4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Module implementing the master-side code."""
24 # pylint: disable=W0201,C0302
26 # W0201 since most LU attributes are defined in CheckPrereq or similar
29 # C0302: since we have waaaay to many lines in this module
45 from ganeti import ssh
46 from ganeti import utils
47 from ganeti import errors
48 from ganeti import hypervisor
49 from ganeti import locking
50 from ganeti import constants
51 from ganeti import objects
52 from ganeti import serializer
53 from ganeti import ssconf
54 from ganeti import uidpool
55 from ganeti import compat
56 from ganeti import masterd
57 from ganeti import netutils
58 from ganeti import query
59 from ganeti import qlang
60 from ganeti import opcodes
63 import ganeti.masterd.instance # pylint: disable=W0611
67 """Data container for LU results with jobs.
69 Instances of this class returned from L{LogicalUnit.Exec} will be recognized
70 by L{mcpu.Processor._ProcessResult}. The latter will then submit the jobs
71 contained in the C{jobs} attribute and include the job IDs in the opcode
75 def __init__(self, jobs, **kwargs):
76 """Initializes this class.
78 Additional return values can be specified as keyword arguments.
80 @type jobs: list of lists of L{opcode.OpCode}
81 @param jobs: A list of lists of opcode objects
88 class LogicalUnit(object):
89 """Logical Unit base class.
91 Subclasses must follow these rules:
92 - implement ExpandNames
93 - implement CheckPrereq (except when tasklets are used)
94 - implement Exec (except when tasklets are used)
95 - implement BuildHooksEnv
96 - implement BuildHooksNodes
97 - redefine HPATH and HTYPE
98 - optionally redefine their run requirements:
99 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
101 Note that all commands require root permissions.
103 @ivar dry_run_result: the value (if any) that will be returned to the caller
104 in dry-run mode (signalled by opcode dry_run parameter)
111 def __init__(self, processor, op, context, rpc):
112 """Constructor for LogicalUnit.
114 This needs to be overridden in derived classes in order to check op
118 self.proc = processor
120 self.cfg = context.cfg
121 self.glm = context.glm
123 self.owned_locks = context.glm.list_owned
124 self.context = context
126 # Dicts used to declare locking needs to mcpu
127 self.needed_locks = None
128 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
130 self.remove_locks = {}
131 # Used to force good behavior when calling helper functions
132 self.recalculate_locks = {}
134 self.Log = processor.Log # pylint: disable=C0103
135 self.LogWarning = processor.LogWarning # pylint: disable=C0103
136 self.LogInfo = processor.LogInfo # pylint: disable=C0103
137 self.LogStep = processor.LogStep # pylint: disable=C0103
138 # support for dry-run
139 self.dry_run_result = None
140 # support for generic debug attribute
141 if (not hasattr(self.op, "debug_level") or
142 not isinstance(self.op.debug_level, int)):
143 self.op.debug_level = 0
148 # Validate opcode parameters and set defaults
149 self.op.Validate(True)
151 self.CheckArguments()
153 def CheckArguments(self):
154 """Check syntactic validity for the opcode arguments.
156 This method is for doing a simple syntactic check and ensure
157 validity of opcode parameters, without any cluster-related
158 checks. While the same can be accomplished in ExpandNames and/or
159 CheckPrereq, doing these separate is better because:
161 - ExpandNames is left as as purely a lock-related function
162 - CheckPrereq is run after we have acquired locks (and possible
165 The function is allowed to change the self.op attribute so that
166 later methods can no longer worry about missing parameters.
171 def ExpandNames(self):
172 """Expand names for this LU.
174 This method is called before starting to execute the opcode, and it should
175 update all the parameters of the opcode to their canonical form (e.g. a
176 short node name must be fully expanded after this method has successfully
177 completed). This way locking, hooks, logging, etc. can work correctly.
179 LUs which implement this method must also populate the self.needed_locks
180 member, as a dict with lock levels as keys, and a list of needed lock names
183 - use an empty dict if you don't need any lock
184 - if you don't need any lock at a particular level omit that level
185 - don't put anything for the BGL level
186 - if you want all locks at a level use locking.ALL_SET as a value
188 If you need to share locks (rather than acquire them exclusively) at one
189 level you can modify self.share_locks, setting a true value (usually 1) for
190 that level. By default locks are not shared.
192 This function can also define a list of tasklets, which then will be
193 executed in order instead of the usual LU-level CheckPrereq and Exec
194 functions, if those are not defined by the LU.
198 # Acquire all nodes and one instance
199 self.needed_locks = {
200 locking.LEVEL_NODE: locking.ALL_SET,
201 locking.LEVEL_INSTANCE: ['instance1.example.com'],
203 # Acquire just two nodes
204 self.needed_locks = {
205 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
208 self.needed_locks = {} # No, you can't leave it to the default value None
211 # The implementation of this method is mandatory only if the new LU is
212 # concurrent, so that old LUs don't need to be changed all at the same
215 self.needed_locks = {} # Exclusive LUs don't need locks.
217 raise NotImplementedError
219 def DeclareLocks(self, level):
220 """Declare LU locking needs for a level
222 While most LUs can just declare their locking needs at ExpandNames time,
223 sometimes there's the need to calculate some locks after having acquired
224 the ones before. This function is called just before acquiring locks at a
225 particular level, but after acquiring the ones at lower levels, and permits
226 such calculations. It can be used to modify self.needed_locks, and by
227 default it does nothing.
229 This function is only called if you have something already set in
230 self.needed_locks for the level.
232 @param level: Locking level which is going to be locked
233 @type level: member of ganeti.locking.LEVELS
237 def CheckPrereq(self):
238 """Check prerequisites for this LU.
240 This method should check that the prerequisites for the execution
241 of this LU are fulfilled. It can do internode communication, but
242 it should be idempotent - no cluster or system changes are
245 The method should raise errors.OpPrereqError in case something is
246 not fulfilled. Its return value is ignored.
248 This method should also update all the parameters of the opcode to
249 their canonical form if it hasn't been done by ExpandNames before.
252 if self.tasklets is not None:
253 for (idx, tl) in enumerate(self.tasklets):
254 logging.debug("Checking prerequisites for tasklet %s/%s",
255 idx + 1, len(self.tasklets))
260 def Exec(self, feedback_fn):
263 This method should implement the actual work. It should raise
264 errors.OpExecError for failures that are somewhat dealt with in
268 if self.tasklets is not None:
269 for (idx, tl) in enumerate(self.tasklets):
270 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
273 raise NotImplementedError
275 def BuildHooksEnv(self):
276 """Build hooks environment for this LU.
279 @return: Dictionary containing the environment that will be used for
280 running the hooks for this LU. The keys of the dict must not be prefixed
281 with "GANETI_"--that'll be added by the hooks runner. The hooks runner
282 will extend the environment with additional variables. If no environment
283 should be defined, an empty dictionary should be returned (not C{None}).
284 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
288 raise NotImplementedError
290 def BuildHooksNodes(self):
291 """Build list of nodes to run LU's hooks.
293 @rtype: tuple; (list, list)
294 @return: Tuple containing a list of node names on which the hook
295 should run before the execution and a list of node names on which the
296 hook should run after the execution. No nodes should be returned as an
297 empty list (and not None).
298 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
302 raise NotImplementedError
304 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
305 """Notify the LU about the results of its hooks.
307 This method is called every time a hooks phase is executed, and notifies
308 the Logical Unit about the hooks' result. The LU can then use it to alter
309 its result based on the hooks. By default the method does nothing and the
310 previous result is passed back unchanged but any LU can define it if it
311 wants to use the local cluster hook-scripts somehow.
313 @param phase: one of L{constants.HOOKS_PHASE_POST} or
314 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
315 @param hook_results: the results of the multi-node hooks rpc call
316 @param feedback_fn: function used send feedback back to the caller
317 @param lu_result: the previous Exec result this LU had, or None
319 @return: the new Exec result, based on the previous result
323 # API must be kept, thus we ignore the unused argument and could
324 # be a function warnings
325 # pylint: disable=W0613,R0201
328 def _ExpandAndLockInstance(self):
329 """Helper function to expand and lock an instance.
331 Many LUs that work on an instance take its name in self.op.instance_name
332 and need to expand it and then declare the expanded name for locking. This
333 function does it, and then updates self.op.instance_name to the expanded
334 name. It also initializes needed_locks as a dict, if this hasn't been done
338 if self.needed_locks is None:
339 self.needed_locks = {}
341 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
342 "_ExpandAndLockInstance called with instance-level locks set"
343 self.op.instance_name = _ExpandInstanceName(self.cfg,
344 self.op.instance_name)
345 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
347 def _LockInstancesNodes(self, primary_only=False):
348 """Helper function to declare instances' nodes for locking.
350 This function should be called after locking one or more instances to lock
351 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
352 with all primary or secondary nodes for instances already locked and
353 present in self.needed_locks[locking.LEVEL_INSTANCE].
355 It should be called from DeclareLocks, and for safety only works if
356 self.recalculate_locks[locking.LEVEL_NODE] is set.
358 In the future it may grow parameters to just lock some instance's nodes, or
359 to just lock primaries or secondary nodes, if needed.
361 If should be called in DeclareLocks in a way similar to::
363 if level == locking.LEVEL_NODE:
364 self._LockInstancesNodes()
366 @type primary_only: boolean
367 @param primary_only: only lock primary nodes of locked instances
370 assert locking.LEVEL_NODE in self.recalculate_locks, \
371 "_LockInstancesNodes helper function called with no nodes to recalculate"
373 # TODO: check if we're really been called with the instance locks held
375 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
376 # future we might want to have different behaviors depending on the value
377 # of self.recalculate_locks[locking.LEVEL_NODE]
379 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
380 for _, instance in self.cfg.GetMultiInstanceInfo(locked_i):
381 wanted_nodes.append(instance.primary_node)
383 wanted_nodes.extend(instance.secondary_nodes)
385 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
386 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
387 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
388 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
390 del self.recalculate_locks[locking.LEVEL_NODE]
393 class NoHooksLU(LogicalUnit): # pylint: disable=W0223
394 """Simple LU which runs no hooks.
396 This LU is intended as a parent for other LogicalUnits which will
397 run no hooks, in order to reduce duplicate code.
403 def BuildHooksEnv(self):
404 """Empty BuildHooksEnv for NoHooksLu.
406 This just raises an error.
409 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
411 def BuildHooksNodes(self):
412 """Empty BuildHooksNodes for NoHooksLU.
415 raise AssertionError("BuildHooksNodes called for NoHooksLU")
419 """Tasklet base class.
421 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
422 they can mix legacy code with tasklets. Locking needs to be done in the LU,
423 tasklets know nothing about locks.
425 Subclasses must follow these rules:
426 - Implement CheckPrereq
430 def __init__(self, lu):
437 def CheckPrereq(self):
438 """Check prerequisites for this tasklets.
440 This method should check whether the prerequisites for the execution of
441 this tasklet are fulfilled. It can do internode communication, but it
442 should be idempotent - no cluster or system changes are allowed.
444 The method should raise errors.OpPrereqError in case something is not
445 fulfilled. Its return value is ignored.
447 This method should also update all parameters to their canonical form if it
448 hasn't been done before.
453 def Exec(self, feedback_fn):
454 """Execute the tasklet.
456 This method should implement the actual work. It should raise
457 errors.OpExecError for failures that are somewhat dealt with in code, or
461 raise NotImplementedError
465 """Base for query utility classes.
468 #: Attribute holding field definitions
471 def __init__(self, filter_, fields, use_locking):
472 """Initializes this class.
475 self.use_locking = use_locking
477 self.query = query.Query(self.FIELDS, fields, filter_=filter_,
479 self.requested_data = self.query.RequestedData()
480 self.names = self.query.RequestedNames()
482 # Sort only if no names were requested
483 self.sort_by_name = not self.names
485 self.do_locking = None
488 def _GetNames(self, lu, all_names, lock_level):
489 """Helper function to determine names asked for in the query.
493 names = lu.owned_locks(lock_level)
497 if self.wanted == locking.ALL_SET:
498 assert not self.names
499 # caller didn't specify names, so ordering is not important
500 return utils.NiceSort(names)
502 # caller specified names and we must keep the same order
504 assert not self.do_locking or lu.glm.is_owned(lock_level)
506 missing = set(self.wanted).difference(names)
508 raise errors.OpExecError("Some items were removed before retrieving"
509 " their data: %s" % missing)
511 # Return expanded names
514 def ExpandNames(self, lu):
515 """Expand names for this query.
517 See L{LogicalUnit.ExpandNames}.
520 raise NotImplementedError()
522 def DeclareLocks(self, lu, level):
523 """Declare locks for this query.
525 See L{LogicalUnit.DeclareLocks}.
528 raise NotImplementedError()
530 def _GetQueryData(self, lu):
531 """Collects all data for this query.
533 @return: Query data object
536 raise NotImplementedError()
538 def NewStyleQuery(self, lu):
539 """Collect data and execute query.
542 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
543 sort_by_name=self.sort_by_name)
545 def OldStyleQuery(self, lu):
546 """Collect data and execute query.
549 return self.query.OldStyleQuery(self._GetQueryData(lu),
550 sort_by_name=self.sort_by_name)
554 """Returns a dict declaring all lock levels shared.
557 return dict.fromkeys(locking.LEVELS, 1)
560 def _CheckInstanceNodeGroups(cfg, instance_name, owned_groups):
561 """Checks if the owned node groups are still correct for an instance.
563 @type cfg: L{config.ConfigWriter}
564 @param cfg: The cluster configuration
565 @type instance_name: string
566 @param instance_name: Instance name
567 @type owned_groups: set or frozenset
568 @param owned_groups: List of currently owned node groups
571 inst_groups = cfg.GetInstanceNodeGroups(instance_name)
573 if not owned_groups.issuperset(inst_groups):
574 raise errors.OpPrereqError("Instance %s's node groups changed since"
575 " locks were acquired, current groups are"
576 " are '%s', owning groups '%s'; retry the"
579 utils.CommaJoin(inst_groups),
580 utils.CommaJoin(owned_groups)),
586 def _CheckNodeGroupInstances(cfg, group_uuid, owned_instances):
587 """Checks if the instances in a node group are still correct.
589 @type cfg: L{config.ConfigWriter}
590 @param cfg: The cluster configuration
591 @type group_uuid: string
592 @param group_uuid: Node group UUID
593 @type owned_instances: set or frozenset
594 @param owned_instances: List of currently owned instances
597 wanted_instances = cfg.GetNodeGroupInstances(group_uuid)
598 if owned_instances != wanted_instances:
599 raise errors.OpPrereqError("Instances in node group '%s' changed since"
600 " locks were acquired, wanted '%s', have '%s';"
601 " retry the operation" %
603 utils.CommaJoin(wanted_instances),
604 utils.CommaJoin(owned_instances)),
607 return wanted_instances
610 def _SupportsOob(cfg, node):
611 """Tells if node supports OOB.
613 @type cfg: L{config.ConfigWriter}
614 @param cfg: The cluster configuration
615 @type node: L{objects.Node}
616 @param node: The node
617 @return: The OOB script if supported or an empty string otherwise
620 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
623 def _GetWantedNodes(lu, nodes):
624 """Returns list of checked and expanded node names.
626 @type lu: L{LogicalUnit}
627 @param lu: the logical unit on whose behalf we execute
629 @param nodes: list of node names or None for all nodes
631 @return: the list of nodes, sorted
632 @raise errors.ProgrammerError: if the nodes parameter is wrong type
636 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
638 return utils.NiceSort(lu.cfg.GetNodeList())
641 def _GetWantedInstances(lu, instances):
642 """Returns list of checked and expanded instance names.
644 @type lu: L{LogicalUnit}
645 @param lu: the logical unit on whose behalf we execute
646 @type instances: list
647 @param instances: list of instance names or None for all instances
649 @return: the list of instances, sorted
650 @raise errors.OpPrereqError: if the instances parameter is wrong type
651 @raise errors.OpPrereqError: if any of the passed instances is not found
655 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
657 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
661 def _GetUpdatedParams(old_params, update_dict,
662 use_default=True, use_none=False):
663 """Return the new version of a parameter dictionary.
665 @type old_params: dict
666 @param old_params: old parameters
667 @type update_dict: dict
668 @param update_dict: dict containing new parameter values, or
669 constants.VALUE_DEFAULT to reset the parameter to its default
671 @param use_default: boolean
672 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
673 values as 'to be deleted' values
674 @param use_none: boolean
675 @type use_none: whether to recognise C{None} values as 'to be
678 @return: the new parameter dictionary
681 params_copy = copy.deepcopy(old_params)
682 for key, val in update_dict.iteritems():
683 if ((use_default and val == constants.VALUE_DEFAULT) or
684 (use_none and val is None)):
690 params_copy[key] = val
694 def _ReleaseLocks(lu, level, names=None, keep=None):
695 """Releases locks owned by an LU.
697 @type lu: L{LogicalUnit}
698 @param level: Lock level
699 @type names: list or None
700 @param names: Names of locks to release
701 @type keep: list or None
702 @param keep: Names of locks to retain
705 assert not (keep is not None and names is not None), \
706 "Only one of the 'names' and the 'keep' parameters can be given"
708 if names is not None:
709 should_release = names.__contains__
711 should_release = lambda name: name not in keep
713 should_release = None
719 # Determine which locks to release
720 for name in lu.owned_locks(level):
721 if should_release(name):
726 assert len(lu.owned_locks(level)) == (len(retain) + len(release))
728 # Release just some locks
729 lu.glm.release(level, names=release)
731 assert frozenset(lu.owned_locks(level)) == frozenset(retain)
734 lu.glm.release(level)
736 assert not lu.glm.is_owned(level), "No locks should be owned"
739 def _MapInstanceDisksToNodes(instances):
740 """Creates a map from (node, volume) to instance name.
742 @type instances: list of L{objects.Instance}
743 @rtype: dict; tuple of (node name, volume name) as key, instance name as value
746 return dict(((node, vol), inst.name)
747 for inst in instances
748 for (node, vols) in inst.MapLVsByNode().items()
752 def _RunPostHook(lu, node_name):
753 """Runs the post-hook for an opcode on a single node.
756 hm = lu.proc.hmclass(lu.rpc.call_hooks_runner, lu)
758 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
760 # pylint: disable=W0702
761 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
764 def _CheckOutputFields(static, dynamic, selected):
765 """Checks whether all selected fields are valid.
767 @type static: L{utils.FieldSet}
768 @param static: static fields set
769 @type dynamic: L{utils.FieldSet}
770 @param dynamic: dynamic fields set
777 delta = f.NonMatching(selected)
779 raise errors.OpPrereqError("Unknown output fields selected: %s"
780 % ",".join(delta), errors.ECODE_INVAL)
783 def _CheckGlobalHvParams(params):
784 """Validates that given hypervisor params are not global ones.
786 This will ensure that instances don't get customised versions of
790 used_globals = constants.HVC_GLOBALS.intersection(params)
792 msg = ("The following hypervisor parameters are global and cannot"
793 " be customized at instance level, please modify them at"
794 " cluster level: %s" % utils.CommaJoin(used_globals))
795 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
798 def _CheckNodeOnline(lu, node, msg=None):
799 """Ensure that a given node is online.
801 @param lu: the LU on behalf of which we make the check
802 @param node: the node to check
803 @param msg: if passed, should be a message to replace the default one
804 @raise errors.OpPrereqError: if the node is offline
808 msg = "Can't use offline node"
809 if lu.cfg.GetNodeInfo(node).offline:
810 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
813 def _CheckNodeNotDrained(lu, node):
814 """Ensure that a given node is not drained.
816 @param lu: the LU on behalf of which we make the check
817 @param node: the node to check
818 @raise errors.OpPrereqError: if the node is drained
821 if lu.cfg.GetNodeInfo(node).drained:
822 raise errors.OpPrereqError("Can't use drained node %s" % node,
826 def _CheckNodeVmCapable(lu, node):
827 """Ensure that a given node is vm capable.
829 @param lu: the LU on behalf of which we make the check
830 @param node: the node to check
831 @raise errors.OpPrereqError: if the node is not vm capable
834 if not lu.cfg.GetNodeInfo(node).vm_capable:
835 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
839 def _CheckNodeHasOS(lu, node, os_name, force_variant):
840 """Ensure that a node supports a given OS.
842 @param lu: the LU on behalf of which we make the check
843 @param node: the node to check
844 @param os_name: the OS to query about
845 @param force_variant: whether to ignore variant errors
846 @raise errors.OpPrereqError: if the node is not supporting the OS
849 result = lu.rpc.call_os_get(node, os_name)
850 result.Raise("OS '%s' not in supported OS list for node %s" %
852 prereq=True, ecode=errors.ECODE_INVAL)
853 if not force_variant:
854 _CheckOSVariant(result.payload, os_name)
857 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
858 """Ensure that a node has the given secondary ip.
860 @type lu: L{LogicalUnit}
861 @param lu: the LU on behalf of which we make the check
863 @param node: the node to check
864 @type secondary_ip: string
865 @param secondary_ip: the ip to check
866 @type prereq: boolean
867 @param prereq: whether to throw a prerequisite or an execute error
868 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
869 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
872 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
873 result.Raise("Failure checking secondary ip on node %s" % node,
874 prereq=prereq, ecode=errors.ECODE_ENVIRON)
875 if not result.payload:
876 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
877 " please fix and re-run this command" % secondary_ip)
879 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
881 raise errors.OpExecError(msg)
884 def _GetClusterDomainSecret():
885 """Reads the cluster domain secret.
888 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
892 def _CheckInstanceDown(lu, instance, reason):
893 """Ensure that an instance is not running."""
894 if instance.admin_up:
895 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
896 (instance.name, reason), errors.ECODE_STATE)
898 pnode = instance.primary_node
899 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
900 ins_l.Raise("Can't contact node %s for instance information" % pnode,
901 prereq=True, ecode=errors.ECODE_ENVIRON)
903 if instance.name in ins_l.payload:
904 raise errors.OpPrereqError("Instance %s is running, %s" %
905 (instance.name, reason), errors.ECODE_STATE)
908 def _ExpandItemName(fn, name, kind):
909 """Expand an item name.
911 @param fn: the function to use for expansion
912 @param name: requested item name
913 @param kind: text description ('Node' or 'Instance')
914 @return: the resolved (full) name
915 @raise errors.OpPrereqError: if the item is not found
919 if full_name is None:
920 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
925 def _ExpandNodeName(cfg, name):
926 """Wrapper over L{_ExpandItemName} for nodes."""
927 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
930 def _ExpandInstanceName(cfg, name):
931 """Wrapper over L{_ExpandItemName} for instance."""
932 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
935 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
936 memory, vcpus, nics, disk_template, disks,
937 bep, hvp, hypervisor_name, tags):
938 """Builds instance related env variables for hooks
940 This builds the hook environment from individual variables.
943 @param name: the name of the instance
944 @type primary_node: string
945 @param primary_node: the name of the instance's primary node
946 @type secondary_nodes: list
947 @param secondary_nodes: list of secondary nodes as strings
948 @type os_type: string
949 @param os_type: the name of the instance's OS
950 @type status: boolean
951 @param status: the should_run status of the instance
953 @param memory: the memory size of the instance
955 @param vcpus: the count of VCPUs the instance has
957 @param nics: list of tuples (ip, mac, mode, link) representing
958 the NICs the instance has
959 @type disk_template: string
960 @param disk_template: the disk template of the instance
962 @param disks: the list of (size, mode) pairs
964 @param bep: the backend parameters for the instance
966 @param hvp: the hypervisor parameters for the instance
967 @type hypervisor_name: string
968 @param hypervisor_name: the hypervisor for the instance
970 @param tags: list of instance tags as strings
972 @return: the hook environment for this instance
981 "INSTANCE_NAME": name,
982 "INSTANCE_PRIMARY": primary_node,
983 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
984 "INSTANCE_OS_TYPE": os_type,
985 "INSTANCE_STATUS": str_status,
986 "INSTANCE_MEMORY": memory,
987 "INSTANCE_VCPUS": vcpus,
988 "INSTANCE_DISK_TEMPLATE": disk_template,
989 "INSTANCE_HYPERVISOR": hypervisor_name,
993 nic_count = len(nics)
994 for idx, (ip, mac, mode, link) in enumerate(nics):
997 env["INSTANCE_NIC%d_IP" % idx] = ip
998 env["INSTANCE_NIC%d_MAC" % idx] = mac
999 env["INSTANCE_NIC%d_MODE" % idx] = mode
1000 env["INSTANCE_NIC%d_LINK" % idx] = link
1001 if mode == constants.NIC_MODE_BRIDGED:
1002 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
1006 env["INSTANCE_NIC_COUNT"] = nic_count
1009 disk_count = len(disks)
1010 for idx, (size, mode) in enumerate(disks):
1011 env["INSTANCE_DISK%d_SIZE" % idx] = size
1012 env["INSTANCE_DISK%d_MODE" % idx] = mode
1016 env["INSTANCE_DISK_COUNT"] = disk_count
1021 env["INSTANCE_TAGS"] = " ".join(tags)
1023 for source, kind in [(bep, "BE"), (hvp, "HV")]:
1024 for key, value in source.items():
1025 env["INSTANCE_%s_%s" % (kind, key)] = value
1030 def _NICListToTuple(lu, nics):
1031 """Build a list of nic information tuples.
1033 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
1034 value in LUInstanceQueryData.
1036 @type lu: L{LogicalUnit}
1037 @param lu: the logical unit on whose behalf we execute
1038 @type nics: list of L{objects.NIC}
1039 @param nics: list of nics to convert to hooks tuples
1043 cluster = lu.cfg.GetClusterInfo()
1047 filled_params = cluster.SimpleFillNIC(nic.nicparams)
1048 mode = filled_params[constants.NIC_MODE]
1049 link = filled_params[constants.NIC_LINK]
1050 hooks_nics.append((ip, mac, mode, link))
1054 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1055 """Builds instance related env variables for hooks from an object.
1057 @type lu: L{LogicalUnit}
1058 @param lu: the logical unit on whose behalf we execute
1059 @type instance: L{objects.Instance}
1060 @param instance: the instance for which we should build the
1062 @type override: dict
1063 @param override: dictionary with key/values that will override
1066 @return: the hook environment dictionary
1069 cluster = lu.cfg.GetClusterInfo()
1070 bep = cluster.FillBE(instance)
1071 hvp = cluster.FillHV(instance)
1073 "name": instance.name,
1074 "primary_node": instance.primary_node,
1075 "secondary_nodes": instance.secondary_nodes,
1076 "os_type": instance.os,
1077 "status": instance.admin_up,
1078 "memory": bep[constants.BE_MEMORY],
1079 "vcpus": bep[constants.BE_VCPUS],
1080 "nics": _NICListToTuple(lu, instance.nics),
1081 "disk_template": instance.disk_template,
1082 "disks": [(disk.size, disk.mode) for disk in instance.disks],
1085 "hypervisor_name": instance.hypervisor,
1086 "tags": instance.tags,
1089 args.update(override)
1090 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
1093 def _AdjustCandidatePool(lu, exceptions):
1094 """Adjust the candidate pool after node operations.
1097 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1099 lu.LogInfo("Promoted nodes to master candidate role: %s",
1100 utils.CommaJoin(node.name for node in mod_list))
1101 for name in mod_list:
1102 lu.context.ReaddNode(name)
1103 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1105 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1109 def _DecideSelfPromotion(lu, exceptions=None):
1110 """Decide whether I should promote myself as a master candidate.
1113 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1114 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1115 # the new node will increase mc_max with one, so:
1116 mc_should = min(mc_should + 1, cp_size)
1117 return mc_now < mc_should
1120 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1121 """Check that the brigdes needed by a list of nics exist.
1124 cluster = lu.cfg.GetClusterInfo()
1125 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1126 brlist = [params[constants.NIC_LINK] for params in paramslist
1127 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1129 result = lu.rpc.call_bridges_exist(target_node, brlist)
1130 result.Raise("Error checking bridges on destination node '%s'" %
1131 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1134 def _CheckInstanceBridgesExist(lu, instance, node=None):
1135 """Check that the brigdes needed by an instance exist.
1139 node = instance.primary_node
1140 _CheckNicsBridgesExist(lu, instance.nics, node)
1143 def _CheckOSVariant(os_obj, name):
1144 """Check whether an OS name conforms to the os variants specification.
1146 @type os_obj: L{objects.OS}
1147 @param os_obj: OS object to check
1149 @param name: OS name passed by the user, to check for validity
1152 variant = objects.OS.GetVariant(name)
1153 if not os_obj.supported_variants:
1155 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'"
1156 " passed)" % (os_obj.name, variant),
1160 raise errors.OpPrereqError("OS name must include a variant",
1163 if variant not in os_obj.supported_variants:
1164 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1167 def _GetNodeInstancesInner(cfg, fn):
1168 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1171 def _GetNodeInstances(cfg, node_name):
1172 """Returns a list of all primary and secondary instances on a node.
1176 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1179 def _GetNodePrimaryInstances(cfg, node_name):
1180 """Returns primary instances on a node.
1183 return _GetNodeInstancesInner(cfg,
1184 lambda inst: node_name == inst.primary_node)
1187 def _GetNodeSecondaryInstances(cfg, node_name):
1188 """Returns secondary instances on a node.
1191 return _GetNodeInstancesInner(cfg,
1192 lambda inst: node_name in inst.secondary_nodes)
1195 def _GetStorageTypeArgs(cfg, storage_type):
1196 """Returns the arguments for a storage type.
1199 # Special case for file storage
1200 if storage_type == constants.ST_FILE:
1201 # storage.FileStorage wants a list of storage directories
1202 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1207 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1210 for dev in instance.disks:
1211 cfg.SetDiskID(dev, node_name)
1213 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1214 result.Raise("Failed to get disk status from node %s" % node_name,
1215 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1217 for idx, bdev_status in enumerate(result.payload):
1218 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1224 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1225 """Check the sanity of iallocator and node arguments and use the
1226 cluster-wide iallocator if appropriate.
1228 Check that at most one of (iallocator, node) is specified. If none is
1229 specified, then the LU's opcode's iallocator slot is filled with the
1230 cluster-wide default iallocator.
1232 @type iallocator_slot: string
1233 @param iallocator_slot: the name of the opcode iallocator slot
1234 @type node_slot: string
1235 @param node_slot: the name of the opcode target node slot
1238 node = getattr(lu.op, node_slot, None)
1239 iallocator = getattr(lu.op, iallocator_slot, None)
1241 if node is not None and iallocator is not None:
1242 raise errors.OpPrereqError("Do not specify both, iallocator and node",
1244 elif node is None and iallocator is None:
1245 default_iallocator = lu.cfg.GetDefaultIAllocator()
1246 if default_iallocator:
1247 setattr(lu.op, iallocator_slot, default_iallocator)
1249 raise errors.OpPrereqError("No iallocator or node given and no"
1250 " cluster-wide default iallocator found;"
1251 " please specify either an iallocator or a"
1252 " node, or set a cluster-wide default"
1256 def _GetDefaultIAllocator(cfg, iallocator):
1257 """Decides on which iallocator to use.
1259 @type cfg: L{config.ConfigWriter}
1260 @param cfg: Cluster configuration object
1261 @type iallocator: string or None
1262 @param iallocator: Iallocator specified in opcode
1264 @return: Iallocator name
1268 # Use default iallocator
1269 iallocator = cfg.GetDefaultIAllocator()
1272 raise errors.OpPrereqError("No iallocator was specified, neither in the"
1273 " opcode nor as a cluster-wide default",
1279 class LUClusterPostInit(LogicalUnit):
1280 """Logical unit for running hooks after cluster initialization.
1283 HPATH = "cluster-init"
1284 HTYPE = constants.HTYPE_CLUSTER
1286 def BuildHooksEnv(self):
1291 "OP_TARGET": self.cfg.GetClusterName(),
1294 def BuildHooksNodes(self):
1295 """Build hooks nodes.
1298 return ([], [self.cfg.GetMasterNode()])
1300 def Exec(self, feedback_fn):
1307 class LUClusterDestroy(LogicalUnit):
1308 """Logical unit for destroying the cluster.
1311 HPATH = "cluster-destroy"
1312 HTYPE = constants.HTYPE_CLUSTER
1314 def BuildHooksEnv(self):
1319 "OP_TARGET": self.cfg.GetClusterName(),
1322 def BuildHooksNodes(self):
1323 """Build hooks nodes.
1328 def CheckPrereq(self):
1329 """Check prerequisites.
1331 This checks whether the cluster is empty.
1333 Any errors are signaled by raising errors.OpPrereqError.
1336 master = self.cfg.GetMasterNode()
1338 nodelist = self.cfg.GetNodeList()
1339 if len(nodelist) != 1 or nodelist[0] != master:
1340 raise errors.OpPrereqError("There are still %d node(s) in"
1341 " this cluster." % (len(nodelist) - 1),
1343 instancelist = self.cfg.GetInstanceList()
1345 raise errors.OpPrereqError("There are still %d instance(s) in"
1346 " this cluster." % len(instancelist),
1349 def Exec(self, feedback_fn):
1350 """Destroys the cluster.
1353 master = self.cfg.GetMasterNode()
1355 # Run post hooks on master node before it's removed
1356 _RunPostHook(self, master)
1358 result = self.rpc.call_node_stop_master(master, False)
1359 result.Raise("Could not disable the master role")
1364 def _VerifyCertificate(filename):
1365 """Verifies a certificate for L{LUClusterVerifyConfig}.
1367 @type filename: string
1368 @param filename: Path to PEM file
1372 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1373 utils.ReadFile(filename))
1374 except Exception, err: # pylint: disable=W0703
1375 return (LUClusterVerifyConfig.ETYPE_ERROR,
1376 "Failed to load X509 certificate %s: %s" % (filename, err))
1379 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1380 constants.SSL_CERT_EXPIRATION_ERROR)
1383 fnamemsg = "While verifying %s: %s" % (filename, msg)
1388 return (None, fnamemsg)
1389 elif errcode == utils.CERT_WARNING:
1390 return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg)
1391 elif errcode == utils.CERT_ERROR:
1392 return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg)
1394 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1397 def _GetAllHypervisorParameters(cluster, instances):
1398 """Compute the set of all hypervisor parameters.
1400 @type cluster: L{objects.Cluster}
1401 @param cluster: the cluster object
1402 @param instances: list of L{objects.Instance}
1403 @param instances: additional instances from which to obtain parameters
1404 @rtype: list of (origin, hypervisor, parameters)
1405 @return: a list with all parameters found, indicating the hypervisor they
1406 apply to, and the origin (can be "cluster", "os X", or "instance Y")
1411 for hv_name in cluster.enabled_hypervisors:
1412 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
1414 for os_name, os_hvp in cluster.os_hvp.items():
1415 for hv_name, hv_params in os_hvp.items():
1417 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
1418 hvp_data.append(("os %s" % os_name, hv_name, full_params))
1420 # TODO: collapse identical parameter values in a single one
1421 for instance in instances:
1422 if instance.hvparams:
1423 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
1424 cluster.FillHV(instance)))
1429 class _VerifyErrors(object):
1430 """Mix-in for cluster/group verify LUs.
1432 It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects
1433 self.op and self._feedback_fn to be available.)
1436 TCLUSTER = "cluster"
1438 TINSTANCE = "instance"
1440 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1441 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1442 ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK")
1443 ECLUSTERDANGLINGNODES = (TNODE, "ECLUSTERDANGLINGNODES")
1444 ECLUSTERDANGLINGINST = (TNODE, "ECLUSTERDANGLINGINST")
1445 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1446 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1447 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1448 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1449 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1450 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1451 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1452 ENODEDRBD = (TNODE, "ENODEDRBD")
1453 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1454 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1455 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1456 ENODEHV = (TNODE, "ENODEHV")
1457 ENODELVM = (TNODE, "ENODELVM")
1458 ENODEN1 = (TNODE, "ENODEN1")
1459 ENODENET = (TNODE, "ENODENET")
1460 ENODEOS = (TNODE, "ENODEOS")
1461 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1462 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1463 ENODERPC = (TNODE, "ENODERPC")
1464 ENODESSH = (TNODE, "ENODESSH")
1465 ENODEVERSION = (TNODE, "ENODEVERSION")
1466 ENODESETUP = (TNODE, "ENODESETUP")
1467 ENODETIME = (TNODE, "ENODETIME")
1468 ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1470 ETYPE_FIELD = "code"
1471 ETYPE_ERROR = "ERROR"
1472 ETYPE_WARNING = "WARNING"
1474 def _Error(self, ecode, item, msg, *args, **kwargs):
1475 """Format an error message.
1477 Based on the opcode's error_codes parameter, either format a
1478 parseable error code, or a simpler error string.
1480 This must be called only from Exec and functions called from Exec.
1483 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1485 # first complete the msg
1488 # then format the whole message
1489 if self.op.error_codes: # This is a mix-in. pylint: disable=E1101
1490 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1496 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1497 # and finally report it via the feedback_fn
1498 self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable=E1101
1500 def _ErrorIf(self, cond, *args, **kwargs):
1501 """Log an error message if the passed condition is True.
1505 or self.op.debug_simulate_errors) # pylint: disable=E1101
1507 self._Error(*args, **kwargs)
1508 # do not mark the operation as failed for WARN cases only
1509 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1510 self.bad = self.bad or cond
1513 class LUClusterVerify(NoHooksLU):
1514 """Submits all jobs necessary to verify the cluster.
1519 def ExpandNames(self):
1520 self.needed_locks = {}
1522 def Exec(self, feedback_fn):
1525 if self.op.group_name:
1526 groups = [self.op.group_name]
1527 depends_fn = lambda: None
1529 groups = self.cfg.GetNodeGroupList()
1531 # Verify global configuration
1532 jobs.append([opcodes.OpClusterVerifyConfig()])
1534 # Always depend on global verification
1535 depends_fn = lambda: [(-len(jobs), [])]
1537 jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,
1538 depends=depends_fn())]
1539 for group in groups)
1541 # Fix up all parameters
1542 for op in itertools.chain(*jobs): # pylint: disable=W0142
1543 op.debug_simulate_errors = self.op.debug_simulate_errors
1544 op.verbose = self.op.verbose
1545 op.error_codes = self.op.error_codes
1547 op.skip_checks = self.op.skip_checks
1548 except AttributeError:
1549 assert not isinstance(op, opcodes.OpClusterVerifyGroup)
1551 return ResultWithJobs(jobs)
1554 class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
1555 """Verifies the cluster config.
1560 def _VerifyHVP(self, hvp_data):
1561 """Verifies locally the syntax of the hypervisor parameters.
1564 for item, hv_name, hv_params in hvp_data:
1565 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
1568 hv_class = hypervisor.GetHypervisor(hv_name)
1569 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1570 hv_class.CheckParameterSyntax(hv_params)
1571 except errors.GenericError, err:
1572 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
1574 def ExpandNames(self):
1575 self.needed_locks = dict.fromkeys(locking.LEVELS, locking.ALL_SET)
1576 self.share_locks = _ShareAll()
1578 def CheckPrereq(self):
1579 """Check prerequisites.
1582 # Retrieve all information
1583 self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
1584 self.all_node_info = self.cfg.GetAllNodesInfo()
1585 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1587 def Exec(self, feedback_fn):
1588 """Verify integrity of cluster, performing various test on nodes.
1592 self._feedback_fn = feedback_fn
1594 feedback_fn("* Verifying cluster config")
1596 for msg in self.cfg.VerifyConfig():
1597 self._ErrorIf(True, self.ECLUSTERCFG, None, msg)
1599 feedback_fn("* Verifying cluster certificate files")
1601 for cert_filename in constants.ALL_CERT_FILES:
1602 (errcode, msg) = _VerifyCertificate(cert_filename)
1603 self._ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
1605 feedback_fn("* Verifying hypervisor parameters")
1607 self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(),
1608 self.all_inst_info.values()))
1610 feedback_fn("* Verifying all nodes belong to an existing group")
1612 # We do this verification here because, should this bogus circumstance
1613 # occur, it would never be caught by VerifyGroup, which only acts on
1614 # nodes/instances reachable from existing node groups.
1616 dangling_nodes = set(node.name for node in self.all_node_info.values()
1617 if node.group not in self.all_group_info)
1619 dangling_instances = {}
1620 no_node_instances = []
1622 for inst in self.all_inst_info.values():
1623 if inst.primary_node in dangling_nodes:
1624 dangling_instances.setdefault(inst.primary_node, []).append(inst.name)
1625 elif inst.primary_node not in self.all_node_info:
1626 no_node_instances.append(inst.name)
1631 utils.CommaJoin(dangling_instances.get(node.name,
1633 for node in dangling_nodes]
1635 self._ErrorIf(bool(dangling_nodes), self.ECLUSTERDANGLINGNODES, None,
1636 "the following nodes (and their instances) belong to a non"
1637 " existing group: %s", utils.CommaJoin(pretty_dangling))
1639 self._ErrorIf(bool(no_node_instances), self.ECLUSTERDANGLINGINST, None,
1640 "the following instances have a non-existing primary-node:"
1641 " %s", utils.CommaJoin(no_node_instances))
1646 class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
1647 """Verifies the status of a node group.
1650 HPATH = "cluster-verify"
1651 HTYPE = constants.HTYPE_CLUSTER
1654 _HOOKS_INDENT_RE = re.compile("^", re.M)
1656 class NodeImage(object):
1657 """A class representing the logical and physical status of a node.
1660 @ivar name: the node name to which this object refers
1661 @ivar volumes: a structure as returned from
1662 L{ganeti.backend.GetVolumeList} (runtime)
1663 @ivar instances: a list of running instances (runtime)
1664 @ivar pinst: list of configured primary instances (config)
1665 @ivar sinst: list of configured secondary instances (config)
1666 @ivar sbp: dictionary of {primary-node: list of instances} for all
1667 instances for which this node is secondary (config)
1668 @ivar mfree: free memory, as reported by hypervisor (runtime)
1669 @ivar dfree: free disk, as reported by the node (runtime)
1670 @ivar offline: the offline status (config)
1671 @type rpc_fail: boolean
1672 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1673 not whether the individual keys were correct) (runtime)
1674 @type lvm_fail: boolean
1675 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1676 @type hyp_fail: boolean
1677 @ivar hyp_fail: whether the RPC call didn't return the instance list
1678 @type ghost: boolean
1679 @ivar ghost: whether this is a known node or not (config)
1680 @type os_fail: boolean
1681 @ivar os_fail: whether the RPC call didn't return valid OS data
1683 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1684 @type vm_capable: boolean
1685 @ivar vm_capable: whether the node can host instances
1688 def __init__(self, offline=False, name=None, vm_capable=True):
1697 self.offline = offline
1698 self.vm_capable = vm_capable
1699 self.rpc_fail = False
1700 self.lvm_fail = False
1701 self.hyp_fail = False
1703 self.os_fail = False
1706 def ExpandNames(self):
1707 # This raises errors.OpPrereqError on its own:
1708 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
1710 # Get instances in node group; this is unsafe and needs verification later
1712 self.cfg.GetNodeGroupInstances(self.group_uuid, primary_only=True)
1714 self.needed_locks = {
1715 locking.LEVEL_INSTANCE: inst_names,
1716 locking.LEVEL_NODEGROUP: [self.group_uuid],
1717 locking.LEVEL_NODE: [],
1720 self.share_locks = _ShareAll()
1722 def DeclareLocks(self, level):
1723 if level == locking.LEVEL_NODE:
1724 # Get members of node group; this is unsafe and needs verification later
1725 nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members)
1727 all_inst_info = self.cfg.GetAllInstancesInfo()
1729 # In Exec(), we warn about mirrored instances that have primary and
1730 # secondary living in separate node groups. To fully verify that
1731 # volumes for these instances are healthy, we will need to do an
1732 # extra call to their secondaries. We ensure here those nodes will
1734 for inst in self.owned_locks(locking.LEVEL_INSTANCE):
1735 # Important: access only the instances whose lock is owned
1736 if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR:
1737 nodes.update(all_inst_info[inst].secondary_nodes)
1739 self.needed_locks[locking.LEVEL_NODE] = nodes
1741 def CheckPrereq(self):
1742 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
1743 self.group_info = self.cfg.GetNodeGroup(self.group_uuid)
1745 group_nodes = set(self.group_info.members)
1747 self.cfg.GetNodeGroupInstances(self.group_uuid, primary_only=True)
1750 group_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
1752 unlocked_instances = \
1753 group_instances.difference(self.owned_locks(locking.LEVEL_INSTANCE))
1756 raise errors.OpPrereqError("Missing lock for nodes: %s" %
1757 utils.CommaJoin(unlocked_nodes),
1760 if unlocked_instances:
1761 raise errors.OpPrereqError("Missing lock for instances: %s" %
1762 utils.CommaJoin(unlocked_instances),
1765 self.all_node_info = self.cfg.GetAllNodesInfo()
1766 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1768 self.my_node_names = utils.NiceSort(group_nodes)
1769 self.my_inst_names = utils.NiceSort(group_instances)
1771 self.my_node_info = dict((name, self.all_node_info[name])
1772 for name in self.my_node_names)
1774 self.my_inst_info = dict((name, self.all_inst_info[name])
1775 for name in self.my_inst_names)
1777 # We detect here the nodes that will need the extra RPC calls for verifying
1778 # split LV volumes; they should be locked.
1779 extra_lv_nodes = set()
1781 for inst in self.my_inst_info.values():
1782 if inst.disk_template in constants.DTS_INT_MIRROR:
1783 for nname in inst.all_nodes:
1784 if self.all_node_info[nname].group != self.group_uuid:
1785 extra_lv_nodes.add(nname)
1787 unlocked_lv_nodes = \
1788 extra_lv_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
1790 if unlocked_lv_nodes:
1791 raise errors.OpPrereqError("Missing node locks for LV check: %s" %
1792 utils.CommaJoin(unlocked_lv_nodes),
1794 self.extra_lv_nodes = list(extra_lv_nodes)
1796 def _VerifyNode(self, ninfo, nresult):
1797 """Perform some basic validation on data returned from a node.
1799 - check the result data structure is well formed and has all the
1801 - check ganeti version
1803 @type ninfo: L{objects.Node}
1804 @param ninfo: the node to check
1805 @param nresult: the results from the node
1807 @return: whether overall this call was successful (and we can expect
1808 reasonable values in the respose)
1812 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1814 # main result, nresult should be a non-empty dict
1815 test = not nresult or not isinstance(nresult, dict)
1816 _ErrorIf(test, self.ENODERPC, node,
1817 "unable to verify node: no data returned")
1821 # compares ganeti version
1822 local_version = constants.PROTOCOL_VERSION
1823 remote_version = nresult.get("version", None)
1824 test = not (remote_version and
1825 isinstance(remote_version, (list, tuple)) and
1826 len(remote_version) == 2)
1827 _ErrorIf(test, self.ENODERPC, node,
1828 "connection to node returned invalid data")
1832 test = local_version != remote_version[0]
1833 _ErrorIf(test, self.ENODEVERSION, node,
1834 "incompatible protocol versions: master %s,"
1835 " node %s", local_version, remote_version[0])
1839 # node seems compatible, we can actually try to look into its results
1841 # full package version
1842 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1843 self.ENODEVERSION, node,
1844 "software version mismatch: master %s, node %s",
1845 constants.RELEASE_VERSION, remote_version[1],
1846 code=self.ETYPE_WARNING)
1848 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1849 if ninfo.vm_capable and isinstance(hyp_result, dict):
1850 for hv_name, hv_result in hyp_result.iteritems():
1851 test = hv_result is not None
1852 _ErrorIf(test, self.ENODEHV, node,
1853 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1855 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1856 if ninfo.vm_capable and isinstance(hvp_result, list):
1857 for item, hv_name, hv_result in hvp_result:
1858 _ErrorIf(True, self.ENODEHV, node,
1859 "hypervisor %s parameter verify failure (source %s): %s",
1860 hv_name, item, hv_result)
1862 test = nresult.get(constants.NV_NODESETUP,
1863 ["Missing NODESETUP results"])
1864 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1869 def _VerifyNodeTime(self, ninfo, nresult,
1870 nvinfo_starttime, nvinfo_endtime):
1871 """Check the node time.
