4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Module implementing the master-side code."""
24 # pylint: disable=W0201,C0302
26 # W0201 since most LU attributes are defined in CheckPrereq or similar
29 # C0302: since we have waaaay too many lines in this module
44 from ganeti import ssh
45 from ganeti import utils
46 from ganeti import errors
47 from ganeti import hypervisor
48 from ganeti import locking
49 from ganeti import constants
50 from ganeti import objects
51 from ganeti import serializer
52 from ganeti import ssconf
53 from ganeti import uidpool
54 from ganeti import compat
55 from ganeti import masterd
56 from ganeti import netutils
57 from ganeti import query
58 from ganeti import qlang
59 from ganeti import opcodes
61 from ganeti import rpc
62 from ganeti import runtime
64 import ganeti.masterd.instance # pylint: disable=W0611
67 #: Size of DRBD meta block device
71 INSTANCE_DOWN = [constants.ADMINST_DOWN]
72 INSTANCE_ONLINE = [constants.ADMINST_DOWN, constants.ADMINST_UP]
73 INSTANCE_NOT_RUNNING = [constants.ADMINST_DOWN, constants.ADMINST_OFFLINE]
75 #: Instance status in which an instance can be marked as offline/online
76 CAN_CHANGE_INSTANCE_OFFLINE = (frozenset(INSTANCE_DOWN) | frozenset([
77 constants.ADMINST_OFFLINE,
82 """Data container for LU results with jobs.
84 Instances of this class returned from L{LogicalUnit.Exec} will be recognized
85 by L{mcpu._ProcessResult}. The latter will then submit the jobs
86 contained in the C{jobs} attribute and include the job IDs in the opcode
90 def __init__(self, jobs, **kwargs):
91 """Initializes this class.
93 Additional return values can be specified as keyword arguments.
95 @type jobs: list of lists of L{opcode.OpCode}
96 @param jobs: A list of lists of opcode objects
103 class LogicalUnit(object):
104 """Logical Unit base class.
106 Subclasses must follow these rules:
107 - implement ExpandNames
108 - implement CheckPrereq (except when tasklets are used)
109 - implement Exec (except when tasklets are used)
110 - implement BuildHooksEnv
111 - implement BuildHooksNodes
112 - redefine HPATH and HTYPE
113 - optionally redefine their run requirements:
114 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
116 Note that all commands require root permissions.
118 @ivar dry_run_result: the value (if any) that will be returned to the caller
119 in dry-run mode (signalled by opcode dry_run parameter)
126 def __init__(self, processor, op, context, rpc_runner):
127 """Constructor for LogicalUnit.
129 This needs to be overridden in derived classes in order to check op
133 self.proc = processor
135 self.cfg = context.cfg
136 self.glm = context.glm
138 self.owned_locks = context.glm.list_owned
139 self.context = context
140 self.rpc = rpc_runner
141 # Dicts used to declare locking needs to mcpu
142 self.needed_locks = None
143 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
145 self.remove_locks = {}
146 # Used to force good behavior when calling helper functions
147 self.recalculate_locks = {}
149 self.Log = processor.Log # pylint: disable=C0103
150 self.LogWarning = processor.LogWarning # pylint: disable=C0103
151 self.LogInfo = processor.LogInfo # pylint: disable=C0103
152 self.LogStep = processor.LogStep # pylint: disable=C0103
153 # support for dry-run
154 self.dry_run_result = None
155 # support for generic debug attribute
156 if (not hasattr(self.op, "debug_level") or
157 not isinstance(self.op.debug_level, int)):
158 self.op.debug_level = 0
163 # Validate opcode parameters and set defaults
164 self.op.Validate(True)
166 self.CheckArguments()
168 def CheckArguments(self):
169 """Check syntactic validity for the opcode arguments.
171 This method is for doing a simple syntactic check and ensure
172 validity of opcode parameters, without any cluster-related
173 checks. While the same can be accomplished in ExpandNames and/or
174 CheckPrereq, doing these separate is better because:
176 - ExpandNames is left as as purely a lock-related function
177 - CheckPrereq is run after we have acquired locks (and possible
180 The function is allowed to change the self.op attribute so that
181 later methods can no longer worry about missing parameters.
186 def ExpandNames(self):
187 """Expand names for this LU.
189 This method is called before starting to execute the opcode, and it should
190 update all the parameters of the opcode to their canonical form (e.g. a
191 short node name must be fully expanded after this method has successfully
192 completed). This way locking, hooks, logging, etc. can work correctly.
194 LUs which implement this method must also populate the self.needed_locks
195 member, as a dict with lock levels as keys, and a list of needed lock names
198 - use an empty dict if you don't need any lock
199 - if you don't need any lock at a particular level omit that
200 level (note that in this case C{DeclareLocks} won't be called
201 at all for that level)
202 - if you need locks at a level, but you can't calculate it in
203 this function, initialise that level with an empty list and do
204 further processing in L{LogicalUnit.DeclareLocks} (see that
205 function's docstring)
206 - don't put anything for the BGL level
207 - if you want all locks at a level use L{locking.ALL_SET} as a value
209 If you need to share locks (rather than acquire them exclusively) at one
210 level you can modify self.share_locks, setting a true value (usually 1) for
211 that level. By default locks are not shared.
213 This function can also define a list of tasklets, which then will be
214 executed in order instead of the usual LU-level CheckPrereq and Exec
215 functions, if those are not defined by the LU.
219 # Acquire all nodes and one instance
220 self.needed_locks = {
221 locking.LEVEL_NODE: locking.ALL_SET,
222 locking.LEVEL_INSTANCE: ['instance1.example.com'],
224 # Acquire just two nodes
225 self.needed_locks = {
226 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
229 self.needed_locks = {} # No, you can't leave it to the default value None
232 # The implementation of this method is mandatory only if the new LU is
233 # concurrent, so that old LUs don't need to be changed all at the same
236 self.needed_locks = {} # Exclusive LUs don't need locks.
238 raise NotImplementedError
240 def DeclareLocks(self, level):
241 """Declare LU locking needs for a level
243 While most LUs can just declare their locking needs at ExpandNames time,
244 sometimes there's the need to calculate some locks after having acquired
245 the ones before. This function is called just before acquiring locks at a
246 particular level, but after acquiring the ones at lower levels, and permits
247 such calculations. It can be used to modify self.needed_locks, and by
248 default it does nothing.
250 This function is only called if you have something already set in
251 self.needed_locks for the level.
253 @param level: Locking level which is going to be locked
254 @type level: member of L{ganeti.locking.LEVELS}
258 def CheckPrereq(self):
259 """Check prerequisites for this LU.
261 This method should check that the prerequisites for the execution
262 of this LU are fulfilled. It can do internode communication, but
263 it should be idempotent - no cluster or system changes are
266 The method should raise errors.OpPrereqError in case something is
267 not fulfilled. Its return value is ignored.
269 This method should also update all the parameters of the opcode to
270 their canonical form if it hasn't been done by ExpandNames before.
273 if self.tasklets is not None:
274 for (idx, tl) in enumerate(self.tasklets):
275 logging.debug("Checking prerequisites for tasklet %s/%s",
276 idx + 1, len(self.tasklets))
281 def Exec(self, feedback_fn):
284 This method should implement the actual work. It should raise
285 errors.OpExecError for failures that are somewhat dealt with in
289 if self.tasklets is not None:
290 for (idx, tl) in enumerate(self.tasklets):
291 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
294 raise NotImplementedError
296 def BuildHooksEnv(self):
297 """Build hooks environment for this LU.
300 @return: Dictionary containing the environment that will be used for
301 running the hooks for this LU. The keys of the dict must not be prefixed
302 with "GANETI_"--that'll be added by the hooks runner. The hooks runner
303 will extend the environment with additional variables. If no environment
304 should be defined, an empty dictionary should be returned (not C{None}).
305 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
309 raise NotImplementedError
311 def BuildHooksNodes(self):
312 """Build list of nodes to run LU's hooks.
314 @rtype: tuple; (list, list)
315 @return: Tuple containing a list of node names on which the hook
316 should run before the execution and a list of node names on which the
317 hook should run after the execution. No nodes should be returned as an
318 empty list (and not None).
319 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
323 raise NotImplementedError
325 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
326 """Notify the LU about the results of its hooks.
328 This method is called every time a hooks phase is executed, and notifies
329 the Logical Unit about the hooks' result. The LU can then use it to alter
330 its result based on the hooks. By default the method does nothing and the
331 previous result is passed back unchanged but any LU can define it if it
332 wants to use the local cluster hook-scripts somehow.
334 @param phase: one of L{constants.HOOKS_PHASE_POST} or
335 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
336 @param hook_results: the results of the multi-node hooks rpc call
337 @param feedback_fn: function used send feedback back to the caller
338 @param lu_result: the previous Exec result this LU had, or None
340 @return: the new Exec result, based on the previous result
344 # API must be kept, thus we ignore the unused argument and could
345 # be a function warnings
346 # pylint: disable=W0613,R0201
349 def _ExpandAndLockInstance(self):
350 """Helper function to expand and lock an instance.
352 Many LUs that work on an instance take its name in self.op.instance_name
353 and need to expand it and then declare the expanded name for locking. This
354 function does it, and then updates self.op.instance_name to the expanded
355 name. It also initializes needed_locks as a dict, if this hasn't been done
359 if self.needed_locks is None:
360 self.needed_locks = {}
362 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
363 "_ExpandAndLockInstance called with instance-level locks set"
364 self.op.instance_name = _ExpandInstanceName(self.cfg,
365 self.op.instance_name)
366 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
368 def _LockInstancesNodes(self, primary_only=False,
369 level=locking.LEVEL_NODE):
370 """Helper function to declare instances' nodes for locking.
372 This function should be called after locking one or more instances to lock
373 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
374 with all primary or secondary nodes for instances already locked and
375 present in self.needed_locks[locking.LEVEL_INSTANCE].
377 It should be called from DeclareLocks, and for safety only works if
378 self.recalculate_locks[locking.LEVEL_NODE] is set.
380 In the future it may grow parameters to just lock some instance's nodes, or
381 to just lock primaries or secondary nodes, if needed.
383 If should be called in DeclareLocks in a way similar to::
385 if level == locking.LEVEL_NODE:
386 self._LockInstancesNodes()
388 @type primary_only: boolean
389 @param primary_only: only lock primary nodes of locked instances
390 @param level: Which lock level to use for locking nodes
393 assert level in self.recalculate_locks, \
394 "_LockInstancesNodes helper function called with no nodes to recalculate"
396 # TODO: check if we're really been called with the instance locks held
398 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
399 # future we might want to have different behaviors depending on the value
400 # of self.recalculate_locks[locking.LEVEL_NODE]
402 locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
403 for _, instance in self.cfg.GetMultiInstanceInfo(locked_i):
404 wanted_nodes.append(instance.primary_node)
406 wanted_nodes.extend(instance.secondary_nodes)
408 if self.recalculate_locks[level] == constants.LOCKS_REPLACE:
409 self.needed_locks[level] = wanted_nodes
410 elif self.recalculate_locks[level] == constants.LOCKS_APPEND:
411 self.needed_locks[level].extend(wanted_nodes)
413 raise errors.ProgrammerError("Unknown recalculation mode")
415 del self.recalculate_locks[level]
418 class NoHooksLU(LogicalUnit): # pylint: disable=W0223
419 """Simple LU which runs no hooks.
421 This LU is intended as a parent for other LogicalUnits which will
422 run no hooks, in order to reduce duplicate code.
428 def BuildHooksEnv(self):
429 """Empty BuildHooksEnv for NoHooksLu.
431 This just raises an error.
434 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
436 def BuildHooksNodes(self):
437 """Empty BuildHooksNodes for NoHooksLU.
440 raise AssertionError("BuildHooksNodes called for NoHooksLU")
444 """Tasklet base class.
446 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
447 they can mix legacy code with tasklets. Locking needs to be done in the LU,
448 tasklets know nothing about locks.
450 Subclasses must follow these rules:
451 - Implement CheckPrereq
455 def __init__(self, lu):
462 def CheckPrereq(self):
463 """Check prerequisites for this tasklets.
465 This method should check whether the prerequisites for the execution of
466 this tasklet are fulfilled. It can do internode communication, but it
467 should be idempotent - no cluster or system changes are allowed.
469 The method should raise errors.OpPrereqError in case something is not
470 fulfilled. Its return value is ignored.
472 This method should also update all parameters to their canonical form if it
473 hasn't been done before.
478 def Exec(self, feedback_fn):
479 """Execute the tasklet.
481 This method should implement the actual work. It should raise
482 errors.OpExecError for failures that are somewhat dealt with in code, or
486 raise NotImplementedError
490 """Base for query utility classes.
493 #: Attribute holding field definitions
499 def __init__(self, qfilter, fields, use_locking):
500 """Initializes this class.
503 self.use_locking = use_locking
505 self.query = query.Query(self.FIELDS, fields, qfilter=qfilter,
506 namefield=self.SORT_FIELD)
507 self.requested_data = self.query.RequestedData()
508 self.names = self.query.RequestedNames()
510 # Sort only if no names were requested
511 self.sort_by_name = not self.names
513 self.do_locking = None
516 def _GetNames(self, lu, all_names, lock_level):
517 """Helper function to determine names asked for in the query.
521 names = lu.owned_locks(lock_level)
525 if self.wanted == locking.ALL_SET:
526 assert not self.names
527 # caller didn't specify names, so ordering is not important
528 return utils.NiceSort(names)
530 # caller specified names and we must keep the same order
532 assert not self.do_locking or lu.glm.is_owned(lock_level)
534 missing = set(self.wanted).difference(names)
536 raise errors.OpExecError("Some items were removed before retrieving"
537 " their data: %s" % missing)
539 # Return expanded names
542 def ExpandNames(self, lu):
543 """Expand names for this query.
545 See L{LogicalUnit.ExpandNames}.
548 raise NotImplementedError()
550 def DeclareLocks(self, lu, level):
551 """Declare locks for this query.
553 See L{LogicalUnit.DeclareLocks}.
556 raise NotImplementedError()
558 def _GetQueryData(self, lu):
559 """Collects all data for this query.
561 @return: Query data object
564 raise NotImplementedError()
566 def NewStyleQuery(self, lu):
567 """Collect data and execute query.
570 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
571 sort_by_name=self.sort_by_name)
573 def OldStyleQuery(self, lu):
574 """Collect data and execute query.
577 return self.query.OldStyleQuery(self._GetQueryData(lu),
578 sort_by_name=self.sort_by_name)
582 """Returns a dict declaring all lock levels shared.
585 return dict.fromkeys(locking.LEVELS, 1)
588 def _MakeLegacyNodeInfo(data):
589 """Formats the data returned by L{rpc.RpcRunner.call_node_info}.
591 Converts the data into a single dictionary. This is fine for most use cases,
592 but some require information from more than one volume group or hypervisor.
595 (bootid, (vg_info, ), (hv_info, )) = data
597 return utils.JoinDisjointDicts(utils.JoinDisjointDicts(vg_info, hv_info), {
602 def _CheckInstancesNodeGroups(cfg, instances, owned_groups, owned_nodes,
604 """Checks if node groups for locked instances are still correct.
606 @type cfg: L{config.ConfigWriter}
607 @param cfg: Cluster configuration
608 @type instances: dict; string as key, L{objects.Instance} as value
609 @param instances: Dictionary, instance name as key, instance object as value
610 @type owned_groups: iterable of string
611 @param owned_groups: List of owned groups
612 @type owned_nodes: iterable of string
613 @param owned_nodes: List of owned nodes
614 @type cur_group_uuid: string or None
615 @param cur_group_uuid: Optional group UUID to check against instance's groups
618 for (name, inst) in instances.items():
619 assert owned_nodes.issuperset(inst.all_nodes), \
620 "Instance %s's nodes changed while we kept the lock" % name
622 inst_groups = _CheckInstanceNodeGroups(cfg, name, owned_groups)
624 assert cur_group_uuid is None or cur_group_uuid in inst_groups, \
625 "Instance %s has no node in group %s" % (name, cur_group_uuid)
628 def _CheckInstanceNodeGroups(cfg, instance_name, owned_groups):
629 """Checks if the owned node groups are still correct for an instance.
631 @type cfg: L{config.ConfigWriter}
632 @param cfg: The cluster configuration
633 @type instance_name: string
634 @param instance_name: Instance name
635 @type owned_groups: set or frozenset
636 @param owned_groups: List of currently owned node groups
639 inst_groups = cfg.GetInstanceNodeGroups(instance_name)
641 if not owned_groups.issuperset(inst_groups):
642 raise errors.OpPrereqError("Instance %s's node groups changed since"
643 " locks were acquired, current groups are"
644 " are '%s', owning groups '%s'; retry the"
647 utils.CommaJoin(inst_groups),
648 utils.CommaJoin(owned_groups)),
654 def _CheckNodeGroupInstances(cfg, group_uuid, owned_instances):
655 """Checks if the instances in a node group are still correct.
657 @type cfg: L{config.ConfigWriter}
658 @param cfg: The cluster configuration
659 @type group_uuid: string
660 @param group_uuid: Node group UUID
661 @type owned_instances: set or frozenset
662 @param owned_instances: List of currently owned instances
665 wanted_instances = cfg.GetNodeGroupInstances(group_uuid)
666 if owned_instances != wanted_instances:
667 raise errors.OpPrereqError("Instances in node group '%s' changed since"
668 " locks were acquired, wanted '%s', have '%s';"
669 " retry the operation" %
671 utils.CommaJoin(wanted_instances),
672 utils.CommaJoin(owned_instances)),
675 return wanted_instances
678 def _SupportsOob(cfg, node):
679 """Tells if node supports OOB.
681 @type cfg: L{config.ConfigWriter}
682 @param cfg: The cluster configuration
683 @type node: L{objects.Node}
684 @param node: The node
685 @return: The OOB script if supported or an empty string otherwise
688 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
691 def _GetWantedNodes(lu, nodes):
692 """Returns list of checked and expanded node names.
694 @type lu: L{LogicalUnit}
695 @param lu: the logical unit on whose behalf we execute
697 @param nodes: list of node names or None for all nodes
699 @return: the list of nodes, sorted
700 @raise errors.ProgrammerError: if the nodes parameter is wrong type
704 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
706 return utils.NiceSort(lu.cfg.GetNodeList())
709 def _GetWantedInstances(lu, instances):
710 """Returns list of checked and expanded instance names.
712 @type lu: L{LogicalUnit}
713 @param lu: the logical unit on whose behalf we execute
714 @type instances: list
715 @param instances: list of instance names or None for all instances
717 @return: the list of instances, sorted
718 @raise errors.OpPrereqError: if the instances parameter is wrong type
719 @raise errors.OpPrereqError: if any of the passed instances is not found
723 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
725 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
729 def _GetUpdatedParams(old_params, update_dict,
730 use_default=True, use_none=False):
731 """Return the new version of a parameter dictionary.
733 @type old_params: dict
734 @param old_params: old parameters
735 @type update_dict: dict
736 @param update_dict: dict containing new parameter values, or
737 constants.VALUE_DEFAULT to reset the parameter to its default
739 @param use_default: boolean
740 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
741 values as 'to be deleted' values
742 @param use_none: boolean
743 @type use_none: whether to recognise C{None} values as 'to be
746 @return: the new parameter dictionary
749 params_copy = copy.deepcopy(old_params)
750 for key, val in update_dict.iteritems():
751 if ((use_default and val == constants.VALUE_DEFAULT) or
752 (use_none and val is None)):
758 params_copy[key] = val
762 def _GetUpdatedIPolicy(old_ipolicy, new_ipolicy, group_policy=False):
763 """Return the new version of a instance policy.
765 @param group_policy: whether this policy applies to a group and thus
766 we should support removal of policy entries
769 use_none = use_default = group_policy
770 ipolicy = copy.deepcopy(old_ipolicy)
771 for key, value in new_ipolicy.items():
772 if key not in constants.IPOLICY_ALL_KEYS:
773 raise errors.OpPrereqError("Invalid key in new ipolicy: %s" % key,
775 if key in constants.IPOLICY_ISPECS:
776 utils.ForceDictType(value, constants.ISPECS_PARAMETER_TYPES)
777 ipolicy[key] = _GetUpdatedParams(old_ipolicy.get(key, {}), value,
779 use_default=use_default)
781 if not value or value == [constants.VALUE_DEFAULT]:
785 raise errors.OpPrereqError("Can't unset ipolicy attribute '%s'"
786 " on the cluster'" % key,
789 if key in constants.IPOLICY_PARAMETERS:
790 # FIXME: we assume all such values are float
792 ipolicy[key] = float(value)
793 except (TypeError, ValueError), err:
794 raise errors.OpPrereqError("Invalid value for attribute"
795 " '%s': '%s', error: %s" %
796 (key, value, err), errors.ECODE_INVAL)
798 # FIXME: we assume all others are lists; this should be redone
800 ipolicy[key] = list(value)
802 objects.InstancePolicy.CheckParameterSyntax(ipolicy)
803 except errors.ConfigurationError, err:
804 raise errors.OpPrereqError("Invalid instance policy: %s" % err,
809 def _UpdateAndVerifySubDict(base, updates, type_check):
810 """Updates and verifies a dict with sub dicts of the same type.
812 @param base: The dict with the old data
813 @param updates: The dict with the new data
814 @param type_check: Dict suitable to ForceDictType to verify correct types
815 @returns: A new dict with updated and verified values
819 new = _GetUpdatedParams(old, value)
820 utils.ForceDictType(new, type_check)
823 ret = copy.deepcopy(base)
824 ret.update(dict((key, fn(base.get(key, {}), value))
825 for key, value in updates.items()))
829 def _MergeAndVerifyHvState(op_input, obj_input):
830 """Combines the hv state from an opcode with the one of the object
832 @param op_input: The input dict from the opcode
833 @param obj_input: The input dict from the objects
834 @return: The verified and updated dict
838 invalid_hvs = set(op_input) - constants.HYPER_TYPES
840 raise errors.OpPrereqError("Invalid hypervisor(s) in hypervisor state:"
841 " %s" % utils.CommaJoin(invalid_hvs),
843 if obj_input is None:
845 type_check = constants.HVSTS_PARAMETER_TYPES
846 return _UpdateAndVerifySubDict(obj_input, op_input, type_check)
851 def _MergeAndVerifyDiskState(op_input, obj_input):
852 """Combines the disk state from an opcode with the one of the object
854 @param op_input: The input dict from the opcode
855 @param obj_input: The input dict from the objects
856 @return: The verified and updated dict
859 invalid_dst = set(op_input) - constants.DS_VALID_TYPES
861 raise errors.OpPrereqError("Invalid storage type(s) in disk state: %s" %
862 utils.CommaJoin(invalid_dst),
864 type_check = constants.DSS_PARAMETER_TYPES
865 if obj_input is None:
867 return dict((key, _UpdateAndVerifySubDict(obj_input.get(key, {}), value,
869 for key, value in op_input.items())
874 def _ReleaseLocks(lu, level, names=None, keep=None):
875 """Releases locks owned by an LU.
877 @type lu: L{LogicalUnit}
878 @param level: Lock level
879 @type names: list or None
880 @param names: Names of locks to release
881 @type keep: list or None
882 @param keep: Names of locks to retain
885 assert not (keep is not None and names is not None), \
886 "Only one of the 'names' and the 'keep' parameters can be given"
888 if names is not None:
889 should_release = names.__contains__
891 should_release = lambda name: name not in keep
893 should_release = None
895 owned = lu.owned_locks(level)
897 # Not owning any lock at this level, do nothing
904 # Determine which locks to release
906 if should_release(name):
911 assert len(lu.owned_locks(level)) == (len(retain) + len(release))
913 # Release just some locks
914 lu.glm.release(level, names=release)
916 assert frozenset(lu.owned_locks(level)) == frozenset(retain)
919 lu.glm.release(level)
921 assert not lu.glm.is_owned(level), "No locks should be owned"
924 def _MapInstanceDisksToNodes(instances):
925 """Creates a map from (node, volume) to instance name.
927 @type instances: list of L{objects.Instance}
928 @rtype: dict; tuple of (node name, volume name) as key, instance name as value
931 return dict(((node, vol), inst.name)
932 for inst in instances
933 for (node, vols) in inst.MapLVsByNode().items()
937 def _RunPostHook(lu, node_name):
938 """Runs the post-hook for an opcode on a single node.
941 hm = lu.proc.BuildHooksManager(lu)
943 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
945 # pylint: disable=W0702
946 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
949 def _CheckOutputFields(static, dynamic, selected):
950 """Checks whether all selected fields are valid.
952 @type static: L{utils.FieldSet}
953 @param static: static fields set
954 @type dynamic: L{utils.FieldSet}
955 @param dynamic: dynamic fields set
962 delta = f.NonMatching(selected)
964 raise errors.OpPrereqError("Unknown output fields selected: %s"
965 % ",".join(delta), errors.ECODE_INVAL)
968 def _CheckGlobalHvParams(params):
969 """Validates that given hypervisor params are not global ones.
971 This will ensure that instances don't get customised versions of
975 used_globals = constants.HVC_GLOBALS.intersection(params)
977 msg = ("The following hypervisor parameters are global and cannot"
978 " be customized at instance level, please modify them at"
979 " cluster level: %s" % utils.CommaJoin(used_globals))
980 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
983 def _CheckNodeOnline(lu, node, msg=None):
984 """Ensure that a given node is online.
986 @param lu: the LU on behalf of which we make the check
987 @param node: the node to check
988 @param msg: if passed, should be a message to replace the default one
989 @raise errors.OpPrereqError: if the node is offline
993 msg = "Can't use offline node"
994 if lu.cfg.GetNodeInfo(node).offline:
995 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
998 def _CheckNodeNotDrained(lu, node):
999 """Ensure that a given node is not drained.
1001 @param lu: the LU on behalf of which we make the check
1002 @param node: the node to check
1003 @raise errors.OpPrereqError: if the node is drained
1006 if lu.cfg.GetNodeInfo(node).drained:
1007 raise errors.OpPrereqError("Can't use drained node %s" % node,
1011 def _CheckNodeVmCapable(lu, node):
1012 """Ensure that a given node is vm capable.
