.IEIO_SCRIPT,
10124 (iobj.disks[idx], idx),
10126 transfers.append(dt)
10129 masterd.instance.TransferInstanceData(self, feedback_fn,
10130 self.op.src_node, pnode_name,
10131 self.pnode.secondary_ip,
10133 if not compat.all(import_result):
10134 self.LogWarning("Some disks for instance %s on node %s were not"
10135 " imported successfully" % (instance, pnode_name))
10137 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
10138 feedback_fn("* preparing remote import...")
10139 # The source cluster will stop the instance before attempting to make a
10140 # connection. In some cases stopping an instance can take a long time,
10141 # hence the shutdown timeout is added to the connection timeout.
10142 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
10143 self.op.source_shutdown_timeout)
10144 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
10146 assert iobj.primary_node == self.pnode.name
10148 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
10149 self.source_x509_ca,
10150 self._cds, timeouts)
10151 if not compat.all(disk_results):
10152 # TODO: Should the instance still be started, even if some disks
10153 # failed to import (valid for local imports, too)?
10154 self.LogWarning("Some disks for instance %s on node %s were not"
10155 " imported successfully" % (instance, pnode_name))
10157 # Run rename script on newly imported instance
10158 assert iobj.name == instance
10159 feedback_fn("Running rename script for %s" % instance)
10160 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
10161 self.source_instance_name,
10162 self.op.debug_level)
10163 if result.fail_msg:
10164 self.LogWarning("Failed to run rename script for %s on node"
10165 " %s: %s" % (instance, pnode_name, result.fail_msg))
10168 # also checked in the prereq part
10169 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
10172 assert not self.owned_locks(locking.LEVEL_NODE_RES)
10175 iobj.admin_state = constants.ADMINST_UP
10176 self.cfg.Update(iobj, feedback_fn)
10177 logging.info("Starting instance %s on node %s", instance, pnode_name)
10178 feedback_fn("* starting instance...")
10179 result = self.rpc.call_instance_start(pnode_name, (iobj, None, None),
10181 result.Raise("Could not start instance")
10183 return list(iobj.all_nodes)
10186 def _CheckRADOSFreeSpace():
10187 """Compute disk size requirements inside the RADOS cluster.
10190 # For the RADOS cluster we assume there is always enough space.
10194 class LUInstanceConsole(NoHooksLU):
10195 """Connect to an instance's console.
10197 This is somewhat special in that it returns the command line that
10198 you need to run on the master node in order to connect to the
10204 def ExpandNames(self):
10205 self.share_locks = _ShareAll()
10206 self._ExpandAndLockInstance()
10208 def CheckPrereq(self):
10209 """Check prerequisites.
10211 This checks that the instance is in the cluster.
10214 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10215 assert self.instance is not None, \
10216 "Cannot retrieve locked instance %s" % self.op.instance_name
10217 _CheckNodeOnline(self, self.instance.primary_node)
10219 def Exec(self, feedback_fn):
10220 """Connect to the console of an instance
10223 instance = self.instance
10224 node = instance.primary_node
10226 node_insts = self.rpc.call_instance_list([node],
10227 [instance.hypervisor])[node]
10228 node_insts.Raise("Can't get node information from %s" % node)
10230 if instance.name not in node_insts.payload:
10231 if instance.admin_state == constants.ADMINST_UP:
10232 state = constants.INSTST_ERRORDOWN
10233 elif instance.admin_state == constants.ADMINST_DOWN:
10234 state = constants.INSTST_ADMINDOWN
10236 state = constants.INSTST_ADMINOFFLINE
10237 raise errors.OpExecError("Instance %s is not running (state %s)" %
10238 (instance.name, state))
10240 logging.debug("Connecting to console of %s on %s", instance.name, node)
10242 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
10245 def _GetInstanceConsole(cluster, instance):
10246 """Returns console information for an instance.
10248 @type cluster: L{objects.Cluster}
10249 @type instance: L{objects.Instance}
10253 hyper = hypervisor.GetHypervisor(instance.hypervisor)
10254 # beparams and hvparams are passed separately, to avoid editing the
10255 # instance and then saving the defaults in the instance itself.
10256 hvparams = cluster.FillHV(instance)
10257 beparams = cluster.FillBE(instance)
10258 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
10260 assert console.instance == instance.name
10261 assert console.Validate()
10263 return console.ToDict()
10266 class LUInstanceReplaceDisks(LogicalUnit):
10267 """Replace the disks of an instance.
10270 HPATH = "mirrors-replace"
10271 HTYPE = constants.HTYPE_INSTANCE
10274 def CheckArguments(self):
10275 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
10276 self.op.iallocator)
10278 def ExpandNames(self):
10279 self._ExpandAndLockInstance()
10281 assert locking.LEVEL_NODE not in self.needed_locks
10282 assert locking.LEVEL_NODE_RES not in self.needed_locks
10283 assert locking.LEVEL_NODEGROUP not in self.needed_locks
10285 assert self.op.iallocator is None or self.op.remote_node is None, \
10286 "Conflicting options"
10288 if self.op.remote_node is not None:
10289 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10291 # Warning: do not remove the locking of the new secondary here
10292 # unless DRBD8.AddChildren is changed to work in parallel;
10293 # currently it doesn't since parallel invocations of
10294 # FindUnusedMinor will conflict
10295 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
10296 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
10298 self.needed_locks[locking.LEVEL_NODE] = []
10299 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10301 if self.op.iallocator is not None:
10302 # iallocator will select a new node in the same group
10303 self.needed_locks[locking.LEVEL_NODEGROUP] = []
10305 self.needed_locks[locking.LEVEL_NODE_RES] = []
10307 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
10308 self.op.iallocator, self.op.remote_node,
10309 self.op.disks, False, self.op.early_release,
10310 self.op.ignore_ipolicy)
10312 self.tasklets = [self.replacer]
10314 def DeclareLocks(self, level):
10315 if level == locking.LEVEL_NODEGROUP:
10316 assert self.op.remote_node is None
10317 assert self.op.iallocator is not None
10318 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
10320 self.share_locks[locking.LEVEL_NODEGROUP] = 1
10321 # Lock all groups used by instance optimistically; this requires going
10322 # via the node before it's locked, requiring verification later on
10323 self.needed_locks[locking.LEVEL_NODEGROUP] = \
10324 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
10326 elif level == locking.LEVEL_NODE:
10327 if self.op.iallocator is not None:
10328 assert self.op.remote_node is None
10329 assert not self.needed_locks[locking.LEVEL_NODE]
10331 # Lock member nodes of all locked groups
10332 self.needed_locks[locking.LEVEL_NODE] = [node_name
10333 for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
10334 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
10336 self._LockInstancesNodes()
10337 elif level == locking.LEVEL_NODE_RES:
10339 self.needed_locks[locking.LEVEL_NODE_RES] = \
10340 self.needed_locks[locking.LEVEL_NODE]
10342 def BuildHooksEnv(self):
10343 """Build hooks env.
10345 This runs on the master, the primary and all the secondaries.
10348 instance = self.replacer.instance
10350 "MODE": self.op.mode,
10351 "NEW_SECONDARY": self.op.remote_node,
10352 "OLD_SECONDARY": instance.secondary_nodes[0],
10354 env.update(_BuildInstanceHookEnvByObject(self, instance))
10357 def BuildHooksNodes(self):
10358 """Build hooks nodes.
10361 instance = self.replacer.instance
10363 self.cfg.GetMasterNode(),
10364 instance.primary_node,
10366 if self.op.remote_node is not None:
10367 nl.append(self.op.remote_node)
10370 def CheckPrereq(self):
10371 """Check prerequisites.
10374 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
10375 self.op.iallocator is None)
10377 # Verify if node group locks are still correct
10378 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
10380 _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
10382 return LogicalUnit.CheckPrereq(self)
10385 class TLReplaceDisks(Tasklet):
10386 """Replaces disks for an instance.
10388 Note: Locking is not within the scope of this class.
10391 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
10392 disks, delay_iallocator, early_release, ignore_ipolicy):
10393 """Initializes this class.
10396 Tasklet.__init__(self, lu)
10399 self.instance_name = instance_name
10401 self.iallocator_name = iallocator_name
10402 self.remote_node = remote_node
10404 self.delay_iallocator = delay_iallocator
10405 self.early_release = early_release
10406 self.ignore_ipolicy = ignore_ipolicy
10409 self.instance = None
10410 self.new_node = None
10411 self.target_node = None
10412 self.other_node = None
10413 self.remote_node_info = None
10414 self.node_secondary_ip = None
10417 def CheckArguments(mode, remote_node, iallocator):
10418 """Helper function for users of this class.
10421 # check for valid parameter combination
10422 if mode == constants.REPLACE_DISK_CHG:
10423 if remote_node is None and iallocator is None:
10424 raise errors.OpPrereqError("When changing the secondary either an"
10425 " iallocator script must be used or the"
10426 " new node given", errors.ECODE_INVAL)
10428 if remote_node is not None and iallocator is not None:
10429 raise errors.OpPrereqError("Give either the iallocator or the new"
10430 " secondary, not both", errors.ECODE_INVAL)
10432 elif remote_node is not None or iallocator is not None:
10433 # Not replacing the secondary
10434 raise errors.OpPrereqError("The iallocator and new node options can"
10435 " only be used when changing the"
10436 " secondary node", errors.ECODE_INVAL)
10439 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
10440 """Compute a new secondary node using an IAllocator.
10443 ial = IAllocator(lu.cfg, lu.rpc,
10444 mode=constants.IALLOCATOR_MODE_RELOC,
10445 name=instance_name,
10446 relocate_from=list(relocate_from))
10448 ial.Run(iallocator_name)
10450 if not ial.success:
10451 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
10452 " %s" % (iallocator_name, ial.info),
10453 errors.ECODE_NORES)
10455 if len(ial.result) != ial.required_nodes:
10456 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
10457 " of nodes (%s), required %s" %
10459 len(ial.result), ial.required_nodes),
10460 errors.ECODE_FAULT)
10462 remote_node_name = ial.result[0]
10464 lu.LogInfo("Selected new secondary for instance '%s': %s",
10465 instance_name, remote_node_name)
10467 return remote_node_name
10469 def _FindFaultyDisks(self, node_name):
10470 """Wrapper for L{_FindFaultyInstanceDisks}.
10473 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
10476 def _CheckDisksActivated(self, instance):
10477 """Checks if the instance disks are activated.
10479 @param instance: The instance to check disks
10480 @return: True if they are activated, False otherwise
10483 nodes = instance.all_nodes
10485 for idx, dev in enumerate(instance.disks):
10487 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
10488 self.cfg.SetDiskID(dev, node)
10490 result = self.rpc.call_blockdev_find(node, dev)
10494 elif result.fail_msg or not result.payload:
10499 def CheckPrereq(self):
10500 """Check prerequisites.
10502 This checks that the instance is in the cluster.
10505 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
10506 assert instance is not None, \
10507 "Cannot retrieve locked instance %s" % self.instance_name
10509 if instance.disk_template != constants.DT_DRBD8:
10510 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
10511 " instances", errors.ECODE_INVAL)
10513 if len(instance.secondary_nodes) != 1:
10514 raise errors.OpPrereqError("The instance has a strange layout,"
10515 " expected one secondary but found %d" %
10516 len(instance.secondary_nodes),
10517 errors.ECODE_FAULT)
10519 if not self.delay_iallocator:
10520 self._CheckPrereq2()
10522 def _CheckPrereq2(self):
10523 """Check prerequisites, second part.
10525 This function should always be part of CheckPrereq. It was separated and is
10526 now called from Exec because during node evacuation iallocator was only
10527 called with an unmodified cluster model, not taking planned changes into
10531 instance = self.instance
10532 secondary_node = instance.secondary_nodes[0]
10534 if self.iallocator_name is None:
10535 remote_node = self.remote_node
10537 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
10538 instance.name, instance.secondary_nodes)
10540 if remote_node is None:
10541 self.remote_node_info = None
10543 assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \
10544 "Remote node '%s' is not locked" % remote_node
10546 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
10547 assert self.remote_node_info is not None, \
10548 "Cannot retrieve locked node %s" % remote_node
10550 if remote_node == self.instance.primary_node:
10551 raise errors.OpPrereqError("The specified node is the primary node of"
10552 " the instance", errors.ECODE_INVAL)
10554 if remote_node == secondary_node:
10555 raise errors.OpPrereqError("The specified node is already the"
10556 " secondary node of the instance",
10557 errors.ECODE_INVAL)
10559 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
10560 constants.REPLACE_DISK_CHG):
10561 raise errors.OpPrereqError("Cannot specify disks to be replaced",
10562 errors.ECODE_INVAL)
10564 if self.mode == constants.REPLACE_DISK_AUTO:
10565 if not self._CheckDisksActivated(instance):
10566 raise errors.OpPrereqError("Please run activate-disks on instance %s"
10567 " first" % self.instance_name,
10568 errors.ECODE_STATE)
10569 faulty_primary = self._FindFaultyDisks(instance.primary_node)
10570 faulty_secondary = self._FindFaultyDisks(secondary_node)
10572 if faulty_primary and faulty_secondary:
10573 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
10574 " one node and can not be repaired"
10575 " automatically" % self.instance_name,
10576 errors.ECODE_STATE)
10579 self.disks = faulty_primary
10580 self.target_node = instance.primary_node
10581 self.other_node = secondary_node
10582 check_nodes = [self.target_node, self.other_node]
10583 elif faulty_secondary:
10584 self.disks = faulty_secondary
10585 self.target_node = secondary_node
10586 self.other_node = instance.primary_node
10587 check_nodes = [self.target_node, self.other_node]
10593 # Non-automatic modes
10594 if self.mode == constants.REPLACE_DISK_PRI:
10595 self.target_node = instance.primary_node
10596 self.other_node = secondary_node
10597 check_nodes = [self.target_node, self.other_node]
10599 elif self.mode == constants.REPLACE_DISK_SEC:
10600 self.target_node = secondary_node
10601 self.other_node = instance.primary_node
10602 check_nodes = [self.target_node, self.other_node]
10604 elif self.mode == constants.REPLACE_DISK_CHG:
10605 self.new_node = remote_node
10606 self.other_node = instance.primary_node
10607 self.target_node = secondary_node
10608 check_nodes = [self.new_node, self.other_node]
10610 _CheckNodeNotDrained(self.lu, remote_node)
10611 _CheckNodeVmCapable(self.lu, remote_node)
10613 old_node_info = self.cfg.GetNodeInfo(secondary_node)
10614 assert old_node_info is not None
10615 if old_node_info.offline and not self.early_release:
10616 # doesn't make sense to delay the release
10617 self.early_release = True
10618 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
10619 " early-release mode", secondary_node)
10622 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
10625 # If not specified all disks should be replaced
10627 self.disks = range(len(self.instance.disks))
10629 # TODO: This is ugly, but right now we can't distinguish between internal
10630 # submitted opcode and external one. We should fix that.
10631 if self.remote_node_info:
10632 # We change the node, lets verify it still meets instance policy
10633 new_group_info = self.cfg.GetNodeGroup(self.remote_node_info.group)
10634 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(),
10636 _CheckTargetNodeIPolicy(self, ipolicy, instance, self.remote_node_info,
10637 ignore=self.ignore_ipolicy)
10639 # TODO: compute disk parameters
10640 primary_node_info = self.cfg.GetNodeInfo(instance.primary_node)
10641 secondary_node_info = self.cfg.GetNodeInfo(secondary_node)
10642 if primary_node_info.group != secondary_node_info.group:
10643 self.lu.LogInfo("The instance primary and secondary nodes are in two"
10644 " different node groups; the disk parameters of the"
10645 " primary node's group will be applied.")
10647 self.diskparams = self.cfg.GetNodeGroup(primary_node_info.group).diskparams
10649 for node in check_nodes:
10650 _CheckNodeOnline(self.lu, node)
10652 touched_nodes = frozenset(node_name for node_name in [self.new_node,
10655 if node_name is not None)
10657 # Release unneeded node and node resource locks
10658 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
10659 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES, keep=touched_nodes)
10661 # Release any owned node group
10662 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
10663 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
10665 # Check whether disks are valid
10666 for disk_idx in self.disks:
10667 instance.FindDisk(disk_idx)
10669 # Get secondary node IP addresses
10670 self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node)
10671 in self.cfg.GetMultiNodeInfo(touched_nodes))
10673 def Exec(self, feedback_fn):
10674 """Execute disk replacement.
10676 This dispatches the disk replacement to the appropriate handler.
10679 if self.delay_iallocator:
10680 self._CheckPrereq2()
10683 # Verify owned locks before starting operation
10684 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
10685 assert set(owned_nodes) == set(self.node_secondary_ip), \
10686 ("Incorrect node locks, owning %s, expected %s" %
10687 (owned_nodes, self.node_secondary_ip.keys()))
10688 assert (self.lu.owned_locks(locking.LEVEL_NODE) ==
10689 self.lu.owned_locks(locking.LEVEL_NODE_RES))
10691 owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
10692 assert list(owned_instances) == [self.instance_name], \
10693 "Instance '%s' not locked" % self.instance_name
10695 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
10696 "Should not own any node group lock at this point"
10699 feedback_fn("No disks need replacement")
10702 feedback_fn("Replacing disk(s) %s for %s" %
10703 (utils.CommaJoin(self.disks), self.instance.name))
10705 activate_disks = (self.instance.admin_state != constants.ADMINST_UP)
10707 # Activate the instance disks if we're replacing them on a down instance
10709 _StartInstanceDisks(self.lu, self.instance, True)
10712 # Should we replace the secondary node?
10713 if self.new_node is not None:
10714 fn = self._ExecDrbd8Secondary
10716 fn = self._ExecDrbd8DiskOnly
10718 result = fn(feedback_fn)
10720 # Deactivate the instance disks if we're replacing them on a
10723 _SafeShutdownInstanceDisks(self.lu, self.instance)
10725 assert not self.lu.owned_locks(locking.LEVEL_NODE)
10728 # Verify owned locks
10729 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE_RES)
10730 nodes = frozenset(self.node_secondary_ip)
10731 assert ((self.early_release and not owned_nodes) or
10732 (not self.early_release and not (set(owned_nodes) - nodes))), \
10733 ("Not owning the correct locks, early_release=%s, owned=%r,"
10734 " nodes=%r" % (self.early_release, owned_nodes, nodes))
10738 def _CheckVolumeGroup(self, nodes):
10739 self.lu.LogInfo("Checking volume groups")
10741 vgname = self.cfg.GetVGName()
10743 # Make sure volume group exists on all involved nodes
10744 results = self.rpc.call_vg_list(nodes)
10746 raise errors.OpExecError("Can't list volume groups on the nodes")
10749 res = results[node]
10750 res.Raise("Error checking node %s" % node)
10751 if vgname not in res.payload:
10752 raise errors.OpExecError("Volume group '%s' not found on node %s" %
10755 def _CheckDisksExistence(self, nodes):
10756 # Check disk existence
10757 for idx, dev in enumerate(self.instance.disks):
10758 if idx not in self.disks:
10762 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
10763 self.cfg.SetDiskID(dev, node)
10765 result = self.rpc.call_blockdev_find(node, dev)
10767 msg = result.fail_msg
10768 if msg or not result.payload:
10770 msg = "disk not found"
10771 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
10774 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
10775 for idx, dev in enumerate(self.instance.disks):
10776 if idx not in self.disks:
10779 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
10782 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
10784 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
10785 " replace disks for instance %s" %
10786 (node_name, self.instance.name))
10788 def _CreateNewStorage(self, node_name):
10789 """Create new storage on the primary or secondary node.