1873 @type ninfo: L{objects.Node}
1874 @param ninfo: the node to check
1875 @param nresult: the remote results for the node
1876 @param nvinfo_starttime: the start time of the RPC call
1877 @param nvinfo_endtime: the end time of the RPC call
1881 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1883 ntime = nresult.get(constants.NV_TIME, None)
1885 ntime_merged = utils.MergeTime(ntime)
1886 except (ValueError, TypeError):
1887 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1890 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1891 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1892 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1893 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1897 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1898 "Node time diverges by at least %s from master node time",
1901 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1902 """Check the node LVM results.
1904 @type ninfo: L{objects.Node}
1905 @param ninfo: the node to check
1906 @param nresult: the remote results for the node
1907 @param vg_name: the configured VG name
1914 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1916 # checks vg existence and size > 20G
1917 vglist = nresult.get(constants.NV_VGLIST, None)
1919 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1921 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1922 constants.MIN_VG_SIZE)
1923 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1926 pvlist = nresult.get(constants.NV_PVLIST, None)
1927 test = pvlist is None
1928 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1930 # check that ':' is not present in PV names, since it's a
1931 # special character for lvcreate (denotes the range of PEs to
1933 for _, pvname, owner_vg in pvlist:
1934 test = ":" in pvname
1935 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1936 " '%s' of VG '%s'", pvname, owner_vg)
1938 def _VerifyNodeBridges(self, ninfo, nresult, bridges):
1939 """Check the node bridges.
1941 @type ninfo: L{objects.Node}
1942 @param ninfo: the node to check
1943 @param nresult: the remote results for the node
1944 @param bridges: the expected list of bridges
1951 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1953 missing = nresult.get(constants.NV_BRIDGES, None)
1954 test = not isinstance(missing, list)
1955 _ErrorIf(test, self.ENODENET, node,
1956 "did not return valid bridge information")
1958 _ErrorIf(bool(missing), self.ENODENET, node, "missing bridges: %s" %
1959 utils.CommaJoin(sorted(missing)))
1961 def _VerifyNodeNetwork(self, ninfo, nresult):
1962 """Check the node network connectivity results.
1964 @type ninfo: L{objects.Node}
1965 @param ninfo: the node to check
1966 @param nresult: the remote results for the node
1970 _ErrorIf = self._ErrorIf # pylint: disable=C0103
1972 test = constants.NV_NODELIST not in nresult
1973 _ErrorIf(test, self.ENODESSH, node,
1974 "node hasn't returned node ssh connectivity data")
1976 if nresult[constants.NV_NODELIST]:
1977 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1978 _ErrorIf(True, self.ENODESSH, node,
1979 "ssh communication with node '%s': %s", a_node, a_msg)
1981 test = constants.NV_NODENETTEST not in nresult
1982 _ErrorIf(test, self.ENODENET, node,
1983 "node hasn't returned node tcp connectivity data")
1985 if nresult[constants.NV_NODENETTEST]:
1986 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1988 _ErrorIf(True, self.ENODENET, node,
1989 "tcp communication with node '%s': %s",
1990 anode, nresult[constants.NV_NODENETTEST][anode])
1992 test = constants.NV_MASTERIP not in nresult
1993 _ErrorIf(test, self.ENODENET, node,
1994 "node hasn't returned node master IP reachability data")
1996 if not nresult[constants.NV_MASTERIP]:
1997 if node == self.master_node:
1998 msg = "the master node cannot reach the master IP (not configured?)"
2000 msg = "cannot reach the master IP"
2001 _ErrorIf(True, self.ENODENET, node, msg)
2003 def _VerifyInstance(self, instance, instanceconfig, node_image,
2005 """Verify an instance.
2007 This function checks to see if the required block devices are
2008 available on the instance's node.
2011 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2012 node_current = instanceconfig.primary_node
2014 node_vol_should = {}
2015 instanceconfig.MapLVsByNode(node_vol_should)
2017 for node in node_vol_should:
2018 n_img = node_image[node]
2019 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2020 # ignore missing volumes on offline or broken nodes
2022 for volume in node_vol_should[node]:
2023 test = volume not in n_img.volumes
2024 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
2025 "volume %s missing on node %s", volume, node)
2027 if instanceconfig.admin_up:
2028 pri_img = node_image[node_current]
2029 test = instance not in pri_img.instances and not pri_img.offline
2030 _ErrorIf(test, self.EINSTANCEDOWN, instance,
2031 "instance not running on its primary node %s",
2034 diskdata = [(nname, success, status, idx)
2035 for (nname, disks) in diskstatus.items()
2036 for idx, (success, status) in enumerate(disks)]
2038 for nname, success, bdev_status, idx in diskdata:
2039 # the 'ghost node' construction in Exec() ensures that we have a
2041 snode = node_image[nname]
2042 bad_snode = snode.ghost or snode.offline
2043 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
2044 self.EINSTANCEFAULTYDISK, instance,
2045 "couldn't retrieve status for disk/%s on %s: %s",
2046 idx, nname, bdev_status)
2047 _ErrorIf((instanceconfig.admin_up and success and
2048 bdev_status.ldisk_status == constants.LDS_FAULTY),
2049 self.EINSTANCEFAULTYDISK, instance,
2050 "disk/%s on %s is faulty", idx, nname)
2052 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
2053 """Verify if there are any unknown volumes in the cluster.
2055 The .os, .swap and backup volumes are ignored. All other volumes are
2056 reported as unknown.
2058 @type reserved: L{ganeti.utils.FieldSet}
2059 @param reserved: a FieldSet of reserved volume names
2062 for node, n_img in node_image.items():
2063 if (n_img.offline or n_img.rpc_fail or n_img.lvm_fail or
2064 self.all_node_info[node].group != self.group_uuid):
2065 # skip non-healthy nodes
2067 for volume in n_img.volumes:
2068 test = ((node not in node_vol_should or
2069 volume not in node_vol_should[node]) and
2070 not reserved.Matches(volume))
2071 self._ErrorIf(test, self.ENODEORPHANLV, node,
2072 "volume %s is unknown", volume)
2074 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
2075 """Verify N+1 Memory Resilience.
2077 Check that if one single node dies we can still start all the
2078 instances it was primary for.
2081 cluster_info = self.cfg.GetClusterInfo()
2082 for node, n_img in node_image.items():
2083 # This code checks that every node which is now listed as
2084 # secondary has enough memory to host all instances it is
2085 # supposed to should a single other node in the cluster fail.
2086 # FIXME: not ready for failover to an arbitrary node
2087 # FIXME: does not support file-backed instances
2088 # WARNING: we currently take into account down instances as well
2089 # as up ones, considering that even if they're down someone
2090 # might want to start them even in the event of a node failure.
2091 if n_img.offline or self.all_node_info[node].group != self.group_uuid:
2092 # we're skipping nodes marked offline and nodes in other groups from
2093 # the N+1 warning, since most likely we don't have good memory
2094 # infromation from them; we already list instances living on such
2095 # nodes, and that's enough warning
2097 for prinode, instances in n_img.sbp.items():
2099 for instance in instances:
2100 bep = cluster_info.FillBE(instance_cfg[instance])
2101 if bep[constants.BE_AUTO_BALANCE]:
2102 needed_mem += bep[constants.BE_MEMORY]
2103 test = n_img.mfree < needed_mem
2104 self._ErrorIf(test, self.ENODEN1, node,
2105 "not enough memory to accomodate instance failovers"
2106 " should node %s fail (%dMiB needed, %dMiB available)",
2107 prinode, needed_mem, n_img.mfree)
2110 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
2111 (files_all, files_all_opt, files_mc, files_vm)):
2112 """Verifies file checksums collected from all nodes.
2114 @param errorif: Callback for reporting errors
2115 @param nodeinfo: List of L{objects.Node} objects
2116 @param master_node: Name of master node
2117 @param all_nvinfo: RPC results
2120 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
2121 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
2122 "Found file listed in more than one file list"
2124 # Define functions determining which nodes to consider for a file
2127 (files_all_opt, None),
2128 (files_mc, lambda node: (node.master_candidate or
2129 node.name == master_node)),
2130 (files_vm, lambda node: node.vm_capable),
2133 # Build mapping from filename to list of nodes which should have the file
2135 for (files, fn) in files2nodefn:
2137 filenodes = nodeinfo
2139 filenodes = filter(fn, nodeinfo)
2140 nodefiles.update((filename,
2141 frozenset(map(operator.attrgetter("name"), filenodes)))
2142 for filename in files)
2144 assert set(nodefiles) == (files_all | files_all_opt | files_mc | files_vm)
2146 fileinfo = dict((filename, {}) for filename in nodefiles)
2147 ignore_nodes = set()
2149 for node in nodeinfo:
2151 ignore_nodes.add(node.name)
2154 nresult = all_nvinfo[node.name]
2156 if nresult.fail_msg or not nresult.payload:
2159 node_files = nresult.payload.get(constants.NV_FILELIST, None)
2161 test = not (node_files and isinstance(node_files, dict))
2162 errorif(test, cls.ENODEFILECHECK, node.name,
2163 "Node did not return file checksum data")
2165 ignore_nodes.add(node.name)
2168 # Build per-checksum mapping from filename to nodes having it
2169 for (filename, checksum) in node_files.items():
2170 assert filename in nodefiles
2171 fileinfo[filename].setdefault(checksum, set()).add(node.name)
2173 for (filename, checksums) in fileinfo.items():
2174 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
2176 # Nodes having the file
2177 with_file = frozenset(node_name
2178 for nodes in fileinfo[filename].values()
2179 for node_name in nodes) - ignore_nodes
2181 expected_nodes = nodefiles[filename] - ignore_nodes
2183 # Nodes missing file
2184 missing_file = expected_nodes - with_file
2186 if filename in files_all_opt:
2188 errorif(missing_file and missing_file != expected_nodes,
2189 cls.ECLUSTERFILECHECK, None,
2190 "File %s is optional, but it must exist on all or no"
2191 " nodes (not found on %s)",
2192 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
2194 # Non-optional files
2195 errorif(missing_file, cls.ECLUSTERFILECHECK, None,
2196 "File %s is missing from node(s) %s", filename,
2197 utils.CommaJoin(utils.NiceSort(missing_file)))
2199 # Warn if a node has a file it shouldn't
2200 unexpected = with_file - expected_nodes
2202 cls.ECLUSTERFILECHECK, None,
2203 "File %s should not exist on node(s) %s",
2204 filename, utils.CommaJoin(utils.NiceSort(unexpected)))
2206 # See if there are multiple versions of the file
2207 test = len(checksums) > 1
2209 variants = ["variant %s on %s" %
2210 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
2211 for (idx, (checksum, nodes)) in
2212 enumerate(sorted(checksums.items()))]
2216 errorif(test, cls.ECLUSTERFILECHECK, None,
2217 "File %s found with %s different checksums (%s)",
2218 filename, len(checksums), "; ".join(variants))
2220 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
2222 """Verifies and the node DRBD status.
2224 @type ninfo: L{objects.Node}
2225 @param ninfo: the node to check
2226 @param nresult: the remote results for the node
2227 @param instanceinfo: the dict of instances
2228 @param drbd_helper: the configured DRBD usermode helper
2229 @param drbd_map: the DRBD map as returned by
2230 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
2234 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2237 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
2238 test = (helper_result == None)
2239 _ErrorIf(test, self.ENODEDRBDHELPER, node,
2240 "no drbd usermode helper returned")
2242 status, payload = helper_result
2244 _ErrorIf(test, self.ENODEDRBDHELPER, node,
2245 "drbd usermode helper check unsuccessful: %s", payload)
2246 test = status and (payload != drbd_helper)
2247 _ErrorIf(test, self.ENODEDRBDHELPER, node,
2248 "wrong drbd usermode helper: %s", payload)
2250 # compute the DRBD minors
2252 for minor, instance in drbd_map[node].items():
2253 test = instance not in instanceinfo
2254 _ErrorIf(test, self.ECLUSTERCFG, None,
2255 "ghost instance '%s' in temporary DRBD map", instance)
2256 # ghost instance should not be running, but otherwise we
2257 # don't give double warnings (both ghost instance and
2258 # unallocated minor in use)
2260 node_drbd[minor] = (instance, False)
2262 instance = instanceinfo[instance]
2263 node_drbd[minor] = (instance.name, instance.admin_up)
2265 # and now check them
2266 used_minors = nresult.get(constants.NV_DRBDLIST, [])
2267 test = not isinstance(used_minors, (tuple, list))
2268 _ErrorIf(test, self.ENODEDRBD, node,
2269 "cannot parse drbd status file: %s", str(used_minors))
2271 # we cannot check drbd status
2274 for minor, (iname, must_exist) in node_drbd.items():
2275 test = minor not in used_minors and must_exist
2276 _ErrorIf(test, self.ENODEDRBD, node,
2277 "drbd minor %d of instance %s is not active", minor, iname)
2278 for minor in used_minors:
2279 test = minor not in node_drbd
2280 _ErrorIf(test, self.ENODEDRBD, node,
2281 "unallocated drbd minor %d is in use", minor)
2283 def _UpdateNodeOS(self, ninfo, nresult, nimg):
2284 """Builds the node OS structures.
2286 @type ninfo: L{objects.Node}
2287 @param ninfo: the node to check
2288 @param nresult: the remote results for the node
2289 @param nimg: the node image object
2293 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2295 remote_os = nresult.get(constants.NV_OSLIST, None)
2296 test = (not isinstance(remote_os, list) or
2297 not compat.all(isinstance(v, list) and len(v) == 7
2298 for v in remote_os))
2300 _ErrorIf(test, self.ENODEOS, node,
2301 "node hasn't returned valid OS data")
2310 for (name, os_path, status, diagnose,
2311 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
2313 if name not in os_dict:
2316 # parameters is a list of lists instead of list of tuples due to
2317 # JSON lacking a real tuple type, fix it:
2318 parameters = [tuple(v) for v in parameters]
2319 os_dict[name].append((os_path, status, diagnose,
2320 set(variants), set(parameters), set(api_ver)))
2322 nimg.oslist = os_dict
2324 def _VerifyNodeOS(self, ninfo, nimg, base):
2325 """Verifies the node OS list.
2327 @type ninfo: L{objects.Node}
2328 @param ninfo: the node to check
2329 @param nimg: the node image object
2330 @param base: the 'template' node we match against (e.g. from the master)
2334 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2336 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
2338 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
2339 for os_name, os_data in nimg.oslist.items():
2340 assert os_data, "Empty OS status for OS %s?!" % os_name
2341 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
2342 _ErrorIf(not f_status, self.ENODEOS, node,
2343 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
2344 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
2345 "OS '%s' has multiple entries (first one shadows the rest): %s",
2346 os_name, utils.CommaJoin([v[0] for v in os_data]))
2347 # comparisons with the 'base' image
2348 test = os_name not in base.oslist
2349 _ErrorIf(test, self.ENODEOS, node,
2350 "Extra OS %s not present on reference node (%s)",
2354 assert base.oslist[os_name], "Base node has empty OS status?"
2355 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
2357 # base OS is invalid, skipping
2359 for kind, a, b in [("API version", f_api, b_api),
2360 ("variants list", f_var, b_var),
2361 ("parameters", beautify_params(f_param),
2362 beautify_params(b_param))]:
2363 _ErrorIf(a != b, self.ENODEOS, node,
2364 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
2365 kind, os_name, base.name,
2366 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
2368 # check any missing OSes
2369 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
2370 _ErrorIf(missing, self.ENODEOS, node,
2371 "OSes present on reference node %s but missing on this node: %s",
2372 base.name, utils.CommaJoin(missing))
2374 def _VerifyOob(self, ninfo, nresult):
2375 """Verifies out of band functionality of a node.
2377 @type ninfo: L{objects.Node}
2378 @param ninfo: the node to check
2379 @param nresult: the remote results for the node
2383 # We just have to verify the paths on master and/or master candidates
2384 # as the oob helper is invoked on the master
2385 if ((ninfo.master_candidate or ninfo.master_capable) and
2386 constants.NV_OOB_PATHS in nresult):
2387 for path_result in nresult[constants.NV_OOB_PATHS]:
2388 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
2390 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
2391 """Verifies and updates the node volume data.
2393 This function will update a L{NodeImage}'s internal structures
2394 with data from the remote call.
2396 @type ninfo: L{objects.Node}
2397 @param ninfo: the node to check
2398 @param nresult: the remote results for the node
2399 @param nimg: the node image object
2400 @param vg_name: the configured VG name
2404 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2406 nimg.lvm_fail = True
2407 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2410 elif isinstance(lvdata, basestring):
2411 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
2412 utils.SafeEncode(lvdata))
2413 elif not isinstance(lvdata, dict):
2414 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
2416 nimg.volumes = lvdata
2417 nimg.lvm_fail = False
2419 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2420 """Verifies and updates the node instance list.
2422 If the listing was successful, then updates this node's instance
2423 list. Otherwise, it marks the RPC call as failed for the instance
2426 @type ninfo: L{objects.Node}
2427 @param ninfo: the node to check
2428 @param nresult: the remote results for the node
2429 @param nimg: the node image object
2432 idata = nresult.get(constants.NV_INSTANCELIST, None)
2433 test = not isinstance(idata, list)
2434 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
2435 " (instancelist): %s", utils.SafeEncode(str(idata)))
2437 nimg.hyp_fail = True
2439 nimg.instances = idata
2441 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2442 """Verifies and computes a node information map
2444 @type ninfo: L{objects.Node}
2445 @param ninfo: the node to check
2446 @param nresult: the remote results for the node
2447 @param nimg: the node image object
2448 @param vg_name: the configured VG name
2452 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2454 # try to read free memory (from the hypervisor)
2455 hv_info = nresult.get(constants.NV_HVINFO, None)
2456 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2457 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
2460 nimg.mfree = int(hv_info["memory_free"])
2461 except (ValueError, TypeError):
2462 _ErrorIf(True, self.ENODERPC, node,
2463 "node returned invalid nodeinfo, check hypervisor")
2465 # FIXME: devise a free space model for file based instances as well
2466 if vg_name is not None:
2467 test = (constants.NV_VGLIST not in nresult or
2468 vg_name not in nresult[constants.NV_VGLIST])
2469 _ErrorIf(test, self.ENODELVM, node,
2470 "node didn't return data for the volume group '%s'"
2471 " - it is either missing or broken", vg_name)
2474 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2475 except (ValueError, TypeError):
2476 _ErrorIf(True, self.ENODERPC, node,
2477 "node returned invalid LVM info, check LVM status")
2479 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2480 """Gets per-disk status information for all instances.
2482 @type nodelist: list of strings
2483 @param nodelist: Node names
2484 @type node_image: dict of (name, L{objects.Node})
2485 @param node_image: Node objects
2486 @type instanceinfo: dict of (name, L{objects.Instance})
2487 @param instanceinfo: Instance objects
2488 @rtype: {instance: {node: [(succes, payload)]}}
2489 @return: a dictionary of per-instance dictionaries with nodes as
2490 keys and disk information as values; the disk information is a
2491 list of tuples (success, payload)
2494 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2497 node_disks_devonly = {}
2498 diskless_instances = set()
2499 diskless = constants.DT_DISKLESS
2501 for nname in nodelist:
2502 node_instances = list(itertools.chain(node_image[nname].pinst,
2503 node_image[nname].sinst))
2504 diskless_instances.update(inst for inst in node_instances
2505 if instanceinfo[inst].disk_template == diskless)
2506 disks = [(inst, disk)
2507 for inst in node_instances
2508 for disk in instanceinfo[inst].disks]
2511 # No need to collect data
2514 node_disks[nname] = disks
2516 # Creating copies as SetDiskID below will modify the objects and that can
2517 # lead to incorrect data returned from nodes
2518 devonly = [dev.Copy() for (_, dev) in disks]
2521 self.cfg.SetDiskID(dev, nname)
2523 node_disks_devonly[nname] = devonly
2525 assert len(node_disks) == len(node_disks_devonly)
2527 # Collect data from all nodes with disks
2528 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2531 assert len(result) == len(node_disks)
2535 for (nname, nres) in result.items():
2536 disks = node_disks[nname]
2539 # No data from this node
2540 data = len(disks) * [(False, "node offline")]
2543 _ErrorIf(msg, self.ENODERPC, nname,
2544 "while getting disk information: %s", msg)
2546 # No data from this node
2547 data = len(disks) * [(False, msg)]
2550 for idx, i in enumerate(nres.payload):
2551 if isinstance(i, (tuple, list)) and len(i) == 2:
2554 logging.warning("Invalid result from node %s, entry %d: %s",
2556 data.append((False, "Invalid result from the remote node"))
2558 for ((inst, _), status) in zip(disks, data):
2559 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2561 # Add empty entries for diskless instances.
2562 for inst in diskless_instances:
2563 assert inst not in instdisk
2566 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2567 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2568 compat.all(isinstance(s, (tuple, list)) and
2569 len(s) == 2 for s in statuses)
2570 for inst, nnames in instdisk.items()
2571 for nname, statuses in nnames.items())
2572 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2577 def _SshNodeSelector(group_uuid, all_nodes):
2578 """Create endless iterators for all potential SSH check hosts.
2581 nodes = [node for node in all_nodes
2582 if (node.group != group_uuid and
2584 keyfunc = operator.attrgetter("group")
2586 return map(itertools.cycle,
2587 [sorted(map(operator.attrgetter("name"), names))
2588 for _, names in itertools.groupby(sorted(nodes, key=keyfunc),
2592 def _SelectSshCheckNodes(cls, group_nodes, group_uuid, all_nodes):
2593 """Choose which nodes should talk to which other nodes.
2595 We will make nodes contact all nodes in their group, and one node from
2598 @warning: This algorithm has a known issue if one node group is much
2599 smaller than others (e.g. just one node). In such a case all other
2600 nodes will talk to the single node.
2603 online_nodes = sorted(node.name for node in group_nodes if not node.offline)
2604 sel = cls._SshNodeSelector(group_uuid, all_nodes)
2606 return (online_nodes,
2607 dict((name, sorted([i.next() for i in sel]))
2608 for name in online_nodes))
2610 def BuildHooksEnv(self):
2613 Cluster-Verify hooks just ran in the post phase and their failure makes
2614 the output be logged in the verify output and the verification to fail.
2618 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2621 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
2622 for node in self.my_node_info.values())
2626 def BuildHooksNodes(self):
2627 """Build hooks nodes.
2630 return ([], self.my_node_names)
2632 def Exec(self, feedback_fn):
2633 """Verify integrity of the node group, performing various test on nodes.
2636 # This method has too many local variables. pylint: disable=R0914
2637 feedback_fn("* Verifying group '%s'" % self.group_info.name)
2639 if not self.my_node_names:
2641 feedback_fn("* Empty node group, skipping verification")
2645 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2646 verbose = self.op.verbose
2647 self._feedback_fn = feedback_fn
2649 vg_name = self.cfg.GetVGName()
2650 drbd_helper = self.cfg.GetDRBDHelper()
2651 cluster = self.cfg.GetClusterInfo()
2652 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2653 hypervisors = cluster.enabled_hypervisors
2654 node_data_list = [self.my_node_info[name] for name in self.my_node_names]
2656 i_non_redundant = [] # Non redundant instances
2657 i_non_a_balanced = [] # Non auto-balanced instances
2658 n_offline = 0 # Count of offline nodes
2659 n_drained = 0 # Count of nodes being drained
2660 node_vol_should = {}
2662 # FIXME: verify OS list
2665 filemap = _ComputeAncillaryFiles(cluster, False)
2667 # do local checksums
2668 master_node = self.master_node = self.cfg.GetMasterNode()
2669 master_ip = self.cfg.GetMasterIP()
2671 feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
2673 node_verify_param = {
2674 constants.NV_FILELIST:
2675 utils.UniqueSequence(filename
2676 for files in filemap
2677 for filename in files),
2678 constants.NV_NODELIST:
2679 self._SelectSshCheckNodes(node_data_list, self.group_uuid,
2680 self.all_node_info.values()),
2681 constants.NV_HYPERVISOR: hypervisors,
2682 constants.NV_HVPARAMS:
2683 _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
2684 constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip)
2685 for node in node_data_list
2686 if not node.offline],
2687 constants.NV_INSTANCELIST: hypervisors,
2688 constants.NV_VERSION: None,
2689 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2690 constants.NV_NODESETUP: None,
2691 constants.NV_TIME: None,
2692 constants.NV_MASTERIP: (master_node, master_ip),
2693 constants.NV_OSLIST: None,
2694 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2697 if vg_name is not None:
2698 node_verify_param[constants.NV_VGLIST] = None
2699 node_verify_param[constants.NV_LVLIST] = vg_name
2700 node_verify_param[constants.NV_PVLIST] = [vg_name]
2701 node_verify_param[constants.NV_DRBDLIST] = None
2704 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2707 # FIXME: this needs to be changed per node-group, not cluster-wide
2709 default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
2710 if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2711 bridges.add(default_nicpp[constants.NIC_LINK])
2712 for instance in self.my_inst_info.values():
2713 for nic in instance.nics:
2714 full_nic = cluster.SimpleFillNIC(nic.nicparams)
2715 if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2716 bridges.add(full_nic[constants.NIC_LINK])
2719 node_verify_param[constants.NV_BRIDGES] = list(bridges)
2721 # Build our expected cluster state
2722 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2724 vm_capable=node.vm_capable))
2725 for node in node_data_list)
2729 for node in self.all_node_info.values():
2730 path = _SupportsOob(self.cfg, node)
2731 if path and path not in oob_paths:
2732 oob_paths.append(path)
2735 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2737 for instance in self.my_inst_names:
2738 inst_config = self.my_inst_info[instance]
2740 for nname in inst_config.all_nodes:
2741 if nname not in node_image:
2742 gnode = self.NodeImage(name=nname)
2743 gnode.ghost = (nname not in self.all_node_info)
2744 node_image[nname] = gnode
2746 inst_config.MapLVsByNode(node_vol_should)
2748 pnode = inst_config.primary_node
2749 node_image[pnode].pinst.append(instance)
2751 for snode in inst_config.secondary_nodes:
2752 nimg = node_image[snode]
2753 nimg.sinst.append(instance)
2754 if pnode not in nimg.sbp:
2755 nimg.sbp[pnode] = []
2756 nimg.sbp[pnode].append(instance)
2758 # At this point, we have the in-memory data structures complete,
2759 # except for the runtime information, which we'll gather next
2761 # Due to the way our RPC system works, exact response times cannot be
2762 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2763 # time before and after executing the request, we can at least have a time
2765 nvinfo_starttime = time.time()
2766 all_nvinfo = self.rpc.call_node_verify(self.my_node_names,
2768 self.cfg.GetClusterName())
2769 nvinfo_endtime = time.time()
2771 if self.extra_lv_nodes and vg_name is not None:
2773 self.rpc.call_node_verify(self.extra_lv_nodes,
2774 {constants.NV_LVLIST: vg_name},
2775 self.cfg.GetClusterName())
2777 extra_lv_nvinfo = {}
2779 all_drbd_map = self.cfg.ComputeDRBDMap()
2781 feedback_fn("* Gathering disk information (%s nodes)" %
2782 len(self.my_node_names))
2783 instdisk = self._CollectDiskInfo(self.my_node_names, node_image,
2786 feedback_fn("* Verifying configuration file consistency")
2788 # If not all nodes are being checked, we need to make sure the master node
2789 # and a non-checked vm_capable node are in the list.
2790 absent_nodes = set(self.all_node_info).difference(self.my_node_info)
2792 vf_nvinfo = all_nvinfo.copy()
2793 vf_node_info = list(self.my_node_info.values())
2794 additional_nodes = []
2795 if master_node not in self.my_node_info:
2796 additional_nodes.append(master_node)
2797 vf_node_info.append(self.all_node_info[master_node])
2798 # Add the first vm_capable node we find which is not included
2799 for node in absent_nodes:
2800 nodeinfo = self.all_node_info[node]
2801 if nodeinfo.vm_capable and not nodeinfo.offline:
2802 additional_nodes.append(node)
2803 vf_node_info.append(self.all_node_info[node])
2805 key = constants.NV_FILELIST
2806 vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes,
2807 {key: node_verify_param[key]},
2808 self.cfg.GetClusterName()))
2810 vf_nvinfo = all_nvinfo
2811 vf_node_info = self.my_node_info.values()
2813 self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap)
2815 feedback_fn("* Verifying node status")
2819 for node_i in node_data_list:
2821 nimg = node_image[node]
2825 feedback_fn("* Skipping offline node %s" % (node,))
2829 if node == master_node:
2831 elif node_i.master_candidate:
2832 ntype = "master candidate"
2833 elif node_i.drained:
2839 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2841 msg = all_nvinfo[node].fail_msg
2842 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2844 nimg.rpc_fail = True
2847 nresult = all_nvinfo[node].payload
2849 nimg.call_ok = self._VerifyNode(node_i, nresult)
2850 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2851 self._VerifyNodeNetwork(node_i, nresult)
2852 self._VerifyOob(node_i, nresult)
2855 self._VerifyNodeLVM(node_i, nresult, vg_name)
2856 self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper,
2859 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2860 self._UpdateNodeInstances(node_i, nresult, nimg)
2861 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2862 self._UpdateNodeOS(node_i, nresult, nimg)
2864 if not nimg.os_fail:
2865 if refos_img is None:
2867 self._VerifyNodeOS(node_i, nimg, refos_img)
2868 self._VerifyNodeBridges(node_i, nresult, bridges)
2870 # Check whether all running instancies are primary for the node. (This
2871 # can no longer be done from _VerifyInstance below, since some of the
2872 # wrong instances could be from other node groups.)
2873 non_primary_inst = set(nimg.instances).difference(nimg.pinst)
2875 for inst in non_primary_inst:
2876 test = inst in self.all_inst_info
2877 _ErrorIf(test, self.EINSTANCEWRONGNODE, inst,
2878 "instance should not run on node %s", node_i.name)
2879 _ErrorIf(not test, self.ENODEORPHANINSTANCE, node_i.name,
2880 "node is running unknown instance %s", inst)
2882 for node, result in extra_lv_nvinfo.items():
2883 self._UpdateNodeVolumes(self.all_node_info[node], result.payload,
2884 node_image[node], vg_name)
2886 feedback_fn("* Verifying instance status")
2887 for instance in self.my_inst_names:
2889 feedback_fn("* Verifying instance %s" % instance)
2890 inst_config = self.my_inst_info[instance]
2891 self._VerifyInstance(instance, inst_config, node_image,
2893 inst_nodes_offline = []
2895 pnode = inst_config.primary_node
2896 pnode_img = node_image[pnode]
2897 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2898 self.ENODERPC, pnode, "instance %s, connection to"
2899 " primary node failed", instance)
2901 _ErrorIf(inst_config.admin_up and pnode_img.offline,
2902 self.EINSTANCEBADNODE, instance,
2903 "instance is marked as running and lives on offline node %s",
2904 inst_config.primary_node)
2906 # If the instance is non-redundant we cannot survive losing its primary
2907 # node, so we are not N+1 compliant. On the other hand we have no disk
2908 # templates with more than one secondary so that situation is not well
2910 # FIXME: does not support file-backed instances
2911 if not inst_config.secondary_nodes:
2912 i_non_redundant.append(instance)
2914 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2915 instance, "instance has multiple secondary nodes: %s",
2916 utils.CommaJoin(inst_config.secondary_nodes),
2917 code=self.ETYPE_WARNING)
2919 if inst_config.disk_template in constants.DTS_INT_MIRROR:
2920 pnode = inst_config.primary_node
2921 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2922 instance_groups = {}
2924 for node in instance_nodes:
2925 instance_groups.setdefault(self.all_node_info[node].group,
2929 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2930 # Sort so that we always list the primary node first.
2931 for group, nodes in sorted(instance_groups.items(),
2932 key=lambda (_, nodes): pnode in nodes,
2935 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2936 instance, "instance has primary and secondary nodes in"
2937 " different groups: %s", utils.CommaJoin(pretty_list),
2938 code=self.ETYPE_WARNING)
2940 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2941 i_non_a_balanced.append(instance)
2943 for snode in inst_config.secondary_nodes:
2944 s_img = node_image[snode]
2945 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2946 "instance %s, connection to secondary node failed", instance)
2949 inst_nodes_offline.append(snode)
2951 # warn that the instance lives on offline nodes
2952 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2953 "instance has offline secondary node(s) %s",
2954 utils.CommaJoin(inst_nodes_offline))
2955 # ... or ghost/non-vm_capable nodes
2956 for node in inst_config.all_nodes:
2957 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2958 "instance lives on ghost node %s", node)
2959 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2960 instance, "instance lives on non-vm_capable node %s", node)
2962 feedback_fn("* Verifying orphan volumes")
2963 reserved = utils.FieldSet(*cluster.reserved_lvs)
2965 # We will get spurious "unknown volume" warnings if any node of this group
2966 # is secondary for an instance whose primary is in another group. To avoid
2967 # them, we find these instances and add their volumes to node_vol_should.
2968 for inst in self.all_inst_info.values():
2969 for secondary in inst.secondary_nodes:
2970 if (secondary in self.my_node_info
2971 and inst.name not in self.my_inst_info):
2972 inst.MapLVsByNode(node_vol_should)
2975 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2977 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2978 feedback_fn("* Verifying N+1 Memory redundancy")
2979 self._VerifyNPlusOneMemory(node_image, self.my_inst_info)
2981 feedback_fn("* Other Notes")
2983 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2984 % len(i_non_redundant))
2986 if i_non_a_balanced:
2987 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2988 % len(i_non_a_balanced))
2991 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2994 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2998 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2999 """Analyze the post-hooks' result
3001 This method analyses the hook result, handles it, and sends some
3002 nicely-formatted feedback back to the user.
3004 @param phase: one of L{constants.HOOKS_PHASE_POST} or
3005 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
3006 @param hooks_results: the results of the multi-node hooks rpc call
3007 @param feedback_fn: function used send feedback back to the caller
3008 @param lu_result: previous Exec result
3009 @return: the new Exec result, based on the previous result
3013 # We only really run POST phase hooks, only for non-empty groups,
3014 # and are only interested in their results
3015 if not self.my_node_names:
3018 elif phase == constants.HOOKS_PHASE_POST:
3019 # Used to change hooks' output to proper indentation
3020 feedback_fn("* Hooks Results")
3021 assert hooks_results, "invalid result from hooks"
3023 for node_name in hooks_results:
3024 res = hooks_results[node_name]
3026 test = msg and not res.offline
3027 self._ErrorIf(test, self.ENODEHOOKS, node_name,
3028 "Communication failure in hooks execution: %s", msg)
3029 if res.offline or msg:
3030 # No need to investigate payload if node is offline or gave
3033 for script, hkr, output in res.payload:
3034 test = hkr == constants.HKR_FAIL
3035 self._ErrorIf(test, self.ENODEHOOKS, node_name,
3036 "Script %s failed, output:", script)
3038 output = self._HOOKS_INDENT_RE.sub(" ", output)
3039 feedback_fn("%s" % output)
3045 class LUClusterVerifyDisks(NoHooksLU):
3046 """Verifies the cluster disks status.
3051 def ExpandNames(self):
3052 self.share_locks = _ShareAll()
3053 self.needed_locks = {
3054 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3057 def Exec(self, feedback_fn):
3058 group_names = self.owned_locks(locking.LEVEL_NODEGROUP)
3060 # Submit one instance of L{opcodes.OpGroupVerifyDisks} per node group
3061 return ResultWithJobs([[opcodes.OpGroupVerifyDisks(group_name=group)]
3062 for group in group_names])
3065 class LUGroupVerifyDisks(NoHooksLU):
3066 """Verifies the status of all disks in a node group.
3071 def ExpandNames(self):
3072 # Raises errors.OpPrereqError on its own if group can't be found
3073 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
3075 self.share_locks = _ShareAll()
3076 self.needed_locks = {
3077 locking.LEVEL_INSTANCE: [],
3078 locking.LEVEL_NODEGROUP: [],
3079 locking.LEVEL_NODE: [],
3082 def DeclareLocks(self, level):
3083 if level == locking.LEVEL_INSTANCE:
3084 assert not self.needed_locks[locking.LEVEL_INSTANCE]
3086 # Lock instances optimistically, needs verification once node and group
3087 # locks have been acquired
3088 self.needed_locks[locking.LEVEL_INSTANCE] = \
3089 self.cfg.GetNodeGroupInstances(self.group_uuid)
3091 elif level == locking.LEVEL_NODEGROUP:
3092 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
3094 self.needed_locks[locking.LEVEL_NODEGROUP] = \
3095 set([self.group_uuid] +
3096 # Lock all groups used by instances optimistically; this requires
3097 # going via the node before it's locked, requiring verification
3100 for instance_name in self.owned_locks(locking.LEVEL_INSTANCE)
3101 for group_uuid in self.cfg.GetInstanceNodeGroups(instance_name)])
3103 elif level == locking.LEVEL_NODE:
3104 # This will only lock the nodes in the group to be verified which contain
3106 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
3107 self._LockInstancesNodes()
3109 # Lock all nodes in group to be verified
3110 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
3111 member_nodes = self.cfg.GetNodeGroup(self.group_uuid).members
3112 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
3114 def CheckPrereq(self):
3115 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
3116 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
3117 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
3119 assert self.group_uuid in owned_groups
3121 # Check if locked instances are still correct
3122 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
3124 # Get instance information
3125 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
3127 # Check if node groups for locked instances are still correct
3128 for (instance_name, inst) in self.instances.items():
3129 assert owned_nodes.issuperset(inst.all_nodes), \
3130 "Instance %s's nodes changed while we kept the lock" % instance_name
3132 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
3135 assert self.group_uuid in inst_groups, \
3136 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
3138 def Exec(self, feedback_fn):
3139 """Verify integrity of cluster disks.
3141 @rtype: tuple of three items
3142 @return: a tuple of (dict of node-to-node_error, list of instances
3143 which need activate-disks, dict of instance: (node, volume) for
3148 res_instances = set()
3151 nv_dict = _MapInstanceDisksToNodes([inst
3152 for inst in self.instances.values()
3156 nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &
3157 set(self.cfg.GetVmCapableNodeList()))
3159 node_lvs = self.rpc.call_lv_list(nodes, [])
3161 for (node, node_res) in node_lvs.items():
3162 if node_res.offline:
3165 msg = node_res.fail_msg
3167 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
3168 res_nodes[node] = msg
3171 for lv_name, (_, _, lv_online) in node_res.payload.items():
3172 inst = nv_dict.pop((node, lv_name), None)
3173 if not (lv_online or inst is None):
3174 res_instances.add(inst)
3176 # any leftover items in nv_dict are missing LVs, let's arrange the data
3178 for key, inst in nv_dict.iteritems():
3179 res_missing.setdefault(inst, []).append(list(key))
3181 return (res_nodes, list(res_instances), res_missing)
3184 class LUClusterRepairDiskSizes(NoHooksLU):
3185 """Verifies the cluster disks sizes.
3190 def ExpandNames(self):
3191 if self.op.instances:
3192 self.wanted_names = _GetWantedInstances(self, self.op.instances)
3193 self.needed_locks = {
3194 locking.LEVEL_NODE: [],
3195 locking.LEVEL_INSTANCE: self.wanted_names,
3197 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3199 self.wanted_names = None
3200 self.needed_locks = {
3201 locking.LEVEL_NODE: locking.ALL_SET,
3202 locking.LEVEL_INSTANCE: locking.ALL_SET,
3204 self.share_locks = {
3205 locking.LEVEL_NODE: 1,
3206 locking.LEVEL_INSTANCE: 0,
3209 def DeclareLocks(self, level):
3210 if level == locking.LEVEL_NODE and self.wanted_names is not None:
3211 self._LockInstancesNodes(primary_only=True)
3213 def CheckPrereq(self):
3214 """Check prerequisites.
3216 This only checks the optional instance list against the existing names.
3219 if self.wanted_names is None:
3220 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
3222 self.wanted_instances = \
3223 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
3225 def _EnsureChildSizes(self, disk):
3226 """Ensure children of the disk have the needed disk size.
3228 This is valid mainly for DRBD8 and fixes an issue where the
3229 children have smaller disk size.
3231 @param disk: an L{ganeti.objects.Disk} object
3234 if disk.dev_type == constants.LD_DRBD8:
3235 assert disk.children, "Empty children for DRBD8?"
3236 fchild = disk.children[0]
3237 mismatch = fchild.size < disk.size
3239 self.LogInfo("Child disk has size %d, parent %d, fixing",
3240 fchild.size, disk.size)
3241 fchild.size = disk.size
3243 # and we recurse on this child only, not on the metadev
3244 return self._EnsureChildSizes(fchild) or mismatch
3248 def Exec(self, feedback_fn):
3249 """Verify the size of cluster disks.
3252 # TODO: check child disks too
3253 # TODO: check differences in size between primary/secondary nodes
3255 for instance in self.wanted_instances:
3256 pnode = instance.primary_node
3257 if pnode not in per_node_disks:
3258 per_node_disks[pnode] = []
3259 for idx, disk in enumerate(instance.disks):
3260 per_node_disks[pnode].append((instance, idx, disk))
3263 for node, dskl in per_node_disks.items():
3264 newl = [v[2].Copy() for v in dskl]
3266 self.cfg.SetDiskID(dsk, node)
3267 result = self.rpc.call_blockdev_getsize(node, newl)
3269 self.LogWarning("Failure in blockdev_getsize call to node"
3270 " %s, ignoring", node)
3272 if len(result.payload) != len(dskl):
3273 logging.warning("Invalid result from node %s: len(dksl)=%d,"
3274 " result.payload=%s", node, len(dskl), result.payload)
3275 self.LogWarning("Invalid result from node %s, ignoring node results",
3278 for ((instance, idx, disk), size) in zip(dskl, result.payload):
3280 self.LogWarning("Disk %d of instance %s did not return size"
3281 " information, ignoring", idx, instance.name)
3283 if not isinstance(size, (int, long)):
3284 self.LogWarning("Disk %d of instance %s did not return valid"
3285 " size information, ignoring", idx, instance.name)
3288 if size != disk.size:
3289 self.LogInfo("Disk %d of instance %s has mismatched size,"
3290 " correcting: recorded %d, actual %d", idx,
3291 instance.name, disk.size, size)
3293 self.cfg.Update(instance, feedback_fn)
3294 changed.append((instance.name, idx, size))
3295 if self._EnsureChildSizes(disk):
3296 self.cfg.Update(instance, feedback_fn)
3297 changed.append((instance.name, idx, disk.size))
3301 class LUClusterRename(LogicalUnit):
3302 """Rename the cluster.
3305 HPATH = "cluster-rename"
3306 HTYPE = constants.HTYPE_CLUSTER
3308 def BuildHooksEnv(self):
3313 "OP_TARGET": self.cfg.GetClusterName(),
3314 "NEW_NAME": self.op.name,
3317 def BuildHooksNodes(self):
3318 """Build hooks nodes.
3321 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
3323 def CheckPrereq(self):
3324 """Verify that the passed name is a valid one.
3327 hostname = netutils.GetHostname(name=self.op.name,
3328 family=self.cfg.GetPrimaryIPFamily())
3330 new_name = hostname.name
3331 self.ip = new_ip = hostname.ip
3332 old_name = self.cfg.GetClusterName()
3333 old_ip = self.cfg.GetMasterIP()
3334 if new_name == old_name and new_ip == old_ip:
3335 raise errors.OpPrereqError("Neither the name nor the IP address of the"
3336 " cluster has changed",
3338 if new_ip != old_ip:
3339 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
3340 raise errors.OpPrereqError("The given cluster IP address (%s) is"
3341 " reachable on the network" %
3342 new_ip, errors.ECODE_NOTUNIQUE)
3344 self.op.name = new_name
3346 def Exec(self, feedback_fn):
3347 """Rename the cluster.
3350 clustername = self.op.name
3353 # shutdown the master IP
3354 master = self.cfg.GetMasterNode()
3355 result = self.rpc.call_node_stop_master(master, False)
3356 result.Raise("Could not disable the master role")
3359 cluster = self.cfg.GetClusterInfo()
3360 cluster.cluster_name = clustername
3361 cluster.master_ip = ip
3362 self.cfg.Update(cluster, feedback_fn)
3364 # update the known hosts file
3365 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
3366 node_list = self.cfg.GetOnlineNodeList()
3368 node_list.remove(master)
3371 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
3373 result = self.rpc.call_node_start_master(master, False, False)
3374 msg = result.fail_msg
3376 self.LogWarning("Could not re-enable the master role on"
3377 " the master, please restart manually: %s", msg)
3382 class LUClusterSetParams(LogicalUnit):
3383 """Change the parameters of the cluster.
3386 HPATH = "cluster-modify"
3387 HTYPE = constants.HTYPE_CLUSTER
3390 def CheckArguments(self):
3394 if self.op.uid_pool:
3395 uidpool.CheckUidPool(self.op.uid_pool)
3397 if self.op.add_uids:
3398 uidpool.CheckUidPool(self.op.add_uids)
3400 if self.op.remove_uids:
3401 uidpool.CheckUidPool(self.op.remove_uids)
3403 def ExpandNames(self):
3404 # FIXME: in the future maybe other cluster params won't require checking on
3405 # all nodes to be modified.
3406 self.needed_locks = {
3407 locking.LEVEL_NODE: locking.ALL_SET,
3409 self.share_locks[locking.LEVEL_NODE] = 1
3411 def BuildHooksEnv(self):
3416 "OP_TARGET": self.cfg.GetClusterName(),
3417 "NEW_VG_NAME": self.op.vg_name,
3420 def BuildHooksNodes(self):
3421 """Build hooks nodes.
3424 mn = self.cfg.GetMasterNode()
3427 def CheckPrereq(self):
3428 """Check prerequisites.
3430 This checks whether the given params don't conflict and
3431 if the given volume group is valid.