1014 @param lu: the LU on behalf of which we make the check
1015 @param node: the node to check
1016 @raise errors.OpPrereqError: if the node is not vm capable
1019 if not lu.cfg.GetNodeInfo(node).vm_capable:
1020 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
1024 def _CheckNodeHasOS(lu, node, os_name, force_variant):
1025 """Ensure that a node supports a given OS.
1027 @param lu: the LU on behalf of which we make the check
1028 @param node: the node to check
1029 @param os_name: the OS to query about
1030 @param force_variant: whether to ignore variant errors
1031 @raise errors.OpPrereqError: if the node is not supporting the OS
1034 result = lu.rpc.call_os_get(node, os_name)
1035 result.Raise("OS '%s' not in supported OS list for node %s" %
1037 prereq=True, ecode=errors.ECODE_INVAL)
1038 if not force_variant:
1039 _CheckOSVariant(result.payload, os_name)
1042 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
1043 """Ensure that a node has the given secondary ip.
1045 @type lu: L{LogicalUnit}
1046 @param lu: the LU on behalf of which we make the check
1048 @param node: the node to check
1049 @type secondary_ip: string
1050 @param secondary_ip: the ip to check
1051 @type prereq: boolean
1052 @param prereq: whether to throw a prerequisite or an execute error
1053 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
1054 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
1057 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
1058 result.Raise("Failure checking secondary ip on node %s" % node,
1059 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1060 if not result.payload:
1061 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
1062 " please fix and re-run this command" % secondary_ip)
1064 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
1066 raise errors.OpExecError(msg)
1069 def _GetClusterDomainSecret():
1070 """Reads the cluster domain secret.
1073 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
1077 def _CheckInstanceState(lu, instance, req_states, msg=None):
1078 """Ensure that an instance is in one of the required states.
1080 @param lu: the LU on behalf of which we make the check
1081 @param instance: the instance to check
1082 @param msg: if passed, should be a message to replace the default one
1083 @raise errors.OpPrereqError: if the instance is not in the required state
1087 msg = "can't use instance from outside %s states" % ", ".join(req_states)
1088 if instance.admin_state not in req_states:
1089 raise errors.OpPrereqError("Instance '%s' is marked to be %s, %s" %
1090 (instance.name, instance.admin_state, msg),
1093 if constants.ADMINST_UP not in req_states:
1094 pnode = instance.primary_node
1095 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
1096 ins_l.Raise("Can't contact node %s for instance information" % pnode,
1097 prereq=True, ecode=errors.ECODE_ENVIRON)
1099 if instance.name in ins_l.payload:
1100 raise errors.OpPrereqError("Instance %s is running, %s" %
1101 (instance.name, msg), errors.ECODE_STATE)
1104 def _ComputeMinMaxSpec(name, ipolicy, value):
1105 """Computes if value is in the desired range.
1107 @param name: name of the parameter for which we perform the check
1108 @param ipolicy: dictionary containing min, max and std values
1109 @param value: actual value that we want to use
1110 @return: None or element not meeting the criteria
1114 if value in [None, constants.VALUE_AUTO]:
1116 max_v = ipolicy[constants.ISPECS_MAX].get(name, value)
1117 min_v = ipolicy[constants.ISPECS_MIN].get(name, value)
1118 if value > max_v or min_v > value:
1119 return ("%s value %s is not in range [%s, %s]" %
1120 (name, value, min_v, max_v))
1124 def _ComputeIPolicySpecViolation(ipolicy, mem_size, cpu_count, disk_count,
1125 nic_count, disk_sizes, spindle_use,
1126 _compute_fn=_ComputeMinMaxSpec):
1127 """Verifies ipolicy against provided specs.
1130 @param ipolicy: The ipolicy
1132 @param mem_size: The memory size
1133 @type cpu_count: int
1134 @param cpu_count: Used cpu cores
1135 @type disk_count: int
1136 @param disk_count: Number of disks used
1137 @type nic_count: int
1138 @param nic_count: Number of nics used
1139 @type disk_sizes: list of ints
1140 @param disk_sizes: Disk sizes of used disk (len must match C{disk_count})
1141 @type spindle_use: int
1142 @param spindle_use: The number of spindles this instance uses
1143 @param _compute_fn: The compute function (unittest only)
1144 @return: A list of violations, or an empty list of no violations are found
1147 assert disk_count == len(disk_sizes)
1150 (constants.ISPEC_MEM_SIZE, mem_size),
1151 (constants.ISPEC_CPU_COUNT, cpu_count),
1152 (constants.ISPEC_DISK_COUNT, disk_count),
1153 (constants.ISPEC_NIC_COUNT, nic_count),
1154 (constants.ISPEC_SPINDLE_USE, spindle_use),
1155 ] + map((lambda d: (constants.ISPEC_DISK_SIZE, d)), disk_sizes)
1158 (_compute_fn(name, ipolicy, value)
1159 for (name, value) in test_settings))
1162 def _ComputeIPolicyInstanceViolation(ipolicy, instance,
1163 _compute_fn=_ComputeIPolicySpecViolation):
1164 """Compute if instance meets the specs of ipolicy.
1167 @param ipolicy: The ipolicy to verify against
1168 @type instance: L{objects.Instance}
1169 @param instance: The instance to verify
1170 @param _compute_fn: The function to verify ipolicy (unittest only)
1171 @see: L{_ComputeIPolicySpecViolation}
1174 mem_size = instance.beparams.get(constants.BE_MAXMEM, None)
1175 cpu_count = instance.beparams.get(constants.BE_VCPUS, None)
1176 spindle_use = instance.beparams.get(constants.BE_SPINDLE_USE, None)
1177 disk_count = len(instance.disks)
1178 disk_sizes = [disk.size for disk in instance.disks]
1179 nic_count = len(instance.nics)
1181 return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
1182 disk_sizes, spindle_use)
1185 def _ComputeIPolicyInstanceSpecViolation(ipolicy, instance_spec,
1186 _compute_fn=_ComputeIPolicySpecViolation):
1187 """Compute if instance specs meets the specs of ipolicy.
1190 @param ipolicy: The ipolicy to verify against
1191 @param instance_spec: dict
1192 @param instance_spec: The instance spec to verify
1193 @param _compute_fn: The function to verify ipolicy (unittest only)
1194 @see: L{_ComputeIPolicySpecViolation}
1197 mem_size = instance_spec.get(constants.ISPEC_MEM_SIZE, None)
1198 cpu_count = instance_spec.get(constants.ISPEC_CPU_COUNT, None)
1199 disk_count = instance_spec.get(constants.ISPEC_DISK_COUNT, 0)
1200 disk_sizes = instance_spec.get(constants.ISPEC_DISK_SIZE, [])
1201 nic_count = instance_spec.get(constants.ISPEC_NIC_COUNT, 0)
1202 spindle_use = instance_spec.get(constants.ISPEC_SPINDLE_USE, None)
1204 return _compute_fn(ipolicy, mem_size, cpu_count, disk_count, nic_count,
1205 disk_sizes, spindle_use)
1208 def _ComputeIPolicyNodeViolation(ipolicy, instance, current_group,
1210 _compute_fn=_ComputeIPolicyInstanceViolation):
1211 """Compute if instance meets the specs of the new target group.
1213 @param ipolicy: The ipolicy to verify
1214 @param instance: The instance object to verify
1215 @param current_group: The current group of the instance
1216 @param target_group: The new group of the instance
1217 @param _compute_fn: The function to verify ipolicy (unittest only)
1218 @see: L{_ComputeIPolicySpecViolation}
1221 if current_group == target_group:
1224 return _compute_fn(ipolicy, instance)
1227 def _CheckTargetNodeIPolicy(lu, ipolicy, instance, node, ignore=False,
1228 _compute_fn=_ComputeIPolicyNodeViolation):
1229 """Checks that the target node is correct in terms of instance policy.
1231 @param ipolicy: The ipolicy to verify
1232 @param instance: The instance object to verify
1233 @param node: The new node to relocate
1234 @param ignore: Ignore violations of the ipolicy
1235 @param _compute_fn: The function to verify ipolicy (unittest only)
1236 @see: L{_ComputeIPolicySpecViolation}
1239 primary_node = lu.cfg.GetNodeInfo(instance.primary_node)
1240 res = _compute_fn(ipolicy, instance, primary_node.group, node.group)
1243 msg = ("Instance does not meet target node group's (%s) instance"
1244 " policy: %s") % (node.group, utils.CommaJoin(res))
1248 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
1251 def _ComputeNewInstanceViolations(old_ipolicy, new_ipolicy, instances):
1252 """Computes a set of any instances that would violate the new ipolicy.
1254 @param old_ipolicy: The current (still in-place) ipolicy
1255 @param new_ipolicy: The new (to become) ipolicy
1256 @param instances: List of instances to verify
1257 @return: A list of instances which violates the new ipolicy but did not before
1260 return (_ComputeViolatingInstances(old_ipolicy, instances) -
1261 _ComputeViolatingInstances(new_ipolicy, instances))
1264 def _ExpandItemName(fn, name, kind):
1265 """Expand an item name.
1267 @param fn: the function to use for expansion
1268 @param name: requested item name
1269 @param kind: text description ('Node' or 'Instance')
1270 @return: the resolved (full) name
1271 @raise errors.OpPrereqError: if the item is not found
1274 full_name = fn(name)
1275 if full_name is None:
1276 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
1281 def _ExpandNodeName(cfg, name):
1282 """Wrapper over L{_ExpandItemName} for nodes."""
1283 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
1286 def _ExpandInstanceName(cfg, name):
1287 """Wrapper over L{_ExpandItemName} for instance."""
1288 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
1291 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
1292 minmem, maxmem, vcpus, nics, disk_template, disks,
1293 bep, hvp, hypervisor_name, tags):
1294 """Builds instance related env variables for hooks
1296 This builds the hook environment from individual variables.
1299 @param name: the name of the instance
1300 @type primary_node: string
1301 @param primary_node: the name of the instance's primary node
1302 @type secondary_nodes: list
1303 @param secondary_nodes: list of secondary nodes as strings
1304 @type os_type: string
1305 @param os_type: the name of the instance's OS
1306 @type status: string
1307 @param status: the desired status of the instance
1308 @type minmem: string
1309 @param minmem: the minimum memory size of the instance
1310 @type maxmem: string
1311 @param maxmem: the maximum memory size of the instance
1313 @param vcpus: the count of VCPUs the instance has
1315 @param nics: list of tuples (ip, mac, mode, link) representing
1316 the NICs the instance has
1317 @type disk_template: string
1318 @param disk_template: the disk template of the instance
1320 @param disks: the list of (size, mode) pairs
1322 @param bep: the backend parameters for the instance
1324 @param hvp: the hypervisor parameters for the instance
1325 @type hypervisor_name: string
1326 @param hypervisor_name: the hypervisor for the instance
1328 @param tags: list of instance tags as strings
1330 @return: the hook environment for this instance
1335 "INSTANCE_NAME": name,
1336 "INSTANCE_PRIMARY": primary_node,
1337 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
1338 "INSTANCE_OS_TYPE": os_type,
1339 "INSTANCE_STATUS": status,
1340 "INSTANCE_MINMEM": minmem,
1341 "INSTANCE_MAXMEM": maxmem,
1342 # TODO(2.7) remove deprecated "memory" value
1343 "INSTANCE_MEMORY": maxmem,
1344 "INSTANCE_VCPUS": vcpus,
1345 "INSTANCE_DISK_TEMPLATE": disk_template,
1346 "INSTANCE_HYPERVISOR": hypervisor_name,
1349 nic_count = len(nics)
1350 for idx, (ip, mac, mode, link) in enumerate(nics):
1353 env["INSTANCE_NIC%d_IP" % idx] = ip
1354 env["INSTANCE_NIC%d_MAC" % idx] = mac
1355 env["INSTANCE_NIC%d_MODE" % idx] = mode
1356 env["INSTANCE_NIC%d_LINK" % idx] = link
1357 if mode == constants.NIC_MODE_BRIDGED:
1358 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
1362 env["INSTANCE_NIC_COUNT"] = nic_count
1365 disk_count = len(disks)
1366 for idx, (size, mode) in enumerate(disks):
1367 env["INSTANCE_DISK%d_SIZE" % idx] = size
1368 env["INSTANCE_DISK%d_MODE" % idx] = mode
1372 env["INSTANCE_DISK_COUNT"] = disk_count
1377 env["INSTANCE_TAGS"] = " ".join(tags)
1379 for source, kind in [(bep, "BE"), (hvp, "HV")]:
1380 for key, value in source.items():
1381 env["INSTANCE_%s_%s" % (kind, key)] = value
1386 def _NICListToTuple(lu, nics):
1387 """Build a list of nic information tuples.
1389 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
1390 value in LUInstanceQueryData.
1392 @type lu: L{LogicalUnit}
1393 @param lu: the logical unit on whose behalf we execute
1394 @type nics: list of L{objects.NIC}
1395 @param nics: list of nics to convert to hooks tuples
1399 cluster = lu.cfg.GetClusterInfo()
1403 filled_params = cluster.SimpleFillNIC(nic.nicparams)
1404 mode = filled_params[constants.NIC_MODE]
1405 link = filled_params[constants.NIC_LINK]
1406 hooks_nics.append((ip, mac, mode, link))
1410 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1411 """Builds instance related env variables for hooks from an object.
1413 @type lu: L{LogicalUnit}
1414 @param lu: the logical unit on whose behalf we execute
1415 @type instance: L{objects.Instance}
1416 @param instance: the instance for which we should build the
1418 @type override: dict
1419 @param override: dictionary with key/values that will override
1422 @return: the hook environment dictionary
1425 cluster = lu.cfg.GetClusterInfo()
1426 bep = cluster.FillBE(instance)
1427 hvp = cluster.FillHV(instance)
1429 "name": instance.name,
1430 "primary_node": instance.primary_node,
1431 "secondary_nodes": instance.secondary_nodes,
1432 "os_type": instance.os,
1433 "status": instance.admin_state,
1434 "maxmem": bep[constants.BE_MAXMEM],
1435 "minmem": bep[constants.BE_MINMEM],
1436 "vcpus": bep[constants.BE_VCPUS],
1437 "nics": _NICListToTuple(lu, instance.nics),
1438 "disk_template": instance.disk_template,
1439 "disks": [(disk.size, disk.mode) for disk in instance.disks],
1442 "hypervisor_name": instance.hypervisor,
1443 "tags": instance.tags,
1446 args.update(override)
1447 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
1450 def _AdjustCandidatePool(lu, exceptions):
1451 """Adjust the candidate pool after node operations.
1454 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1456 lu.LogInfo("Promoted nodes to master candidate role: %s",
1457 utils.CommaJoin(node.name for node in mod_list))
1458 for name in mod_list:
1459 lu.context.ReaddNode(name)
1460 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1462 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1466 def _DecideSelfPromotion(lu, exceptions=None):
1467 """Decide whether I should promote myself as a master candidate.
1470 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1471 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1472 # the new node will increase mc_max with one, so:
1473 mc_should = min(mc_should + 1, cp_size)
1474 return mc_now < mc_should
1477 def _CalculateGroupIPolicy(cluster, group):
1478 """Calculate instance policy for group.
1481 return cluster.SimpleFillIPolicy(group.ipolicy)
1484 def _ComputeViolatingInstances(ipolicy, instances):
1485 """Computes a set of instances who violates given ipolicy.
1487 @param ipolicy: The ipolicy to verify
1488 @type instances: object.Instance
1489 @param instances: List of instances to verify
1490 @return: A frozenset of instance names violating the ipolicy
1493 return frozenset([inst.name for inst in instances
1494 if _ComputeIPolicyInstanceViolation(ipolicy, inst)])
1497 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1498 """Check that the brigdes needed by a list of nics exist.
1501 cluster = lu.cfg.GetClusterInfo()
1502 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1503 brlist = [params[constants.NIC_LINK] for params in paramslist
1504 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1506 result = lu.rpc.call_bridges_exist(target_node, brlist)
1507 result.Raise("Error checking bridges on destination node '%s'" %
1508 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1511 def _CheckInstanceBridgesExist(lu, instance, node=None):
1512 """Check that the brigdes needed by an instance exist.
1516 node = instance.primary_node
1517 _CheckNicsBridgesExist(lu, instance.nics, node)
1520 def _CheckOSVariant(os_obj, name):
1521 """Check whether an OS name conforms to the os variants specification.
1523 @type os_obj: L{objects.OS}
1524 @param os_obj: OS object to check
1526 @param name: OS name passed by the user, to check for validity
1529 variant = objects.OS.GetVariant(name)
1530 if not os_obj.supported_variants:
1532 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'"
1533 " passed)" % (os_obj.name, variant),
1537 raise errors.OpPrereqError("OS name must include a variant",
1540 if variant not in os_obj.supported_variants:
1541 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1544 def _GetNodeInstancesInner(cfg, fn):
1545 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1548 def _GetNodeInstances(cfg, node_name):
1549 """Returns a list of all primary and secondary instances on a node.
1553 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1556 def _GetNodePrimaryInstances(cfg, node_name):
1557 """Returns primary instances on a node.
1560 return _GetNodeInstancesInner(cfg,
1561 lambda inst: node_name == inst.primary_node)
1564 def _GetNodeSecondaryInstances(cfg, node_name):
1565 """Returns secondary instances on a node.
1568 return _GetNodeInstancesInner(cfg,
1569 lambda inst: node_name in inst.secondary_nodes)
1572 def _GetStorageTypeArgs(cfg, storage_type):
1573 """Returns the arguments for a storage type.
1576 # Special case for file storage
1577 if storage_type == constants.ST_FILE:
1578 # storage.FileStorage wants a list of storage directories
1579 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1584 def _FindFaultyInstanceDisks(cfg, rpc_runner, instance, node_name, prereq):
1587 for dev in instance.disks:
1588 cfg.SetDiskID(dev, node_name)
1590 result = rpc_runner.call_blockdev_getmirrorstatus(node_name, instance.disks)
1591 result.Raise("Failed to get disk status from node %s" % node_name,
1592 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1594 for idx, bdev_status in enumerate(result.payload):
1595 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1601 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1602 """Check the sanity of iallocator and node arguments and use the
1603 cluster-wide iallocator if appropriate.
1605 Check that at most one of (iallocator, node) is specified. If none is
1606 specified, then the LU's opcode's iallocator slot is filled with the
1607 cluster-wide default iallocator.
1609 @type iallocator_slot: string
1610 @param iallocator_slot: the name of the opcode iallocator slot
1611 @type node_slot: string
1612 @param node_slot: the name of the opcode target node slot
1615 node = getattr(lu.op, node_slot, None)
1616 iallocator = getattr(lu.op, iallocator_slot, None)
1618 if node is not None and iallocator is not None:
1619 raise errors.OpPrereqError("Do not specify both, iallocator and node",
1621 elif node is None and iallocator is None:
1622 default_iallocator = lu.cfg.GetDefaultIAllocator()
1623 if default_iallocator:
1624 setattr(lu.op, iallocator_slot, default_iallocator)
1626 raise errors.OpPrereqError("No iallocator or node given and no"
1627 " cluster-wide default iallocator found;"
1628 " please specify either an iallocator or a"
1629 " node, or set a cluster-wide default"
1633 def _GetDefaultIAllocator(cfg, iallocator):
1634 """Decides on which iallocator to use.
1636 @type cfg: L{config.ConfigWriter}
1637 @param cfg: Cluster configuration object
1638 @type iallocator: string or None
1639 @param iallocator: Iallocator specified in opcode
1641 @return: Iallocator name
1645 # Use default iallocator
1646 iallocator = cfg.GetDefaultIAllocator()
1649 raise errors.OpPrereqError("No iallocator was specified, neither in the"
1650 " opcode nor as a cluster-wide default",
1656 class LUClusterPostInit(LogicalUnit):
1657 """Logical unit for running hooks after cluster initialization.
1660 HPATH = "cluster-init"
1661 HTYPE = constants.HTYPE_CLUSTER
1663 def BuildHooksEnv(self):
1668 "OP_TARGET": self.cfg.GetClusterName(),
1671 def BuildHooksNodes(self):
1672 """Build hooks nodes.
1675 return ([], [self.cfg.GetMasterNode()])
1677 def Exec(self, feedback_fn):
1684 class LUClusterDestroy(LogicalUnit):
1685 """Logical unit for destroying the cluster.
1688 HPATH = "cluster-destroy"
1689 HTYPE = constants.HTYPE_CLUSTER
1691 def BuildHooksEnv(self):
1696 "OP_TARGET": self.cfg.GetClusterName(),
1699 def BuildHooksNodes(self):
1700 """Build hooks nodes.
1705 def CheckPrereq(self):
1706 """Check prerequisites.
1708 This checks whether the cluster is empty.
1710 Any errors are signaled by raising errors.OpPrereqError.
1713 master = self.cfg.GetMasterNode()
1715 nodelist = self.cfg.GetNodeList()
1716 if len(nodelist) != 1 or nodelist[0] != master:
1717 raise errors.OpPrereqError("There are still %d node(s) in"
1718 " this cluster." % (len(nodelist) - 1),
1720 instancelist = self.cfg.GetInstanceList()
1722 raise errors.OpPrereqError("There are still %d instance(s) in"
1723 " this cluster." % len(instancelist),
1726 def Exec(self, feedback_fn):
1727 """Destroys the cluster.
1730 master_params = self.cfg.GetMasterNetworkParameters()
1732 # Run post hooks on master node before it's removed
1733 _RunPostHook(self, master_params.name)
1735 ems = self.cfg.GetUseExternalMipScript()
1736 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
1739 self.LogWarning("Error disabling the master IP address: %s",
1742 return master_params.name
1745 def _VerifyCertificate(filename):
1746 """Verifies a certificate for L{LUClusterVerifyConfig}.
1748 @type filename: string
1749 @param filename: Path to PEM file
1753 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1754 utils.ReadFile(filename))
1755 except Exception, err: # pylint: disable=W0703
1756 return (LUClusterVerifyConfig.ETYPE_ERROR,
1757 "Failed to load X509 certificate %s: %s" % (filename, err))
1760 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1761 constants.SSL_CERT_EXPIRATION_ERROR)
1764 fnamemsg = "While verifying %s: %s" % (filename, msg)
1769 return (None, fnamemsg)
1770 elif errcode == utils.CERT_WARNING:
1771 return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg)
1772 elif errcode == utils.CERT_ERROR:
1773 return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg)
1775 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1778 def _GetAllHypervisorParameters(cluster, instances):
1779 """Compute the set of all hypervisor parameters.
1781 @type cluster: L{objects.Cluster}
1782 @param cluster: the cluster object
1783 @param instances: list of L{objects.Instance}
1784 @param instances: additional instances from which to obtain parameters
1785 @rtype: list of (origin, hypervisor, parameters)
1786 @return: a list with all parameters found, indicating the hypervisor they
1787 apply to, and the origin (can be "cluster", "os X", or "instance Y")
1792 for hv_name in cluster.enabled_hypervisors:
1793 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
1795 for os_name, os_hvp in cluster.os_hvp.items():
1796 for hv_name, hv_params in os_hvp.items():
1798 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
1799 hvp_data.append(("os %s" % os_name, hv_name, full_params))
1801 # TODO: collapse identical parameter values in a single one
1802 for instance in instances:
1803 if instance.hvparams:
1804 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
1805 cluster.FillHV(instance)))
1810 class _VerifyErrors(object):
1811 """Mix-in for cluster/group verify LUs.
1813 It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects
1814 self.op and self._feedback_fn to be available.)
1818 ETYPE_FIELD = "code"
1819 ETYPE_ERROR = "ERROR"
1820 ETYPE_WARNING = "WARNING"
1822 def _Error(self, ecode, item, msg, *args, **kwargs):
1823 """Format an error message.
1825 Based on the opcode's error_codes parameter, either format a
1826 parseable error code, or a simpler error string.
1828 This must be called only from Exec and functions called from Exec.
1831 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1832 itype, etxt, _ = ecode
1833 # first complete the msg
1836 # then format the whole message
1837 if self.op.error_codes: # This is a mix-in. pylint: disable=E1101
1838 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1844 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1845 # and finally report it via the feedback_fn
1846 self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable=E1101
1848 def _ErrorIf(self, cond, ecode, *args, **kwargs):
1849 """Log an error message if the passed condition is True.
1853 or self.op.debug_simulate_errors) # pylint: disable=E1101
1855 # If the error code is in the list of ignored errors, demote the error to a
1857 (_, etxt, _) = ecode
1858 if etxt in self.op.ignore_errors: # pylint: disable=E1101
1859 kwargs[self.ETYPE_FIELD] = self.ETYPE_WARNING
1862 self._Error(ecode, *args, **kwargs)
1864 # do not mark the operation as failed for WARN cases only
1865 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1866 self.bad = self.bad or cond
1869 class LUClusterVerify(NoHooksLU):
1870 """Submits all jobs necessary to verify the cluster.
1875 def ExpandNames(self):
1876 self.needed_locks = {}
1878 def Exec(self, feedback_fn):
1881 if self.op.group_name:
1882 groups = [self.op.group_name]
1883 depends_fn = lambda: None
1885 groups = self.cfg.GetNodeGroupList()
1887 # Verify global configuration
1889 opcodes.OpClusterVerifyConfig(ignore_errors=self.op.ignore_errors)
1892 # Always depend on global verification
1893 depends_fn = lambda: [(-len(jobs), [])]
1895 jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,
1896 ignore_errors=self.op.ignore_errors,
1897 depends=depends_fn())]
1898 for group in groups)
1900 # Fix up all parameters
1901 for op in itertools.chain(*jobs): # pylint: disable=W0142
1902 op.debug_simulate_errors = self.op.debug_simulate_errors
1903 op.verbose = self.op.verbose
1904 op.error_codes = self.op.error_codes
1906 op.skip_checks = self.op.skip_checks
1907 except AttributeError:
1908 assert not isinstance(op, opcodes.OpClusterVerifyGroup)
1910 return ResultWithJobs(jobs)
1913 class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
1914 """Verifies the cluster config.