10791 This is only used for same-node replaces, not for changing the
10792 secondary node, hence we don't want to modify the existing disk.
10797 for idx, dev in enumerate(self.instance.disks):
10798 if idx not in self.disks:
10801 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
10803 self.cfg.SetDiskID(dev, node_name)
10805 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
10806 names = _GenerateUniqueNames(self.lu, lv_names)
10808 _, data_p, meta_p = _ComputeLDParams(constants.DT_DRBD8, self.diskparams)
10810 vg_data = dev.children[0].logical_id[0]
10811 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
10812 logical_id=(vg_data, names[0]), params=data_p)
10813 vg_meta = dev.children[1].logical_id[0]
10814 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,
10815 logical_id=(vg_meta, names[1]), params=meta_p)
10817 new_lvs = [lv_data, lv_meta]
10818 old_lvs = [child.Copy() for child in dev.children]
10819 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
10821 # we pass force_create=True to force the LVM creation
10822 for new_lv in new_lvs:
10823 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
10824 _GetInstanceInfoText(self.instance), False)
10828 def _CheckDevices(self, node_name, iv_names):
10829 for name, (dev, _, _) in iv_names.iteritems():
10830 self.cfg.SetDiskID(dev, node_name)
10832 result = self.rpc.call_blockdev_find(node_name, dev)
10834 msg = result.fail_msg
10835 if msg or not result.payload:
10837 msg = "disk not found"
10838 raise errors.OpExecError("Can't find DRBD device %s: %s" %
10841 if result.payload.is_degraded:
10842 raise errors.OpExecError("DRBD device %s is degraded!" % name)
10844 def _RemoveOldStorage(self, node_name, iv_names):
10845 for name, (_, old_lvs, _) in iv_names.iteritems():
10846 self.lu.LogInfo("Remove logical volumes for %s" % name)
10849 self.cfg.SetDiskID(lv, node_name)
10851 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
10853 self.lu.LogWarning("Can't remove old LV: %s" % msg,
10854 hint="remove unused LVs manually")
10856 def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
10857 """Replace a disk on the primary or secondary for DRBD 8.
10859 The algorithm for replace is quite complicated:
10861 1. for each disk to be replaced:
10863 1. create new LVs on the target node with unique names
10864 1. detach old LVs from the drbd device
10865 1. rename old LVs to name_replaced.<time_t>
10866 1. rename new LVs to old LVs
10867 1. attach the new LVs (with the old names now) to the drbd device
10869 1. wait for sync across all devices
10871 1. for each modified disk:
10873 1. remove old LVs (which have the name name_replaces.<time_t>)
10875 Failures are not very well handled.
10880 # Step: check device activation
10881 self.lu.LogStep(1, steps_total, "Check device existence")
10882 self._CheckDisksExistence([self.other_node, self.target_node])
10883 self._CheckVolumeGroup([self.target_node, self.other_node])
10885 # Step: check other node consistency
10886 self.lu.LogStep(2, steps_total, "Check peer consistency")
10887 self._CheckDisksConsistency(self.other_node,
10888 self.other_node == self.instance.primary_node,
10891 # Step: create new storage
10892 self.lu.LogStep(3, steps_total, "Allocate new storage")
10893 iv_names = self._CreateNewStorage(self.target_node)
10895 # Step: for each lv, detach+rename*2+attach
10896 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
10897 for dev, old_lvs, new_lvs in iv_names.itervalues():
10898 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
10900 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
10902 result.Raise("Can't detach drbd from local storage on node"
10903 " %s for device %s" % (self.target_node, dev.iv_name))
10905 #cfg.Update(instance)
10907 # ok, we created the new LVs, so now we know we have the needed
10908 # storage; as such, we proceed on the target node to rename
10909 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
10910 # using the assumption that logical_id == physical_id (which in
10911 # turn is the unique_id on that node)
10913 # FIXME(iustin): use a better name for the replaced LVs
10914 temp_suffix = int(time.time())
10915 ren_fn = lambda d, suff: (d.physical_id[0],
10916 d.physical_id[1] + "_replaced-%s" % suff)
10918 # Build the rename list based on what LVs exist on the node
10919 rename_old_to_new = []
10920 for to_ren in old_lvs:
10921 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
10922 if not result.fail_msg and result.payload:
10924 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
10926 self.lu.LogInfo("Renaming the old LVs on the target node")
10927 result = self.rpc.call_blockdev_rename(self.target_node,
10929 result.Raise("Can't rename old LVs on node %s" % self.target_node)
10931 # Now we rename the new LVs to the old LVs
10932 self.lu.LogInfo("Renaming the new LVs on the target node")
10933 rename_new_to_old = [(new, old.physical_id)
10934 for old, new in zip(old_lvs, new_lvs)]
10935 result = self.rpc.call_blockdev_rename(self.target_node,
10937 result.Raise("Can't rename new LVs on node %s" % self.target_node)
10939 # Intermediate steps of in memory modifications
10940 for old, new in zip(old_lvs, new_lvs):
10941 new.logical_id = old.logical_id
10942 self.cfg.SetDiskID(new, self.target_node)
10944 # We need to modify old_lvs so that removal later removes the
10945 # right LVs, not the newly added ones; note that old_lvs is a
10947 for disk in old_lvs:
10948 disk.logical_id = ren_fn(disk, temp_suffix)
10949 self.cfg.SetDiskID(disk, self.target_node)
10951 # Now that the new lvs have the old name, we can add them to the device
10952 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
10953 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
10955 msg = result.fail_msg
10957 for new_lv in new_lvs:
10958 msg2 = self.rpc.call_blockdev_remove(self.target_node,
10961 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
10962 hint=("cleanup manually the unused logical"
10964 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
10966 cstep = itertools.count(5)
10968 if self.early_release:
10969 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10970 self._RemoveOldStorage(self.target_node, iv_names)
10971 # TODO: Check if releasing locks early still makes sense
10972 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
10974 # Release all resource locks except those used by the instance
10975 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
10976 keep=self.node_secondary_ip.keys())
10978 # Release all node locks while waiting for sync
10979 _ReleaseLocks(self.lu, locking.LEVEL_NODE)
10981 # TODO: Can the instance lock be downgraded here? Take the optional disk
10982 # shutdown in the caller into consideration.
10985 # This can fail as the old devices are degraded and _WaitForSync
10986 # does a combined result over all disks, so we don't check its return value
10987 self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
10988 _WaitForSync(self.lu, self.instance)
10990 # Check all devices manually
10991 self._CheckDevices(self.instance.primary_node, iv_names)
10993 # Step: remove old storage
10994 if not self.early_release:
10995 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10996 self._RemoveOldStorage(self.target_node, iv_names)
10998 def _ExecDrbd8Secondary(self, feedback_fn):
10999 """Replace the secondary node for DRBD 8.
11001 The algorithm for replace is quite complicated:
11002 - for all disks of the instance:
11003 - create new LVs on the new node with same names
11004 - shutdown the drbd device on the old secondary
11005 - disconnect the drbd network on the primary
11006 - create the drbd device on the new secondary
11007 - network attach the drbd on the primary, using an artifice:
11008 the drbd code for Attach() will connect to the network if it
11009 finds a device which is connected to the good local disks but
11010 not network enabled
11011 - wait for sync across all devices
11012 - remove all disks from the old secondary
11014 Failures are not very well handled.
11019 pnode = self.instance.primary_node
11021 # Step: check device activation
11022 self.lu.LogStep(1, steps_total, "Check device existence")
11023 self._CheckDisksExistence([self.instance.primary_node])
11024 self._CheckVolumeGroup([self.instance.primary_node])
11026 # Step: check other node consistency
11027 self.lu.LogStep(2, steps_total, "Check peer consistency")
11028 self._CheckDisksConsistency(self.instance.primary_node, True, True)
11030 # Step: create new storage
11031 self.lu.LogStep(3, steps_total, "Allocate new storage")
11032 for idx, dev in enumerate(self.instance.disks):
11033 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
11034 (self.new_node, idx))
11035 # we pass force_create=True to force LVM creation
11036 for new_lv in dev.children:
11037 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
11038 _GetInstanceInfoText(self.instance), False)
11040 # Step 4: dbrd minors and drbd setups changes
11041 # after this, we must manually remove the drbd minors on both the
11042 # error and the success paths
11043 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
11044 minors = self.cfg.AllocateDRBDMinor([self.new_node
11045 for dev in self.instance.disks],
11046 self.instance.name)
11047 logging.debug("Allocated minors %r", minors)
11050 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
11051 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
11052 (self.new_node, idx))
11053 # create new devices on new_node; note that we create two IDs:
11054 # one without port, so the drbd will be activated without
11055 # networking information on the new node at this stage, and one
11056 # with network, for the latter activation in step 4
11057 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
11058 if self.instance.primary_node == o_node1:
11061 assert self.instance.primary_node == o_node2, "Three-node instance?"
11064 new_alone_id = (self.instance.primary_node, self.new_node, None,
11065 p_minor, new_minor, o_secret)
11066 new_net_id = (self.instance.primary_node, self.new_node, o_port,
11067 p_minor, new_minor, o_secret)
11069 iv_names[idx] = (dev, dev.children, new_net_id)
11070 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
11072 drbd_params, _, _ = _ComputeLDParams(constants.DT_DRBD8, self.diskparams)
11073 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
11074 logical_id=new_alone_id,
11075 children=dev.children,
11077 params=drbd_params)
11079 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
11080 _GetInstanceInfoText(self.instance), False)
11081 except errors.GenericError:
11082 self.cfg.ReleaseDRBDMinors(self.instance.name)
11085 # We have new devices, shutdown the drbd on the old secondary
11086 for idx, dev in enumerate(self.instance.disks):
11087 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
11088 self.cfg.SetDiskID(dev, self.target_node)
11089 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
11091 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
11092 "node: %s" % (idx, msg),
11093 hint=("Please cleanup this device manually as"
11094 " soon as possible"))
11096 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
11097 result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
11098 self.instance.disks)[pnode]
11100 msg = result.fail_msg
11102 # detaches didn't succeed (unlikely)
11103 self.cfg.ReleaseDRBDMinors(self.instance.name)
11104 raise errors.OpExecError("Can't detach the disks from the network on"
11105 " old node: %s" % (msg,))
11107 # if we managed to detach at least one, we update all the disks of
11108 # the instance to point to the new secondary
11109 self.lu.LogInfo("Updating instance configuration")
11110 for dev, _, new_logical_id in iv_names.itervalues():
11111 dev.logical_id = new_logical_id
11112 self.cfg.SetDiskID(dev, self.instance.primary_node)
11114 self.cfg.Update(self.instance, feedback_fn)
11116 # Release all node locks (the configuration has been updated)
11117 _ReleaseLocks(self.lu, locking.LEVEL_NODE)
11119 # and now perform the drbd attach
11120 self.lu.LogInfo("Attaching primary drbds to new secondary"
11121 " (standalone => connected)")
11122 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
11124 self.node_secondary_ip,
11125 self.instance.disks,
11126 self.instance.name,
11128 for to_node, to_result in result.items():
11129 msg = to_result.fail_msg
11131 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
11133 hint=("please do a gnt-instance info to see the"
11134 " status of disks"))
11136 cstep = itertools.count(5)
11138 if self.early_release:
11139 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
11140 self._RemoveOldStorage(self.target_node, iv_names)
11141 # TODO: Check if releasing locks early still makes sense
11142 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
11144 # Release all resource locks except those used by the instance
11145 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
11146 keep=self.node_secondary_ip.keys())
11148 # TODO: Can the instance lock be downgraded here? Take the optional disk
11149 # shutdown in the caller into consideration.
11152 # This can fail as the old devices are degraded and _WaitForSync
11153 # does a combined result over all disks, so we don't check its return value
11154 self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
11155 _WaitForSync(self.lu, self.instance)
11157 # Check all devices manually
11158 self._CheckDevices(self.instance.primary_node, iv_names)
11160 # Step: remove old storage
11161 if not self.early_release:
11162 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
11163 self._RemoveOldStorage(self.target_node, iv_names)
11166 class LURepairNodeStorage(NoHooksLU):
11167 """Repairs the volume group on a node.
11172 def CheckArguments(self):
11173 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
11175 storage_type = self.op.storage_type
11177 if (constants.SO_FIX_CONSISTENCY not in
11178 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
11179 raise errors.OpPrereqError("Storage units of type '%s' can not be"
11180 " repaired" % storage_type,
11181 errors.ECODE_INVAL)
11183 def ExpandNames(self):
11184 self.needed_locks = {
11185 locking.LEVEL_NODE: [self.op.node_name],
11188 def _CheckFaultyDisks(self, instance, node_name):
11189 """Ensure faulty disks abort the opcode or at least warn."""
11191 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
11193 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
11194 " node '%s'" % (instance.name, node_name),
11195 errors.ECODE_STATE)
11196 except errors.OpPrereqError, err:
11197 if self.op.ignore_consistency:
11198 self.proc.LogWarning(str(err.args[0]))
11202 def CheckPrereq(self):
11203 """Check prerequisites.
11206 # Check whether any instance on this node has faulty disks
11207 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
11208 if inst.admin_state != constants.ADMINST_UP:
11210 check_nodes = set(inst.all_nodes)
11211 check_nodes.discard(self.op.node_name)
11212 for inst_node_name in check_nodes:
11213 self._CheckFaultyDisks(inst, inst_node_name)
11215 def Exec(self, feedback_fn):
11216 feedback_fn("Repairing storage unit '%s' on %s ..." %
11217 (self.op.name, self.op.node_name))
11219 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
11220 result = self.rpc.call_storage_execute(self.op.node_name,
11221 self.op.storage_type, st_args,
11223 constants.SO_FIX_CONSISTENCY)
11224 result.Raise("Failed to repair storage unit '%s' on %s" %
11225 (self.op.name, self.op.node_name))
11228 class LUNodeEvacuate(NoHooksLU):
11229 """Evacuates instances off a list of nodes.
11234 _MODE2IALLOCATOR = {
11235 constants.NODE_EVAC_PRI: constants.IALLOCATOR_NEVAC_PRI,
11236 constants.NODE_EVAC_SEC: constants.IALLOCATOR_NEVAC_SEC,
11237 constants.NODE_EVAC_ALL: constants.IALLOCATOR_NEVAC_ALL,
11239 assert frozenset(_MODE2IALLOCATOR.keys()) == constants.NODE_EVAC_MODES
11240 assert (frozenset(_MODE2IALLOCATOR.values()) ==
11241 constants.IALLOCATOR_NEVAC_MODES)
11243 def CheckArguments(self):
11244 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
11246 def ExpandNames(self):
11247 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
11249 if self.op.remote_node is not None:
11250 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
11251 assert self.op.remote_node
11253 if self.op.remote_node == self.op.node_name:
11254 raise errors.OpPrereqError("Can not use evacuated node as a new"
11255 " secondary node", errors.ECODE_INVAL)
11257 if self.op.mode != constants.NODE_EVAC_SEC:
11258 raise errors.OpPrereqError("Without the use of an iallocator only"
11259 " secondary instances can be evacuated",
11260 errors.ECODE_INVAL)
11263 self.share_locks = _ShareAll()
11264 self.needed_locks = {
11265 locking.LEVEL_INSTANCE: [],
11266 locking.LEVEL_NODEGROUP: [],
11267 locking.LEVEL_NODE: [],
11270 # Determine nodes (via group) optimistically, needs verification once locks
11271 # have been acquired
11272 self.lock_nodes = self._DetermineNodes()
11274 def _DetermineNodes(self):
11275 """Gets the list of nodes to operate on.
11278 if self.op.remote_node is None:
11279 # Iallocator will choose any node(s) in the same group
11280 group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name])
11282 group_nodes = frozenset([self.op.remote_node])
11284 # Determine nodes to be locked
11285 return set([self.op.node_name]) | group_nodes
11287 def _DetermineInstances(self):
11288 """Builds list of instances to operate on.