3434 if self.op.vg_name is not None and not self.op.vg_name:
3435 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
3436 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
3437 " instances exist", errors.ECODE_INVAL)
3439 if self.op.drbd_helper is not None and not self.op.drbd_helper:
3440 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
3441 raise errors.OpPrereqError("Cannot disable drbd helper while"
3442 " drbd-based instances exist",
3445 node_list = self.owned_locks(locking.LEVEL_NODE)
3447 # if vg_name not None, checks given volume group on all nodes
3449 vglist = self.rpc.call_vg_list(node_list)
3450 for node in node_list:
3451 msg = vglist[node].fail_msg
3453 # ignoring down node
3454 self.LogWarning("Error while gathering data on node %s"
3455 " (ignoring node): %s", node, msg)
3457 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
3459 constants.MIN_VG_SIZE)
3461 raise errors.OpPrereqError("Error on node '%s': %s" %
3462 (node, vgstatus), errors.ECODE_ENVIRON)
3464 if self.op.drbd_helper:
3465 # checks given drbd helper on all nodes
3466 helpers = self.rpc.call_drbd_helper(node_list)
3467 for (node, ninfo) in self.cfg.GetMultiNodeInfo(node_list):
3469 self.LogInfo("Not checking drbd helper on offline node %s", node)
3471 msg = helpers[node].fail_msg
3473 raise errors.OpPrereqError("Error checking drbd helper on node"
3474 " '%s': %s" % (node, msg),
3475 errors.ECODE_ENVIRON)
3476 node_helper = helpers[node].payload
3477 if node_helper != self.op.drbd_helper:
3478 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
3479 (node, node_helper), errors.ECODE_ENVIRON)
3481 self.cluster = cluster = self.cfg.GetClusterInfo()
3482 # validate params changes
3483 if self.op.beparams:
3484 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
3485 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
3487 if self.op.ndparams:
3488 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
3489 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
3491 # TODO: we need a more general way to handle resetting
3492 # cluster-level parameters to default values
3493 if self.new_ndparams["oob_program"] == "":
3494 self.new_ndparams["oob_program"] = \
3495 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
3497 if self.op.nicparams:
3498 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
3499 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
3500 objects.NIC.CheckParameterSyntax(self.new_nicparams)
3503 # check all instances for consistency
3504 for instance in self.cfg.GetAllInstancesInfo().values():
3505 for nic_idx, nic in enumerate(instance.nics):
3506 params_copy = copy.deepcopy(nic.nicparams)
3507 params_filled = objects.FillDict(self.new_nicparams, params_copy)
3509 # check parameter syntax
3511 objects.NIC.CheckParameterSyntax(params_filled)
3512 except errors.ConfigurationError, err:
3513 nic_errors.append("Instance %s, nic/%d: %s" %
3514 (instance.name, nic_idx, err))
3516 # if we're moving instances to routed, check that they have an ip
3517 target_mode = params_filled[constants.NIC_MODE]
3518 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
3519 nic_errors.append("Instance %s, nic/%d: routed NIC with no ip"
3520 " address" % (instance.name, nic_idx))
3522 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
3523 "\n".join(nic_errors))
3525 # hypervisor list/parameters
3526 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
3527 if self.op.hvparams:
3528 for hv_name, hv_dict in self.op.hvparams.items():
3529 if hv_name not in self.new_hvparams:
3530 self.new_hvparams[hv_name] = hv_dict
3532 self.new_hvparams[hv_name].update(hv_dict)
3534 # os hypervisor parameters
3535 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
3537 for os_name, hvs in self.op.os_hvp.items():
3538 if os_name not in self.new_os_hvp:
3539 self.new_os_hvp[os_name] = hvs
3541 for hv_name, hv_dict in hvs.items():
3542 if hv_name not in self.new_os_hvp[os_name]:
3543 self.new_os_hvp[os_name][hv_name] = hv_dict
3545 self.new_os_hvp[os_name][hv_name].update(hv_dict)
3548 self.new_osp = objects.FillDict(cluster.osparams, {})
3549 if self.op.osparams:
3550 for os_name, osp in self.op.osparams.items():
3551 if os_name not in self.new_osp:
3552 self.new_osp[os_name] = {}
3554 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
3557 if not self.new_osp[os_name]:
3558 # we removed all parameters
3559 del self.new_osp[os_name]
3561 # check the parameter validity (remote check)
3562 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
3563 os_name, self.new_osp[os_name])
3565 # changes to the hypervisor list
3566 if self.op.enabled_hypervisors is not None:
3567 self.hv_list = self.op.enabled_hypervisors
3568 for hv in self.hv_list:
3569 # if the hypervisor doesn't already exist in the cluster
3570 # hvparams, we initialize it to empty, and then (in both
3571 # cases) we make sure to fill the defaults, as we might not
3572 # have a complete defaults list if the hypervisor wasn't
3574 if hv not in new_hvp:
3576 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
3577 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
3579 self.hv_list = cluster.enabled_hypervisors
3581 if self.op.hvparams or self.op.enabled_hypervisors is not None:
3582 # either the enabled list has changed, or the parameters have, validate
3583 for hv_name, hv_params in self.new_hvparams.items():
3584 if ((self.op.hvparams and hv_name in self.op.hvparams) or
3585 (self.op.enabled_hypervisors and
3586 hv_name in self.op.enabled_hypervisors)):
3587 # either this is a new hypervisor, or its parameters have changed
3588 hv_class = hypervisor.GetHypervisor(hv_name)
3589 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3590 hv_class.CheckParameterSyntax(hv_params)
3591 _CheckHVParams(self, node_list, hv_name, hv_params)
3594 # no need to check any newly-enabled hypervisors, since the
3595 # defaults have already been checked in the above code-block
3596 for os_name, os_hvp in self.new_os_hvp.items():
3597 for hv_name, hv_params in os_hvp.items():
3598 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3599 # we need to fill in the new os_hvp on top of the actual hv_p
3600 cluster_defaults = self.new_hvparams.get(hv_name, {})
3601 new_osp = objects.FillDict(cluster_defaults, hv_params)
3602 hv_class = hypervisor.GetHypervisor(hv_name)
3603 hv_class.CheckParameterSyntax(new_osp)
3604 _CheckHVParams(self, node_list, hv_name, new_osp)
3606 if self.op.default_iallocator:
3607 alloc_script = utils.FindFile(self.op.default_iallocator,
3608 constants.IALLOCATOR_SEARCH_PATH,
3610 if alloc_script is None:
3611 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
3612 " specified" % self.op.default_iallocator,
3615 def Exec(self, feedback_fn):
3616 """Change the parameters of the cluster.
3619 if self.op.vg_name is not None:
3620 new_volume = self.op.vg_name
3623 if new_volume != self.cfg.GetVGName():
3624 self.cfg.SetVGName(new_volume)
3626 feedback_fn("Cluster LVM configuration already in desired"
3627 " state, not changing")
3628 if self.op.drbd_helper is not None:
3629 new_helper = self.op.drbd_helper
3632 if new_helper != self.cfg.GetDRBDHelper():
3633 self.cfg.SetDRBDHelper(new_helper)
3635 feedback_fn("Cluster DRBD helper already in desired state,"
3637 if self.op.hvparams:
3638 self.cluster.hvparams = self.new_hvparams
3640 self.cluster.os_hvp = self.new_os_hvp
3641 if self.op.enabled_hypervisors is not None:
3642 self.cluster.hvparams = self.new_hvparams
3643 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
3644 if self.op.beparams:
3645 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
3646 if self.op.nicparams:
3647 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
3648 if self.op.osparams:
3649 self.cluster.osparams = self.new_osp
3650 if self.op.ndparams:
3651 self.cluster.ndparams = self.new_ndparams
3653 if self.op.candidate_pool_size is not None:
3654 self.cluster.candidate_pool_size = self.op.candidate_pool_size
3655 # we need to update the pool size here, otherwise the save will fail
3656 _AdjustCandidatePool(self, [])
3658 if self.op.maintain_node_health is not None:
3659 self.cluster.maintain_node_health = self.op.maintain_node_health
3661 if self.op.prealloc_wipe_disks is not None:
3662 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
3664 if self.op.add_uids is not None:
3665 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
3667 if self.op.remove_uids is not None:
3668 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
3670 if self.op.uid_pool is not None:
3671 self.cluster.uid_pool = self.op.uid_pool
3673 if self.op.default_iallocator is not None:
3674 self.cluster.default_iallocator = self.op.default_iallocator
3676 if self.op.reserved_lvs is not None:
3677 self.cluster.reserved_lvs = self.op.reserved_lvs
3679 def helper_os(aname, mods, desc):
3681 lst = getattr(self.cluster, aname)
3682 for key, val in mods:
3683 if key == constants.DDM_ADD:
3685 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3688 elif key == constants.DDM_REMOVE:
3692 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3694 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3696 if self.op.hidden_os:
3697 helper_os("hidden_os", self.op.hidden_os, "hidden")
3699 if self.op.blacklisted_os:
3700 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3702 if self.op.master_netdev:
3703 master = self.cfg.GetMasterNode()
3704 feedback_fn("Shutting down master ip on the current netdev (%s)" %
3705 self.cluster.master_netdev)
3706 result = self.rpc.call_node_stop_master(master, False)
3707 result.Raise("Could not disable the master ip")
3708 feedback_fn("Changing master_netdev from %s to %s" %
3709 (self.cluster.master_netdev, self.op.master_netdev))
3710 self.cluster.master_netdev = self.op.master_netdev
3712 self.cfg.Update(self.cluster, feedback_fn)
3714 if self.op.master_netdev:
3715 feedback_fn("Starting the master ip on the new master netdev (%s)" %
3716 self.op.master_netdev)
3717 result = self.rpc.call_node_start_master(master, False, False)
3719 self.LogWarning("Could not re-enable the master ip on"
3720 " the master, please restart manually: %s",
3724 def _UploadHelper(lu, nodes, fname):
3725 """Helper for uploading a file and showing warnings.
3728 if os.path.exists(fname):
3729 result = lu.rpc.call_upload_file(nodes, fname)
3730 for to_node, to_result in result.items():
3731 msg = to_result.fail_msg
3733 msg = ("Copy of file %s to node %s failed: %s" %
3734 (fname, to_node, msg))
3735 lu.proc.LogWarning(msg)
3738 def _ComputeAncillaryFiles(cluster, redist):
3739 """Compute files external to Ganeti which need to be consistent.
3741 @type redist: boolean
3742 @param redist: Whether to include files which need to be redistributed
3745 # Compute files for all nodes
3747 constants.SSH_KNOWN_HOSTS_FILE,
3748 constants.CONFD_HMAC_KEY,
3749 constants.CLUSTER_DOMAIN_SECRET_FILE,
3753 files_all.update(constants.ALL_CERT_FILES)
3754 files_all.update(ssconf.SimpleStore().GetFileList())
3756 # we need to ship at least the RAPI certificate
3757 files_all.add(constants.RAPI_CERT_FILE)
3759 if cluster.modify_etc_hosts:
3760 files_all.add(constants.ETC_HOSTS)
3762 # Files which must either exist on all nodes or on none
3763 files_all_opt = set([
3764 constants.RAPI_USERS_FILE,
3767 # Files which should only be on master candidates
3770 files_mc.add(constants.CLUSTER_CONF_FILE)
3772 # Files which should only be on VM-capable nodes
3773 files_vm = set(filename
3774 for hv_name in cluster.enabled_hypervisors
3775 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles())
3777 # Filenames must be unique
3778 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
3779 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
3780 "Found file listed in more than one file list"
3782 return (files_all, files_all_opt, files_mc, files_vm)
3785 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3786 """Distribute additional files which are part of the cluster configuration.
3788 ConfigWriter takes care of distributing the config and ssconf files, but
3789 there are more files which should be distributed to all nodes. This function
3790 makes sure those are copied.
3792 @param lu: calling logical unit
3793 @param additional_nodes: list of nodes not in the config to distribute to
3794 @type additional_vm: boolean
3795 @param additional_vm: whether the additional nodes are vm-capable or not
3798 # Gather target nodes
3799 cluster = lu.cfg.GetClusterInfo()
3800 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3802 online_nodes = lu.cfg.GetOnlineNodeList()
3803 vm_nodes = lu.cfg.GetVmCapableNodeList()
3805 if additional_nodes is not None:
3806 online_nodes.extend(additional_nodes)
3808 vm_nodes.extend(additional_nodes)
3810 # Never distribute to master node
3811 for nodelist in [online_nodes, vm_nodes]:
3812 if master_info.name in nodelist:
3813 nodelist.remove(master_info.name)
3816 (files_all, files_all_opt, files_mc, files_vm) = \
3817 _ComputeAncillaryFiles(cluster, True)
3819 # Never re-distribute configuration file from here
3820 assert not (constants.CLUSTER_CONF_FILE in files_all or
3821 constants.CLUSTER_CONF_FILE in files_vm)
3822 assert not files_mc, "Master candidates not handled in this function"
3825 (online_nodes, files_all),
3826 (online_nodes, files_all_opt),
3827 (vm_nodes, files_vm),
3831 for (node_list, files) in filemap:
3833 _UploadHelper(lu, node_list, fname)
3836 class LUClusterRedistConf(NoHooksLU):
3837 """Force the redistribution of cluster configuration.
3839 This is a very simple LU.
3844 def ExpandNames(self):
3845 self.needed_locks = {
3846 locking.LEVEL_NODE: locking.ALL_SET,
3848 self.share_locks[locking.LEVEL_NODE] = 1
3850 def Exec(self, feedback_fn):
3851 """Redistribute the configuration.
3854 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3855 _RedistributeAncillaryFiles(self)
3858 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3859 """Sleep and poll for an instance's disk to sync.
3862 if not instance.disks or disks is not None and not disks:
3865 disks = _ExpandCheckDisks(instance, disks)
3868 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3870 node = instance.primary_node
3873 lu.cfg.SetDiskID(dev, node)
3875 # TODO: Convert to utils.Retry
3878 degr_retries = 10 # in seconds, as we sleep 1 second each time
3882 cumul_degraded = False
3883 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3884 msg = rstats.fail_msg
3886 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3889 raise errors.RemoteError("Can't contact node %s for mirror data,"
3890 " aborting." % node)
3893 rstats = rstats.payload
3895 for i, mstat in enumerate(rstats):
3897 lu.LogWarning("Can't compute data for node %s/%s",
3898 node, disks[i].iv_name)
3901 cumul_degraded = (cumul_degraded or
3902 (mstat.is_degraded and mstat.sync_percent is None))
3903 if mstat.sync_percent is not None:
3905 if mstat.estimated_time is not None:
3906 rem_time = ("%s remaining (estimated)" %
3907 utils.FormatSeconds(mstat.estimated_time))
3908 max_time = mstat.estimated_time
3910 rem_time = "no time estimate"
3911 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3912 (disks[i].iv_name, mstat.sync_percent, rem_time))
3914 # if we're done but degraded, let's do a few small retries, to
3915 # make sure we see a stable and not transient situation; therefore
3916 # we force restart of the loop
3917 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3918 logging.info("Degraded disks found, %d retries left", degr_retries)
3926 time.sleep(min(60, max_time))
3929 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3930 return not cumul_degraded
3933 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3934 """Check that mirrors are not degraded.
3936 The ldisk parameter, if True, will change the test from the
3937 is_degraded attribute (which represents overall non-ok status for
3938 the device(s)) to the ldisk (representing the local storage status).
3941 lu.cfg.SetDiskID(dev, node)
3945 if on_primary or dev.AssembleOnSecondary():
3946 rstats = lu.rpc.call_blockdev_find(node, dev)
3947 msg = rstats.fail_msg
3949 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3951 elif not rstats.payload:
3952 lu.LogWarning("Can't find disk on node %s", node)
3956 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3958 result = result and not rstats.payload.is_degraded
3961 for child in dev.children:
3962 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3967 class LUOobCommand(NoHooksLU):
3968 """Logical unit for OOB handling.
3972 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
3974 def ExpandNames(self):
3975 """Gather locks we need.
3978 if self.op.node_names:
3979 self.op.node_names = _GetWantedNodes(self, self.op.node_names)
3980 lock_names = self.op.node_names
3982 lock_names = locking.ALL_SET
3984 self.needed_locks = {
3985 locking.LEVEL_NODE: lock_names,
3988 def CheckPrereq(self):
3989 """Check prerequisites.
3992 - the node exists in the configuration
3995 Any errors are signaled by raising errors.OpPrereqError.
3999 self.master_node = self.cfg.GetMasterNode()
4001 assert self.op.power_delay >= 0.0
4003 if self.op.node_names:
4004 if (self.op.command in self._SKIP_MASTER and
4005 self.master_node in self.op.node_names):
4006 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
4007 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
4009 if master_oob_handler:
4010 additional_text = ("run '%s %s %s' if you want to operate on the"
4011 " master regardless") % (master_oob_handler,
4015 additional_text = "it does not support out-of-band operations"
4017 raise errors.OpPrereqError(("Operating on the master node %s is not"
4018 " allowed for %s; %s") %
4019 (self.master_node, self.op.command,
4020 additional_text), errors.ECODE_INVAL)
4022 self.op.node_names = self.cfg.GetNodeList()
4023 if self.op.command in self._SKIP_MASTER:
4024 self.op.node_names.remove(self.master_node)
4026 if self.op.command in self._SKIP_MASTER:
4027 assert self.master_node not in self.op.node_names
4029 for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names):
4031 raise errors.OpPrereqError("Node %s not found" % node_name,
4034 self.nodes.append(node)
4036 if (not self.op.ignore_status and
4037 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
4038 raise errors.OpPrereqError(("Cannot power off node %s because it is"
4039 " not marked offline") % node_name,
4042 def Exec(self, feedback_fn):
4043 """Execute OOB and return result if we expect any.
4046 master_node = self.master_node
4049 for idx, node in enumerate(utils.NiceSort(self.nodes,
4050 key=lambda node: node.name)):
4051 node_entry = [(constants.RS_NORMAL, node.name)]
4052 ret.append(node_entry)
4054 oob_program = _SupportsOob(self.cfg, node)
4057 node_entry.append((constants.RS_UNAVAIL, None))
4060 logging.info("Executing out-of-band command '%s' using '%s' on %s",
4061 self.op.command, oob_program, node.name)
4062 result = self.rpc.call_run_oob(master_node, oob_program,
4063 self.op.command, node.name,
4067 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
4068 node.name, result.fail_msg)
4069 node_entry.append((constants.RS_NODATA, None))
4072 self._CheckPayload(result)
4073 except errors.OpExecError, err:
4074 self.LogWarning("Payload returned by node '%s' is not valid: %s",
4076 node_entry.append((constants.RS_NODATA, None))
4078 if self.op.command == constants.OOB_HEALTH:
4079 # For health we should log important events
4080 for item, status in result.payload:
4081 if status in [constants.OOB_STATUS_WARNING,
4082 constants.OOB_STATUS_CRITICAL]:
4083 self.LogWarning("Item '%s' on node '%s' has status '%s'",
4084 item, node.name, status)
4086 if self.op.command == constants.OOB_POWER_ON:
4088 elif self.op.command == constants.OOB_POWER_OFF:
4089 node.powered = False
4090 elif self.op.command == constants.OOB_POWER_STATUS:
4091 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
4092 if powered != node.powered:
4093 logging.warning(("Recorded power state (%s) of node '%s' does not"
4094 " match actual power state (%s)"), node.powered,
4097 # For configuration changing commands we should update the node
4098 if self.op.command in (constants.OOB_POWER_ON,
4099 constants.OOB_POWER_OFF):
4100 self.cfg.Update(node, feedback_fn)
4102 node_entry.append((constants.RS_NORMAL, result.payload))
4104 if (self.op.command == constants.OOB_POWER_ON and
4105 idx < len(self.nodes) - 1):
4106 time.sleep(self.op.power_delay)
4110 def _CheckPayload(self, result):
4111 """Checks if the payload is valid.
4113 @param result: RPC result
4114 @raises errors.OpExecError: If payload is not valid
4118 if self.op.command == constants.OOB_HEALTH:
4119 if not isinstance(result.payload, list):
4120 errs.append("command 'health' is expected to return a list but got %s" %
4121 type(result.payload))
4123 for item, status in result.payload:
4124 if status not in constants.OOB_STATUSES:
4125 errs.append("health item '%s' has invalid status '%s'" %
4128 if self.op.command == constants.OOB_POWER_STATUS:
4129 if not isinstance(result.payload, dict):
4130 errs.append("power-status is expected to return a dict but got %s" %
4131 type(result.payload))
4133 if self.op.command in [
4134 constants.OOB_POWER_ON,
4135 constants.OOB_POWER_OFF,
4136 constants.OOB_POWER_CYCLE,
4138 if result.payload is not None:
4139 errs.append("%s is expected to not return payload but got '%s'" %
4140 (self.op.command, result.payload))
4143 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
4144 utils.CommaJoin(errs))
4147 class _OsQuery(_QueryBase):
4148 FIELDS = query.OS_FIELDS
4150 def ExpandNames(self, lu):
4151 # Lock all nodes in shared mode
4152 # Temporary removal of locks, should be reverted later
4153 # TODO: reintroduce locks when they are lighter-weight
4154 lu.needed_locks = {}
4155 #self.share_locks[locking.LEVEL_NODE] = 1
4156 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4158 # The following variables interact with _QueryBase._GetNames
4160 self.wanted = self.names
4162 self.wanted = locking.ALL_SET
4164 self.do_locking = self.use_locking
4166 def DeclareLocks(self, lu, level):
4170 def _DiagnoseByOS(rlist):
4171 """Remaps a per-node return list into an a per-os per-node dictionary
4173 @param rlist: a map with node names as keys and OS objects as values
4176 @return: a dictionary with osnames as keys and as value another
4177 map, with nodes as keys and tuples of (path, status, diagnose,
4178 variants, parameters, api_versions) as values, eg::
4180 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
4181 (/srv/..., False, "invalid api")],
4182 "node2": [(/srv/..., True, "", [], [])]}
4187 # we build here the list of nodes that didn't fail the RPC (at RPC
4188 # level), so that nodes with a non-responding node daemon don't
4189 # make all OSes invalid
4190 good_nodes = [node_name for node_name in rlist
4191 if not rlist[node_name].fail_msg]
4192 for node_name, nr in rlist.items():
4193 if nr.fail_msg or not nr.payload:
4195 for (name, path, status, diagnose, variants,
4196 params, api_versions) in nr.payload:
4197 if name not in all_os:
4198 # build a list of nodes for this os containing empty lists
4199 # for each node in node_list
4201 for nname in good_nodes:
4202 all_os[name][nname] = []
4203 # convert params from [name, help] to (name, help)
4204 params = [tuple(v) for v in params]
4205 all_os[name][node_name].append((path, status, diagnose,
4206 variants, params, api_versions))
4209 def _GetQueryData(self, lu):
4210 """Computes the list of nodes and their attributes.
4213 # Locking is not used
4214 assert not (compat.any(lu.glm.is_owned(level)
4215 for level in locking.LEVELS
4216 if level != locking.LEVEL_CLUSTER) or
4217 self.do_locking or self.use_locking)
4219 valid_nodes = [node.name
4220 for node in lu.cfg.GetAllNodesInfo().values()
4221 if not node.offline and node.vm_capable]
4222 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
4223 cluster = lu.cfg.GetClusterInfo()
4227 for (os_name, os_data) in pol.items():
4228 info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
4229 hidden=(os_name in cluster.hidden_os),
4230 blacklisted=(os_name in cluster.blacklisted_os))
4234 api_versions = set()
4236 for idx, osl in enumerate(os_data.values()):
4237 info.valid = bool(info.valid and osl and osl[0][1])
4241 (node_variants, node_params, node_api) = osl[0][3:6]
4244 variants.update(node_variants)
4245 parameters.update(node_params)
4246 api_versions.update(node_api)
4248 # Filter out inconsistent values
4249 variants.intersection_update(node_variants)
4250 parameters.intersection_update(node_params)
4251 api_versions.intersection_update(node_api)
4253 info.variants = list(variants)
4254 info.parameters = list(parameters)
4255 info.api_versions = list(api_versions)
4257 data[os_name] = info
4259 # Prepare data in requested order
4260 return [data[name] for name in self._GetNames(lu, pol.keys(), None)
4264 class LUOsDiagnose(NoHooksLU):
4265 """Logical unit for OS diagnose/query.
4271 def _BuildFilter(fields, names):
4272 """Builds a filter for querying OSes.
4275 name_filter = qlang.MakeSimpleFilter("name", names)
4277 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
4278 # respective field is not requested
4279 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]]
4280 for fname in ["hidden", "blacklisted"]
4281 if fname not in fields]
4282 if "valid" not in fields:
4283 status_filter.append([qlang.OP_TRUE, "valid"])
4286 status_filter.insert(0, qlang.OP_AND)
4288 status_filter = None
4290 if name_filter and status_filter:
4291 return [qlang.OP_AND, name_filter, status_filter]
4295 return status_filter
4297 def CheckArguments(self):
4298 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names),
4299 self.op.output_fields, False)
4301 def ExpandNames(self):
4302 self.oq.ExpandNames(self)
4304 def Exec(self, feedback_fn):
4305 return self.oq.OldStyleQuery(self)
4308 class LUNodeRemove(LogicalUnit):
4309 """Logical unit for removing a node.
4312 HPATH = "node-remove"
4313 HTYPE = constants.HTYPE_NODE
4315 def BuildHooksEnv(self):
4318 This doesn't run on the target node in the pre phase as a failed
4319 node would then be impossible to remove.
4323 "OP_TARGET": self.op.node_name,
4324 "NODE_NAME": self.op.node_name,
4327 def BuildHooksNodes(self):
4328 """Build hooks nodes.
4331 all_nodes = self.cfg.GetNodeList()
4333 all_nodes.remove(self.op.node_name)
4335 logging.warning("Node '%s', which is about to be removed, was not found"
4336 " in the list of all nodes", self.op.node_name)
4337 return (all_nodes, all_nodes)
4339 def CheckPrereq(self):
4340 """Check prerequisites.
4343 - the node exists in the configuration
4344 - it does not have primary or secondary instances
4345 - it's not the master
4347 Any errors are signaled by raising errors.OpPrereqError.
4350 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4351 node = self.cfg.GetNodeInfo(self.op.node_name)
4352 assert node is not None
4354 masternode = self.cfg.GetMasterNode()
4355 if node.name == masternode:
4356 raise errors.OpPrereqError("Node is the master node, failover to another"
4357 " node is required", errors.ECODE_INVAL)
4359 for instance_name, instance in self.cfg.GetAllInstancesInfo().items():
4360 if node.name in instance.all_nodes:
4361 raise errors.OpPrereqError("Instance %s is still running on the node,"
4362 " please remove first" % instance_name,
4364 self.op.node_name = node.name
4367 def Exec(self, feedback_fn):
4368 """Removes the node from the cluster.
4372 logging.info("Stopping the node daemon and removing configs from node %s",
4375 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
4377 # Promote nodes to master candidate as needed
4378 _AdjustCandidatePool(self, exceptions=[node.name])
4379 self.context.RemoveNode(node.name)
4381 # Run post hooks on the node before it's removed
4382 _RunPostHook(self, node.name)
4384 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
4385 msg = result.fail_msg
4387 self.LogWarning("Errors encountered on the remote node while leaving"
4388 " the cluster: %s", msg)
4390 # Remove node from our /etc/hosts
4391 if self.cfg.GetClusterInfo().modify_etc_hosts:
4392 master_node = self.cfg.GetMasterNode()
4393 result = self.rpc.call_etc_hosts_modify(master_node,
4394 constants.ETC_HOSTS_REMOVE,
4396 result.Raise("Can't update hosts file with new host data")
4397 _RedistributeAncillaryFiles(self)
4400 class _NodeQuery(_QueryBase):
4401 FIELDS = query.NODE_FIELDS
4403 def ExpandNames(self, lu):
4404 lu.needed_locks = {}
4405 lu.share_locks = _ShareAll()
4408 self.wanted = _GetWantedNodes(lu, self.names)
4410 self.wanted = locking.ALL_SET
4412 self.do_locking = (self.use_locking and
4413 query.NQ_LIVE in self.requested_data)
4416 # If any non-static field is requested we need to lock the nodes
4417 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
4419 def DeclareLocks(self, lu, level):
4422 def _GetQueryData(self, lu):
4423 """Computes the list of nodes and their attributes.
4426 all_info = lu.cfg.GetAllNodesInfo()
4428 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
4430 # Gather data as requested
4431 if query.NQ_LIVE in self.requested_data:
4432 # filter out non-vm_capable nodes
4433 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
4435 node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
4436 lu.cfg.GetHypervisorType())
4437 live_data = dict((name, nresult.payload)
4438 for (name, nresult) in node_data.items()
4439 if not nresult.fail_msg and nresult.payload)
4443 if query.NQ_INST in self.requested_data:
4444 node_to_primary = dict([(name, set()) for name in nodenames])
4445 node_to_secondary = dict([(name, set()) for name in nodenames])
4447 inst_data = lu.cfg.GetAllInstancesInfo()
4449 for inst in inst_data.values():
4450 if inst.primary_node in node_to_primary:
4451 node_to_primary[inst.primary_node].add(inst.name)
4452 for secnode in inst.secondary_nodes:
4453 if secnode in node_to_secondary:
4454 node_to_secondary[secnode].add(inst.name)
4456 node_to_primary = None
4457 node_to_secondary = None
4459 if query.NQ_OOB in self.requested_data:
4460 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
4461 for name, node in all_info.iteritems())
4465 if query.NQ_GROUP in self.requested_data:
4466 groups = lu.cfg.GetAllNodeGroupsInfo()
4470 return query.NodeQueryData([all_info[name] for name in nodenames],
4471 live_data, lu.cfg.GetMasterNode(),
4472 node_to_primary, node_to_secondary, groups,
4473 oob_support, lu.cfg.GetClusterInfo())
4476 class LUNodeQuery(NoHooksLU):
4477 """Logical unit for querying nodes.
4480 # pylint: disable=W0142
4483 def CheckArguments(self):
4484 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
4485 self.op.output_fields, self.op.use_locking)
4487 def ExpandNames(self):
4488 self.nq.ExpandNames(self)
4490 def Exec(self, feedback_fn):
4491 return self.nq.OldStyleQuery(self)
4494 class LUNodeQueryvols(NoHooksLU):
4495 """Logical unit for getting volumes on node(s).
4499 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
4500 _FIELDS_STATIC = utils.FieldSet("node")
4502 def CheckArguments(self):
4503 _CheckOutputFields(static=self._FIELDS_STATIC,
4504 dynamic=self._FIELDS_DYNAMIC,
4505 selected=self.op.output_fields)
4507 def ExpandNames(self):
4508 self.needed_locks = {}
4509 self.share_locks[locking.LEVEL_NODE] = 1
4510 if not self.op.nodes:
4511 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4513 self.needed_locks[locking.LEVEL_NODE] = \
4514 _GetWantedNodes(self, self.op.nodes)
4516 def Exec(self, feedback_fn):
4517 """Computes the list of nodes and their attributes.
4520 nodenames = self.owned_locks(locking.LEVEL_NODE)
4521 volumes = self.rpc.call_node_volumes(nodenames)
4523 ilist = self.cfg.GetAllInstancesInfo()
4524 vol2inst = _MapInstanceDisksToNodes(ilist.values())
4527 for node in nodenames:
4528 nresult = volumes[node]
4531 msg = nresult.fail_msg
4533 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
4536 node_vols = sorted(nresult.payload,
4537 key=operator.itemgetter("dev"))
4539 for vol in node_vols:
4541 for field in self.op.output_fields:
4544 elif field == "phys":
4548 elif field == "name":
4550 elif field == "size":
4551 val = int(float(vol["size"]))
4552 elif field == "instance":
4553 val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-")
4555 raise errors.ParameterError(field)
4556 node_output.append(str(val))
4558 output.append(node_output)
4563 class LUNodeQueryStorage(NoHooksLU):
4564 """Logical unit for getting information on storage units on node(s).
4567 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
4570 def CheckArguments(self):
4571 _CheckOutputFields(static=self._FIELDS_STATIC,
4572 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
4573 selected=self.op.output_fields)
4575 def ExpandNames(self):
4576 self.needed_locks = {}
4577 self.share_locks[locking.LEVEL_NODE] = 1
4580 self.needed_locks[locking.LEVEL_NODE] = \
4581 _GetWantedNodes(self, self.op.nodes)
4583 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4585 def Exec(self, feedback_fn):
4586 """Computes the list of nodes and their attributes.
4589 self.nodes = self.owned_locks(locking.LEVEL_NODE)
4591 # Always get name to sort by
4592 if constants.SF_NAME in self.op.output_fields:
4593 fields = self.op.output_fields[:]
4595 fields = [constants.SF_NAME] + self.op.output_fields
4597 # Never ask for node or type as it's only known to the LU
4598 for extra in [constants.SF_NODE, constants.SF_TYPE]:
4599 while extra in fields:
4600 fields.remove(extra)
4602 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
4603 name_idx = field_idx[constants.SF_NAME]
4605 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4606 data = self.rpc.call_storage_list(self.nodes,
4607 self.op.storage_type, st_args,
4608 self.op.name, fields)
4612 for node in utils.NiceSort(self.nodes):
4613 nresult = data[node]
4617 msg = nresult.fail_msg
4619 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
4622 rows = dict([(row[name_idx], row) for row in nresult.payload])
4624 for name in utils.NiceSort(rows.keys()):
4629 for field in self.op.output_fields:
4630 if field == constants.SF_NODE:
4632 elif field == constants.SF_TYPE:
4633 val = self.op.storage_type
4634 elif field in field_idx:
4635 val = row[field_idx[field]]
4637 raise errors.ParameterError(field)
4646 class _InstanceQuery(_QueryBase):
4647 FIELDS = query.INSTANCE_FIELDS
4649 def ExpandNames(self, lu):
4650 lu.needed_locks = {}
4651 lu.share_locks = _ShareAll()
4654 self.wanted = _GetWantedInstances(lu, self.names)
4656 self.wanted = locking.ALL_SET
4658 self.do_locking = (self.use_locking and
4659 query.IQ_LIVE in self.requested_data)
4661 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4662 lu.needed_locks[locking.LEVEL_NODEGROUP] = []
4663 lu.needed_locks[locking.LEVEL_NODE] = []
4664 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4666 self.do_grouplocks = (self.do_locking and
4667 query.IQ_NODES in self.requested_data)
4669 def DeclareLocks(self, lu, level):
4671 if level == locking.LEVEL_NODEGROUP and self.do_grouplocks:
4672 assert not lu.needed_locks[locking.LEVEL_NODEGROUP]
4674 # Lock all groups used by instances optimistically; this requires going
4675 # via the node before it's locked, requiring verification later on
4676 lu.needed_locks[locking.LEVEL_NODEGROUP] = \
4678 for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
4679 for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
4680 elif level == locking.LEVEL_NODE:
4681 lu._LockInstancesNodes() # pylint: disable=W0212
4684 def _CheckGroupLocks(lu):
4685 owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE))
4686 owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP))
4688 # Check if node groups for locked instances are still correct
4689 for instance_name in owned_instances:
4690 _CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups)
4692 def _GetQueryData(self, lu):
4693 """Computes the list of instances and their attributes.
4696 if self.do_grouplocks:
4697 self._CheckGroupLocks(lu)
4699 cluster = lu.cfg.GetClusterInfo()
4700 all_info = lu.cfg.GetAllInstancesInfo()
4702 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
4704 instance_list = [all_info[name] for name in instance_names]
4705 nodes = frozenset(itertools.chain(*(inst.all_nodes
4706 for inst in instance_list)))
4707 hv_list = list(set([inst.hypervisor for inst in instance_list]))
4710 wrongnode_inst = set()
4712 # Gather data as requested
4713 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
4715 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
4717 result = node_data[name]
4719 # offline nodes will be in both lists
4720 assert result.fail_msg
4721 offline_nodes.append(name)
4723 bad_nodes.append(name)
4724 elif result.payload:
4725 for inst in result.payload:
4726 if inst in all_info:
4727 if all_info[inst].primary_node == name:
4728 live_data.update(result.payload)
4730 wrongnode_inst.add(inst)
4732 # orphan instance; we don't list it here as we don't
4733 # handle this case yet in the output of instance listing
4734 logging.warning("Orphan instance '%s' found on node %s",
4736 # else no instance is alive
4740 if query.IQ_DISKUSAGE in self.requested_data:
4741 disk_usage = dict((inst.name,
4742 _ComputeDiskSize(inst.disk_template,
4743 [{constants.IDISK_SIZE: disk.size}
4744 for disk in inst.disks]))
4745 for inst in instance_list)
4749 if query.IQ_CONSOLE in self.requested_data:
4751 for inst in instance_list:
4752 if inst.name in live_data:
4753 # Instance is running
4754 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
4756 consinfo[inst.name] = None
4757 assert set(consinfo.keys()) == set(instance_names)
4761 if query.IQ_NODES in self.requested_data:
4762 node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"),
4764 nodes = dict(lu.cfg.GetMultiNodeInfo(node_names))
4765 groups = dict((uuid, lu.cfg.GetNodeGroup(uuid))
4766 for uuid in set(map(operator.attrgetter("group"),
4772 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
4773 disk_usage, offline_nodes, bad_nodes,
4774 live_data, wrongnode_inst, consinfo,
4778 class LUQuery(NoHooksLU):
4779 """Query for resources/items of a certain kind.
4782 # pylint: disable=W0142
4785 def CheckArguments(self):
4786 qcls = _GetQueryImplementation(self.op.what)
4788 self.impl = qcls(self.op.filter, self.op.fields, self.op.use_locking)
4790 def ExpandNames(self):
4791 self.impl.ExpandNames(self)
4793 def DeclareLocks(self, level):
4794 self.impl.DeclareLocks(self, level)
4796 def Exec(self, feedback_fn):
4797 return self.impl.NewStyleQuery(self)
4800 class LUQueryFields(NoHooksLU):
4801 """Query for resources/items of a certain kind.
4804 # pylint: disable=W0142
4807 def CheckArguments(self):
4808 self.qcls = _GetQueryImplementation(self.op.what)
4810 def ExpandNames(self):
4811 self.needed_locks = {}
4813 def Exec(self, feedback_fn):
4814 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
4817 class LUNodeModifyStorage(NoHooksLU):
4818 """Logical unit for modifying a storage volume on a node.
4823 def CheckArguments(self):
4824 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4826 storage_type = self.op.storage_type
4829 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4831 raise errors.OpPrereqError("Storage units of type '%s' can not be"
4832 " modified" % storage_type,
4835 diff = set(self.op.changes.keys()) - modifiable
4837 raise errors.OpPrereqError("The following fields can not be modified for"
4838 " storage units of type '%s': %r" %
4839 (storage_type, list(diff)),
4842 def ExpandNames(self):
4843 self.needed_locks = {
4844 locking.LEVEL_NODE: self.op.node_name,
4847 def Exec(self, feedback_fn):
4848 """Computes the list of nodes and their attributes.
4851 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4852 result = self.rpc.call_storage_modify(self.op.node_name,
4853 self.op.storage_type, st_args,
4854 self.op.name, self.op.changes)
4855 result.Raise("Failed to modify storage unit '%s' on %s" %
4856 (self.op.name, self.op.node_name))
4859 class LUNodeAdd(LogicalUnit):
4860 """Logical unit for adding node to the cluster.
4864 HTYPE = constants.HTYPE_NODE
4865 _NFLAGS = ["master_capable", "vm_capable"]
4867 def CheckArguments(self):
4868 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4869 # validate/normalize the node name
4870 self.hostname = netutils.GetHostname(name=self.op.node_name,
4871 family=self.primary_ip_family)
4872 self.op.node_name = self.hostname.name
4874 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
4875 raise errors.OpPrereqError("Cannot readd the master node",
4878 if self.op.readd and self.op.group:
4879 raise errors.OpPrereqError("Cannot pass a node group when a node is"
4880 " being readded", errors.ECODE_INVAL)
4882 def BuildHooksEnv(self):
4885 This will run on all nodes before, and on all nodes + the new node after.
4889 "OP_TARGET": self.op.node_name,
4890 "NODE_NAME": self.op.node_name,
4891 "NODE_PIP": self.op.primary_ip,
4892 "NODE_SIP": self.op.secondary_ip,
4893 "MASTER_CAPABLE": str(self.op.master_capable),
4894 "VM_CAPABLE": str(self.op.vm_capable),
4897 def BuildHooksNodes(self):
4898 """Build hooks nodes.
4901 # Exclude added node
4902 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name]))
4903 post_nodes = pre_nodes + [self.op.node_name, ]
4905 return (pre_nodes, post_nodes)
4907 def CheckPrereq(self):
4908 """Check prerequisites.
4911 - the new node is not already in the config
4913 - its parameters (single/dual homed) matches the cluster
4915 Any errors are signaled by raising errors.OpPrereqError.
4919 hostname = self.hostname
4920 node = hostname.name
4921 primary_ip = self.op.primary_ip = hostname.ip
4922 if self.op.secondary_ip is None:
4923 if self.primary_ip_family == netutils.IP6Address.family:
4924 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4925 " IPv4 address must be given as secondary",
4927 self.op.secondary_ip = primary_ip
4929 secondary_ip = self.op.secondary_ip
4930 if not netutils.IP4Address.IsValid(secondary_ip):
4931 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4932 " address" % secondary_ip, errors.ECODE_INVAL)
4934 node_list = cfg.GetNodeList()
4935 if not self.op.readd and node in node_list:
4936 raise errors.OpPrereqError("Node %s is already in the configuration" %
4937 node, errors.ECODE_EXISTS)
4938 elif self.op.readd and node not in node_list:
4939 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4942 self.changed_primary_ip = False
4944 for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list):
4945 if self.op.readd and node == existing_node_name:
4946 if existing_node.secondary_ip != secondary_ip:
4947 raise errors.OpPrereqError("Readded node doesn't have the same IP"
4948 " address configuration as before",
4950 if existing_node.primary_ip != primary_ip:
4951 self.changed_primary_ip = True
4955 if (existing_node.primary_ip == primary_ip or
4956 existing_node.secondary_ip == primary_ip or
4957 existing_node.primary_ip == secondary_ip or
4958 existing_node.secondary_ip == secondary_ip):
4959 raise errors.OpPrereqError("New node ip address(es) conflict with"
4960 " existing node %s" % existing_node.name,
4961 errors.ECODE_NOTUNIQUE)
4963 # After this 'if' block, None is no longer a valid value for the
4964 # _capable op attributes
4966 old_node = self.cfg.GetNodeInfo(node)
4967 assert old_node is not None, "Can't retrieve locked node %s" % node
4968 for attr in self._NFLAGS:
4969 if getattr(self.op, attr) is None:
4970 setattr(self.op, attr, getattr(old_node, attr))
4972 for attr in self._NFLAGS:
4973 if getattr(self.op, attr) is None:
4974 setattr(self.op, attr, True)
4976 if self.op.readd and not self.op.vm_capable:
4977 pri, sec = cfg.GetNodeInstances(node)
4979 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4980 " flag set to false, but it already holds"
4981 " instances" % node,
4984 # check that the type of the node (single versus dual homed) is the
4985 # same as for the master
4986 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4987 master_singlehomed = myself.secondary_ip == myself.primary_ip
4988 newbie_singlehomed = secondary_ip == primary_ip
4989 if master_singlehomed != newbie_singlehomed:
4990 if master_singlehomed:
4991 raise errors.OpPrereqError("The master has no secondary ip but the"
4992 " new node has one",
4995 raise errors.OpPrereqError("The master has a secondary ip but the"
4996 " new node doesn't have one",
4999 # checks reachability
5000 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
5001 raise errors.OpPrereqError("Node not reachable by ping",
5002 errors.ECODE_ENVIRON)
5004 if not newbie_singlehomed:
5005 # check reachability from my secondary ip to newbie's secondary ip
5006 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
5007 source=myself.secondary_ip):
5008 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5009 " based ping to node daemon port",
5010 errors.ECODE_ENVIRON)
5017 if self.op.master_capable:
5018 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
5020 self.master_candidate = False
5023 self.new_node = old_node
5025 node_group = cfg.LookupNodeGroup(self.op.group)
5026 self.new_node = objects.Node(name=node,
5027 primary_ip=primary_ip,
5028 secondary_ip=secondary_ip,
5029 master_candidate=self.master_candidate,
5030 offline=False, drained=False,
5033 if self.op.ndparams:
5034 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
5036 def Exec(self, feedback_fn):
5037 """Adds the new node to the cluster.
5040 new_node = self.new_node
5041 node = new_node.name
5043 # We adding a new node so we assume it's powered
5044 new_node.powered = True
5046 # for re-adds, reset the offline/drained/master-candidate flags;
5047 # we need to reset here, otherwise offline would prevent RPC calls
5048 # later in the procedure; this also means that if the re-add
5049 # fails, we are left with a non-offlined, broken node
5051 new_node.drained = new_node.offline = False # pylint: disable=W0201
5052 self.LogInfo("Readding a node, the offline/drained flags were reset")
5053 # if we demote the node, we do cleanup later in the procedure
5054 new_node.master_candidate = self.master_candidate
5055 if self.changed_primary_ip:
5056 new_node.primary_ip = self.op.primary_ip
5058 # copy the master/vm_capable flags
5059 for attr in self._NFLAGS:
5060 setattr(new_node, attr, getattr(self.op, attr))
5062 # notify the user about any possible mc promotion
5063 if new_node.master_candidate:
5064 self.LogInfo("Node will be a master candidate")
5066 if self.op.ndparams:
5067 new_node.ndparams = self.op.ndparams
5069 new_node.ndparams = {}
5071 # check connectivity
5072 result = self.rpc.call_version([node])[node]
5073 result.Raise("Can't get version information from node %s" % node)
5074 if constants.PROTOCOL_VERSION == result.payload:
5075 logging.info("Communication to node %s fine, sw version %s match",
5076 node, result.payload)
5078 raise errors.OpExecError("Version mismatch master version %s,"
5079 " node version %s" %
5080 (constants.PROTOCOL_VERSION, result.payload))
5082 # Add node to our /etc/hosts, and add key to known_hosts
5083 if self.cfg.GetClusterInfo().modify_etc_hosts:
5084 master_node = self.cfg.GetMasterNode()
5085 result = self.rpc.call_etc_hosts_modify(master_node,
5086 constants.ETC_HOSTS_ADD,
5089 result.Raise("Can't update hosts file with new host data")
5091 if new_node.secondary_ip != new_node.primary_ip:
5092 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
5095 node_verify_list = [self.cfg.GetMasterNode()]
5096 node_verify_param = {
5097 constants.NV_NODELIST: ([node], {}),
5098 # TODO: do a node-net-test as well?
5101 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
5102 self.cfg.GetClusterName())
5103 for verifier in node_verify_list:
5104 result[verifier].Raise("Cannot communicate with node %s" % verifier)
5105 nl_payload = result[verifier].payload[constants.NV_NODELIST]
5107 for failed in nl_payload:
5108 feedback_fn("ssh/hostname verification failed"
5109 " (checking from %s): %s" %
5110 (verifier, nl_payload[failed]))
5111 raise errors.OpExecError("ssh/hostname verification failed")
5114 _RedistributeAncillaryFiles(self)
5115 self.context.ReaddNode(new_node)
5116 # make sure we redistribute the config
5117 self.cfg.Update(new_node, feedback_fn)
5118 # and make sure the new node will not have old files around
5119 if not new_node.master_candidate:
5120 result = self.rpc.call_node_demote_from_mc(new_node.name)
5121 msg = result.fail_msg
5123 self.LogWarning("Node failed to demote itself from master"
5124 " candidate status: %s" % msg)
5126 _RedistributeAncillaryFiles(self, additional_nodes=[node],
5127 additional_vm=self.op.vm_capable)
5128 self.context.AddNode(new_node, self.proc.GetECId())
5131 class LUNodeSetParams(LogicalUnit):
5132 """Modifies the parameters of a node.
5134 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
5135 to the node role (as _ROLE_*)
5136 @cvar _R2F: a dictionary from node role to tuples of flags
5137 @cvar _FLAGS: a list of attribute names corresponding to the flags
5140 HPATH = "node-modify"
5141 HTYPE = constants.HTYPE_NODE
5143 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
5145 (True, False, False): _ROLE_CANDIDATE,
5146 (False, True, False): _ROLE_DRAINED,
5147 (False, False, True): _ROLE_OFFLINE,
5148 (False, False, False): _ROLE_REGULAR,
5150 _R2F = dict((v, k) for k, v in _F2R.items())
5151 _FLAGS = ["master_candidate", "drained", "offline"]
5153 def CheckArguments(self):
5154 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5155 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
5156 self.op.master_capable, self.op.vm_capable,
5157 self.op.secondary_ip, self.op.ndparams]
5158 if all_mods.count(None) == len(all_mods):
5159 raise errors.OpPrereqError("Please pass at least one modification",
5161 if all_mods.count(True) > 1:
5162 raise errors.OpPrereqError("Can't set the node into more than one"
5163 " state at the same time",
5166 # Boolean value that tells us whether we might be demoting from MC
5167 self.might_demote = (self.op.master_candidate == False or
5168 self.op.offline == True or
5169 self.op.drained == True or
5170 self.op.master_capable == False)
5172 if self.op.secondary_ip:
5173 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
5174 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5175 " address" % self.op.secondary_ip,
5178 self.lock_all = self.op.auto_promote and self.might_demote
5179 self.lock_instances = self.op.secondary_ip is not None
5181 def ExpandNames(self):
5183 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
5185 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
5187 if self.lock_instances:
5188 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5190 def DeclareLocks(self, level):
5191 # If we have locked all instances, before waiting to lock nodes, release
5192 # all the ones living on nodes unrelated to the current operation.