1919 def _VerifyHVP(self, hvp_data):
1920 """Verifies locally the syntax of the hypervisor parameters.
1923 for item, hv_name, hv_params in hvp_data:
1924 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
1927 hv_class = hypervisor.GetHypervisor(hv_name)
1928 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1929 hv_class.CheckParameterSyntax(hv_params)
1930 except errors.GenericError, err:
1931 self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg % str(err))
1933 def ExpandNames(self):
1934 self.needed_locks = dict.fromkeys(locking.LEVELS, locking.ALL_SET)
1935 self.share_locks = _ShareAll()
1937 def CheckPrereq(self):
1938 """Check prerequisites.
1941 # Retrieve all information
1942 self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
1943 self.all_node_info = self.cfg.GetAllNodesInfo()
1944 self.all_inst_info = self.cfg.GetAllInstancesInfo()
1946 def Exec(self, feedback_fn):
1947 """Verify integrity of cluster, performing various test on nodes.
1951 self._feedback_fn = feedback_fn
1953 feedback_fn("* Verifying cluster config")
1955 for msg in self.cfg.VerifyConfig():
1956 self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg)
1958 feedback_fn("* Verifying cluster certificate files")
1960 for cert_filename in constants.ALL_CERT_FILES:
1961 (errcode, msg) = _VerifyCertificate(cert_filename)
1962 self._ErrorIf(errcode, constants.CV_ECLUSTERCERT, None, msg, code=errcode)
1964 feedback_fn("* Verifying hypervisor parameters")
1966 self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(),
1967 self.all_inst_info.values()))
1969 feedback_fn("* Verifying all nodes belong to an existing group")
1971 # We do this verification here because, should this bogus circumstance
1972 # occur, it would never be caught by VerifyGroup, which only acts on
1973 # nodes/instances reachable from existing node groups.
1975 dangling_nodes = set(node.name for node in self.all_node_info.values()
1976 if node.group not in self.all_group_info)
1978 dangling_instances = {}
1979 no_node_instances = []
1981 for inst in self.all_inst_info.values():
1982 if inst.primary_node in dangling_nodes:
1983 dangling_instances.setdefault(inst.primary_node, []).append(inst.name)
1984 elif inst.primary_node not in self.all_node_info:
1985 no_node_instances.append(inst.name)
1990 utils.CommaJoin(dangling_instances.get(node.name,
1992 for node in dangling_nodes]
1994 self._ErrorIf(bool(dangling_nodes), constants.CV_ECLUSTERDANGLINGNODES,
1996 "the following nodes (and their instances) belong to a non"
1997 " existing group: %s", utils.CommaJoin(pretty_dangling))
1999 self._ErrorIf(bool(no_node_instances), constants.CV_ECLUSTERDANGLINGINST,
2001 "the following instances have a non-existing primary-node:"
2002 " %s", utils.CommaJoin(no_node_instances))
2007 class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
2008 """Verifies the status of a node group.
2011 HPATH = "cluster-verify"
2012 HTYPE = constants.HTYPE_CLUSTER
2015 _HOOKS_INDENT_RE = re.compile("^", re.M)
2017 class NodeImage(object):
2018 """A class representing the logical and physical status of a node.
2021 @ivar name: the node name to which this object refers
2022 @ivar volumes: a structure as returned from
2023 L{ganeti.backend.GetVolumeList} (runtime)
2024 @ivar instances: a list of running instances (runtime)
2025 @ivar pinst: list of configured primary instances (config)
2026 @ivar sinst: list of configured secondary instances (config)
2027 @ivar sbp: dictionary of {primary-node: list of instances} for all
2028 instances for which this node is secondary (config)
2029 @ivar mfree: free memory, as reported by hypervisor (runtime)
2030 @ivar dfree: free disk, as reported by the node (runtime)
2031 @ivar offline: the offline status (config)
2032 @type rpc_fail: boolean
2033 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
2034 not whether the individual keys were correct) (runtime)
2035 @type lvm_fail: boolean
2036 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
2037 @type hyp_fail: boolean
2038 @ivar hyp_fail: whether the RPC call didn't return the instance list
2039 @type ghost: boolean
2040 @ivar ghost: whether this is a known node or not (config)
2041 @type os_fail: boolean
2042 @ivar os_fail: whether the RPC call didn't return valid OS data
2044 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
2045 @type vm_capable: boolean
2046 @ivar vm_capable: whether the node can host instances
2049 def __init__(self, offline=False, name=None, vm_capable=True):
2058 self.offline = offline
2059 self.vm_capable = vm_capable
2060 self.rpc_fail = False
2061 self.lvm_fail = False
2062 self.hyp_fail = False
2064 self.os_fail = False
2067 def ExpandNames(self):
2068 # This raises errors.OpPrereqError on its own:
2069 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
2071 # Get instances in node group; this is unsafe and needs verification later
2073 self.cfg.GetNodeGroupInstances(self.group_uuid, primary_only=True)
2075 self.needed_locks = {
2076 locking.LEVEL_INSTANCE: inst_names,
2077 locking.LEVEL_NODEGROUP: [self.group_uuid],
2078 locking.LEVEL_NODE: [],
2081 self.share_locks = _ShareAll()
2083 def DeclareLocks(self, level):
2084 if level == locking.LEVEL_NODE:
2085 # Get members of node group; this is unsafe and needs verification later
2086 nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members)
2088 all_inst_info = self.cfg.GetAllInstancesInfo()
2090 # In Exec(), we warn about mirrored instances that have primary and
2091 # secondary living in separate node groups. To fully verify that
2092 # volumes for these instances are healthy, we will need to do an
2093 # extra call to their secondaries. We ensure here those nodes will
2095 for inst in self.owned_locks(locking.LEVEL_INSTANCE):
2096 # Important: access only the instances whose lock is owned
2097 if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR:
2098 nodes.update(all_inst_info[inst].secondary_nodes)
2100 self.needed_locks[locking.LEVEL_NODE] = nodes
2102 def CheckPrereq(self):
2103 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
2104 self.group_info = self.cfg.GetNodeGroup(self.group_uuid)
2106 group_nodes = set(self.group_info.members)
2108 self.cfg.GetNodeGroupInstances(self.group_uuid, primary_only=True)
2111 group_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
2113 unlocked_instances = \
2114 group_instances.difference(self.owned_locks(locking.LEVEL_INSTANCE))
2117 raise errors.OpPrereqError("Missing lock for nodes: %s" %
2118 utils.CommaJoin(unlocked_nodes),
2121 if unlocked_instances:
2122 raise errors.OpPrereqError("Missing lock for instances: %s" %
2123 utils.CommaJoin(unlocked_instances),
2126 self.all_node_info = self.cfg.GetAllNodesInfo()
2127 self.all_inst_info = self.cfg.GetAllInstancesInfo()
2129 self.my_node_names = utils.NiceSort(group_nodes)
2130 self.my_inst_names = utils.NiceSort(group_instances)
2132 self.my_node_info = dict((name, self.all_node_info[name])
2133 for name in self.my_node_names)
2135 self.my_inst_info = dict((name, self.all_inst_info[name])
2136 for name in self.my_inst_names)
2138 # We detect here the nodes that will need the extra RPC calls for verifying
2139 # split LV volumes; they should be locked.
2140 extra_lv_nodes = set()
2142 for inst in self.my_inst_info.values():
2143 if inst.disk_template in constants.DTS_INT_MIRROR:
2144 for nname in inst.all_nodes:
2145 if self.all_node_info[nname].group != self.group_uuid:
2146 extra_lv_nodes.add(nname)
2148 unlocked_lv_nodes = \
2149 extra_lv_nodes.difference(self.owned_locks(locking.LEVEL_NODE))
2151 if unlocked_lv_nodes:
2152 raise errors.OpPrereqError("Missing node locks for LV check: %s" %
2153 utils.CommaJoin(unlocked_lv_nodes),
2155 self.extra_lv_nodes = list(extra_lv_nodes)
2157 def _VerifyNode(self, ninfo, nresult):
2158 """Perform some basic validation on data returned from a node.
2160 - check the result data structure is well formed and has all the
2162 - check ganeti version
2164 @type ninfo: L{objects.Node}
2165 @param ninfo: the node to check
2166 @param nresult: the results from the node
2168 @return: whether overall this call was successful (and we can expect
2169 reasonable values in the respose)
2173 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2175 # main result, nresult should be a non-empty dict
2176 test = not nresult or not isinstance(nresult, dict)
2177 _ErrorIf(test, constants.CV_ENODERPC, node,
2178 "unable to verify node: no data returned")
2182 # compares ganeti version
2183 local_version = constants.PROTOCOL_VERSION
2184 remote_version = nresult.get("version", None)
2185 test = not (remote_version and
2186 isinstance(remote_version, (list, tuple)) and
2187 len(remote_version) == 2)
2188 _ErrorIf(test, constants.CV_ENODERPC, node,
2189 "connection to node returned invalid data")
2193 test = local_version != remote_version[0]
2194 _ErrorIf(test, constants.CV_ENODEVERSION, node,
2195 "incompatible protocol versions: master %s,"
2196 " node %s", local_version, remote_version[0])
2200 # node seems compatible, we can actually try to look into its results
2202 # full package version
2203 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
2204 constants.CV_ENODEVERSION, node,
2205 "software version mismatch: master %s, node %s",
2206 constants.RELEASE_VERSION, remote_version[1],
2207 code=self.ETYPE_WARNING)
2209 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
2210 if ninfo.vm_capable and isinstance(hyp_result, dict):
2211 for hv_name, hv_result in hyp_result.iteritems():
2212 test = hv_result is not None
2213 _ErrorIf(test, constants.CV_ENODEHV, node,
2214 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
2216 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
2217 if ninfo.vm_capable and isinstance(hvp_result, list):
2218 for item, hv_name, hv_result in hvp_result:
2219 _ErrorIf(True, constants.CV_ENODEHV, node,
2220 "hypervisor %s parameter verify failure (source %s): %s",
2221 hv_name, item, hv_result)
2223 test = nresult.get(constants.NV_NODESETUP,
2224 ["Missing NODESETUP results"])
2225 _ErrorIf(test, constants.CV_ENODESETUP, node, "node setup error: %s",
2230 def _VerifyNodeTime(self, ninfo, nresult,
2231 nvinfo_starttime, nvinfo_endtime):
2232 """Check the node time.
2234 @type ninfo: L{objects.Node}
2235 @param ninfo: the node to check
2236 @param nresult: the remote results for the node
2237 @param nvinfo_starttime: the start time of the RPC call
2238 @param nvinfo_endtime: the end time of the RPC call
2242 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2244 ntime = nresult.get(constants.NV_TIME, None)
2246 ntime_merged = utils.MergeTime(ntime)
2247 except (ValueError, TypeError):
2248 _ErrorIf(True, constants.CV_ENODETIME, node, "Node returned invalid time")
2251 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
2252 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
2253 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
2254 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
2258 _ErrorIf(ntime_diff is not None, constants.CV_ENODETIME, node,
2259 "Node time diverges by at least %s from master node time",
2262 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
2263 """Check the node LVM results.
2265 @type ninfo: L{objects.Node}
2266 @param ninfo: the node to check
2267 @param nresult: the remote results for the node
2268 @param vg_name: the configured VG name
2275 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2277 # checks vg existence and size > 20G
2278 vglist = nresult.get(constants.NV_VGLIST, None)
2280 _ErrorIf(test, constants.CV_ENODELVM, node, "unable to check volume groups")
2282 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
2283 constants.MIN_VG_SIZE)
2284 _ErrorIf(vgstatus, constants.CV_ENODELVM, node, vgstatus)
2287 pvlist = nresult.get(constants.NV_PVLIST, None)
2288 test = pvlist is None
2289 _ErrorIf(test, constants.CV_ENODELVM, node, "Can't get PV list from node")
2291 # check that ':' is not present in PV names, since it's a
2292 # special character for lvcreate (denotes the range of PEs to
2294 for _, pvname, owner_vg in pvlist:
2295 test = ":" in pvname
2296 _ErrorIf(test, constants.CV_ENODELVM, node,
2297 "Invalid character ':' in PV '%s' of VG '%s'",
2300 def _VerifyNodeBridges(self, ninfo, nresult, bridges):
2301 """Check the node bridges.
2303 @type ninfo: L{objects.Node}
2304 @param ninfo: the node to check
2305 @param nresult: the remote results for the node
2306 @param bridges: the expected list of bridges
2313 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2315 missing = nresult.get(constants.NV_BRIDGES, None)
2316 test = not isinstance(missing, list)
2317 _ErrorIf(test, constants.CV_ENODENET, node,
2318 "did not return valid bridge information")
2320 _ErrorIf(bool(missing), constants.CV_ENODENET, node,
2321 "missing bridges: %s" % utils.CommaJoin(sorted(missing)))
2323 def _VerifyNodeUserScripts(self, ninfo, nresult):
2324 """Check the results of user scripts presence and executability on the node
2326 @type ninfo: L{objects.Node}
2327 @param ninfo: the node to check
2328 @param nresult: the remote results for the node
2333 test = not constants.NV_USERSCRIPTS in nresult
2334 self._ErrorIf(test, constants.CV_ENODEUSERSCRIPTS, node,
2335 "did not return user scripts information")
2337 broken_scripts = nresult.get(constants.NV_USERSCRIPTS, None)
2339 self._ErrorIf(broken_scripts, constants.CV_ENODEUSERSCRIPTS, node,
2340 "user scripts not present or not executable: %s" %
2341 utils.CommaJoin(sorted(broken_scripts)))
2343 def _VerifyNodeNetwork(self, ninfo, nresult):
2344 """Check the node network connectivity results.
2346 @type ninfo: L{objects.Node}
2347 @param ninfo: the node to check
2348 @param nresult: the remote results for the node
2352 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2354 test = constants.NV_NODELIST not in nresult
2355 _ErrorIf(test, constants.CV_ENODESSH, node,
2356 "node hasn't returned node ssh connectivity data")
2358 if nresult[constants.NV_NODELIST]:
2359 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
2360 _ErrorIf(True, constants.CV_ENODESSH, node,
2361 "ssh communication with node '%s': %s", a_node, a_msg)
2363 test = constants.NV_NODENETTEST not in nresult
2364 _ErrorIf(test, constants.CV_ENODENET, node,
2365 "node hasn't returned node tcp connectivity data")
2367 if nresult[constants.NV_NODENETTEST]:
2368 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
2370 _ErrorIf(True, constants.CV_ENODENET, node,
2371 "tcp communication with node '%s': %s",
2372 anode, nresult[constants.NV_NODENETTEST][anode])
2374 test = constants.NV_MASTERIP not in nresult
2375 _ErrorIf(test, constants.CV_ENODENET, node,
2376 "node hasn't returned node master IP reachability data")
2378 if not nresult[constants.NV_MASTERIP]:
2379 if node == self.master_node:
2380 msg = "the master node cannot reach the master IP (not configured?)"
2382 msg = "cannot reach the master IP"
2383 _ErrorIf(True, constants.CV_ENODENET, node, msg)
2385 def _VerifyInstance(self, instance, instanceconfig, node_image,
2387 """Verify an instance.
2389 This function checks to see if the required block devices are
2390 available on the instance's node.
2393 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2394 node_current = instanceconfig.primary_node
2396 node_vol_should = {}
2397 instanceconfig.MapLVsByNode(node_vol_should)
2399 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(), self.group_info)
2400 err = _ComputeIPolicyInstanceViolation(ipolicy, instanceconfig)
2401 _ErrorIf(err, constants.CV_EINSTANCEPOLICY, instance, err)
2403 for node in node_vol_should:
2404 n_img = node_image[node]
2405 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
2406 # ignore missing volumes on offline or broken nodes
2408 for volume in node_vol_should[node]:
2409 test = volume not in n_img.volumes
2410 _ErrorIf(test, constants.CV_EINSTANCEMISSINGDISK, instance,
2411 "volume %s missing on node %s", volume, node)
2413 if instanceconfig.admin_state == constants.ADMINST_UP:
2414 pri_img = node_image[node_current]
2415 test = instance not in pri_img.instances and not pri_img.offline
2416 _ErrorIf(test, constants.CV_EINSTANCEDOWN, instance,
2417 "instance not running on its primary node %s",
2420 diskdata = [(nname, success, status, idx)
2421 for (nname, disks) in diskstatus.items()
2422 for idx, (success, status) in enumerate(disks)]
2424 for nname, success, bdev_status, idx in diskdata:
2425 # the 'ghost node' construction in Exec() ensures that we have a
2427 snode = node_image[nname]
2428 bad_snode = snode.ghost or snode.offline
2429 _ErrorIf(instanceconfig.admin_state == constants.ADMINST_UP and
2430 not success and not bad_snode,
2431 constants.CV_EINSTANCEFAULTYDISK, instance,
2432 "couldn't retrieve status for disk/%s on %s: %s",
2433 idx, nname, bdev_status)
2434 _ErrorIf((instanceconfig.admin_state == constants.ADMINST_UP and
2435 success and bdev_status.ldisk_status == constants.LDS_FAULTY),
2436 constants.CV_EINSTANCEFAULTYDISK, instance,
2437 "disk/%s on %s is faulty", idx, nname)
2439 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
2440 """Verify if there are any unknown volumes in the cluster.
2442 The .os, .swap and backup volumes are ignored. All other volumes are
2443 reported as unknown.
2445 @type reserved: L{ganeti.utils.FieldSet}
2446 @param reserved: a FieldSet of reserved volume names
2449 for node, n_img in node_image.items():
2450 if (n_img.offline or n_img.rpc_fail or n_img.lvm_fail or
2451 self.all_node_info[node].group != self.group_uuid):
2452 # skip non-healthy nodes
2454 for volume in n_img.volumes:
2455 test = ((node not in node_vol_should or
2456 volume not in node_vol_should[node]) and
2457 not reserved.Matches(volume))
2458 self._ErrorIf(test, constants.CV_ENODEORPHANLV, node,
2459 "volume %s is unknown", volume)
2461 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
2462 """Verify N+1 Memory Resilience.
2464 Check that if one single node dies we can still start all the
2465 instances it was primary for.
2468 cluster_info = self.cfg.GetClusterInfo()
2469 for node, n_img in node_image.items():
2470 # This code checks that every node which is now listed as
2471 # secondary has enough memory to host all instances it is
2472 # supposed to should a single other node in the cluster fail.
2473 # FIXME: not ready for failover to an arbitrary node
2474 # FIXME: does not support file-backed instances
2475 # WARNING: we currently take into account down instances as well
2476 # as up ones, considering that even if they're down someone
2477 # might want to start them even in the event of a node failure.
2478 if n_img.offline or self.all_node_info[node].group != self.group_uuid:
2479 # we're skipping nodes marked offline and nodes in other groups from
2480 # the N+1 warning, since most likely we don't have good memory
2481 # infromation from them; we already list instances living on such
2482 # nodes, and that's enough warning
2484 #TODO(dynmem): also consider ballooning out other instances
2485 for prinode, instances in n_img.sbp.items():
2487 for instance in instances:
2488 bep = cluster_info.FillBE(instance_cfg[instance])
2489 if bep[constants.BE_AUTO_BALANCE]:
2490 needed_mem += bep[constants.BE_MINMEM]
2491 test = n_img.mfree < needed_mem
2492 self._ErrorIf(test, constants.CV_ENODEN1, node,
2493 "not enough memory to accomodate instance failovers"
2494 " should node %s fail (%dMiB needed, %dMiB available)",
2495 prinode, needed_mem, n_img.mfree)
2498 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
2499 (files_all, files_opt, files_mc, files_vm)):
2500 """Verifies file checksums collected from all nodes.
2502 @param errorif: Callback for reporting errors
2503 @param nodeinfo: List of L{objects.Node} objects
2504 @param master_node: Name of master node
2505 @param all_nvinfo: RPC results
2508 # Define functions determining which nodes to consider for a file
2511 (files_mc, lambda node: (node.master_candidate or
2512 node.name == master_node)),
2513 (files_vm, lambda node: node.vm_capable),
2516 # Build mapping from filename to list of nodes which should have the file
2518 for (files, fn) in files2nodefn:
2520 filenodes = nodeinfo
2522 filenodes = filter(fn, nodeinfo)
2523 nodefiles.update((filename,
2524 frozenset(map(operator.attrgetter("name"), filenodes)))
2525 for filename in files)
2527 assert set(nodefiles) == (files_all | files_mc | files_vm)
2529 fileinfo = dict((filename, {}) for filename in nodefiles)
2530 ignore_nodes = set()
2532 for node in nodeinfo:
2534 ignore_nodes.add(node.name)
2537 nresult = all_nvinfo[node.name]
2539 if nresult.fail_msg or not nresult.payload:
2542 node_files = nresult.payload.get(constants.NV_FILELIST, None)
2544 test = not (node_files and isinstance(node_files, dict))
2545 errorif(test, constants.CV_ENODEFILECHECK, node.name,
2546 "Node did not return file checksum data")
2548 ignore_nodes.add(node.name)
2551 # Build per-checksum mapping from filename to nodes having it
2552 for (filename, checksum) in node_files.items():
2553 assert filename in nodefiles
2554 fileinfo[filename].setdefault(checksum, set()).add(node.name)
2556 for (filename, checksums) in fileinfo.items():
2557 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
2559 # Nodes having the file
2560 with_file = frozenset(node_name
2561 for nodes in fileinfo[filename].values()
2562 for node_name in nodes) - ignore_nodes
2564 expected_nodes = nodefiles[filename] - ignore_nodes
2566 # Nodes missing file
2567 missing_file = expected_nodes - with_file
2569 if filename in files_opt:
2571 errorif(missing_file and missing_file != expected_nodes,
2572 constants.CV_ECLUSTERFILECHECK, None,
2573 "File %s is optional, but it must exist on all or no"
2574 " nodes (not found on %s)",
2575 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
2577 errorif(missing_file, constants.CV_ECLUSTERFILECHECK, None,
2578 "File %s is missing from node(s) %s", filename,
2579 utils.CommaJoin(utils.NiceSort(missing_file)))
2581 # Warn if a node has a file it shouldn't
2582 unexpected = with_file - expected_nodes
2584 constants.CV_ECLUSTERFILECHECK, None,
2585 "File %s should not exist on node(s) %s",
2586 filename, utils.CommaJoin(utils.NiceSort(unexpected)))
2588 # See if there are multiple versions of the file
2589 test = len(checksums) > 1
2591 variants = ["variant %s on %s" %
2592 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
2593 for (idx, (checksum, nodes)) in
2594 enumerate(sorted(checksums.items()))]
2598 errorif(test, constants.CV_ECLUSTERFILECHECK, None,
2599 "File %s found with %s different checksums (%s)",
2600 filename, len(checksums), "; ".join(variants))
2602 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
2604 """Verifies and the node DRBD status.
2606 @type ninfo: L{objects.Node}
2607 @param ninfo: the node to check
2608 @param nresult: the remote results for the node
2609 @param instanceinfo: the dict of instances
2610 @param drbd_helper: the configured DRBD usermode helper
2611 @param drbd_map: the DRBD map as returned by
2612 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
2616 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2619 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
2620 test = (helper_result == None)
2621 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2622 "no drbd usermode helper returned")
2624 status, payload = helper_result
2626 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2627 "drbd usermode helper check unsuccessful: %s", payload)
2628 test = status and (payload != drbd_helper)
2629 _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,
2630 "wrong drbd usermode helper: %s", payload)
2632 # compute the DRBD minors
2634 for minor, instance in drbd_map[node].items():
2635 test = instance not in instanceinfo
2636 _ErrorIf(test, constants.CV_ECLUSTERCFG, None,
2637 "ghost instance '%s' in temporary DRBD map", instance)
2638 # ghost instance should not be running, but otherwise we
2639 # don't give double warnings (both ghost instance and
2640 # unallocated minor in use)
2642 node_drbd[minor] = (instance, False)
2644 instance = instanceinfo[instance]
2645 node_drbd[minor] = (instance.name,
2646 instance.admin_state == constants.ADMINST_UP)
2648 # and now check them
2649 used_minors = nresult.get(constants.NV_DRBDLIST, [])
2650 test = not isinstance(used_minors, (tuple, list))
2651 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2652 "cannot parse drbd status file: %s", str(used_minors))
2654 # we cannot check drbd status
2657 for minor, (iname, must_exist) in node_drbd.items():
2658 test = minor not in used_minors and must_exist
2659 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2660 "drbd minor %d of instance %s is not active", minor, iname)
2661 for minor in used_minors:
2662 test = minor not in node_drbd
2663 _ErrorIf(test, constants.CV_ENODEDRBD, node,
2664 "unallocated drbd minor %d is in use", minor)
2666 def _UpdateNodeOS(self, ninfo, nresult, nimg):
2667 """Builds the node OS structures.
2669 @type ninfo: L{objects.Node}
2670 @param ninfo: the node to check
2671 @param nresult: the remote results for the node
2672 @param nimg: the node image object
2676 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2678 remote_os = nresult.get(constants.NV_OSLIST, None)
2679 test = (not isinstance(remote_os, list) or
2680 not compat.all(isinstance(v, list) and len(v) == 7
2681 for v in remote_os))
2683 _ErrorIf(test, constants.CV_ENODEOS, node,
2684 "node hasn't returned valid OS data")
2693 for (name, os_path, status, diagnose,
2694 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
2696 if name not in os_dict:
2699 # parameters is a list of lists instead of list of tuples due to
2700 # JSON lacking a real tuple type, fix it:
2701 parameters = [tuple(v) for v in parameters]
2702 os_dict[name].append((os_path, status, diagnose,
2703 set(variants), set(parameters), set(api_ver)))
2705 nimg.oslist = os_dict
2707 def _VerifyNodeOS(self, ninfo, nimg, base):
2708 """Verifies the node OS list.