11291 assert self.op.mode in constants.NODE_EVAC_MODES
11293 if self.op.mode == constants.NODE_EVAC_PRI:
11294 # Primary instances only
11295 inst_fn = _GetNodePrimaryInstances
11296 assert self.op.remote_node is None, \
11297 "Evacuating primary instances requires iallocator"
11298 elif self.op.mode == constants.NODE_EVAC_SEC:
11299 # Secondary instances only
11300 inst_fn = _GetNodeSecondaryInstances
11303 assert self.op.mode == constants.NODE_EVAC_ALL
11304 inst_fn = _GetNodeInstances
11305 # TODO: In 2.6, change the iallocator interface to take an evacuation mode
11307 raise errors.OpPrereqError("Due to an issue with the iallocator"
11308 " interface it is not possible to evacuate"
11309 " all instances at once; specify explicitly"
11310 " whether to evacuate primary or secondary"
11312 errors.ECODE_INVAL)
11314 return inst_fn(self.cfg, self.op.node_name)
11316 def DeclareLocks(self, level):
11317 if level == locking.LEVEL_INSTANCE:
11318 # Lock instances optimistically, needs verification once node and group
11319 # locks have been acquired
11320 self.needed_locks[locking.LEVEL_INSTANCE] = \
11321 set(i.name for i in self._DetermineInstances())
11323 elif level == locking.LEVEL_NODEGROUP:
11324 # Lock node groups for all potential target nodes optimistically, needs
11325 # verification once nodes have been acquired
11326 self.needed_locks[locking.LEVEL_NODEGROUP] = \
11327 self.cfg.GetNodeGroupsFromNodes(self.lock_nodes)
11329 elif level == locking.LEVEL_NODE:
11330 self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes
11332 def CheckPrereq(self):
11334 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
11335 owned_nodes = self.owned_locks(locking.LEVEL_NODE)
11336 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
11338 need_nodes = self._DetermineNodes()
11340 if not owned_nodes.issuperset(need_nodes):
11341 raise errors.OpPrereqError("Nodes in same group as '%s' changed since"
11342 " locks were acquired, current nodes are"
11343 " are '%s', used to be '%s'; retry the"
11345 (self.op.node_name,
11346 utils.CommaJoin(need_nodes),
11347 utils.CommaJoin(owned_nodes)),
11348 errors.ECODE_STATE)
11350 wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
11351 if owned_groups != wanted_groups:
11352 raise errors.OpExecError("Node groups changed since locks were acquired,"
11353 " current groups are '%s', used to be '%s';"
11354 " retry the operation" %
11355 (utils.CommaJoin(wanted_groups),
11356 utils.CommaJoin(owned_groups)))
11358 # Determine affected instances
11359 self.instances = self._DetermineInstances()
11360 self.instance_names = [i.name for i in self.instances]
11362 if set(self.instance_names) != owned_instances:
11363 raise errors.OpExecError("Instances on node '%s' changed since locks"
11364 " were acquired, current instances are '%s',"
11365 " used to be '%s'; retry the operation" %
11366 (self.op.node_name,
11367 utils.CommaJoin(self.instance_names),
11368 utils.CommaJoin(owned_instances)))
11370 if self.instance_names:
11371 self.LogInfo("Evacuating instances from node '%s': %s",
11373 utils.CommaJoin(utils.NiceSort(self.instance_names)))
11375 self.LogInfo("No instances to evacuate from node '%s'",
11378 if self.op.remote_node is not None:
11379 for i in self.instances:
11380 if i.primary_node == self.op.remote_node:
11381 raise errors.OpPrereqError("Node %s is the primary node of"
11382 " instance %s, cannot use it as"
11384 (self.op.remote_node, i.name),
11385 errors.ECODE_INVAL)
11387 def Exec(self, feedback_fn):
11388 assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None)
11390 if not self.instance_names:
11391 # No instances to evacuate
11394 elif self.op.iallocator is not None:
11395 # TODO: Implement relocation to other group
11396 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_NODE_EVAC,
11397 evac_mode=self._MODE2IALLOCATOR[self.op.mode],
11398 instances=list(self.instance_names))
11400 ial.Run(self.op.iallocator)
11402 if not ial.success:
11403 raise errors.OpPrereqError("Can't compute node evacuation using"
11404 " iallocator '%s': %s" %
11405 (self.op.iallocator, ial.info),
11406 errors.ECODE_NORES)
11408 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True)
11410 elif self.op.remote_node is not None:
11411 assert self.op.mode == constants.NODE_EVAC_SEC
11413 [opcodes.OpInstanceReplaceDisks(instance_name=instance_name,
11414 remote_node=self.op.remote_node,
11416 mode=constants.REPLACE_DISK_CHG,
11417 early_release=self.op.early_release)]
11418 for instance_name in self.instance_names
11422 raise errors.ProgrammerError("No iallocator or remote node")
11424 return ResultWithJobs(jobs)
11427 def _SetOpEarlyRelease(early_release, op):
11428 """Sets C{early_release} flag on opcodes if available.
11432 op.early_release = early_release
11433 except AttributeError:
11434 assert not isinstance(op, opcodes.OpInstanceReplaceDisks)
11439 def _NodeEvacDest(use_nodes, group, nodes):
11440 """Returns group or nodes depending on caller's choice.
11444 return utils.CommaJoin(nodes)
11449 def _LoadNodeEvacResult(lu, alloc_result, early_release, use_nodes):
11450 """Unpacks the result of change-group and node-evacuate iallocator requests.
11452 Iallocator modes L{constants.IALLOCATOR_MODE_NODE_EVAC} and
11453 L{constants.IALLOCATOR_MODE_CHG_GROUP}.
11455 @type lu: L{LogicalUnit}
11456 @param lu: Logical unit instance
11457 @type alloc_result: tuple/list
11458 @param alloc_result: Result from iallocator
11459 @type early_release: bool
11460 @param early_release: Whether to release locks early if possible
11461 @type use_nodes: bool
11462 @param use_nodes: Whether to display node names instead of groups
11465 (moved, failed, jobs) = alloc_result
11468 failreason = utils.CommaJoin("%s (%s)" % (name, reason)
11469 for (name, reason) in failed)
11470 lu.LogWarning("Unable to evacuate instances %s", failreason)
11471 raise errors.OpExecError("Unable to evacuate instances %s" % failreason)
11474 lu.LogInfo("Instances to be moved: %s",
11475 utils.CommaJoin("%s (to %s)" %
11476 (name, _NodeEvacDest(use_nodes, group, nodes))
11477 for (name, group, nodes) in moved))
11479 return [map(compat.partial(_SetOpEarlyRelease, early_release),
11480 map(opcodes.OpCode.LoadOpCode, ops))
11484 class LUInstanceGrowDisk(LogicalUnit):
11485 """Grow a disk of an instance.
11488 HPATH = "disk-grow"
11489 HTYPE = constants.HTYPE_INSTANCE
11492 def ExpandNames(self):
11493 self._ExpandAndLockInstance()
11494 self.needed_locks[locking.LEVEL_NODE] = []
11495 self.needed_locks[locking.LEVEL_NODE_RES] = []
11496 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11497 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
11499 def DeclareLocks(self, level):
11500 if level == locking.LEVEL_NODE:
11501 self._LockInstancesNodes()
11502 elif level == locking.LEVEL_NODE_RES:
11504 self.needed_locks[locking.LEVEL_NODE_RES] = \
11505 self.needed_locks[locking.LEVEL_NODE][:]
11507 def BuildHooksEnv(self):
11508 """Build hooks env.
11510 This runs on the master, the primary and all the secondaries.
11514 "DISK": self.op.disk,
11515 "AMOUNT": self.op.amount,
11517 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11520 def BuildHooksNodes(self):
11521 """Build hooks nodes.
11524 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
11527 def CheckPrereq(self):
11528 """Check prerequisites.
11530 This checks that the instance is in the cluster.
11533 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
11534 assert instance is not None, \
11535 "Cannot retrieve locked instance %s" % self.op.instance_name
11536 nodenames = list(instance.all_nodes)
11537 for node in nodenames:
11538 _CheckNodeOnline(self, node)
11540 self.instance = instance
11542 if instance.disk_template not in constants.DTS_GROWABLE:
11543 raise errors.OpPrereqError("Instance's disk layout does not support"
11544 " growing", errors.ECODE_INVAL)
11546 self.disk = instance.FindDisk(self.op.disk)
11548 if instance.disk_template not in (constants.DT_FILE,
11549 constants.DT_SHARED_FILE,
11551 # TODO: check the free disk space for file, when that feature will be
11553 _CheckNodesFreeDiskPerVG(self, nodenames,
11554 self.disk.ComputeGrowth(self.op.amount))
11556 def Exec(self, feedback_fn):
11557 """Execute disk grow.
11560 instance = self.instance
11563 assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
11564 assert (self.owned_locks(locking.LEVEL_NODE) ==
11565 self.owned_locks(locking.LEVEL_NODE_RES))
11567 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
11569 raise errors.OpExecError("Cannot activate block device to grow")
11571 feedback_fn("Growing disk %s of instance '%s' by %s" %
11572 (self.op.disk, instance.name,
11573 utils.FormatUnit(self.op.amount, "h")))
11575 # First run all grow ops in dry-run mode
11576 for node in instance.all_nodes:
11577 self.cfg.SetDiskID(disk, node)
11578 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, True)
11579 result.Raise("Grow request failed to node %s" % node)
11581 # We know that (as far as we can test) operations across different
11582 # nodes will succeed, time to run it for real
11583 for node in instance.all_nodes:
11584 self.cfg.SetDiskID(disk, node)
11585 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, False)
11586 result.Raise("Grow request failed to node %s" % node)
11588 # TODO: Rewrite code to work properly
11589 # DRBD goes into sync mode for a short amount of time after executing the
11590 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
11591 # calling "resize" in sync mode fails. Sleeping for a short amount of
11592 # time is a work-around.
11595 disk.RecordGrow(self.op.amount)
11596 self.cfg.Update(instance, feedback_fn)
11598 # Changes have been recorded, release node lock
11599 _ReleaseLocks(self, locking.LEVEL_NODE)
11601 # Downgrade lock while waiting for sync
11602 self.glm.downgrade(locking.LEVEL_INSTANCE)
11604 if self.op.wait_for_sync:
11605 disk_abort = not _WaitForSync(self, instance, disks=[disk])
11607 self.proc.LogWarning("Disk sync-ing has not returned a good"
11608 " status; please check the instance")
11609 if instance.admin_state != constants.ADMINST_UP:
11610 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
11611 elif instance.admin_state != constants.ADMINST_UP:
11612 self.proc.LogWarning("Not shutting down the disk even if the instance is"
11613 " not supposed to be running because no wait for"
11614 " sync mode was requested")
11616 assert self.owned_locks(locking.LEVEL_NODE_RES)
11617 assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
11620 class LUInstanceQueryData(NoHooksLU):
11621 """Query runtime instance data.
11626 def ExpandNames(self):
11627 self.needed_locks = {}
11629 # Use locking if requested or when non-static information is wanted
11630 if not (self.op.static or self.op.use_locking):
11631 self.LogWarning("Non-static data requested, locks need to be acquired")
11632 self.op.use_locking = True
11634 if self.op.instances or not self.op.use_locking:
11635 # Expand instance names right here
11636 self.wanted_names = _GetWantedInstances(self, self.op.instances)
11638 # Will use acquired locks
11639 self.wanted_names = None
11641 if self.op.use_locking:
11642 self.share_locks = _ShareAll()
11644 if self.wanted_names is None:
11645 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
11647 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
11649 self.needed_locks[locking.LEVEL_NODE] = []
11650 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11652 def DeclareLocks(self, level):
11653 if self.op.use_locking and level == locking.LEVEL_NODE:
11654 self._LockInstancesNodes()
11656 def CheckPrereq(self):
11657 """Check prerequisites.
11659 This only checks the optional instance list against the existing names.
11662 if self.wanted_names is None:
11663 assert self.op.use_locking, "Locking was not used"
11664 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
11666 self.wanted_instances = \
11667 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
11669 def _ComputeBlockdevStatus(self, node, instance_name, dev):
11670 """Returns the status of a block device
11673 if self.op.static or not node:
11676 self.cfg.SetDiskID(dev, node)
11678 result = self.rpc.call_blockdev_find(node, dev)
11682 result.Raise("Can't compute disk status for %s" % instance_name)
11684 status = result.payload
11688 return (status.dev_path, status.major, status.minor,
11689 status.sync_percent, status.estimated_time,
11690 status.is_degraded, status.ldisk_status)
11692 def _ComputeDiskStatus(self, instance, snode, dev):
11693 """Compute block device status.
11696 if dev.dev_type in constants.LDS_DRBD:
11697 # we change the snode then (otherwise we use the one passed in)
11698 if dev.logical_id[0] == instance.primary_node:
11699 snode = dev.logical_id[1]
11701 snode = dev.logical_id[0]
11703 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
11704 instance.name, dev)
11705 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
11708 dev_children = map(compat.partial(self._ComputeDiskStatus,
11715 "iv_name": dev.iv_name,
11716 "dev_type": dev.dev_type,
11717 "logical_id": dev.logical_id,
11718 "physical_id": dev.physical_id,
11719 "pstatus": dev_pstatus,
11720 "sstatus": dev_sstatus,
11721 "children": dev_children,
11726 def Exec(self, feedback_fn):
11727 """Gather and return data"""
11730 cluster = self.cfg.GetClusterInfo()
11732 pri_nodes = self.cfg.GetMultiNodeInfo(i.primary_node
11733 for i in self.wanted_instances)
11734 for instance, (_, pnode) in zip(self.wanted_instances, pri_nodes):
11735 if self.op.static or pnode.offline:
11736 remote_state = None
11738 self.LogWarning("Primary node %s is marked offline, returning static"
11739 " information only for instance %s" %
11740 (pnode.name, instance.name))
11742 remote_info = self.rpc.call_instance_info(instance.primary_node,
11744 instance.hypervisor)
11745 remote_info.Raise("Error checking node %s" % instance.primary_node)
11746 remote_info = remote_info.payload
11747 if remote_info and "state" in remote_info:
11748 remote_state = "up"
11750 if instance.admin_state == constants.ADMINST_UP:
11751 remote_state = "down"
11753 remote_state = instance.admin_state
11755 disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
11758 result[instance.name] = {
11759 "name": instance.name,
11760 "config_state": instance.admin_state,
11761 "run_state": remote_state,
11762 "pnode": instance.primary_node,
11763 "snodes": instance.secondary_nodes,
11765 # this happens to be the same format used for hooks
11766 "nics": _NICListToTuple(self, instance.nics),
11767 "disk_template": instance.disk_template,
11769 "hypervisor": instance.hypervisor,
11770 "network_port": instance.network_port,
11771 "hv_instance": instance.hvparams,
11772 "hv_actual": cluster.FillHV(instance, skip_globals=True),
11773 "be_instance": instance.beparams,
11774 "be_actual": cluster.FillBE(instance),
11775 "os_instance": instance.osparams,
11776 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
11777 "serial_no": instance.serial_no,
11778 "mtime": instance.mtime,
11779 "ctime": instance.ctime,
11780 "uuid": instance.uuid,
11786 def PrepareContainerMods(mods, private_fn):
11787 """Prepares a list of container modifications by adding a private data field.
11789 @type mods: list of tuples; (operation, index, parameters)
11790 @param mods: List of modifications
11791 @type private_fn: callable or None
11792 @param private_fn: Callable for constructing a private data field for a
11797 if private_fn is None:
11802 return [(op, idx, params, fn()) for (op, idx, params) in mods]
11805 #: Type description for changes as returned by L{ApplyContainerMods}'s
11807 _TApplyContModsCbChanges = \
11808 ht.TMaybeListOf(ht.TAnd(ht.TIsLength(2), ht.TItems([
11809 ht.TNonEmptyString,
11814 def ApplyContainerMods(kind, container, chgdesc, mods,
11815 create_fn, modify_fn, remove_fn):
11816 """Applies descriptions in C{mods} to C{container}.
11819 @param kind: One-word item description
11820 @type container: list
11821 @param container: Container to modify
11822 @type chgdesc: None or list
11823 @param chgdesc: List of applied changes
11825 @param mods: Modifications as returned by L{PrepareContainerMods}
11826 @type create_fn: callable
11827 @param create_fn: Callback for creating a new item (L{constants.DDM_ADD});
11828 receives absolute item index, parameters and private data object as added
11829 by L{PrepareContainerMods}, returns tuple containing new item and changes
11831 @type modify_fn: callable
11832 @param modify_fn: Callback for modifying an existing item
11833 (L{constants.DDM_MODIFY}); receives absolute item index, item, parameters
11834 and private data object as added by L{PrepareContainerMods}, returns
11836 @type remove_fn: callable
11837 @param remove_fn: Callback on removing item; receives absolute item index,
11838 item and private data object as added by L{PrepareContainerMods}
11841 for (op, idx, params, private) in mods:
11844 absidx = len(container) - 1
11846 raise IndexError("Not accepting negative indices other than -1")
11847 elif idx > len(container):
11848 raise IndexError("Got %s index %s, but there are only %s" %
11849 (kind, idx, len(container)))
11855 if op == constants.DDM_ADD:
11856 # Calculate where item will be added
11858 addidx = len(container)
11862 if create_fn is None:
11865 (item, changes) = create_fn(addidx, params, private)
11868 container.append(item)
11871 assert idx <= len(container)
11872 # list.insert does so before the specified index
11873 container.insert(idx, item)
11875 # Retrieve existing item
11877 item = container[absidx]
11879 raise IndexError("Invalid %s index %s" % (kind, idx))
11881 if op == constants.DDM_REMOVE:
11884 if remove_fn is not None:
11885 remove_fn(absidx, item, private)
11887 changes = [("%s/%s" % (kind, absidx), "remove")]
11889 assert container[absidx] == item
11890 del container[absidx]
11891 elif op == constants.DDM_MODIFY:
11892 if modify_fn is not None:
11893 changes = modify_fn(absidx, item, params, private)
11895 raise errors.ProgrammerError("Unhandled operation '%s'" % op)
11897 assert _TApplyContModsCbChanges(changes)
11899 if not (chgdesc is None or changes is None):
11900 chgdesc.extend(changes)
11903 def _UpdateIvNames(base_index, disks):
11904 """Updates the C{iv_name} attribute of disks.
11906 @type disks: list of L{objects.Disk}
11909 for (idx, disk) in enumerate(disks):
11910 disk.iv_name = "disk/%s" % (base_index + idx, )
11913 class _InstNicModPrivate:
11914 """Data structure for network interface modifications.
11916 Used by L{LUInstanceSetParams}.
11919 def __init__(self):
11924 class LUInstanceSetParams(LogicalUnit):
11925 """Modifies an instances's parameters.
11928 HPATH = "instance-modify"
11929 HTYPE = constants.HTYPE_INSTANCE
11933 def _UpgradeDiskNicMods(kind, mods, verify_fn):
11934 assert ht.TList(mods)
11935 assert not mods or len(mods[0]) in (2, 3)
11937 if mods and len(mods[0]) == 2:
11941 for op, params in mods:
11942 if op in (constants.DDM_ADD, constants.DDM_REMOVE):
11943 result.append((op, -1, params))
11947 raise errors.OpPrereqError("Only one %s add or remove operation is"
11948 " supported at a time" % kind,
11949 errors.ECODE_INVAL)
11951 result.append((constants.DDM_MODIFY, op, params))
11953 assert verify_fn(result)
11960 def _CheckMods(kind, mods, key_types, item_fn):
11961 """Ensures requested disk/NIC modifications are valid.
11964 for (op, _, params) in mods:
11965 assert ht.TDict(params)
11967 utils.ForceDictType(params, key_types)
11969 if op == constants.DDM_REMOVE:
11971 raise errors.OpPrereqError("No settings should be passed when"
11972 " removing a %s" % kind,
11973 errors.ECODE_INVAL)
11974 elif op in (constants.DDM_ADD, constants.DDM_MODIFY):
11975 item_fn(op, params)
11977 raise errors.ProgrammerError("Unhandled operation '%s'" % op)
11980 def _VerifyDiskModification(op, params):
11981 """Verifies a disk modification.