5193 if level == locking.LEVEL_NODE and self.lock_instances:
5194 self.affected_instances = []
5195 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
5198 # Build list of instances to release
5199 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
5200 for instance_name, instance in self.cfg.GetMultiInstanceInfo(locked_i):
5201 if (instance.disk_template in constants.DTS_INT_MIRROR and
5202 self.op.node_name in instance.all_nodes):
5203 instances_keep.append(instance_name)
5204 self.affected_instances.append(instance)
5206 _ReleaseLocks(self, locking.LEVEL_INSTANCE, keep=instances_keep)
5208 assert (set(self.owned_locks(locking.LEVEL_INSTANCE)) ==
5209 set(instances_keep))
5211 def BuildHooksEnv(self):
5214 This runs on the master node.
5218 "OP_TARGET": self.op.node_name,
5219 "MASTER_CANDIDATE": str(self.op.master_candidate),
5220 "OFFLINE": str(self.op.offline),
5221 "DRAINED": str(self.op.drained),
5222 "MASTER_CAPABLE": str(self.op.master_capable),
5223 "VM_CAPABLE": str(self.op.vm_capable),
5226 def BuildHooksNodes(self):
5227 """Build hooks nodes.
5230 nl = [self.cfg.GetMasterNode(), self.op.node_name]
5233 def CheckPrereq(self):
5234 """Check prerequisites.
5236 This only checks the instance list against the existing names.
5239 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
5241 if (self.op.master_candidate is not None or
5242 self.op.drained is not None or
5243 self.op.offline is not None):
5244 # we can't change the master's node flags
5245 if self.op.node_name == self.cfg.GetMasterNode():
5246 raise errors.OpPrereqError("The master role can be changed"
5247 " only via master-failover",
5250 if self.op.master_candidate and not node.master_capable:
5251 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
5252 " it a master candidate" % node.name,
5255 if self.op.vm_capable == False:
5256 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
5258 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
5259 " the vm_capable flag" % node.name,
5262 if node.master_candidate and self.might_demote and not self.lock_all:
5263 assert not self.op.auto_promote, "auto_promote set but lock_all not"
5264 # check if after removing the current node, we're missing master
5266 (mc_remaining, mc_should, _) = \
5267 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
5268 if mc_remaining < mc_should:
5269 raise errors.OpPrereqError("Not enough master candidates, please"
5270 " pass auto promote option to allow"
5271 " promotion", errors.ECODE_STATE)
5273 self.old_flags = old_flags = (node.master_candidate,
5274 node.drained, node.offline)
5275 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
5276 self.old_role = old_role = self._F2R[old_flags]
5278 # Check for ineffective changes
5279 for attr in self._FLAGS:
5280 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
5281 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
5282 setattr(self.op, attr, None)
5284 # Past this point, any flag change to False means a transition
5285 # away from the respective state, as only real changes are kept
5287 # TODO: We might query the real power state if it supports OOB
5288 if _SupportsOob(self.cfg, node):
5289 if self.op.offline is False and not (node.powered or
5290 self.op.powered == True):
5291 raise errors.OpPrereqError(("Node %s needs to be turned on before its"
5292 " offline status can be reset") %
5294 elif self.op.powered is not None:
5295 raise errors.OpPrereqError(("Unable to change powered state for node %s"
5296 " as it does not support out-of-band"
5297 " handling") % self.op.node_name)
5299 # If we're being deofflined/drained, we'll MC ourself if needed
5300 if (self.op.drained == False or self.op.offline == False or
5301 (self.op.master_capable and not node.master_capable)):
5302 if _DecideSelfPromotion(self):
5303 self.op.master_candidate = True
5304 self.LogInfo("Auto-promoting node to master candidate")
5306 # If we're no longer master capable, we'll demote ourselves from MC
5307 if self.op.master_capable == False and node.master_candidate:
5308 self.LogInfo("Demoting from master candidate")
5309 self.op.master_candidate = False
5312 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
5313 if self.op.master_candidate:
5314 new_role = self._ROLE_CANDIDATE
5315 elif self.op.drained:
5316 new_role = self._ROLE_DRAINED
5317 elif self.op.offline:
5318 new_role = self._ROLE_OFFLINE
5319 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
5320 # False is still in new flags, which means we're un-setting (the
5322 new_role = self._ROLE_REGULAR
5323 else: # no new flags, nothing, keep old role
5326 self.new_role = new_role
5328 if old_role == self._ROLE_OFFLINE and new_role != old_role:
5329 # Trying to transition out of offline status
5330 result = self.rpc.call_version([node.name])[node.name]
5332 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
5333 " to report its version: %s" %
5334 (node.name, result.fail_msg),
5337 self.LogWarning("Transitioning node from offline to online state"
5338 " without using re-add. Please make sure the node"
5341 if self.op.secondary_ip:
5342 # Ok even without locking, because this can't be changed by any LU
5343 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
5344 master_singlehomed = master.secondary_ip == master.primary_ip
5345 if master_singlehomed and self.op.secondary_ip:
5346 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
5347 " homed cluster", errors.ECODE_INVAL)
5350 if self.affected_instances:
5351 raise errors.OpPrereqError("Cannot change secondary ip: offline"
5352 " node has instances (%s) configured"
5353 " to use it" % self.affected_instances)
5355 # On online nodes, check that no instances are running, and that
5356 # the node has the new ip and we can reach it.
5357 for instance in self.affected_instances:
5358 _CheckInstanceDown(self, instance, "cannot change secondary ip")
5360 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
5361 if master.name != node.name:
5362 # check reachability from master secondary ip to new secondary ip
5363 if not netutils.TcpPing(self.op.secondary_ip,
5364 constants.DEFAULT_NODED_PORT,
5365 source=master.secondary_ip):
5366 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5367 " based ping to node daemon port",
5368 errors.ECODE_ENVIRON)
5370 if self.op.ndparams:
5371 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
5372 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
5373 self.new_ndparams = new_ndparams
5375 def Exec(self, feedback_fn):
5380 old_role = self.old_role
5381 new_role = self.new_role
5385 if self.op.ndparams:
5386 node.ndparams = self.new_ndparams
5388 if self.op.powered is not None:
5389 node.powered = self.op.powered
5391 for attr in ["master_capable", "vm_capable"]:
5392 val = getattr(self.op, attr)
5394 setattr(node, attr, val)
5395 result.append((attr, str(val)))
5397 if new_role != old_role:
5398 # Tell the node to demote itself, if no longer MC and not offline
5399 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
5400 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
5402 self.LogWarning("Node failed to demote itself: %s", msg)
5404 new_flags = self._R2F[new_role]
5405 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
5407 result.append((desc, str(nf)))
5408 (node.master_candidate, node.drained, node.offline) = new_flags
5410 # we locked all nodes, we adjust the CP before updating this node
5412 _AdjustCandidatePool(self, [node.name])
5414 if self.op.secondary_ip:
5415 node.secondary_ip = self.op.secondary_ip
5416 result.append(("secondary_ip", self.op.secondary_ip))
5418 # this will trigger configuration file update, if needed
5419 self.cfg.Update(node, feedback_fn)
5421 # this will trigger job queue propagation or cleanup if the mc
5423 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
5424 self.context.ReaddNode(node)
5429 class LUNodePowercycle(NoHooksLU):
5430 """Powercycles a node.
5435 def CheckArguments(self):
5436 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5437 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
5438 raise errors.OpPrereqError("The node is the master and the force"
5439 " parameter was not set",
5442 def ExpandNames(self):
5443 """Locking for PowercycleNode.
5445 This is a last-resort option and shouldn't block on other
5446 jobs. Therefore, we grab no locks.
5449 self.needed_locks = {}
5451 def Exec(self, feedback_fn):
5455 result = self.rpc.call_node_powercycle(self.op.node_name,
5456 self.cfg.GetHypervisorType())
5457 result.Raise("Failed to schedule the reboot")
5458 return result.payload
5461 class LUClusterQuery(NoHooksLU):
5462 """Query cluster configuration.
5467 def ExpandNames(self):
5468 self.needed_locks = {}
5470 def Exec(self, feedback_fn):
5471 """Return cluster config.
5474 cluster = self.cfg.GetClusterInfo()
5477 # Filter just for enabled hypervisors
5478 for os_name, hv_dict in cluster.os_hvp.items():
5479 os_hvp[os_name] = {}
5480 for hv_name, hv_params in hv_dict.items():
5481 if hv_name in cluster.enabled_hypervisors:
5482 os_hvp[os_name][hv_name] = hv_params
5484 # Convert ip_family to ip_version
5485 primary_ip_version = constants.IP4_VERSION
5486 if cluster.primary_ip_family == netutils.IP6Address.family:
5487 primary_ip_version = constants.IP6_VERSION
5490 "software_version": constants.RELEASE_VERSION,
5491 "protocol_version": constants.PROTOCOL_VERSION,
5492 "config_version": constants.CONFIG_VERSION,
5493 "os_api_version": max(constants.OS_API_VERSIONS),
5494 "export_version": constants.EXPORT_VERSION,
5495 "architecture": (platform.architecture()[0], platform.machine()),
5496 "name": cluster.cluster_name,
5497 "master": cluster.master_node,
5498 "default_hypervisor": cluster.enabled_hypervisors[0],
5499 "enabled_hypervisors": cluster.enabled_hypervisors,
5500 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
5501 for hypervisor_name in cluster.enabled_hypervisors]),
5503 "beparams": cluster.beparams,
5504 "osparams": cluster.osparams,
5505 "nicparams": cluster.nicparams,
5506 "ndparams": cluster.ndparams,
5507 "candidate_pool_size": cluster.candidate_pool_size,
5508 "master_netdev": cluster.master_netdev,
5509 "volume_group_name": cluster.volume_group_name,
5510 "drbd_usermode_helper": cluster.drbd_usermode_helper,
5511 "file_storage_dir": cluster.file_storage_dir,
5512 "shared_file_storage_dir": cluster.shared_file_storage_dir,
5513 "maintain_node_health": cluster.maintain_node_health,
5514 "ctime": cluster.ctime,
5515 "mtime": cluster.mtime,
5516 "uuid": cluster.uuid,
5517 "tags": list(cluster.GetTags()),
5518 "uid_pool": cluster.uid_pool,
5519 "default_iallocator": cluster.default_iallocator,
5520 "reserved_lvs": cluster.reserved_lvs,
5521 "primary_ip_version": primary_ip_version,
5522 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
5523 "hidden_os": cluster.hidden_os,
5524 "blacklisted_os": cluster.blacklisted_os,
5530 class LUClusterConfigQuery(NoHooksLU):
5531 """Return configuration values.
5535 _FIELDS_DYNAMIC = utils.FieldSet()
5536 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
5537 "watcher_pause", "volume_group_name")
5539 def CheckArguments(self):
5540 _CheckOutputFields(static=self._FIELDS_STATIC,
5541 dynamic=self._FIELDS_DYNAMIC,
5542 selected=self.op.output_fields)
5544 def ExpandNames(self):
5545 self.needed_locks = {}
5547 def Exec(self, feedback_fn):
5548 """Dump a representation of the cluster config to the standard output.
5552 for field in self.op.output_fields:
5553 if field == "cluster_name":
5554 entry = self.cfg.GetClusterName()
5555 elif field == "master_node":
5556 entry = self.cfg.GetMasterNode()
5557 elif field == "drain_flag":
5558 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
5559 elif field == "watcher_pause":
5560 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
5561 elif field == "volume_group_name":
5562 entry = self.cfg.GetVGName()
5564 raise errors.ParameterError(field)
5565 values.append(entry)
5569 class LUInstanceActivateDisks(NoHooksLU):
5570 """Bring up an instance's disks.
5575 def ExpandNames(self):
5576 self._ExpandAndLockInstance()
5577 self.needed_locks[locking.LEVEL_NODE] = []
5578 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5580 def DeclareLocks(self, level):
5581 if level == locking.LEVEL_NODE:
5582 self._LockInstancesNodes()
5584 def CheckPrereq(self):
5585 """Check prerequisites.
5587 This checks that the instance is in the cluster.
5590 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5591 assert self.instance is not None, \
5592 "Cannot retrieve locked instance %s" % self.op.instance_name
5593 _CheckNodeOnline(self, self.instance.primary_node)
5595 def Exec(self, feedback_fn):
5596 """Activate the disks.
5599 disks_ok, disks_info = \
5600 _AssembleInstanceDisks(self, self.instance,
5601 ignore_size=self.op.ignore_size)
5603 raise errors.OpExecError("Cannot activate block devices")
5608 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
5610 """Prepare the block devices for an instance.
5612 This sets up the block devices on all nodes.
5614 @type lu: L{LogicalUnit}
5615 @param lu: the logical unit on whose behalf we execute
5616 @type instance: L{objects.Instance}
5617 @param instance: the instance for whose disks we assemble
5618 @type disks: list of L{objects.Disk} or None
5619 @param disks: which disks to assemble (or all, if None)
5620 @type ignore_secondaries: boolean
5621 @param ignore_secondaries: if true, errors on secondary nodes
5622 won't result in an error return from the function
5623 @type ignore_size: boolean
5624 @param ignore_size: if true, the current known size of the disk
5625 will not be used during the disk activation, useful for cases
5626 when the size is wrong
5627 @return: False if the operation failed, otherwise a list of
5628 (host, instance_visible_name, node_visible_name)
5629 with the mapping from node devices to instance devices
5634 iname = instance.name
5635 disks = _ExpandCheckDisks(instance, disks)
5637 # With the two passes mechanism we try to reduce the window of
5638 # opportunity for the race condition of switching DRBD to primary
5639 # before handshaking occured, but we do not eliminate it
5641 # The proper fix would be to wait (with some limits) until the
5642 # connection has been made and drbd transitions from WFConnection
5643 # into any other network-connected state (Connected, SyncTarget,
5646 # 1st pass, assemble on all nodes in secondary mode
5647 for idx, inst_disk in enumerate(disks):
5648 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
5650 node_disk = node_disk.Copy()
5651 node_disk.UnsetSize()
5652 lu.cfg.SetDiskID(node_disk, node)
5653 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
5654 msg = result.fail_msg
5656 lu.proc.LogWarning("Could not prepare block device %s on node %s"
5657 " (is_primary=False, pass=1): %s",
5658 inst_disk.iv_name, node, msg)
5659 if not ignore_secondaries:
5662 # FIXME: race condition on drbd migration to primary
5664 # 2nd pass, do only the primary node
5665 for idx, inst_disk in enumerate(disks):
5668 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
5669 if node != instance.primary_node:
5672 node_disk = node_disk.Copy()
5673 node_disk.UnsetSize()
5674 lu.cfg.SetDiskID(node_disk, node)
5675 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
5676 msg = result.fail_msg
5678 lu.proc.LogWarning("Could not prepare block device %s on node %s"
5679 " (is_primary=True, pass=2): %s",
5680 inst_disk.iv_name, node, msg)
5683 dev_path = result.payload
5685 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
5687 # leave the disks configured for the primary node
5688 # this is a workaround that would be fixed better by
5689 # improving the logical/physical id handling
5691 lu.cfg.SetDiskID(disk, instance.primary_node)
5693 return disks_ok, device_info
5696 def _StartInstanceDisks(lu, instance, force):
5697 """Start the disks of an instance.
5700 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
5701 ignore_secondaries=force)
5703 _ShutdownInstanceDisks(lu, instance)
5704 if force is not None and not force:
5705 lu.proc.LogWarning("", hint="If the message above refers to a"
5707 " you can retry the operation using '--force'.")
5708 raise errors.OpExecError("Disk consistency error")
5711 class LUInstanceDeactivateDisks(NoHooksLU):
5712 """Shutdown an instance's disks.
5717 def ExpandNames(self):
5718 self._ExpandAndLockInstance()
5719 self.needed_locks[locking.LEVEL_NODE] = []
5720 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5722 def DeclareLocks(self, level):
5723 if level == locking.LEVEL_NODE:
5724 self._LockInstancesNodes()
5726 def CheckPrereq(self):
5727 """Check prerequisites.
5729 This checks that the instance is in the cluster.
5732 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5733 assert self.instance is not None, \
5734 "Cannot retrieve locked instance %s" % self.op.instance_name
5736 def Exec(self, feedback_fn):
5737 """Deactivate the disks
5740 instance = self.instance
5742 _ShutdownInstanceDisks(self, instance)
5744 _SafeShutdownInstanceDisks(self, instance)
5747 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
5748 """Shutdown block devices of an instance.
5750 This function checks if an instance is running, before calling
5751 _ShutdownInstanceDisks.
5754 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
5755 _ShutdownInstanceDisks(lu, instance, disks=disks)
5758 def _ExpandCheckDisks(instance, disks):
5759 """Return the instance disks selected by the disks list
5761 @type disks: list of L{objects.Disk} or None
5762 @param disks: selected disks
5763 @rtype: list of L{objects.Disk}
5764 @return: selected instance disks to act on
5768 return instance.disks
5770 if not set(disks).issubset(instance.disks):
5771 raise errors.ProgrammerError("Can only act on disks belonging to the"
5776 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
5777 """Shutdown block devices of an instance.
5779 This does the shutdown on all nodes of the instance.
5781 If the ignore_primary is false, errors on the primary node are
5786 disks = _ExpandCheckDisks(instance, disks)
5789 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
5790 lu.cfg.SetDiskID(top_disk, node)
5791 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
5792 msg = result.fail_msg
5794 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
5795 disk.iv_name, node, msg)
5796 if ((node == instance.primary_node and not ignore_primary) or
5797 (node != instance.primary_node and not result.offline)):
5802 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
5803 """Checks if a node has enough free memory.
5805 This function check if a given node has the needed amount of free
5806 memory. In case the node has less memory or we cannot get the
5807 information from the node, this function raise an OpPrereqError
5810 @type lu: C{LogicalUnit}
5811 @param lu: a logical unit from which we get configuration data
5813 @param node: the node to check
5814 @type reason: C{str}
5815 @param reason: string to use in the error message
5816 @type requested: C{int}
5817 @param requested: the amount of memory in MiB to check for
5818 @type hypervisor_name: C{str}
5819 @param hypervisor_name: the hypervisor to ask for memory stats
5820 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
5821 we cannot check the node
5824 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
5825 nodeinfo[node].Raise("Can't get data from node %s" % node,
5826 prereq=True, ecode=errors.ECODE_ENVIRON)
5827 free_mem = nodeinfo[node].payload.get("memory_free", None)
5828 if not isinstance(free_mem, int):
5829 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
5830 " was '%s'" % (node, free_mem),
5831 errors.ECODE_ENVIRON)
5832 if requested > free_mem:
5833 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
5834 " needed %s MiB, available %s MiB" %
5835 (node, reason, requested, free_mem),
5839 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5840 """Checks if nodes have enough free disk space in the all VGs.
5842 This function check if all given nodes have the needed amount of
5843 free disk. In case any node has less disk or we cannot get the
5844 information from the node, this function raise an OpPrereqError
5847 @type lu: C{LogicalUnit}
5848 @param lu: a logical unit from which we get configuration data
5849 @type nodenames: C{list}
5850 @param nodenames: the list of node names to check
5851 @type req_sizes: C{dict}
5852 @param req_sizes: the hash of vg and corresponding amount of disk in
5854 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5855 or we cannot check the node
5858 for vg, req_size in req_sizes.items():
5859 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5862 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5863 """Checks if nodes have enough free disk space in the specified VG.
5865 This function check if all given nodes have the needed amount of
5866 free disk. In case any node has less disk or we cannot get the
5867 information from the node, this function raise an OpPrereqError
5870 @type lu: C{LogicalUnit}
5871 @param lu: a logical unit from which we get configuration data
5872 @type nodenames: C{list}
5873 @param nodenames: the list of node names to check
5875 @param vg: the volume group to check
5876 @type requested: C{int}
5877 @param requested: the amount of disk in MiB to check for
5878 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5879 or we cannot check the node
5882 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
5883 for node in nodenames:
5884 info = nodeinfo[node]
5885 info.Raise("Cannot get current information from node %s" % node,
5886 prereq=True, ecode=errors.ECODE_ENVIRON)
5887 vg_free = info.payload.get("vg_free", None)
5888 if not isinstance(vg_free, int):
5889 raise errors.OpPrereqError("Can't compute free disk space on node"
5890 " %s for vg %s, result was '%s'" %
5891 (node, vg, vg_free), errors.ECODE_ENVIRON)
5892 if requested > vg_free:
5893 raise errors.OpPrereqError("Not enough disk space on target node %s"
5894 " vg %s: required %d MiB, available %d MiB" %
5895 (node, vg, requested, vg_free),
5899 class LUInstanceStartup(LogicalUnit):
5900 """Starts an instance.
5903 HPATH = "instance-start"
5904 HTYPE = constants.HTYPE_INSTANCE
5907 def CheckArguments(self):
5909 if self.op.beparams:
5910 # fill the beparams dict
5911 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5913 def ExpandNames(self):
5914 self._ExpandAndLockInstance()
5916 def BuildHooksEnv(self):
5919 This runs on master, primary and secondary nodes of the instance.
5923 "FORCE": self.op.force,
5926 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5930 def BuildHooksNodes(self):
5931 """Build hooks nodes.
5934 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5937 def CheckPrereq(self):
5938 """Check prerequisites.
5940 This checks that the instance is in the cluster.
5943 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5944 assert self.instance is not None, \
5945 "Cannot retrieve locked instance %s" % self.op.instance_name
5948 if self.op.hvparams:
5949 # check hypervisor parameter syntax (locally)
5950 cluster = self.cfg.GetClusterInfo()
5951 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5952 filled_hvp = cluster.FillHV(instance)
5953 filled_hvp.update(self.op.hvparams)
5954 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5955 hv_type.CheckParameterSyntax(filled_hvp)
5956 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5958 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5960 if self.primary_offline and self.op.ignore_offline_nodes:
5961 self.proc.LogWarning("Ignoring offline primary node")
5963 if self.op.hvparams or self.op.beparams:
5964 self.proc.LogWarning("Overridden parameters are ignored")
5966 _CheckNodeOnline(self, instance.primary_node)
5968 bep = self.cfg.GetClusterInfo().FillBE(instance)
5970 # check bridges existence
5971 _CheckInstanceBridgesExist(self, instance)
5973 remote_info = self.rpc.call_instance_info(instance.primary_node,
5975 instance.hypervisor)
5976 remote_info.Raise("Error checking node %s" % instance.primary_node,
5977 prereq=True, ecode=errors.ECODE_ENVIRON)
5978 if not remote_info.payload: # not running already
5979 _CheckNodeFreeMemory(self, instance.primary_node,
5980 "starting instance %s" % instance.name,
5981 bep[constants.BE_MEMORY], instance.hypervisor)
5983 def Exec(self, feedback_fn):
5984 """Start the instance.
5987 instance = self.instance
5988 force = self.op.force
5990 if not self.op.no_remember:
5991 self.cfg.MarkInstanceUp(instance.name)
5993 if self.primary_offline:
5994 assert self.op.ignore_offline_nodes
5995 self.proc.LogInfo("Primary node offline, marked instance as started")
5997 node_current = instance.primary_node
5999 _StartInstanceDisks(self, instance, force)
6001 result = self.rpc.call_instance_start(node_current, instance,
6002 self.op.hvparams, self.op.beparams,
6003 self.op.startup_paused)
6004 msg = result.fail_msg
6006 _ShutdownInstanceDisks(self, instance)
6007 raise errors.OpExecError("Could not start instance: %s" % msg)
6010 class LUInstanceReboot(LogicalUnit):
6011 """Reboot an instance.
6014 HPATH = "instance-reboot"
6015 HTYPE = constants.HTYPE_INSTANCE
6018 def ExpandNames(self):
6019 self._ExpandAndLockInstance()
6021 def BuildHooksEnv(self):
6024 This runs on master, primary and secondary nodes of the instance.
6028 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
6029 "REBOOT_TYPE": self.op.reboot_type,
6030 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6033 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6037 def BuildHooksNodes(self):
6038 """Build hooks nodes.
6041 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6044 def CheckPrereq(self):
6045 """Check prerequisites.
6047 This checks that the instance is in the cluster.
6050 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6051 assert self.instance is not None, \
6052 "Cannot retrieve locked instance %s" % self.op.instance_name
6054 _CheckNodeOnline(self, instance.primary_node)
6056 # check bridges existence
6057 _CheckInstanceBridgesExist(self, instance)
6059 def Exec(self, feedback_fn):
6060 """Reboot the instance.
6063 instance = self.instance
6064 ignore_secondaries = self.op.ignore_secondaries
6065 reboot_type = self.op.reboot_type
6067 remote_info = self.rpc.call_instance_info(instance.primary_node,
6069 instance.hypervisor)
6070 remote_info.Raise("Error checking node %s" % instance.primary_node)
6071 instance_running = bool(remote_info.payload)
6073 node_current = instance.primary_node
6075 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
6076 constants.INSTANCE_REBOOT_HARD]:
6077 for disk in instance.disks:
6078 self.cfg.SetDiskID(disk, node_current)
6079 result = self.rpc.call_instance_reboot(node_current, instance,
6081 self.op.shutdown_timeout)
6082 result.Raise("Could not reboot instance")
6084 if instance_running:
6085 result = self.rpc.call_instance_shutdown(node_current, instance,
6086 self.op.shutdown_timeout)
6087 result.Raise("Could not shutdown instance for full reboot")
6088 _ShutdownInstanceDisks(self, instance)
6090 self.LogInfo("Instance %s was already stopped, starting now",
6092 _StartInstanceDisks(self, instance, ignore_secondaries)
6093 result = self.rpc.call_instance_start(node_current, instance,
6095 msg = result.fail_msg
6097 _ShutdownInstanceDisks(self, instance)
6098 raise errors.OpExecError("Could not start instance for"
6099 " full reboot: %s" % msg)
6101 self.cfg.MarkInstanceUp(instance.name)
6104 class LUInstanceShutdown(LogicalUnit):
6105 """Shutdown an instance.
6108 HPATH = "instance-stop"
6109 HTYPE = constants.HTYPE_INSTANCE
6112 def ExpandNames(self):
6113 self._ExpandAndLockInstance()
6115 def BuildHooksEnv(self):
6118 This runs on master, primary and secondary nodes of the instance.
6121 env = _BuildInstanceHookEnvByObject(self, self.instance)
6122 env["TIMEOUT"] = self.op.timeout
6125 def BuildHooksNodes(self):
6126 """Build hooks nodes.
6129 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6132 def CheckPrereq(self):
6133 """Check prerequisites.
6135 This checks that the instance is in the cluster.
6138 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6139 assert self.instance is not None, \
6140 "Cannot retrieve locked instance %s" % self.op.instance_name
6142 self.primary_offline = \
6143 self.cfg.GetNodeInfo(self.instance.primary_node).offline
6145 if self.primary_offline and self.op.ignore_offline_nodes:
6146 self.proc.LogWarning("Ignoring offline primary node")
6148 _CheckNodeOnline(self, self.instance.primary_node)
6150 def Exec(self, feedback_fn):
6151 """Shutdown the instance.
6154 instance = self.instance
6155 node_current = instance.primary_node
6156 timeout = self.op.timeout
6158 if not self.op.no_remember:
6159 self.cfg.MarkInstanceDown(instance.name)
6161 if self.primary_offline:
6162 assert self.op.ignore_offline_nodes
6163 self.proc.LogInfo("Primary node offline, marked instance as stopped")
6165 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
6166 msg = result.fail_msg
6168 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
6170 _ShutdownInstanceDisks(self, instance)
6173 class LUInstanceReinstall(LogicalUnit):
6174 """Reinstall an instance.
6177 HPATH = "instance-reinstall"
6178 HTYPE = constants.HTYPE_INSTANCE
6181 def ExpandNames(self):
6182 self._ExpandAndLockInstance()
6184 def BuildHooksEnv(self):
6187 This runs on master, primary and secondary nodes of the instance.
6190 return _BuildInstanceHookEnvByObject(self, self.instance)
6192 def BuildHooksNodes(self):
6193 """Build hooks nodes.
6196 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6199 def CheckPrereq(self):
6200 """Check prerequisites.
6202 This checks that the instance is in the cluster and is not running.
6205 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6206 assert instance is not None, \
6207 "Cannot retrieve locked instance %s" % self.op.instance_name
6208 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
6209 " offline, cannot reinstall")
6210 for node in instance.secondary_nodes:
6211 _CheckNodeOnline(self, node, "Instance secondary node offline,"
6212 " cannot reinstall")
6214 if instance.disk_template == constants.DT_DISKLESS:
6215 raise errors.OpPrereqError("Instance '%s' has no disks" %
6216 self.op.instance_name,
6218 _CheckInstanceDown(self, instance, "cannot reinstall")
6220 if self.op.os_type is not None:
6222 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
6223 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
6224 instance_os = self.op.os_type
6226 instance_os = instance.os
6228 nodelist = list(instance.all_nodes)
6230 if self.op.osparams:
6231 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
6232 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
6233 self.os_inst = i_osdict # the new dict (without defaults)
6237 self.instance = instance
6239 def Exec(self, feedback_fn):
6240 """Reinstall the instance.
6243 inst = self.instance
6245 if self.op.os_type is not None:
6246 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
6247 inst.os = self.op.os_type
6248 # Write to configuration
6249 self.cfg.Update(inst, feedback_fn)
6251 _StartInstanceDisks(self, inst, None)
6253 feedback_fn("Running the instance OS create scripts...")
6254 # FIXME: pass debug option from opcode to backend
6255 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
6256 self.op.debug_level,
6257 osparams=self.os_inst)
6258 result.Raise("Could not install OS for instance %s on node %s" %
6259 (inst.name, inst.primary_node))
6261 _ShutdownInstanceDisks(self, inst)
6264 class LUInstanceRecreateDisks(LogicalUnit):
6265 """Recreate an instance's missing disks.
6268 HPATH = "instance-recreate-disks"
6269 HTYPE = constants.HTYPE_INSTANCE
6272 def CheckArguments(self):
6273 # normalise the disk list
6274 self.op.disks = sorted(frozenset(self.op.disks))
6276 def ExpandNames(self):
6277 self._ExpandAndLockInstance()
6278 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6280 self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes]
6281 self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes)
6283 self.needed_locks[locking.LEVEL_NODE] = []
6285 def DeclareLocks(self, level):
6286 if level == locking.LEVEL_NODE:
6287 # if we replace the nodes, we only need to lock the old primary,
6288 # otherwise we need to lock all nodes for disk re-creation
6289 primary_only = bool(self.op.nodes)
6290 self._LockInstancesNodes(primary_only=primary_only)
6292 def BuildHooksEnv(self):
6295 This runs on master, primary and secondary nodes of the instance.
6298 return _BuildInstanceHookEnvByObject(self, self.instance)
6300 def BuildHooksNodes(self):
6301 """Build hooks nodes.
6304 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6307 def CheckPrereq(self):
6308 """Check prerequisites.
6310 This checks that the instance is in the cluster and is not running.
6313 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6314 assert instance is not None, \
6315 "Cannot retrieve locked instance %s" % self.op.instance_name
6317 if len(self.op.nodes) != len(instance.all_nodes):
6318 raise errors.OpPrereqError("Instance %s currently has %d nodes, but"
6319 " %d replacement nodes were specified" %
6320 (instance.name, len(instance.all_nodes),
6321 len(self.op.nodes)),
6323 assert instance.disk_template != constants.DT_DRBD8 or \
6324 len(self.op.nodes) == 2
6325 assert instance.disk_template != constants.DT_PLAIN or \
6326 len(self.op.nodes) == 1
6327 primary_node = self.op.nodes[0]
6329 primary_node = instance.primary_node
6330 _CheckNodeOnline(self, primary_node)
6332 if instance.disk_template == constants.DT_DISKLESS:
6333 raise errors.OpPrereqError("Instance '%s' has no disks" %
6334 self.op.instance_name, errors.ECODE_INVAL)
6335 # if we replace nodes *and* the old primary is offline, we don't
6337 assert instance.primary_node in self.needed_locks[locking.LEVEL_NODE]
6338 old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
6339 if not (self.op.nodes and old_pnode.offline):
6340 _CheckInstanceDown(self, instance, "cannot recreate disks")
6342 if not self.op.disks:
6343 self.op.disks = range(len(instance.disks))
6345 for idx in self.op.disks:
6346 if idx >= len(instance.disks):
6347 raise errors.OpPrereqError("Invalid disk index '%s'" % idx,
6349 if self.op.disks != range(len(instance.disks)) and self.op.nodes:
6350 raise errors.OpPrereqError("Can't recreate disks partially and"
6351 " change the nodes at the same time",
6353 self.instance = instance
6355 def Exec(self, feedback_fn):
6356 """Recreate the disks.
6359 instance = self.instance
6362 mods = [] # keeps track of needed logical_id changes
6364 for idx, disk in enumerate(instance.disks):
6365 if idx not in self.op.disks: # disk idx has not been passed in
6368 # update secondaries for disks, if needed
6370 if disk.dev_type == constants.LD_DRBD8:
6371 # need to update the nodes and minors
6372 assert len(self.op.nodes) == 2
6373 assert len(disk.logical_id) == 6 # otherwise disk internals
6375 (_, _, old_port, _, _, old_secret) = disk.logical_id
6376 new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name)
6377 new_id = (self.op.nodes[0], self.op.nodes[1], old_port,
6378 new_minors[0], new_minors[1], old_secret)
6379 assert len(disk.logical_id) == len(new_id)
6380 mods.append((idx, new_id))
6382 # now that we have passed all asserts above, we can apply the mods
6383 # in a single run (to avoid partial changes)
6384 for idx, new_id in mods:
6385 instance.disks[idx].logical_id = new_id
6387 # change primary node, if needed
6389 instance.primary_node = self.op.nodes[0]
6390 self.LogWarning("Changing the instance's nodes, you will have to"
6391 " remove any disks left on the older nodes manually")
6394 self.cfg.Update(instance, feedback_fn)
6396 _CreateDisks(self, instance, to_skip=to_skip)
6399 class LUInstanceRename(LogicalUnit):
6400 """Rename an instance.
6403 HPATH = "instance-rename"
6404 HTYPE = constants.HTYPE_INSTANCE
6406 def CheckArguments(self):
6410 if self.op.ip_check and not self.op.name_check:
6411 # TODO: make the ip check more flexible and not depend on the name check
6412 raise errors.OpPrereqError("IP address check requires a name check",
6415 def BuildHooksEnv(self):
6418 This runs on master, primary and secondary nodes of the instance.
6421 env = _BuildInstanceHookEnvByObject(self, self.instance)
6422 env["INSTANCE_NEW_NAME"] = self.op.new_name
6425 def BuildHooksNodes(self):
6426 """Build hooks nodes.
6429 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6432 def CheckPrereq(self):
6433 """Check prerequisites.
6435 This checks that the instance is in the cluster and is not running.
6438 self.op.instance_name = _ExpandInstanceName(self.cfg,
6439 self.op.instance_name)
6440 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6441 assert instance is not None
6442 _CheckNodeOnline(self, instance.primary_node)
6443 _CheckInstanceDown(self, instance, "cannot rename")
6444 self.instance = instance
6446 new_name = self.op.new_name
6447 if self.op.name_check:
6448 hostname = netutils.GetHostname(name=new_name)
6449 if hostname.name != new_name:
6450 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
6452 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
6453 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
6454 " same as given hostname '%s'") %
6455 (hostname.name, self.op.new_name),
6457 new_name = self.op.new_name = hostname.name
6458 if (self.op.ip_check and
6459 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
6460 raise errors.OpPrereqError("IP %s of instance %s already in use" %
6461 (hostname.ip, new_name),
6462 errors.ECODE_NOTUNIQUE)
6464 instance_list = self.cfg.GetInstanceList()
6465 if new_name in instance_list and new_name != instance.name:
6466 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6467 new_name, errors.ECODE_EXISTS)
6469 def Exec(self, feedback_fn):
6470 """Rename the instance.
6473 inst = self.instance
6474 old_name = inst.name
6476 rename_file_storage = False
6477 if (inst.disk_template in constants.DTS_FILEBASED and
6478 self.op.new_name != inst.name):
6479 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
6480 rename_file_storage = True
6482 self.cfg.RenameInstance(inst.name, self.op.new_name)
6483 # Change the instance lock. This is definitely safe while we hold the BGL.
6484 # Otherwise the new lock would have to be added in acquired mode.
6486 self.glm.remove(locking.LEVEL_INSTANCE, old_name)
6487 self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
6489 # re-read the instance from the configuration after rename
6490 inst = self.cfg.GetInstanceInfo(self.op.new_name)
6492 if rename_file_storage:
6493 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
6494 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
6495 old_file_storage_dir,
6496 new_file_storage_dir)
6497 result.Raise("Could not rename on node %s directory '%s' to '%s'"
6498 " (but the instance has been renamed in Ganeti)" %
6499 (inst.primary_node, old_file_storage_dir,
6500 new_file_storage_dir))
6502 _StartInstanceDisks(self, inst, None)
6504 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
6505 old_name, self.op.debug_level)
6506 msg = result.fail_msg
6508 msg = ("Could not run OS rename script for instance %s on node %s"
6509 " (but the instance has been renamed in Ganeti): %s" %
6510 (inst.name, inst.primary_node, msg))
6511 self.proc.LogWarning(msg)
6513 _ShutdownInstanceDisks(self, inst)
6518 class LUInstanceRemove(LogicalUnit):
6519 """Remove an instance.
6522 HPATH = "instance-remove"
6523 HTYPE = constants.HTYPE_INSTANCE
6526 def ExpandNames(self):
6527 self._ExpandAndLockInstance()
6528 self.needed_locks[locking.LEVEL_NODE] = []
6529 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6531 def DeclareLocks(self, level):
6532 if level == locking.LEVEL_NODE:
6533 self._LockInstancesNodes()
6535 def BuildHooksEnv(self):
6538 This runs on master, primary and secondary nodes of the instance.
6541 env = _BuildInstanceHookEnvByObject(self, self.instance)
6542 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
6545 def BuildHooksNodes(self):
6546 """Build hooks nodes.
6549 nl = [self.cfg.GetMasterNode()]
6550 nl_post = list(self.instance.all_nodes) + nl
6551 return (nl, nl_post)
6553 def CheckPrereq(self):
6554 """Check prerequisites.
6556 This checks that the instance is in the cluster.
6559 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6560 assert self.instance is not None, \
6561 "Cannot retrieve locked instance %s" % self.op.instance_name
6563 def Exec(self, feedback_fn):
6564 """Remove the instance.
6567 instance = self.instance
6568 logging.info("Shutting down instance %s on node %s",
6569 instance.name, instance.primary_node)
6571 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
6572 self.op.shutdown_timeout)
6573 msg = result.fail_msg
6575 if self.op.ignore_failures:
6576 feedback_fn("Warning: can't shutdown instance: %s" % msg)
6578 raise errors.OpExecError("Could not shutdown instance %s on"
6580 (instance.name, instance.primary_node, msg))
6582 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
6585 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
6586 """Utility function to remove an instance.
6589 logging.info("Removing block devices for instance %s", instance.name)
6591 if not _RemoveDisks(lu, instance):
6592 if not ignore_failures:
6593 raise errors.OpExecError("Can't remove instance's disks")
6594 feedback_fn("Warning: can't remove instance's disks")
6596 logging.info("Removing instance %s out of cluster config", instance.name)
6598 lu.cfg.RemoveInstance(instance.name)
6600 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
6601 "Instance lock removal conflict"
6603 # Remove lock for the instance
6604 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
6607 class LUInstanceQuery(NoHooksLU):
6608 """Logical unit for querying instances.
6611 # pylint: disable=W0142
6614 def CheckArguments(self):
6615 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
6616 self.op.output_fields, self.op.use_locking)
6618 def ExpandNames(self):
6619 self.iq.ExpandNames(self)
6621 def DeclareLocks(self, level):
6622 self.iq.DeclareLocks(self, level)
6624 def Exec(self, feedback_fn):
6625 return self.iq.OldStyleQuery(self)
6628 class LUInstanceFailover(LogicalUnit):
6629 """Failover an instance.
6632 HPATH = "instance-failover"
6633 HTYPE = constants.HTYPE_INSTANCE
6636 def CheckArguments(self):
6637 """Check the arguments.
6640 self.iallocator = getattr(self.op, "iallocator", None)
6641 self.target_node = getattr(self.op, "target_node", None)
6643 def ExpandNames(self):
6644 self._ExpandAndLockInstance()
6646 if self.op.target_node is not None:
6647 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6649 self.needed_locks[locking.LEVEL_NODE] = []
6650 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6652 ignore_consistency = self.op.ignore_consistency
6653 shutdown_timeout = self.op.shutdown_timeout
6654 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6657 ignore_consistency=ignore_consistency,
6658 shutdown_timeout=shutdown_timeout)
6659 self.tasklets = [self._migrater]
6661 def DeclareLocks(self, level):
6662 if level == locking.LEVEL_NODE:
6663 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6664 if instance.disk_template in constants.DTS_EXT_MIRROR:
6665 if self.op.target_node is None:
6666 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6668 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6669 self.op.target_node]
6670 del self.recalculate_locks[locking.LEVEL_NODE]
6672 self._LockInstancesNodes()
6674 def BuildHooksEnv(self):
6677 This runs on master, primary and secondary nodes of the instance.
6680 instance = self._migrater.instance
6681 source_node = instance.primary_node
6682 target_node = self.op.target_node
6684 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
6685 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6686 "OLD_PRIMARY": source_node,
6687 "NEW_PRIMARY": target_node,
6690 if instance.disk_template in constants.DTS_INT_MIRROR:
6691 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
6692 env["NEW_SECONDARY"] = source_node
6694 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
6696 env.update(_BuildInstanceHookEnvByObject(self, instance))
6700 def BuildHooksNodes(self):
6701 """Build hooks nodes.
6704 instance = self._migrater.instance
6705 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6706 return (nl, nl + [instance.primary_node])
6709 class LUInstanceMigrate(LogicalUnit):
6710 """Migrate an instance.
6712 This is migration without shutting down, compared to the failover,
6713 which is done with shutdown.
6716 HPATH = "instance-migrate"
6717 HTYPE = constants.HTYPE_INSTANCE
6720 def ExpandNames(self):
6721 self._ExpandAndLockInstance()
6723 if self.op.target_node is not None:
6724 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6726 self.needed_locks[locking.LEVEL_NODE] = []
6727 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6729 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6730 cleanup=self.op.cleanup,
6732 fallback=self.op.allow_failover)
6733 self.tasklets = [self._migrater]
6735 def DeclareLocks(self, level):
6736 if level == locking.LEVEL_NODE:
6737 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6738 if instance.disk_template in constants.DTS_EXT_MIRROR:
6739 if self.op.target_node is None:
6740 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6742 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6743 self.op.target_node]
6744 del self.recalculate_locks[locking.LEVEL_NODE]
6746 self._LockInstancesNodes()
6748 def BuildHooksEnv(self):
6751 This runs on master, primary and secondary nodes of the instance.
6754 instance = self._migrater.instance
6755 source_node = instance.primary_node
6756 target_node = self.op.target_node
6757 env = _BuildInstanceHookEnvByObject(self, instance)
6759 "MIGRATE_LIVE": self._migrater.live,
6760 "MIGRATE_CLEANUP": self.op.cleanup,
6761 "OLD_PRIMARY": source_node,
6762 "NEW_PRIMARY": target_node,
6765 if instance.disk_template in constants.DTS_INT_MIRROR:
6766 env["OLD_SECONDARY"] = target_node
6767 env["NEW_SECONDARY"] = source_node
6769 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
6773 def BuildHooksNodes(self):
6774 """Build hooks nodes.
6777 instance = self._migrater.instance
6778 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6779 return (nl, nl + [instance.primary_node])
6782 class LUInstanceMove(LogicalUnit):
6783 """Move an instance by data-copying.
6786 HPATH = "instance-move"
6787 HTYPE = constants.HTYPE_INSTANCE
6790 def ExpandNames(self):
6791 self._ExpandAndLockInstance()
6792 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6793 self.op.target_node = target_node
6794 self.needed_locks[locking.LEVEL_NODE] = [target_node]
6795 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6797 def DeclareLocks(self, level):
6798 if level == locking.LEVEL_NODE:
6799 self._LockInstancesNodes(primary_only=True)
6801 def BuildHooksEnv(self):
6804 This runs on master, primary and secondary nodes of the instance.
6808 "TARGET_NODE": self.op.target_node,
6809 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6811 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6814 def BuildHooksNodes(self):
6815 """Build hooks nodes.
6819 self.cfg.GetMasterNode(),
6820 self.instance.primary_node,
6821 self.op.target_node,
6825 def CheckPrereq(self):
6826 """Check prerequisites.
6828 This checks that the instance is in the cluster.
6831 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6832 assert self.instance is not None, \
6833 "Cannot retrieve locked instance %s" % self.op.instance_name
6835 node = self.cfg.GetNodeInfo(self.op.target_node)
6836 assert node is not None, \
6837 "Cannot retrieve locked node %s" % self.op.target_node
6839 self.target_node = target_node = node.name
6841 if target_node == instance.primary_node:
6842 raise errors.OpPrereqError("Instance %s is already on the node %s" %
6843 (instance.name, target_node),
6846 bep = self.cfg.GetClusterInfo().FillBE(instance)
6848 for idx, dsk in enumerate(instance.disks):
6849 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
6850 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
6851 " cannot copy" % idx, errors.ECODE_STATE)
6853 _CheckNodeOnline(self, target_node)
6854 _CheckNodeNotDrained(self, target_node)
6855 _CheckNodeVmCapable(self, target_node)
6857 if instance.admin_up:
6858 # check memory requirements on the secondary node
6859 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
6860 instance.name, bep[constants.BE_MEMORY],
6861 instance.hypervisor)
6863 self.LogInfo("Not checking memory on the secondary node as"
6864 " instance will not be started")
6866 # check bridge existance
6867 _CheckInstanceBridgesExist(self, instance, node=target_node)
6869 def Exec(self, feedback_fn):
6870 """Move an instance.
6872 The move is done by shutting it down on its present node, copying
6873 the data over (slow) and starting it on the new node.
6876 instance = self.instance
6878 source_node = instance.primary_node
6879 target_node = self.target_node
6881 self.LogInfo("Shutting down instance %s on source node %s",
6882 instance.name, source_node)
6884 result = self.rpc.call_instance_shutdown(source_node, instance,
6885 self.op.shutdown_timeout)
6886 msg = result.fail_msg
6888 if self.op.ignore_consistency:
6889 self.proc.LogWarning("Could not shutdown instance %s on node %s."
6890 " Proceeding anyway. Please make sure node"
6891 " %s is down. Error details: %s",
6892 instance.name, source_node, source_node, msg)
6894 raise errors.OpExecError("Could not shutdown instance %s on"
6896 (instance.name, source_node, msg))
6898 # create the target disks
6900 _CreateDisks(self, instance, target_node=target_node)
6901 except errors.OpExecError:
6902 self.LogWarning("Device creation failed, reverting...")