2710 @type ninfo: L{objects.Node}
2711 @param ninfo: the node to check
2712 @param nimg: the node image object
2713 @param base: the 'template' node we match against (e.g. from the master)
2717 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2719 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
2721 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
2722 for os_name, os_data in nimg.oslist.items():
2723 assert os_data, "Empty OS status for OS %s?!" % os_name
2724 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
2725 _ErrorIf(not f_status, constants.CV_ENODEOS, node,
2726 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
2727 _ErrorIf(len(os_data) > 1, constants.CV_ENODEOS, node,
2728 "OS '%s' has multiple entries (first one shadows the rest): %s",
2729 os_name, utils.CommaJoin([v[0] for v in os_data]))
2730 # comparisons with the 'base' image
2731 test = os_name not in base.oslist
2732 _ErrorIf(test, constants.CV_ENODEOS, node,
2733 "Extra OS %s not present on reference node (%s)",
2737 assert base.oslist[os_name], "Base node has empty OS status?"
2738 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
2740 # base OS is invalid, skipping
2742 for kind, a, b in [("API version", f_api, b_api),
2743 ("variants list", f_var, b_var),
2744 ("parameters", beautify_params(f_param),
2745 beautify_params(b_param))]:
2746 _ErrorIf(a != b, constants.CV_ENODEOS, node,
2747 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
2748 kind, os_name, base.name,
2749 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
2751 # check any missing OSes
2752 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
2753 _ErrorIf(missing, constants.CV_ENODEOS, node,
2754 "OSes present on reference node %s but missing on this node: %s",
2755 base.name, utils.CommaJoin(missing))
2757 def _VerifyOob(self, ninfo, nresult):
2758 """Verifies out of band functionality of a node.
2760 @type ninfo: L{objects.Node}
2761 @param ninfo: the node to check
2762 @param nresult: the remote results for the node
2766 # We just have to verify the paths on master and/or master candidates
2767 # as the oob helper is invoked on the master
2768 if ((ninfo.master_candidate or ninfo.master_capable) and
2769 constants.NV_OOB_PATHS in nresult):
2770 for path_result in nresult[constants.NV_OOB_PATHS]:
2771 self._ErrorIf(path_result, constants.CV_ENODEOOBPATH, node, path_result)
2773 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
2774 """Verifies and updates the node volume data.
2776 This function will update a L{NodeImage}'s internal structures
2777 with data from the remote call.
2779 @type ninfo: L{objects.Node}
2780 @param ninfo: the node to check
2781 @param nresult: the remote results for the node
2782 @param nimg: the node image object
2783 @param vg_name: the configured VG name
2787 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2789 nimg.lvm_fail = True
2790 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2793 elif isinstance(lvdata, basestring):
2794 _ErrorIf(True, constants.CV_ENODELVM, node, "LVM problem on node: %s",
2795 utils.SafeEncode(lvdata))
2796 elif not isinstance(lvdata, dict):
2797 _ErrorIf(True, constants.CV_ENODELVM, node,
2798 "rpc call to node failed (lvlist)")
2800 nimg.volumes = lvdata
2801 nimg.lvm_fail = False
2803 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2804 """Verifies and updates the node instance list.
2806 If the listing was successful, then updates this node's instance
2807 list. Otherwise, it marks the RPC call as failed for the instance
2810 @type ninfo: L{objects.Node}
2811 @param ninfo: the node to check
2812 @param nresult: the remote results for the node
2813 @param nimg: the node image object
2816 idata = nresult.get(constants.NV_INSTANCELIST, None)
2817 test = not isinstance(idata, list)
2818 self._ErrorIf(test, constants.CV_ENODEHV, ninfo.name,
2819 "rpc call to node failed (instancelist): %s",
2820 utils.SafeEncode(str(idata)))
2822 nimg.hyp_fail = True
2824 nimg.instances = idata
2826 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2827 """Verifies and computes a node information map
2829 @type ninfo: L{objects.Node}
2830 @param ninfo: the node to check
2831 @param nresult: the remote results for the node
2832 @param nimg: the node image object
2833 @param vg_name: the configured VG name
2837 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2839 # try to read free memory (from the hypervisor)
2840 hv_info = nresult.get(constants.NV_HVINFO, None)
2841 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2842 _ErrorIf(test, constants.CV_ENODEHV, node,
2843 "rpc call to node failed (hvinfo)")
2846 nimg.mfree = int(hv_info["memory_free"])
2847 except (ValueError, TypeError):
2848 _ErrorIf(True, constants.CV_ENODERPC, node,
2849 "node returned invalid nodeinfo, check hypervisor")
2851 # FIXME: devise a free space model for file based instances as well
2852 if vg_name is not None:
2853 test = (constants.NV_VGLIST not in nresult or
2854 vg_name not in nresult[constants.NV_VGLIST])
2855 _ErrorIf(test, constants.CV_ENODELVM, node,
2856 "node didn't return data for the volume group '%s'"
2857 " - it is either missing or broken", vg_name)
2860 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2861 except (ValueError, TypeError):
2862 _ErrorIf(True, constants.CV_ENODERPC, node,
2863 "node returned invalid LVM info, check LVM status")
2865 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2866 """Gets per-disk status information for all instances.
2868 @type nodelist: list of strings
2869 @param nodelist: Node names
2870 @type node_image: dict of (name, L{objects.Node})
2871 @param node_image: Node objects
2872 @type instanceinfo: dict of (name, L{objects.Instance})
2873 @param instanceinfo: Instance objects
2874 @rtype: {instance: {node: [(succes, payload)]}}
2875 @return: a dictionary of per-instance dictionaries with nodes as
2876 keys and disk information as values; the disk information is a
2877 list of tuples (success, payload)
2880 _ErrorIf = self._ErrorIf # pylint: disable=C0103
2883 node_disks_devonly = {}
2884 diskless_instances = set()
2885 diskless = constants.DT_DISKLESS
2887 for nname in nodelist:
2888 node_instances = list(itertools.chain(node_image[nname].pinst,
2889 node_image[nname].sinst))
2890 diskless_instances.update(inst for inst in node_instances
2891 if instanceinfo[inst].disk_template == diskless)
2892 disks = [(inst, disk)
2893 for inst in node_instances
2894 for disk in instanceinfo[inst].disks]
2897 # No need to collect data
2900 node_disks[nname] = disks
2902 # Creating copies as SetDiskID below will modify the objects and that can
2903 # lead to incorrect data returned from nodes
2904 devonly = [dev.Copy() for (_, dev) in disks]
2907 self.cfg.SetDiskID(dev, nname)
2909 node_disks_devonly[nname] = devonly
2911 assert len(node_disks) == len(node_disks_devonly)
2913 # Collect data from all nodes with disks
2914 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2917 assert len(result) == len(node_disks)
2921 for (nname, nres) in result.items():
2922 disks = node_disks[nname]
2925 # No data from this node
2926 data = len(disks) * [(False, "node offline")]
2929 _ErrorIf(msg, constants.CV_ENODERPC, nname,
2930 "while getting disk information: %s", msg)
2932 # No data from this node
2933 data = len(disks) * [(False, msg)]
2936 for idx, i in enumerate(nres.payload):
2937 if isinstance(i, (tuple, list)) and len(i) == 2:
2940 logging.warning("Invalid result from node %s, entry %d: %s",
2942 data.append((False, "Invalid result from the remote node"))
2944 for ((inst, _), status) in zip(disks, data):
2945 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2947 # Add empty entries for diskless instances.
2948 for inst in diskless_instances:
2949 assert inst not in instdisk
2952 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2953 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2954 compat.all(isinstance(s, (tuple, list)) and
2955 len(s) == 2 for s in statuses)
2956 for inst, nnames in instdisk.items()
2957 for nname, statuses in nnames.items())
2958 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2963 def _SshNodeSelector(group_uuid, all_nodes):
2964 """Create endless iterators for all potential SSH check hosts.
2967 nodes = [node for node in all_nodes
2968 if (node.group != group_uuid and
2970 keyfunc = operator.attrgetter("group")
2972 return map(itertools.cycle,
2973 [sorted(map(operator.attrgetter("name"), names))
2974 for _, names in itertools.groupby(sorted(nodes, key=keyfunc),
2978 def _SelectSshCheckNodes(cls, group_nodes, group_uuid, all_nodes):
2979 """Choose which nodes should talk to which other nodes.
2981 We will make nodes contact all nodes in their group, and one node from
2984 @warning: This algorithm has a known issue if one node group is much
2985 smaller than others (e.g. just one node). In such a case all other
2986 nodes will talk to the single node.
2989 online_nodes = sorted(node.name for node in group_nodes if not node.offline)
2990 sel = cls._SshNodeSelector(group_uuid, all_nodes)
2992 return (online_nodes,
2993 dict((name, sorted([i.next() for i in sel]))
2994 for name in online_nodes))
2996 def BuildHooksEnv(self):
2999 Cluster-Verify hooks just ran in the post phase and their failure makes
3000 the output be logged in the verify output and the verification to fail.
3004 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
3007 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
3008 for node in self.my_node_info.values())
3012 def BuildHooksNodes(self):
3013 """Build hooks nodes.
3016 return ([], self.my_node_names)
3018 def Exec(self, feedback_fn):
3019 """Verify integrity of the node group, performing various test on nodes.
3022 # This method has too many local variables. pylint: disable=R0914
3023 feedback_fn("* Verifying group '%s'" % self.group_info.name)
3025 if not self.my_node_names:
3027 feedback_fn("* Empty node group, skipping verification")
3031 _ErrorIf = self._ErrorIf # pylint: disable=C0103
3032 verbose = self.op.verbose
3033 self._feedback_fn = feedback_fn
3035 vg_name = self.cfg.GetVGName()
3036 drbd_helper = self.cfg.GetDRBDHelper()
3037 cluster = self.cfg.GetClusterInfo()
3038 groupinfo = self.cfg.GetAllNodeGroupsInfo()
3039 hypervisors = cluster.enabled_hypervisors
3040 node_data_list = [self.my_node_info[name] for name in self.my_node_names]
3042 i_non_redundant = [] # Non redundant instances
3043 i_non_a_balanced = [] # Non auto-balanced instances
3044 i_offline = 0 # Count of offline instances
3045 n_offline = 0 # Count of offline nodes
3046 n_drained = 0 # Count of nodes being drained
3047 node_vol_should = {}
3049 # FIXME: verify OS list
3052 filemap = _ComputeAncillaryFiles(cluster, False)
3054 # do local checksums
3055 master_node = self.master_node = self.cfg.GetMasterNode()
3056 master_ip = self.cfg.GetMasterIP()
3058 feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
3061 if self.cfg.GetUseExternalMipScript():
3062 user_scripts.append(constants.EXTERNAL_MASTER_SETUP_SCRIPT)
3064 node_verify_param = {
3065 constants.NV_FILELIST:
3066 utils.UniqueSequence(filename
3067 for files in filemap
3068 for filename in files),
3069 constants.NV_NODELIST:
3070 self._SelectSshCheckNodes(node_data_list, self.group_uuid,
3071 self.all_node_info.values()),
3072 constants.NV_HYPERVISOR: hypervisors,
3073 constants.NV_HVPARAMS:
3074 _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
3075 constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip)
3076 for node in node_data_list
3077 if not node.offline],
3078 constants.NV_INSTANCELIST: hypervisors,
3079 constants.NV_VERSION: None,
3080 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
3081 constants.NV_NODESETUP: None,
3082 constants.NV_TIME: None,
3083 constants.NV_MASTERIP: (master_node, master_ip),
3084 constants.NV_OSLIST: None,
3085 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
3086 constants.NV_USERSCRIPTS: user_scripts,
3089 if vg_name is not None:
3090 node_verify_param[constants.NV_VGLIST] = None
3091 node_verify_param[constants.NV_LVLIST] = vg_name
3092 node_verify_param[constants.NV_PVLIST] = [vg_name]
3093 node_verify_param[constants.NV_DRBDLIST] = None
3096 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
3099 # FIXME: this needs to be changed per node-group, not cluster-wide
3101 default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
3102 if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
3103 bridges.add(default_nicpp[constants.NIC_LINK])
3104 for instance in self.my_inst_info.values():
3105 for nic in instance.nics:
3106 full_nic = cluster.SimpleFillNIC(nic.nicparams)
3107 if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
3108 bridges.add(full_nic[constants.NIC_LINK])
3111 node_verify_param[constants.NV_BRIDGES] = list(bridges)
3113 # Build our expected cluster state
3114 node_image = dict((node.name, self.NodeImage(offline=node.offline,
3116 vm_capable=node.vm_capable))
3117 for node in node_data_list)
3121 for node in self.all_node_info.values():
3122 path = _SupportsOob(self.cfg, node)
3123 if path and path not in oob_paths:
3124 oob_paths.append(path)
3127 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
3129 for instance in self.my_inst_names:
3130 inst_config = self.my_inst_info[instance]
3132 for nname in inst_config.all_nodes:
3133 if nname not in node_image:
3134 gnode = self.NodeImage(name=nname)
3135 gnode.ghost = (nname not in self.all_node_info)
3136 node_image[nname] = gnode
3138 inst_config.MapLVsByNode(node_vol_should)
3140 pnode = inst_config.primary_node
3141 node_image[pnode].pinst.append(instance)
3143 for snode in inst_config.secondary_nodes:
3144 nimg = node_image[snode]
3145 nimg.sinst.append(instance)
3146 if pnode not in nimg.sbp:
3147 nimg.sbp[pnode] = []
3148 nimg.sbp[pnode].append(instance)
3150 # At this point, we have the in-memory data structures complete,
3151 # except for the runtime information, which we'll gather next
3153 # Due to the way our RPC system works, exact response times cannot be
3154 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
3155 # time before and after executing the request, we can at least have a time
3157 nvinfo_starttime = time.time()
3158 all_nvinfo = self.rpc.call_node_verify(self.my_node_names,
3160 self.cfg.GetClusterName())
3161 nvinfo_endtime = time.time()
3163 if self.extra_lv_nodes and vg_name is not None:
3165 self.rpc.call_node_verify(self.extra_lv_nodes,
3166 {constants.NV_LVLIST: vg_name},
3167 self.cfg.GetClusterName())
3169 extra_lv_nvinfo = {}
3171 all_drbd_map = self.cfg.ComputeDRBDMap()
3173 feedback_fn("* Gathering disk information (%s nodes)" %
3174 len(self.my_node_names))
3175 instdisk = self._CollectDiskInfo(self.my_node_names, node_image,
3178 feedback_fn("* Verifying configuration file consistency")
3180 # If not all nodes are being checked, we need to make sure the master node
3181 # and a non-checked vm_capable node are in the list.
3182 absent_nodes = set(self.all_node_info).difference(self.my_node_info)
3184 vf_nvinfo = all_nvinfo.copy()
3185 vf_node_info = list(self.my_node_info.values())
3186 additional_nodes = []
3187 if master_node not in self.my_node_info:
3188 additional_nodes.append(master_node)
3189 vf_node_info.append(self.all_node_info[master_node])
3190 # Add the first vm_capable node we find which is not included
3191 for node in absent_nodes:
3192 nodeinfo = self.all_node_info[node]
3193 if nodeinfo.vm_capable and not nodeinfo.offline:
3194 additional_nodes.append(node)
3195 vf_node_info.append(self.all_node_info[node])
3197 key = constants.NV_FILELIST
3198 vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes,
3199 {key: node_verify_param[key]},
3200 self.cfg.GetClusterName()))
3202 vf_nvinfo = all_nvinfo
3203 vf_node_info = self.my_node_info.values()
3205 self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap)
3207 feedback_fn("* Verifying node status")
3211 for node_i in node_data_list:
3213 nimg = node_image[node]
3217 feedback_fn("* Skipping offline node %s" % (node,))
3221 if node == master_node:
3223 elif node_i.master_candidate:
3224 ntype = "master candidate"
3225 elif node_i.drained:
3231 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
3233 msg = all_nvinfo[node].fail_msg
3234 _ErrorIf(msg, constants.CV_ENODERPC, node, "while contacting node: %s",
3237 nimg.rpc_fail = True
3240 nresult = all_nvinfo[node].payload
3242 nimg.call_ok = self._VerifyNode(node_i, nresult)
3243 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
3244 self._VerifyNodeNetwork(node_i, nresult)
3245 self._VerifyNodeUserScripts(node_i, nresult)
3246 self._VerifyOob(node_i, nresult)
3249 self._VerifyNodeLVM(node_i, nresult, vg_name)
3250 self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper,
3253 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
3254 self._UpdateNodeInstances(node_i, nresult, nimg)
3255 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
3256 self._UpdateNodeOS(node_i, nresult, nimg)
3258 if not nimg.os_fail:
3259 if refos_img is None:
3261 self._VerifyNodeOS(node_i, nimg, refos_img)
3262 self._VerifyNodeBridges(node_i, nresult, bridges)
3264 # Check whether all running instancies are primary for the node. (This
3265 # can no longer be done from _VerifyInstance below, since some of the
3266 # wrong instances could be from other node groups.)
3267 non_primary_inst = set(nimg.instances).difference(nimg.pinst)
3269 for inst in non_primary_inst:
3270 # FIXME: investigate best way to handle offline insts
3271 if inst.admin_state == constants.ADMINST_OFFLINE:
3273 feedback_fn("* Skipping offline instance %s" % inst.name)
3276 test = inst in self.all_inst_info
3277 _ErrorIf(test, constants.CV_EINSTANCEWRONGNODE, inst,
3278 "instance should not run on node %s", node_i.name)
3279 _ErrorIf(not test, constants.CV_ENODEORPHANINSTANCE, node_i.name,
3280 "node is running unknown instance %s", inst)
3282 for node, result in extra_lv_nvinfo.items():
3283 self._UpdateNodeVolumes(self.all_node_info[node], result.payload,
3284 node_image[node], vg_name)
3286 feedback_fn("* Verifying instance status")
3287 for instance in self.my_inst_names:
3289 feedback_fn("* Verifying instance %s" % instance)
3290 inst_config = self.my_inst_info[instance]
3291 self._VerifyInstance(instance, inst_config, node_image,
3293 inst_nodes_offline = []
3295 pnode = inst_config.primary_node
3296 pnode_img = node_image[pnode]
3297 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
3298 constants.CV_ENODERPC, pnode, "instance %s, connection to"
3299 " primary node failed", instance)
3301 _ErrorIf(inst_config.admin_state == constants.ADMINST_UP and
3303 constants.CV_EINSTANCEBADNODE, instance,
3304 "instance is marked as running and lives on offline node %s",
3305 inst_config.primary_node)
3307 # If the instance is non-redundant we cannot survive losing its primary
3308 # node, so we are not N+1 compliant. On the other hand we have no disk
3309 # templates with more than one secondary so that situation is not well
3311 # FIXME: does not support file-backed instances
3312 if not inst_config.secondary_nodes:
3313 i_non_redundant.append(instance)
3315 _ErrorIf(len(inst_config.secondary_nodes) > 1,
3316 constants.CV_EINSTANCELAYOUT,
3317 instance, "instance has multiple secondary nodes: %s",
3318 utils.CommaJoin(inst_config.secondary_nodes),
3319 code=self.ETYPE_WARNING)
3321 if inst_config.disk_template in constants.DTS_INT_MIRROR:
3322 pnode = inst_config.primary_node
3323 instance_nodes = utils.NiceSort(inst_config.all_nodes)
3324 instance_groups = {}
3326 for node in instance_nodes:
3327 instance_groups.setdefault(self.all_node_info[node].group,
3331 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
3332 # Sort so that we always list the primary node first.
3333 for group, nodes in sorted(instance_groups.items(),
3334 key=lambda (_, nodes): pnode in nodes,
3337 self._ErrorIf(len(instance_groups) > 1,
3338 constants.CV_EINSTANCESPLITGROUPS,
3339 instance, "instance has primary and secondary nodes in"
3340 " different groups: %s", utils.CommaJoin(pretty_list),
3341 code=self.ETYPE_WARNING)
3343 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
3344 i_non_a_balanced.append(instance)
3346 for snode in inst_config.secondary_nodes:
3347 s_img = node_image[snode]
3348 _ErrorIf(s_img.rpc_fail and not s_img.offline, constants.CV_ENODERPC,
3349 snode, "instance %s, connection to secondary node failed",
3353 inst_nodes_offline.append(snode)
3355 # warn that the instance lives on offline nodes
3356 _ErrorIf(inst_nodes_offline, constants.CV_EINSTANCEBADNODE, instance,
3357 "instance has offline secondary node(s) %s",
3358 utils.CommaJoin(inst_nodes_offline))
3359 # ... or ghost/non-vm_capable nodes
3360 for node in inst_config.all_nodes:
3361 _ErrorIf(node_image[node].ghost, constants.CV_EINSTANCEBADNODE,
3362 instance, "instance lives on ghost node %s", node)
3363 _ErrorIf(not node_image[node].vm_capable, constants.CV_EINSTANCEBADNODE,
3364 instance, "instance lives on non-vm_capable node %s", node)
3366 feedback_fn("* Verifying orphan volumes")
3367 reserved = utils.FieldSet(*cluster.reserved_lvs)
3369 # We will get spurious "unknown volume" warnings if any node of this group
3370 # is secondary for an instance whose primary is in another group. To avoid
3371 # them, we find these instances and add their volumes to node_vol_should.
3372 for inst in self.all_inst_info.values():
3373 for secondary in inst.secondary_nodes:
3374 if (secondary in self.my_node_info
3375 and inst.name not in self.my_inst_info):
3376 inst.MapLVsByNode(node_vol_should)
3379 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
3381 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
3382 feedback_fn("* Verifying N+1 Memory redundancy")
3383 self._VerifyNPlusOneMemory(node_image, self.my_inst_info)
3385 feedback_fn("* Other Notes")
3387 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
3388 % len(i_non_redundant))
3390 if i_non_a_balanced:
3391 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
3392 % len(i_non_a_balanced))
3395 feedback_fn(" - NOTICE: %d offline instance(s) found." % i_offline)
3398 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
3401 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
3405 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
3406 """Analyze the post-hooks' result
3408 This method analyses the hook result, handles it, and sends some
3409 nicely-formatted feedback back to the user.
3411 @param phase: one of L{constants.HOOKS_PHASE_POST} or
3412 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
3413 @param hooks_results: the results of the multi-node hooks rpc call
3414 @param feedback_fn: function used send feedback back to the caller
3415 @param lu_result: previous Exec result
3416 @return: the new Exec result, based on the previous result
3420 # We only really run POST phase hooks, only for non-empty groups,
3421 # and are only interested in their results
3422 if not self.my_node_names:
3425 elif phase == constants.HOOKS_PHASE_POST:
3426 # Used to change hooks' output to proper indentation
3427 feedback_fn("* Hooks Results")
3428 assert hooks_results, "invalid result from hooks"
3430 for node_name in hooks_results:
3431 res = hooks_results[node_name]
3433 test = msg and not res.offline
3434 self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,
3435 "Communication failure in hooks execution: %s", msg)
3436 if res.offline or msg:
3437 # No need to investigate payload if node is offline or gave
3440 for script, hkr, output in res.payload:
3441 test = hkr == constants.HKR_FAIL
3442 self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,
3443 "Script %s failed, output:", script)
3445 output = self._HOOKS_INDENT_RE.sub(" ", output)
3446 feedback_fn("%s" % output)
3452 class LUClusterVerifyDisks(NoHooksLU):
3453 """Verifies the cluster disks status.
3458 def ExpandNames(self):
3459 self.share_locks = _ShareAll()
3460 self.needed_locks = {
3461 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3464 def Exec(self, feedback_fn):
3465 group_names = self.owned_locks(locking.LEVEL_NODEGROUP)
3467 # Submit one instance of L{opcodes.OpGroupVerifyDisks} per node group
3468 return ResultWithJobs([[opcodes.OpGroupVerifyDisks(group_name=group)]
3469 for group in group_names])
3472 class LUGroupVerifyDisks(NoHooksLU):
3473 """Verifies the status of all disks in a node group.
3478 def ExpandNames(self):
3479 # Raises errors.OpPrereqError on its own if group can't be found
3480 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
3482 self.share_locks = _ShareAll()
3483 self.needed_locks = {
3484 locking.LEVEL_INSTANCE: [],
3485 locking.LEVEL_NODEGROUP: [],
3486 locking.LEVEL_NODE: [],
3489 def DeclareLocks(self, level):
3490 if level == locking.LEVEL_INSTANCE:
3491 assert not self.needed_locks[locking.LEVEL_INSTANCE]
3493 # Lock instances optimistically, needs verification once node and group
3494 # locks have been acquired
3495 self.needed_locks[locking.LEVEL_INSTANCE] = \
3496 self.cfg.GetNodeGroupInstances(self.group_uuid)
3498 elif level == locking.LEVEL_NODEGROUP:
3499 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
3501 self.needed_locks[locking.LEVEL_NODEGROUP] = \
3502 set([self.group_uuid] +
3503 # Lock all groups used by instances optimistically; this requires
3504 # going via the node before it's locked, requiring verification
3507 for instance_name in self.owned_locks(locking.LEVEL_INSTANCE)
3508 for group_uuid in self.cfg.GetInstanceNodeGroups(instance_name)])
3510 elif level == locking.LEVEL_NODE:
3511 # This will only lock the nodes in the group to be verified which contain
3513 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
3514 self._LockInstancesNodes()
3516 # Lock all nodes in group to be verified
3517 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
3518 member_nodes = self.cfg.GetNodeGroup(self.group_uuid).members
3519 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
3521 def CheckPrereq(self):
3522 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
3523 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
3524 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
3526 assert self.group_uuid in owned_groups
3528 # Check if locked instances are still correct
3529 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
3531 # Get instance information
3532 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
3534 # Check if node groups for locked instances are still correct
3535 _CheckInstancesNodeGroups(self.cfg, self.instances,
3536 owned_groups, owned_nodes, self.group_uuid)
3538 def Exec(self, feedback_fn):
3539 """Verify integrity of cluster disks.