11984 if op == constants.DDM_ADD:
11985 mode = params.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
11986 if mode not in constants.DISK_ACCESS_SET:
11987 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
11988 errors.ECODE_INVAL)
11990 size = params.get(constants.IDISK_SIZE, None)
11992 raise errors.OpPrereqError("Required disk parameter '%s' missing" %
11993 constants.IDISK_SIZE, errors.ECODE_INVAL)
11997 except (TypeError, ValueError), err:
11998 raise errors.OpPrereqError("Invalid disk size parameter: %s" % err,
11999 errors.ECODE_INVAL)
12001 params[constants.IDISK_SIZE] = size
12003 elif op == constants.DDM_MODIFY and constants.IDISK_SIZE in params:
12004 raise errors.OpPrereqError("Disk size change not possible, use"
12005 " grow-disk", errors.ECODE_INVAL)
12008 def _VerifyNicModification(op, params):
12009 """Verifies a network interface modification.
12012 if op in (constants.DDM_ADD, constants.DDM_MODIFY):
12013 ip = params.get(constants.INIC_IP, None)
12016 elif ip.lower() == constants.VALUE_NONE:
12017 params[constants.INIC_IP] = None
12018 elif not netutils.IPAddress.IsValid(ip):
12019 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
12020 errors.ECODE_INVAL)
12022 bridge = params.get("bridge", None)
12023 link = params.get(constants.INIC_LINK, None)
12024 if bridge and link:
12025 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
12026 " at the same time", errors.ECODE_INVAL)
12027 elif bridge and bridge.lower() == constants.VALUE_NONE:
12028 params["bridge"] = None
12029 elif link and link.lower() == constants.VALUE_NONE:
12030 params[constants.INIC_LINK] = None
12032 if op == constants.DDM_ADD:
12033 macaddr = params.get(constants.INIC_MAC, None)
12034 if macaddr is None:
12035 params[constants.INIC_MAC] = constants.VALUE_AUTO
12037 if constants.INIC_MAC in params:
12038 macaddr = params[constants.INIC_MAC]
12039 if macaddr not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
12040 macaddr = utils.NormalizeAndValidateMac(macaddr)
12042 if op == constants.DDM_MODIFY and macaddr == constants.VALUE_AUTO:
12043 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
12044 " modifying an existing NIC",
12045 errors.ECODE_INVAL)
12047 def CheckArguments(self):
12048 if not (self.op.nics or self.op.disks or self.op.disk_template or
12049 self.op.hvparams or self.op.beparams or self.op.os_name or
12050 self.op.offline is not None or self.op.runtime_mem):
12051 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
12053 if self.op.hvparams:
12054 _CheckGlobalHvParams(self.op.hvparams)
12057 self._UpgradeDiskNicMods("disk", self.op.disks,
12058 opcodes.OpInstanceSetParams.TestDiskModifications)
12060 self._UpgradeDiskNicMods("NIC", self.op.nics,
12061 opcodes.OpInstanceSetParams.TestNicModifications)
12063 # Check disk modifications
12064 self._CheckMods("disk", self.op.disks, constants.IDISK_PARAMS_TYPES,
12065 self._VerifyDiskModification)
12067 if self.op.disks and self.op.disk_template is not None:
12068 raise errors.OpPrereqError("Disk template conversion and other disk"
12069 " changes not supported at the same time",
12070 errors.ECODE_INVAL)
12072 if (self.op.disk_template and
12073 self.op.disk_template in constants.DTS_INT_MIRROR and
12074 self.op.remote_node is None):
12075 raise errors.OpPrereqError("Changing the disk template to a mirrored"
12076 " one requires specifying a secondary node",
12077 errors.ECODE_INVAL)
12079 # Check NIC modifications
12080 self._CheckMods("NIC", self.op.nics, constants.INIC_PARAMS_TYPES,
12081 self._VerifyNicModification)
12083 def ExpandNames(self):
12084 self._ExpandAndLockInstance()
12085 # Can't even acquire node locks in shared mode as upcoming changes in
12086 # Ganeti 2.6 will start to modify the node object on disk conversion
12087 self.needed_locks[locking.LEVEL_NODE] = []
12088 self.needed_locks[locking.LEVEL_NODE_RES] = []
12089 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
12091 def DeclareLocks(self, level):
12092 # TODO: Acquire group lock in shared mode (disk parameters)
12093 if level == locking.LEVEL_NODE:
12094 self._LockInstancesNodes()
12095 if self.op.disk_template and self.op.remote_node:
12096 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
12097 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
12098 elif level == locking.LEVEL_NODE_RES and self.op.disk_template:
12100 self.needed_locks[locking.LEVEL_NODE_RES] = \
12101 self.needed_locks[locking.LEVEL_NODE][:]
12103 def BuildHooksEnv(self):
12104 """Build hooks env.
12106 This runs on the master, primary and secondaries.
12110 if constants.BE_MINMEM in self.be_new:
12111 args["minmem"] = self.be_new[constants.BE_MINMEM]
12112 if constants.BE_MAXMEM in self.be_new:
12113 args["maxmem"] = self.be_new[constants.BE_MAXMEM]
12114 if constants.BE_VCPUS in self.be_new:
12115 args["vcpus"] = self.be_new[constants.BE_VCPUS]
12116 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
12117 # information at all.
12119 if self._new_nics is not None:
12122 for nic in self._new_nics:
12123 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
12124 mode = nicparams[constants.NIC_MODE]
12125 link = nicparams[constants.NIC_LINK]
12126 nics.append((nic.ip, nic.mac, mode, link))
12128 args["nics"] = nics
12130 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
12131 if self.op.disk_template:
12132 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
12133 if self.op.runtime_mem:
12134 env["RUNTIME_MEMORY"] = self.op.runtime_mem
12138 def BuildHooksNodes(self):
12139 """Build hooks nodes.
12142 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
12145 def _PrepareNicModification(self, params, private, old_ip, old_params,
12147 update_params_dict = dict([(key, params[key])
12148 for key in constants.NICS_PARAMETERS
12151 if "bridge" in params:
12152 update_params_dict[constants.NIC_LINK] = params["bridge"]
12154 new_params = _GetUpdatedParams(old_params, update_params_dict)
12155 utils.ForceDictType(new_params, constants.NICS_PARAMETER_TYPES)
12157 new_filled_params = cluster.SimpleFillNIC(new_params)
12158 objects.NIC.CheckParameterSyntax(new_filled_params)
12160 new_mode = new_filled_params[constants.NIC_MODE]
12161 if new_mode == constants.NIC_MODE_BRIDGED:
12162 bridge = new_filled_params[constants.NIC_LINK]
12163 msg = self.rpc.call_bridges_exist(pnode, [bridge]).fail_msg
12165 msg = "Error checking bridges on node '%s': %s" % (pnode, msg)
12167 self.warn.append(msg)
12169 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
12171 elif new_mode == constants.NIC_MODE_ROUTED:
12172 ip = params.get(constants.INIC_IP, old_ip)
12174 raise errors.OpPrereqError("Cannot set the NIC IP address to None"
12175 " on a routed NIC", errors.ECODE_INVAL)
12177 if constants.INIC_MAC in params:
12178 mac = params[constants.INIC_MAC]
12180 raise errors.OpPrereqError("Cannot unset the NIC MAC address",
12181 errors.ECODE_INVAL)
12182 elif mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
12183 # otherwise generate the MAC address
12184 params[constants.INIC_MAC] = \
12185 self.cfg.GenerateMAC(self.proc.GetECId())
12187 # or validate/reserve the current one
12189 self.cfg.ReserveMAC(mac, self.proc.GetECId())
12190 except errors.ReservationError:
12191 raise errors.OpPrereqError("MAC address '%s' already in use"
12192 " in cluster" % mac,
12193 errors.ECODE_NOTUNIQUE)
12195 private.params = new_params
12196 private.filled = new_filled_params
12198 return (None, None)
12200 def CheckPrereq(self):
12201 """Check prerequisites.
12203 This only checks the instance list against the existing names.
12206 # checking the new params on the primary/secondary nodes
12208 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
12209 cluster = self.cluster = self.cfg.GetClusterInfo()
12210 assert self.instance is not None, \
12211 "Cannot retrieve locked instance %s" % self.op.instance_name
12212 pnode = instance.primary_node
12213 nodelist = list(instance.all_nodes)
12214 pnode_info = self.cfg.GetNodeInfo(pnode)
12215 self.diskparams = self.cfg.GetNodeGroup(pnode_info.group).diskparams
12217 # Prepare disk/NIC modifications
12218 self.diskmod = PrepareContainerMods(self.op.disks, None)
12219 self.nicmod = PrepareContainerMods(self.op.nics, _InstNicModPrivate)
12222 if self.op.os_name and not self.op.force:
12223 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
12224 self.op.force_variant)
12225 instance_os = self.op.os_name
12227 instance_os = instance.os
12229 assert not (self.op.disk_template and self.op.disks), \
12230 "Can't modify disk template and apply disk changes at the same time"
12232 if self.op.disk_template:
12233 if instance.disk_template == self.op.disk_template:
12234 raise errors.OpPrereqError("Instance already has disk template %s" %
12235 instance.disk_template, errors.ECODE_INVAL)
12237 if (instance.disk_template,
12238 self.op.disk_template) not in self._DISK_CONVERSIONS:
12239 raise errors.OpPrereqError("Unsupported disk template conversion from"
12240 " %s to %s" % (instance.disk_template,
12241 self.op.disk_template),
12242 errors.ECODE_INVAL)
12243 _CheckInstanceState(self, instance, INSTANCE_DOWN,
12244 msg="cannot change disk template")
12245 if self.op.disk_template in constants.DTS_INT_MIRROR:
12246 if self.op.remote_node == pnode:
12247 raise errors.OpPrereqError("Given new secondary node %s is the same"
12248 " as the primary node of the instance" %
12249 self.op.remote_node, errors.ECODE_STATE)
12250 _CheckNodeOnline(self, self.op.remote_node)
12251 _CheckNodeNotDrained(self, self.op.remote_node)
12252 # FIXME: here we assume that the old instance type is DT_PLAIN
12253 assert instance.disk_template == constants.DT_PLAIN
12254 disks = [{constants.IDISK_SIZE: d.size,
12255 constants.IDISK_VG: d.logical_id[0]}
12256 for d in instance.disks]
12257 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
12258 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
12260 snode_info = self.cfg.GetNodeInfo(self.op.remote_node)
12261 snode_group = self.cfg.GetNodeGroup(snode_info.group)
12262 ipolicy = _CalculateGroupIPolicy(cluster, snode_group)
12263 _CheckTargetNodeIPolicy(self, ipolicy, instance, snode_info,
12264 ignore=self.op.ignore_ipolicy)
12265 if pnode_info.group != snode_info.group:
12266 self.LogWarning("The primary and secondary nodes are in two"
12267 " different node groups; the disk parameters"
12268 " from the first disk's node group will be"
12271 # hvparams processing
12272 if self.op.hvparams:
12273 hv_type = instance.hypervisor
12274 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
12275 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
12276 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
12279 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
12280 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
12281 self.hv_proposed = self.hv_new = hv_new # the new actual values
12282 self.hv_inst = i_hvdict # the new dict (without defaults)
12284 self.hv_proposed = cluster.SimpleFillHV(instance.hypervisor, instance.os,
12286 self.hv_new = self.hv_inst = {}
12288 # beparams processing
12289 if self.op.beparams:
12290 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
12292 objects.UpgradeBeParams(i_bedict)
12293 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
12294 be_new = cluster.SimpleFillBE(i_bedict)
12295 self.be_proposed = self.be_new = be_new # the new actual values
12296 self.be_inst = i_bedict # the new dict (without defaults)
12298 self.be_new = self.be_inst = {}
12299 self.be_proposed = cluster.SimpleFillBE(instance.beparams)
12300 be_old = cluster.FillBE(instance)
12302 # CPU param validation -- checking every time a paramtere is
12303 # changed to cover all cases where either CPU mask or vcpus have
12305 if (constants.BE_VCPUS in self.be_proposed and
12306 constants.HV_CPU_MASK in self.hv_proposed):
12308 utils.ParseMultiCpuMask(self.hv_proposed[constants.HV_CPU_MASK])
12309 # Verify mask is consistent with number of vCPUs. Can skip this
12310 # test if only 1 entry in the CPU mask, which means same mask
12311 # is applied to all vCPUs.
12312 if (len(cpu_list) > 1 and
12313 len(cpu_list) != self.be_proposed[constants.BE_VCPUS]):
12314 raise errors.OpPrereqError("Number of vCPUs [%d] does not match the"
12316 (self.be_proposed[constants.BE_VCPUS],
12317 self.hv_proposed[constants.HV_CPU_MASK]),
12318 errors.ECODE_INVAL)
12320 # Only perform this test if a new CPU mask is given
12321 if constants.HV_CPU_MASK in self.hv_new:
12322 # Calculate the largest CPU number requested
12323 max_requested_cpu = max(map(max, cpu_list))
12324 # Check that all of the instance's nodes have enough physical CPUs to
12325 # satisfy the requested CPU mask
12326 _CheckNodesPhysicalCPUs(self, instance.all_nodes,
12327 max_requested_cpu + 1, instance.hypervisor)
12329 # osparams processing
12330 if self.op.osparams:
12331 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
12332 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
12333 self.os_inst = i_osdict # the new dict (without defaults)
12339 #TODO(dynmem): do the appropriate check involving MINMEM
12340 if (constants.BE_MAXMEM in self.op.beparams and not self.op.force and
12341 be_new[constants.BE_MAXMEM] > be_old[constants.BE_MAXMEM]):
12342 mem_check_list = [pnode]
12343 if be_new[constants.BE_AUTO_BALANCE]:
12344 # either we changed auto_balance to yes or it was from before
12345 mem_check_list.extend(instance.secondary_nodes)
12346 instance_info = self.rpc.call_instance_info(pnode, instance.name,
12347 instance.hypervisor)
12348 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
12349 [instance.hypervisor])
12350 pninfo = nodeinfo[pnode]
12351 msg = pninfo.fail_msg
12353 # Assume the primary node is unreachable and go ahead
12354 self.warn.append("Can't get info from primary node %s: %s" %
12357 (_, _, (pnhvinfo, )) = pninfo.payload
12358 if not isinstance(pnhvinfo.get("memory_free", None), int):
12359 self.warn.append("Node data from primary node %s doesn't contain"
12360 " free memory information" % pnode)
12361 elif instance_info.fail_msg:
12362 self.warn.append("Can't get instance runtime information: %s" %
12363 instance_info.fail_msg)
12365 if instance_info.payload:
12366 current_mem = int(instance_info.payload["memory"])
12368 # Assume instance not running
12369 # (there is a slight race condition here, but it's not very
12370 # probable, and we have no other way to check)
12371 # TODO: Describe race condition
12373 #TODO(dynmem): do the appropriate check involving MINMEM
12374 miss_mem = (be_new[constants.BE_MAXMEM] - current_mem -
12375 pnhvinfo["memory_free"])
12377 raise errors.OpPrereqError("This change will prevent the instance"
12378 " from starting, due to %d MB of memory"
12379 " missing on its primary node" %
12381 errors.ECODE_NORES)
12383 if be_new[constants.BE_AUTO_BALANCE]:
12384 for node, nres in nodeinfo.items():
12385 if node not in instance.secondary_nodes:
12387 nres.Raise("Can't get info from secondary node %s" % node,
12388 prereq=True, ecode=errors.ECODE_STATE)
12389 (_, _, (nhvinfo, )) = nres.payload
12390 if not isinstance(nhvinfo.get("memory_free", None), int):
12391 raise errors.OpPrereqError("Secondary node %s didn't return free"
12392 " memory information" % node,
12393 errors.ECODE_STATE)
12394 #TODO(dynmem): do the appropriate check involving MINMEM
12395 elif be_new[constants.BE_MAXMEM] > nhvinfo["memory_free"]:
12396 raise errors.OpPrereqError("This change will prevent the instance"
12397 " from failover to its secondary node"
12398 " %s, due to not enough memory" % node,
12399 errors.ECODE_STATE)
12401 if self.op.runtime_mem:
12402 remote_info = self.rpc.call_instance_info(instance.primary_node,
12404 instance.hypervisor)
12405 remote_info.Raise("Error checking node %s" % instance.primary_node)
12406 if not remote_info.payload: # not running already
12407 raise errors.OpPrereqError("Instance %s is not running" % instance.name,
12408 errors.ECODE_STATE)
12410 current_memory = remote_info.payload["memory"]
12411 if (not self.op.force and
12412 (self.op.runtime_mem > self.be_proposed[constants.BE_MAXMEM] or
12413 self.op.runtime_mem < self.be_proposed[constants.BE_MINMEM])):
12414 raise errors.OpPrereqError("Instance %s must have memory between %d"
12415 " and %d MB of memory unless --force is"
12416 " given" % (instance.name,
12417 self.be_proposed[constants.BE_MINMEM],
12418 self.be_proposed[constants.BE_MAXMEM]),
12419 errors.ECODE_INVAL)
12421 if self.op.runtime_mem > current_memory:
12422 _CheckNodeFreeMemory(self, instance.primary_node,
12423 "ballooning memory for instance %s" %
12425 self.op.memory - current_memory,
12426 instance.hypervisor)
12428 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
12429 raise errors.OpPrereqError("Disk operations not supported for"
12430 " diskless instances",
12431 errors.ECODE_INVAL)
12433 def _PrepareNicCreate(_, params, private):
12434 return self._PrepareNicModification(params, private, None, {},
12437 def _PrepareNicMod(_, nic, params, private):
12438 return self._PrepareNicModification(params, private, nic.ip,
12439 nic.nicparams, cluster, pnode)
12441 # Verify NIC changes (operating on copy)
12442 nics = instance.nics[:]
12443 ApplyContainerMods("NIC", nics, None, self.nicmod,
12444 _PrepareNicCreate, _PrepareNicMod, None)
12445 if len(nics) > constants.MAX_NICS:
12446 raise errors.OpPrereqError("Instance has too many network interfaces"
12447 " (%d), cannot add more" % constants.MAX_NICS,
12448 errors.ECODE_STATE)
12450 # Verify disk changes (operating on a copy)
12451 disks = instance.disks[:]
12452 ApplyContainerMods("disk", disks, None, self.diskmod, None, None, None)
12453 if len(disks) > constants.MAX_DISKS:
12454 raise errors.OpPrereqError("Instance has too many disks (%d), cannot add"
12455 " more" % constants.MAX_DISKS,
12456 errors.ECODE_STATE)
12458 if self.op.offline is not None:
12459 if self.op.offline:
12460 msg = "can't change to offline"
12462 msg = "can't change to online"
12463 _CheckInstanceState(self, instance, CAN_CHANGE_INSTANCE_OFFLINE, msg=msg)
12465 # Pre-compute NIC changes (necessary to use result in hooks)
12466 self._nic_chgdesc = []
12468 # Operate on copies as this is still in prereq
12469 nics = [nic.Copy() for nic in instance.nics]
12470 ApplyContainerMods("NIC", nics, self._nic_chgdesc, self.nicmod,
12471 self._CreateNewNic, self._ApplyNicMods, None)
12472 self._new_nics = nics
12474 self._new_nics = None
12476 def _ConvertPlainToDrbd(self, feedback_fn):
12477 """Converts an instance from plain to drbd.