6904 _RemoveDisks(self, instance, target_node=target_node)
6906 self.cfg.ReleaseDRBDMinors(instance.name)
6909 cluster_name = self.cfg.GetClusterInfo().cluster_name
6912 # activate, get path, copy the data over
6913 for idx, disk in enumerate(instance.disks):
6914 self.LogInfo("Copying data for disk %d", idx)
6915 result = self.rpc.call_blockdev_assemble(target_node, disk,
6916 instance.name, True, idx)
6918 self.LogWarning("Can't assemble newly created disk %d: %s",
6919 idx, result.fail_msg)
6920 errs.append(result.fail_msg)
6922 dev_path = result.payload
6923 result = self.rpc.call_blockdev_export(source_node, disk,
6924 target_node, dev_path,
6927 self.LogWarning("Can't copy data over for disk %d: %s",
6928 idx, result.fail_msg)
6929 errs.append(result.fail_msg)
6933 self.LogWarning("Some disks failed to copy, aborting")
6935 _RemoveDisks(self, instance, target_node=target_node)
6937 self.cfg.ReleaseDRBDMinors(instance.name)
6938 raise errors.OpExecError("Errors during disk copy: %s" %
6941 instance.primary_node = target_node
6942 self.cfg.Update(instance, feedback_fn)
6944 self.LogInfo("Removing the disks on the original node")
6945 _RemoveDisks(self, instance, target_node=source_node)
6947 # Only start the instance if it's marked as up
6948 if instance.admin_up:
6949 self.LogInfo("Starting instance %s on node %s",
6950 instance.name, target_node)
6952 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6953 ignore_secondaries=True)
6955 _ShutdownInstanceDisks(self, instance)
6956 raise errors.OpExecError("Can't activate the instance's disks")
6958 result = self.rpc.call_instance_start(target_node, instance,
6960 msg = result.fail_msg
6962 _ShutdownInstanceDisks(self, instance)
6963 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6964 (instance.name, target_node, msg))
6967 class LUNodeMigrate(LogicalUnit):
6968 """Migrate all instances from a node.
6971 HPATH = "node-migrate"
6972 HTYPE = constants.HTYPE_NODE
6975 def CheckArguments(self):
6978 def ExpandNames(self):
6979 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6981 self.share_locks = _ShareAll()
6982 self.needed_locks = {
6983 locking.LEVEL_NODE: [self.op.node_name],
6986 def BuildHooksEnv(self):
6989 This runs on the master, the primary and all the secondaries.
6993 "NODE_NAME": self.op.node_name,
6996 def BuildHooksNodes(self):
6997 """Build hooks nodes.
7000 nl = [self.cfg.GetMasterNode()]
7003 def CheckPrereq(self):
7006 def Exec(self, feedback_fn):
7007 # Prepare jobs for migration instances
7009 [opcodes.OpInstanceMigrate(instance_name=inst.name,
7012 iallocator=self.op.iallocator,
7013 target_node=self.op.target_node)]
7014 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name)
7017 # TODO: Run iallocator in this opcode and pass correct placement options to
7018 # OpInstanceMigrate. Since other jobs can modify the cluster between
7019 # running the iallocator and the actual migration, a good consistency model
7020 # will have to be found.
7022 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
7023 frozenset([self.op.node_name]))
7025 return ResultWithJobs(jobs)
7028 class TLMigrateInstance(Tasklet):
7029 """Tasklet class for instance migration.
7032 @ivar live: whether the migration will be done live or non-live;
7033 this variable is initalized only after CheckPrereq has run
7034 @type cleanup: boolean
7035 @ivar cleanup: Wheater we cleanup from a failed migration
7036 @type iallocator: string
7037 @ivar iallocator: The iallocator used to determine target_node
7038 @type target_node: string
7039 @ivar target_node: If given, the target_node to reallocate the instance to
7040 @type failover: boolean
7041 @ivar failover: Whether operation results in failover or migration
7042 @type fallback: boolean
7043 @ivar fallback: Whether fallback to failover is allowed if migration not
7045 @type ignore_consistency: boolean
7046 @ivar ignore_consistency: Wheter we should ignore consistency between source
7048 @type shutdown_timeout: int
7049 @ivar shutdown_timeout: In case of failover timeout of the shutdown
7052 def __init__(self, lu, instance_name, cleanup=False,
7053 failover=False, fallback=False,
7054 ignore_consistency=False,
7055 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT):
7056 """Initializes this class.
7059 Tasklet.__init__(self, lu)
7062 self.instance_name = instance_name
7063 self.cleanup = cleanup
7064 self.live = False # will be overridden later
7065 self.failover = failover
7066 self.fallback = fallback
7067 self.ignore_consistency = ignore_consistency
7068 self.shutdown_timeout = shutdown_timeout
7070 def CheckPrereq(self):
7071 """Check prerequisites.
7073 This checks that the instance is in the cluster.
7076 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
7077 instance = self.cfg.GetInstanceInfo(instance_name)
7078 assert instance is not None
7079 self.instance = instance
7081 if (not self.cleanup and not instance.admin_up and not self.failover and
7083 self.lu.LogInfo("Instance is marked down, fallback allowed, switching"
7085 self.failover = True
7087 if instance.disk_template not in constants.DTS_MIRRORED:
7092 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
7093 " %s" % (instance.disk_template, text),
7096 if instance.disk_template in constants.DTS_EXT_MIRROR:
7097 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
7099 if self.lu.op.iallocator:
7100 self._RunAllocator()
7102 # We set set self.target_node as it is required by
7104 self.target_node = self.lu.op.target_node
7106 # self.target_node is already populated, either directly or by the
7108 target_node = self.target_node
7109 if self.target_node == instance.primary_node:
7110 raise errors.OpPrereqError("Cannot migrate instance %s"
7111 " to its primary (%s)" %
7112 (instance.name, instance.primary_node))
7114 if len(self.lu.tasklets) == 1:
7115 # It is safe to release locks only when we're the only tasklet
7117 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
7118 keep=[instance.primary_node, self.target_node])
7121 secondary_nodes = instance.secondary_nodes
7122 if not secondary_nodes:
7123 raise errors.ConfigurationError("No secondary node but using"
7124 " %s disk template" %
7125 instance.disk_template)
7126 target_node = secondary_nodes[0]
7127 if self.lu.op.iallocator or (self.lu.op.target_node and
7128 self.lu.op.target_node != target_node):
7130 text = "failed over"
7133 raise errors.OpPrereqError("Instances with disk template %s cannot"
7134 " be %s to arbitrary nodes"
7135 " (neither an iallocator nor a target"
7136 " node can be passed)" %
7137 (instance.disk_template, text),
7140 i_be = self.cfg.GetClusterInfo().FillBE(instance)
7142 # check memory requirements on the secondary node
7143 if not self.cleanup and (not self.failover or instance.admin_up):
7144 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
7145 instance.name, i_be[constants.BE_MEMORY],
7146 instance.hypervisor)
7148 self.lu.LogInfo("Not checking memory on the secondary node as"
7149 " instance will not be started")
7151 # check bridge existance
7152 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
7154 if not self.cleanup:
7155 _CheckNodeNotDrained(self.lu, target_node)
7156 if not self.failover:
7157 result = self.rpc.call_instance_migratable(instance.primary_node,
7159 if result.fail_msg and self.fallback:
7160 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
7162 self.failover = True
7164 result.Raise("Can't migrate, please use failover",
7165 prereq=True, ecode=errors.ECODE_STATE)
7167 assert not (self.failover and self.cleanup)
7169 if not self.failover:
7170 if self.lu.op.live is not None and self.lu.op.mode is not None:
7171 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
7172 " parameters are accepted",
7174 if self.lu.op.live is not None:
7176 self.lu.op.mode = constants.HT_MIGRATION_LIVE
7178 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
7179 # reset the 'live' parameter to None so that repeated
7180 # invocations of CheckPrereq do not raise an exception
7181 self.lu.op.live = None
7182 elif self.lu.op.mode is None:
7183 # read the default value from the hypervisor
7184 i_hv = self.cfg.GetClusterInfo().FillHV(self.instance,
7186 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
7188 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
7190 # Failover is never live
7193 def _RunAllocator(self):
7194 """Run the allocator based on input opcode.
7197 ial = IAllocator(self.cfg, self.rpc,
7198 mode=constants.IALLOCATOR_MODE_RELOC,
7199 name=self.instance_name,
7200 # TODO See why hail breaks with a single node below
7201 relocate_from=[self.instance.primary_node,
7202 self.instance.primary_node],
7205 ial.Run(self.lu.op.iallocator)
7208 raise errors.OpPrereqError("Can't compute nodes using"
7209 " iallocator '%s': %s" %
7210 (self.lu.op.iallocator, ial.info),
7212 if len(ial.result) != ial.required_nodes:
7213 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7214 " of nodes (%s), required %s" %
7215 (self.lu.op.iallocator, len(ial.result),
7216 ial.required_nodes), errors.ECODE_FAULT)
7217 self.target_node = ial.result[0]
7218 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7219 self.instance_name, self.lu.op.iallocator,
7220 utils.CommaJoin(ial.result))
7222 def _WaitUntilSync(self):
7223 """Poll with custom rpc for disk sync.
7225 This uses our own step-based rpc call.
7228 self.feedback_fn("* wait until resync is done")
7232 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
7234 self.instance.disks)
7236 for node, nres in result.items():
7237 nres.Raise("Cannot resync disks on node %s" % node)
7238 node_done, node_percent = nres.payload
7239 all_done = all_done and node_done
7240 if node_percent is not None:
7241 min_percent = min(min_percent, node_percent)
7243 if min_percent < 100:
7244 self.feedback_fn(" - progress: %.1f%%" % min_percent)
7247 def _EnsureSecondary(self, node):
7248 """Demote a node to secondary.
7251 self.feedback_fn("* switching node %s to secondary mode" % node)
7253 for dev in self.instance.disks:
7254 self.cfg.SetDiskID(dev, node)
7256 result = self.rpc.call_blockdev_close(node, self.instance.name,
7257 self.instance.disks)
7258 result.Raise("Cannot change disk to secondary on node %s" % node)
7260 def _GoStandalone(self):
7261 """Disconnect from the network.
7264 self.feedback_fn("* changing into standalone mode")
7265 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
7266 self.instance.disks)
7267 for node, nres in result.items():
7268 nres.Raise("Cannot disconnect disks node %s" % node)
7270 def _GoReconnect(self, multimaster):
7271 """Reconnect to the network.
7277 msg = "single-master"
7278 self.feedback_fn("* changing disks into %s mode" % msg)
7279 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
7280 self.instance.disks,
7281 self.instance.name, multimaster)
7282 for node, nres in result.items():
7283 nres.Raise("Cannot change disks config on node %s" % node)
7285 def _ExecCleanup(self):
7286 """Try to cleanup after a failed migration.
7288 The cleanup is done by:
7289 - check that the instance is running only on one node
7290 (and update the config if needed)
7291 - change disks on its secondary node to secondary
7292 - wait until disks are fully synchronized
7293 - disconnect from the network
7294 - change disks into single-master mode
7295 - wait again until disks are fully synchronized
7298 instance = self.instance
7299 target_node = self.target_node
7300 source_node = self.source_node
7302 # check running on only one node
7303 self.feedback_fn("* checking where the instance actually runs"
7304 " (if this hangs, the hypervisor might be in"
7306 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
7307 for node, result in ins_l.items():
7308 result.Raise("Can't contact node %s" % node)
7310 runningon_source = instance.name in ins_l[source_node].payload
7311 runningon_target = instance.name in ins_l[target_node].payload
7313 if runningon_source and runningon_target:
7314 raise errors.OpExecError("Instance seems to be running on two nodes,"
7315 " or the hypervisor is confused; you will have"
7316 " to ensure manually that it runs only on one"
7317 " and restart this operation")
7319 if not (runningon_source or runningon_target):
7320 raise errors.OpExecError("Instance does not seem to be running at all;"
7321 " in this case it's safer to repair by"
7322 " running 'gnt-instance stop' to ensure disk"
7323 " shutdown, and then restarting it")
7325 if runningon_target:
7326 # the migration has actually succeeded, we need to update the config
7327 self.feedback_fn("* instance running on secondary node (%s),"
7328 " updating config" % target_node)
7329 instance.primary_node = target_node
7330 self.cfg.Update(instance, self.feedback_fn)
7331 demoted_node = source_node
7333 self.feedback_fn("* instance confirmed to be running on its"
7334 " primary node (%s)" % source_node)
7335 demoted_node = target_node
7337 if instance.disk_template in constants.DTS_INT_MIRROR:
7338 self._EnsureSecondary(demoted_node)
7340 self._WaitUntilSync()
7341 except errors.OpExecError:
7342 # we ignore here errors, since if the device is standalone, it
7343 # won't be able to sync
7345 self._GoStandalone()
7346 self._GoReconnect(False)
7347 self._WaitUntilSync()
7349 self.feedback_fn("* done")
7351 def _RevertDiskStatus(self):
7352 """Try to revert the disk status after a failed migration.
7355 target_node = self.target_node
7356 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
7360 self._EnsureSecondary(target_node)
7361 self._GoStandalone()
7362 self._GoReconnect(False)
7363 self._WaitUntilSync()
7364 except errors.OpExecError, err:
7365 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
7366 " please try to recover the instance manually;"
7367 " error '%s'" % str(err))
7369 def _AbortMigration(self):
7370 """Call the hypervisor code to abort a started migration.
7373 instance = self.instance
7374 target_node = self.target_node
7375 migration_info = self.migration_info
7377 abort_result = self.rpc.call_finalize_migration(target_node,
7381 abort_msg = abort_result.fail_msg
7383 logging.error("Aborting migration failed on target node %s: %s",
7384 target_node, abort_msg)
7385 # Don't raise an exception here, as we stil have to try to revert the
7386 # disk status, even if this step failed.
7388 def _ExecMigration(self):
7389 """Migrate an instance.
7391 The migrate is done by:
7392 - change the disks into dual-master mode
7393 - wait until disks are fully synchronized again
7394 - migrate the instance
7395 - change disks on the new secondary node (the old primary) to secondary
7396 - wait until disks are fully synchronized
7397 - change disks into single-master mode
7400 instance = self.instance
7401 target_node = self.target_node
7402 source_node = self.source_node
7404 self.feedback_fn("* checking disk consistency between source and target")
7405 for dev in instance.disks:
7406 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
7407 raise errors.OpExecError("Disk %s is degraded or not fully"
7408 " synchronized on target node,"
7409 " aborting migration" % dev.iv_name)
7411 # First get the migration information from the remote node
7412 result = self.rpc.call_migration_info(source_node, instance)
7413 msg = result.fail_msg
7415 log_err = ("Failed fetching source migration information from %s: %s" %
7417 logging.error(log_err)
7418 raise errors.OpExecError(log_err)
7420 self.migration_info = migration_info = result.payload
7422 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
7423 # Then switch the disks to master/master mode
7424 self._EnsureSecondary(target_node)
7425 self._GoStandalone()
7426 self._GoReconnect(True)
7427 self._WaitUntilSync()
7429 self.feedback_fn("* preparing %s to accept the instance" % target_node)
7430 result = self.rpc.call_accept_instance(target_node,
7433 self.nodes_ip[target_node])
7435 msg = result.fail_msg
7437 logging.error("Instance pre-migration failed, trying to revert"
7438 " disk status: %s", msg)
7439 self.feedback_fn("Pre-migration failed, aborting")
7440 self._AbortMigration()
7441 self._RevertDiskStatus()
7442 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
7443 (instance.name, msg))
7445 self.feedback_fn("* migrating instance to %s" % target_node)
7446 result = self.rpc.call_instance_migrate(source_node, instance,
7447 self.nodes_ip[target_node],
7449 msg = result.fail_msg
7451 logging.error("Instance migration failed, trying to revert"
7452 " disk status: %s", msg)
7453 self.feedback_fn("Migration failed, aborting")
7454 self._AbortMigration()
7455 self._RevertDiskStatus()
7456 raise errors.OpExecError("Could not migrate instance %s: %s" %
7457 (instance.name, msg))
7459 instance.primary_node = target_node
7460 # distribute new instance config to the other nodes
7461 self.cfg.Update(instance, self.feedback_fn)
7463 result = self.rpc.call_finalize_migration(target_node,
7467 msg = result.fail_msg
7469 logging.error("Instance migration succeeded, but finalization failed:"
7471 raise errors.OpExecError("Could not finalize instance migration: %s" %
7474 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
7475 self._EnsureSecondary(source_node)
7476 self._WaitUntilSync()
7477 self._GoStandalone()
7478 self._GoReconnect(False)
7479 self._WaitUntilSync()
7481 self.feedback_fn("* done")
7483 def _ExecFailover(self):
7484 """Failover an instance.
7486 The failover is done by shutting it down on its present node and
7487 starting it on the secondary.
7490 instance = self.instance
7491 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
7493 source_node = instance.primary_node
7494 target_node = self.target_node
7496 if instance.admin_up:
7497 self.feedback_fn("* checking disk consistency between source and target")
7498 for dev in instance.disks:
7499 # for drbd, these are drbd over lvm
7500 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
7501 if primary_node.offline:
7502 self.feedback_fn("Node %s is offline, ignoring degraded disk %s on"
7504 (primary_node.name, dev.iv_name, target_node))
7505 elif not self.ignore_consistency:
7506 raise errors.OpExecError("Disk %s is degraded on target node,"
7507 " aborting failover" % dev.iv_name)
7509 self.feedback_fn("* not checking disk consistency as instance is not"
7512 self.feedback_fn("* shutting down instance on source node")
7513 logging.info("Shutting down instance %s on node %s",
7514 instance.name, source_node)
7516 result = self.rpc.call_instance_shutdown(source_node, instance,
7517 self.shutdown_timeout)
7518 msg = result.fail_msg
7520 if self.ignore_consistency or primary_node.offline:
7521 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
7522 " proceeding anyway; please make sure node"
7523 " %s is down; error details: %s",
7524 instance.name, source_node, source_node, msg)
7526 raise errors.OpExecError("Could not shutdown instance %s on"
7528 (instance.name, source_node, msg))
7530 self.feedback_fn("* deactivating the instance's disks on source node")
7531 if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True):
7532 raise errors.OpExecError("Can't shut down the instance's disks")
7534 instance.primary_node = target_node
7535 # distribute new instance config to the other nodes
7536 self.cfg.Update(instance, self.feedback_fn)
7538 # Only start the instance if it's marked as up
7539 if instance.admin_up:
7540 self.feedback_fn("* activating the instance's disks on target node %s" %
7542 logging.info("Starting instance %s on node %s",
7543 instance.name, target_node)
7545 disks_ok, _ = _AssembleInstanceDisks(self.lu, instance,
7546 ignore_secondaries=True)
7548 _ShutdownInstanceDisks(self.lu, instance)
7549 raise errors.OpExecError("Can't activate the instance's disks")
7551 self.feedback_fn("* starting the instance on the target node %s" %
7553 result = self.rpc.call_instance_start(target_node, instance, None, None,
7555 msg = result.fail_msg
7557 _ShutdownInstanceDisks(self.lu, instance)
7558 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
7559 (instance.name, target_node, msg))
7561 def Exec(self, feedback_fn):
7562 """Perform the migration.
7565 self.feedback_fn = feedback_fn
7566 self.source_node = self.instance.primary_node
7568 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
7569 if self.instance.disk_template in constants.DTS_INT_MIRROR:
7570 self.target_node = self.instance.secondary_nodes[0]
7571 # Otherwise self.target_node has been populated either
7572 # directly, or through an iallocator.
7574 self.all_nodes = [self.source_node, self.target_node]
7575 self.nodes_ip = dict((name, node.secondary_ip) for (name, node)
7576 in self.cfg.GetMultiNodeInfo(self.all_nodes))
7579 feedback_fn("Failover instance %s" % self.instance.name)
7580 self._ExecFailover()
7582 feedback_fn("Migrating instance %s" % self.instance.name)
7585 return self._ExecCleanup()
7587 return self._ExecMigration()
7590 def _CreateBlockDev(lu, node, instance, device, force_create,
7592 """Create a tree of block devices on a given node.
7594 If this device type has to be created on secondaries, create it and
7597 If not, just recurse to children keeping the same 'force' value.
7599 @param lu: the lu on whose behalf we execute
7600 @param node: the node on which to create the device
7601 @type instance: L{objects.Instance}
7602 @param instance: the instance which owns the device
7603 @type device: L{objects.Disk}
7604 @param device: the device to create
7605 @type force_create: boolean
7606 @param force_create: whether to force creation of this device; this
7607 will be change to True whenever we find a device which has
7608 CreateOnSecondary() attribute
7609 @param info: the extra 'metadata' we should attach to the device
7610 (this will be represented as a LVM tag)
7611 @type force_open: boolean
7612 @param force_open: this parameter will be passes to the
7613 L{backend.BlockdevCreate} function where it specifies
7614 whether we run on primary or not, and it affects both
7615 the child assembly and the device own Open() execution
7618 if device.CreateOnSecondary():
7622 for child in device.children:
7623 _CreateBlockDev(lu, node, instance, child, force_create,
7626 if not force_create:
7629 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
7632 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
7633 """Create a single block device on a given node.
7635 This will not recurse over children of the device, so they must be
7638 @param lu: the lu on whose behalf we execute
7639 @param node: the node on which to create the device
7640 @type instance: L{objects.Instance}
7641 @param instance: the instance which owns the device
7642 @type device: L{objects.Disk}
7643 @param device: the device to create
7644 @param info: the extra 'metadata' we should attach to the device
7645 (this will be represented as a LVM tag)
7646 @type force_open: boolean
7647 @param force_open: this parameter will be passes to the
7648 L{backend.BlockdevCreate} function where it specifies
7649 whether we run on primary or not, and it affects both
7650 the child assembly and the device own Open() execution
7653 lu.cfg.SetDiskID(device, node)
7654 result = lu.rpc.call_blockdev_create(node, device, device.size,
7655 instance.name, force_open, info)
7656 result.Raise("Can't create block device %s on"
7657 " node %s for instance %s" % (device, node, instance.name))
7658 if device.physical_id is None:
7659 device.physical_id = result.payload
7662 def _GenerateUniqueNames(lu, exts):
7663 """Generate a suitable LV name.
7665 This will generate a logical volume name for the given instance.
7670 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
7671 results.append("%s%s" % (new_id, val))
7675 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
7676 iv_name, p_minor, s_minor):
7677 """Generate a drbd8 device complete with its children.
7680 assert len(vgnames) == len(names) == 2
7681 port = lu.cfg.AllocatePort()
7682 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
7683 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
7684 logical_id=(vgnames[0], names[0]))
7685 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7686 logical_id=(vgnames[1], names[1]))
7687 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
7688 logical_id=(primary, secondary, port,
7691 children=[dev_data, dev_meta],
7696 def _GenerateDiskTemplate(lu, template_name,
7697 instance_name, primary_node,
7698 secondary_nodes, disk_info,
7699 file_storage_dir, file_driver,
7700 base_index, feedback_fn):
7701 """Generate the entire disk layout for a given template type.
7704 #TODO: compute space requirements
7706 vgname = lu.cfg.GetVGName()
7707 disk_count = len(disk_info)
7709 if template_name == constants.DT_DISKLESS:
7711 elif template_name == constants.DT_PLAIN:
7712 if len(secondary_nodes) != 0:
7713 raise errors.ProgrammerError("Wrong template configuration")
7715 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7716 for i in range(disk_count)])
7717 for idx, disk in enumerate(disk_info):
7718 disk_index = idx + base_index
7719 vg = disk.get(constants.IDISK_VG, vgname)
7720 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
7721 disk_dev = objects.Disk(dev_type=constants.LD_LV,
7722 size=disk[constants.IDISK_SIZE],
7723 logical_id=(vg, names[idx]),
7724 iv_name="disk/%d" % disk_index,
7725 mode=disk[constants.IDISK_MODE])
7726 disks.append(disk_dev)
7727 elif template_name == constants.DT_DRBD8:
7728 if len(secondary_nodes) != 1:
7729 raise errors.ProgrammerError("Wrong template configuration")
7730 remote_node = secondary_nodes[0]
7731 minors = lu.cfg.AllocateDRBDMinor(
7732 [primary_node, remote_node] * len(disk_info), instance_name)
7735 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7736 for i in range(disk_count)]):
7737 names.append(lv_prefix + "_data")
7738 names.append(lv_prefix + "_meta")
7739 for idx, disk in enumerate(disk_info):
7740 disk_index = idx + base_index
7741 data_vg = disk.get(constants.IDISK_VG, vgname)
7742 meta_vg = disk.get(constants.IDISK_METAVG, data_vg)
7743 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
7744 disk[constants.IDISK_SIZE],
7746 names[idx * 2:idx * 2 + 2],
7747 "disk/%d" % disk_index,
7748 minors[idx * 2], minors[idx * 2 + 1])
7749 disk_dev.mode = disk[constants.IDISK_MODE]
7750 disks.append(disk_dev)
7751 elif template_name == constants.DT_FILE:
7752 if len(secondary_nodes) != 0:
7753 raise errors.ProgrammerError("Wrong template configuration")
7755 opcodes.RequireFileStorage()
7757 for idx, disk in enumerate(disk_info):
7758 disk_index = idx + base_index
7759 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7760 size=disk[constants.IDISK_SIZE],
7761 iv_name="disk/%d" % disk_index,
7762 logical_id=(file_driver,
7763 "%s/disk%d" % (file_storage_dir,
7765 mode=disk[constants.IDISK_MODE])
7766 disks.append(disk_dev)
7767 elif template_name == constants.DT_SHARED_FILE:
7768 if len(secondary_nodes) != 0:
7769 raise errors.ProgrammerError("Wrong template configuration")
7771 opcodes.RequireSharedFileStorage()
7773 for idx, disk in enumerate(disk_info):
7774 disk_index = idx + base_index
7775 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7776 size=disk[constants.IDISK_SIZE],
7777 iv_name="disk/%d" % disk_index,
7778 logical_id=(file_driver,
7779 "%s/disk%d" % (file_storage_dir,
7781 mode=disk[constants.IDISK_MODE])
7782 disks.append(disk_dev)
7783 elif template_name == constants.DT_BLOCK:
7784 if len(secondary_nodes) != 0:
7785 raise errors.ProgrammerError("Wrong template configuration")
7787 for idx, disk in enumerate(disk_info):
7788 disk_index = idx + base_index
7789 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV,
7790 size=disk[constants.IDISK_SIZE],
7791 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL,
7792 disk[constants.IDISK_ADOPT]),
7793 iv_name="disk/%d" % disk_index,
7794 mode=disk[constants.IDISK_MODE])
7795 disks.append(disk_dev)
7798 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
7802 def _GetInstanceInfoText(instance):
7803 """Compute that text that should be added to the disk's metadata.
7806 return "originstname+%s" % instance.name
7809 def _CalcEta(time_taken, written, total_size):
7810 """Calculates the ETA based on size written and total size.
7812 @param time_taken: The time taken so far
7813 @param written: amount written so far
7814 @param total_size: The total size of data to be written
7815 @return: The remaining time in seconds
7818 avg_time = time_taken / float(written)
7819 return (total_size - written) * avg_time
7822 def _WipeDisks(lu, instance):
7823 """Wipes instance disks.
7825 @type lu: L{LogicalUnit}
7826 @param lu: the logical unit on whose behalf we execute
7827 @type instance: L{objects.Instance}
7828 @param instance: the instance whose disks we should create
7829 @return: the success of the wipe
7832 node = instance.primary_node
7834 for device in instance.disks:
7835 lu.cfg.SetDiskID(device, node)
7837 logging.info("Pause sync of instance %s disks", instance.name)
7838 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
7840 for idx, success in enumerate(result.payload):
7842 logging.warn("pause-sync of instance %s for disks %d failed",
7846 for idx, device in enumerate(instance.disks):
7847 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
7848 # MAX_WIPE_CHUNK at max
7849 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
7850 constants.MIN_WIPE_CHUNK_PERCENT)
7851 # we _must_ make this an int, otherwise rounding errors will
7853 wipe_chunk_size = int(wipe_chunk_size)
7855 lu.LogInfo("* Wiping disk %d", idx)
7856 logging.info("Wiping disk %d for instance %s, node %s using"
7857 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
7862 start_time = time.time()
7864 while offset < size:
7865 wipe_size = min(wipe_chunk_size, size - offset)
7866 logging.debug("Wiping disk %d, offset %s, chunk %s",
7867 idx, offset, wipe_size)
7868 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
7869 result.Raise("Could not wipe disk %d at offset %d for size %d" %
7870 (idx, offset, wipe_size))
7873 if now - last_output >= 60:
7874 eta = _CalcEta(now - start_time, offset, size)
7875 lu.LogInfo(" - done: %.1f%% ETA: %s" %
7876 (offset / float(size) * 100, utils.FormatSeconds(eta)))
7879 logging.info("Resume sync of instance %s disks", instance.name)
7881 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
7883 for idx, success in enumerate(result.payload):
7885 lu.LogWarning("Resume sync of disk %d failed, please have a"
7886 " look at the status and troubleshoot the issue", idx)
7887 logging.warn("resume-sync of instance %s for disks %d failed",
7891 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
7892 """Create all disks for an instance.
7894 This abstracts away some work from AddInstance.
7896 @type lu: L{LogicalUnit}
7897 @param lu: the logical unit on whose behalf we execute
7898 @type instance: L{objects.Instance}
7899 @param instance: the instance whose disks we should create
7901 @param to_skip: list of indices to skip
7902 @type target_node: string
7903 @param target_node: if passed, overrides the target node for creation
7905 @return: the success of the creation
7908 info = _GetInstanceInfoText(instance)
7909 if target_node is None:
7910 pnode = instance.primary_node
7911 all_nodes = instance.all_nodes
7916 if instance.disk_template in constants.DTS_FILEBASED:
7917 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7918 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
7920 result.Raise("Failed to create directory '%s' on"
7921 " node %s" % (file_storage_dir, pnode))
7923 # Note: this needs to be kept in sync with adding of disks in
7924 # LUInstanceSetParams
7925 for idx, device in enumerate(instance.disks):
7926 if to_skip and idx in to_skip:
7928 logging.info("Creating volume %s for instance %s",
7929 device.iv_name, instance.name)
7931 for node in all_nodes:
7932 f_create = node == pnode
7933 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
7936 def _RemoveDisks(lu, instance, target_node=None):
7937 """Remove all disks for an instance.
7939 This abstracts away some work from `AddInstance()` and
7940 `RemoveInstance()`. Note that in case some of the devices couldn't
7941 be removed, the removal will continue with the other ones (compare
7942 with `_CreateDisks()`).
7944 @type lu: L{LogicalUnit}
7945 @param lu: the logical unit on whose behalf we execute
7946 @type instance: L{objects.Instance}
7947 @param instance: the instance whose disks we should remove
7948 @type target_node: string
7949 @param target_node: used to override the node on which to remove the disks
7951 @return: the success of the removal
7954 logging.info("Removing block devices for instance %s", instance.name)
7957 for device in instance.disks:
7959 edata = [(target_node, device)]
7961 edata = device.ComputeNodeTree(instance.primary_node)
7962 for node, disk in edata:
7963 lu.cfg.SetDiskID(disk, node)
7964 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
7966 lu.LogWarning("Could not remove block device %s on node %s,"
7967 " continuing anyway: %s", device.iv_name, node, msg)
7970 # if this is a DRBD disk, return its port to the pool
7971 if device.dev_type in constants.LDS_DRBD:
7972 tcp_port = device.logical_id[2]
7973 lu.cfg.AddTcpUdpPort(tcp_port)
7975 if instance.disk_template == constants.DT_FILE:
7976 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7980 tgt = instance.primary_node
7981 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
7983 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
7984 file_storage_dir, instance.primary_node, result.fail_msg)
7990 def _ComputeDiskSizePerVG(disk_template, disks):
7991 """Compute disk size requirements in the volume group
7994 def _compute(disks, payload):
7995 """Universal algorithm.
8000 vgs[disk[constants.IDISK_VG]] = \
8001 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
8005 # Required free disk space as a function of disk and swap space
8007 constants.DT_DISKLESS: {},
8008 constants.DT_PLAIN: _compute(disks, 0),
8009 # 128 MB are added for drbd metadata for each disk
8010 constants.DT_DRBD8: _compute(disks, 128),
8011 constants.DT_FILE: {},
8012 constants.DT_SHARED_FILE: {},
8015 if disk_template not in req_size_dict:
8016 raise errors.ProgrammerError("Disk template '%s' size requirement"
8017 " is unknown" % disk_template)
8019 return req_size_dict[disk_template]
8022 def _ComputeDiskSize(disk_template, disks):
8023 """Compute disk size requirements in the volume group
8026 # Required free disk space as a function of disk and swap space
8028 constants.DT_DISKLESS: None,
8029 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
8030 # 128 MB are added for drbd metadata for each disk
8031 constants.DT_DRBD8: sum(d[constants.IDISK_SIZE] + 128 for d in disks),
8032 constants.DT_FILE: None,
8033 constants.DT_SHARED_FILE: 0,
8034 constants.DT_BLOCK: 0,
8037 if disk_template not in req_size_dict:
8038 raise errors.ProgrammerError("Disk template '%s' size requirement"
8039 " is unknown" % disk_template)
8041 return req_size_dict[disk_template]
8044 def _FilterVmNodes(lu, nodenames):
8045 """Filters out non-vm_capable nodes from a list.
8047 @type lu: L{LogicalUnit}
8048 @param lu: the logical unit for which we check
8049 @type nodenames: list
8050 @param nodenames: the list of nodes on which we should check
8052 @return: the list of vm-capable nodes
8055 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
8056 return [name for name in nodenames if name not in vm_nodes]
8059 def _CheckHVParams(lu, nodenames, hvname, hvparams):
8060 """Hypervisor parameter validation.
8062 This function abstract the hypervisor parameter validation to be
8063 used in both instance create and instance modify.
8065 @type lu: L{LogicalUnit}
8066 @param lu: the logical unit for which we check
8067 @type nodenames: list
8068 @param nodenames: the list of nodes on which we should check
8069 @type hvname: string
8070 @param hvname: the name of the hypervisor we should use
8071 @type hvparams: dict
8072 @param hvparams: the parameters which we need to check
8073 @raise errors.OpPrereqError: if the parameters are not valid
8076 nodenames = _FilterVmNodes(lu, nodenames)
8077 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
8080 for node in nodenames:
8084 info.Raise("Hypervisor parameter validation failed on node %s" % node)
8087 def _CheckOSParams(lu, required, nodenames, osname, osparams):
8088 """OS parameters validation.
8090 @type lu: L{LogicalUnit}
8091 @param lu: the logical unit for which we check
8092 @type required: boolean
8093 @param required: whether the validation should fail if the OS is not
8095 @type nodenames: list
8096 @param nodenames: the list of nodes on which we should check
8097 @type osname: string
8098 @param osname: the name of the hypervisor we should use
8099 @type osparams: dict
8100 @param osparams: the parameters which we need to check
8101 @raise errors.OpPrereqError: if the parameters are not valid
8104 nodenames = _FilterVmNodes(lu, nodenames)
8105 result = lu.rpc.call_os_validate(required, nodenames, osname,
8106 [constants.OS_VALIDATE_PARAMETERS],
8108 for node, nres in result.items():
8109 # we don't check for offline cases since this should be run only
8110 # against the master node and/or an instance's nodes
8111 nres.Raise("OS Parameters validation failed on node %s" % node)
8112 if not nres.payload:
8113 lu.LogInfo("OS %s not found on node %s, validation skipped",
8117 class LUInstanceCreate(LogicalUnit):
8118 """Create an instance.
8121 HPATH = "instance-add"
8122 HTYPE = constants.HTYPE_INSTANCE
8125 def CheckArguments(self):
8129 # do not require name_check to ease forward/backward compatibility
8131 if self.op.no_install and self.op.start:
8132 self.LogInfo("No-installation mode selected, disabling startup")
8133 self.op.start = False
8134 # validate/normalize the instance name
8135 self.op.instance_name = \
8136 netutils.Hostname.GetNormalizedName(self.op.instance_name)
8138 if self.op.ip_check and not self.op.name_check:
8139 # TODO: make the ip check more flexible and not depend on the name check
8140 raise errors.OpPrereqError("Cannot do IP address check without a name"
8141 " check", errors.ECODE_INVAL)
8143 # check nics' parameter names
8144 for nic in self.op.nics:
8145 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
8147 # check disks. parameter names and consistent adopt/no-adopt strategy
8148 has_adopt = has_no_adopt = False
8149 for disk in self.op.disks:
8150 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
8151 if constants.IDISK_ADOPT in disk:
8155 if has_adopt and has_no_adopt:
8156 raise errors.OpPrereqError("Either all disks are adopted or none is",
8159 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
8160 raise errors.OpPrereqError("Disk adoption is not supported for the"
8161 " '%s' disk template" %
8162 self.op.disk_template,
8164 if self.op.iallocator is not None:
8165 raise errors.OpPrereqError("Disk adoption not allowed with an"
8166 " iallocator script", errors.ECODE_INVAL)
8167 if self.op.mode == constants.INSTANCE_IMPORT:
8168 raise errors.OpPrereqError("Disk adoption not allowed for"
8169 " instance import", errors.ECODE_INVAL)
8171 if self.op.disk_template in constants.DTS_MUST_ADOPT:
8172 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
8173 " but no 'adopt' parameter given" %
8174 self.op.disk_template,
8177 self.adopt_disks = has_adopt
8179 # instance name verification
8180 if self.op.name_check:
8181 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
8182 self.op.instance_name = self.hostname1.name
8183 # used in CheckPrereq for ip ping check
8184 self.check_ip = self.hostname1.ip
8186 self.check_ip = None
8188 # file storage checks
8189 if (self.op.file_driver and
8190 not self.op.file_driver in constants.FILE_DRIVER):
8191 raise errors.OpPrereqError("Invalid file driver name '%s'" %
8192 self.op.file_driver, errors.ECODE_INVAL)
8194 if self.op.disk_template == constants.DT_FILE:
8195 opcodes.RequireFileStorage()
8196 elif self.op.disk_template == constants.DT_SHARED_FILE:
8197 opcodes.RequireSharedFileStorage()
8199 ### Node/iallocator related checks
8200 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
8202 if self.op.pnode is not None:
8203 if self.op.disk_template in constants.DTS_INT_MIRROR:
8204 if self.op.snode is None:
8205 raise errors.OpPrereqError("The networked disk templates need"
8206 " a mirror node", errors.ECODE_INVAL)
8208 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
8210 self.op.snode = None
8212 self._cds = _GetClusterDomainSecret()
8214 if self.op.mode == constants.INSTANCE_IMPORT:
8215 # On import force_variant must be True, because if we forced it at
8216 # initial install, our only chance when importing it back is that it
8218 self.op.force_variant = True
8220 if self.op.no_install:
8221 self.LogInfo("No-installation mode has no effect during import")
8223 elif self.op.mode == constants.INSTANCE_CREATE:
8224 if self.op.os_type is None:
8225 raise errors.OpPrereqError("No guest OS specified",
8227 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
8228 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
8229 " installation" % self.op.os_type,
8231 if self.op.disk_template is None:
8232 raise errors.OpPrereqError("No disk template specified",
8235 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
8236 # Check handshake to ensure both clusters have the same domain secret
8237 src_handshake = self.op.source_handshake
8238 if not src_handshake:
8239 raise errors.OpPrereqError("Missing source handshake",
8242 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
8245 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
8248 # Load and check source CA
8249 self.source_x509_ca_pem = self.op.source_x509_ca
8250 if not self.source_x509_ca_pem:
8251 raise errors.OpPrereqError("Missing source X509 CA",
8255 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
8257 except OpenSSL.crypto.Error, err:
8258 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
8259 (err, ), errors.ECODE_INVAL)
8261 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
8262 if errcode is not None:
8263 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
8266 self.source_x509_ca = cert
8268 src_instance_name = self.op.source_instance_name
8269 if not src_instance_name:
8270 raise errors.OpPrereqError("Missing source instance name",
8273 self.source_instance_name = \
8274 netutils.GetHostname(name=src_instance_name).name
8277 raise errors.OpPrereqError("Invalid instance creation mode %r" %
8278 self.op.mode, errors.ECODE_INVAL)
8280 def ExpandNames(self):
8281 """ExpandNames for CreateInstance.
8283 Figure out the right locks for instance creation.
8286 self.needed_locks = {}
8288 instance_name = self.op.instance_name
8289 # this is just a preventive check, but someone might still add this
8290 # instance in the meantime, and creation will fail at lock-add time
8291 if instance_name in self.cfg.GetInstanceList():
8292 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
8293 instance_name, errors.ECODE_EXISTS)
8295 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
8297 if self.op.iallocator:
8298 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8300 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
8301 nodelist = [self.op.pnode]
8302 if self.op.snode is not None:
8303 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
8304 nodelist.append(self.op.snode)
8305 self.needed_locks[locking.LEVEL_NODE] = nodelist
8307 # in case of import lock the source node too
8308 if self.op.mode == constants.INSTANCE_IMPORT:
8309 src_node = self.op.src_node
8310 src_path = self.op.src_path
8312 if src_path is None:
8313 self.op.src_path = src_path = self.op.instance_name
8315 if src_node is None:
8316 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8317 self.op.src_node = None
8318 if os.path.isabs(src_path):
8319 raise errors.OpPrereqError("Importing an instance from a path"
8320 " requires a source node option",
8323 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
8324 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
8325 self.needed_locks[locking.LEVEL_NODE].append(src_node)
8326 if not os.path.isabs(src_path):
8327 self.op.src_path = src_path = \
8328 utils.PathJoin(constants.EXPORT_DIR, src_path)
8330 def _RunAllocator(self):
8331 """Run the allocator based on input opcode.
8334 nics = [n.ToDict() for n in self.nics]
8335 ial = IAllocator(self.cfg, self.rpc,
8336 mode=constants.IALLOCATOR_MODE_ALLOC,
8337 name=self.op.instance_name,
8338 disk_template=self.op.disk_template,
8341 vcpus=self.be_full[constants.BE_VCPUS],
8342 memory=self.be_full[constants.BE_MEMORY],
8345 hypervisor=self.op.hypervisor,
8348 ial.Run(self.op.iallocator)
8351 raise errors.OpPrereqError("Can't compute nodes using"
8352 " iallocator '%s': %s" %
8353 (self.op.iallocator, ial.info),
8355 if len(ial.result) != ial.required_nodes:
8356 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8357 " of nodes (%s), required %s" %
8358 (self.op.iallocator, len(ial.result),
8359 ial.required_nodes), errors.ECODE_FAULT)
8360 self.op.pnode = ial.result[0]
8361 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
8362 self.op.instance_name, self.op.iallocator,
8363 utils.CommaJoin(ial.result))
8364 if ial.required_nodes == 2:
8365 self.op.snode = ial.result[1]
8367 def BuildHooksEnv(self):
8370 This runs on master, primary and secondary nodes of the instance.
8374 "ADD_MODE": self.op.mode,
8376 if self.op.mode == constants.INSTANCE_IMPORT:
8377 env["SRC_NODE"] = self.op.src_node
8378 env["SRC_PATH"] = self.op.src_path
8379 env["SRC_IMAGES"] = self.src_images
8381 env.update(_BuildInstanceHookEnv(
8382 name=self.op.instance_name,
8383 primary_node=self.op.pnode,
8384 secondary_nodes=self.secondaries,
8385 status=self.op.start,
8386 os_type=self.op.os_type,
8387 memory=self.be_full[constants.BE_MEMORY],
8388 vcpus=self.be_full[constants.BE_VCPUS],
8389 nics=_NICListToTuple(self, self.nics),
8390 disk_template=self.op.disk_template,
8391 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
8392 for d in self.disks],
8395 hypervisor_name=self.op.hypervisor,
8401 def BuildHooksNodes(self):
8402 """Build hooks nodes.
8405 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
8408 def _ReadExportInfo(self):
8409 """Reads the export information from disk.
8411 It will override the opcode source node and path with the actual
8412 information, if these two were not specified before.
8414 @return: the export information
8417 assert self.op.mode == constants.INSTANCE_IMPORT
8419 src_node = self.op.src_node
8420 src_path = self.op.src_path
8422 if src_node is None:
8423 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
8424 exp_list = self.rpc.call_export_list(locked_nodes)
8426 for node in exp_list:
8427 if exp_list[node].fail_msg:
8429 if src_path in exp_list[node].payload:
8431 self.op.src_node = src_node = node
8432 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
8436 raise errors.OpPrereqError("No export found for relative path %s" %
8437 src_path, errors.ECODE_INVAL)
8439 _CheckNodeOnline(self, src_node)
8440 result = self.rpc.call_export_info(src_node, src_path)
8441 result.Raise("No export or invalid export found in dir %s" % src_path)
8443 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
8444 if not export_info.has_section(constants.INISECT_EXP):
8445 raise errors.ProgrammerError("Corrupted export config",
8446 errors.ECODE_ENVIRON)
8448 ei_version = export_info.get(constants.INISECT_EXP, "version")
8449 if (int(ei_version) != constants.EXPORT_VERSION):
8450 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
8451 (ei_version, constants.EXPORT_VERSION),
8452 errors.ECODE_ENVIRON)
8455 def _ReadExportParams(self, einfo):
8456 """Use export parameters as defaults.
8458 In case the opcode doesn't specify (as in override) some instance
8459 parameters, then try to use them from the export information, if
8463 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
8465 if self.op.disk_template is None:
8466 if einfo.has_option(constants.INISECT_INS, "disk_template"):
8467 self.op.disk_template = einfo.get(constants.INISECT_INS,
8470 raise errors.OpPrereqError("No disk template specified and the export"
8471 " is missing the disk_template information",
8474 if not self.op.disks:
8475 if einfo.has_option(constants.INISECT_INS, "disk_count"):
8477 # TODO: import the disk iv_name too
8478 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
8479 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
8480 disks.append({constants.IDISK_SIZE: disk_sz})
8481 self.op.disks = disks
8483 raise errors.OpPrereqError("No disk info specified and the export"
8484 " is missing the disk information",
8487 if (not self.op.nics and
8488 einfo.has_option(constants.INISECT_INS, "nic_count")):
8490 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
8492 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
8493 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
8498 if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):
8499 self.op.tags = einfo.get(constants.INISECT_INS, "tags").split()
8501 if (self.op.hypervisor is None and
8502 einfo.has_option(constants.INISECT_INS, "hypervisor")):
8503 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
8505 if einfo.has_section(constants.INISECT_HYP):
8506 # use the export parameters but do not override the ones
8507 # specified by the user
8508 for name, value in einfo.items(constants.INISECT_HYP):
8509 if name not in self.op.hvparams:
8510 self.op.hvparams[name] = value
8512 if einfo.has_section(constants.INISECT_BEP):
8513 # use the parameters, without overriding
8514 for name, value in einfo.items(constants.INISECT_BEP):
8515 if name not in self.op.beparams:
8516 self.op.beparams[name] = value
8518 # try to read the parameters old style, from the main section
8519 for name in constants.BES_PARAMETERS:
8520 if (name not in self.op.beparams and
8521 einfo.has_option(constants.INISECT_INS, name)):
8522 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
8524 if einfo.has_section(constants.INISECT_OSP):
8525 # use the parameters, without overriding
8526 for name, value in einfo.items(constants.INISECT_OSP):
8527 if name not in self.op.osparams:
8528 self.op.osparams[name] = value
8530 def _RevertToDefaults(self, cluster):
8531 """Revert the instance parameters to the default values.