3541 @rtype: tuple of three items
3542 @return: a tuple of (dict of node-to-node_error, list of instances
3543 which need activate-disks, dict of instance: (node, volume) for
3548 res_instances = set()
3551 nv_dict = _MapInstanceDisksToNodes([inst
3552 for inst in self.instances.values()
3553 if inst.admin_state == constants.ADMINST_UP])
3556 nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &
3557 set(self.cfg.GetVmCapableNodeList()))
3559 node_lvs = self.rpc.call_lv_list(nodes, [])
3561 for (node, node_res) in node_lvs.items():
3562 if node_res.offline:
3565 msg = node_res.fail_msg
3567 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
3568 res_nodes[node] = msg
3571 for lv_name, (_, _, lv_online) in node_res.payload.items():
3572 inst = nv_dict.pop((node, lv_name), None)
3573 if not (lv_online or inst is None):
3574 res_instances.add(inst)
3576 # any leftover items in nv_dict are missing LVs, let's arrange the data
3578 for key, inst in nv_dict.iteritems():
3579 res_missing.setdefault(inst, []).append(list(key))
3581 return (res_nodes, list(res_instances), res_missing)
3584 class LUClusterRepairDiskSizes(NoHooksLU):
3585 """Verifies the cluster disks sizes.
3590 def ExpandNames(self):
3591 if self.op.instances:
3592 self.wanted_names = _GetWantedInstances(self, self.op.instances)
3593 self.needed_locks = {
3594 locking.LEVEL_NODE_RES: [],
3595 locking.LEVEL_INSTANCE: self.wanted_names,
3597 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
3599 self.wanted_names = None
3600 self.needed_locks = {
3601 locking.LEVEL_NODE_RES: locking.ALL_SET,
3602 locking.LEVEL_INSTANCE: locking.ALL_SET,
3604 self.share_locks = {
3605 locking.LEVEL_NODE_RES: 1,
3606 locking.LEVEL_INSTANCE: 0,
3609 def DeclareLocks(self, level):
3610 if level == locking.LEVEL_NODE_RES and self.wanted_names is not None:
3611 self._LockInstancesNodes(primary_only=True, level=level)
3613 def CheckPrereq(self):
3614 """Check prerequisites.
3616 This only checks the optional instance list against the existing names.
3619 if self.wanted_names is None:
3620 self.wanted_names = self.owned_locks(locking.LEVEL_INSTANCE)
3622 self.wanted_instances = \
3623 map(compat.snd, self.cfg.GetMultiInstanceInfo(self.wanted_names))
3625 def _EnsureChildSizes(self, disk):
3626 """Ensure children of the disk have the needed disk size.
3628 This is valid mainly for DRBD8 and fixes an issue where the
3629 children have smaller disk size.
3631 @param disk: an L{ganeti.objects.Disk} object
3634 if disk.dev_type == constants.LD_DRBD8:
3635 assert disk.children, "Empty children for DRBD8?"
3636 fchild = disk.children[0]
3637 mismatch = fchild.size < disk.size
3639 self.LogInfo("Child disk has size %d, parent %d, fixing",
3640 fchild.size, disk.size)
3641 fchild.size = disk.size
3643 # and we recurse on this child only, not on the metadev
3644 return self._EnsureChildSizes(fchild) or mismatch
3648 def Exec(self, feedback_fn):
3649 """Verify the size of cluster disks.
3652 # TODO: check child disks too
3653 # TODO: check differences in size between primary/secondary nodes
3655 for instance in self.wanted_instances:
3656 pnode = instance.primary_node
3657 if pnode not in per_node_disks:
3658 per_node_disks[pnode] = []
3659 for idx, disk in enumerate(instance.disks):
3660 per_node_disks[pnode].append((instance, idx, disk))
3662 assert not (frozenset(per_node_disks.keys()) -
3663 self.owned_locks(locking.LEVEL_NODE_RES)), \
3664 "Not owning correct locks"
3665 assert not self.owned_locks(locking.LEVEL_NODE)
3668 for node, dskl in per_node_disks.items():
3669 newl = [v[2].Copy() for v in dskl]
3671 self.cfg.SetDiskID(dsk, node)
3672 result = self.rpc.call_blockdev_getsize(node, newl)
3674 self.LogWarning("Failure in blockdev_getsize call to node"
3675 " %s, ignoring", node)
3677 if len(result.payload) != len(dskl):
3678 logging.warning("Invalid result from node %s: len(dksl)=%d,"
3679 " result.payload=%s", node, len(dskl), result.payload)
3680 self.LogWarning("Invalid result from node %s, ignoring node results",
3683 for ((instance, idx, disk), size) in zip(dskl, result.payload):
3685 self.LogWarning("Disk %d of instance %s did not return size"
3686 " information, ignoring", idx, instance.name)
3688 if not isinstance(size, (int, long)):
3689 self.LogWarning("Disk %d of instance %s did not return valid"
3690 " size information, ignoring", idx, instance.name)
3693 if size != disk.size:
3694 self.LogInfo("Disk %d of instance %s has mismatched size,"
3695 " correcting: recorded %d, actual %d", idx,
3696 instance.name, disk.size, size)
3698 self.cfg.Update(instance, feedback_fn)
3699 changed.append((instance.name, idx, size))
3700 if self._EnsureChildSizes(disk):
3701 self.cfg.Update(instance, feedback_fn)
3702 changed.append((instance.name, idx, disk.size))
3706 class LUClusterRename(LogicalUnit):
3707 """Rename the cluster.
3710 HPATH = "cluster-rename"
3711 HTYPE = constants.HTYPE_CLUSTER
3713 def BuildHooksEnv(self):
3718 "OP_TARGET": self.cfg.GetClusterName(),
3719 "NEW_NAME": self.op.name,
3722 def BuildHooksNodes(self):
3723 """Build hooks nodes.
3726 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
3728 def CheckPrereq(self):
3729 """Verify that the passed name is a valid one.
3732 hostname = netutils.GetHostname(name=self.op.name,
3733 family=self.cfg.GetPrimaryIPFamily())
3735 new_name = hostname.name
3736 self.ip = new_ip = hostname.ip
3737 old_name = self.cfg.GetClusterName()
3738 old_ip = self.cfg.GetMasterIP()
3739 if new_name == old_name and new_ip == old_ip:
3740 raise errors.OpPrereqError("Neither the name nor the IP address of the"
3741 " cluster has changed",
3743 if new_ip != old_ip:
3744 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
3745 raise errors.OpPrereqError("The given cluster IP address (%s) is"
3746 " reachable on the network" %
3747 new_ip, errors.ECODE_NOTUNIQUE)
3749 self.op.name = new_name
3751 def Exec(self, feedback_fn):
3752 """Rename the cluster.
3755 clustername = self.op.name
3758 # shutdown the master IP
3759 master_params = self.cfg.GetMasterNetworkParameters()
3760 ems = self.cfg.GetUseExternalMipScript()
3761 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
3763 result.Raise("Could not disable the master role")
3766 cluster = self.cfg.GetClusterInfo()
3767 cluster.cluster_name = clustername
3768 cluster.master_ip = new_ip
3769 self.cfg.Update(cluster, feedback_fn)
3771 # update the known hosts file
3772 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
3773 node_list = self.cfg.GetOnlineNodeList()
3775 node_list.remove(master_params.name)
3778 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
3780 master_params.ip = new_ip
3781 result = self.rpc.call_node_activate_master_ip(master_params.name,
3783 msg = result.fail_msg
3785 self.LogWarning("Could not re-enable the master role on"
3786 " the master, please restart manually: %s", msg)
3791 def _ValidateNetmask(cfg, netmask):
3792 """Checks if a netmask is valid.
3794 @type cfg: L{config.ConfigWriter}
3795 @param cfg: The cluster configuration
3797 @param netmask: the netmask to be verified
3798 @raise errors.OpPrereqError: if the validation fails
3801 ip_family = cfg.GetPrimaryIPFamily()
3803 ipcls = netutils.IPAddress.GetClassFromIpFamily(ip_family)
3804 except errors.ProgrammerError:
3805 raise errors.OpPrereqError("Invalid primary ip family: %s." %
3807 if not ipcls.ValidateNetmask(netmask):
3808 raise errors.OpPrereqError("CIDR netmask (%s) not valid" %
3812 class LUClusterSetParams(LogicalUnit):
3813 """Change the parameters of the cluster.
3816 HPATH = "cluster-modify"
3817 HTYPE = constants.HTYPE_CLUSTER
3820 def CheckArguments(self):
3824 if self.op.uid_pool:
3825 uidpool.CheckUidPool(self.op.uid_pool)
3827 if self.op.add_uids:
3828 uidpool.CheckUidPool(self.op.add_uids)
3830 if self.op.remove_uids:
3831 uidpool.CheckUidPool(self.op.remove_uids)
3833 if self.op.master_netmask is not None:
3834 _ValidateNetmask(self.cfg, self.op.master_netmask)
3836 if self.op.diskparams:
3837 for dt_params in self.op.diskparams.values():
3838 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
3840 def ExpandNames(self):
3841 # FIXME: in the future maybe other cluster params won't require checking on
3842 # all nodes to be modified.
3843 self.needed_locks = {
3844 locking.LEVEL_NODE: locking.ALL_SET,
3845 locking.LEVEL_INSTANCE: locking.ALL_SET,
3846 locking.LEVEL_NODEGROUP: locking.ALL_SET,
3848 self.share_locks = {
3849 locking.LEVEL_NODE: 1,
3850 locking.LEVEL_INSTANCE: 1,
3851 locking.LEVEL_NODEGROUP: 1,
3854 def BuildHooksEnv(self):
3859 "OP_TARGET": self.cfg.GetClusterName(),
3860 "NEW_VG_NAME": self.op.vg_name,
3863 def BuildHooksNodes(self):
3864 """Build hooks nodes.
3867 mn = self.cfg.GetMasterNode()
3870 def CheckPrereq(self):
3871 """Check prerequisites.
3873 This checks whether the given params don't conflict and
3874 if the given volume group is valid.
3877 if self.op.vg_name is not None and not self.op.vg_name:
3878 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
3879 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
3880 " instances exist", errors.ECODE_INVAL)
3882 if self.op.drbd_helper is not None and not self.op.drbd_helper:
3883 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
3884 raise errors.OpPrereqError("Cannot disable drbd helper while"
3885 " drbd-based instances exist",
3888 node_list = self.owned_locks(locking.LEVEL_NODE)
3890 # if vg_name not None, checks given volume group on all nodes
3892 vglist = self.rpc.call_vg_list(node_list)
3893 for node in node_list:
3894 msg = vglist[node].fail_msg
3896 # ignoring down node
3897 self.LogWarning("Error while gathering data on node %s"
3898 " (ignoring node): %s", node, msg)
3900 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
3902 constants.MIN_VG_SIZE)
3904 raise errors.OpPrereqError("Error on node '%s': %s" %
3905 (node, vgstatus), errors.ECODE_ENVIRON)
3907 if self.op.drbd_helper:
3908 # checks given drbd helper on all nodes
3909 helpers = self.rpc.call_drbd_helper(node_list)
3910 for (node, ninfo) in self.cfg.GetMultiNodeInfo(node_list):
3912 self.LogInfo("Not checking drbd helper on offline node %s", node)
3914 msg = helpers[node].fail_msg
3916 raise errors.OpPrereqError("Error checking drbd helper on node"
3917 " '%s': %s" % (node, msg),
3918 errors.ECODE_ENVIRON)
3919 node_helper = helpers[node].payload
3920 if node_helper != self.op.drbd_helper:
3921 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
3922 (node, node_helper), errors.ECODE_ENVIRON)
3924 self.cluster = cluster = self.cfg.GetClusterInfo()
3925 # validate params changes
3926 if self.op.beparams:
3927 objects.UpgradeBeParams(self.op.beparams)
3928 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
3929 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
3931 if self.op.ndparams:
3932 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
3933 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
3935 # TODO: we need a more general way to handle resetting
3936 # cluster-level parameters to default values
3937 if self.new_ndparams["oob_program"] == "":
3938 self.new_ndparams["oob_program"] = \
3939 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
3941 if self.op.hv_state:
3942 new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
3943 self.cluster.hv_state_static)
3944 self.new_hv_state = dict((hv, cluster.SimpleFillHvState(values))
3945 for hv, values in new_hv_state.items())
3947 if self.op.disk_state:
3948 new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state,
3949 self.cluster.disk_state_static)
3950 self.new_disk_state = \
3951 dict((storage, dict((name, cluster.SimpleFillDiskState(values))
3952 for name, values in svalues.items()))
3953 for storage, svalues in new_disk_state.items())
3956 self.new_ipolicy = _GetUpdatedIPolicy(cluster.ipolicy, self.op.ipolicy,
3959 all_instances = self.cfg.GetAllInstancesInfo().values()
3961 for group in self.cfg.GetAllNodeGroupsInfo().values():
3962 instances = frozenset([inst for inst in all_instances
3963 if compat.any(node in group.members
3964 for node in inst.all_nodes)])
3965 new_ipolicy = objects.FillIPolicy(self.new_ipolicy, group.ipolicy)
3966 new = _ComputeNewInstanceViolations(_CalculateGroupIPolicy(cluster,
3968 new_ipolicy, instances)
3970 violations.update(new)
3973 self.LogWarning("After the ipolicy change the following instances"
3974 " violate them: %s",
3975 utils.CommaJoin(violations))
3977 if self.op.nicparams:
3978 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
3979 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
3980 objects.NIC.CheckParameterSyntax(self.new_nicparams)
3983 # check all instances for consistency
3984 for instance in self.cfg.GetAllInstancesInfo().values():
3985 for nic_idx, nic in enumerate(instance.nics):
3986 params_copy = copy.deepcopy(nic.nicparams)
3987 params_filled = objects.FillDict(self.new_nicparams, params_copy)
3989 # check parameter syntax
3991 objects.NIC.CheckParameterSyntax(params_filled)
3992 except errors.ConfigurationError, err:
3993 nic_errors.append("Instance %s, nic/%d: %s" %
3994 (instance.name, nic_idx, err))
3996 # if we're moving instances to routed, check that they have an ip
3997 target_mode = params_filled[constants.NIC_MODE]
3998 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
3999 nic_errors.append("Instance %s, nic/%d: routed NIC with no ip"
4000 " address" % (instance.name, nic_idx))
4002 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
4003 "\n".join(nic_errors))
4005 # hypervisor list/parameters
4006 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
4007 if self.op.hvparams:
4008 for hv_name, hv_dict in self.op.hvparams.items():
4009 if hv_name not in self.new_hvparams:
4010 self.new_hvparams[hv_name] = hv_dict
4012 self.new_hvparams[hv_name].update(hv_dict)
4014 # disk template parameters
4015 self.new_diskparams = objects.FillDict(cluster.diskparams, {})
4016 if self.op.diskparams:
4017 for dt_name, dt_params in self.op.diskparams.items():
4018 if dt_name not in self.op.diskparams:
4019 self.new_diskparams[dt_name] = dt_params
4021 self.new_diskparams[dt_name].update(dt_params)
4023 # os hypervisor parameters
4024 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
4026 for os_name, hvs in self.op.os_hvp.items():
4027 if os_name not in self.new_os_hvp:
4028 self.new_os_hvp[os_name] = hvs
4030 for hv_name, hv_dict in hvs.items():
4031 if hv_name not in self.new_os_hvp[os_name]:
4032 self.new_os_hvp[os_name][hv_name] = hv_dict
4034 self.new_os_hvp[os_name][hv_name].update(hv_dict)
4037 self.new_osp = objects.FillDict(cluster.osparams, {})
4038 if self.op.osparams:
4039 for os_name, osp in self.op.osparams.items():
4040 if os_name not in self.new_osp:
4041 self.new_osp[os_name] = {}
4043 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
4046 if not self.new_osp[os_name]:
4047 # we removed all parameters
4048 del self.new_osp[os_name]
4050 # check the parameter validity (remote check)
4051 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
4052 os_name, self.new_osp[os_name])
4054 # changes to the hypervisor list
4055 if self.op.enabled_hypervisors is not None:
4056 self.hv_list = self.op.enabled_hypervisors
4057 for hv in self.hv_list:
4058 # if the hypervisor doesn't already exist in the cluster
4059 # hvparams, we initialize it to empty, and then (in both
4060 # cases) we make sure to fill the defaults, as we might not
4061 # have a complete defaults list if the hypervisor wasn't
4063 if hv not in new_hvp:
4065 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
4066 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
4068 self.hv_list = cluster.enabled_hypervisors
4070 if self.op.hvparams or self.op.enabled_hypervisors is not None:
4071 # either the enabled list has changed, or the parameters have, validate
4072 for hv_name, hv_params in self.new_hvparams.items():
4073 if ((self.op.hvparams and hv_name in self.op.hvparams) or
4074 (self.op.enabled_hypervisors and
4075 hv_name in self.op.enabled_hypervisors)):
4076 # either this is a new hypervisor, or its parameters have changed
4077 hv_class = hypervisor.GetHypervisor(hv_name)
4078 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
4079 hv_class.CheckParameterSyntax(hv_params)
4080 _CheckHVParams(self, node_list, hv_name, hv_params)
4083 # no need to check any newly-enabled hypervisors, since the
4084 # defaults have already been checked in the above code-block
4085 for os_name, os_hvp in self.new_os_hvp.items():
4086 for hv_name, hv_params in os_hvp.items():
4087 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
4088 # we need to fill in the new os_hvp on top of the actual hv_p
4089 cluster_defaults = self.new_hvparams.get(hv_name, {})
4090 new_osp = objects.FillDict(cluster_defaults, hv_params)
4091 hv_class = hypervisor.GetHypervisor(hv_name)
4092 hv_class.CheckParameterSyntax(new_osp)
4093 _CheckHVParams(self, node_list, hv_name, new_osp)
4095 if self.op.default_iallocator:
4096 alloc_script = utils.FindFile(self.op.default_iallocator,
4097 constants.IALLOCATOR_SEARCH_PATH,
4099 if alloc_script is None:
4100 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
4101 " specified" % self.op.default_iallocator,
4104 def Exec(self, feedback_fn):
4105 """Change the parameters of the cluster.
4108 if self.op.vg_name is not None:
4109 new_volume = self.op.vg_name
4112 if new_volume != self.cfg.GetVGName():
4113 self.cfg.SetVGName(new_volume)
4115 feedback_fn("Cluster LVM configuration already in desired"
4116 " state, not changing")
4117 if self.op.drbd_helper is not None:
4118 new_helper = self.op.drbd_helper
4121 if new_helper != self.cfg.GetDRBDHelper():
4122 self.cfg.SetDRBDHelper(new_helper)
4124 feedback_fn("Cluster DRBD helper already in desired state,"
4126 if self.op.hvparams:
4127 self.cluster.hvparams = self.new_hvparams
4129 self.cluster.os_hvp = self.new_os_hvp
4130 if self.op.enabled_hypervisors is not None:
4131 self.cluster.hvparams = self.new_hvparams
4132 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
4133 if self.op.beparams:
4134 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
4135 if self.op.nicparams:
4136 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
4138 self.cluster.ipolicy = self.new_ipolicy
4139 if self.op.osparams:
4140 self.cluster.osparams = self.new_osp
4141 if self.op.ndparams:
4142 self.cluster.ndparams = self.new_ndparams
4143 if self.op.diskparams:
4144 self.cluster.diskparams = self.new_diskparams
4145 if self.op.hv_state:
4146 self.cluster.hv_state_static = self.new_hv_state
4147 if self.op.disk_state:
4148 self.cluster.disk_state_static = self.new_disk_state
4150 if self.op.candidate_pool_size is not None:
4151 self.cluster.candidate_pool_size = self.op.candidate_pool_size
4152 # we need to update the pool size here, otherwise the save will fail
4153 _AdjustCandidatePool(self, [])
4155 if self.op.maintain_node_health is not None:
4156 if self.op.maintain_node_health and not constants.ENABLE_CONFD:
4157 feedback_fn("Note: CONFD was disabled at build time, node health"
4158 " maintenance is not useful (still enabling it)")
4159 self.cluster.maintain_node_health = self.op.maintain_node_health
4161 if self.op.prealloc_wipe_disks is not None:
4162 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
4164 if self.op.add_uids is not None:
4165 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
4167 if self.op.remove_uids is not None:
4168 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
4170 if self.op.uid_pool is not None:
4171 self.cluster.uid_pool = self.op.uid_pool
4173 if self.op.default_iallocator is not None:
4174 self.cluster.default_iallocator = self.op.default_iallocator
4176 if self.op.reserved_lvs is not None:
4177 self.cluster.reserved_lvs = self.op.reserved_lvs
4179 if self.op.use_external_mip_script is not None:
4180 self.cluster.use_external_mip_script = self.op.use_external_mip_script
4182 def helper_os(aname, mods, desc):
4184 lst = getattr(self.cluster, aname)
4185 for key, val in mods:
4186 if key == constants.DDM_ADD:
4188 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
4191 elif key == constants.DDM_REMOVE:
4195 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
4197 raise errors.ProgrammerError("Invalid modification '%s'" % key)
4199 if self.op.hidden_os:
4200 helper_os("hidden_os", self.op.hidden_os, "hidden")
4202 if self.op.blacklisted_os:
4203 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
4205 if self.op.master_netdev:
4206 master_params = self.cfg.GetMasterNetworkParameters()
4207 ems = self.cfg.GetUseExternalMipScript()
4208 feedback_fn("Shutting down master ip on the current netdev (%s)" %
4209 self.cluster.master_netdev)
4210 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
4212 result.Raise("Could not disable the master ip")
4213 feedback_fn("Changing master_netdev from %s to %s" %
4214 (master_params.netdev, self.op.master_netdev))
4215 self.cluster.master_netdev = self.op.master_netdev
4217 if self.op.master_netmask:
4218 master_params = self.cfg.GetMasterNetworkParameters()
4219 feedback_fn("Changing master IP netmask to %s" % self.op.master_netmask)
4220 result = self.rpc.call_node_change_master_netmask(master_params.name,
4221 master_params.netmask,
4222 self.op.master_netmask,
4224 master_params.netdev)
4226 msg = "Could not change the master IP netmask: %s" % result.fail_msg
4229 self.cluster.master_netmask = self.op.master_netmask
4231 self.cfg.Update(self.cluster, feedback_fn)
4233 if self.op.master_netdev:
4234 master_params = self.cfg.GetMasterNetworkParameters()
4235 feedback_fn("Starting the master ip on the new master netdev (%s)" %
4236 self.op.master_netdev)
4237 ems = self.cfg.GetUseExternalMipScript()
4238 result = self.rpc.call_node_activate_master_ip(master_params.name,
4241 self.LogWarning("Could not re-enable the master ip on"
4242 " the master, please restart manually: %s",
4246 def _UploadHelper(lu, nodes, fname):
4247 """Helper for uploading a file and showing warnings.
4250 if os.path.exists(fname):
4251 result = lu.rpc.call_upload_file(nodes, fname)
4252 for to_node, to_result in result.items():
4253 msg = to_result.fail_msg
4255 msg = ("Copy of file %s to node %s failed: %s" %
4256 (fname, to_node, msg))
4257 lu.proc.LogWarning(msg)
4260 def _ComputeAncillaryFiles(cluster, redist):
4261 """Compute files external to Ganeti which need to be consistent.
4263 @type redist: boolean
4264 @param redist: Whether to include files which need to be redistributed
4267 # Compute files for all nodes
4269 constants.SSH_KNOWN_HOSTS_FILE,
4270 constants.CONFD_HMAC_KEY,
4271 constants.CLUSTER_DOMAIN_SECRET_FILE,
4272 constants.SPICE_CERT_FILE,
4273 constants.SPICE_CACERT_FILE,
4274 constants.RAPI_USERS_FILE,
4278 files_all.update(constants.ALL_CERT_FILES)
4279 files_all.update(ssconf.SimpleStore().GetFileList())
4281 # we need to ship at least the RAPI certificate
4282 files_all.add(constants.RAPI_CERT_FILE)
4284 if cluster.modify_etc_hosts:
4285 files_all.add(constants.ETC_HOSTS)
4287 # Files which are optional, these must:
4288 # - be present in one other category as well
4289 # - either exist or not exist on all nodes of that category (mc, vm all)
4291 constants.RAPI_USERS_FILE,
4294 # Files which should only be on master candidates
4298 files_mc.add(constants.CLUSTER_CONF_FILE)
4300 # FIXME: this should also be replicated but Ganeti doesn't support files_mc
4302 files_mc.add(constants.DEFAULT_MASTER_SETUP_SCRIPT)
4304 # Files which should only be on VM-capable nodes
4305 files_vm = set(filename
4306 for hv_name in cluster.enabled_hypervisors
4307 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[0])
4309 files_opt |= set(filename
4310 for hv_name in cluster.enabled_hypervisors
4311 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[1])
4313 # Filenames in each category must be unique
4314 all_files_set = files_all | files_mc | files_vm
4315 assert (len(all_files_set) ==
4316 sum(map(len, [files_all, files_mc, files_vm]))), \
4317 "Found file listed in more than one file list"
4319 # Optional files must be present in one other category
4320 assert all_files_set.issuperset(files_opt), \
4321 "Optional file not in a different required list"
4323 return (files_all, files_opt, files_mc, files_vm)
4326 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
4327 """Distribute additional files which are part of the cluster configuration.
4329 ConfigWriter takes care of distributing the config and ssconf files, but
4330 there are more files which should be distributed to all nodes. This function
4331 makes sure those are copied.
4333 @param lu: calling logical unit
4334 @param additional_nodes: list of nodes not in the config to distribute to
4335 @type additional_vm: boolean
4336 @param additional_vm: whether the additional nodes are vm-capable or not
4339 # Gather target nodes
4340 cluster = lu.cfg.GetClusterInfo()
4341 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
4343 online_nodes = lu.cfg.GetOnlineNodeList()
4344 vm_nodes = lu.cfg.GetVmCapableNodeList()
4346 if additional_nodes is not None:
4347 online_nodes.extend(additional_nodes)
4349 vm_nodes.extend(additional_nodes)
4351 # Never distribute to master node
4352 for nodelist in [online_nodes, vm_nodes]:
4353 if master_info.name in nodelist:
4354 nodelist.remove(master_info.name)
4357 (files_all, _, files_mc, files_vm) = \
4358 _ComputeAncillaryFiles(cluster, True)
4360 # Never re-distribute configuration file from here
4361 assert not (constants.CLUSTER_CONF_FILE in files_all or
4362 constants.CLUSTER_CONF_FILE in files_vm)
4363 assert not files_mc, "Master candidates not handled in this function"
4366 (online_nodes, files_all),
4367 (vm_nodes, files_vm),
4371 for (node_list, files) in filemap:
4373 _UploadHelper(lu, node_list, fname)
4376 class LUClusterRedistConf(NoHooksLU):
4377 """Force the redistribution of cluster configuration.