12480 feedback_fn("Converting template to drbd")
12481 instance = self.instance
12482 pnode = instance.primary_node
12483 snode = self.op.remote_node
12485 assert instance.disk_template == constants.DT_PLAIN
12487 # create a fake disk info for _GenerateDiskTemplate
12488 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
12489 constants.IDISK_VG: d.logical_id[0]}
12490 for d in instance.disks]
12491 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
12492 instance.name, pnode, [snode],
12493 disk_info, None, None, 0, feedback_fn,
12495 info = _GetInstanceInfoText(instance)
12496 feedback_fn("Creating aditional volumes...")
12497 # first, create the missing data and meta devices
12498 for disk in new_disks:
12499 # unfortunately this is... not too nice
12500 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
12502 for child in disk.children:
12503 _CreateSingleBlockDev(self, snode, instance, child, info, True)
12504 # at this stage, all new LVs have been created, we can rename the
12506 feedback_fn("Renaming original volumes...")
12507 rename_list = [(o, n.children[0].logical_id)
12508 for (o, n) in zip(instance.disks, new_disks)]
12509 result = self.rpc.call_blockdev_rename(pnode, rename_list)
12510 result.Raise("Failed to rename original LVs")
12512 feedback_fn("Initializing DRBD devices...")
12513 # all child devices are in place, we can now create the DRBD devices
12514 for disk in new_disks:
12515 for node in [pnode, snode]:
12516 f_create = node == pnode
12517 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
12519 # at this point, the instance has been modified
12520 instance.disk_template = constants.DT_DRBD8
12521 instance.disks = new_disks
12522 self.cfg.Update(instance, feedback_fn)
12524 # Release node locks while waiting for sync
12525 _ReleaseLocks(self, locking.LEVEL_NODE)
12527 # disks are created, waiting for sync
12528 disk_abort = not _WaitForSync(self, instance,
12529 oneshot=not self.op.wait_for_sync)
12531 raise errors.OpExecError("There are some degraded disks for"
12532 " this instance, please cleanup manually")
12534 # Node resource locks will be released by caller
12536 def _ConvertDrbdToPlain(self, feedback_fn):
12537 """Converts an instance from drbd to plain.
12540 instance = self.instance
12542 assert len(instance.secondary_nodes) == 1
12543 assert instance.disk_template == constants.DT_DRBD8
12545 pnode = instance.primary_node
12546 snode = instance.secondary_nodes[0]
12547 feedback_fn("Converting template to plain")
12549 old_disks = instance.disks
12550 new_disks = [d.children[0] for d in old_disks]
12552 # copy over size and mode
12553 for parent, child in zip(old_disks, new_disks):
12554 child.size = parent.size
12555 child.mode = parent.mode
12557 # update instance structure
12558 instance.disks = new_disks
12559 instance.disk_template = constants.DT_PLAIN
12560 self.cfg.Update(instance, feedback_fn)
12562 # Release locks in case removing disks takes a while
12563 _ReleaseLocks(self, locking.LEVEL_NODE)
12565 feedback_fn("Removing volumes on the secondary node...")
12566 for disk in old_disks:
12567 self.cfg.SetDiskID(disk, snode)
12568 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
12570 self.LogWarning("Could not remove block device %s on node %s,"
12571 " continuing anyway: %s", disk.iv_name, snode, msg)
12573 feedback_fn("Removing unneeded volumes on the primary node...")
12574 for idx, disk in enumerate(old_disks):
12575 meta = disk.children[1]
12576 self.cfg.SetDiskID(meta, pnode)
12577 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
12579 self.LogWarning("Could not remove metadata for disk %d on node %s,"
12580 " continuing anyway: %s", idx, pnode, msg)
12582 # this is a DRBD disk, return its port to the pool
12583 for disk in old_disks:
12584 tcp_port = disk.logical_id[2]
12585 self.cfg.AddTcpUdpPort(tcp_port)
12587 # Node resource locks will be released by caller
12589 def _CreateNewDisk(self, idx, params, _):
12590 """Creates a new disk.
12593 instance = self.instance
12596 if instance.disk_template in constants.DTS_FILEBASED:
12597 (file_driver, file_path) = instance.disks[0].logical_id
12598 file_path = os.path.dirname(file_path)
12600 file_driver = file_path = None
12603 _GenerateDiskTemplate(self, instance.disk_template, instance.name,
12604 instance.primary_node, instance.secondary_nodes,
12605 [params], file_path, file_driver, idx,
12606 self.Log, self.diskparams)[0]
12608 info = _GetInstanceInfoText(instance)
12610 logging.info("Creating volume %s for instance %s",
12611 disk.iv_name, instance.name)
12612 # Note: this needs to be kept in sync with _CreateDisks
12614 for node in instance.all_nodes:
12615 f_create = (node == instance.primary_node)
12617 _CreateBlockDev(self, node, instance, disk, f_create, info, f_create)
12618 except errors.OpExecError, err:
12619 self.LogWarning("Failed to create volume %s (%s) on node '%s': %s",
12620 disk.iv_name, disk, node, err)
12623 ("disk/%d" % idx, "add:size=%s,mode=%s" % (disk.size, disk.mode)),
12627 def _ModifyDisk(idx, disk, params, _):
12628 """Modifies a disk.
12631 disk.mode = params[constants.IDISK_MODE]
12634 ("disk.mode/%d" % idx, disk.mode),
12637 def _RemoveDisk(self, idx, root, _):
12641 for node, disk in root.ComputeNodeTree(self.instance.primary_node):
12642 self.cfg.SetDiskID(disk, node)
12643 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
12645 self.LogWarning("Could not remove disk/%d on node '%s': %s,"
12646 " continuing anyway", idx, node, msg)
12648 # if this is a DRBD disk, return its port to the pool
12649 if root.dev_type in constants.LDS_DRBD:
12650 self.cfg.AddTcpUdpPort(root.logical_id[2])
12653 def _CreateNewNic(idx, params, private):
12654 """Creates data structure for a new network interface.
12657 mac = params[constants.INIC_MAC]
12658 ip = params.get(constants.INIC_IP, None)
12659 nicparams = private.params
12661 return (objects.NIC(mac=mac, ip=ip, nicparams=nicparams), [
12663 "add:mac=%s,ip=%s,mode=%s,link=%s" %
12664 (mac, ip, private.filled[constants.NIC_MODE],
12665 private.filled[constants.NIC_LINK])),
12669 def _ApplyNicMods(idx, nic, params, private):
12670 """Modifies a network interface.
12675 for key in [constants.INIC_MAC, constants.INIC_IP]:
12677 changes.append(("nic.%s/%d" % (key, idx), params[key]))
12678 setattr(nic, key, params[key])
12681 nic.nicparams = private.params
12683 for (key, val) in params.items():
12684 changes.append(("nic.%s/%d" % (key, idx), val))
12688 def Exec(self, feedback_fn):
12689 """Modifies an instance.
12691 All parameters take effect only at the next restart of the instance.
12694 # Process here the warnings from CheckPrereq, as we don't have a
12695 # feedback_fn there.
12696 # TODO: Replace with self.LogWarning
12697 for warn in self.warn:
12698 feedback_fn("WARNING: %s" % warn)
12700 assert ((self.op.disk_template is None) ^
12701 bool(self.owned_locks(locking.LEVEL_NODE_RES))), \
12702 "Not owning any node resource locks"
12705 instance = self.instance
12708 if self.op.runtime_mem:
12709 rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
12711 self.op.runtime_mem)
12712 rpcres.Raise("Cannot modify instance runtime memory")
12713 result.append(("runtime_memory", self.op.runtime_mem))
12715 # Apply disk changes
12716 ApplyContainerMods("disk", instance.disks, result, self.diskmod,
12717 self._CreateNewDisk, self._ModifyDisk, self._RemoveDisk)
12718 _UpdateIvNames(0, instance.disks)
12720 if self.op.disk_template:
12722 check_nodes = set(instance.all_nodes)
12723 if self.op.remote_node:
12724 check_nodes.add(self.op.remote_node)
12725 for level in [locking.LEVEL_NODE, locking.LEVEL_NODE_RES]:
12726 owned = self.owned_locks(level)
12727 assert not (check_nodes - owned), \
12728 ("Not owning the correct locks, owning %r, expected at least %r" %
12729 (owned, check_nodes))
12731 r_shut = _ShutdownInstanceDisks(self, instance)
12733 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
12734 " proceed with disk template conversion")
12735 mode = (instance.disk_template, self.op.disk_template)
12737 self._DISK_CONVERSIONS[mode](self, feedback_fn)
12739 self.cfg.ReleaseDRBDMinors(instance.name)
12741 result.append(("disk_template", self.op.disk_template))
12743 assert instance.disk_template == self.op.disk_template, \
12744 ("Expected disk template '%s', found '%s'" %
12745 (self.op.disk_template, instance.disk_template))
12747 # Release node and resource locks if there are any (they might already have
12748 # been released during disk conversion)
12749 _ReleaseLocks(self, locking.LEVEL_NODE)
12750 _ReleaseLocks(self, locking.LEVEL_NODE_RES)
12752 # Apply NIC changes
12753 if self._new_nics is not None:
12754 instance.nics = self._new_nics
12755 result.extend(self._nic_chgdesc)
12758 if self.op.hvparams:
12759 instance.hvparams = self.hv_inst
12760 for key, val in self.op.hvparams.iteritems():
12761 result.append(("hv/%s" % key, val))
12764 if self.op.beparams:
12765 instance.beparams = self.be_inst
12766 for key, val in self.op.beparams.iteritems():
12767 result.append(("be/%s" % key, val))
12770 if self.op.os_name:
12771 instance.os = self.op.os_name
12774 if self.op.osparams:
12775 instance.osparams = self.os_inst
12776 for key, val in self.op.osparams.iteritems():
12777 result.append(("os/%s" % key, val))
12779 if self.op.offline is None:
12782 elif self.op.offline:
12783 # Mark instance as offline
12784 self.cfg.MarkInstanceOffline(instance.name)
12785 result.append(("admin_state", constants.ADMINST_OFFLINE))
12787 # Mark instance as online, but stopped
12788 self.cfg.MarkInstanceDown(instance.name)
12789 result.append(("admin_state", constants.ADMINST_DOWN))
12791 self.cfg.Update(instance, feedback_fn)
12793 assert not (self.owned_locks(locking.LEVEL_NODE_RES) or
12794 self.owned_locks(locking.LEVEL_NODE)), \
12795 "All node locks should have been released by now"
12799 _DISK_CONVERSIONS = {
12800 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
12801 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
12805 class LUInstanceChangeGroup(LogicalUnit):
12806 HPATH = "instance-change-group"
12807 HTYPE = constants.HTYPE_INSTANCE
12810 def ExpandNames(self):
12811 self.share_locks = _ShareAll()
12812 self.needed_locks = {
12813 locking.LEVEL_NODEGROUP: [],
12814 locking.LEVEL_NODE: [],
12817 self._ExpandAndLockInstance()
12819 if self.op.target_groups:
12820 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
12821 self.op.target_groups)
12823 self.req_target_uuids = None
12825 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
12827 def DeclareLocks(self, level):
12828 if level == locking.LEVEL_NODEGROUP:
12829 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
12831 if self.req_target_uuids:
12832 lock_groups = set(self.req_target_uuids)
12834 # Lock all groups used by instance optimistically; this requires going
12835 # via the node before it's locked, requiring verification later on
12836 instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
12837 lock_groups.update(instance_groups)
12839 # No target groups, need to lock all of them
12840 lock_groups = locking.ALL_SET
12842 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
12844 elif level == locking.LEVEL_NODE:
12845 if self.req_target_uuids:
12846 # Lock all nodes used by instances
12847 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
12848 self._LockInstancesNodes()
12850 # Lock all nodes in all potential target groups
12851 lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) -
12852 self.cfg.GetInstanceNodeGroups(self.op.instance_name))
12853 member_nodes = [node_name
12854 for group in lock_groups
12855 for node_name in self.cfg.GetNodeGroup(group).members]
12856 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
12858 # Lock all nodes as all groups are potential targets
12859 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12861 def CheckPrereq(self):
12862 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
12863 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12864 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
12866 assert (self.req_target_uuids is None or
12867 owned_groups.issuperset(self.req_target_uuids))
12868 assert owned_instances == set([self.op.instance_name])
12870 # Get instance information
12871 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
12873 # Check if node groups for locked instance are still correct
12874 assert owned_nodes.issuperset(self.instance.all_nodes), \
12875 ("Instance %s's nodes changed while we kept the lock" %
12876 self.op.instance_name)
12878 inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name,
12881 if self.req_target_uuids:
12882 # User requested specific target groups
12883 self.target_uuids = self.req_target_uuids
12885 # All groups except those used by the instance are potential targets
12886 self.target_uuids = owned_groups - inst_groups
12888 conflicting_groups = self.target_uuids & inst_groups
12889 if conflicting_groups:
12890 raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are"
12891 " used by the instance '%s'" %
12892 (utils.CommaJoin(conflicting_groups),
12893 self.op.instance_name),
12894 errors.ECODE_INVAL)
12896 if not self.target_uuids:
12897 raise errors.OpPrereqError("There are no possible target groups",
12898 errors.ECODE_INVAL)
12900 def BuildHooksEnv(self):
12901 """Build hooks env.
12904 assert self.target_uuids
12907 "TARGET_GROUPS": " ".join(self.target_uuids),
12910 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
12914 def BuildHooksNodes(self):
12915 """Build hooks nodes.
12918 mn = self.cfg.GetMasterNode()
12919 return ([mn], [mn])
12921 def Exec(self, feedback_fn):
12922 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
12924 assert instances == [self.op.instance_name], "Instance not locked"
12926 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
12927 instances=instances, target_groups=list(self.target_uuids))
12929 ial.Run(self.op.iallocator)
12931 if not ial.success:
12932 raise errors.OpPrereqError("Can't compute solution for changing group of"
12933 " instance '%s' using iallocator '%s': %s" %
12934 (self.op.instance_name, self.op.iallocator,
12936 errors.ECODE_NORES)
12938 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
12940 self.LogInfo("Iallocator returned %s job(s) for changing group of"
12941 " instance '%s'", len(jobs), self.op.instance_name)
12943 return ResultWithJobs(jobs)
12946 class LUBackupQuery(NoHooksLU):
12947 """Query the exports list
12952 def ExpandNames(self):
12953 self.needed_locks = {}
12954 self.share_locks[locking.LEVEL_NODE] = 1
12955 if not self.op.nodes:
12956 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12958 self.needed_locks[locking.LEVEL_NODE] = \
12959 _GetWantedNodes(self, self.op.nodes)
12961 def Exec(self, feedback_fn):
12962 """Compute the list of all the exported system images.
12965 @return: a dictionary with the structure node->(export-list)
12966 where export-list is a list of the instances exported on
12970 self.nodes = self.owned_locks(locking.LEVEL_NODE)
12971 rpcresult = self.rpc.call_export_list(self.nodes)
12973 for node in rpcresult:
12974 if rpcresult[node].fail_msg:
12975 result[node] = False
12977 result[node] = rpcresult[node].payload
12982 class LUBackupPrepare(NoHooksLU):
12983 """Prepares an instance for an export and returns useful information.
12988 def ExpandNames(self):
12989 self._ExpandAndLockInstance()
12991 def CheckPrereq(self):
12992 """Check prerequisites.
12995 instance_name = self.op.instance_name
12997 self.instance = self.cfg.GetInstanceInfo(instance_name)
12998 assert self.instance is not None, \
12999 "Cannot retrieve locked instance %s" % self.op.instance_name
13000 _CheckNodeOnline(self, self.instance.primary_node)
13002 self._cds = _GetClusterDomainSecret()
13004 def Exec(self, feedback_fn):
13005 """Prepares an instance for an export.
13008 instance = self.instance
13010 if self.op.mode == constants.EXPORT_MODE_REMOTE:
13011 salt = utils.GenerateSecret(8)
13013 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
13014 result = self.rpc.call_x509_cert_create(instance.primary_node,
13015 constants.RIE_CERT_VALIDITY)
13016 result.Raise("Can't create X509 key and certificate on %s" % result.node)
13018 (name, cert_pem) = result.payload
13020 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
13024 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
13025 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
13027 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
13033 class LUBackupExport(LogicalUnit):
13034 """Export an instance to an image in the cluster.
13037 HPATH = "instance-export"
13038 HTYPE = constants.HTYPE_INSTANCE
13041 def CheckArguments(self):
13042 """Check the arguments.
13045 self.x509_key_name = self.op.x509_key_name
13046 self.dest_x509_ca_pem = self.op.destination_x509_ca
13048 if self.op.mode == constants.EXPORT_MODE_REMOTE:
13049 if not self.x509_key_name:
13050 raise errors.OpPrereqError("Missing X509 key name for encryption",
13051 errors.ECODE_INVAL)
13053 if not self.dest_x509_ca_pem:
13054 raise errors.OpPrereqError("Missing destination X509 CA",
13055 errors.ECODE_INVAL)
13057 def ExpandNames(self):
13058 self._ExpandAndLockInstance()
13060 # Lock all nodes for local exports
13061 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13062 # FIXME: lock only instance primary and destination node
13064 # Sad but true, for now we have do lock all nodes, as we don't know where
13065 # the previous export might be, and in this LU we search for it and
13066 # remove it from its current node. In the future we could fix this by:
13067 # - making a tasklet to search (share-lock all), then create the
13068 # new one, then one to remove, after
13069 # - removing the removal operation altogether
13070 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
13072 def DeclareLocks(self, level):
13073 """Last minute lock declaration."""
13074 # All nodes are locked anyway, so nothing to do here.
13076 def BuildHooksEnv(self):
13077 """Build hooks env.
13079 This will run on the master, primary node and target node.
13083 "EXPORT_MODE": self.op.mode,
13084 "EXPORT_NODE": self.op.target_node,
13085 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
13086 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
13087 # TODO: Generic function for boolean env variables
13088 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
13091 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
13095 def BuildHooksNodes(self):
13096 """Build hooks nodes.
13099 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
13101 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13102 nl.append(self.op.target_node)
13106 def CheckPrereq(self):
13107 """Check prerequisites.