8535 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
8536 for name in self.op.hvparams.keys():
8537 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
8538 del self.op.hvparams[name]
8540 be_defs = cluster.SimpleFillBE({})
8541 for name in self.op.beparams.keys():
8542 if name in be_defs and be_defs[name] == self.op.beparams[name]:
8543 del self.op.beparams[name]
8545 nic_defs = cluster.SimpleFillNIC({})
8546 for nic in self.op.nics:
8547 for name in constants.NICS_PARAMETERS:
8548 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
8551 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
8552 for name in self.op.osparams.keys():
8553 if name in os_defs and os_defs[name] == self.op.osparams[name]:
8554 del self.op.osparams[name]
8556 def _CalculateFileStorageDir(self):
8557 """Calculate final instance file storage dir.
8560 # file storage dir calculation/check
8561 self.instance_file_storage_dir = None
8562 if self.op.disk_template in constants.DTS_FILEBASED:
8563 # build the full file storage dir path
8566 if self.op.disk_template == constants.DT_SHARED_FILE:
8567 get_fsd_fn = self.cfg.GetSharedFileStorageDir
8569 get_fsd_fn = self.cfg.GetFileStorageDir
8571 cfg_storagedir = get_fsd_fn()
8572 if not cfg_storagedir:
8573 raise errors.OpPrereqError("Cluster file storage dir not defined")
8574 joinargs.append(cfg_storagedir)
8576 if self.op.file_storage_dir is not None:
8577 joinargs.append(self.op.file_storage_dir)
8579 joinargs.append(self.op.instance_name)
8581 # pylint: disable=W0142
8582 self.instance_file_storage_dir = utils.PathJoin(*joinargs)
8584 def CheckPrereq(self):
8585 """Check prerequisites.
8588 self._CalculateFileStorageDir()
8590 if self.op.mode == constants.INSTANCE_IMPORT:
8591 export_info = self._ReadExportInfo()
8592 self._ReadExportParams(export_info)
8594 if (not self.cfg.GetVGName() and
8595 self.op.disk_template not in constants.DTS_NOT_LVM):
8596 raise errors.OpPrereqError("Cluster does not support lvm-based"
8597 " instances", errors.ECODE_STATE)
8599 if self.op.hypervisor is None:
8600 self.op.hypervisor = self.cfg.GetHypervisorType()
8602 cluster = self.cfg.GetClusterInfo()
8603 enabled_hvs = cluster.enabled_hypervisors
8604 if self.op.hypervisor not in enabled_hvs:
8605 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
8606 " cluster (%s)" % (self.op.hypervisor,
8607 ",".join(enabled_hvs)),
8610 # Check tag validity
8611 for tag in self.op.tags:
8612 objects.TaggableObject.ValidateTag(tag)
8614 # check hypervisor parameter syntax (locally)
8615 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
8616 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
8618 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
8619 hv_type.CheckParameterSyntax(filled_hvp)
8620 self.hv_full = filled_hvp
8621 # check that we don't specify global parameters on an instance
8622 _CheckGlobalHvParams(self.op.hvparams)
8624 # fill and remember the beparams dict
8625 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
8626 self.be_full = cluster.SimpleFillBE(self.op.beparams)
8628 # build os parameters
8629 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
8631 # now that hvp/bep are in final format, let's reset to defaults,
8633 if self.op.identify_defaults:
8634 self._RevertToDefaults(cluster)
8638 for idx, nic in enumerate(self.op.nics):
8639 nic_mode_req = nic.get(constants.INIC_MODE, None)
8640 nic_mode = nic_mode_req
8641 if nic_mode is None:
8642 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
8644 # in routed mode, for the first nic, the default ip is 'auto'
8645 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
8646 default_ip_mode = constants.VALUE_AUTO
8648 default_ip_mode = constants.VALUE_NONE
8650 # ip validity checks
8651 ip = nic.get(constants.INIC_IP, default_ip_mode)
8652 if ip is None or ip.lower() == constants.VALUE_NONE:
8654 elif ip.lower() == constants.VALUE_AUTO:
8655 if not self.op.name_check:
8656 raise errors.OpPrereqError("IP address set to auto but name checks"
8657 " have been skipped",
8659 nic_ip = self.hostname1.ip
8661 if not netutils.IPAddress.IsValid(ip):
8662 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
8666 # TODO: check the ip address for uniqueness
8667 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
8668 raise errors.OpPrereqError("Routed nic mode requires an ip address",
8671 # MAC address verification
8672 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
8673 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8674 mac = utils.NormalizeAndValidateMac(mac)
8677 self.cfg.ReserveMAC(mac, self.proc.GetECId())
8678 except errors.ReservationError:
8679 raise errors.OpPrereqError("MAC address %s already in use"
8680 " in cluster" % mac,
8681 errors.ECODE_NOTUNIQUE)
8683 # Build nic parameters
8684 link = nic.get(constants.INIC_LINK, None)
8687 nicparams[constants.NIC_MODE] = nic_mode_req
8689 nicparams[constants.NIC_LINK] = link
8691 check_params = cluster.SimpleFillNIC(nicparams)
8692 objects.NIC.CheckParameterSyntax(check_params)
8693 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
8695 # disk checks/pre-build
8696 default_vg = self.cfg.GetVGName()
8698 for disk in self.op.disks:
8699 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
8700 if mode not in constants.DISK_ACCESS_SET:
8701 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
8702 mode, errors.ECODE_INVAL)
8703 size = disk.get(constants.IDISK_SIZE, None)
8705 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
8708 except (TypeError, ValueError):
8709 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
8712 data_vg = disk.get(constants.IDISK_VG, default_vg)
8714 constants.IDISK_SIZE: size,
8715 constants.IDISK_MODE: mode,
8716 constants.IDISK_VG: data_vg,
8717 constants.IDISK_METAVG: disk.get(constants.IDISK_METAVG, data_vg),
8719 if constants.IDISK_ADOPT in disk:
8720 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
8721 self.disks.append(new_disk)
8723 if self.op.mode == constants.INSTANCE_IMPORT:
8725 # Check that the new instance doesn't have less disks than the export
8726 instance_disks = len(self.disks)
8727 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
8728 if instance_disks < export_disks:
8729 raise errors.OpPrereqError("Not enough disks to import."
8730 " (instance: %d, export: %d)" %
8731 (instance_disks, export_disks),
8735 for idx in range(export_disks):
8736 option = "disk%d_dump" % idx
8737 if export_info.has_option(constants.INISECT_INS, option):
8738 # FIXME: are the old os-es, disk sizes, etc. useful?
8739 export_name = export_info.get(constants.INISECT_INS, option)
8740 image = utils.PathJoin(self.op.src_path, export_name)
8741 disk_images.append(image)
8743 disk_images.append(False)
8745 self.src_images = disk_images
8747 old_name = export_info.get(constants.INISECT_INS, "name")
8749 exp_nic_count = export_info.getint(constants.INISECT_INS, "nic_count")
8750 except (TypeError, ValueError), err:
8751 raise errors.OpPrereqError("Invalid export file, nic_count is not"
8752 " an integer: %s" % str(err),
8754 if self.op.instance_name == old_name:
8755 for idx, nic in enumerate(self.nics):
8756 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
8757 nic_mac_ini = "nic%d_mac" % idx
8758 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
8760 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
8762 # ip ping checks (we use the same ip that was resolved in ExpandNames)
8763 if self.op.ip_check:
8764 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
8765 raise errors.OpPrereqError("IP %s of instance %s already in use" %
8766 (self.check_ip, self.op.instance_name),
8767 errors.ECODE_NOTUNIQUE)
8769 #### mac address generation
8770 # By generating here the mac address both the allocator and the hooks get
8771 # the real final mac address rather than the 'auto' or 'generate' value.
8772 # There is a race condition between the generation and the instance object
8773 # creation, which means that we know the mac is valid now, but we're not
8774 # sure it will be when we actually add the instance. If things go bad
8775 # adding the instance will abort because of a duplicate mac, and the
8776 # creation job will fail.
8777 for nic in self.nics:
8778 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8779 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
8783 if self.op.iallocator is not None:
8784 self._RunAllocator()
8786 #### node related checks
8788 # check primary node
8789 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
8790 assert self.pnode is not None, \
8791 "Cannot retrieve locked node %s" % self.op.pnode
8793 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
8794 pnode.name, errors.ECODE_STATE)
8796 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
8797 pnode.name, errors.ECODE_STATE)
8798 if not pnode.vm_capable:
8799 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
8800 " '%s'" % pnode.name, errors.ECODE_STATE)
8802 self.secondaries = []
8804 # mirror node verification
8805 if self.op.disk_template in constants.DTS_INT_MIRROR:
8806 if self.op.snode == pnode.name:
8807 raise errors.OpPrereqError("The secondary node cannot be the"
8808 " primary node", errors.ECODE_INVAL)
8809 _CheckNodeOnline(self, self.op.snode)
8810 _CheckNodeNotDrained(self, self.op.snode)
8811 _CheckNodeVmCapable(self, self.op.snode)
8812 self.secondaries.append(self.op.snode)
8814 nodenames = [pnode.name] + self.secondaries
8816 if not self.adopt_disks:
8817 # Check lv size requirements, if not adopting
8818 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
8819 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
8821 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
8822 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
8823 disk[constants.IDISK_ADOPT])
8824 for disk in self.disks])
8825 if len(all_lvs) != len(self.disks):
8826 raise errors.OpPrereqError("Duplicate volume names given for adoption",
8828 for lv_name in all_lvs:
8830 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
8831 # to ReserveLV uses the same syntax
8832 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
8833 except errors.ReservationError:
8834 raise errors.OpPrereqError("LV named %s used by another instance" %
8835 lv_name, errors.ECODE_NOTUNIQUE)
8837 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
8838 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
8840 node_lvs = self.rpc.call_lv_list([pnode.name],
8841 vg_names.payload.keys())[pnode.name]
8842 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
8843 node_lvs = node_lvs.payload
8845 delta = all_lvs.difference(node_lvs.keys())
8847 raise errors.OpPrereqError("Missing logical volume(s): %s" %
8848 utils.CommaJoin(delta),
8850 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
8852 raise errors.OpPrereqError("Online logical volumes found, cannot"
8853 " adopt: %s" % utils.CommaJoin(online_lvs),
8855 # update the size of disk based on what is found
8856 for dsk in self.disks:
8857 dsk[constants.IDISK_SIZE] = \
8858 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
8859 dsk[constants.IDISK_ADOPT])][0]))
8861 elif self.op.disk_template == constants.DT_BLOCK:
8862 # Normalize and de-duplicate device paths
8863 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
8864 for disk in self.disks])
8865 if len(all_disks) != len(self.disks):
8866 raise errors.OpPrereqError("Duplicate disk names given for adoption",
8868 baddisks = [d for d in all_disks
8869 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
8871 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
8872 " cannot be adopted" %
8873 (", ".join(baddisks),
8874 constants.ADOPTABLE_BLOCKDEV_ROOT),
8877 node_disks = self.rpc.call_bdev_sizes([pnode.name],
8878 list(all_disks))[pnode.name]
8879 node_disks.Raise("Cannot get block device information from node %s" %
8881 node_disks = node_disks.payload
8882 delta = all_disks.difference(node_disks.keys())
8884 raise errors.OpPrereqError("Missing block device(s): %s" %
8885 utils.CommaJoin(delta),
8887 for dsk in self.disks:
8888 dsk[constants.IDISK_SIZE] = \
8889 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
8891 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
8893 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
8894 # check OS parameters (remotely)
8895 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
8897 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
8899 # memory check on primary node
8901 _CheckNodeFreeMemory(self, self.pnode.name,
8902 "creating instance %s" % self.op.instance_name,
8903 self.be_full[constants.BE_MEMORY],
8906 self.dry_run_result = list(nodenames)
8908 def Exec(self, feedback_fn):
8909 """Create and add the instance to the cluster.
8912 instance = self.op.instance_name
8913 pnode_name = self.pnode.name
8915 ht_kind = self.op.hypervisor
8916 if ht_kind in constants.HTS_REQ_PORT:
8917 network_port = self.cfg.AllocatePort()
8921 disks = _GenerateDiskTemplate(self,
8922 self.op.disk_template,
8923 instance, pnode_name,
8926 self.instance_file_storage_dir,
8927 self.op.file_driver,
8931 iobj = objects.Instance(name=instance, os=self.op.os_type,
8932 primary_node=pnode_name,
8933 nics=self.nics, disks=disks,
8934 disk_template=self.op.disk_template,
8936 network_port=network_port,
8937 beparams=self.op.beparams,
8938 hvparams=self.op.hvparams,
8939 hypervisor=self.op.hypervisor,
8940 osparams=self.op.osparams,
8944 for tag in self.op.tags:
8947 if self.adopt_disks:
8948 if self.op.disk_template == constants.DT_PLAIN:
8949 # rename LVs to the newly-generated names; we need to construct
8950 # 'fake' LV disks with the old data, plus the new unique_id
8951 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
8953 for t_dsk, a_dsk in zip(tmp_disks, self.disks):
8954 rename_to.append(t_dsk.logical_id)
8955 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
8956 self.cfg.SetDiskID(t_dsk, pnode_name)
8957 result = self.rpc.call_blockdev_rename(pnode_name,
8958 zip(tmp_disks, rename_to))
8959 result.Raise("Failed to rename adoped LVs")
8961 feedback_fn("* creating instance disks...")
8963 _CreateDisks(self, iobj)
8964 except errors.OpExecError:
8965 self.LogWarning("Device creation failed, reverting...")
8967 _RemoveDisks(self, iobj)
8969 self.cfg.ReleaseDRBDMinors(instance)
8972 feedback_fn("adding instance %s to cluster config" % instance)
8974 self.cfg.AddInstance(iobj, self.proc.GetECId())
8976 # Declare that we don't want to remove the instance lock anymore, as we've
8977 # added the instance to the config
8978 del self.remove_locks[locking.LEVEL_INSTANCE]
8980 if self.op.mode == constants.INSTANCE_IMPORT:
8981 # Release unused nodes
8982 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
8985 _ReleaseLocks(self, locking.LEVEL_NODE)
8988 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
8989 feedback_fn("* wiping instance disks...")
8991 _WipeDisks(self, iobj)
8992 except errors.OpExecError, err:
8993 logging.exception("Wiping disks failed")
8994 self.LogWarning("Wiping instance disks failed (%s)", err)
8998 # Something is already wrong with the disks, don't do anything else
9000 elif self.op.wait_for_sync:
9001 disk_abort = not _WaitForSync(self, iobj)
9002 elif iobj.disk_template in constants.DTS_INT_MIRROR:
9003 # make sure the disks are not degraded (still sync-ing is ok)
9004 feedback_fn("* checking mirrors status")
9005 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
9010 _RemoveDisks(self, iobj)
9011 self.cfg.RemoveInstance(iobj.name)
9012 # Make sure the instance lock gets removed
9013 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
9014 raise errors.OpExecError("There are some degraded disks for"
9017 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
9018 if self.op.mode == constants.INSTANCE_CREATE:
9019 if not self.op.no_install:
9020 pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and
9021 not self.op.wait_for_sync)
9023 feedback_fn("* pausing disk sync to install instance OS")
9024 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
9026 for idx, success in enumerate(result.payload):
9028 logging.warn("pause-sync of instance %s for disk %d failed",
9031 feedback_fn("* running the instance OS create scripts...")
9032 # FIXME: pass debug option from opcode to backend
9033 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
9034 self.op.debug_level)
9036 feedback_fn("* resuming disk sync")
9037 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
9039 for idx, success in enumerate(result.payload):
9041 logging.warn("resume-sync of instance %s for disk %d failed",
9044 result.Raise("Could not add os for instance %s"
9045 " on node %s" % (instance, pnode_name))
9047 elif self.op.mode == constants.INSTANCE_IMPORT:
9048 feedback_fn("* running the instance OS import scripts...")
9052 for idx, image in enumerate(self.src_images):
9056 # FIXME: pass debug option from opcode to backend
9057 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
9058 constants.IEIO_FILE, (image, ),
9059 constants.IEIO_SCRIPT,
9060 (iobj.disks[idx], idx),
9062 transfers.append(dt)
9065 masterd.instance.TransferInstanceData(self, feedback_fn,
9066 self.op.src_node, pnode_name,
9067 self.pnode.secondary_ip,
9069 if not compat.all(import_result):
9070 self.LogWarning("Some disks for instance %s on node %s were not"
9071 " imported successfully" % (instance, pnode_name))
9073 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
9074 feedback_fn("* preparing remote import...")
9075 # The source cluster will stop the instance before attempting to make a
9076 # connection. In some cases stopping an instance can take a long time,
9077 # hence the shutdown timeout is added to the connection timeout.
9078 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
9079 self.op.source_shutdown_timeout)
9080 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9082 assert iobj.primary_node == self.pnode.name
9084 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
9085 self.source_x509_ca,
9086 self._cds, timeouts)
9087 if not compat.all(disk_results):
9088 # TODO: Should the instance still be started, even if some disks
9089 # failed to import (valid for local imports, too)?
9090 self.LogWarning("Some disks for instance %s on node %s were not"
9091 " imported successfully" % (instance, pnode_name))
9093 # Run rename script on newly imported instance
9094 assert iobj.name == instance
9095 feedback_fn("Running rename script for %s" % instance)
9096 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
9097 self.source_instance_name,
9098 self.op.debug_level)
9100 self.LogWarning("Failed to run rename script for %s on node"
9101 " %s: %s" % (instance, pnode_name, result.fail_msg))
9104 # also checked in the prereq part
9105 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
9109 iobj.admin_up = True
9110 self.cfg.Update(iobj, feedback_fn)
9111 logging.info("Starting instance %s on node %s", instance, pnode_name)
9112 feedback_fn("* starting instance...")
9113 result = self.rpc.call_instance_start(pnode_name, iobj,
9115 result.Raise("Could not start instance")
9117 return list(iobj.all_nodes)
9120 class LUInstanceConsole(NoHooksLU):
9121 """Connect to an instance's console.
9123 This is somewhat special in that it returns the command line that
9124 you need to run on the master node in order to connect to the
9130 def ExpandNames(self):
9131 self._ExpandAndLockInstance()
9133 def CheckPrereq(self):
9134 """Check prerequisites.
9136 This checks that the instance is in the cluster.
9139 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9140 assert self.instance is not None, \
9141 "Cannot retrieve locked instance %s" % self.op.instance_name
9142 _CheckNodeOnline(self, self.instance.primary_node)
9144 def Exec(self, feedback_fn):
9145 """Connect to the console of an instance
9148 instance = self.instance
9149 node = instance.primary_node
9151 node_insts = self.rpc.call_instance_list([node],
9152 [instance.hypervisor])[node]
9153 node_insts.Raise("Can't get node information from %s" % node)
9155 if instance.name not in node_insts.payload:
9156 if instance.admin_up:
9157 state = constants.INSTST_ERRORDOWN
9159 state = constants.INSTST_ADMINDOWN
9160 raise errors.OpExecError("Instance %s is not running (state %s)" %
9161 (instance.name, state))
9163 logging.debug("Connecting to console of %s on %s", instance.name, node)
9165 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
9168 def _GetInstanceConsole(cluster, instance):
9169 """Returns console information for an instance.
9171 @type cluster: L{objects.Cluster}
9172 @type instance: L{objects.Instance}
9176 hyper = hypervisor.GetHypervisor(instance.hypervisor)
9177 # beparams and hvparams are passed separately, to avoid editing the
9178 # instance and then saving the defaults in the instance itself.
9179 hvparams = cluster.FillHV(instance)
9180 beparams = cluster.FillBE(instance)
9181 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
9183 assert console.instance == instance.name
9184 assert console.Validate()
9186 return console.ToDict()
9189 class LUInstanceReplaceDisks(LogicalUnit):
9190 """Replace the disks of an instance.
9193 HPATH = "mirrors-replace"
9194 HTYPE = constants.HTYPE_INSTANCE
9197 def CheckArguments(self):
9198 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
9201 def ExpandNames(self):
9202 self._ExpandAndLockInstance()
9204 assert locking.LEVEL_NODE not in self.needed_locks
9205 assert locking.LEVEL_NODEGROUP not in self.needed_locks
9207 assert self.op.iallocator is None or self.op.remote_node is None, \
9208 "Conflicting options"
9210 if self.op.remote_node is not None:
9211 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9213 # Warning: do not remove the locking of the new secondary here
9214 # unless DRBD8.AddChildren is changed to work in parallel;
9215 # currently it doesn't since parallel invocations of
9216 # FindUnusedMinor will conflict
9217 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
9218 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
9220 self.needed_locks[locking.LEVEL_NODE] = []
9221 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9223 if self.op.iallocator is not None:
9224 # iallocator will select a new node in the same group
9225 self.needed_locks[locking.LEVEL_NODEGROUP] = []
9227 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
9228 self.op.iallocator, self.op.remote_node,
9229 self.op.disks, False, self.op.early_release)
9231 self.tasklets = [self.replacer]
9233 def DeclareLocks(self, level):
9234 if level == locking.LEVEL_NODEGROUP:
9235 assert self.op.remote_node is None
9236 assert self.op.iallocator is not None
9237 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
9239 self.share_locks[locking.LEVEL_NODEGROUP] = 1
9240 self.needed_locks[locking.LEVEL_NODEGROUP] = \
9241 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
9243 elif level == locking.LEVEL_NODE:
9244 if self.op.iallocator is not None:
9245 assert self.op.remote_node is None
9246 assert not self.needed_locks[locking.LEVEL_NODE]
9248 # Lock member nodes of all locked groups
9249 self.needed_locks[locking.LEVEL_NODE] = [node_name
9250 for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
9251 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
9253 self._LockInstancesNodes()
9255 def BuildHooksEnv(self):
9258 This runs on the master, the primary and all the secondaries.
9261 instance = self.replacer.instance
9263 "MODE": self.op.mode,
9264 "NEW_SECONDARY": self.op.remote_node,
9265 "OLD_SECONDARY": instance.secondary_nodes[0],
9267 env.update(_BuildInstanceHookEnvByObject(self, instance))
9270 def BuildHooksNodes(self):
9271 """Build hooks nodes.
9274 instance = self.replacer.instance
9276 self.cfg.GetMasterNode(),
9277 instance.primary_node,
9279 if self.op.remote_node is not None:
9280 nl.append(self.op.remote_node)
9283 def CheckPrereq(self):
9284 """Check prerequisites.
9287 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
9288 self.op.iallocator is None)
9290 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
9292 _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
9294 return LogicalUnit.CheckPrereq(self)
9297 class TLReplaceDisks(Tasklet):
9298 """Replaces disks for an instance.
9300 Note: Locking is not within the scope of this class.
9303 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
9304 disks, delay_iallocator, early_release):
9305 """Initializes this class.
9308 Tasklet.__init__(self, lu)
9311 self.instance_name = instance_name
9313 self.iallocator_name = iallocator_name
9314 self.remote_node = remote_node
9316 self.delay_iallocator = delay_iallocator
9317 self.early_release = early_release
9320 self.instance = None
9321 self.new_node = None
9322 self.target_node = None
9323 self.other_node = None
9324 self.remote_node_info = None
9325 self.node_secondary_ip = None
9328 def CheckArguments(mode, remote_node, iallocator):
9329 """Helper function for users of this class.
9332 # check for valid parameter combination
9333 if mode == constants.REPLACE_DISK_CHG:
9334 if remote_node is None and iallocator is None:
9335 raise errors.OpPrereqError("When changing the secondary either an"
9336 " iallocator script must be used or the"
9337 " new node given", errors.ECODE_INVAL)
9339 if remote_node is not None and iallocator is not None:
9340 raise errors.OpPrereqError("Give either the iallocator or the new"
9341 " secondary, not both", errors.ECODE_INVAL)
9343 elif remote_node is not None or iallocator is not None:
9344 # Not replacing the secondary
9345 raise errors.OpPrereqError("The iallocator and new node options can"
9346 " only be used when changing the"
9347 " secondary node", errors.ECODE_INVAL)
9350 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
9351 """Compute a new secondary node using an IAllocator.
9354 ial = IAllocator(lu.cfg, lu.rpc,
9355 mode=constants.IALLOCATOR_MODE_RELOC,
9357 relocate_from=list(relocate_from))
9359 ial.Run(iallocator_name)
9362 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
9363 " %s" % (iallocator_name, ial.info),
9366 if len(ial.result) != ial.required_nodes:
9367 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
9368 " of nodes (%s), required %s" %
9370 len(ial.result), ial.required_nodes),
9373 remote_node_name = ial.result[0]
9375 lu.LogInfo("Selected new secondary for instance '%s': %s",
9376 instance_name, remote_node_name)
9378 return remote_node_name
9380 def _FindFaultyDisks(self, node_name):
9381 """Wrapper for L{_FindFaultyInstanceDisks}.
9384 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
9387 def _CheckDisksActivated(self, instance):
9388 """Checks if the instance disks are activated.
9390 @param instance: The instance to check disks
9391 @return: True if they are activated, False otherwise
9394 nodes = instance.all_nodes
9396 for idx, dev in enumerate(instance.disks):
9398 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
9399 self.cfg.SetDiskID(dev, node)
9401 result = self.rpc.call_blockdev_find(node, dev)
9405 elif result.fail_msg or not result.payload:
9410 def CheckPrereq(self):
9411 """Check prerequisites.
9413 This checks that the instance is in the cluster.
9416 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
9417 assert instance is not None, \
9418 "Cannot retrieve locked instance %s" % self.instance_name
9420 if instance.disk_template != constants.DT_DRBD8:
9421 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
9422 " instances", errors.ECODE_INVAL)
9424 if len(instance.secondary_nodes) != 1:
9425 raise errors.OpPrereqError("The instance has a strange layout,"
9426 " expected one secondary but found %d" %
9427 len(instance.secondary_nodes),
9430 if not self.delay_iallocator:
9431 self._CheckPrereq2()
9433 def _CheckPrereq2(self):
9434 """Check prerequisites, second part.
9436 This function should always be part of CheckPrereq. It was separated and is
9437 now called from Exec because during node evacuation iallocator was only
9438 called with an unmodified cluster model, not taking planned changes into
9442 instance = self.instance
9443 secondary_node = instance.secondary_nodes[0]
9445 if self.iallocator_name is None:
9446 remote_node = self.remote_node
9448 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
9449 instance.name, instance.secondary_nodes)
9451 if remote_node is None:
9452 self.remote_node_info = None
9454 assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \
9455 "Remote node '%s' is not locked" % remote_node
9457 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
9458 assert self.remote_node_info is not None, \
9459 "Cannot retrieve locked node %s" % remote_node
9461 if remote_node == self.instance.primary_node:
9462 raise errors.OpPrereqError("The specified node is the primary node of"
9463 " the instance", errors.ECODE_INVAL)
9465 if remote_node == secondary_node:
9466 raise errors.OpPrereqError("The specified node is already the"
9467 " secondary node of the instance",
9470 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
9471 constants.REPLACE_DISK_CHG):
9472 raise errors.OpPrereqError("Cannot specify disks to be replaced",
9475 if self.mode == constants.REPLACE_DISK_AUTO:
9476 if not self._CheckDisksActivated(instance):
9477 raise errors.OpPrereqError("Please run activate-disks on instance %s"
9478 " first" % self.instance_name,
9480 faulty_primary = self._FindFaultyDisks(instance.primary_node)
9481 faulty_secondary = self._FindFaultyDisks(secondary_node)
9483 if faulty_primary and faulty_secondary:
9484 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
9485 " one node and can not be repaired"
9486 " automatically" % self.instance_name,
9490 self.disks = faulty_primary
9491 self.target_node = instance.primary_node
9492 self.other_node = secondary_node
9493 check_nodes = [self.target_node, self.other_node]
9494 elif faulty_secondary:
9495 self.disks = faulty_secondary
9496 self.target_node = secondary_node
9497 self.other_node = instance.primary_node
9498 check_nodes = [self.target_node, self.other_node]
9504 # Non-automatic modes
9505 if self.mode == constants.REPLACE_DISK_PRI:
9506 self.target_node = instance.primary_node
9507 self.other_node = secondary_node
9508 check_nodes = [self.target_node, self.other_node]
9510 elif self.mode == constants.REPLACE_DISK_SEC:
9511 self.target_node = secondary_node
9512 self.other_node = instance.primary_node
9513 check_nodes = [self.target_node, self.other_node]
9515 elif self.mode == constants.REPLACE_DISK_CHG:
9516 self.new_node = remote_node
9517 self.other_node = instance.primary_node
9518 self.target_node = secondary_node
9519 check_nodes = [self.new_node, self.other_node]
9521 _CheckNodeNotDrained(self.lu, remote_node)
9522 _CheckNodeVmCapable(self.lu, remote_node)
9524 old_node_info = self.cfg.GetNodeInfo(secondary_node)
9525 assert old_node_info is not None
9526 if old_node_info.offline and not self.early_release:
9527 # doesn't make sense to delay the release
9528 self.early_release = True
9529 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
9530 " early-release mode", secondary_node)
9533 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
9536 # If not specified all disks should be replaced
9538 self.disks = range(len(self.instance.disks))
9540 for node in check_nodes:
9541 _CheckNodeOnline(self.lu, node)
9543 touched_nodes = frozenset(node_name for node_name in [self.new_node,
9546 if node_name is not None)
9548 # Release unneeded node locks
9549 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
9551 # Release any owned node group
9552 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
9553 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
9555 # Check whether disks are valid
9556 for disk_idx in self.disks:
9557 instance.FindDisk(disk_idx)
9559 # Get secondary node IP addresses
9560 self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node)
9561 in self.cfg.GetMultiNodeInfo(touched_nodes))
9563 def Exec(self, feedback_fn):
9564 """Execute disk replacement.
9566 This dispatches the disk replacement to the appropriate handler.
9569 if self.delay_iallocator:
9570 self._CheckPrereq2()
9573 # Verify owned locks before starting operation
9574 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
9575 assert set(owned_nodes) == set(self.node_secondary_ip), \
9576 ("Incorrect node locks, owning %s, expected %s" %
9577 (owned_nodes, self.node_secondary_ip.keys()))
9579 owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
9580 assert list(owned_instances) == [self.instance_name], \
9581 "Instance '%s' not locked" % self.instance_name
9583 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
9584 "Should not own any node group lock at this point"
9587 feedback_fn("No disks need replacement")
9590 feedback_fn("Replacing disk(s) %s for %s" %
9591 (utils.CommaJoin(self.disks), self.instance.name))
9593 activate_disks = (not self.instance.admin_up)
9595 # Activate the instance disks if we're replacing them on a down instance
9597 _StartInstanceDisks(self.lu, self.instance, True)
9600 # Should we replace the secondary node?
9601 if self.new_node is not None:
9602 fn = self._ExecDrbd8Secondary
9604 fn = self._ExecDrbd8DiskOnly
9606 result = fn(feedback_fn)
9608 # Deactivate the instance disks if we're replacing them on a
9611 _SafeShutdownInstanceDisks(self.lu, self.instance)
9614 # Verify owned locks
9615 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
9616 nodes = frozenset(self.node_secondary_ip)
9617 assert ((self.early_release and not owned_nodes) or
9618 (not self.early_release and not (set(owned_nodes) - nodes))), \
9619 ("Not owning the correct locks, early_release=%s, owned=%r,"
9620 " nodes=%r" % (self.early_release, owned_nodes, nodes))
9624 def _CheckVolumeGroup(self, nodes):
9625 self.lu.LogInfo("Checking volume groups")
9627 vgname = self.cfg.GetVGName()
9629 # Make sure volume group exists on all involved nodes
9630 results = self.rpc.call_vg_list(nodes)
9632 raise errors.OpExecError("Can't list volume groups on the nodes")
9636 res.Raise("Error checking node %s" % node)
9637 if vgname not in res.payload:
9638 raise errors.OpExecError("Volume group '%s' not found on node %s" %
9641 def _CheckDisksExistence(self, nodes):
9642 # Check disk existence
9643 for idx, dev in enumerate(self.instance.disks):
9644 if idx not in self.disks:
9648 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
9649 self.cfg.SetDiskID(dev, node)
9651 result = self.rpc.call_blockdev_find(node, dev)
9653 msg = result.fail_msg
9654 if msg or not result.payload:
9656 msg = "disk not found"
9657 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
9660 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
9661 for idx, dev in enumerate(self.instance.disks):
9662 if idx not in self.disks:
9665 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
9668 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
9670 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
9671 " replace disks for instance %s" %
9672 (node_name, self.instance.name))
9674 def _CreateNewStorage(self, node_name):
9675 """Create new storage on the primary or secondary node.
9677 This is only used for same-node replaces, not for changing the
9678 secondary node, hence we don't want to modify the existing disk.
9683 for idx, dev in enumerate(self.instance.disks):
9684 if idx not in self.disks:
9687 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
9689 self.cfg.SetDiskID(dev, node_name)
9691 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
9692 names = _GenerateUniqueNames(self.lu, lv_names)
9694 vg_data = dev.children[0].logical_id[0]
9695 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
9696 logical_id=(vg_data, names[0]))
9697 vg_meta = dev.children[1].logical_id[0]
9698 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
9699 logical_id=(vg_meta, names[1]))
9701 new_lvs = [lv_data, lv_meta]
9702 old_lvs = [child.Copy() for child in dev.children]
9703 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
9705 # we pass force_create=True to force the LVM creation
9706 for new_lv in new_lvs:
9707 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
9708 _GetInstanceInfoText(self.instance), False)
9712 def _CheckDevices(self, node_name, iv_names):
9713 for name, (dev, _, _) in iv_names.iteritems():
9714 self.cfg.SetDiskID(dev, node_name)
9716 result = self.rpc.call_blockdev_find(node_name, dev)
9718 msg = result.fail_msg
9719 if msg or not result.payload:
9721 msg = "disk not found"
9722 raise errors.OpExecError("Can't find DRBD device %s: %s" %
9725 if result.payload.is_degraded:
9726 raise errors.OpExecError("DRBD device %s is degraded!" % name)
9728 def _RemoveOldStorage(self, node_name, iv_names):
9729 for name, (_, old_lvs, _) in iv_names.iteritems():
9730 self.lu.LogInfo("Remove logical volumes for %s" % name)
9733 self.cfg.SetDiskID(lv, node_name)
9735 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
9737 self.lu.LogWarning("Can't remove old LV: %s" % msg,
9738 hint="remove unused LVs manually")
9740 def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
9741 """Replace a disk on the primary or secondary for DRBD 8.
9743 The algorithm for replace is quite complicated:
9745 1. for each disk to be replaced:
9747 1. create new LVs on the target node with unique names
9748 1. detach old LVs from the drbd device
9749 1. rename old LVs to name_replaced.<time_t>
9750 1. rename new LVs to old LVs
9751 1. attach the new LVs (with the old names now) to the drbd device
9753 1. wait for sync across all devices
9755 1. for each modified disk:
9757 1. remove old LVs (which have the name name_replaces.<time_t>)
9759 Failures are not very well handled.
9764 # Step: check device activation
9765 self.lu.LogStep(1, steps_total, "Check device existence")
9766 self._CheckDisksExistence([self.other_node, self.target_node])
9767 self._CheckVolumeGroup([self.target_node, self.other_node])
9769 # Step: check other node consistency
9770 self.lu.LogStep(2, steps_total, "Check peer consistency")
9771 self._CheckDisksConsistency(self.other_node,
9772 self.other_node == self.instance.primary_node,
9775 # Step: create new storage
9776 self.lu.LogStep(3, steps_total, "Allocate new storage")
9777 iv_names = self._CreateNewStorage(self.target_node)
9779 # Step: for each lv, detach+rename*2+attach
9780 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9781 for dev, old_lvs, new_lvs in iv_names.itervalues():
9782 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
9784 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
9786 result.Raise("Can't detach drbd from local storage on node"
9787 " %s for device %s" % (self.target_node, dev.iv_name))
9789 #cfg.Update(instance)
9791 # ok, we created the new LVs, so now we know we have the needed
9792 # storage; as such, we proceed on the target node to rename
9793 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
9794 # using the assumption that logical_id == physical_id (which in
9795 # turn is the unique_id on that node)
9797 # FIXME(iustin): use a better name for the replaced LVs
9798 temp_suffix = int(time.time())
9799 ren_fn = lambda d, suff: (d.physical_id[0],
9800 d.physical_id[1] + "_replaced-%s" % suff)
9802 # Build the rename list based on what LVs exist on the node
9803 rename_old_to_new = []
9804 for to_ren in old_lvs:
9805 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
9806 if not result.fail_msg and result.payload:
9808 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
9810 self.lu.LogInfo("Renaming the old LVs on the target node")
9811 result = self.rpc.call_blockdev_rename(self.target_node,
9813 result.Raise("Can't rename old LVs on node %s" % self.target_node)
9815 # Now we rename the new LVs to the old LVs
9816 self.lu.LogInfo("Renaming the new LVs on the target node")
9817 rename_new_to_old = [(new, old.physical_id)
9818 for old, new in zip(old_lvs, new_lvs)]
9819 result = self.rpc.call_blockdev_rename(self.target_node,
9821 result.Raise("Can't rename new LVs on node %s" % self.target_node)
9823 # Intermediate steps of in memory modifications
9824 for old, new in zip(old_lvs, new_lvs):
9825 new.logical_id = old.logical_id
9826 self.cfg.SetDiskID(new, self.target_node)
9828 # We need to modify old_lvs so that removal later removes the
9829 # right LVs, not the newly added ones; note that old_lvs is a
9831 for disk in old_lvs:
9832 disk.logical_id = ren_fn(disk, temp_suffix)
9833 self.cfg.SetDiskID(disk, self.target_node)
9835 # Now that the new lvs have the old name, we can add them to the device
9836 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
9837 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
9839 msg = result.fail_msg
9841 for new_lv in new_lvs:
9842 msg2 = self.rpc.call_blockdev_remove(self.target_node,
9845 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
9846 hint=("cleanup manually the unused logical"
9848 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
9851 if self.early_release:
9852 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9854 self._RemoveOldStorage(self.target_node, iv_names)
9855 # WARNING: we release both node locks here, do not do other RPCs
9856 # than WaitForSync to the primary node
9857 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9858 names=[self.target_node, self.other_node])
9861 # This can fail as the old devices are degraded and _WaitForSync
9862 # does a combined result over all disks, so we don't check its return value
9863 self.lu.LogStep(cstep, steps_total, "Sync devices")
9865 _WaitForSync(self.lu, self.instance)
9867 # Check all devices manually
9868 self._CheckDevices(self.instance.primary_node, iv_names)
9870 # Step: remove old storage
9871 if not self.early_release:
9872 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9874 self._RemoveOldStorage(self.target_node, iv_names)
9876 def _ExecDrbd8Secondary(self, feedback_fn):
9877 """Replace the secondary node for DRBD 8.
9879 The algorithm for replace is quite complicated:
9880 - for all disks of the instance:
9881 - create new LVs on the new node with same names
9882 - shutdown the drbd device on the old secondary
9883 - disconnect the drbd network on the primary
9884 - create the drbd device on the new secondary
9885 - network attach the drbd on the primary, using an artifice:
9886 the drbd code for Attach() will connect to the network if it
9887 finds a device which is connected to the good local disks but
9889 - wait for sync across all devices
9890 - remove all disks from the old secondary
9892 Failures are not very well handled.
9897 pnode = self.instance.primary_node
9899 # Step: check device activation
9900 self.lu.LogStep(1, steps_total, "Check device existence")
9901 self._CheckDisksExistence([self.instance.primary_node])
9902 self._CheckVolumeGroup([self.instance.primary_node])
9904 # Step: check other node consistency
9905 self.lu.LogStep(2, steps_total, "Check peer consistency")
9906 self._CheckDisksConsistency(self.instance.primary_node, True, True)
9908 # Step: create new storage
9909 self.lu.LogStep(3, steps_total, "Allocate new storage")
9910 for idx, dev in enumerate(self.instance.disks):
9911 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
9912 (self.new_node, idx))
9913 # we pass force_create=True to force LVM creation
9914 for new_lv in dev.children:
9915 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
9916 _GetInstanceInfoText(self.instance), False)
9918 # Step 4: dbrd minors and drbd setups changes
9919 # after this, we must manually remove the drbd minors on both the
9920 # error and the success paths
9921 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9922 minors = self.cfg.AllocateDRBDMinor([self.new_node
9923 for dev in self.instance.disks],
9925 logging.debug("Allocated minors %r", minors)
9928 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
9929 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
9930 (self.new_node, idx))
9931 # create new devices on new_node; note that we create two IDs:
9932 # one without port, so the drbd will be activated without
9933 # networking information on the new node at this stage, and one
9934 # with network, for the latter activation in step 4
9935 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
9936 if self.instance.primary_node == o_node1:
9939 assert self.instance.primary_node == o_node2, "Three-node instance?"
9942 new_alone_id = (self.instance.primary_node, self.new_node, None,
9943 p_minor, new_minor, o_secret)
9944 new_net_id = (self.instance.primary_node, self.new_node, o_port,
9945 p_minor, new_minor, o_secret)
9947 iv_names[idx] = (dev, dev.children, new_net_id)
9948 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
9950 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
9951 logical_id=new_alone_id,
9952 children=dev.children,
9955 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
9956 _GetInstanceInfoText(self.instance), False)
9957 except errors.GenericError:
9958 self.cfg.ReleaseDRBDMinors(self.instance.name)
9961 # We have new devices, shutdown the drbd on the old secondary
9962 for idx, dev in enumerate(self.instance.disks):
9963 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
9964 self.cfg.SetDiskID(dev, self.target_node)
9965 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
9967 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
9968 "node: %s" % (idx, msg),
9969 hint=("Please cleanup this device manually as"
9970 " soon as possible"))
9972 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
9973 result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
9974 self.instance.disks)[pnode]
9976 msg = result.fail_msg
9978 # detaches didn't succeed (unlikely)
9979 self.cfg.ReleaseDRBDMinors(self.instance.name)
9980 raise errors.OpExecError("Can't detach the disks from the network on"
9981 " old node: %s" % (msg,))
9983 # if we managed to detach at least one, we update all the disks of
9984 # the instance to point to the new secondary
9985 self.lu.LogInfo("Updating instance configuration")
9986 for dev, _, new_logical_id in iv_names.itervalues():
9987 dev.logical_id = new_logical_id
9988 self.cfg.SetDiskID(dev, self.instance.primary_node)
9990 self.cfg.Update(self.instance, feedback_fn)
9992 # and now perform the drbd attach
9993 self.lu.LogInfo("Attaching primary drbds to new secondary"
9994 " (standalone => connected)")
9995 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
9997 self.node_secondary_ip,
9998 self.instance.disks,
10001 for to_node, to_result in result.items():
10002 msg = to_result.fail_msg
10004 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
10006 hint=("please do a gnt-instance info to see the"
10007 " status of disks"))
10009 if self.early_release:
10010 self.lu.LogStep(cstep, steps_total, "Removing old storage")
10012 self._RemoveOldStorage(self.target_node, iv_names)
10013 # WARNING: we release all node locks here, do not do other RPCs
10014 # than WaitForSync to the primary node
10015 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
10016 names=[self.instance.primary_node,
10021 # This can fail as the old devices are degraded and _WaitForSync
10022 # does a combined result over all disks, so we don't check its return value
10023 self.lu.LogStep(cstep, steps_total, "Sync devices")
10025 _WaitForSync(self.lu, self.instance)
10027 # Check all devices manually
10028 self._CheckDevices(self.instance.primary_node, iv_names)
10030 # Step: remove old storage
10031 if not self.early_release:
10032 self.lu.LogStep(cstep, steps_total, "Removing old storage")
10033 self._RemoveOldStorage(self.target_node, iv_names)
10036 class LURepairNodeStorage(NoHooksLU):
10037 """Repairs the volume group on a node.
10042 def CheckArguments(self):
10043 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
10045 storage_type = self.op.storage_type
10047 if (constants.SO_FIX_CONSISTENCY not in
10048 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
10049 raise errors.OpPrereqError("Storage units of type '%s' can not be"
10050 " repaired" % storage_type,
10051 errors.ECODE_INVAL)
10053 def ExpandNames(self):
10054 self.needed_locks = {
10055 locking.LEVEL_NODE: [self.op.node_name],
10058 def _CheckFaultyDisks(self, instance, node_name):
10059 """Ensure faulty disks abort the opcode or at least warn."""
10061 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
10063 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
10064 " node '%s'" % (instance.name, node_name),
10065 errors.ECODE_STATE)
10066 except errors.OpPrereqError, err:
10067 if self.op.ignore_consistency:
10068 self.proc.LogWarning(str(err.args[0]))
10072 def CheckPrereq(self):
10073 """Check prerequisites.
10076 # Check whether any instance on this node has faulty disks
10077 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
10078 if not inst.admin_up:
10080 check_nodes = set(inst.all_nodes)
10081 check_nodes.discard(self.op.node_name)
10082 for inst_node_name in check_nodes:
10083 self._CheckFaultyDisks(inst, inst_node_name)
10085 def Exec(self, feedback_fn):
10086 feedback_fn("Repairing storage unit '%s' on %s ..." %
10087 (self.op.name, self.op.node_name))
10089 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
10090 result = self.rpc.call_storage_execute(self.op.node_name,
10091 self.op.storage_type, st_args,
10093 constants.SO_FIX_CONSISTENCY)
10094 result.Raise("Failed to repair storage unit '%s' on %s" %
10095 (self.op.name, self.op.node_name))
10098 class LUNodeEvacuate(NoHooksLU):
10099 """Evacuates instances off a list of nodes.
10104 def CheckArguments(self):
10105 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
10107 def ExpandNames(self):
10108 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
10110 if self.op.remote_node is not None:
10111 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10112 assert self.op.remote_node
10114 if self.op.remote_node == self.op.node_name:
10115 raise errors.OpPrereqError("Can not use evacuated node as a new"
10116 " secondary node", errors.ECODE_INVAL)
10118 if self.op.mode != constants.IALLOCATOR_NEVAC_SEC:
10119 raise errors.OpPrereqError("Without the use of an iallocator only"
10120 " secondary instances can be evacuated",
10121 errors.ECODE_INVAL)
10124 self.share_locks = _ShareAll()
10125 self.needed_locks = {
10126 locking.LEVEL_INSTANCE: [],
10127 locking.LEVEL_NODEGROUP: [],
10128 locking.LEVEL_NODE: [],
10131 # Determine nodes (via group) optimistically, needs verification once locks
10132 # have been acquired
10133 self.lock_nodes = self._DetermineNodes()
10135 def _DetermineNodes(self):
10136 """Gets the list of nodes to operate on.
10139 if self.op.remote_node is None:
10140 # Iallocator will choose any node(s) in the same group
10141 group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name])
10143 group_nodes = frozenset([self.op.remote_node])
10145 # Determine nodes to be locked
10146 return set([self.op.node_name]) | group_nodes
10148 def _DetermineInstances(self):
10149 """Builds list of instances to operate on.