4379 This is a very simple LU.
4384 def ExpandNames(self):
4385 self.needed_locks = {
4386 locking.LEVEL_NODE: locking.ALL_SET,
4388 self.share_locks[locking.LEVEL_NODE] = 1
4390 def Exec(self, feedback_fn):
4391 """Redistribute the configuration.
4394 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
4395 _RedistributeAncillaryFiles(self)
4398 class LUClusterActivateMasterIp(NoHooksLU):
4399 """Activate the master IP on the master node.
4402 def Exec(self, feedback_fn):
4403 """Activate the master IP.
4406 master_params = self.cfg.GetMasterNetworkParameters()
4407 ems = self.cfg.GetUseExternalMipScript()
4408 result = self.rpc.call_node_activate_master_ip(master_params.name,
4410 result.Raise("Could not activate the master IP")
4413 class LUClusterDeactivateMasterIp(NoHooksLU):
4414 """Deactivate the master IP on the master node.
4417 def Exec(self, feedback_fn):
4418 """Deactivate the master IP.
4421 master_params = self.cfg.GetMasterNetworkParameters()
4422 ems = self.cfg.GetUseExternalMipScript()
4423 result = self.rpc.call_node_deactivate_master_ip(master_params.name,
4425 result.Raise("Could not deactivate the master IP")
4428 def _WaitForSync(lu, instance, disks=None, oneshot=False):
4429 """Sleep and poll for an instance's disk to sync.
4432 if not instance.disks or disks is not None and not disks:
4435 disks = _ExpandCheckDisks(instance, disks)
4438 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
4440 node = instance.primary_node
4443 lu.cfg.SetDiskID(dev, node)
4445 # TODO: Convert to utils.Retry
4448 degr_retries = 10 # in seconds, as we sleep 1 second each time
4452 cumul_degraded = False
4453 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
4454 msg = rstats.fail_msg
4456 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
4459 raise errors.RemoteError("Can't contact node %s for mirror data,"
4460 " aborting." % node)
4463 rstats = rstats.payload
4465 for i, mstat in enumerate(rstats):
4467 lu.LogWarning("Can't compute data for node %s/%s",
4468 node, disks[i].iv_name)
4471 cumul_degraded = (cumul_degraded or
4472 (mstat.is_degraded and mstat.sync_percent is None))
4473 if mstat.sync_percent is not None:
4475 if mstat.estimated_time is not None:
4476 rem_time = ("%s remaining (estimated)" %
4477 utils.FormatSeconds(mstat.estimated_time))
4478 max_time = mstat.estimated_time
4480 rem_time = "no time estimate"
4481 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
4482 (disks[i].iv_name, mstat.sync_percent, rem_time))
4484 # if we're done but degraded, let's do a few small retries, to
4485 # make sure we see a stable and not transient situation; therefore
4486 # we force restart of the loop
4487 if (done or oneshot) and cumul_degraded and degr_retries > 0:
4488 logging.info("Degraded disks found, %d retries left", degr_retries)
4496 time.sleep(min(60, max_time))
4499 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
4500 return not cumul_degraded
4503 def _CheckDiskConsistency(lu, instance, dev, node, on_primary, ldisk=False):
4504 """Check that mirrors are not degraded.
4506 The ldisk parameter, if True, will change the test from the
4507 is_degraded attribute (which represents overall non-ok status for
4508 the device(s)) to the ldisk (representing the local storage status).
4511 lu.cfg.SetDiskID(dev, node)
4515 if on_primary or dev.AssembleOnSecondary():
4516 rstats = lu.rpc.call_blockdev_find(node, dev)
4517 msg = rstats.fail_msg
4519 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
4521 elif not rstats.payload:
4522 lu.LogWarning("Can't find disk on node %s", node)
4526 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
4528 result = result and not rstats.payload.is_degraded
4531 for child in dev.children:
4532 result = result and _CheckDiskConsistency(lu, instance, child, node,
4538 class LUOobCommand(NoHooksLU):
4539 """Logical unit for OOB handling.
4543 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
4545 def ExpandNames(self):
4546 """Gather locks we need.
4549 if self.op.node_names:
4550 self.op.node_names = _GetWantedNodes(self, self.op.node_names)
4551 lock_names = self.op.node_names
4553 lock_names = locking.ALL_SET
4555 self.needed_locks = {
4556 locking.LEVEL_NODE: lock_names,
4559 def CheckPrereq(self):
4560 """Check prerequisites.
4563 - the node exists in the configuration
4566 Any errors are signaled by raising errors.OpPrereqError.
4570 self.master_node = self.cfg.GetMasterNode()
4572 assert self.op.power_delay >= 0.0
4574 if self.op.node_names:
4575 if (self.op.command in self._SKIP_MASTER and
4576 self.master_node in self.op.node_names):
4577 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
4578 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
4580 if master_oob_handler:
4581 additional_text = ("run '%s %s %s' if you want to operate on the"
4582 " master regardless") % (master_oob_handler,
4586 additional_text = "it does not support out-of-band operations"
4588 raise errors.OpPrereqError(("Operating on the master node %s is not"
4589 " allowed for %s; %s") %
4590 (self.master_node, self.op.command,
4591 additional_text), errors.ECODE_INVAL)
4593 self.op.node_names = self.cfg.GetNodeList()
4594 if self.op.command in self._SKIP_MASTER:
4595 self.op.node_names.remove(self.master_node)
4597 if self.op.command in self._SKIP_MASTER:
4598 assert self.master_node not in self.op.node_names
4600 for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names):
4602 raise errors.OpPrereqError("Node %s not found" % node_name,
4605 self.nodes.append(node)
4607 if (not self.op.ignore_status and
4608 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
4609 raise errors.OpPrereqError(("Cannot power off node %s because it is"
4610 " not marked offline") % node_name,
4613 def Exec(self, feedback_fn):
4614 """Execute OOB and return result if we expect any.
4617 master_node = self.master_node
4620 for idx, node in enumerate(utils.NiceSort(self.nodes,
4621 key=lambda node: node.name)):
4622 node_entry = [(constants.RS_NORMAL, node.name)]
4623 ret.append(node_entry)
4625 oob_program = _SupportsOob(self.cfg, node)
4628 node_entry.append((constants.RS_UNAVAIL, None))
4631 logging.info("Executing out-of-band command '%s' using '%s' on %s",
4632 self.op.command, oob_program, node.name)
4633 result = self.rpc.call_run_oob(master_node, oob_program,
4634 self.op.command, node.name,
4638 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
4639 node.name, result.fail_msg)
4640 node_entry.append((constants.RS_NODATA, None))
4643 self._CheckPayload(result)
4644 except errors.OpExecError, err:
4645 self.LogWarning("Payload returned by node '%s' is not valid: %s",
4647 node_entry.append((constants.RS_NODATA, None))
4649 if self.op.command == constants.OOB_HEALTH:
4650 # For health we should log important events
4651 for item, status in result.payload:
4652 if status in [constants.OOB_STATUS_WARNING,
4653 constants.OOB_STATUS_CRITICAL]:
4654 self.LogWarning("Item '%s' on node '%s' has status '%s'",
4655 item, node.name, status)
4657 if self.op.command == constants.OOB_POWER_ON:
4659 elif self.op.command == constants.OOB_POWER_OFF:
4660 node.powered = False
4661 elif self.op.command == constants.OOB_POWER_STATUS:
4662 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
4663 if powered != node.powered:
4664 logging.warning(("Recorded power state (%s) of node '%s' does not"
4665 " match actual power state (%s)"), node.powered,
4668 # For configuration changing commands we should update the node
4669 if self.op.command in (constants.OOB_POWER_ON,
4670 constants.OOB_POWER_OFF):
4671 self.cfg.Update(node, feedback_fn)
4673 node_entry.append((constants.RS_NORMAL, result.payload))
4675 if (self.op.command == constants.OOB_POWER_ON and
4676 idx < len(self.nodes) - 1):
4677 time.sleep(self.op.power_delay)
4681 def _CheckPayload(self, result):
4682 """Checks if the payload is valid.
4684 @param result: RPC result
4685 @raises errors.OpExecError: If payload is not valid
4689 if self.op.command == constants.OOB_HEALTH:
4690 if not isinstance(result.payload, list):
4691 errs.append("command 'health' is expected to return a list but got %s" %
4692 type(result.payload))
4694 for item, status in result.payload:
4695 if status not in constants.OOB_STATUSES:
4696 errs.append("health item '%s' has invalid status '%s'" %
4699 if self.op.command == constants.OOB_POWER_STATUS:
4700 if not isinstance(result.payload, dict):
4701 errs.append("power-status is expected to return a dict but got %s" %
4702 type(result.payload))
4704 if self.op.command in [
4705 constants.OOB_POWER_ON,
4706 constants.OOB_POWER_OFF,
4707 constants.OOB_POWER_CYCLE,
4709 if result.payload is not None:
4710 errs.append("%s is expected to not return payload but got '%s'" %
4711 (self.op.command, result.payload))
4714 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
4715 utils.CommaJoin(errs))
4718 class _OsQuery(_QueryBase):
4719 FIELDS = query.OS_FIELDS
4721 def ExpandNames(self, lu):
4722 # Lock all nodes in shared mode
4723 # Temporary removal of locks, should be reverted later
4724 # TODO: reintroduce locks when they are lighter-weight
4725 lu.needed_locks = {}
4726 #self.share_locks[locking.LEVEL_NODE] = 1
4727 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4729 # The following variables interact with _QueryBase._GetNames
4731 self.wanted = self.names
4733 self.wanted = locking.ALL_SET
4735 self.do_locking = self.use_locking
4737 def DeclareLocks(self, lu, level):
4741 def _DiagnoseByOS(rlist):
4742 """Remaps a per-node return list into an a per-os per-node dictionary
4744 @param rlist: a map with node names as keys and OS objects as values
4747 @return: a dictionary with osnames as keys and as value another
4748 map, with nodes as keys and tuples of (path, status, diagnose,
4749 variants, parameters, api_versions) as values, eg::
4751 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
4752 (/srv/..., False, "invalid api")],
4753 "node2": [(/srv/..., True, "", [], [])]}
4758 # we build here the list of nodes that didn't fail the RPC (at RPC
4759 # level), so that nodes with a non-responding node daemon don't
4760 # make all OSes invalid
4761 good_nodes = [node_name for node_name in rlist
4762 if not rlist[node_name].fail_msg]
4763 for node_name, nr in rlist.items():
4764 if nr.fail_msg or not nr.payload:
4766 for (name, path, status, diagnose, variants,
4767 params, api_versions) in nr.payload:
4768 if name not in all_os:
4769 # build a list of nodes for this os containing empty lists
4770 # for each node in node_list
4772 for nname in good_nodes:
4773 all_os[name][nname] = []
4774 # convert params from [name, help] to (name, help)
4775 params = [tuple(v) for v in params]
4776 all_os[name][node_name].append((path, status, diagnose,
4777 variants, params, api_versions))
4780 def _GetQueryData(self, lu):
4781 """Computes the list of nodes and their attributes.
4784 # Locking is not used
4785 assert not (compat.any(lu.glm.is_owned(level)
4786 for level in locking.LEVELS
4787 if level != locking.LEVEL_CLUSTER) or
4788 self.do_locking or self.use_locking)
4790 valid_nodes = [node.name
4791 for node in lu.cfg.GetAllNodesInfo().values()
4792 if not node.offline and node.vm_capable]
4793 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
4794 cluster = lu.cfg.GetClusterInfo()
4798 for (os_name, os_data) in pol.items():
4799 info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
4800 hidden=(os_name in cluster.hidden_os),
4801 blacklisted=(os_name in cluster.blacklisted_os))
4805 api_versions = set()
4807 for idx, osl in enumerate(os_data.values()):
4808 info.valid = bool(info.valid and osl and osl[0][1])
4812 (node_variants, node_params, node_api) = osl[0][3:6]
4815 variants.update(node_variants)
4816 parameters.update(node_params)
4817 api_versions.update(node_api)
4819 # Filter out inconsistent values
4820 variants.intersection_update(node_variants)
4821 parameters.intersection_update(node_params)
4822 api_versions.intersection_update(node_api)
4824 info.variants = list(variants)
4825 info.parameters = list(parameters)
4826 info.api_versions = list(api_versions)
4828 data[os_name] = info
4830 # Prepare data in requested order
4831 return [data[name] for name in self._GetNames(lu, pol.keys(), None)
4835 class LUOsDiagnose(NoHooksLU):
4836 """Logical unit for OS diagnose/query.
4842 def _BuildFilter(fields, names):
4843 """Builds a filter for querying OSes.
4846 name_filter = qlang.MakeSimpleFilter("name", names)
4848 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
4849 # respective field is not requested
4850 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]]
4851 for fname in ["hidden", "blacklisted"]
4852 if fname not in fields]
4853 if "valid" not in fields:
4854 status_filter.append([qlang.OP_TRUE, "valid"])
4857 status_filter.insert(0, qlang.OP_AND)
4859 status_filter = None
4861 if name_filter and status_filter:
4862 return [qlang.OP_AND, name_filter, status_filter]
4866 return status_filter
4868 def CheckArguments(self):
4869 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names),
4870 self.op.output_fields, False)
4872 def ExpandNames(self):
4873 self.oq.ExpandNames(self)
4875 def Exec(self, feedback_fn):
4876 return self.oq.OldStyleQuery(self)
4879 class LUNodeRemove(LogicalUnit):
4880 """Logical unit for removing a node.
4883 HPATH = "node-remove"
4884 HTYPE = constants.HTYPE_NODE
4886 def BuildHooksEnv(self):
4891 "OP_TARGET": self.op.node_name,
4892 "NODE_NAME": self.op.node_name,
4895 def BuildHooksNodes(self):
4896 """Build hooks nodes.
4898 This doesn't run on the target node in the pre phase as a failed
4899 node would then be impossible to remove.
4902 all_nodes = self.cfg.GetNodeList()
4904 all_nodes.remove(self.op.node_name)
4907 return (all_nodes, all_nodes)
4909 def CheckPrereq(self):
4910 """Check prerequisites.
4913 - the node exists in the configuration
4914 - it does not have primary or secondary instances
4915 - it's not the master
4917 Any errors are signaled by raising errors.OpPrereqError.
4920 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4921 node = self.cfg.GetNodeInfo(self.op.node_name)
4922 assert node is not None
4924 masternode = self.cfg.GetMasterNode()
4925 if node.name == masternode:
4926 raise errors.OpPrereqError("Node is the master node, failover to another"
4927 " node is required", errors.ECODE_INVAL)
4929 for instance_name, instance in self.cfg.GetAllInstancesInfo().items():
4930 if node.name in instance.all_nodes:
4931 raise errors.OpPrereqError("Instance %s is still running on the node,"
4932 " please remove first" % instance_name,
4934 self.op.node_name = node.name
4937 def Exec(self, feedback_fn):
4938 """Removes the node from the cluster.
4942 logging.info("Stopping the node daemon and removing configs from node %s",
4945 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
4947 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
4950 # Promote nodes to master candidate as needed
4951 _AdjustCandidatePool(self, exceptions=[node.name])
4952 self.context.RemoveNode(node.name)
4954 # Run post hooks on the node before it's removed
4955 _RunPostHook(self, node.name)
4957 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
4958 msg = result.fail_msg
4960 self.LogWarning("Errors encountered on the remote node while leaving"
4961 " the cluster: %s", msg)
4963 # Remove node from our /etc/hosts
4964 if self.cfg.GetClusterInfo().modify_etc_hosts:
4965 master_node = self.cfg.GetMasterNode()
4966 result = self.rpc.call_etc_hosts_modify(master_node,
4967 constants.ETC_HOSTS_REMOVE,
4969 result.Raise("Can't update hosts file with new host data")
4970 _RedistributeAncillaryFiles(self)
4973 class _NodeQuery(_QueryBase):
4974 FIELDS = query.NODE_FIELDS
4976 def ExpandNames(self, lu):
4977 lu.needed_locks = {}
4978 lu.share_locks = _ShareAll()
4981 self.wanted = _GetWantedNodes(lu, self.names)
4983 self.wanted = locking.ALL_SET
4985 self.do_locking = (self.use_locking and
4986 query.NQ_LIVE in self.requested_data)
4989 # If any non-static field is requested we need to lock the nodes
4990 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
4992 def DeclareLocks(self, lu, level):
4995 def _GetQueryData(self, lu):
4996 """Computes the list of nodes and their attributes.
4999 all_info = lu.cfg.GetAllNodesInfo()
5001 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
5003 # Gather data as requested
5004 if query.NQ_LIVE in self.requested_data:
5005 # filter out non-vm_capable nodes
5006 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
5008 node_data = lu.rpc.call_node_info(toquery_nodes, [lu.cfg.GetVGName()],
5009 [lu.cfg.GetHypervisorType()])
5010 live_data = dict((name, _MakeLegacyNodeInfo(nresult.payload))
5011 for (name, nresult) in node_data.items()
5012 if not nresult.fail_msg and nresult.payload)
5016 if query.NQ_INST in self.requested_data:
5017 node_to_primary = dict([(name, set()) for name in nodenames])
5018 node_to_secondary = dict([(name, set()) for name in nodenames])
5020 inst_data = lu.cfg.GetAllInstancesInfo()
5022 for inst in inst_data.values():
5023 if inst.primary_node in node_to_primary:
5024 node_to_primary[inst.primary_node].add(inst.name)
5025 for secnode in inst.secondary_nodes:
5026 if secnode in node_to_secondary:
5027 node_to_secondary[secnode].add(inst.name)
5029 node_to_primary = None
5030 node_to_secondary = None
5032 if query.NQ_OOB in self.requested_data:
5033 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
5034 for name, node in all_info.iteritems())
5038 if query.NQ_GROUP in self.requested_data:
5039 groups = lu.cfg.GetAllNodeGroupsInfo()
5043 return query.NodeQueryData([all_info[name] for name in nodenames],
5044 live_data, lu.cfg.GetMasterNode(),
5045 node_to_primary, node_to_secondary, groups,
5046 oob_support, lu.cfg.GetClusterInfo())
5049 class LUNodeQuery(NoHooksLU):
5050 """Logical unit for querying nodes.
5053 # pylint: disable=W0142
5056 def CheckArguments(self):
5057 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
5058 self.op.output_fields, self.op.use_locking)
5060 def ExpandNames(self):
5061 self.nq.ExpandNames(self)
5063 def DeclareLocks(self, level):
5064 self.nq.DeclareLocks(self, level)
5066 def Exec(self, feedback_fn):
5067 return self.nq.OldStyleQuery(self)
5070 class LUNodeQueryvols(NoHooksLU):
5071 """Logical unit for getting volumes on node(s).
5075 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
5076 _FIELDS_STATIC = utils.FieldSet("node")
5078 def CheckArguments(self):
5079 _CheckOutputFields(static=self._FIELDS_STATIC,
5080 dynamic=self._FIELDS_DYNAMIC,
5081 selected=self.op.output_fields)
5083 def ExpandNames(self):
5084 self.share_locks = _ShareAll()
5085 self.needed_locks = {}
5087 if not self.op.nodes:
5088 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5090 self.needed_locks[locking.LEVEL_NODE] = \
5091 _GetWantedNodes(self, self.op.nodes)
5093 def Exec(self, feedback_fn):
5094 """Computes the list of nodes and their attributes.
5097 nodenames = self.owned_locks(locking.LEVEL_NODE)
5098 volumes = self.rpc.call_node_volumes(nodenames)
5100 ilist = self.cfg.GetAllInstancesInfo()
5101 vol2inst = _MapInstanceDisksToNodes(ilist.values())
5104 for node in nodenames:
5105 nresult = volumes[node]
5108 msg = nresult.fail_msg
5110 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
5113 node_vols = sorted(nresult.payload,
5114 key=operator.itemgetter("dev"))
5116 for vol in node_vols:
5118 for field in self.op.output_fields:
5121 elif field == "phys":
5125 elif field == "name":
5127 elif field == "size":
5128 val = int(float(vol["size"]))
5129 elif field == "instance":
5130 val = vol2inst.get((node, vol["vg"] + "/" + vol["name"]), "-")
5132 raise errors.ParameterError(field)
5133 node_output.append(str(val))
5135 output.append(node_output)
5140 class LUNodeQueryStorage(NoHooksLU):
5141 """Logical unit for getting information on storage units on node(s).
5144 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
5147 def CheckArguments(self):
5148 _CheckOutputFields(static=self._FIELDS_STATIC,
5149 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
5150 selected=self.op.output_fields)
5152 def ExpandNames(self):
5153 self.share_locks = _ShareAll()
5154 self.needed_locks = {}
5157 self.needed_locks[locking.LEVEL_NODE] = \
5158 _GetWantedNodes(self, self.op.nodes)
5160 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5162 def Exec(self, feedback_fn):
5163 """Computes the list of nodes and their attributes.
5166 self.nodes = self.owned_locks(locking.LEVEL_NODE)
5168 # Always get name to sort by
5169 if constants.SF_NAME in self.op.output_fields:
5170 fields = self.op.output_fields[:]
5172 fields = [constants.SF_NAME] + self.op.output_fields
5174 # Never ask for node or type as it's only known to the LU
5175 for extra in [constants.SF_NODE, constants.SF_TYPE]:
5176 while extra in fields:
5177 fields.remove(extra)
5179 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
5180 name_idx = field_idx[constants.SF_NAME]
5182 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
5183 data = self.rpc.call_storage_list(self.nodes,
5184 self.op.storage_type, st_args,
5185 self.op.name, fields)
5189 for node in utils.NiceSort(self.nodes):
5190 nresult = data[node]
5194 msg = nresult.fail_msg
5196 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
5199 rows = dict([(row[name_idx], row) for row in nresult.payload])
5201 for name in utils.NiceSort(rows.keys()):
5206 for field in self.op.output_fields:
5207 if field == constants.SF_NODE:
5209 elif field == constants.SF_TYPE:
5210 val = self.op.storage_type
5211 elif field in field_idx:
5212 val = row[field_idx[field]]
5214 raise errors.ParameterError(field)
5223 class _InstanceQuery(_QueryBase):
5224 FIELDS = query.INSTANCE_FIELDS
5226 def ExpandNames(self, lu):
5227 lu.needed_locks = {}
5228 lu.share_locks = _ShareAll()
5231 self.wanted = _GetWantedInstances(lu, self.names)
5233 self.wanted = locking.ALL_SET
5235 self.do_locking = (self.use_locking and
5236 query.IQ_LIVE in self.requested_data)
5238 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5239 lu.needed_locks[locking.LEVEL_NODEGROUP] = []
5240 lu.needed_locks[locking.LEVEL_NODE] = []
5241 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5243 self.do_grouplocks = (self.do_locking and
5244 query.IQ_NODES in self.requested_data)
5246 def DeclareLocks(self, lu, level):
5248 if level == locking.LEVEL_NODEGROUP and self.do_grouplocks:
5249 assert not lu.needed_locks[locking.LEVEL_NODEGROUP]
5251 # Lock all groups used by instances optimistically; this requires going
5252 # via the node before it's locked, requiring verification later on
5253 lu.needed_locks[locking.LEVEL_NODEGROUP] = \
5255 for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
5256 for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
5257 elif level == locking.LEVEL_NODE:
5258 lu._LockInstancesNodes() # pylint: disable=W0212
5261 def _CheckGroupLocks(lu):
5262 owned_instances = frozenset(lu.owned_locks(locking.LEVEL_INSTANCE))
5263 owned_groups = frozenset(lu.owned_locks(locking.LEVEL_NODEGROUP))
5265 # Check if node groups for locked instances are still correct
5266 for instance_name in owned_instances:
5267 _CheckInstanceNodeGroups(lu.cfg, instance_name, owned_groups)
5269 def _GetQueryData(self, lu):
5270 """Computes the list of instances and their attributes.
5273 if self.do_grouplocks:
5274 self._CheckGroupLocks(lu)
5276 cluster = lu.cfg.GetClusterInfo()
5277 all_info = lu.cfg.GetAllInstancesInfo()
5279 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
5281 instance_list = [all_info[name] for name in instance_names]
5282 nodes = frozenset(itertools.chain(*(inst.all_nodes
5283 for inst in instance_list)))
5284 hv_list = list(set([inst.hypervisor for inst in instance_list]))
5287 wrongnode_inst = set()
5289 # Gather data as requested
5290 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
5292 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
5294 result = node_data[name]
5296 # offline nodes will be in both lists
5297 assert result.fail_msg
5298 offline_nodes.append(name)
5300 bad_nodes.append(name)
5301 elif result.payload:
5302 for inst in result.payload:
5303 if inst in all_info:
5304 if all_info[inst].primary_node == name:
5305 live_data.update(result.payload)
5307 wrongnode_inst.add(inst)
5309 # orphan instance; we don't list it here as we don't
5310 # handle this case yet in the output of instance listing
5311 logging.warning("Orphan instance '%s' found on node %s",
5313 # else no instance is alive
5317 if query.IQ_DISKUSAGE in self.requested_data:
5318 disk_usage = dict((inst.name,
5319 _ComputeDiskSize(inst.disk_template,
5320 [{constants.IDISK_SIZE: disk.size}
5321 for disk in inst.disks]))
5322 for inst in instance_list)
5326 if query.IQ_CONSOLE in self.requested_data:
5328 for inst in instance_list:
5329 if inst.name in live_data:
5330 # Instance is running
5331 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
5333 consinfo[inst.name] = None
5334 assert set(consinfo.keys()) == set(instance_names)
5338 if query.IQ_NODES in self.requested_data:
5339 node_names = set(itertools.chain(*map(operator.attrgetter("all_nodes"),
5341 nodes = dict(lu.cfg.GetMultiNodeInfo(node_names))
5342 groups = dict((uuid, lu.cfg.GetNodeGroup(uuid))
5343 for uuid in set(map(operator.attrgetter("group"),
5349 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
5350 disk_usage, offline_nodes, bad_nodes,
5351 live_data, wrongnode_inst, consinfo,
5355 class LUQuery(NoHooksLU):
5356 """Query for resources/items of a certain kind.