13109 This checks that the instance and node names are valid.
13112 instance_name = self.op.instance_name
13114 self.instance = self.cfg.GetInstanceInfo(instance_name)
13115 assert self.instance is not None, \
13116 "Cannot retrieve locked instance %s" % self.op.instance_name
13117 _CheckNodeOnline(self, self.instance.primary_node)
13119 if (self.op.remove_instance and
13120 self.instance.admin_state == constants.ADMINST_UP and
13121 not self.op.shutdown):
13122 raise errors.OpPrereqError("Can not remove instance without shutting it"
13125 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13126 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
13127 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
13128 assert self.dst_node is not None
13130 _CheckNodeOnline(self, self.dst_node.name)
13131 _CheckNodeNotDrained(self, self.dst_node.name)
13134 self.dest_disk_info = None
13135 self.dest_x509_ca = None
13137 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
13138 self.dst_node = None
13140 if len(self.op.target_node) != len(self.instance.disks):
13141 raise errors.OpPrereqError(("Received destination information for %s"
13142 " disks, but instance %s has %s disks") %
13143 (len(self.op.target_node), instance_name,
13144 len(self.instance.disks)),
13145 errors.ECODE_INVAL)
13147 cds = _GetClusterDomainSecret()
13149 # Check X509 key name
13151 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
13152 except (TypeError, ValueError), err:
13153 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
13155 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
13156 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
13157 errors.ECODE_INVAL)
13159 # Load and verify CA
13161 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
13162 except OpenSSL.crypto.Error, err:
13163 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
13164 (err, ), errors.ECODE_INVAL)
13166 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
13167 if errcode is not None:
13168 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
13169 (msg, ), errors.ECODE_INVAL)
13171 self.dest_x509_ca = cert
13173 # Verify target information
13175 for idx, disk_data in enumerate(self.op.target_node):
13177 (host, port, magic) = \
13178 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
13179 except errors.GenericError, err:
13180 raise errors.OpPrereqError("Target info for disk %s: %s" %
13181 (idx, err), errors.ECODE_INVAL)
13183 disk_info.append((host, port, magic))
13185 assert len(disk_info) == len(self.op.target_node)
13186 self.dest_disk_info = disk_info
13189 raise errors.ProgrammerError("Unhandled export mode %r" %
13192 # instance disk type verification
13193 # TODO: Implement export support for file-based disks
13194 for disk in self.instance.disks:
13195 if disk.dev_type == constants.LD_FILE:
13196 raise errors.OpPrereqError("Export not supported for instances with"
13197 " file-based disks", errors.ECODE_INVAL)
13199 def _CleanupExports(self, feedback_fn):
13200 """Removes exports of current instance from all other nodes.
13202 If an instance in a cluster with nodes A..D was exported to node C, its
13203 exports will be removed from the nodes A, B and D.
13206 assert self.op.mode != constants.EXPORT_MODE_REMOTE
13208 nodelist = self.cfg.GetNodeList()
13209 nodelist.remove(self.dst_node.name)
13211 # on one-node clusters nodelist will be empty after the removal
13212 # if we proceed the backup would be removed because OpBackupQuery
13213 # substitutes an empty list with the full cluster node list.
13214 iname = self.instance.name
13216 feedback_fn("Removing old exports for instance %s" % iname)
13217 exportlist = self.rpc.call_export_list(nodelist)
13218 for node in exportlist:
13219 if exportlist[node].fail_msg:
13221 if iname in exportlist[node].payload:
13222 msg = self.rpc.call_export_remove(node, iname).fail_msg
13224 self.LogWarning("Could not remove older export for instance %s"
13225 " on node %s: %s", iname, node, msg)
13227 def Exec(self, feedback_fn):
13228 """Export an instance to an image in the cluster.
13231 assert self.op.mode in constants.EXPORT_MODES
13233 instance = self.instance
13234 src_node = instance.primary_node
13236 if self.op.shutdown:
13237 # shutdown the instance, but not the disks
13238 feedback_fn("Shutting down instance %s" % instance.name)
13239 result = self.rpc.call_instance_shutdown(src_node, instance,
13240 self.op.shutdown_timeout)
13241 # TODO: Maybe ignore failures if ignore_remove_failures is set
13242 result.Raise("Could not shutdown instance %s on"
13243 " node %s" % (instance.name, src_node))
13245 # set the disks ID correctly since call_instance_start needs the
13246 # correct drbd minor to create the symlinks
13247 for disk in instance.disks:
13248 self.cfg.SetDiskID(disk, src_node)
13250 activate_disks = (instance.admin_state != constants.ADMINST_UP)
13253 # Activate the instance disks if we'exporting a stopped instance
13254 feedback_fn("Activating disks for %s" % instance.name)
13255 _StartInstanceDisks(self, instance, None)
13258 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
13261 helper.CreateSnapshots()
13263 if (self.op.shutdown and
13264 instance.admin_state == constants.ADMINST_UP and
13265 not self.op.remove_instance):
13266 assert not activate_disks
13267 feedback_fn("Starting instance %s" % instance.name)
13268 result = self.rpc.call_instance_start(src_node,
13269 (instance, None, None), False)
13270 msg = result.fail_msg
13272 feedback_fn("Failed to start instance: %s" % msg)
13273 _ShutdownInstanceDisks(self, instance)
13274 raise errors.OpExecError("Could not start instance: %s" % msg)
13276 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13277 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
13278 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
13279 connect_timeout = constants.RIE_CONNECT_TIMEOUT
13280 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
13282 (key_name, _, _) = self.x509_key_name
13285 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
13288 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
13289 key_name, dest_ca_pem,
13294 # Check for backwards compatibility
13295 assert len(dresults) == len(instance.disks)
13296 assert compat.all(isinstance(i, bool) for i in dresults), \
13297 "Not all results are boolean: %r" % dresults
13301 feedback_fn("Deactivating disks for %s" % instance.name)
13302 _ShutdownInstanceDisks(self, instance)
13304 if not (compat.all(dresults) and fin_resu):
13307 failures.append("export finalization")
13308 if not compat.all(dresults):
13309 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
13311 failures.append("disk export: disk(s) %s" % fdsk)
13313 raise errors.OpExecError("Export failed, errors in %s" %
13314 utils.CommaJoin(failures))
13316 # At this point, the export was successful, we can cleanup/finish
13318 # Remove instance if requested
13319 if self.op.remove_instance:
13320 feedback_fn("Removing instance %s" % instance.name)
13321 _RemoveInstance(self, feedback_fn, instance,
13322 self.op.ignore_remove_failures)
13324 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13325 self._CleanupExports(feedback_fn)
13327 return fin_resu, dresults
13330 class LUBackupRemove(NoHooksLU):
13331 """Remove exports related to the named instance.
13336 def ExpandNames(self):
13337 self.needed_locks = {}
13338 # We need all nodes to be locked in order for RemoveExport to work, but we
13339 # don't need to lock the instance itself, as nothing will happen to it (and
13340 # we can remove exports also for a removed instance)
13341 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
13343 def Exec(self, feedback_fn):
13344 """Remove any export.
13347 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
13348 # If the instance was not found we'll try with the name that was passed in.
13349 # This will only work if it was an FQDN, though.
13351 if not instance_name:
13353 instance_name = self.op.instance_name
13355 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
13356 exportlist = self.rpc.call_export_list(locked_nodes)
13358 for node in exportlist:
13359 msg = exportlist[node].fail_msg
13361 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
13363 if instance_name in exportlist[node].payload:
13365 result = self.rpc.call_export_remove(node, instance_name)
13366 msg = result.fail_msg
13368 logging.error("Could not remove export for instance %s"
13369 " on node %s: %s", instance_name, node, msg)
13371 if fqdn_warn and not found:
13372 feedback_fn("Export not found. If trying to remove an export belonging"
13373 " to a deleted instance please use its Fully Qualified"
13377 class LUGroupAdd(LogicalUnit):
13378 """Logical unit for creating node groups.
13381 HPATH = "group-add"
13382 HTYPE = constants.HTYPE_GROUP
13385 def ExpandNames(self):
13386 # We need the new group's UUID here so that we can create and acquire the
13387 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
13388 # that it should not check whether the UUID exists in the configuration.
13389 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
13390 self.needed_locks = {}
13391 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
13393 def CheckPrereq(self):
13394 """Check prerequisites.
13396 This checks that the given group name is not an existing node group
13401 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13402 except errors.OpPrereqError:
13405 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
13406 " node group (UUID: %s)" %
13407 (self.op.group_name, existing_uuid),
13408 errors.ECODE_EXISTS)
13410 if self.op.ndparams:
13411 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
13413 if self.op.hv_state:
13414 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None)
13416 self.new_hv_state = None
13418 if self.op.disk_state:
13419 self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None)
13421 self.new_disk_state = None
13423 if self.op.diskparams:
13424 for templ in constants.DISK_TEMPLATES:
13425 if templ not in self.op.diskparams:
13426 self.op.diskparams[templ] = {}
13427 utils.ForceDictType(self.op.diskparams[templ], constants.DISK_DT_TYPES)
13429 self.op.diskparams = self.cfg.GetClusterInfo().diskparams
13431 if self.op.ipolicy:
13432 cluster = self.cfg.GetClusterInfo()
13433 full_ipolicy = cluster.SimpleFillIPolicy(self.op.ipolicy)
13435 objects.InstancePolicy.CheckParameterSyntax(full_ipolicy)
13436 except errors.ConfigurationError, err:
13437 raise errors.OpPrereqError("Invalid instance policy: %s" % err,
13438 errors.ECODE_INVAL)
13440 def BuildHooksEnv(self):
13441 """Build hooks env.
13445 "GROUP_NAME": self.op.group_name,
13448 def BuildHooksNodes(self):
13449 """Build hooks nodes.
13452 mn = self.cfg.GetMasterNode()
13453 return ([mn], [mn])
13455 def Exec(self, feedback_fn):
13456 """Add the node group to the cluster.
13459 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
13460 uuid=self.group_uuid,
13461 alloc_policy=self.op.alloc_policy,
13462 ndparams=self.op.ndparams,
13463 diskparams=self.op.diskparams,
13464 ipolicy=self.op.ipolicy,
13465 hv_state_static=self.new_hv_state,
13466 disk_state_static=self.new_disk_state)
13468 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
13469 del self.remove_locks[locking.LEVEL_NODEGROUP]
13472 class LUGroupAssignNodes(NoHooksLU):
13473 """Logical unit for assigning nodes to groups.
13478 def ExpandNames(self):
13479 # These raise errors.OpPrereqError on their own:
13480 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13481 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
13483 # We want to lock all the affected nodes and groups. We have readily
13484 # available the list of nodes, and the *destination* group. To gather the
13485 # list of "source" groups, we need to fetch node information later on.
13486 self.needed_locks = {
13487 locking.LEVEL_NODEGROUP: set([self.group_uuid]),
13488 locking.LEVEL_NODE: self.op.nodes,
13491 def DeclareLocks(self, level):
13492 if level == locking.LEVEL_NODEGROUP:
13493 assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1
13495 # Try to get all affected nodes' groups without having the group or node
13496 # lock yet. Needs verification later in the code flow.
13497 groups = self.cfg.GetNodeGroupsFromNodes(self.op.nodes)
13499 self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)
13501 def CheckPrereq(self):
13502 """Check prerequisites.
13505 assert self.needed_locks[locking.LEVEL_NODEGROUP]
13506 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
13507 frozenset(self.op.nodes))
13509 expected_locks = (set([self.group_uuid]) |
13510 self.cfg.GetNodeGroupsFromNodes(self.op.nodes))
13511 actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP)
13512 if actual_locks != expected_locks:
13513 raise errors.OpExecError("Nodes changed groups since locks were acquired,"
13514 " current groups are '%s', used to be '%s'" %
13515 (utils.CommaJoin(expected_locks),
13516 utils.CommaJoin(actual_locks)))
13518 self.node_data = self.cfg.GetAllNodesInfo()
13519 self.group = self.cfg.GetNodeGroup(self.group_uuid)
13520 instance_data = self.cfg.GetAllInstancesInfo()
13522 if self.group is None:
13523 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13524 (self.op.group_name, self.group_uuid))
13526 (new_splits, previous_splits) = \
13527 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
13528 for node in self.op.nodes],
13529 self.node_data, instance_data)
13532 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
13534 if not self.op.force:
13535 raise errors.OpExecError("The following instances get split by this"
13536 " change and --force was not given: %s" %
13539 self.LogWarning("This operation will split the following instances: %s",
13542 if previous_splits:
13543 self.LogWarning("In addition, these already-split instances continue"
13544 " to be split across groups: %s",
13545 utils.CommaJoin(utils.NiceSort(previous_splits)))
13547 def Exec(self, feedback_fn):
13548 """Assign nodes to a new group.
13551 mods = [(node_name, self.group_uuid) for node_name in self.op.nodes]
13553 self.cfg.AssignGroupNodes(mods)
13556 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
13557 """Check for split instances after a node assignment.
13559 This method considers a series of node assignments as an atomic operation,
13560 and returns information about split instances after applying the set of
13563 In particular, it returns information about newly split instances, and
13564 instances that were already split, and remain so after the change.
13566 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
13569 @type changes: list of (node_name, new_group_uuid) pairs.
13570 @param changes: list of node assignments to consider.
13571 @param node_data: a dict with data for all nodes
13572 @param instance_data: a dict with all instances to consider
13573 @rtype: a two-tuple
13574 @return: a list of instances that were previously okay and result split as a
13575 consequence of this change, and a list of instances that were previously
13576 split and this change does not fix.
13579 changed_nodes = dict((node, group) for node, group in changes
13580 if node_data[node].group != group)
13582 all_split_instances = set()
13583 previously_split_instances = set()
13585 def InstanceNodes(instance):
13586 return [instance.primary_node] + list(instance.secondary_nodes)
13588 for inst in instance_data.values():
13589 if inst.disk_template not in constants.DTS_INT_MIRROR:
13592 instance_nodes = InstanceNodes(inst)
13594 if len(set(node_data[node].group for node in instance_nodes)) > 1:
13595 previously_split_instances.add(inst.name)
13597 if len(set(changed_nodes.get(node, node_data[node].group)
13598 for node in instance_nodes)) > 1:
13599 all_split_instances.add(inst.name)
13601 return (list(all_split_instances - previously_split_instances),
13602 list(previously_split_instances & all_split_instances))
13605 class _GroupQuery(_QueryBase):
13606 FIELDS = query.GROUP_FIELDS
13608 def ExpandNames(self, lu):
13609 lu.needed_locks = {}
13611 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
13612 self._cluster = lu.cfg.GetClusterInfo()
13613 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
13616 self.wanted = [name_to_uuid[name]
13617 for name in utils.NiceSort(name_to_uuid.keys())]
13619 # Accept names to be either names or UUIDs.
13622 all_uuid = frozenset(self._all_groups.keys())
13624 for name in self.names:
13625 if name in all_uuid:
13626 self.wanted.append(name)
13627 elif name in name_to_uuid:
13628 self.wanted.append(name_to_uuid[name])
13630 missing.append(name)
13633 raise errors.OpPrereqError("Some groups do not exist: %s" %
13634 utils.CommaJoin(missing),
13635 errors.ECODE_NOENT)
13637 def DeclareLocks(self, lu, level):
13640 def _GetQueryData(self, lu):
13641 """Computes the list of node groups and their attributes.
13644 do_nodes = query.GQ_NODE in self.requested_data
13645 do_instances = query.GQ_INST in self.requested_data
13647 group_to_nodes = None
13648 group_to_instances = None
13650 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
13651 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
13652 # latter GetAllInstancesInfo() is not enough, for we have to go through
13653 # instance->node. Hence, we will need to process nodes even if we only need
13654 # instance information.
13655 if do_nodes or do_instances:
13656 all_nodes = lu.cfg.GetAllNodesInfo()
13657 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
13660 for node in all_nodes.values():
13661 if node.group in group_to_nodes:
13662 group_to_nodes[node.group].append(node.name)
13663 node_to_group[node.name] = node.group
13666 all_instances = lu.cfg.GetAllInstancesInfo()
13667 group_to_instances = dict((uuid, []) for uuid in self.wanted)
13669 for instance in all_instances.values():
13670 node = instance.primary_node
13671 if node in node_to_group:
13672 group_to_instances[node_to_group[node]].append(instance.name)
13675 # Do not pass on node information if it was not requested.
13676 group_to_nodes = None
13678 return query.GroupQueryData(self._cluster,
13679 [self._all_groups[uuid]
13680 for uuid in self.wanted],
13681 group_to_nodes, group_to_instances)
13684 class LUGroupQuery(NoHooksLU):
13685 """Logical unit for querying node groups.
13690 def CheckArguments(self):
13691 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
13692 self.op.output_fields, False)
13694 def ExpandNames(self):
13695 self.gq.ExpandNames(self)
13697 def DeclareLocks(self, level):
13698 self.gq.DeclareLocks(self, level)
13700 def Exec(self, feedback_fn):
13701 return self.gq.OldStyleQuery(self)
13704 class LUGroupSetParams(LogicalUnit):
13705 """Modifies the parameters of a node group.
13708 HPATH = "group-modify"
13709 HTYPE = constants.HTYPE_GROUP
13712 def CheckArguments(self):
13715 self.op.diskparams,
13716 self.op.alloc_policy,
13718 self.op.disk_state,
13722 if all_changes.count(None) == len(all_changes):
13723 raise errors.OpPrereqError("Please pass at least one modification",
13724 errors.ECODE_INVAL)
13726 def ExpandNames(self):
13727 # This raises errors.OpPrereqError on its own:
13728 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13730 self.needed_locks = {
13731 locking.LEVEL_INSTANCE: [],
13732 locking.LEVEL_NODEGROUP: [self.group_uuid],
13735 self.share_locks[locking.LEVEL_INSTANCE] = 1
13737 def DeclareLocks(self, level):
13738 if level == locking.LEVEL_INSTANCE:
13739 assert not self.needed_locks[locking.LEVEL_INSTANCE]
13741 # Lock instances optimistically, needs verification once group lock has
13743 self.needed_locks[locking.LEVEL_INSTANCE] = \
13744 self.cfg.GetNodeGroupInstances(self.group_uuid)
13746 def CheckPrereq(self):
13747 """Check prerequisites.