10152 assert self.op.mode in constants.IALLOCATOR_NEVAC_MODES
10154 if self.op.mode == constants.IALLOCATOR_NEVAC_PRI:
10155 # Primary instances only
10156 inst_fn = _GetNodePrimaryInstances
10157 assert self.op.remote_node is None, \
10158 "Evacuating primary instances requires iallocator"
10159 elif self.op.mode == constants.IALLOCATOR_NEVAC_SEC:
10160 # Secondary instances only
10161 inst_fn = _GetNodeSecondaryInstances
10164 assert self.op.mode == constants.IALLOCATOR_NEVAC_ALL
10165 inst_fn = _GetNodeInstances
10166 # TODO: In 2.6, change the iallocator interface to take an evacuation mode
10168 raise errors.OpPrereqError("Due to an issue with the iallocator"
10169 " interface it is not possible to evacuate"
10170 " all instances at once; specify explicitly"
10171 " whether to evacuate primary or secondary"
10173 errors.ECODE_INVAL)
10175 return inst_fn(self.cfg, self.op.node_name)
10177 def DeclareLocks(self, level):
10178 if level == locking.LEVEL_INSTANCE:
10179 # Lock instances optimistically, needs verification once node and group
10180 # locks have been acquired
10181 self.needed_locks[locking.LEVEL_INSTANCE] = \
10182 set(i.name for i in self._DetermineInstances())
10184 elif level == locking.LEVEL_NODEGROUP:
10185 # Lock node groups for all potential target nodes optimistically, needs
10186 # verification once nodes have been acquired
10187 self.needed_locks[locking.LEVEL_NODEGROUP] = \
10188 self.cfg.GetNodeGroupsFromNodes(self.lock_nodes)
10190 elif level == locking.LEVEL_NODE:
10191 self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes
10193 def CheckPrereq(self):
10195 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
10196 owned_nodes = self.owned_locks(locking.LEVEL_NODE)
10197 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
10199 need_nodes = self._DetermineNodes()
10201 if not owned_nodes.issuperset(need_nodes):
10202 raise errors.OpPrereqError("Nodes in same group as '%s' changed since"
10203 " locks were acquired, current nodes are"
10204 " are '%s', used to be '%s'; retry the"
10206 (self.op.node_name,
10207 utils.CommaJoin(need_nodes),
10208 utils.CommaJoin(owned_nodes)),
10209 errors.ECODE_STATE)
10211 wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
10212 if owned_groups != wanted_groups:
10213 raise errors.OpExecError("Node groups changed since locks were acquired,"
10214 " current groups are '%s', used to be '%s';"
10215 " retry the operation" %
10216 (utils.CommaJoin(wanted_groups),
10217 utils.CommaJoin(owned_groups)))
10219 # Determine affected instances
10220 self.instances = self._DetermineInstances()
10221 self.instance_names = [i.name for i in self.instances]
10223 if set(self.instance_names) != owned_instances:
10224 raise errors.OpExecError("Instances on node '%s' changed since locks"
10225 " were acquired, current instances are '%s',"
10226 " used to be '%s'; retry the operation" %
10227 (self.op.node_name,
10228 utils.CommaJoin(self.instance_names),
10229 utils.CommaJoin(owned_instances)))
10231 if self.instance_names:
10232 self.LogInfo("Evacuating instances from node '%s': %s",
10234 utils.CommaJoin(utils.NiceSort(self.instance_names)))
10236 self.LogInfo("No instances to evacuate from node '%s'",
10239 if self.op.remote_node is not None:
10240 for i in self.instances:
10241 if i.primary_node == self.op.remote_node:
10242 raise errors.OpPrereqError("Node %s is the primary node of"
10243 " instance %s, cannot use it as"
10245 (self.op.remote_node, i.name),
10246 errors.ECODE_INVAL)
10248 def Exec(self, feedback_fn):
10249 assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None)
10251 if not self.instance_names:
10252 # No instances to evacuate
10255 elif self.op.iallocator is not None:
10256 # TODO: Implement relocation to other group
10257 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_NODE_EVAC,
10258 evac_mode=self.op.mode,
10259 instances=list(self.instance_names))
10261 ial.Run(self.op.iallocator)
10263 if not ial.success:
10264 raise errors.OpPrereqError("Can't compute node evacuation using"
10265 " iallocator '%s': %s" %
10266 (self.op.iallocator, ial.info),
10267 errors.ECODE_NORES)
10269 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True)
10271 elif self.op.remote_node is not None:
10272 assert self.op.mode == constants.IALLOCATOR_NEVAC_SEC
10274 [opcodes.OpInstanceReplaceDisks(instance_name=instance_name,
10275 remote_node=self.op.remote_node,
10277 mode=constants.REPLACE_DISK_CHG,
10278 early_release=self.op.early_release)]
10279 for instance_name in self.instance_names
10283 raise errors.ProgrammerError("No iallocator or remote node")
10285 return ResultWithJobs(jobs)
10288 def _SetOpEarlyRelease(early_release, op):
10289 """Sets C{early_release} flag on opcodes if available.
10293 op.early_release = early_release
10294 except AttributeError:
10295 assert not isinstance(op, opcodes.OpInstanceReplaceDisks)
10300 def _NodeEvacDest(use_nodes, group, nodes):
10301 """Returns group or nodes depending on caller's choice.
10305 return utils.CommaJoin(nodes)
10310 def _LoadNodeEvacResult(lu, alloc_result, early_release, use_nodes):
10311 """Unpacks the result of change-group and node-evacuate iallocator requests.
10313 Iallocator modes L{constants.IALLOCATOR_MODE_NODE_EVAC} and
10314 L{constants.IALLOCATOR_MODE_CHG_GROUP}.
10316 @type lu: L{LogicalUnit}
10317 @param lu: Logical unit instance
10318 @type alloc_result: tuple/list
10319 @param alloc_result: Result from iallocator
10320 @type early_release: bool
10321 @param early_release: Whether to release locks early if possible
10322 @type use_nodes: bool
10323 @param use_nodes: Whether to display node names instead of groups
10326 (moved, failed, jobs) = alloc_result
10329 failreason = utils.CommaJoin("%s (%s)" % (name, reason)
10330 for (name, reason) in failed)
10331 lu.LogWarning("Unable to evacuate instances %s", failreason)
10332 raise errors.OpExecError("Unable to evacuate instances %s" % failreason)
10335 lu.LogInfo("Instances to be moved: %s",
10336 utils.CommaJoin("%s (to %s)" %
10337 (name, _NodeEvacDest(use_nodes, group, nodes))
10338 for (name, group, nodes) in moved))
10340 return [map(compat.partial(_SetOpEarlyRelease, early_release),
10341 map(opcodes.OpCode.LoadOpCode, ops))
10345 class LUInstanceGrowDisk(LogicalUnit):
10346 """Grow a disk of an instance.
10349 HPATH = "disk-grow"
10350 HTYPE = constants.HTYPE_INSTANCE
10353 def ExpandNames(self):
10354 self._ExpandAndLockInstance()
10355 self.needed_locks[locking.LEVEL_NODE] = []
10356 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10358 def DeclareLocks(self, level):
10359 if level == locking.LEVEL_NODE:
10360 self._LockInstancesNodes()
10362 def BuildHooksEnv(self):
10363 """Build hooks env.
10365 This runs on the master, the primary and all the secondaries.
10369 "DISK": self.op.disk,
10370 "AMOUNT": self.op.amount,
10372 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
10375 def BuildHooksNodes(self):
10376 """Build hooks nodes.
10379 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
10382 def CheckPrereq(self):
10383 """Check prerequisites.
10385 This checks that the instance is in the cluster.
10388 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10389 assert instance is not None, \
10390 "Cannot retrieve locked instance %s" % self.op.instance_name
10391 nodenames = list(instance.all_nodes)
10392 for node in nodenames:
10393 _CheckNodeOnline(self, node)
10395 self.instance = instance
10397 if instance.disk_template not in constants.DTS_GROWABLE:
10398 raise errors.OpPrereqError("Instance's disk layout does not support"
10399 " growing", errors.ECODE_INVAL)
10401 self.disk = instance.FindDisk(self.op.disk)
10403 if instance.disk_template not in (constants.DT_FILE,
10404 constants.DT_SHARED_FILE):
10405 # TODO: check the free disk space for file, when that feature will be
10407 _CheckNodesFreeDiskPerVG(self, nodenames,
10408 self.disk.ComputeGrowth(self.op.amount))
10410 def Exec(self, feedback_fn):
10411 """Execute disk grow.
10414 instance = self.instance
10417 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
10419 raise errors.OpExecError("Cannot activate block device to grow")
10421 # First run all grow ops in dry-run mode
10422 for node in instance.all_nodes:
10423 self.cfg.SetDiskID(disk, node)
10424 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, True)
10425 result.Raise("Grow request failed to node %s" % node)
10427 # We know that (as far as we can test) operations across different
10428 # nodes will succeed, time to run it for real
10429 for node in instance.all_nodes:
10430 self.cfg.SetDiskID(disk, node)
10431 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount, False)
10432 result.Raise("Grow request failed to node %s" % node)
10434 # TODO: Rewrite code to work properly
10435 # DRBD goes into sync mode for a short amount of time after executing the
10436 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
10437 # calling "resize" in sync mode fails. Sleeping for a short amount of
10438 # time is a work-around.
10441 disk.RecordGrow(self.op.amount)
10442 self.cfg.Update(instance, feedback_fn)
10443 if self.op.wait_for_sync:
10444 disk_abort = not _WaitForSync(self, instance, disks=[disk])
10446 self.proc.LogWarning("Disk sync-ing has not returned a good"
10447 " status; please check the instance")
10448 if not instance.admin_up:
10449 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
10450 elif not instance.admin_up:
10451 self.proc.LogWarning("Not shutting down the disk even if the instance is"
10452 " not supposed to be running because no wait for"
10453 " sync mode was requested")
10456 class LUInstanceQueryData(NoHooksLU):
10457 """Query runtime instance data.
10462 def ExpandNames(self):
10463 self.needed_locks = {}
10465 # Use locking if requested or when non-static information is wanted
10466 if not (self.op.static or self.op.use_locking):
10467 self.LogWarning("Non-static data requested, locks need to be acquired")
10468 self.op.use_locking = True
10470 if self.op.instances or not self.op.use_locking:
10471 # Expand instance names right here
10472 self.wanted_names = _GetWantedInstances(self, self.op.instances)
10474 # Will use acquired locks
10475 self.wanted_names = None
10477 if self.op.use_locking:
10478 self.share_locks = _ShareAll()
10480 if self.wanted_names is None:
10481 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
10483 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
10485 self.needed_locks[locking.LEVEL_NODE] = []
10486 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10488 def DeclareLocks(self, level):
10489 if self.op.use_locking and level == locking.LEVEL_NODE:
10490 self._LockInstancesNodes()
10492 def CheckPrereq(self):
10493 """Check prerequisites.
10495 This only checks the optional instance list against the existing names.
10498 if self.wanted_names is None:
10499 assert self.op.use_locking, "Locking was not used"
10500 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
10502 self.wanted_instances = \
10503 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
10505 def _ComputeBlockdevStatus(self, node, instance_name, dev):
10506 """Returns the status of a block device
10509 if self.op.static or not node:
10512 self.cfg.SetDiskID(dev, node)
10514 result = self.rpc.call_blockdev_find(node, dev)
10518 result.Raise("Can't compute disk status for %s" % instance_name)
10520 status = result.payload
10524 return (status.dev_path, status.major, status.minor,
10525 status.sync_percent, status.estimated_time,
10526 status.is_degraded, status.ldisk_status)
10528 def _ComputeDiskStatus(self, instance, snode, dev):
10529 """Compute block device status.
10532 if dev.dev_type in constants.LDS_DRBD:
10533 # we change the snode then (otherwise we use the one passed in)
10534 if dev.logical_id[0] == instance.primary_node:
10535 snode = dev.logical_id[1]
10537 snode = dev.logical_id[0]
10539 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
10540 instance.name, dev)
10541 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
10544 dev_children = map(compat.partial(self._ComputeDiskStatus,
10551 "iv_name": dev.iv_name,
10552 "dev_type": dev.dev_type,
10553 "logical_id": dev.logical_id,
10554 "physical_id": dev.physical_id,
10555 "pstatus": dev_pstatus,
10556 "sstatus": dev_sstatus,
10557 "children": dev_children,
10562 def Exec(self, feedback_fn):
10563 """Gather and return data"""
10566 cluster = self.cfg.GetClusterInfo()
10568 pri_nodes = self.cfg.GetMultiNodeInfo(i.primary_node
10569 for i in self.wanted_instances)
10570 for instance, (_, pnode) in zip(self.wanted_instances, pri_nodes):
10571 if self.op.static or pnode.offline:
10572 remote_state = None
10574 self.LogWarning("Primary node %s is marked offline, returning static"
10575 " information only for instance %s" %
10576 (pnode.name, instance.name))
10578 remote_info = self.rpc.call_instance_info(instance.primary_node,
10580 instance.hypervisor)
10581 remote_info.Raise("Error checking node %s" % instance.primary_node)
10582 remote_info = remote_info.payload
10583 if remote_info and "state" in remote_info:
10584 remote_state = "up"
10586 remote_state = "down"
10588 if instance.admin_up:
10589 config_state = "up"
10591 config_state = "down"
10593 disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
10596 result[instance.name] = {
10597 "name": instance.name,
10598 "config_state": config_state,
10599 "run_state": remote_state,
10600 "pnode": instance.primary_node,
10601 "snodes": instance.secondary_nodes,
10603 # this happens to be the same format used for hooks
10604 "nics": _NICListToTuple(self, instance.nics),
10605 "disk_template": instance.disk_template,
10607 "hypervisor": instance.hypervisor,
10608 "network_port": instance.network_port,
10609 "hv_instance": instance.hvparams,
10610 "hv_actual": cluster.FillHV(instance, skip_globals=True),
10611 "be_instance": instance.beparams,
10612 "be_actual": cluster.FillBE(instance),
10613 "os_instance": instance.osparams,
10614 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
10615 "serial_no": instance.serial_no,
10616 "mtime": instance.mtime,
10617 "ctime": instance.ctime,
10618 "uuid": instance.uuid,
10624 class LUInstanceSetParams(LogicalUnit):
10625 """Modifies an instances's parameters.
10628 HPATH = "instance-modify"
10629 HTYPE = constants.HTYPE_INSTANCE
10632 def CheckArguments(self):
10633 if not (self.op.nics or self.op.disks or self.op.disk_template or
10634 self.op.hvparams or self.op.beparams or self.op.os_name):
10635 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
10637 if self.op.hvparams:
10638 _CheckGlobalHvParams(self.op.hvparams)
10642 for disk_op, disk_dict in self.op.disks:
10643 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
10644 if disk_op == constants.DDM_REMOVE:
10645 disk_addremove += 1
10647 elif disk_op == constants.DDM_ADD:
10648 disk_addremove += 1
10650 if not isinstance(disk_op, int):
10651 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
10652 if not isinstance(disk_dict, dict):
10653 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
10654 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
10656 if disk_op == constants.DDM_ADD:
10657 mode = disk_dict.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
10658 if mode not in constants.DISK_ACCESS_SET:
10659 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
10660 errors.ECODE_INVAL)
10661 size = disk_dict.get(constants.IDISK_SIZE, None)
10663 raise errors.OpPrereqError("Required disk parameter size missing",
10664 errors.ECODE_INVAL)
10667 except (TypeError, ValueError), err:
10668 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
10669 str(err), errors.ECODE_INVAL)
10670 disk_dict[constants.IDISK_SIZE] = size
10672 # modification of disk
10673 if constants.IDISK_SIZE in disk_dict:
10674 raise errors.OpPrereqError("Disk size change not possible, use"
10675 " grow-disk", errors.ECODE_INVAL)
10677 if disk_addremove > 1:
10678 raise errors.OpPrereqError("Only one disk add or remove operation"
10679 " supported at a time", errors.ECODE_INVAL)
10681 if self.op.disks and self.op.disk_template is not None:
10682 raise errors.OpPrereqError("Disk template conversion and other disk"
10683 " changes not supported at the same time",
10684 errors.ECODE_INVAL)
10686 if (self.op.disk_template and
10687 self.op.disk_template in constants.DTS_INT_MIRROR and
10688 self.op.remote_node is None):
10689 raise errors.OpPrereqError("Changing the disk template to a mirrored"
10690 " one requires specifying a secondary node",
10691 errors.ECODE_INVAL)
10695 for nic_op, nic_dict in self.op.nics:
10696 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
10697 if nic_op == constants.DDM_REMOVE:
10700 elif nic_op == constants.DDM_ADD:
10703 if not isinstance(nic_op, int):
10704 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
10705 if not isinstance(nic_dict, dict):
10706 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
10707 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
10709 # nic_dict should be a dict
10710 nic_ip = nic_dict.get(constants.INIC_IP, None)
10711 if nic_ip is not None:
10712 if nic_ip.lower() == constants.VALUE_NONE:
10713 nic_dict[constants.INIC_IP] = None
10715 if not netutils.IPAddress.IsValid(nic_ip):
10716 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
10717 errors.ECODE_INVAL)
10719 nic_bridge = nic_dict.get("bridge", None)
10720 nic_link = nic_dict.get(constants.INIC_LINK, None)
10721 if nic_bridge and nic_link:
10722 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
10723 " at the same time", errors.ECODE_INVAL)
10724 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
10725 nic_dict["bridge"] = None
10726 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
10727 nic_dict[constants.INIC_LINK] = None
10729 if nic_op == constants.DDM_ADD:
10730 nic_mac = nic_dict.get(constants.INIC_MAC, None)
10731 if nic_mac is None:
10732 nic_dict[constants.INIC_MAC] = constants.VALUE_AUTO
10734 if constants.INIC_MAC in nic_dict:
10735 nic_mac = nic_dict[constants.INIC_MAC]
10736 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
10737 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
10739 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
10740 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
10741 " modifying an existing nic",
10742 errors.ECODE_INVAL)
10744 if nic_addremove > 1:
10745 raise errors.OpPrereqError("Only one NIC add or remove operation"
10746 " supported at a time", errors.ECODE_INVAL)
10748 def ExpandNames(self):
10749 self._ExpandAndLockInstance()
10750 self.needed_locks[locking.LEVEL_NODE] = []
10751 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10753 def DeclareLocks(self, level):
10754 if level == locking.LEVEL_NODE:
10755 self._LockInstancesNodes()
10756 if self.op.disk_template and self.op.remote_node:
10757 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10758 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
10760 def BuildHooksEnv(self):
10761 """Build hooks env.
10763 This runs on the master, primary and secondaries.
10767 if constants.BE_MEMORY in self.be_new:
10768 args["memory"] = self.be_new[constants.BE_MEMORY]
10769 if constants.BE_VCPUS in self.be_new:
10770 args["vcpus"] = self.be_new[constants.BE_VCPUS]
10771 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
10772 # information at all.
10775 nic_override = dict(self.op.nics)
10776 for idx, nic in enumerate(self.instance.nics):
10777 if idx in nic_override:
10778 this_nic_override = nic_override[idx]
10780 this_nic_override = {}
10781 if constants.INIC_IP in this_nic_override:
10782 ip = this_nic_override[constants.INIC_IP]
10785 if constants.INIC_MAC in this_nic_override:
10786 mac = this_nic_override[constants.INIC_MAC]
10789 if idx in self.nic_pnew:
10790 nicparams = self.nic_pnew[idx]
10792 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
10793 mode = nicparams[constants.NIC_MODE]
10794 link = nicparams[constants.NIC_LINK]
10795 args["nics"].append((ip, mac, mode, link))
10796 if constants.DDM_ADD in nic_override:
10797 ip = nic_override[constants.DDM_ADD].get(constants.INIC_IP, None)
10798 mac = nic_override[constants.DDM_ADD][constants.INIC_MAC]
10799 nicparams = self.nic_pnew[constants.DDM_ADD]
10800 mode = nicparams[constants.NIC_MODE]
10801 link = nicparams[constants.NIC_LINK]
10802 args["nics"].append((ip, mac, mode, link))
10803 elif constants.DDM_REMOVE in nic_override:
10804 del args["nics"][-1]
10806 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
10807 if self.op.disk_template:
10808 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
10812 def BuildHooksNodes(self):
10813 """Build hooks nodes.
10816 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
10819 def CheckPrereq(self):
10820 """Check prerequisites.
10822 This only checks the instance list against the existing names.
10825 # checking the new params on the primary/secondary nodes
10827 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10828 cluster = self.cluster = self.cfg.GetClusterInfo()
10829 assert self.instance is not None, \
10830 "Cannot retrieve locked instance %s" % self.op.instance_name
10831 pnode = instance.primary_node
10832 nodelist = list(instance.all_nodes)
10835 if self.op.os_name and not self.op.force:
10836 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
10837 self.op.force_variant)
10838 instance_os = self.op.os_name
10840 instance_os = instance.os
10842 if self.op.disk_template:
10843 if instance.disk_template == self.op.disk_template:
10844 raise errors.OpPrereqError("Instance already has disk template %s" %
10845 instance.disk_template, errors.ECODE_INVAL)
10847 if (instance.disk_template,
10848 self.op.disk_template) not in self._DISK_CONVERSIONS:
10849 raise errors.OpPrereqError("Unsupported disk template conversion from"
10850 " %s to %s" % (instance.disk_template,
10851 self.op.disk_template),
10852 errors.ECODE_INVAL)
10853 _CheckInstanceDown(self, instance, "cannot change disk template")
10854 if self.op.disk_template in constants.DTS_INT_MIRROR:
10855 if self.op.remote_node == pnode:
10856 raise errors.OpPrereqError("Given new secondary node %s is the same"
10857 " as the primary node of the instance" %
10858 self.op.remote_node, errors.ECODE_STATE)
10859 _CheckNodeOnline(self, self.op.remote_node)
10860 _CheckNodeNotDrained(self, self.op.remote_node)
10861 # FIXME: here we assume that the old instance type is DT_PLAIN
10862 assert instance.disk_template == constants.DT_PLAIN
10863 disks = [{constants.IDISK_SIZE: d.size,
10864 constants.IDISK_VG: d.logical_id[0]}
10865 for d in instance.disks]
10866 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
10867 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
10869 # hvparams processing
10870 if self.op.hvparams:
10871 hv_type = instance.hypervisor
10872 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
10873 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
10874 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
10877 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
10878 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
10879 self.hv_new = hv_new # the new actual values
10880 self.hv_inst = i_hvdict # the new dict (without defaults)
10882 self.hv_new = self.hv_inst = {}
10884 # beparams processing
10885 if self.op.beparams:
10886 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
10888 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
10889 be_new = cluster.SimpleFillBE(i_bedict)
10890 self.be_new = be_new # the new actual values
10891 self.be_inst = i_bedict # the new dict (without defaults)
10893 self.be_new = self.be_inst = {}
10894 be_old = cluster.FillBE(instance)
10896 # osparams processing
10897 if self.op.osparams:
10898 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
10899 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
10900 self.os_inst = i_osdict # the new dict (without defaults)
10906 if (constants.BE_MEMORY in self.op.beparams and not self.op.force and
10907 be_new[constants.BE_MEMORY] > be_old[constants.BE_MEMORY]):
10908 mem_check_list = [pnode]
10909 if be_new[constants.BE_AUTO_BALANCE]:
10910 # either we changed auto_balance to yes or it was from before
10911 mem_check_list.extend(instance.secondary_nodes)
10912 instance_info = self.rpc.call_instance_info(pnode, instance.name,
10913 instance.hypervisor)
10914 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
10915 instance.hypervisor)
10916 pninfo = nodeinfo[pnode]
10917 msg = pninfo.fail_msg
10919 # Assume the primary node is unreachable and go ahead
10920 self.warn.append("Can't get info from primary node %s: %s" %
10922 elif not isinstance(pninfo.payload.get("memory_free", None), int):
10923 self.warn.append("Node data from primary node %s doesn't contain"
10924 " free memory information" % pnode)
10925 elif instance_info.fail_msg:
10926 self.warn.append("Can't get instance runtime information: %s" %
10927 instance_info.fail_msg)
10929 if instance_info.payload:
10930 current_mem = int(instance_info.payload["memory"])
10932 # Assume instance not running
10933 # (there is a slight race condition here, but it's not very probable,
10934 # and we have no other way to check)
10936 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
10937 pninfo.payload["memory_free"])
10939 raise errors.OpPrereqError("This change will prevent the instance"
10940 " from starting, due to %d MB of memory"
10941 " missing on its primary node" % miss_mem,
10942 errors.ECODE_NORES)
10944 if be_new[constants.BE_AUTO_BALANCE]:
10945 for node, nres in nodeinfo.items():
10946 if node not in instance.secondary_nodes:
10948 nres.Raise("Can't get info from secondary node %s" % node,
10949 prereq=True, ecode=errors.ECODE_STATE)
10950 if not isinstance(nres.payload.get("memory_free", None), int):
10951 raise errors.OpPrereqError("Secondary node %s didn't return free"
10952 " memory information" % node,
10953 errors.ECODE_STATE)
10954 elif be_new[constants.BE_MEMORY] > nres.payload["memory_free"]:
10955 raise errors.OpPrereqError("This change will prevent the instance"
10956 " from failover to its secondary node"
10957 " %s, due to not enough memory" % node,
10958 errors.ECODE_STATE)
10962 self.nic_pinst = {}
10963 for nic_op, nic_dict in self.op.nics:
10964 if nic_op == constants.DDM_REMOVE:
10965 if not instance.nics:
10966 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
10967 errors.ECODE_INVAL)
10969 if nic_op != constants.DDM_ADD:
10971 if not instance.nics:
10972 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
10973 " no NICs" % nic_op,
10974 errors.ECODE_INVAL)
10975 if nic_op < 0 or nic_op >= len(instance.nics):
10976 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
10978 (nic_op, len(instance.nics) - 1),
10979 errors.ECODE_INVAL)
10980 old_nic_params = instance.nics[nic_op].nicparams
10981 old_nic_ip = instance.nics[nic_op].ip
10983 old_nic_params = {}
10986 update_params_dict = dict([(key, nic_dict[key])
10987 for key in constants.NICS_PARAMETERS
10988 if key in nic_dict])
10990 if "bridge" in nic_dict:
10991 update_params_dict[constants.NIC_LINK] = nic_dict["bridge"]
10993 new_nic_params = _GetUpdatedParams(old_nic_params,
10994 update_params_dict)
10995 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
10996 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
10997 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
10998 self.nic_pinst[nic_op] = new_nic_params
10999 self.nic_pnew[nic_op] = new_filled_nic_params
11000 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
11002 if new_nic_mode == constants.NIC_MODE_BRIDGED:
11003 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
11004 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
11006 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
11008 self.warn.append(msg)
11010 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
11011 if new_nic_mode == constants.NIC_MODE_ROUTED:
11012 if constants.INIC_IP in nic_dict:
11013 nic_ip = nic_dict[constants.INIC_IP]
11015 nic_ip = old_nic_ip
11017 raise errors.OpPrereqError("Cannot set the nic ip to None"
11018 " on a routed nic", errors.ECODE_INVAL)
11019 if constants.INIC_MAC in nic_dict:
11020 nic_mac = nic_dict[constants.INIC_MAC]
11021 if nic_mac is None:
11022 raise errors.OpPrereqError("Cannot set the nic mac to None",
11023 errors.ECODE_INVAL)
11024 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
11025 # otherwise generate the mac
11026 nic_dict[constants.INIC_MAC] = \
11027 self.cfg.GenerateMAC(self.proc.GetECId())
11029 # or validate/reserve the current one
11031 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
11032 except errors.ReservationError:
11033 raise errors.OpPrereqError("MAC address %s already in use"
11034 " in cluster" % nic_mac,
11035 errors.ECODE_NOTUNIQUE)
11038 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
11039 raise errors.OpPrereqError("Disk operations not supported for"
11040 " diskless instances",
11041 errors.ECODE_INVAL)
11042 for disk_op, _ in self.op.disks:
11043 if disk_op == constants.DDM_REMOVE:
11044 if len(instance.disks) == 1:
11045 raise errors.OpPrereqError("Cannot remove the last disk of"
11046 " an instance", errors.ECODE_INVAL)
11047 _CheckInstanceDown(self, instance, "cannot remove disks")
11049 if (disk_op == constants.DDM_ADD and
11050 len(instance.disks) >= constants.MAX_DISKS):
11051 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
11052 " add more" % constants.MAX_DISKS,
11053 errors.ECODE_STATE)
11054 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
11056 if disk_op < 0 or disk_op >= len(instance.disks):
11057 raise errors.OpPrereqError("Invalid disk index %s, valid values"
11059 (disk_op, len(instance.disks)),
11060 errors.ECODE_INVAL)
11064 def _ConvertPlainToDrbd(self, feedback_fn):
11065 """Converts an instance from plain to drbd.
11068 feedback_fn("Converting template to drbd")
11069 instance = self.instance
11070 pnode = instance.primary_node
11071 snode = self.op.remote_node
11073 # create a fake disk info for _GenerateDiskTemplate
11074 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
11075 constants.IDISK_VG: d.logical_id[0]}
11076 for d in instance.disks]
11077 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
11078 instance.name, pnode, [snode],
11079 disk_info, None, None, 0, feedback_fn)
11080 info = _GetInstanceInfoText(instance)
11081 feedback_fn("Creating aditional volumes...")
11082 # first, create the missing data and meta devices
11083 for disk in new_disks:
11084 # unfortunately this is... not too nice
11085 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
11087 for child in disk.children:
11088 _CreateSingleBlockDev(self, snode, instance, child, info, True)
11089 # at this stage, all new LVs have been created, we can rename the
11091 feedback_fn("Renaming original volumes...")
11092 rename_list = [(o, n.children[0].logical_id)
11093 for (o, n) in zip(instance.disks, new_disks)]
11094 result = self.rpc.call_blockdev_rename(pnode, rename_list)
11095 result.Raise("Failed to rename original LVs")
11097 feedback_fn("Initializing DRBD devices...")
11098 # all child devices are in place, we can now create the DRBD devices
11099 for disk in new_disks:
11100 for node in [pnode, snode]:
11101 f_create = node == pnode
11102 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
11104 # at this point, the instance has been modified
11105 instance.disk_template = constants.DT_DRBD8
11106 instance.disks = new_disks
11107 self.cfg.Update(instance, feedback_fn)
11109 # disks are created, waiting for sync
11110 disk_abort = not _WaitForSync(self, instance,
11111 oneshot=not self.op.wait_for_sync)
11113 raise errors.OpExecError("There are some degraded disks for"
11114 " this instance, please cleanup manually")
11116 def _ConvertDrbdToPlain(self, feedback_fn):
11117 """Converts an instance from drbd to plain.
11120 instance = self.instance
11121 assert len(instance.secondary_nodes) == 1
11122 pnode = instance.primary_node
11123 snode = instance.secondary_nodes[0]
11124 feedback_fn("Converting template to plain")
11126 old_disks = instance.disks
11127 new_disks = [d.children[0] for d in old_disks]
11129 # copy over size and mode
11130 for parent, child in zip(old_disks, new_disks):
11131 child.size = parent.size
11132 child.mode = parent.mode
11134 # update instance structure
11135 instance.disks = new_disks
11136 instance.disk_template = constants.DT_PLAIN
11137 self.cfg.Update(instance, feedback_fn)
11139 feedback_fn("Removing volumes on the secondary node...")
11140 for disk in old_disks:
11141 self.cfg.SetDiskID(disk, snode)
11142 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
11144 self.LogWarning("Could not remove block device %s on node %s,"
11145 " continuing anyway: %s", disk.iv_name, snode, msg)
11147 feedback_fn("Removing unneeded volumes on the primary node...")
11148 for idx, disk in enumerate(old_disks):
11149 meta = disk.children[1]
11150 self.cfg.SetDiskID(meta, pnode)
11151 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
11153 self.LogWarning("Could not remove metadata for disk %d on node %s,"
11154 " continuing anyway: %s", idx, pnode, msg)
11156 # this is a DRBD disk, return its port to the pool
11157 for disk in old_disks:
11158 tcp_port = disk.logical_id[2]
11159 self.cfg.AddTcpUdpPort(tcp_port)
11161 def Exec(self, feedback_fn):
11162 """Modifies an instance.
11164 All parameters take effect only at the next restart of the instance.
11167 # Process here the warnings from CheckPrereq, as we don't have a
11168 # feedback_fn there.
11169 for warn in self.warn:
11170 feedback_fn("WARNING: %s" % warn)
11173 instance = self.instance
11175 for disk_op, disk_dict in self.op.disks:
11176 if disk_op == constants.DDM_REMOVE:
11177 # remove the last disk
11178 device = instance.disks.pop()
11179 device_idx = len(instance.disks)
11180 for node, disk in device.ComputeNodeTree(instance.primary_node):
11181 self.cfg.SetDiskID(disk, node)
11182 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
11184 self.LogWarning("Could not remove disk/%d on node %s: %s,"
11185 " continuing anyway", device_idx, node, msg)
11186 result.append(("disk/%d" % device_idx, "remove"))
11188 # if this is a DRBD disk, return its port to the pool
11189 if device.dev_type in constants.LDS_DRBD:
11190 tcp_port = device.logical_id[2]
11191 self.cfg.AddTcpUdpPort(tcp_port)
11192 elif disk_op == constants.DDM_ADD:
11194 if instance.disk_template in (constants.DT_FILE,
11195 constants.DT_SHARED_FILE):
11196 file_driver, file_path = instance.disks[0].logical_id
11197 file_path = os.path.dirname(file_path)
11199 file_driver = file_path = None
11200 disk_idx_base = len(instance.disks)
11201 new_disk = _GenerateDiskTemplate(self,
11202 instance.disk_template,
11203 instance.name, instance.primary_node,
11204 instance.secondary_nodes,
11208 disk_idx_base, feedback_fn)[0]
11209 instance.disks.append(new_disk)
11210 info = _GetInstanceInfoText(instance)
11212 logging.info("Creating volume %s for instance %s",
11213 new_disk.iv_name, instance.name)
11214 # Note: this needs to be kept in sync with _CreateDisks
11216 for node in instance.all_nodes:
11217 f_create = node == instance.primary_node
11219 _CreateBlockDev(self, node, instance, new_disk,
11220 f_create, info, f_create)
11221 except errors.OpExecError, err:
11222 self.LogWarning("Failed to create volume %s (%s) on"
11224 new_disk.iv_name, new_disk, node, err)
11225 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
11226 (new_disk.size, new_disk.mode)))
11228 # change a given disk
11229 instance.disks[disk_op].mode = disk_dict[constants.IDISK_MODE]
11230 result.append(("disk.mode/%d" % disk_op,
11231 disk_dict[constants.IDISK_MODE]))
11233 if self.op.disk_template:
11234 r_shut = _ShutdownInstanceDisks(self, instance)
11236 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
11237 " proceed with disk template conversion")
11238 mode = (instance.disk_template, self.op.disk_template)
11240 self._DISK_CONVERSIONS[mode](self, feedback_fn)
11242 self.cfg.ReleaseDRBDMinors(instance.name)
11244 result.append(("disk_template", self.op.disk_template))
11247 for nic_op, nic_dict in self.op.nics:
11248 if nic_op == constants.DDM_REMOVE:
11249 # remove the last nic
11250 del instance.nics[-1]
11251 result.append(("nic.%d" % len(instance.nics), "remove"))
11252 elif nic_op == constants.DDM_ADD:
11253 # mac and bridge should be set, by now
11254 mac = nic_dict[constants.INIC_MAC]
11255 ip = nic_dict.get(constants.INIC_IP, None)
11256 nicparams = self.nic_pinst[constants.DDM_ADD]
11257 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
11258 instance.nics.append(new_nic)
11259 result.append(("nic.%d" % (len(instance.nics) - 1),
11260 "add:mac=%s,ip=%s,mode=%s,link=%s" %
11261 (new_nic.mac, new_nic.ip,
11262 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
11263 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
11266 for key in (constants.INIC_MAC, constants.INIC_IP):
11267 if key in nic_dict:
11268 setattr(instance.nics[nic_op], key, nic_dict[key])
11269 if nic_op in self.nic_pinst:
11270 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
11271 for key, val in nic_dict.iteritems():
11272 result.append(("nic.%s/%d" % (key, nic_op), val))
11275 if self.op.hvparams:
11276 instance.hvparams = self.hv_inst
11277 for key, val in self.op.hvparams.iteritems():
11278 result.append(("hv/%s" % key, val))
11281 if self.op.beparams:
11282 instance.beparams = self.be_inst
11283 for key, val in self.op.beparams.iteritems():
11284 result.append(("be/%s" % key, val))
11287 if self.op.os_name:
11288 instance.os = self.op.os_name
11291 if self.op.osparams:
11292 instance.osparams = self.os_inst
11293 for key, val in self.op.osparams.iteritems():
11294 result.append(("os/%s" % key, val))
11296 self.cfg.Update(instance, feedback_fn)
11300 _DISK_CONVERSIONS = {
11301 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
11302 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
11306 class LUInstanceChangeGroup(LogicalUnit):
11307 HPATH = "instance-change-group"
11308 HTYPE = constants.HTYPE_INSTANCE
11311 def ExpandNames(self):
11312 self.share_locks = _ShareAll()
11313 self.needed_locks = {
11314 locking.LEVEL_NODEGROUP: [],
11315 locking.LEVEL_NODE: [],
11318 self._ExpandAndLockInstance()
11320 if self.op.target_groups:
11321 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
11322 self.op.target_groups)
11324 self.req_target_uuids = None
11326 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
11328 def DeclareLocks(self, level):
11329 if level == locking.LEVEL_NODEGROUP:
11330 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
11332 if self.req_target_uuids:
11333 lock_groups = set(self.req_target_uuids)
11335 # Lock all groups used by instance optimistically; this requires going
11336 # via the node before it's locked, requiring verification later on
11337 instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
11338 lock_groups.update(instance_groups)
11340 # No target groups, need to lock all of them
11341 lock_groups = locking.ALL_SET
11343 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
11345 elif level == locking.LEVEL_NODE:
11346 if self.req_target_uuids:
11347 # Lock all nodes used by instances
11348 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
11349 self._LockInstancesNodes()
11351 # Lock all nodes in all potential target groups
11352 lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) -
11353 self.cfg.GetInstanceNodeGroups(self.op.instance_name))
11354 member_nodes = [node_name
11355 for group in lock_groups
11356 for node_name in self.cfg.GetNodeGroup(group).members]
11357 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
11359 # Lock all nodes as all groups are potential targets
11360 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11362 def CheckPrereq(self):
11363 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
11364 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
11365 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
11367 assert (self.req_target_uuids is None or
11368 owned_groups.issuperset(self.req_target_uuids))
11369 assert owned_instances == set([self.op.instance_name])
11371 # Get instance information
11372 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
11374 # Check if node groups for locked instance are still correct
11375 assert owned_nodes.issuperset(self.instance.all_nodes), \
11376 ("Instance %s's nodes changed while we kept the lock" %
11377 self.op.instance_name)
11379 inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name,
11382 if self.req_target_uuids:
11383 # User requested specific target groups
11384 self.target_uuids = frozenset(self.req_target_uuids)
11386 # All groups except those used by the instance are potential targets
11387 self.target_uuids = owned_groups - inst_groups
11389 conflicting_groups = self.target_uuids & inst_groups
11390 if conflicting_groups:
11391 raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are"
11392 " used by the instance '%s'" %
11393 (utils.CommaJoin(conflicting_groups),
11394 self.op.instance_name),
11395 errors.ECODE_INVAL)
11397 if not self.target_uuids:
11398 raise errors.OpPrereqError("There are no possible target groups",
11399 errors.ECODE_INVAL)
11401 def BuildHooksEnv(self):
11402 """Build hooks env.
11405 assert self.target_uuids
11408 "TARGET_GROUPS": " ".join(self.target_uuids),
11411 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11415 def BuildHooksNodes(self):
11416 """Build hooks nodes.
11419 mn = self.cfg.GetMasterNode()
11420 return ([mn], [mn])
11422 def Exec(self, feedback_fn):
11423 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
11425 assert instances == [self.op.instance_name], "Instance not locked"
11427 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
11428 instances=instances, target_groups=list(self.target_uuids))
11430 ial.Run(self.op.iallocator)
11432 if not ial.success:
11433 raise errors.OpPrereqError("Can't compute solution for changing group of"
11434 " instance '%s' using iallocator '%s': %s" %
11435 (self.op.instance_name, self.op.iallocator,
11437 errors.ECODE_NORES)
11439 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
11441 self.LogInfo("Iallocator returned %s job(s) for changing group of"
11442 " instance '%s'", len(jobs), self.op.instance_name)
11444 return ResultWithJobs(jobs)
11447 class LUBackupQuery(NoHooksLU):
11448 """Query the exports list
11453 def ExpandNames(self):
11454 self.needed_locks = {}
11455 self.share_locks[locking.LEVEL_NODE] = 1
11456 if not self.op.nodes:
11457 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11459 self.needed_locks[locking.LEVEL_NODE] = \
11460 _GetWantedNodes(self, self.op.nodes)
11462 def Exec(self, feedback_fn):
11463 """Compute the list of all the exported system images.
11466 @return: a dictionary with the structure node->(export-list)
11467 where export-list is a list of the instances exported on
11471 self.nodes = self.owned_locks(locking.LEVEL_NODE)
11472 rpcresult = self.rpc.call_export_list(self.nodes)
11474 for node in rpcresult:
11475 if rpcresult[node].fail_msg:
11476 result[node] = False
11478 result[node] = rpcresult[node].payload
11483 class LUBackupPrepare(NoHooksLU):
11484 """Prepares an instance for an export and returns useful information.
11489 def ExpandNames(self):
11490 self._ExpandAndLockInstance()
11492 def CheckPrereq(self):
11493 """Check prerequisites.
11496 instance_name = self.op.instance_name
11498 self.instance = self.cfg.GetInstanceInfo(instance_name)
11499 assert self.instance is not None, \
11500 "Cannot retrieve locked instance %s" % self.op.instance_name
11501 _CheckNodeOnline(self, self.instance.primary_node)
11503 self._cds = _GetClusterDomainSecret()
11505 def Exec(self, feedback_fn):
11506 """Prepares an instance for an export.
11509 instance = self.instance
11511 if self.op.mode == constants.EXPORT_MODE_REMOTE:
11512 salt = utils.GenerateSecret(8)
11514 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
11515 result = self.rpc.call_x509_cert_create(instance.primary_node,
11516 constants.RIE_CERT_VALIDITY)
11517 result.Raise("Can't create X509 key and certificate on %s" % result.node)
11519 (name, cert_pem) = result.payload
11521 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
11525 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
11526 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
11528 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
11534 class LUBackupExport(LogicalUnit):
11535 """Export an instance to an image in the cluster.
11538 HPATH = "instance-export"
11539 HTYPE = constants.HTYPE_INSTANCE
11542 def CheckArguments(self):
11543 """Check the arguments.
11546 self.x509_key_name = self.op.x509_key_name
11547 self.dest_x509_ca_pem = self.op.destination_x509_ca
11549 if self.op.mode == constants.EXPORT_MODE_REMOTE:
11550 if not self.x509_key_name:
11551 raise errors.OpPrereqError("Missing X509 key name for encryption",
11552 errors.ECODE_INVAL)
11554 if not self.dest_x509_ca_pem:
11555 raise errors.OpPrereqError("Missing destination X509 CA",
11556 errors.ECODE_INVAL)
11558 def ExpandNames(self):
11559 self._ExpandAndLockInstance()
11561 # Lock all nodes for local exports
11562 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11563 # FIXME: lock only instance primary and destination node
11565 # Sad but true, for now we have do lock all nodes, as we don't know where
11566 # the previous export might be, and in this LU we search for it and
11567 # remove it from its current node. In the future we could fix this by:
11568 # - making a tasklet to search (share-lock all), then create the
11569 # new one, then one to remove, after
11570 # - removing the removal operation altogether
11571 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11573 def DeclareLocks(self, level):
11574 """Last minute lock declaration."""
11575 # All nodes are locked anyway, so nothing to do here.
11577 def BuildHooksEnv(self):
11578 """Build hooks env.
11580 This will run on the master, primary node and target node.
11584 "EXPORT_MODE": self.op.mode,
11585 "EXPORT_NODE": self.op.target_node,
11586 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
11587 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
11588 # TODO: Generic function for boolean env variables
11589 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
11592 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11596 def BuildHooksNodes(self):
11597 """Build hooks nodes.
11600 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
11602 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11603 nl.append(self.op.target_node)
11607 def CheckPrereq(self):
11608 """Check prerequisites.
11610 This checks that the instance and node names are valid.
11613 instance_name = self.op.instance_name
11615 self.instance = self.cfg.GetInstanceInfo(instance_name)
11616 assert self.instance is not None, \
11617 "Cannot retrieve locked instance %s" % self.op.instance_name
11618 _CheckNodeOnline(self, self.instance.primary_node)
11620 if (self.op.remove_instance and self.instance.admin_up and
11621 not self.op.shutdown):
11622 raise errors.OpPrereqError("Can not remove instance without shutting it"
11625 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11626 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
11627 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
11628 assert self.dst_node is not None
11630 _CheckNodeOnline(self, self.dst_node.name)
11631 _CheckNodeNotDrained(self, self.dst_node.name)
11634 self.dest_disk_info = None
11635 self.dest_x509_ca = None
11637 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
11638 self.dst_node = None
11640 if len(self.op.target_node) != len(self.instance.disks):
11641 raise errors.OpPrereqError(("Received destination information for %s"
11642 " disks, but instance %s has %s disks") %
11643 (len(self.op.target_node), instance_name,
11644 len(self.instance.disks)),
11645 errors.ECODE_INVAL)
11647 cds = _GetClusterDomainSecret()
11649 # Check X509 key name
11651 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
11652 except (TypeError, ValueError), err:
11653 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
11655 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
11656 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
11657 errors.ECODE_INVAL)
11659 # Load and verify CA
11661 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
11662 except OpenSSL.crypto.Error, err:
11663 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
11664 (err, ), errors.ECODE_INVAL)
11666 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
11667 if errcode is not None:
11668 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
11669 (msg, ), errors.ECODE_INVAL)
11671 self.dest_x509_ca = cert
11673 # Verify target information
11675 for idx, disk_data in enumerate(self.op.target_node):
11677 (host, port, magic) = \
11678 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
11679 except errors.GenericError, err:
11680 raise errors.OpPrereqError("Target info for disk %s: %s" %
11681 (idx, err), errors.ECODE_INVAL)
11683 disk_info.append((host, port, magic))
11685 assert len(disk_info) == len(self.op.target_node)
11686 self.dest_disk_info = disk_info
11689 raise errors.ProgrammerError("Unhandled export mode %r" %
11692 # instance disk type verification
11693 # TODO: Implement export support for file-based disks
11694 for disk in self.instance.disks:
11695 if disk.dev_type == constants.LD_FILE:
11696 raise errors.OpPrereqError("Export not supported for instances with"
11697 " file-based disks", errors.ECODE_INVAL)
11699 def _CleanupExports(self, feedback_fn):
11700 """Removes exports of current instance from all other nodes.
11702 If an instance in a cluster with nodes A..D was exported to node C, its
11703 exports will be removed from the nodes A, B and D.
11706 assert self.op.mode != constants.EXPORT_MODE_REMOTE
11708 nodelist = self.cfg.GetNodeList()
11709 nodelist.remove(self.dst_node.name)
11711 # on one-node clusters nodelist will be empty after the removal
11712 # if we proceed the backup would be removed because OpBackupQuery
11713 # substitutes an empty list with the full cluster node list.
11714 iname = self.instance.name
11716 feedback_fn("Removing old exports for instance %s" % iname)
11717 exportlist = self.rpc.call_export_list(nodelist)
11718 for node in exportlist:
11719 if exportlist[node].fail_msg:
11721 if iname in exportlist[node].payload:
11722 msg = self.rpc.call_export_remove(node, iname).fail_msg
11724 self.LogWarning("Could not remove older export for instance %s"
11725 " on node %s: %s", iname, node, msg)
11727 def Exec(self, feedback_fn):
11728 """Export an instance to an image in the cluster.