5359 # pylint: disable=W0142
5362 def CheckArguments(self):
5363 qcls = _GetQueryImplementation(self.op.what)
5365 self.impl = qcls(self.op.qfilter, self.op.fields, self.op.use_locking)
5367 def ExpandNames(self):
5368 self.impl.ExpandNames(self)
5370 def DeclareLocks(self, level):
5371 self.impl.DeclareLocks(self, level)
5373 def Exec(self, feedback_fn):
5374 return self.impl.NewStyleQuery(self)
5377 class LUQueryFields(NoHooksLU):
5378 """Query for resources/items of a certain kind.
5381 # pylint: disable=W0142
5384 def CheckArguments(self):
5385 self.qcls = _GetQueryImplementation(self.op.what)
5387 def ExpandNames(self):
5388 self.needed_locks = {}
5390 def Exec(self, feedback_fn):
5391 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
5394 class LUNodeModifyStorage(NoHooksLU):
5395 """Logical unit for modifying a storage volume on a node.
5400 def CheckArguments(self):
5401 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5403 storage_type = self.op.storage_type
5406 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
5408 raise errors.OpPrereqError("Storage units of type '%s' can not be"
5409 " modified" % storage_type,
5412 diff = set(self.op.changes.keys()) - modifiable
5414 raise errors.OpPrereqError("The following fields can not be modified for"
5415 " storage units of type '%s': %r" %
5416 (storage_type, list(diff)),
5419 def ExpandNames(self):
5420 self.needed_locks = {
5421 locking.LEVEL_NODE: self.op.node_name,
5424 def Exec(self, feedback_fn):
5425 """Computes the list of nodes and their attributes.
5428 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
5429 result = self.rpc.call_storage_modify(self.op.node_name,
5430 self.op.storage_type, st_args,
5431 self.op.name, self.op.changes)
5432 result.Raise("Failed to modify storage unit '%s' on %s" %
5433 (self.op.name, self.op.node_name))
5436 class LUNodeAdd(LogicalUnit):
5437 """Logical unit for adding node to the cluster.
5441 HTYPE = constants.HTYPE_NODE
5442 _NFLAGS = ["master_capable", "vm_capable"]
5444 def CheckArguments(self):
5445 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
5446 # validate/normalize the node name
5447 self.hostname = netutils.GetHostname(name=self.op.node_name,
5448 family=self.primary_ip_family)
5449 self.op.node_name = self.hostname.name
5451 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
5452 raise errors.OpPrereqError("Cannot readd the master node",
5455 if self.op.readd and self.op.group:
5456 raise errors.OpPrereqError("Cannot pass a node group when a node is"
5457 " being readded", errors.ECODE_INVAL)
5459 def BuildHooksEnv(self):
5462 This will run on all nodes before, and on all nodes + the new node after.
5466 "OP_TARGET": self.op.node_name,
5467 "NODE_NAME": self.op.node_name,
5468 "NODE_PIP": self.op.primary_ip,
5469 "NODE_SIP": self.op.secondary_ip,
5470 "MASTER_CAPABLE": str(self.op.master_capable),
5471 "VM_CAPABLE": str(self.op.vm_capable),
5474 def BuildHooksNodes(self):
5475 """Build hooks nodes.
5478 # Exclude added node
5479 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name]))
5480 post_nodes = pre_nodes + [self.op.node_name, ]
5482 return (pre_nodes, post_nodes)
5484 def CheckPrereq(self):
5485 """Check prerequisites.
5488 - the new node is not already in the config
5490 - its parameters (single/dual homed) matches the cluster
5492 Any errors are signaled by raising errors.OpPrereqError.
5496 hostname = self.hostname
5497 node = hostname.name
5498 primary_ip = self.op.primary_ip = hostname.ip
5499 if self.op.secondary_ip is None:
5500 if self.primary_ip_family == netutils.IP6Address.family:
5501 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
5502 " IPv4 address must be given as secondary",
5504 self.op.secondary_ip = primary_ip
5506 secondary_ip = self.op.secondary_ip
5507 if not netutils.IP4Address.IsValid(secondary_ip):
5508 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5509 " address" % secondary_ip, errors.ECODE_INVAL)
5511 node_list = cfg.GetNodeList()
5512 if not self.op.readd and node in node_list:
5513 raise errors.OpPrereqError("Node %s is already in the configuration" %
5514 node, errors.ECODE_EXISTS)
5515 elif self.op.readd and node not in node_list:
5516 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
5519 self.changed_primary_ip = False
5521 for existing_node_name, existing_node in cfg.GetMultiNodeInfo(node_list):
5522 if self.op.readd and node == existing_node_name:
5523 if existing_node.secondary_ip != secondary_ip:
5524 raise errors.OpPrereqError("Readded node doesn't have the same IP"
5525 " address configuration as before",
5527 if existing_node.primary_ip != primary_ip:
5528 self.changed_primary_ip = True
5532 if (existing_node.primary_ip == primary_ip or
5533 existing_node.secondary_ip == primary_ip or
5534 existing_node.primary_ip == secondary_ip or
5535 existing_node.secondary_ip == secondary_ip):
5536 raise errors.OpPrereqError("New node ip address(es) conflict with"
5537 " existing node %s" % existing_node.name,
5538 errors.ECODE_NOTUNIQUE)
5540 # After this 'if' block, None is no longer a valid value for the
5541 # _capable op attributes
5543 old_node = self.cfg.GetNodeInfo(node)
5544 assert old_node is not None, "Can't retrieve locked node %s" % node
5545 for attr in self._NFLAGS:
5546 if getattr(self.op, attr) is None:
5547 setattr(self.op, attr, getattr(old_node, attr))
5549 for attr in self._NFLAGS:
5550 if getattr(self.op, attr) is None:
5551 setattr(self.op, attr, True)
5553 if self.op.readd and not self.op.vm_capable:
5554 pri, sec = cfg.GetNodeInstances(node)
5556 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
5557 " flag set to false, but it already holds"
5558 " instances" % node,
5561 # check that the type of the node (single versus dual homed) is the
5562 # same as for the master
5563 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
5564 master_singlehomed = myself.secondary_ip == myself.primary_ip
5565 newbie_singlehomed = secondary_ip == primary_ip
5566 if master_singlehomed != newbie_singlehomed:
5567 if master_singlehomed:
5568 raise errors.OpPrereqError("The master has no secondary ip but the"
5569 " new node has one",
5572 raise errors.OpPrereqError("The master has a secondary ip but the"
5573 " new node doesn't have one",
5576 # checks reachability
5577 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
5578 raise errors.OpPrereqError("Node not reachable by ping",
5579 errors.ECODE_ENVIRON)
5581 if not newbie_singlehomed:
5582 # check reachability from my secondary ip to newbie's secondary ip
5583 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
5584 source=myself.secondary_ip):
5585 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5586 " based ping to node daemon port",
5587 errors.ECODE_ENVIRON)
5594 if self.op.master_capable:
5595 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
5597 self.master_candidate = False
5600 self.new_node = old_node
5602 node_group = cfg.LookupNodeGroup(self.op.group)
5603 self.new_node = objects.Node(name=node,
5604 primary_ip=primary_ip,
5605 secondary_ip=secondary_ip,
5606 master_candidate=self.master_candidate,
5607 offline=False, drained=False,
5610 if self.op.ndparams:
5611 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
5613 if self.op.hv_state:
5614 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None)
5616 if self.op.disk_state:
5617 self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None)
5619 # TODO: If we need to have multiple DnsOnlyRunner we probably should make
5620 # it a property on the base class.
5621 result = rpc.DnsOnlyRunner().call_version([node])[node]
5622 result.Raise("Can't get version information from node %s" % node)
5623 if constants.PROTOCOL_VERSION == result.payload:
5624 logging.info("Communication to node %s fine, sw version %s match",
5625 node, result.payload)
5627 raise errors.OpPrereqError("Version mismatch master version %s,"
5628 " node version %s" %
5629 (constants.PROTOCOL_VERSION, result.payload),
5630 errors.ECODE_ENVIRON)
5632 def Exec(self, feedback_fn):
5633 """Adds the new node to the cluster.
5636 new_node = self.new_node
5637 node = new_node.name
5639 assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
5642 # We adding a new node so we assume it's powered
5643 new_node.powered = True
5645 # for re-adds, reset the offline/drained/master-candidate flags;
5646 # we need to reset here, otherwise offline would prevent RPC calls
5647 # later in the procedure; this also means that if the re-add
5648 # fails, we are left with a non-offlined, broken node
5650 new_node.drained = new_node.offline = False # pylint: disable=W0201
5651 self.LogInfo("Readding a node, the offline/drained flags were reset")
5652 # if we demote the node, we do cleanup later in the procedure
5653 new_node.master_candidate = self.master_candidate
5654 if self.changed_primary_ip:
5655 new_node.primary_ip = self.op.primary_ip
5657 # copy the master/vm_capable flags
5658 for attr in self._NFLAGS:
5659 setattr(new_node, attr, getattr(self.op, attr))
5661 # notify the user about any possible mc promotion
5662 if new_node.master_candidate:
5663 self.LogInfo("Node will be a master candidate")
5665 if self.op.ndparams:
5666 new_node.ndparams = self.op.ndparams
5668 new_node.ndparams = {}
5670 if self.op.hv_state:
5671 new_node.hv_state_static = self.new_hv_state
5673 if self.op.disk_state:
5674 new_node.disk_state_static = self.new_disk_state
5676 # Add node to our /etc/hosts, and add key to known_hosts
5677 if self.cfg.GetClusterInfo().modify_etc_hosts:
5678 master_node = self.cfg.GetMasterNode()
5679 result = self.rpc.call_etc_hosts_modify(master_node,
5680 constants.ETC_HOSTS_ADD,
5683 result.Raise("Can't update hosts file with new host data")
5685 if new_node.secondary_ip != new_node.primary_ip:
5686 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
5689 node_verify_list = [self.cfg.GetMasterNode()]
5690 node_verify_param = {
5691 constants.NV_NODELIST: ([node], {}),
5692 # TODO: do a node-net-test as well?
5695 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
5696 self.cfg.GetClusterName())
5697 for verifier in node_verify_list:
5698 result[verifier].Raise("Cannot communicate with node %s" % verifier)
5699 nl_payload = result[verifier].payload[constants.NV_NODELIST]
5701 for failed in nl_payload:
5702 feedback_fn("ssh/hostname verification failed"
5703 " (checking from %s): %s" %
5704 (verifier, nl_payload[failed]))
5705 raise errors.OpExecError("ssh/hostname verification failed")
5708 _RedistributeAncillaryFiles(self)
5709 self.context.ReaddNode(new_node)
5710 # make sure we redistribute the config
5711 self.cfg.Update(new_node, feedback_fn)
5712 # and make sure the new node will not have old files around
5713 if not new_node.master_candidate:
5714 result = self.rpc.call_node_demote_from_mc(new_node.name)
5715 msg = result.fail_msg
5717 self.LogWarning("Node failed to demote itself from master"
5718 " candidate status: %s" % msg)
5720 _RedistributeAncillaryFiles(self, additional_nodes=[node],
5721 additional_vm=self.op.vm_capable)
5722 self.context.AddNode(new_node, self.proc.GetECId())
5725 class LUNodeSetParams(LogicalUnit):
5726 """Modifies the parameters of a node.
5728 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
5729 to the node role (as _ROLE_*)
5730 @cvar _R2F: a dictionary from node role to tuples of flags
5731 @cvar _FLAGS: a list of attribute names corresponding to the flags
5734 HPATH = "node-modify"
5735 HTYPE = constants.HTYPE_NODE
5737 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
5739 (True, False, False): _ROLE_CANDIDATE,
5740 (False, True, False): _ROLE_DRAINED,
5741 (False, False, True): _ROLE_OFFLINE,
5742 (False, False, False): _ROLE_REGULAR,
5744 _R2F = dict((v, k) for k, v in _F2R.items())
5745 _FLAGS = ["master_candidate", "drained", "offline"]
5747 def CheckArguments(self):
5748 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5749 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
5750 self.op.master_capable, self.op.vm_capable,
5751 self.op.secondary_ip, self.op.ndparams, self.op.hv_state,
5753 if all_mods.count(None) == len(all_mods):
5754 raise errors.OpPrereqError("Please pass at least one modification",
5756 if all_mods.count(True) > 1:
5757 raise errors.OpPrereqError("Can't set the node into more than one"
5758 " state at the same time",
5761 # Boolean value that tells us whether we might be demoting from MC
5762 self.might_demote = (self.op.master_candidate == False or
5763 self.op.offline == True or
5764 self.op.drained == True or
5765 self.op.master_capable == False)
5767 if self.op.secondary_ip:
5768 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
5769 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
5770 " address" % self.op.secondary_ip,
5773 self.lock_all = self.op.auto_promote and self.might_demote
5774 self.lock_instances = self.op.secondary_ip is not None
5776 def _InstanceFilter(self, instance):
5777 """Filter for getting affected instances.
5780 return (instance.disk_template in constants.DTS_INT_MIRROR and
5781 self.op.node_name in instance.all_nodes)
5783 def ExpandNames(self):
5785 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
5787 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
5789 # Since modifying a node can have severe effects on currently running
5790 # operations the resource lock is at least acquired in shared mode
5791 self.needed_locks[locking.LEVEL_NODE_RES] = \
5792 self.needed_locks[locking.LEVEL_NODE]
5794 # Get node resource and instance locks in shared mode; they are not used
5795 # for anything but read-only access
5796 self.share_locks[locking.LEVEL_NODE_RES] = 1
5797 self.share_locks[locking.LEVEL_INSTANCE] = 1
5799 if self.lock_instances:
5800 self.needed_locks[locking.LEVEL_INSTANCE] = \
5801 frozenset(self.cfg.GetInstancesInfoByFilter(self._InstanceFilter))
5803 def BuildHooksEnv(self):
5806 This runs on the master node.
5810 "OP_TARGET": self.op.node_name,
5811 "MASTER_CANDIDATE": str(self.op.master_candidate),
5812 "OFFLINE": str(self.op.offline),
5813 "DRAINED": str(self.op.drained),
5814 "MASTER_CAPABLE": str(self.op.master_capable),
5815 "VM_CAPABLE": str(self.op.vm_capable),
5818 def BuildHooksNodes(self):
5819 """Build hooks nodes.
5822 nl = [self.cfg.GetMasterNode(), self.op.node_name]
5825 def CheckPrereq(self):
5826 """Check prerequisites.
5828 This only checks the instance list against the existing names.
5831 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
5833 if self.lock_instances:
5834 affected_instances = \
5835 self.cfg.GetInstancesInfoByFilter(self._InstanceFilter)
5837 # Verify instance locks
5838 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
5839 wanted_instances = frozenset(affected_instances.keys())
5840 if wanted_instances - owned_instances:
5841 raise errors.OpPrereqError("Instances affected by changing node %s's"
5842 " secondary IP address have changed since"
5843 " locks were acquired, wanted '%s', have"
5844 " '%s'; retry the operation" %
5846 utils.CommaJoin(wanted_instances),
5847 utils.CommaJoin(owned_instances)),
5850 affected_instances = None
5852 if (self.op.master_candidate is not None or
5853 self.op.drained is not None or
5854 self.op.offline is not None):
5855 # we can't change the master's node flags
5856 if self.op.node_name == self.cfg.GetMasterNode():
5857 raise errors.OpPrereqError("The master role can be changed"
5858 " only via master-failover",
5861 if self.op.master_candidate and not node.master_capable:
5862 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
5863 " it a master candidate" % node.name,
5866 if self.op.vm_capable == False:
5867 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
5869 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
5870 " the vm_capable flag" % node.name,
5873 if node.master_candidate and self.might_demote and not self.lock_all:
5874 assert not self.op.auto_promote, "auto_promote set but lock_all not"
5875 # check if after removing the current node, we're missing master
5877 (mc_remaining, mc_should, _) = \
5878 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
5879 if mc_remaining < mc_should:
5880 raise errors.OpPrereqError("Not enough master candidates, please"
5881 " pass auto promote option to allow"
5882 " promotion", errors.ECODE_STATE)
5884 self.old_flags = old_flags = (node.master_candidate,
5885 node.drained, node.offline)
5886 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
5887 self.old_role = old_role = self._F2R[old_flags]
5889 # Check for ineffective changes
5890 for attr in self._FLAGS:
5891 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
5892 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
5893 setattr(self.op, attr, None)
5895 # Past this point, any flag change to False means a transition
5896 # away from the respective state, as only real changes are kept
5898 # TODO: We might query the real power state if it supports OOB
5899 if _SupportsOob(self.cfg, node):
5900 if self.op.offline is False and not (node.powered or
5901 self.op.powered == True):
5902 raise errors.OpPrereqError(("Node %s needs to be turned on before its"
5903 " offline status can be reset") %
5905 elif self.op.powered is not None:
5906 raise errors.OpPrereqError(("Unable to change powered state for node %s"
5907 " as it does not support out-of-band"
5908 " handling") % self.op.node_name)
5910 # If we're being deofflined/drained, we'll MC ourself if needed
5911 if (self.op.drained == False or self.op.offline == False or
5912 (self.op.master_capable and not node.master_capable)):
5913 if _DecideSelfPromotion(self):
5914 self.op.master_candidate = True
5915 self.LogInfo("Auto-promoting node to master candidate")
5917 # If we're no longer master capable, we'll demote ourselves from MC
5918 if self.op.master_capable == False and node.master_candidate:
5919 self.LogInfo("Demoting from master candidate")
5920 self.op.master_candidate = False
5923 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
5924 if self.op.master_candidate:
5925 new_role = self._ROLE_CANDIDATE
5926 elif self.op.drained:
5927 new_role = self._ROLE_DRAINED
5928 elif self.op.offline:
5929 new_role = self._ROLE_OFFLINE
5930 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
5931 # False is still in new flags, which means we're un-setting (the
5933 new_role = self._ROLE_REGULAR
5934 else: # no new flags, nothing, keep old role
5937 self.new_role = new_role
5939 if old_role == self._ROLE_OFFLINE and new_role != old_role:
5940 # Trying to transition out of offline status
5941 result = self.rpc.call_version([node.name])[node.name]
5943 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
5944 " to report its version: %s" %
5945 (node.name, result.fail_msg),
5948 self.LogWarning("Transitioning node from offline to online state"
5949 " without using re-add. Please make sure the node"
5952 if self.op.secondary_ip:
5953 # Ok even without locking, because this can't be changed by any LU
5954 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
5955 master_singlehomed = master.secondary_ip == master.primary_ip
5956 if master_singlehomed and self.op.secondary_ip:
5957 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
5958 " homed cluster", errors.ECODE_INVAL)
5960 assert not (frozenset(affected_instances) -
5961 self.owned_locks(locking.LEVEL_INSTANCE))
5964 if affected_instances:
5965 raise errors.OpPrereqError("Cannot change secondary IP address:"
5966 " offline node has instances (%s)"
5967 " configured to use it" %
5968 utils.CommaJoin(affected_instances.keys()))
5970 # On online nodes, check that no instances are running, and that
5971 # the node has the new ip and we can reach it.
5972 for instance in affected_instances.values():
5973 _CheckInstanceState(self, instance, INSTANCE_DOWN,
5974 msg="cannot change secondary ip")
5976 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
5977 if master.name != node.name:
5978 # check reachability from master secondary ip to new secondary ip
5979 if not netutils.TcpPing(self.op.secondary_ip,
5980 constants.DEFAULT_NODED_PORT,
5981 source=master.secondary_ip):
5982 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
5983 " based ping to node daemon port",
5984 errors.ECODE_ENVIRON)
5986 if self.op.ndparams:
5987 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
5988 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
5989 self.new_ndparams = new_ndparams
5991 if self.op.hv_state:
5992 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
5993 self.node.hv_state_static)
5995 if self.op.disk_state:
5996 self.new_disk_state = \
5997 _MergeAndVerifyDiskState(self.op.disk_state,
5998 self.node.disk_state_static)
6000 def Exec(self, feedback_fn):
6005 old_role = self.old_role
6006 new_role = self.new_role
6010 if self.op.ndparams:
6011 node.ndparams = self.new_ndparams
6013 if self.op.powered is not None:
6014 node.powered = self.op.powered
6016 if self.op.hv_state:
6017 node.hv_state_static = self.new_hv_state
6019 if self.op.disk_state:
6020 node.disk_state_static = self.new_disk_state
6022 for attr in ["master_capable", "vm_capable"]:
6023 val = getattr(self.op, attr)
6025 setattr(node, attr, val)
6026 result.append((attr, str(val)))
6028 if new_role != old_role:
6029 # Tell the node to demote itself, if no longer MC and not offline
6030 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
6031 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
6033 self.LogWarning("Node failed to demote itself: %s", msg)
6035 new_flags = self._R2F[new_role]
6036 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
6038 result.append((desc, str(nf)))
6039 (node.master_candidate, node.drained, node.offline) = new_flags
6041 # we locked all nodes, we adjust the CP before updating this node
6043 _AdjustCandidatePool(self, [node.name])
6045 if self.op.secondary_ip:
6046 node.secondary_ip = self.op.secondary_ip
6047 result.append(("secondary_ip", self.op.secondary_ip))
6049 # this will trigger configuration file update, if needed
6050 self.cfg.Update(node, feedback_fn)
6052 # this will trigger job queue propagation or cleanup if the mc
6054 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
6055 self.context.ReaddNode(node)
6060 class LUNodePowercycle(NoHooksLU):
6061 """Powercycles a node.
6066 def CheckArguments(self):
6067 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6068 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
6069 raise errors.OpPrereqError("The node is the master and the force"
6070 " parameter was not set",
6073 def ExpandNames(self):
6074 """Locking for PowercycleNode.
6076 This is a last-resort option and shouldn't block on other
6077 jobs. Therefore, we grab no locks.
6080 self.needed_locks = {}
6082 def Exec(self, feedback_fn):
6086 result = self.rpc.call_node_powercycle(self.op.node_name,
6087 self.cfg.GetHypervisorType())
6088 result.Raise("Failed to schedule the reboot")
6089 return result.payload
6092 class LUClusterQuery(NoHooksLU):
6093 """Query cluster configuration.
6098 def ExpandNames(self):
6099 self.needed_locks = {}
6101 def Exec(self, feedback_fn):
6102 """Return cluster config.
6105 cluster = self.cfg.GetClusterInfo()
6108 # Filter just for enabled hypervisors
6109 for os_name, hv_dict in cluster.os_hvp.items():
6110 os_hvp[os_name] = {}
6111 for hv_name, hv_params in hv_dict.items():
6112 if hv_name in cluster.enabled_hypervisors:
6113 os_hvp[os_name][hv_name] = hv_params
6115 # Convert ip_family to ip_version
6116 primary_ip_version = constants.IP4_VERSION
6117 if cluster.primary_ip_family == netutils.IP6Address.family:
6118 primary_ip_version = constants.IP6_VERSION
6121 "software_version": constants.RELEASE_VERSION,
6122 "protocol_version": constants.PROTOCOL_VERSION,
6123 "config_version": constants.CONFIG_VERSION,
6124 "os_api_version": max(constants.OS_API_VERSIONS),
6125 "export_version": constants.EXPORT_VERSION,
6126 "architecture": runtime.GetArchInfo(),
6127 "name": cluster.cluster_name,
6128 "master": cluster.master_node,
6129 "default_hypervisor": cluster.primary_hypervisor,
6130 "enabled_hypervisors": cluster.enabled_hypervisors,
6131 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
6132 for hypervisor_name in cluster.enabled_hypervisors]),
6134 "beparams": cluster.beparams,
6135 "osparams": cluster.osparams,
6136 "ipolicy": cluster.ipolicy,
6137 "nicparams": cluster.nicparams,
6138 "ndparams": cluster.ndparams,
6139 "candidate_pool_size": cluster.candidate_pool_size,
6140 "master_netdev": cluster.master_netdev,
6141 "master_netmask": cluster.master_netmask,
6142 "use_external_mip_script": cluster.use_external_mip_script,
6143 "volume_group_name": cluster.volume_group_name,
6144 "drbd_usermode_helper": cluster.drbd_usermode_helper,
6145 "file_storage_dir": cluster.file_storage_dir,
6146 "shared_file_storage_dir": cluster.shared_file_storage_dir,
6147 "maintain_node_health": cluster.maintain_node_health,
6148 "ctime": cluster.ctime,
6149 "mtime": cluster.mtime,
6150 "uuid": cluster.uuid,
6151 "tags": list(cluster.GetTags()),
6152 "uid_pool": cluster.uid_pool,
6153 "default_iallocator": cluster.default_iallocator,
6154 "reserved_lvs": cluster.reserved_lvs,
6155 "primary_ip_version": primary_ip_version,
6156 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
6157 "hidden_os": cluster.hidden_os,
6158 "blacklisted_os": cluster.blacklisted_os,
6164 class LUClusterConfigQuery(NoHooksLU):
6165 """Return configuration values.
6170 def CheckArguments(self):
6171 self.cq = _ClusterQuery(None, self.op.output_fields, False)
6173 def ExpandNames(self):
6174 self.cq.ExpandNames(self)
6176 def DeclareLocks(self, level):
6177 self.cq.DeclareLocks(self, level)
6179 def Exec(self, feedback_fn):
6180 result = self.cq.OldStyleQuery(self)
6182 assert len(result) == 1
6187 class _ClusterQuery(_QueryBase):
6188 FIELDS = query.CLUSTER_FIELDS
6190 #: Do not sort (there is only one item)
6193 def ExpandNames(self, lu):
6194 lu.needed_locks = {}
6196 # The following variables interact with _QueryBase._GetNames
6197 self.wanted = locking.ALL_SET
6198 self.do_locking = self.use_locking
6201 raise errors.OpPrereqError("Can not use locking for cluster queries",
6204 def DeclareLocks(self, lu, level):
6207 def _GetQueryData(self, lu):
6208 """Computes the list of nodes and their attributes.