13750 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
13752 # Check if locked instances are still correct
13753 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
13755 self.group = self.cfg.GetNodeGroup(self.group_uuid)
13756 cluster = self.cfg.GetClusterInfo()
13758 if self.group is None:
13759 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13760 (self.op.group_name, self.group_uuid))
13762 if self.op.ndparams:
13763 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
13764 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
13765 self.new_ndparams = new_ndparams
13767 if self.op.diskparams:
13768 self.new_diskparams = dict()
13769 for templ in constants.DISK_TEMPLATES:
13770 if templ not in self.op.diskparams:
13771 self.op.diskparams[templ] = {}
13772 new_templ_params = _GetUpdatedParams(self.group.diskparams[templ],
13773 self.op.diskparams[templ])
13774 utils.ForceDictType(new_templ_params, constants.DISK_DT_TYPES)
13775 self.new_diskparams[templ] = new_templ_params
13777 if self.op.hv_state:
13778 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
13779 self.group.hv_state_static)
13781 if self.op.disk_state:
13782 self.new_disk_state = \
13783 _MergeAndVerifyDiskState(self.op.disk_state,
13784 self.group.disk_state_static)
13786 if self.op.ipolicy:
13787 self.new_ipolicy = _GetUpdatedIPolicy(self.group.ipolicy,
13791 new_ipolicy = cluster.SimpleFillIPolicy(self.new_ipolicy)
13792 inst_filter = lambda inst: inst.name in owned_instances
13793 instances = self.cfg.GetInstancesInfoByFilter(inst_filter).values()
13795 _ComputeNewInstanceViolations(_CalculateGroupIPolicy(cluster,
13797 new_ipolicy, instances)
13800 self.LogWarning("After the ipolicy change the following instances"
13801 " violate them: %s",
13802 utils.CommaJoin(violations))
13804 def BuildHooksEnv(self):
13805 """Build hooks env.
13809 "GROUP_NAME": self.op.group_name,
13810 "NEW_ALLOC_POLICY": self.op.alloc_policy,
13813 def BuildHooksNodes(self):
13814 """Build hooks nodes.
13817 mn = self.cfg.GetMasterNode()
13818 return ([mn], [mn])
13820 def Exec(self, feedback_fn):
13821 """Modifies the node group.
13826 if self.op.ndparams:
13827 self.group.ndparams = self.new_ndparams
13828 result.append(("ndparams", str(self.group.ndparams)))
13830 if self.op.diskparams:
13831 self.group.diskparams = self.new_diskparams
13832 result.append(("diskparams", str(self.group.diskparams)))
13834 if self.op.alloc_policy:
13835 self.group.alloc_policy = self.op.alloc_policy
13837 if self.op.hv_state:
13838 self.group.hv_state_static = self.new_hv_state
13840 if self.op.disk_state:
13841 self.group.disk_state_static = self.new_disk_state
13843 if self.op.ipolicy:
13844 self.group.ipolicy = self.new_ipolicy
13846 self.cfg.Update(self.group, feedback_fn)
13850 class LUGroupRemove(LogicalUnit):
13851 HPATH = "group-remove"
13852 HTYPE = constants.HTYPE_GROUP
13855 def ExpandNames(self):
13856 # This will raises errors.OpPrereqError on its own:
13857 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13858 self.needed_locks = {
13859 locking.LEVEL_NODEGROUP: [self.group_uuid],
13862 def CheckPrereq(self):
13863 """Check prerequisites.
13865 This checks that the given group name exists as a node group, that is
13866 empty (i.e., contains no nodes), and that is not the last group of the
13870 # Verify that the group is empty.
13871 group_nodes = [node.name
13872 for node in self.cfg.GetAllNodesInfo().values()
13873 if node.group == self.group_uuid]
13876 raise errors.OpPrereqError("Group '%s' not empty, has the following"
13878 (self.op.group_name,
13879 utils.CommaJoin(utils.NiceSort(group_nodes))),
13880 errors.ECODE_STATE)
13882 # Verify the cluster would not be left group-less.
13883 if len(self.cfg.GetNodeGroupList()) == 1:
13884 raise errors.OpPrereqError("Group '%s' is the only group,"
13885 " cannot be removed" %
13886 self.op.group_name,
13887 errors.ECODE_STATE)
13889 def BuildHooksEnv(self):
13890 """Build hooks env.
13894 "GROUP_NAME": self.op.group_name,
13897 def BuildHooksNodes(self):
13898 """Build hooks nodes.
13901 mn = self.cfg.GetMasterNode()
13902 return ([mn], [mn])
13904 def Exec(self, feedback_fn):
13905 """Remove the node group.
13909 self.cfg.RemoveNodeGroup(self.group_uuid)
13910 except errors.ConfigurationError:
13911 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
13912 (self.op.group_name, self.group_uuid))
13914 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
13917 class LUGroupRename(LogicalUnit):
13918 HPATH = "group-rename"
13919 HTYPE = constants.HTYPE_GROUP
13922 def ExpandNames(self):
13923 # This raises errors.OpPrereqError on its own:
13924 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13926 self.needed_locks = {
13927 locking.LEVEL_NODEGROUP: [self.group_uuid],
13930 def CheckPrereq(self):
13931 """Check prerequisites.
13933 Ensures requested new name is not yet used.
13937 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
13938 except errors.OpPrereqError:
13941 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
13942 " node group (UUID: %s)" %
13943 (self.op.new_name, new_name_uuid),
13944 errors.ECODE_EXISTS)
13946 def BuildHooksEnv(self):
13947 """Build hooks env.
13951 "OLD_NAME": self.op.group_name,
13952 "NEW_NAME": self.op.new_name,
13955 def BuildHooksNodes(self):
13956 """Build hooks nodes.
13959 mn = self.cfg.GetMasterNode()
13961 all_nodes = self.cfg.GetAllNodesInfo()
13962 all_nodes.pop(mn, None)
13965 run_nodes.extend(node.name for node in all_nodes.values()
13966 if node.group == self.group_uuid)
13968 return (run_nodes, run_nodes)
13970 def Exec(self, feedback_fn):
13971 """Rename the node group.
13974 group = self.cfg.GetNodeGroup(self.group_uuid)
13977 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13978 (self.op.group_name, self.group_uuid))
13980 group.name = self.op.new_name
13981 self.cfg.Update(group, feedback_fn)
13983 return self.op.new_name
13986 class LUGroupEvacuate(LogicalUnit):
13987 HPATH = "group-evacuate"
13988 HTYPE = constants.HTYPE_GROUP
13991 def ExpandNames(self):
13992 # This raises errors.OpPrereqError on its own:
13993 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13995 if self.op.target_groups:
13996 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
13997 self.op.target_groups)
13999 self.req_target_uuids = []
14001 if self.group_uuid in self.req_target_uuids:
14002 raise errors.OpPrereqError("Group to be evacuated (%s) can not be used"
14003 " as a target group (targets are %s)" %
14005 utils.CommaJoin(self.req_target_uuids)),
14006 errors.ECODE_INVAL)
14008 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
14010 self.share_locks = _ShareAll()
14011 self.needed_locks = {
14012 locking.LEVEL_INSTANCE: [],
14013 locking.LEVEL_NODEGROUP: [],
14014 locking.LEVEL_NODE: [],
14017 def DeclareLocks(self, level):
14018 if level == locking.LEVEL_INSTANCE:
14019 assert not self.needed_locks[locking.LEVEL_INSTANCE]
14021 # Lock instances optimistically, needs verification once node and group
14022 # locks have been acquired
14023 self.needed_locks[locking.LEVEL_INSTANCE] = \
14024 self.cfg.GetNodeGroupInstances(self.group_uuid)
14026 elif level == locking.LEVEL_NODEGROUP:
14027 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
14029 if self.req_target_uuids:
14030 lock_groups = set([self.group_uuid] + self.req_target_uuids)
14032 # Lock all groups used by instances optimistically; this requires going
14033 # via the node before it's locked, requiring verification later on
14034 lock_groups.update(group_uuid
14035 for instance_name in
14036 self.owned_locks(locking.LEVEL_INSTANCE)
14038 self.cfg.GetInstanceNodeGroups(instance_name))
14040 # No target groups, need to lock all of them
14041 lock_groups = locking.ALL_SET
14043 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
14045 elif level == locking.LEVEL_NODE:
14046 # This will only lock the nodes in the group to be evacuated which
14047 # contain actual instances
14048 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
14049 self._LockInstancesNodes()
14051 # Lock all nodes in group to be evacuated and target groups
14052 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
14053 assert self.group_uuid in owned_groups
14054 member_nodes = [node_name
14055 for group in owned_groups
14056 for node_name in self.cfg.GetNodeGroup(group).members]
14057 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
14059 def CheckPrereq(self):
14060 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
14061 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
14062 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
14064 assert owned_groups.issuperset(self.req_target_uuids)
14065 assert self.group_uuid in owned_groups
14067 # Check if locked instances are still correct
14068 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
14070 # Get instance information
14071 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
14073 # Check if node groups for locked instances are still correct
14074 for instance_name in owned_instances:
14075 inst = self.instances[instance_name]
14076 assert owned_nodes.issuperset(inst.all_nodes), \
14077 "Instance %s's nodes changed while we kept the lock" % instance_name
14079 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
14082 assert self.group_uuid in inst_groups, \
14083 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
14085 if self.req_target_uuids:
14086 # User requested specific target groups
14087 self.target_uuids = self.req_target_uuids
14089 # All groups except the one to be evacuated are potential targets
14090 self.target_uuids = [group_uuid for group_uuid in owned_groups
14091 if group_uuid != self.group_uuid]
14093 if not self.target_uuids:
14094 raise errors.OpPrereqError("There are no possible target groups",
14095 errors.ECODE_INVAL)
14097 def BuildHooksEnv(self):
14098 """Build hooks env.
14102 "GROUP_NAME": self.op.group_name,
14103 "TARGET_GROUPS": " ".join(self.target_uuids),
14106 def BuildHooksNodes(self):
14107 """Build hooks nodes.
14110 mn = self.cfg.GetMasterNode()
14112 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
14114 run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members
14116 return (run_nodes, run_nodes)
14118 def Exec(self, feedback_fn):
14119 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
14121 assert self.group_uuid not in self.target_uuids
14123 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
14124 instances=instances, target_groups=self.target_uuids)
14126 ial.Run(self.op.iallocator)
14128 if not ial.success:
14129 raise errors.OpPrereqError("Can't compute group evacuation using"
14130 " iallocator '%s': %s" %
14131 (self.op.iallocator, ial.info),
14132 errors.ECODE_NORES)
14134 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
14136 self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s",
14137 len(jobs), self.op.group_name)
14139 return ResultWithJobs(jobs)
14142 class TagsLU(NoHooksLU): # pylint: disable=W0223
14143 """Generic tags LU.
14145 This is an abstract class which is the parent of all the other tags LUs.
14148 def ExpandNames(self):
14149 self.group_uuid = None
14150 self.needed_locks = {}
14151 if self.op.kind == constants.TAG_NODE:
14152 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
14153 self.needed_locks[locking.LEVEL_NODE] = self.op.name
14154 elif self.op.kind == constants.TAG_INSTANCE:
14155 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
14156 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
14157 elif self.op.kind == constants.TAG_NODEGROUP:
14158 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
14160 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
14161 # not possible to acquire the BGL based on opcode parameters)
14163 def CheckPrereq(self):
14164 """Check prerequisites.
14167 if self.op.kind == constants.TAG_CLUSTER:
14168 self.target = self.cfg.GetClusterInfo()
14169 elif self.op.kind == constants.TAG_NODE:
14170 self.target = self.cfg.GetNodeInfo(self.op.name)
14171 elif self.op.kind == constants.TAG_INSTANCE:
14172 self.target = self.cfg.GetInstanceInfo(self.op.name)
14173 elif self.op.kind == constants.TAG_NODEGROUP:
14174 self.target = self.cfg.GetNodeGroup(self.group_uuid)
14176 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
14177 str(self.op.kind), errors.ECODE_INVAL)
14180 class LUTagsGet(TagsLU):
14181 """Returns the tags of a given object.
14186 def ExpandNames(self):
14187 TagsLU.ExpandNames(self)
14189 # Share locks as this is only a read operation
14190 self.share_locks = _ShareAll()
14192 def Exec(self, feedback_fn):
14193 """Returns the tag list.
14196 return list(self.target.GetTags())
14199 class LUTagsSearch(NoHooksLU):
14200 """Searches the tags for a given pattern.
14205 def ExpandNames(self):
14206 self.needed_locks = {}
14208 def CheckPrereq(self):
14209 """Check prerequisites.
14211 This checks the pattern passed for validity by compiling it.
14215 self.re = re.compile(self.op.pattern)
14216 except re.error, err:
14217 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
14218 (self.op.pattern, err), errors.ECODE_INVAL)
14220 def Exec(self, feedback_fn):
14221 """Returns the tag list.
14225 tgts = [("/cluster", cfg.GetClusterInfo())]
14226 ilist = cfg.GetAllInstancesInfo().values()
14227 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
14228 nlist = cfg.GetAllNodesInfo().values()
14229 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
14230 tgts.extend(("/nodegroup/%s" % n.name, n)
14231 for n in cfg.GetAllNodeGroupsInfo().values())
14233 for path, target in tgts:
14234 for tag in target.GetTags():
14235 if self.re.search(tag):
14236 results.append((path, tag))
14240 class LUTagsSet(TagsLU):
14241 """Sets a tag on a given object.
14246 def CheckPrereq(self):
14247 """Check prerequisites.
14249 This checks the type and length of the tag name and value.
14252 TagsLU.CheckPrereq(self)
14253 for tag in self.op.tags:
14254 objects.TaggableObject.ValidateTag(tag)
14256 def Exec(self, feedback_fn):
14261 for tag in self.op.tags:
14262 self.target.AddTag(tag)
14263 except errors.TagError, err:
14264 raise errors.OpExecError("Error while setting tag: %s" % str(err))
14265 self.cfg.Update(self.target, feedback_fn)
14268 class LUTagsDel(TagsLU):
14269 """Delete a list of tags from a given object.
14274 def CheckPrereq(self):
14275 """Check prerequisites.
14277 This checks that we have the given tag.
14280 TagsLU.CheckPrereq(self)
14281 for tag in self.op.tags:
14282 objects.TaggableObject.ValidateTag(tag)
14283 del_tags = frozenset(self.op.tags)
14284 cur_tags = self.target.GetTags()
14286 diff_tags = del_tags - cur_tags
14288 diff_names = ("'%s'" % i for i in sorted(diff_tags))
14289 raise errors.OpPrereqError("Tag(s) %s not found" %
14290 (utils.CommaJoin(diff_names), ),
14291 errors.ECODE_NOENT)
14293 def Exec(self, feedback_fn):
14294 """Remove the tag from the object.
14297 for tag in self.op.tags:
14298 self.target.RemoveTag(tag)
14299 self.cfg.Update(self.target, feedback_fn)
14302 class LUTestDelay(NoHooksLU):
14303 """Sleep for a specified amount of time.
14305 This LU sleeps on the master and/or nodes for a specified amount of
14311 def ExpandNames(self):
14312 """Expand names and set required locks.
14314 This expands the node list, if any.
14317 self.needed_locks = {}
14318 if self.op.on_nodes:
14319 # _GetWantedNodes can be used here, but is not always appropriate to use
14320 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
14321 # more information.
14322 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
14323 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
14325 def _TestDelay(self):
14326 """Do the actual sleep.
14329 if self.op.on_master:
14330 if not utils.TestDelay(self.op.duration):
14331 raise errors.OpExecError("Error during master delay test")
14332 if self.op.on_nodes:
14333 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
14334 for node, node_result in result.items():
14335 node_result.Raise("Failure during rpc call to node %s" % node)
14337 def Exec(self, feedback_fn):
14338 """Execute the test delay opcode, with the wanted repetitions.
14341 if self.op.repeat == 0:
14344 top_value = self.op.repeat - 1
14345 for i in range(self.op.repeat):
14346 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
14350 class LUTestJqueue(NoHooksLU):
14351 """Utility LU to test some aspects of the job queue.
14356 # Must be lower than default timeout for WaitForJobChange to see whether it
14357 # notices changed jobs
14358 _CLIENT_CONNECT_TIMEOUT = 20.0
14359 _CLIENT_CONFIRM_TIMEOUT = 60.0
14362 def _NotifyUsingSocket(cls, cb, errcls):
14363 """Opens a Unix socket and waits for another program to connect.
14366 @param cb: Callback to send socket name to client
14367 @type errcls: class
14368 @param errcls: Exception class to use for errors
14371 # Using a temporary directory as there's no easy way to create temporary
14372 # sockets without writing a custom loop around tempfile.mktemp and
14374 tmpdir = tempfile.mkdtemp()
14376 tmpsock = utils.PathJoin(tmpdir, "sock")
14378 logging.debug("Creating temporary socket at %s", tmpsock)
14379 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
14384 # Send details to client
14387 # Wait for client to connect before continuing
14388 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
14390 (conn, _) = sock.accept()
14391 except socket.error, err:
14392 raise errcls("Client didn't connect in time (%s)" % err)
14396 # Remove as soon as client is connected
14397 shutil.rmtree(tmpdir)
14399 # Wait for client to close
14402 # pylint: disable=E1101
14403 # Instance of '_socketobject' has no ... member
14404 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
14406 except socket.error, err:
14407 raise errcls("Client failed to confirm notification (%s)" % err)
14411 def _SendNotification(self, test, arg, sockname):
14412 """Sends a notification to the client.
14415 @param test: Test name
14416 @param arg: Test argument (depends on test)
14417 @type sockname: string
14418 @param sockname: Socket path
14421 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
14423 def _Notify(self, prereq, test, arg):
14424 """Notifies the client of a test.
14427 @param prereq: Whether this is a prereq-phase test
14429 @param test: Test name
14430 @param arg: Test argument (depends on test)
14434 errcls = errors.OpPrereqError
14436 errcls = errors.OpExecError
14438 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
14442 def CheckArguments(self):
14443 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
14444 self.expandnames_calls = 0
14446 def ExpandNames(self):
14447 checkargs_calls = getattr(self, "checkargs_calls", 0)
14448 if checkargs_calls < 1:
14449 raise errors.ProgrammerError("CheckArguments was not called")
14451 self.expandnames_calls += 1
14453 if self.op.notify_waitlock:
14454 self._Notify(True, constants.JQT_EXPANDNAMES, None)
14456 self.LogInfo("Expanding names")
14458 # Get lock on master node (just to get a lock, not for a particular reason)
14459 self.needed_locks = {
14460 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
14463 def Exec(self, feedback_fn):
14464 if self.expandnames_calls < 1:
14465 raise errors.ProgrammerError("ExpandNames was not called")
14467 if self.op.notify_exec:
14468 self._Notify(False, constants.JQT_EXEC, None)
14470 self.LogInfo("Executing")
14472 if self.op.log_messages:
14473 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
14474 for idx, msg in enumerate(self.op.log_messages):
14475 self.LogInfo("Sending log message %s", idx + 1)
14476 feedback_fn(constants.JQT_MSGPREFIX + msg)
14477 # Report how many test messages have been sent
14478 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
14481 raise errors.OpExecError("Opcode failure was requested")
14486 class IAllocator(object):
14487 """IAllocator framework.