11731 assert self.op.mode in constants.EXPORT_MODES
11733 instance = self.instance
11734 src_node = instance.primary_node
11736 if self.op.shutdown:
11737 # shutdown the instance, but not the disks
11738 feedback_fn("Shutting down instance %s" % instance.name)
11739 result = self.rpc.call_instance_shutdown(src_node, instance,
11740 self.op.shutdown_timeout)
11741 # TODO: Maybe ignore failures if ignore_remove_failures is set
11742 result.Raise("Could not shutdown instance %s on"
11743 " node %s" % (instance.name, src_node))
11745 # set the disks ID correctly since call_instance_start needs the
11746 # correct drbd minor to create the symlinks
11747 for disk in instance.disks:
11748 self.cfg.SetDiskID(disk, src_node)
11750 activate_disks = (not instance.admin_up)
11753 # Activate the instance disks if we'exporting a stopped instance
11754 feedback_fn("Activating disks for %s" % instance.name)
11755 _StartInstanceDisks(self, instance, None)
11758 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
11761 helper.CreateSnapshots()
11763 if (self.op.shutdown and instance.admin_up and
11764 not self.op.remove_instance):
11765 assert not activate_disks
11766 feedback_fn("Starting instance %s" % instance.name)
11767 result = self.rpc.call_instance_start(src_node, instance,
11769 msg = result.fail_msg
11771 feedback_fn("Failed to start instance: %s" % msg)
11772 _ShutdownInstanceDisks(self, instance)
11773 raise errors.OpExecError("Could not start instance: %s" % msg)
11775 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11776 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
11777 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
11778 connect_timeout = constants.RIE_CONNECT_TIMEOUT
11779 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
11781 (key_name, _, _) = self.x509_key_name
11784 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
11787 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
11788 key_name, dest_ca_pem,
11793 # Check for backwards compatibility
11794 assert len(dresults) == len(instance.disks)
11795 assert compat.all(isinstance(i, bool) for i in dresults), \
11796 "Not all results are boolean: %r" % dresults
11800 feedback_fn("Deactivating disks for %s" % instance.name)
11801 _ShutdownInstanceDisks(self, instance)
11803 if not (compat.all(dresults) and fin_resu):
11806 failures.append("export finalization")
11807 if not compat.all(dresults):
11808 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
11810 failures.append("disk export: disk(s) %s" % fdsk)
11812 raise errors.OpExecError("Export failed, errors in %s" %
11813 utils.CommaJoin(failures))
11815 # At this point, the export was successful, we can cleanup/finish
11817 # Remove instance if requested
11818 if self.op.remove_instance:
11819 feedback_fn("Removing instance %s" % instance.name)
11820 _RemoveInstance(self, feedback_fn, instance,
11821 self.op.ignore_remove_failures)
11823 if self.op.mode == constants.EXPORT_MODE_LOCAL:
11824 self._CleanupExports(feedback_fn)
11826 return fin_resu, dresults
11829 class LUBackupRemove(NoHooksLU):
11830 """Remove exports related to the named instance.
11835 def ExpandNames(self):
11836 self.needed_locks = {}
11837 # We need all nodes to be locked in order for RemoveExport to work, but we
11838 # don't need to lock the instance itself, as nothing will happen to it (and
11839 # we can remove exports also for a removed instance)
11840 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
11842 def Exec(self, feedback_fn):
11843 """Remove any export.
11846 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
11847 # If the instance was not found we'll try with the name that was passed in.
11848 # This will only work if it was an FQDN, though.
11850 if not instance_name:
11852 instance_name = self.op.instance_name
11854 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
11855 exportlist = self.rpc.call_export_list(locked_nodes)
11857 for node in exportlist:
11858 msg = exportlist[node].fail_msg
11860 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
11862 if instance_name in exportlist[node].payload:
11864 result = self.rpc.call_export_remove(node, instance_name)
11865 msg = result.fail_msg
11867 logging.error("Could not remove export for instance %s"
11868 " on node %s: %s", instance_name, node, msg)
11870 if fqdn_warn and not found:
11871 feedback_fn("Export not found. If trying to remove an export belonging"
11872 " to a deleted instance please use its Fully Qualified"
11876 class LUGroupAdd(LogicalUnit):
11877 """Logical unit for creating node groups.
11880 HPATH = "group-add"
11881 HTYPE = constants.HTYPE_GROUP
11884 def ExpandNames(self):
11885 # We need the new group's UUID here so that we can create and acquire the
11886 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
11887 # that it should not check whether the UUID exists in the configuration.
11888 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
11889 self.needed_locks = {}
11890 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
11892 def CheckPrereq(self):
11893 """Check prerequisites.
11895 This checks that the given group name is not an existing node group
11900 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11901 except errors.OpPrereqError:
11904 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
11905 " node group (UUID: %s)" %
11906 (self.op.group_name, existing_uuid),
11907 errors.ECODE_EXISTS)
11909 if self.op.ndparams:
11910 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
11912 def BuildHooksEnv(self):
11913 """Build hooks env.
11917 "GROUP_NAME": self.op.group_name,
11920 def BuildHooksNodes(self):
11921 """Build hooks nodes.
11924 mn = self.cfg.GetMasterNode()
11925 return ([mn], [mn])
11927 def Exec(self, feedback_fn):
11928 """Add the node group to the cluster.
11931 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
11932 uuid=self.group_uuid,
11933 alloc_policy=self.op.alloc_policy,
11934 ndparams=self.op.ndparams)
11936 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
11937 del self.remove_locks[locking.LEVEL_NODEGROUP]
11940 class LUGroupAssignNodes(NoHooksLU):
11941 """Logical unit for assigning nodes to groups.
11946 def ExpandNames(self):
11947 # These raise errors.OpPrereqError on their own:
11948 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11949 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
11951 # We want to lock all the affected nodes and groups. We have readily
11952 # available the list of nodes, and the *destination* group. To gather the
11953 # list of "source" groups, we need to fetch node information later on.
11954 self.needed_locks = {
11955 locking.LEVEL_NODEGROUP: set([self.group_uuid]),
11956 locking.LEVEL_NODE: self.op.nodes,
11959 def DeclareLocks(self, level):
11960 if level == locking.LEVEL_NODEGROUP:
11961 assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1
11963 # Try to get all affected nodes' groups without having the group or node
11964 # lock yet. Needs verification later in the code flow.
11965 groups = self.cfg.GetNodeGroupsFromNodes(self.op.nodes)
11967 self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)
11969 def CheckPrereq(self):
11970 """Check prerequisites.
11973 assert self.needed_locks[locking.LEVEL_NODEGROUP]
11974 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
11975 frozenset(self.op.nodes))
11977 expected_locks = (set([self.group_uuid]) |
11978 self.cfg.GetNodeGroupsFromNodes(self.op.nodes))
11979 actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP)
11980 if actual_locks != expected_locks:
11981 raise errors.OpExecError("Nodes changed groups since locks were acquired,"
11982 " current groups are '%s', used to be '%s'" %
11983 (utils.CommaJoin(expected_locks),
11984 utils.CommaJoin(actual_locks)))
11986 self.node_data = self.cfg.GetAllNodesInfo()
11987 self.group = self.cfg.GetNodeGroup(self.group_uuid)
11988 instance_data = self.cfg.GetAllInstancesInfo()
11990 if self.group is None:
11991 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
11992 (self.op.group_name, self.group_uuid))
11994 (new_splits, previous_splits) = \
11995 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
11996 for node in self.op.nodes],
11997 self.node_data, instance_data)
12000 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
12002 if not self.op.force:
12003 raise errors.OpExecError("The following instances get split by this"
12004 " change and --force was not given: %s" %
12007 self.LogWarning("This operation will split the following instances: %s",
12010 if previous_splits:
12011 self.LogWarning("In addition, these already-split instances continue"
12012 " to be split across groups: %s",
12013 utils.CommaJoin(utils.NiceSort(previous_splits)))
12015 def Exec(self, feedback_fn):
12016 """Assign nodes to a new group.
12019 mods = [(node_name, self.group_uuid) for node_name in self.op.nodes]
12021 self.cfg.AssignGroupNodes(mods)
12024 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
12025 """Check for split instances after a node assignment.
12027 This method considers a series of node assignments as an atomic operation,
12028 and returns information about split instances after applying the set of
12031 In particular, it returns information about newly split instances, and
12032 instances that were already split, and remain so after the change.
12034 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
12037 @type changes: list of (node_name, new_group_uuid) pairs.
12038 @param changes: list of node assignments to consider.
12039 @param node_data: a dict with data for all nodes
12040 @param instance_data: a dict with all instances to consider
12041 @rtype: a two-tuple
12042 @return: a list of instances that were previously okay and result split as a
12043 consequence of this change, and a list of instances that were previously
12044 split and this change does not fix.
12047 changed_nodes = dict((node, group) for node, group in changes
12048 if node_data[node].group != group)
12050 all_split_instances = set()
12051 previously_split_instances = set()
12053 def InstanceNodes(instance):
12054 return [instance.primary_node] + list(instance.secondary_nodes)
12056 for inst in instance_data.values():
12057 if inst.disk_template not in constants.DTS_INT_MIRROR:
12060 instance_nodes = InstanceNodes(inst)
12062 if len(set(node_data[node].group for node in instance_nodes)) > 1:
12063 previously_split_instances.add(inst.name)
12065 if len(set(changed_nodes.get(node, node_data[node].group)
12066 for node in instance_nodes)) > 1:
12067 all_split_instances.add(inst.name)
12069 return (list(all_split_instances - previously_split_instances),
12070 list(previously_split_instances & all_split_instances))
12073 class _GroupQuery(_QueryBase):
12074 FIELDS = query.GROUP_FIELDS
12076 def ExpandNames(self, lu):
12077 lu.needed_locks = {}
12079 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
12080 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
12083 self.wanted = [name_to_uuid[name]
12084 for name in utils.NiceSort(name_to_uuid.keys())]
12086 # Accept names to be either names or UUIDs.
12089 all_uuid = frozenset(self._all_groups.keys())
12091 for name in self.names:
12092 if name in all_uuid:
12093 self.wanted.append(name)
12094 elif name in name_to_uuid:
12095 self.wanted.append(name_to_uuid[name])
12097 missing.append(name)
12100 raise errors.OpPrereqError("Some groups do not exist: %s" %
12101 utils.CommaJoin(missing),
12102 errors.ECODE_NOENT)
12104 def DeclareLocks(self, lu, level):
12107 def _GetQueryData(self, lu):
12108 """Computes the list of node groups and their attributes.
12111 do_nodes = query.GQ_NODE in self.requested_data
12112 do_instances = query.GQ_INST in self.requested_data
12114 group_to_nodes = None
12115 group_to_instances = None
12117 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
12118 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
12119 # latter GetAllInstancesInfo() is not enough, for we have to go through
12120 # instance->node. Hence, we will need to process nodes even if we only need
12121 # instance information.
12122 if do_nodes or do_instances:
12123 all_nodes = lu.cfg.GetAllNodesInfo()
12124 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
12127 for node in all_nodes.values():
12128 if node.group in group_to_nodes:
12129 group_to_nodes[node.group].append(node.name)
12130 node_to_group[node.name] = node.group
12133 all_instances = lu.cfg.GetAllInstancesInfo()
12134 group_to_instances = dict((uuid, []) for uuid in self.wanted)
12136 for instance in all_instances.values():
12137 node = instance.primary_node
12138 if node in node_to_group:
12139 group_to_instances[node_to_group[node]].append(instance.name)
12142 # Do not pass on node information if it was not requested.
12143 group_to_nodes = None
12145 return query.GroupQueryData([self._all_groups[uuid]
12146 for uuid in self.wanted],
12147 group_to_nodes, group_to_instances)
12150 class LUGroupQuery(NoHooksLU):
12151 """Logical unit for querying node groups.
12156 def CheckArguments(self):
12157 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
12158 self.op.output_fields, False)
12160 def ExpandNames(self):
12161 self.gq.ExpandNames(self)
12163 def DeclareLocks(self, level):
12164 self.gq.DeclareLocks(self, level)
12166 def Exec(self, feedback_fn):
12167 return self.gq.OldStyleQuery(self)
12170 class LUGroupSetParams(LogicalUnit):
12171 """Modifies the parameters of a node group.
12174 HPATH = "group-modify"
12175 HTYPE = constants.HTYPE_GROUP
12178 def CheckArguments(self):
12181 self.op.alloc_policy,
12184 if all_changes.count(None) == len(all_changes):
12185 raise errors.OpPrereqError("Please pass at least one modification",
12186 errors.ECODE_INVAL)
12188 def ExpandNames(self):
12189 # This raises errors.OpPrereqError on its own:
12190 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12192 self.needed_locks = {
12193 locking.LEVEL_NODEGROUP: [self.group_uuid],
12196 def CheckPrereq(self):
12197 """Check prerequisites.
12200 self.group = self.cfg.GetNodeGroup(self.group_uuid)
12202 if self.group is None:
12203 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
12204 (self.op.group_name, self.group_uuid))
12206 if self.op.ndparams:
12207 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
12208 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
12209 self.new_ndparams = new_ndparams
12211 def BuildHooksEnv(self):
12212 """Build hooks env.
12216 "GROUP_NAME": self.op.group_name,
12217 "NEW_ALLOC_POLICY": self.op.alloc_policy,
12220 def BuildHooksNodes(self):
12221 """Build hooks nodes.
12224 mn = self.cfg.GetMasterNode()
12225 return ([mn], [mn])
12227 def Exec(self, feedback_fn):
12228 """Modifies the node group.
12233 if self.op.ndparams:
12234 self.group.ndparams = self.new_ndparams
12235 result.append(("ndparams", str(self.group.ndparams)))
12237 if self.op.alloc_policy:
12238 self.group.alloc_policy = self.op.alloc_policy
12240 self.cfg.Update(self.group, feedback_fn)
12244 class LUGroupRemove(LogicalUnit):
12245 HPATH = "group-remove"
12246 HTYPE = constants.HTYPE_GROUP
12249 def ExpandNames(self):
12250 # This will raises errors.OpPrereqError on its own:
12251 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12252 self.needed_locks = {
12253 locking.LEVEL_NODEGROUP: [self.group_uuid],
12256 def CheckPrereq(self):
12257 """Check prerequisites.
12259 This checks that the given group name exists as a node group, that is
12260 empty (i.e., contains no nodes), and that is not the last group of the
12264 # Verify that the group is empty.
12265 group_nodes = [node.name
12266 for node in self.cfg.GetAllNodesInfo().values()
12267 if node.group == self.group_uuid]
12270 raise errors.OpPrereqError("Group '%s' not empty, has the following"
12272 (self.op.group_name,
12273 utils.CommaJoin(utils.NiceSort(group_nodes))),
12274 errors.ECODE_STATE)
12276 # Verify the cluster would not be left group-less.
12277 if len(self.cfg.GetNodeGroupList()) == 1:
12278 raise errors.OpPrereqError("Group '%s' is the only group,"
12279 " cannot be removed" %
12280 self.op.group_name,
12281 errors.ECODE_STATE)
12283 def BuildHooksEnv(self):
12284 """Build hooks env.
12288 "GROUP_NAME": self.op.group_name,
12291 def BuildHooksNodes(self):
12292 """Build hooks nodes.
12295 mn = self.cfg.GetMasterNode()
12296 return ([mn], [mn])
12298 def Exec(self, feedback_fn):
12299 """Remove the node group.
12303 self.cfg.RemoveNodeGroup(self.group_uuid)
12304 except errors.ConfigurationError:
12305 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
12306 (self.op.group_name, self.group_uuid))
12308 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
12311 class LUGroupRename(LogicalUnit):
12312 HPATH = "group-rename"
12313 HTYPE = constants.HTYPE_GROUP
12316 def ExpandNames(self):
12317 # This raises errors.OpPrereqError on its own:
12318 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12320 self.needed_locks = {
12321 locking.LEVEL_NODEGROUP: [self.group_uuid],
12324 def CheckPrereq(self):
12325 """Check prerequisites.
12327 Ensures requested new name is not yet used.
12331 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
12332 except errors.OpPrereqError:
12335 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
12336 " node group (UUID: %s)" %
12337 (self.op.new_name, new_name_uuid),
12338 errors.ECODE_EXISTS)
12340 def BuildHooksEnv(self):
12341 """Build hooks env.
12345 "OLD_NAME": self.op.group_name,
12346 "NEW_NAME": self.op.new_name,
12349 def BuildHooksNodes(self):
12350 """Build hooks nodes.
12353 mn = self.cfg.GetMasterNode()
12355 all_nodes = self.cfg.GetAllNodesInfo()
12356 all_nodes.pop(mn, None)
12359 run_nodes.extend(node.name for node in all_nodes.values()
12360 if node.group == self.group_uuid)
12362 return (run_nodes, run_nodes)
12364 def Exec(self, feedback_fn):
12365 """Rename the node group.
12368 group = self.cfg.GetNodeGroup(self.group_uuid)
12371 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
12372 (self.op.group_name, self.group_uuid))
12374 group.name = self.op.new_name
12375 self.cfg.Update(group, feedback_fn)
12377 return self.op.new_name
12380 class LUGroupEvacuate(LogicalUnit):
12381 HPATH = "group-evacuate"
12382 HTYPE = constants.HTYPE_GROUP
12385 def ExpandNames(self):
12386 # This raises errors.OpPrereqError on its own:
12387 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
12389 if self.op.target_groups:
12390 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
12391 self.op.target_groups)
12393 self.req_target_uuids = []
12395 if self.group_uuid in self.req_target_uuids:
12396 raise errors.OpPrereqError("Group to be evacuated (%s) can not be used"
12397 " as a target group (targets are %s)" %
12399 utils.CommaJoin(self.req_target_uuids)),
12400 errors.ECODE_INVAL)
12402 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
12404 self.share_locks = _ShareAll()
12405 self.needed_locks = {
12406 locking.LEVEL_INSTANCE: [],
12407 locking.LEVEL_NODEGROUP: [],
12408 locking.LEVEL_NODE: [],
12411 def DeclareLocks(self, level):
12412 if level == locking.LEVEL_INSTANCE:
12413 assert not self.needed_locks[locking.LEVEL_INSTANCE]
12415 # Lock instances optimistically, needs verification once node and group
12416 # locks have been acquired
12417 self.needed_locks[locking.LEVEL_INSTANCE] = \
12418 self.cfg.GetNodeGroupInstances(self.group_uuid)
12420 elif level == locking.LEVEL_NODEGROUP:
12421 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
12423 if self.req_target_uuids:
12424 lock_groups = set([self.group_uuid] + self.req_target_uuids)
12426 # Lock all groups used by instances optimistically; this requires going
12427 # via the node before it's locked, requiring verification later on
12428 lock_groups.update(group_uuid
12429 for instance_name in
12430 self.owned_locks(locking.LEVEL_INSTANCE)
12432 self.cfg.GetInstanceNodeGroups(instance_name))
12434 # No target groups, need to lock all of them
12435 lock_groups = locking.ALL_SET
12437 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
12439 elif level == locking.LEVEL_NODE:
12440 # This will only lock the nodes in the group to be evacuated which
12441 # contain actual instances
12442 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
12443 self._LockInstancesNodes()
12445 # Lock all nodes in group to be evacuated and target groups
12446 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12447 assert self.group_uuid in owned_groups
12448 member_nodes = [node_name
12449 for group in owned_groups
12450 for node_name in self.cfg.GetNodeGroup(group).members]
12451 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
12453 def CheckPrereq(self):
12454 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
12455 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12456 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
12458 assert owned_groups.issuperset(self.req_target_uuids)
12459 assert self.group_uuid in owned_groups
12461 # Check if locked instances are still correct
12462 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
12464 # Get instance information
12465 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
12467 # Check if node groups for locked instances are still correct
12468 for instance_name in owned_instances:
12469 inst = self.instances[instance_name]
12470 assert owned_nodes.issuperset(inst.all_nodes), \
12471 "Instance %s's nodes changed while we kept the lock" % instance_name
12473 inst_groups = _CheckInstanceNodeGroups(self.cfg, instance_name,
12476 assert self.group_uuid in inst_groups, \
12477 "Instance %s has no node in group %s" % (instance_name, self.group_uuid)
12479 if self.req_target_uuids:
12480 # User requested specific target groups
12481 self.target_uuids = self.req_target_uuids
12483 # All groups except the one to be evacuated are potential targets
12484 self.target_uuids = [group_uuid for group_uuid in owned_groups
12485 if group_uuid != self.group_uuid]
12487 if not self.target_uuids:
12488 raise errors.OpPrereqError("There are no possible target groups",
12489 errors.ECODE_INVAL)
12491 def BuildHooksEnv(self):
12492 """Build hooks env.
12496 "GROUP_NAME": self.op.group_name,
12497 "TARGET_GROUPS": " ".join(self.target_uuids),
12500 def BuildHooksNodes(self):
12501 """Build hooks nodes.
12504 mn = self.cfg.GetMasterNode()
12506 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
12508 run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members
12510 return (run_nodes, run_nodes)
12512 def Exec(self, feedback_fn):
12513 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
12515 assert self.group_uuid not in self.target_uuids
12517 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
12518 instances=instances, target_groups=self.target_uuids)
12520 ial.Run(self.op.iallocator)
12522 if not ial.success:
12523 raise errors.OpPrereqError("Can't compute group evacuation using"
12524 " iallocator '%s': %s" %
12525 (self.op.iallocator, ial.info),
12526 errors.ECODE_NORES)
12528 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
12530 self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s",
12531 len(jobs), self.op.group_name)
12533 return ResultWithJobs(jobs)
12536 class TagsLU(NoHooksLU): # pylint: disable=W0223
12537 """Generic tags LU.
12539 This is an abstract class which is the parent of all the other tags LUs.
12542 def ExpandNames(self):
12543 self.group_uuid = None
12544 self.needed_locks = {}
12545 if self.op.kind == constants.TAG_NODE:
12546 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
12547 self.needed_locks[locking.LEVEL_NODE] = self.op.name
12548 elif self.op.kind == constants.TAG_INSTANCE:
12549 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
12550 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
12551 elif self.op.kind == constants.TAG_NODEGROUP:
12552 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
12554 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
12555 # not possible to acquire the BGL based on opcode parameters)
12557 def CheckPrereq(self):
12558 """Check prerequisites.
12561 if self.op.kind == constants.TAG_CLUSTER:
12562 self.target = self.cfg.GetClusterInfo()
12563 elif self.op.kind == constants.TAG_NODE:
12564 self.target = self.cfg.GetNodeInfo(self.op.name)
12565 elif self.op.kind == constants.TAG_INSTANCE:
12566 self.target = self.cfg.GetInstanceInfo(self.op.name)
12567 elif self.op.kind == constants.TAG_NODEGROUP:
12568 self.target = self.cfg.GetNodeGroup(self.group_uuid)
12570 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
12571 str(self.op.kind), errors.ECODE_INVAL)
12574 class LUTagsGet(TagsLU):
12575 """Returns the tags of a given object.
12580 def ExpandNames(self):
12581 TagsLU.ExpandNames(self)
12583 # Share locks as this is only a read operation
12584 self.share_locks = _ShareAll()
12586 def Exec(self, feedback_fn):
12587 """Returns the tag list.
12590 return list(self.target.GetTags())
12593 class LUTagsSearch(NoHooksLU):
12594 """Searches the tags for a given pattern.
12599 def ExpandNames(self):
12600 self.needed_locks = {}
12602 def CheckPrereq(self):
12603 """Check prerequisites.
12605 This checks the pattern passed for validity by compiling it.
12609 self.re = re.compile(self.op.pattern)
12610 except re.error, err:
12611 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
12612 (self.op.pattern, err), errors.ECODE_INVAL)
12614 def Exec(self, feedback_fn):
12615 """Returns the tag list.
12619 tgts = [("/cluster", cfg.GetClusterInfo())]
12620 ilist = cfg.GetAllInstancesInfo().values()
12621 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
12622 nlist = cfg.GetAllNodesInfo().values()
12623 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
12624 tgts.extend(("/nodegroup/%s" % n.name, n)
12625 for n in cfg.GetAllNodeGroupsInfo().values())
12627 for path, target in tgts:
12628 for tag in target.GetTags():
12629 if self.re.search(tag):
12630 results.append((path, tag))
12634 class LUTagsSet(TagsLU):
12635 """Sets a tag on a given object.
12640 def CheckPrereq(self):
12641 """Check prerequisites.
12643 This checks the type and length of the tag name and value.
12646 TagsLU.CheckPrereq(self)
12647 for tag in self.op.tags:
12648 objects.TaggableObject.ValidateTag(tag)
12650 def Exec(self, feedback_fn):
12655 for tag in self.op.tags:
12656 self.target.AddTag(tag)
12657 except errors.TagError, err:
12658 raise errors.OpExecError("Error while setting tag: %s" % str(err))
12659 self.cfg.Update(self.target, feedback_fn)
12662 class LUTagsDel(TagsLU):
12663 """Delete a list of tags from a given object.
12668 def CheckPrereq(self):
12669 """Check prerequisites.
12671 This checks that we have the given tag.
12674 TagsLU.CheckPrereq(self)
12675 for tag in self.op.tags:
12676 objects.TaggableObject.ValidateTag(tag)
12677 del_tags = frozenset(self.op.tags)
12678 cur_tags = self.target.GetTags()
12680 diff_tags = del_tags - cur_tags
12682 diff_names = ("'%s'" % i for i in sorted(diff_tags))
12683 raise errors.OpPrereqError("Tag(s) %s not found" %
12684 (utils.CommaJoin(diff_names), ),
12685 errors.ECODE_NOENT)
12687 def Exec(self, feedback_fn):
12688 """Remove the tag from the object.
12691 for tag in self.op.tags:
12692 self.target.RemoveTag(tag)
12693 self.cfg.Update(self.target, feedback_fn)
12696 class LUTestDelay(NoHooksLU):
12697 """Sleep for a specified amount of time.
12699 This LU sleeps on the master and/or nodes for a specified amount of
12705 def ExpandNames(self):
12706 """Expand names and set required locks.
12708 This expands the node list, if any.
12711 self.needed_locks = {}
12712 if self.op.on_nodes:
12713 # _GetWantedNodes can be used here, but is not always appropriate to use
12714 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
12715 # more information.
12716 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
12717 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
12719 def _TestDelay(self):
12720 """Do the actual sleep.
12723 if self.op.on_master:
12724 if not utils.TestDelay(self.op.duration):
12725 raise errors.OpExecError("Error during master delay test")
12726 if self.op.on_nodes:
12727 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
12728 for node, node_result in result.items():
12729 node_result.Raise("Failure during rpc call to node %s" % node)
12731 def Exec(self, feedback_fn):
12732 """Execute the test delay opcode, with the wanted repetitions.
12735 if self.op.repeat == 0:
12738 top_value = self.op.repeat - 1
12739 for i in range(self.op.repeat):
12740 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
12744 class LUTestJqueue(NoHooksLU):
12745 """Utility LU to test some aspects of the job queue.
12750 # Must be lower than default timeout for WaitForJobChange to see whether it
12751 # notices changed jobs
12752 _CLIENT_CONNECT_TIMEOUT = 20.0
12753 _CLIENT_CONFIRM_TIMEOUT = 60.0
12756 def _NotifyUsingSocket(cls, cb, errcls):
12757 """Opens a Unix socket and waits for another program to connect.
12760 @param cb: Callback to send socket name to client
12761 @type errcls: class
12762 @param errcls: Exception class to use for errors
12765 # Using a temporary directory as there's no easy way to create temporary
12766 # sockets without writing a custom loop around tempfile.mktemp and
12768 tmpdir = tempfile.mkdtemp()
12770 tmpsock = utils.PathJoin(tmpdir, "sock")
12772 logging.debug("Creating temporary socket at %s", tmpsock)
12773 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
12778 # Send details to client
12781 # Wait for client to connect before continuing
12782 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
12784 (conn, _) = sock.accept()
12785 except socket.error, err:
12786 raise errcls("Client didn't connect in time (%s)" % err)
12790 # Remove as soon as client is connected
12791 shutil.rmtree(tmpdir)
12793 # Wait for client to close
12796 # pylint: disable=E1101
12797 # Instance of '_socketobject' has no ... member
12798 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
12800 except socket.error, err:
12801 raise errcls("Client failed to confirm notification (%s)" % err)
12805 def _SendNotification(self, test, arg, sockname):
12806 """Sends a notification to the client.
12809 @param test: Test name
12810 @param arg: Test argument (depends on test)
12811 @type sockname: string
12812 @param sockname: Socket path
12815 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
12817 def _Notify(self, prereq, test, arg):
12818 """Notifies the client of a test.
12821 @param prereq: Whether this is a prereq-phase test
12823 @param test: Test name
12824 @param arg: Test argument (depends on test)
12828 errcls = errors.OpPrereqError
12830 errcls = errors.OpExecError
12832 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
12836 def CheckArguments(self):
12837 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
12838 self.expandnames_calls = 0
12840 def ExpandNames(self):
12841 checkargs_calls = getattr(self, "checkargs_calls", 0)
12842 if checkargs_calls < 1:
12843 raise errors.ProgrammerError("CheckArguments was not called")
12845 self.expandnames_calls += 1
12847 if self.op.notify_waitlock:
12848 self._Notify(True, constants.JQT_EXPANDNAMES, None)
12850 self.LogInfo("Expanding names")
12852 # Get lock on master node (just to get a lock, not for a particular reason)
12853 self.needed_locks = {
12854 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
12857 def Exec(self, feedback_fn):
12858 if self.expandnames_calls < 1:
12859 raise errors.ProgrammerError("ExpandNames was not called")
12861 if self.op.notify_exec:
12862 self._Notify(False, constants.JQT_EXEC, None)
12864 self.LogInfo("Executing")
12866 if self.op.log_messages:
12867 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
12868 for idx, msg in enumerate(self.op.log_messages):
12869 self.LogInfo("Sending log message %s", idx + 1)
12870 feedback_fn(constants.JQT_MSGPREFIX + msg)
12871 # Report how many test messages have been sent
12872 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
12875 raise errors.OpExecError("Opcode failure was requested")
12880 class IAllocator(object):
12881 """IAllocator framework.
12883 An IAllocator instance has three sets of attributes:
12884 - cfg that is needed to query the cluster
12885 - input data (all members of the _KEYS class attribute are required)
12886 - four buffer attributes (in|out_data|text), that represent the
12887 input (to the external script) in text and data structure format,
12888 and the output from it, again in two formats
12889 - the result variables from the script (success, info, nodes) for
12893 # pylint: disable=R0902
12894 # lots of instance attributes
12896 def __init__(self, cfg, rpc, mode, **kwargs):
12899 # init buffer variables
12900 self.in_text = self.out_text = self.in_data = self.out_data = None
12901 # init all input fields so that pylint is happy
12903 self.memory = self.disks = self.disk_template = None
12904 self.os = self.tags = self.nics = self.vcpus = None
12905 self.hypervisor = None
12906 self.relocate_from = None
12908 self.instances = None
12909 self.evac_mode = None
12910 self.target_groups = []
12912 self.required_nodes = None
12913 # init result fields
12914 self.success = self.info = self.result = None
12917 (fn, keydata, self._result_check) = self._MODE_DATA[self.mode]
12919 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
12920 " IAllocator" % self.mode)
12922 keyset = [n for (n, _) in keydata]
12925 if key not in keyset:
12926 raise errors.ProgrammerError("Invalid input parameter '%s' to"
12927 " IAllocator" % key)
12928 setattr(self, key, kwargs[key])
12931 if key not in kwargs:
12932 raise errors.ProgrammerError("Missing input parameter '%s' to"
12933 " IAllocator" % key)
12934 self._BuildInputData(compat.partial(fn, self), keydata)
12936 def _ComputeClusterData(self):
12937 """Compute the generic allocator input data.
12939 This is the data that is independent of the actual operation.
12943 cluster_info = cfg.GetClusterInfo()
12946 "version": constants.IALLOCATOR_VERSION,
12947 "cluster_name": cfg.GetClusterName(),
12948 "cluster_tags": list(cluster_info.GetTags()),
12949 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
12950 # we don't have job IDs
12952 ninfo = cfg.GetAllNodesInfo()
12953 iinfo = cfg.GetAllInstancesInfo().values()
12954 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
12957 node_list = [n.name for n in ninfo.values() if n.vm_capable]
12959 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
12960 hypervisor_name = self.hypervisor
12961 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
12962 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
12964 hypervisor_name = cluster_info.enabled_hypervisors[0]
12966 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
12969 self.rpc.call_all_instances_info(node_list,
12970 cluster_info.enabled_hypervisors)
12972 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
12974 config_ndata = self._ComputeBasicNodeData(ninfo)
12975 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
12976 i_list, config_ndata)
12977 assert len(data["nodes"]) == len(ninfo), \
12978 "Incomplete node data computed"
12980 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
12982 self.in_data = data
12985 def _ComputeNodeGroupData(cfg):
12986 """Compute node groups data.
12989 ng = dict((guuid, {
12990 "name": gdata.name,
12991 "alloc_policy": gdata.alloc_policy,
12993 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items())
12998 def _ComputeBasicNodeData(node_cfg):
12999 """Compute global node data.
13002 @returns: a dict of name: (node dict, node config)
13005 # fill in static (config-based) values
13006 node_results = dict((ninfo.name, {
13007 "tags": list(ninfo.GetTags()),
13008 "primary_ip": ninfo.primary_ip,
13009 "secondary_ip": ninfo.secondary_ip,
13010 "offline": ninfo.offline,
13011 "drained": ninfo.drained,
13012 "master_candidate": ninfo.master_candidate,
13013 "group": ninfo.group,
13014 "master_capable": ninfo.master_capable,
13015 "vm_capable": ninfo.vm_capable,
13017 for ninfo in node_cfg.values())
13019 return node_results
13022 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
13024 """Compute global node data.
13026 @param node_results: the basic node structures as filled from the config
13029 # make a copy of the current dict
13030 node_results = dict(node_results)
13031 for nname, nresult in node_data.items():
13032 assert nname in node_results, "Missing basic data for node %s" % nname
13033 ninfo = node_cfg[nname]
13035 if not (ninfo.offline or ninfo.drained):
13036 nresult.Raise("Can't get data for node %s" % nname)
13037 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
13039 remote_info = nresult.payload
13041 for attr in ["memory_total", "memory_free", "memory_dom0",
13042 "vg_size", "vg_free", "cpu_total"]:
13043 if attr not in remote_info:
13044 raise errors.OpExecError("Node '%s' didn't return attribute"
13045 " '%s'" % (nname, attr))
13046 if not isinstance(remote_info[attr], int):
13047 raise errors.OpExecError("Node '%s' returned invalid value"
13049 (nname, attr, remote_info[attr]))
13050 # compute memory used by primary instances
13051 i_p_mem = i_p_up_mem = 0
13052 for iinfo, beinfo in i_list:
13053 if iinfo.primary_node == nname:
13054 i_p_mem += beinfo[constants.BE_MEMORY]
13055 if iinfo.name not in node_iinfo[nname].payload:
13058 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]["memory"])
13059 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
13060 remote_info["memory_free"] -= max(0, i_mem_diff)
13063 i_p_up_mem += beinfo[constants.BE_MEMORY]
13065 # compute memory used by instances
13067 "total_memory": remote_info["memory_total"],
13068 "reserved_memory": remote_info["memory_dom0"],
13069 "free_memory": remote_info["memory_free"],
13070 "total_disk": remote_info["vg_size"],
13071 "free_disk": remote_info["vg_free"],
13072 "total_cpus": remote_info["cpu_total"],
13073 "i_pri_memory": i_p_mem,
13074 "i_pri_up_memory": i_p_up_mem,
13076 pnr_dyn.update(node_results[nname])
13077 node_results[nname] = pnr_dyn
13079 return node_results
13082 def _ComputeInstanceData(cluster_info, i_list):
13083 """Compute global instance data.
13087 for iinfo, beinfo in i_list:
13089 for nic in iinfo.nics:
13090 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
13094 "mode": filled_params[constants.NIC_MODE],
13095 "link": filled_params[constants.NIC_LINK],
13097 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
13098 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
13099 nic_data.append(nic_dict)
13101 "tags": list(iinfo.GetTags()),
13102 "admin_up": iinfo.admin_up,
13103 "vcpus": beinfo[constants.BE_VCPUS],
13104 "memory": beinfo[constants.BE_MEMORY],
13106 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
13108 "disks": [{constants.IDISK_SIZE: dsk.size,
13109 constants.IDISK_MODE: dsk.mode}
13110 for dsk in iinfo.disks],
13111 "disk_template": iinfo.disk_template,
13112 "hypervisor": iinfo.hypervisor,
13114 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
13116 instance_data[iinfo.name] = pir
13118 return instance_data
13120 def _AddNewInstance(self):
13121 """Add new instance data to allocator structure.
13123 This in combination with _AllocatorGetClusterData will create the
13124 correct structure needed as input for the allocator.
13126 The checks for the completeness of the opcode must have already been
13130 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
13132 if self.disk_template in constants.DTS_INT_MIRROR:
13133 self.required_nodes = 2
13135 self.required_nodes = 1
13139 "disk_template": self.disk_template,
13142 "vcpus": self.vcpus,
13143 "memory": self.memory,
13144 "disks": self.disks,
13145 "disk_space_total": disk_space,
13147 "required_nodes": self.required_nodes,
13148 "hypervisor": self.hypervisor,
13153 def _AddRelocateInstance(self):
13154 """Add relocate instance data to allocator structure.
13156 This in combination with _IAllocatorGetClusterData will create the
13157 correct structure needed as input for the allocator.
13159 The checks for the completeness of the opcode must have already been
13163 instance = self.cfg.GetInstanceInfo(self.name)
13164 if instance is None:
13165 raise errors.ProgrammerError("Unknown instance '%s' passed to"
13166 " IAllocator" % self.name)
13168 if instance.disk_template not in constants.DTS_MIRRORED:
13169 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
13170 errors.ECODE_INVAL)
13172 if instance.disk_template in constants.DTS_INT_MIRROR and \
13173 len(instance.secondary_nodes) != 1:
13174 raise errors.OpPrereqError("Instance has not exactly one secondary node",
13175 errors.ECODE_STATE)
13177 self.required_nodes = 1
13178 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
13179 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
13183 "disk_space_total": disk_space,
13184 "required_nodes": self.required_nodes,
13185 "relocate_from": self.relocate_from,
13189 def _AddNodeEvacuate(self):
13190 """Get data for node-evacuate requests.
13194 "instances": self.instances,
13195 "evac_mode": self.evac_mode,
13198 def _AddChangeGroup(self):
13199 """Get data for node-evacuate requests.
13203 "instances": self.instances,
13204 "target_groups": self.target_groups,
13207 def _BuildInputData(self, fn, keydata):
13208 """Build input data structures.
13211 self._ComputeClusterData()
13214 request["type"] = self.mode
13215 for keyname, keytype in keydata:
13216 if keyname not in request:
13217 raise errors.ProgrammerError("Request parameter %s is missing" %
13219 val = request[keyname]
13220 if not keytype(val):
13221 raise errors.ProgrammerError("Request parameter %s doesn't pass"
13222 " validation, value %s, expected"
13223 " type %s" % (keyname, val, keytype))
13224 self.in_data["request"] = request
13226 self.in_text = serializer.Dump(self.in_data)
13228 _STRING_LIST = ht.TListOf(ht.TString)
13229 _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
13230 # pylint: disable=E1101
13231 # Class '...' has no 'OP_ID' member
13232 "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
13233 opcodes.OpInstanceMigrate.OP_ID,
13234 opcodes.OpInstanceReplaceDisks.OP_ID])
13238 ht.TListOf(ht.TAnd(ht.TIsLength(3),
13239 ht.TItems([ht.TNonEmptyString,
13240 ht.TNonEmptyString,
13241 ht.TListOf(ht.TNonEmptyString),
13244 ht.TListOf(ht.TAnd(ht.TIsLength(2),
13245 ht.TItems([ht.TNonEmptyString,
13248 _NEVAC_RESULT = ht.TAnd(ht.TIsLength(3),
13249 ht.TItems([_NEVAC_MOVED, _NEVAC_FAILED, _JOB_LIST]))
13252 constants.IALLOCATOR_MODE_ALLOC:
13255 ("name", ht.TString),
13256 ("memory", ht.TInt),
13257 ("disks", ht.TListOf(ht.TDict)),
13258 ("disk_template", ht.TString),
13259 ("os", ht.TString),
13260 ("tags", _STRING_LIST),
13261 ("nics", ht.TListOf(ht.TDict)),
13262 ("vcpus", ht.TInt),
13263 ("hypervisor", ht.TString),
13265 constants.IALLOCATOR_MODE_RELOC:
13266 (_AddRelocateInstance,
13267 [("name", ht.TString), ("relocate_from", _STRING_LIST)],
13269 constants.IALLOCATOR_MODE_NODE_EVAC:
13270 (_AddNodeEvacuate, [
13271 ("instances", _STRING_LIST),
13272 ("evac_mode", ht.TElemOf(constants.IALLOCATOR_NEVAC_MODES)),
13274 constants.IALLOCATOR_MODE_CHG_GROUP:
13275 (_AddChangeGroup, [
13276 ("instances", _STRING_LIST),
13277 ("target_groups", _STRING_LIST),
13281 def Run(self, name, validate=True, call_fn=None):
13282 """Run an instance allocator and return the results.
13285 if call_fn is None:
13286 call_fn = self.rpc.call_iallocator_runner
13288 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
13289 result.Raise("Failure while running the iallocator script")
13291 self.out_text = result.payload
13293 self._ValidateResult()
13295 def _ValidateResult(self):
13296 """Process the allocator results.
13298 This will process and if successful save the result in
13299 self.out_data and the other parameters.
13303 rdict = serializer.Load(self.out_text)
13304 except Exception, err:
13305 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
13307 if not isinstance(rdict, dict):
13308 raise errors.OpExecError("Can't parse iallocator results: not a dict")
13310 # TODO: remove backwards compatiblity in later versions
13311 if "nodes" in rdict and "result" not in rdict:
13312 rdict["result"] = rdict["nodes"]
13315 for key in "success", "info", "result":
13316 if key not in rdict:
13317 raise errors.OpExecError("Can't parse iallocator results:"
13318 " missing key '%s'" % key)
13319 setattr(self, key, rdict[key])
13321 if not self._result_check(self.result):
13322 raise errors.OpExecError("Iallocator returned invalid result,"
13323 " expected %s, got %s" %
13324 (self._result_check, self.result),
13325 errors.ECODE_INVAL)
13327 if self.mode == constants.IALLOCATOR_MODE_RELOC:
13328 assert self.relocate_from is not None
13329 assert self.required_nodes == 1
13331 node2group = dict((name, ndata["group"])
13332 for (name, ndata) in self.in_data["nodes"].items())
13334 fn = compat.partial(self._NodesToGroups, node2group,
13335 self.in_data["nodegroups"])
13337 instance = self.cfg.GetInstanceInfo(self.name)
13338 request_groups = fn(self.relocate_from + [instance.primary_node])
13339 result_groups = fn(rdict["result"] + [instance.primary_node])
13341 if self.success and not set(result_groups).issubset(request_groups):
13342 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
13343 " differ from original groups (%s)" %
13344 (utils.CommaJoin(result_groups),
13345 utils.CommaJoin(request_groups)))
13347 elif self.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
13348 assert self.evac_mode in constants.IALLOCATOR_NEVAC_MODES
13350 self.out_data = rdict
13353 def _NodesToGroups(node2group, groups, nodes):
13354 """Returns a list of unique group names for a list of nodes.
13356 @type node2group: dict
13357 @param node2group: Map from node name to group UUID
13359 @param groups: Group information
13361 @param nodes: Node names
13368 group_uuid = node2group[node]
13370 # Ignore unknown node
13374 group = groups[group_uuid]
13376 # Can't find group, let's use UUID
13377 group_name = group_uuid
13379 group_name = group["name"]
13381 result.add(group_name)
13383 return sorted(result)
13386 class LUTestAllocator(NoHooksLU):
13387 """Run allocator tests.
13389 This LU runs the allocator tests
13392 def CheckPrereq(self):
13393 """Check prerequisites.
13395 This checks the opcode parameters depending on the director and mode test.
13398 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
13399 for attr in ["memory", "disks", "disk_template",
13400 "os", "tags", "nics", "vcpus"]:
13401 if not hasattr(self.op, attr):
13402 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
13403 attr, errors.ECODE_INVAL)
13404 iname = self.cfg.ExpandInstanceName(self.op.name)
13405 if iname is not None:
13406 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
13407 iname, errors.ECODE_EXISTS)
13408 if not isinstance(self.op.nics, list):
13409 raise errors.OpPrereqError("Invalid parameter 'nics'",
13410 errors.ECODE_INVAL)
13411 if not isinstance(self.op.disks, list):
13412 raise errors.OpPrereqError("Invalid parameter 'disks'",
13413 errors.ECODE_INVAL)
13414 for row in self.op.disks:
13415 if (not isinstance(row, dict) or
13416 constants.IDISK_SIZE not in row or
13417 not isinstance(row[constants.IDISK_SIZE], int) or
13418 constants.IDISK_MODE not in row or
13419 row[constants.IDISK_MODE] not in constants.DISK_ACCESS_SET):
13420 raise errors.OpPrereqError("Invalid contents of the 'disks'"
13421 " parameter", errors.ECODE_INVAL)
13422 if self.op.hypervisor is None:
13423 self.op.hypervisor = self.cfg.GetHypervisorType()
13424 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
13425 fname = _ExpandInstanceName(self.cfg, self.op.name)
13426 self.op.name = fname
13427 self.relocate_from = \
13428 list(self.cfg.GetInstanceInfo(fname).secondary_nodes)
13429 elif self.op.mode in (constants.IALLOCATOR_MODE_CHG_GROUP,
13430 constants.IALLOCATOR_MODE_NODE_EVAC):
13431 if not self.op.instances:
13432 raise errors.OpPrereqError("Missing instances", errors.ECODE_INVAL)
13433 self.op.instances = _GetWantedInstances(self, self.op.instances)
13435 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
13436 self.op.mode, errors.ECODE_INVAL)
13438 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
13439 if self.op.allocator is None:
13440 raise errors.OpPrereqError("Missing allocator name",
13441 errors.ECODE_INVAL)
13442 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
13443 raise errors.OpPrereqError("Wrong allocator test '%s'" %
13444 self.op.direction, errors.ECODE_INVAL)
13446 def Exec(self, feedback_fn):
13447 """Run the allocator test.
13450 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
13451 ial = IAllocator(self.cfg, self.rpc,
13454 memory=self.op.memory,
13455 disks=self.op.disks,
13456 disk_template=self.op.disk_template,
13460 vcpus=self.op.vcpus,
13461 hypervisor=self.op.hypervisor,
13463 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
13464 ial = IAllocator(self.cfg, self.rpc,
13467 relocate_from=list(self.relocate_from),
13469 elif self.op.mode == constants.IALLOCATOR_MODE_CHG_GROUP:
13470 ial = IAllocator(self.cfg, self.rpc,
13472 instances=self.op.instances,
13473 target_groups=self.op.target_groups)
13474 elif self.op.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
13475 ial = IAllocator(self.cfg, self.rpc,
13477 instances=self.op.instances,
13478 evac_mode=self.op.evac_mode)
13480 raise errors.ProgrammerError("Uncatched mode %s in"
13481 " LUTestAllocator.Exec", self.op.mode)
13483 if self.op.direction == constants.IALLOCATOR_DIR_IN:
13484 result = ial.in_text
13486 ial.Run(self.op.allocator, validate=False)
13487 result = ial.out_text
13491 #: Query type implementations
13493 constants.QR_INSTANCE: _InstanceQuery,
13494 constants.QR_NODE: _NodeQuery,
13495 constants.QR_GROUP: _GroupQuery,
13496 constants.QR_OS: _OsQuery,
13499 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
13502 def _GetQueryImplementation(name):
13503 """Returns the implemtnation for a query type.
13505 @param name: Query type, must be one of L{constants.QR_VIA_OP}
13509 return _QUERY_IMPL[name]
13511 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
13512 errors.ECODE_INVAL)