14489 An IAllocator instance has three sets of attributes:
14490 - cfg that is needed to query the cluster
14491 - input data (all members of the _KEYS class attribute are required)
14492 - four buffer attributes (in|out_data|text), that represent the
14493 input (to the external script) in text and data structure format,
14494 and the output from it, again in two formats
14495 - the result variables from the script (success, info, nodes) for
14499 # pylint: disable=R0902
14500 # lots of instance attributes
14502 def __init__(self, cfg, rpc_runner, mode, **kwargs):
14504 self.rpc = rpc_runner
14505 # init buffer variables
14506 self.in_text = self.out_text = self.in_data = self.out_data = None
14507 # init all input fields so that pylint is happy
14509 self.memory = self.disks = self.disk_template = None
14510 self.os = self.tags = self.nics = self.vcpus = None
14511 self.hypervisor = None
14512 self.relocate_from = None
14514 self.instances = None
14515 self.evac_mode = None
14516 self.target_groups = []
14518 self.required_nodes = None
14519 # init result fields
14520 self.success = self.info = self.result = None
14523 (fn, keydata, self._result_check) = self._MODE_DATA[self.mode]
14525 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
14526 " IAllocator" % self.mode)
14528 keyset = [n for (n, _) in keydata]
14531 if key not in keyset:
14532 raise errors.ProgrammerError("Invalid input parameter '%s' to"
14533 " IAllocator" % key)
14534 setattr(self, key, kwargs[key])
14537 if key not in kwargs:
14538 raise errors.ProgrammerError("Missing input parameter '%s' to"
14539 " IAllocator" % key)
14540 self._BuildInputData(compat.partial(fn, self), keydata)
14542 def _ComputeClusterData(self):
14543 """Compute the generic allocator input data.
14545 This is the data that is independent of the actual operation.
14549 cluster_info = cfg.GetClusterInfo()
14552 "version": constants.IALLOCATOR_VERSION,
14553 "cluster_name": cfg.GetClusterName(),
14554 "cluster_tags": list(cluster_info.GetTags()),
14555 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
14556 "ipolicy": cluster_info.ipolicy,
14558 ninfo = cfg.GetAllNodesInfo()
14559 iinfo = cfg.GetAllInstancesInfo().values()
14560 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
14563 node_list = [n.name for n in ninfo.values() if n.vm_capable]
14565 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
14566 hypervisor_name = self.hypervisor
14567 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
14568 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
14570 hypervisor_name = cluster_info.primary_hypervisor
14572 node_data = self.rpc.call_node_info(node_list, [cfg.GetVGName()],
14575 self.rpc.call_all_instances_info(node_list,
14576 cluster_info.enabled_hypervisors)
14578 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
14580 config_ndata = self._ComputeBasicNodeData(cfg, ninfo)
14581 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
14582 i_list, config_ndata)
14583 assert len(data["nodes"]) == len(ninfo), \
14584 "Incomplete node data computed"
14586 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
14588 self.in_data = data
14591 def _ComputeNodeGroupData(cfg):
14592 """Compute node groups data.
14595 cluster = cfg.GetClusterInfo()
14596 ng = dict((guuid, {
14597 "name": gdata.name,
14598 "alloc_policy": gdata.alloc_policy,
14599 "ipolicy": _CalculateGroupIPolicy(cluster, gdata),
14601 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items())
14606 def _ComputeBasicNodeData(cfg, node_cfg):
14607 """Compute global node data.
14610 @returns: a dict of name: (node dict, node config)
14613 # fill in static (config-based) values
14614 node_results = dict((ninfo.name, {
14615 "tags": list(ninfo.GetTags()),
14616 "primary_ip": ninfo.primary_ip,
14617 "secondary_ip": ninfo.secondary_ip,
14618 "offline": ninfo.offline,
14619 "drained": ninfo.drained,
14620 "master_candidate": ninfo.master_candidate,
14621 "group": ninfo.group,
14622 "master_capable": ninfo.master_capable,
14623 "vm_capable": ninfo.vm_capable,
14624 "ndparams": cfg.GetNdParams(ninfo),
14626 for ninfo in node_cfg.values())
14628 return node_results
14631 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
14633 """Compute global node data.
14635 @param node_results: the basic node structures as filled from the config
14638 #TODO(dynmem): compute the right data on MAX and MIN memory
14639 # make a copy of the current dict
14640 node_results = dict(node_results)
14641 for nname, nresult in node_data.items():
14642 assert nname in node_results, "Missing basic data for node %s" % nname
14643 ninfo = node_cfg[nname]
14645 if not (ninfo.offline or ninfo.drained):
14646 nresult.Raise("Can't get data for node %s" % nname)
14647 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
14649 remote_info = _MakeLegacyNodeInfo(nresult.payload)
14651 for attr in ["memory_total", "memory_free", "memory_dom0",
14652 "vg_size", "vg_free", "cpu_total"]:
14653 if attr not in remote_info:
14654 raise errors.OpExecError("Node '%s' didn't return attribute"
14655 " '%s'" % (nname, attr))
14656 if not isinstance(remote_info[attr], int):
14657 raise errors.OpExecError("Node '%s' returned invalid value"
14659 (nname, attr, remote_info[attr]))
14660 # compute memory used by primary instances
14661 i_p_mem = i_p_up_mem = 0
14662 for iinfo, beinfo in i_list:
14663 if iinfo.primary_node == nname:
14664 i_p_mem += beinfo[constants.BE_MAXMEM]
14665 if iinfo.name not in node_iinfo[nname].payload:
14668 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]["memory"])
14669 i_mem_diff = beinfo[constants.BE_MAXMEM] - i_used_mem
14670 remote_info["memory_free"] -= max(0, i_mem_diff)
14672 if iinfo.admin_state == constants.ADMINST_UP:
14673 i_p_up_mem += beinfo[constants.BE_MAXMEM]
14675 # compute memory used by instances
14677 "total_memory": remote_info["memory_total"],
14678 "reserved_memory": remote_info["memory_dom0"],
14679 "free_memory": remote_info["memory_free"],
14680 "total_disk": remote_info["vg_size"],
14681 "free_disk": remote_info["vg_free"],
14682 "total_cpus": remote_info["cpu_total"],
14683 "i_pri_memory": i_p_mem,
14684 "i_pri_up_memory": i_p_up_mem,
14686 pnr_dyn.update(node_results[nname])
14687 node_results[nname] = pnr_dyn
14689 return node_results
14692 def _ComputeInstanceData(cluster_info, i_list):
14693 """Compute global instance data.
14697 for iinfo, beinfo in i_list:
14699 for nic in iinfo.nics:
14700 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
14704 "mode": filled_params[constants.NIC_MODE],
14705 "link": filled_params[constants.NIC_LINK],
14707 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
14708 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
14709 nic_data.append(nic_dict)
14711 "tags": list(iinfo.GetTags()),
14712 "admin_state": iinfo.admin_state,
14713 "vcpus": beinfo[constants.BE_VCPUS],
14714 "memory": beinfo[constants.BE_MAXMEM],
14716 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
14718 "disks": [{constants.IDISK_SIZE: dsk.size,
14719 constants.IDISK_MODE: dsk.mode}
14720 for dsk in iinfo.disks],
14721 "disk_template": iinfo.disk_template,
14722 "hypervisor": iinfo.hypervisor,
14724 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
14726 instance_data[iinfo.name] = pir
14728 return instance_data
14730 def _AddNewInstance(self):
14731 """Add new instance data to allocator structure.
14733 This in combination with _AllocatorGetClusterData will create the
14734 correct structure needed as input for the allocator.
14736 The checks for the completeness of the opcode must have already been
14740 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
14742 if self.disk_template in constants.DTS_INT_MIRROR:
14743 self.required_nodes = 2
14745 self.required_nodes = 1
14749 "disk_template": self.disk_template,
14752 "vcpus": self.vcpus,
14753 "memory": self.memory,
14754 "disks": self.disks,
14755 "disk_space_total": disk_space,
14757 "required_nodes": self.required_nodes,
14758 "hypervisor": self.hypervisor,
14763 def _AddRelocateInstance(self):
14764 """Add relocate instance data to allocator structure.
14766 This in combination with _IAllocatorGetClusterData will create the
14767 correct structure needed as input for the allocator.
14769 The checks for the completeness of the opcode must have already been
14773 instance = self.cfg.GetInstanceInfo(self.name)
14774 if instance is None:
14775 raise errors.ProgrammerError("Unknown instance '%s' passed to"
14776 " IAllocator" % self.name)
14778 if instance.disk_template not in constants.DTS_MIRRORED:
14779 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
14780 errors.ECODE_INVAL)
14782 if instance.disk_template in constants.DTS_INT_MIRROR and \
14783 len(instance.secondary_nodes) != 1:
14784 raise errors.OpPrereqError("Instance has not exactly one secondary node",
14785 errors.ECODE_STATE)
14787 self.required_nodes = 1
14788 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
14789 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
14793 "disk_space_total": disk_space,
14794 "required_nodes": self.required_nodes,
14795 "relocate_from": self.relocate_from,
14799 def _AddNodeEvacuate(self):
14800 """Get data for node-evacuate requests.
14804 "instances": self.instances,
14805 "evac_mode": self.evac_mode,
14808 def _AddChangeGroup(self):
14809 """Get data for node-evacuate requests.
14813 "instances": self.instances,
14814 "target_groups": self.target_groups,
14817 def _BuildInputData(self, fn, keydata):
14818 """Build input data structures.
14821 self._ComputeClusterData()
14824 request["type"] = self.mode
14825 for keyname, keytype in keydata:
14826 if keyname not in request:
14827 raise errors.ProgrammerError("Request parameter %s is missing" %
14829 val = request[keyname]
14830 if not keytype(val):
14831 raise errors.ProgrammerError("Request parameter %s doesn't pass"
14832 " validation, value %s, expected"
14833 " type %s" % (keyname, val, keytype))
14834 self.in_data["request"] = request
14836 self.in_text = serializer.Dump(self.in_data)
14838 _STRING_LIST = ht.TListOf(ht.TString)
14839 _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
14840 # pylint: disable=E1101
14841 # Class '...' has no 'OP_ID' member
14842 "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
14843 opcodes.OpInstanceMigrate.OP_ID,
14844 opcodes.OpInstanceReplaceDisks.OP_ID])
14848 ht.TListOf(ht.TAnd(ht.TIsLength(3),
14849 ht.TItems([ht.TNonEmptyString,
14850 ht.TNonEmptyString,
14851 ht.TListOf(ht.TNonEmptyString),
14854 ht.TListOf(ht.TAnd(ht.TIsLength(2),
14855 ht.TItems([ht.TNonEmptyString,
14858 _NEVAC_RESULT = ht.TAnd(ht.TIsLength(3),
14859 ht.TItems([_NEVAC_MOVED, _NEVAC_FAILED, _JOB_LIST]))
14862 constants.IALLOCATOR_MODE_ALLOC:
14865 ("name", ht.TString),
14866 ("memory", ht.TInt),
14867 ("disks", ht.TListOf(ht.TDict)),
14868 ("disk_template", ht.TString),
14869 ("os", ht.TString),
14870 ("tags", _STRING_LIST),
14871 ("nics", ht.TListOf(ht.TDict)),
14872 ("vcpus", ht.TInt),
14873 ("hypervisor", ht.TString),
14875 constants.IALLOCATOR_MODE_RELOC:
14876 (_AddRelocateInstance,
14877 [("name", ht.TString), ("relocate_from", _STRING_LIST)],
14879 constants.IALLOCATOR_MODE_NODE_EVAC:
14880 (_AddNodeEvacuate, [
14881 ("instances", _STRING_LIST),
14882 ("evac_mode", ht.TElemOf(constants.IALLOCATOR_NEVAC_MODES)),
14884 constants.IALLOCATOR_MODE_CHG_GROUP:
14885 (_AddChangeGroup, [
14886 ("instances", _STRING_LIST),
14887 ("target_groups", _STRING_LIST),
14891 def Run(self, name, validate=True, call_fn=None):
14892 """Run an instance allocator and return the results.
14895 if call_fn is None:
14896 call_fn = self.rpc.call_iallocator_runner
14898 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
14899 result.Raise("Failure while running the iallocator script")
14901 self.out_text = result.payload
14903 self._ValidateResult()
14905 def _ValidateResult(self):
14906 """Process the allocator results.
14908 This will process and if successful save the result in
14909 self.out_data and the other parameters.
14913 rdict = serializer.Load(self.out_text)
14914 except Exception, err:
14915 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
14917 if not isinstance(rdict, dict):
14918 raise errors.OpExecError("Can't parse iallocator results: not a dict")
14920 # TODO: remove backwards compatiblity in later versions
14921 if "nodes" in rdict and "result" not in rdict:
14922 rdict["result"] = rdict["nodes"]
14925 for key in "success", "info", "result":
14926 if key not in rdict:
14927 raise errors.OpExecError("Can't parse iallocator results:"
14928 " missing key '%s'" % key)
14929 setattr(self, key, rdict[key])
14931 if not self._result_check(self.result):
14932 raise errors.OpExecError("Iallocator returned invalid result,"
14933 " expected %s, got %s" %
14934 (self._result_check, self.result),
14935 errors.ECODE_INVAL)
14937 if self.mode == constants.IALLOCATOR_MODE_RELOC:
14938 assert self.relocate_from is not None
14939 assert self.required_nodes == 1
14941 node2group = dict((name, ndata["group"])
14942 for (name, ndata) in self.in_data["nodes"].items())
14944 fn = compat.partial(self._NodesToGroups, node2group,
14945 self.in_data["nodegroups"])
14947 instance = self.cfg.GetInstanceInfo(self.name)
14948 request_groups = fn(self.relocate_from + [instance.primary_node])
14949 result_groups = fn(rdict["result"] + [instance.primary_node])
14951 if self.success and not set(result_groups).issubset(request_groups):
14952 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
14953 " differ from original groups (%s)" %
14954 (utils.CommaJoin(result_groups),
14955 utils.CommaJoin(request_groups)))
14957 elif self.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
14958 assert self.evac_mode in constants.IALLOCATOR_NEVAC_MODES
14960 self.out_data = rdict
14963 def _NodesToGroups(node2group, groups, nodes):
14964 """Returns a list of unique group names for a list of nodes.
14966 @type node2group: dict
14967 @param node2group: Map from node name to group UUID
14969 @param groups: Group information
14971 @param nodes: Node names
14978 group_uuid = node2group[node]
14980 # Ignore unknown node
14984 group = groups[group_uuid]
14986 # Can't find group, let's use UUID
14987 group_name = group_uuid
14989 group_name = group["name"]
14991 result.add(group_name)
14993 return sorted(result)
14996 class LUTestAllocator(NoHooksLU):
14997 """Run allocator tests.
14999 This LU runs the allocator tests
15002 def CheckPrereq(self):
15003 """Check prerequisites.
15005 This checks the opcode parameters depending on the director and mode test.
15008 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
15009 for attr in ["memory", "disks", "disk_template",
15010 "os", "tags", "nics", "vcpus"]:
15011 if not hasattr(self.op, attr):
15012 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
15013 attr, errors.ECODE_INVAL)
15014 iname = self.cfg.ExpandInstanceName(self.op.name)
15015 if iname is not None:
15016 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
15017 iname, errors.ECODE_EXISTS)
15018 if not isinstance(self.op.nics, list):
15019 raise errors.OpPrereqError("Invalid parameter 'nics'",
15020 errors.ECODE_INVAL)
15021 if not isinstance(self.op.disks, list):
15022 raise errors.OpPrereqError("Invalid parameter 'disks'",
15023 errors.ECODE_INVAL)
15024 for row in self.op.disks:
15025 if (not isinstance(row, dict) or
15026 constants.IDISK_SIZE not in row or
15027 not isinstance(row[constants.IDISK_SIZE], int) or
15028 constants.IDISK_MODE not in row or
15029 row[constants.IDISK_MODE] not in constants.DISK_ACCESS_SET):
15030 raise errors.OpPrereqError("Invalid contents of the 'disks'"
15031 " parameter", errors.ECODE_INVAL)
15032 if self.op.hypervisor is None:
15033 self.op.hypervisor = self.cfg.GetHypervisorType()
15034 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
15035 fname = _ExpandInstanceName(self.cfg, self.op.name)
15036 self.op.name = fname
15037 self.relocate_from = \
15038 list(self.cfg.GetInstanceInfo(fname).secondary_nodes)
15039 elif self.op.mode in (constants.IALLOCATOR_MODE_CHG_GROUP,
15040 constants.IALLOCATOR_MODE_NODE_EVAC):
15041 if not self.op.instances:
15042 raise errors.OpPrereqError("Missing instances", errors.ECODE_INVAL)
15043 self.op.instances = _GetWantedInstances(self, self.op.instances)
15045 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
15046 self.op.mode, errors.ECODE_INVAL)
15048 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
15049 if self.op.allocator is None:
15050 raise errors.OpPrereqError("Missing allocator name",
15051 errors.ECODE_INVAL)
15052 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
15053 raise errors.OpPrereqError("Wrong allocator test '%s'" %
15054 self.op.direction, errors.ECODE_INVAL)
15056 def Exec(self, feedback_fn):
15057 """Run the allocator test.
15060 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
15061 ial = IAllocator(self.cfg, self.rpc,
15064 memory=self.op.memory,
15065 disks=self.op.disks,
15066 disk_template=self.op.disk_template,
15070 vcpus=self.op.vcpus,
15071 hypervisor=self.op.hypervisor,
15073 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
15074 ial = IAllocator(self.cfg, self.rpc,
15077 relocate_from=list(self.relocate_from),
15079 elif self.op.mode == constants.IALLOCATOR_MODE_CHG_GROUP:
15080 ial = IAllocator(self.cfg, self.rpc,
15082 instances=self.op.instances,
15083 target_groups=self.op.target_groups)
15084 elif self.op.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
15085 ial = IAllocator(self.cfg, self.rpc,
15087 instances=self.op.instances,
15088 evac_mode=self.op.evac_mode)
15090 raise errors.ProgrammerError("Uncatched mode %s in"
15091 " LUTestAllocator.Exec", self.op.mode)
15093 if self.op.direction == constants.IALLOCATOR_DIR_IN:
15094 result = ial.in_text
15096 ial.Run(self.op.allocator, validate=False)
15097 result = ial.out_text
15101 #: Query type implementations
15103 constants.QR_INSTANCE: _InstanceQuery,
15104 constants.QR_NODE: _NodeQuery,
15105 constants.QR_GROUP: _GroupQuery,
15106 constants.QR_OS: _OsQuery,
15109 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
15112 def _GetQueryImplementation(name):
15113 """Returns the implemtnation for a query type.
15115 @param name: Query type, must be one of L{constants.QR_VIA_OP}
15119 return _QUERY_IMPL[name]
15121 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
15122 errors.ECODE_INVAL)