6211 # Locking is not used
6212 assert not (compat.any(lu.glm.is_owned(level)
6213 for level in locking.LEVELS
6214 if level != locking.LEVEL_CLUSTER) or
6215 self.do_locking or self.use_locking)
6217 if query.CQ_CONFIG in self.requested_data:
6218 cluster = lu.cfg.GetClusterInfo()
6220 cluster = NotImplemented
6222 if query.CQ_QUEUE_DRAINED in self.requested_data:
6223 drain_flag = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
6225 drain_flag = NotImplemented
6227 if query.CQ_WATCHER_PAUSE in self.requested_data:
6228 watcher_pause = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
6230 watcher_pause = NotImplemented
6232 return query.ClusterQueryData(cluster, drain_flag, watcher_pause)
6235 class LUInstanceActivateDisks(NoHooksLU):
6236 """Bring up an instance's disks.
6241 def ExpandNames(self):
6242 self._ExpandAndLockInstance()
6243 self.needed_locks[locking.LEVEL_NODE] = []
6244 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6246 def DeclareLocks(self, level):
6247 if level == locking.LEVEL_NODE:
6248 self._LockInstancesNodes()
6250 def CheckPrereq(self):
6251 """Check prerequisites.
6253 This checks that the instance is in the cluster.
6256 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6257 assert self.instance is not None, \
6258 "Cannot retrieve locked instance %s" % self.op.instance_name
6259 _CheckNodeOnline(self, self.instance.primary_node)
6261 def Exec(self, feedback_fn):
6262 """Activate the disks.
6265 disks_ok, disks_info = \
6266 _AssembleInstanceDisks(self, self.instance,
6267 ignore_size=self.op.ignore_size)
6269 raise errors.OpExecError("Cannot activate block devices")
6274 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
6276 """Prepare the block devices for an instance.
6278 This sets up the block devices on all nodes.
6280 @type lu: L{LogicalUnit}
6281 @param lu: the logical unit on whose behalf we execute
6282 @type instance: L{objects.Instance}
6283 @param instance: the instance for whose disks we assemble
6284 @type disks: list of L{objects.Disk} or None
6285 @param disks: which disks to assemble (or all, if None)
6286 @type ignore_secondaries: boolean
6287 @param ignore_secondaries: if true, errors on secondary nodes
6288 won't result in an error return from the function
6289 @type ignore_size: boolean
6290 @param ignore_size: if true, the current known size of the disk
6291 will not be used during the disk activation, useful for cases
6292 when the size is wrong
6293 @return: False if the operation failed, otherwise a list of
6294 (host, instance_visible_name, node_visible_name)
6295 with the mapping from node devices to instance devices
6300 iname = instance.name
6301 disks = _ExpandCheckDisks(instance, disks)
6303 # With the two passes mechanism we try to reduce the window of
6304 # opportunity for the race condition of switching DRBD to primary
6305 # before handshaking occured, but we do not eliminate it
6307 # The proper fix would be to wait (with some limits) until the
6308 # connection has been made and drbd transitions from WFConnection
6309 # into any other network-connected state (Connected, SyncTarget,
6312 # 1st pass, assemble on all nodes in secondary mode
6313 for idx, inst_disk in enumerate(disks):
6314 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
6316 node_disk = node_disk.Copy()
6317 node_disk.UnsetSize()
6318 lu.cfg.SetDiskID(node_disk, node)
6319 result = lu.rpc.call_blockdev_assemble(node, (node_disk, instance), iname,
6321 msg = result.fail_msg
6323 lu.proc.LogWarning("Could not prepare block device %s on node %s"
6324 " (is_primary=False, pass=1): %s",
6325 inst_disk.iv_name, node, msg)
6326 if not ignore_secondaries:
6329 # FIXME: race condition on drbd migration to primary
6331 # 2nd pass, do only the primary node
6332 for idx, inst_disk in enumerate(disks):
6335 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
6336 if node != instance.primary_node:
6339 node_disk = node_disk.Copy()
6340 node_disk.UnsetSize()
6341 lu.cfg.SetDiskID(node_disk, node)
6342 result = lu.rpc.call_blockdev_assemble(node, (node_disk, instance), iname,
6344 msg = result.fail_msg
6346 lu.proc.LogWarning("Could not prepare block device %s on node %s"
6347 " (is_primary=True, pass=2): %s",
6348 inst_disk.iv_name, node, msg)
6351 dev_path = result.payload
6353 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
6355 # leave the disks configured for the primary node
6356 # this is a workaround that would be fixed better by
6357 # improving the logical/physical id handling
6359 lu.cfg.SetDiskID(disk, instance.primary_node)
6361 return disks_ok, device_info
6364 def _StartInstanceDisks(lu, instance, force):
6365 """Start the disks of an instance.
6368 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
6369 ignore_secondaries=force)
6371 _ShutdownInstanceDisks(lu, instance)
6372 if force is not None and not force:
6373 lu.proc.LogWarning("", hint="If the message above refers to a"
6375 " you can retry the operation using '--force'.")
6376 raise errors.OpExecError("Disk consistency error")
6379 class LUInstanceDeactivateDisks(NoHooksLU):
6380 """Shutdown an instance's disks.
6385 def ExpandNames(self):
6386 self._ExpandAndLockInstance()
6387 self.needed_locks[locking.LEVEL_NODE] = []
6388 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6390 def DeclareLocks(self, level):
6391 if level == locking.LEVEL_NODE:
6392 self._LockInstancesNodes()
6394 def CheckPrereq(self):
6395 """Check prerequisites.
6397 This checks that the instance is in the cluster.
6400 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6401 assert self.instance is not None, \
6402 "Cannot retrieve locked instance %s" % self.op.instance_name
6404 def Exec(self, feedback_fn):
6405 """Deactivate the disks
6408 instance = self.instance
6410 _ShutdownInstanceDisks(self, instance)
6412 _SafeShutdownInstanceDisks(self, instance)
6415 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
6416 """Shutdown block devices of an instance.
6418 This function checks if an instance is running, before calling
6419 _ShutdownInstanceDisks.
6422 _CheckInstanceState(lu, instance, INSTANCE_DOWN, msg="cannot shutdown disks")
6423 _ShutdownInstanceDisks(lu, instance, disks=disks)
6426 def _ExpandCheckDisks(instance, disks):
6427 """Return the instance disks selected by the disks list
6429 @type disks: list of L{objects.Disk} or None
6430 @param disks: selected disks
6431 @rtype: list of L{objects.Disk}
6432 @return: selected instance disks to act on
6436 return instance.disks
6438 if not set(disks).issubset(instance.disks):
6439 raise errors.ProgrammerError("Can only act on disks belonging to the"
6444 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
6445 """Shutdown block devices of an instance.
6447 This does the shutdown on all nodes of the instance.
6449 If the ignore_primary is false, errors on the primary node are
6454 disks = _ExpandCheckDisks(instance, disks)
6457 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
6458 lu.cfg.SetDiskID(top_disk, node)
6459 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
6460 msg = result.fail_msg
6462 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
6463 disk.iv_name, node, msg)
6464 if ((node == instance.primary_node and not ignore_primary) or
6465 (node != instance.primary_node and not result.offline)):
6470 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
6471 """Checks if a node has enough free memory.
6473 This function check if a given node has the needed amount of free
6474 memory. In case the node has less memory or we cannot get the
6475 information from the node, this function raise an OpPrereqError
6478 @type lu: C{LogicalUnit}
6479 @param lu: a logical unit from which we get configuration data
6481 @param node: the node to check
6482 @type reason: C{str}
6483 @param reason: string to use in the error message
6484 @type requested: C{int}
6485 @param requested: the amount of memory in MiB to check for
6486 @type hypervisor_name: C{str}
6487 @param hypervisor_name: the hypervisor to ask for memory stats
6489 @return: node current free memory
6490 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
6491 we cannot check the node
6494 nodeinfo = lu.rpc.call_node_info([node], None, [hypervisor_name])
6495 nodeinfo[node].Raise("Can't get data from node %s" % node,
6496 prereq=True, ecode=errors.ECODE_ENVIRON)
6497 (_, _, (hv_info, )) = nodeinfo[node].payload
6499 free_mem = hv_info.get("memory_free", None)
6500 if not isinstance(free_mem, int):
6501 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
6502 " was '%s'" % (node, free_mem),
6503 errors.ECODE_ENVIRON)
6504 if requested > free_mem:
6505 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
6506 " needed %s MiB, available %s MiB" %
6507 (node, reason, requested, free_mem),
6512 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
6513 """Checks if nodes have enough free disk space in the all VGs.
6515 This function check if all given nodes have the needed amount of
6516 free disk. In case any node has less disk or we cannot get the
6517 information from the node, this function raise an OpPrereqError
6520 @type lu: C{LogicalUnit}
6521 @param lu: a logical unit from which we get configuration data
6522 @type nodenames: C{list}
6523 @param nodenames: the list of node names to check
6524 @type req_sizes: C{dict}
6525 @param req_sizes: the hash of vg and corresponding amount of disk in
6527 @raise errors.OpPrereqError: if the node doesn't have enough disk,
6528 or we cannot check the node
6531 for vg, req_size in req_sizes.items():
6532 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
6535 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
6536 """Checks if nodes have enough free disk space in the specified VG.
6538 This function check if all given nodes have the needed amount of
6539 free disk. In case any node has less disk or we cannot get the
6540 information from the node, this function raise an OpPrereqError
6543 @type lu: C{LogicalUnit}
6544 @param lu: a logical unit from which we get configuration data
6545 @type nodenames: C{list}
6546 @param nodenames: the list of node names to check
6548 @param vg: the volume group to check
6549 @type requested: C{int}
6550 @param requested: the amount of disk in MiB to check for
6551 @raise errors.OpPrereqError: if the node doesn't have enough disk,
6552 or we cannot check the node
6555 nodeinfo = lu.rpc.call_node_info(nodenames, [vg], None)
6556 for node in nodenames:
6557 info = nodeinfo[node]
6558 info.Raise("Cannot get current information from node %s" % node,
6559 prereq=True, ecode=errors.ECODE_ENVIRON)
6560 (_, (vg_info, ), _) = info.payload
6561 vg_free = vg_info.get("vg_free", None)
6562 if not isinstance(vg_free, int):
6563 raise errors.OpPrereqError("Can't compute free disk space on node"
6564 " %s for vg %s, result was '%s'" %
6565 (node, vg, vg_free), errors.ECODE_ENVIRON)
6566 if requested > vg_free:
6567 raise errors.OpPrereqError("Not enough disk space on target node %s"
6568 " vg %s: required %d MiB, available %d MiB" %
6569 (node, vg, requested, vg_free),
6573 def _CheckNodesPhysicalCPUs(lu, nodenames, requested, hypervisor_name):
6574 """Checks if nodes have enough physical CPUs
6576 This function checks if all given nodes have the needed number of
6577 physical CPUs. In case any node has less CPUs or we cannot get the
6578 information from the node, this function raises an OpPrereqError
6581 @type lu: C{LogicalUnit}
6582 @param lu: a logical unit from which we get configuration data
6583 @type nodenames: C{list}
6584 @param nodenames: the list of node names to check
6585 @type requested: C{int}
6586 @param requested: the minimum acceptable number of physical CPUs
6587 @raise errors.OpPrereqError: if the node doesn't have enough CPUs,
6588 or we cannot check the node
6591 nodeinfo = lu.rpc.call_node_info(nodenames, None, [hypervisor_name])
6592 for node in nodenames:
6593 info = nodeinfo[node]
6594 info.Raise("Cannot get current information from node %s" % node,
6595 prereq=True, ecode=errors.ECODE_ENVIRON)
6596 (_, _, (hv_info, )) = info.payload
6597 num_cpus = hv_info.get("cpu_total", None)
6598 if not isinstance(num_cpus, int):
6599 raise errors.OpPrereqError("Can't compute the number of physical CPUs"
6600 " on node %s, result was '%s'" %
6601 (node, num_cpus), errors.ECODE_ENVIRON)
6602 if requested > num_cpus:
6603 raise errors.OpPrereqError("Node %s has %s physical CPUs, but %s are "
6604 "required" % (node, num_cpus, requested),
6608 class LUInstanceStartup(LogicalUnit):
6609 """Starts an instance.
6612 HPATH = "instance-start"
6613 HTYPE = constants.HTYPE_INSTANCE
6616 def CheckArguments(self):
6618 if self.op.beparams:
6619 # fill the beparams dict
6620 objects.UpgradeBeParams(self.op.beparams)
6621 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6623 def ExpandNames(self):
6624 self._ExpandAndLockInstance()
6625 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
6627 def DeclareLocks(self, level):
6628 if level == locking.LEVEL_NODE_RES:
6629 self._LockInstancesNodes(primary_only=True, level=locking.LEVEL_NODE_RES)
6631 def BuildHooksEnv(self):
6634 This runs on master, primary and secondary nodes of the instance.
6638 "FORCE": self.op.force,
6641 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6645 def BuildHooksNodes(self):
6646 """Build hooks nodes.
6649 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6652 def CheckPrereq(self):
6653 """Check prerequisites.
6655 This checks that the instance is in the cluster.
6658 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6659 assert self.instance is not None, \
6660 "Cannot retrieve locked instance %s" % self.op.instance_name
6663 if self.op.hvparams:
6664 # check hypervisor parameter syntax (locally)
6665 cluster = self.cfg.GetClusterInfo()
6666 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6667 filled_hvp = cluster.FillHV(instance)
6668 filled_hvp.update(self.op.hvparams)
6669 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
6670 hv_type.CheckParameterSyntax(filled_hvp)
6671 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
6673 _CheckInstanceState(self, instance, INSTANCE_ONLINE)
6675 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
6677 if self.primary_offline and self.op.ignore_offline_nodes:
6678 self.proc.LogWarning("Ignoring offline primary node")
6680 if self.op.hvparams or self.op.beparams:
6681 self.proc.LogWarning("Overridden parameters are ignored")
6683 _CheckNodeOnline(self, instance.primary_node)
6685 bep = self.cfg.GetClusterInfo().FillBE(instance)
6686 bep.update(self.op.beparams)
6688 # check bridges existence
6689 _CheckInstanceBridgesExist(self, instance)
6691 remote_info = self.rpc.call_instance_info(instance.primary_node,
6693 instance.hypervisor)
6694 remote_info.Raise("Error checking node %s" % instance.primary_node,
6695 prereq=True, ecode=errors.ECODE_ENVIRON)
6696 if not remote_info.payload: # not running already
6697 _CheckNodeFreeMemory(self, instance.primary_node,
6698 "starting instance %s" % instance.name,
6699 bep[constants.BE_MINMEM], instance.hypervisor)
6701 def Exec(self, feedback_fn):
6702 """Start the instance.
6705 instance = self.instance
6706 force = self.op.force
6708 if not self.op.no_remember:
6709 self.cfg.MarkInstanceUp(instance.name)
6711 if self.primary_offline:
6712 assert self.op.ignore_offline_nodes
6713 self.proc.LogInfo("Primary node offline, marked instance as started")
6715 node_current = instance.primary_node
6717 _StartInstanceDisks(self, instance, force)
6720 self.rpc.call_instance_start(node_current,
6721 (instance, self.op.hvparams,
6723 self.op.startup_paused)
6724 msg = result.fail_msg
6726 _ShutdownInstanceDisks(self, instance)
6727 raise errors.OpExecError("Could not start instance: %s" % msg)
6730 class LUInstanceReboot(LogicalUnit):
6731 """Reboot an instance.
6734 HPATH = "instance-reboot"
6735 HTYPE = constants.HTYPE_INSTANCE
6738 def ExpandNames(self):
6739 self._ExpandAndLockInstance()
6741 def BuildHooksEnv(self):
6744 This runs on master, primary and secondary nodes of the instance.
6748 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
6749 "REBOOT_TYPE": self.op.reboot_type,
6750 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6753 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6757 def BuildHooksNodes(self):
6758 """Build hooks nodes.
6761 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6764 def CheckPrereq(self):
6765 """Check prerequisites.
6767 This checks that the instance is in the cluster.
6770 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6771 assert self.instance is not None, \
6772 "Cannot retrieve locked instance %s" % self.op.instance_name
6773 _CheckInstanceState(self, instance, INSTANCE_ONLINE)
6774 _CheckNodeOnline(self, instance.primary_node)
6776 # check bridges existence
6777 _CheckInstanceBridgesExist(self, instance)
6779 def Exec(self, feedback_fn):
6780 """Reboot the instance.
6783 instance = self.instance
6784 ignore_secondaries = self.op.ignore_secondaries
6785 reboot_type = self.op.reboot_type
6787 remote_info = self.rpc.call_instance_info(instance.primary_node,
6789 instance.hypervisor)
6790 remote_info.Raise("Error checking node %s" % instance.primary_node)
6791 instance_running = bool(remote_info.payload)
6793 node_current = instance.primary_node
6795 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
6796 constants.INSTANCE_REBOOT_HARD]:
6797 for disk in instance.disks:
6798 self.cfg.SetDiskID(disk, node_current)
6799 result = self.rpc.call_instance_reboot(node_current, instance,
6801 self.op.shutdown_timeout)
6802 result.Raise("Could not reboot instance")
6804 if instance_running:
6805 result = self.rpc.call_instance_shutdown(node_current, instance,
6806 self.op.shutdown_timeout)
6807 result.Raise("Could not shutdown instance for full reboot")
6808 _ShutdownInstanceDisks(self, instance)
6810 self.LogInfo("Instance %s was already stopped, starting now",
6812 _StartInstanceDisks(self, instance, ignore_secondaries)
6813 result = self.rpc.call_instance_start(node_current,
6814 (instance, None, None), False)
6815 msg = result.fail_msg
6817 _ShutdownInstanceDisks(self, instance)
6818 raise errors.OpExecError("Could not start instance for"
6819 " full reboot: %s" % msg)
6821 self.cfg.MarkInstanceUp(instance.name)
6824 class LUInstanceShutdown(LogicalUnit):
6825 """Shutdown an instance.
6828 HPATH = "instance-stop"
6829 HTYPE = constants.HTYPE_INSTANCE
6832 def ExpandNames(self):
6833 self._ExpandAndLockInstance()
6835 def BuildHooksEnv(self):
6838 This runs on master, primary and secondary nodes of the instance.
6841 env = _BuildInstanceHookEnvByObject(self, self.instance)
6842 env["TIMEOUT"] = self.op.timeout
6845 def BuildHooksNodes(self):
6846 """Build hooks nodes.
6849 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6852 def CheckPrereq(self):
6853 """Check prerequisites.
6855 This checks that the instance is in the cluster.
6858 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6859 assert self.instance is not None, \
6860 "Cannot retrieve locked instance %s" % self.op.instance_name
6862 _CheckInstanceState(self, self.instance, INSTANCE_ONLINE)
6864 self.primary_offline = \
6865 self.cfg.GetNodeInfo(self.instance.primary_node).offline
6867 if self.primary_offline and self.op.ignore_offline_nodes:
6868 self.proc.LogWarning("Ignoring offline primary node")
6870 _CheckNodeOnline(self, self.instance.primary_node)
6872 def Exec(self, feedback_fn):
6873 """Shutdown the instance.
6876 instance = self.instance
6877 node_current = instance.primary_node
6878 timeout = self.op.timeout
6880 if not self.op.no_remember:
6881 self.cfg.MarkInstanceDown(instance.name)
6883 if self.primary_offline:
6884 assert self.op.ignore_offline_nodes
6885 self.proc.LogInfo("Primary node offline, marked instance as stopped")
6887 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
6888 msg = result.fail_msg
6890 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
6892 _ShutdownInstanceDisks(self, instance)
6895 class LUInstanceReinstall(LogicalUnit):
6896 """Reinstall an instance.
6899 HPATH = "instance-reinstall"
6900 HTYPE = constants.HTYPE_INSTANCE
6903 def ExpandNames(self):
6904 self._ExpandAndLockInstance()
6906 def BuildHooksEnv(self):
6909 This runs on master, primary and secondary nodes of the instance.
6912 return _BuildInstanceHookEnvByObject(self, self.instance)
6914 def BuildHooksNodes(self):
6915 """Build hooks nodes.
6918 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6921 def CheckPrereq(self):
6922 """Check prerequisites.
6924 This checks that the instance is in the cluster and is not running.
6927 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6928 assert instance is not None, \
6929 "Cannot retrieve locked instance %s" % self.op.instance_name
6930 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
6931 " offline, cannot reinstall")
6932 for node in instance.secondary_nodes:
6933 _CheckNodeOnline(self, node, "Instance secondary node offline,"
6934 " cannot reinstall")
6936 if instance.disk_template == constants.DT_DISKLESS:
6937 raise errors.OpPrereqError("Instance '%s' has no disks" %
6938 self.op.instance_name,
6940 _CheckInstanceState(self, instance, INSTANCE_DOWN, msg="cannot reinstall")
6942 if self.op.os_type is not None:
6944 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
6945 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
6946 instance_os = self.op.os_type
6948 instance_os = instance.os
6950 nodelist = list(instance.all_nodes)
6952 if self.op.osparams:
6953 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
6954 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
6955 self.os_inst = i_osdict # the new dict (without defaults)
6959 self.instance = instance
6961 def Exec(self, feedback_fn):
6962 """Reinstall the instance.
6965 inst = self.instance
6967 if self.op.os_type is not None:
6968 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
6969 inst.os = self.op.os_type
6970 # Write to configuration
6971 self.cfg.Update(inst, feedback_fn)
6973 _StartInstanceDisks(self, inst, None)
6975 feedback_fn("Running the instance OS create scripts...")
6976 # FIXME: pass debug option from opcode to backend
6977 result = self.rpc.call_instance_os_add(inst.primary_node,
6978 (inst, self.os_inst), True,
6979 self.op.debug_level)
6980 result.Raise("Could not install OS for instance %s on node %s" %
6981 (inst.name, inst.primary_node))
6983 _ShutdownInstanceDisks(self, inst)
6986 class LUInstanceRecreateDisks(LogicalUnit):
6987 """Recreate an instance's missing disks.
6990 HPATH = "instance-recreate-disks"
6991 HTYPE = constants.HTYPE_INSTANCE
6994 _MODIFYABLE = frozenset([
6995 constants.IDISK_SIZE,
6996 constants.IDISK_MODE,
6999 # New or changed disk parameters may have different semantics
7000 assert constants.IDISK_PARAMS == (_MODIFYABLE | frozenset([
7001 constants.IDISK_ADOPT,
7003 # TODO: Implement support changing VG while recreating
7005 constants.IDISK_METAVG,
7008 def CheckArguments(self):
7009 if self.op.disks and ht.TPositiveInt(self.op.disks[0]):
7010 # Normalize and convert deprecated list of disk indices
7011 self.op.disks = [(idx, {}) for idx in sorted(frozenset(self.op.disks))]
7013 duplicates = utils.FindDuplicates(map(compat.fst, self.op.disks))
7015 raise errors.OpPrereqError("Some disks have been specified more than"
7016 " once: %s" % utils.CommaJoin(duplicates),
7019 for (idx, params) in self.op.disks:
7020 utils.ForceDictType(params, constants.IDISK_PARAMS_TYPES)
7021 unsupported = frozenset(params.keys()) - self._MODIFYABLE
7023 raise errors.OpPrereqError("Parameters for disk %s try to change"
7024 " unmodifyable parameter(s): %s" %
7025 (idx, utils.CommaJoin(unsupported)),
7028 def ExpandNames(self):
7029 self._ExpandAndLockInstance()
7030 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7032 self.op.nodes = [_ExpandNodeName(self.cfg, n) for n in self.op.nodes]
7033 self.needed_locks[locking.LEVEL_NODE] = list(self.op.nodes)
7035 self.needed_locks[locking.LEVEL_NODE] = []
7036 self.needed_locks[locking.LEVEL_NODE_RES] = []
7038 def DeclareLocks(self, level):
7039 if level == locking.LEVEL_NODE:
7040 # if we replace the nodes, we only need to lock the old primary,
7041 # otherwise we need to lock all nodes for disk re-creation
7042 primary_only = bool(self.op.nodes)
7043 self._LockInstancesNodes(primary_only=primary_only)
7044 elif level == locking.LEVEL_NODE_RES:
7046 self.needed_locks[locking.LEVEL_NODE_RES] = \
7047 self.needed_locks[locking.LEVEL_NODE][:]
7049 def BuildHooksEnv(self):
7052 This runs on master, primary and secondary nodes of the instance.
7055 return _BuildInstanceHookEnvByObject(self, self.instance)
7057 def BuildHooksNodes(self):
7058 """Build hooks nodes.
7061 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7064 def CheckPrereq(self):
7065 """Check prerequisites.
7067 This checks that the instance is in the cluster and is not running.
7070 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7071 assert instance is not None, \
7072 "Cannot retrieve locked instance %s" % self.op.instance_name
7074 if len(self.op.nodes) != len(instance.all_nodes):
7075 raise errors.OpPrereqError("Instance %s currently has %d nodes, but"
7076 " %d replacement nodes were specified" %
7077 (instance.name, len(instance.all_nodes),
7078 len(self.op.nodes)),
7080 assert instance.disk_template != constants.DT_DRBD8 or \
7081 len(self.op.nodes) == 2
7082 assert instance.disk_template != constants.DT_PLAIN or \
7083 len(self.op.nodes) == 1
7084 primary_node = self.op.nodes[0]
7086 primary_node = instance.primary_node
7087 _CheckNodeOnline(self, primary_node)
7089 if instance.disk_template == constants.DT_DISKLESS:
7090 raise errors.OpPrereqError("Instance '%s' has no disks" %
7091 self.op.instance_name, errors.ECODE_INVAL)
7093 # if we replace nodes *and* the old primary is offline, we don't
7095 assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE)
7096 assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE_RES)
7097 old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
7098 if not (self.op.nodes and old_pnode.offline):
7099 _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
7100 msg="cannot recreate disks")
7103 self.disks = dict(self.op.disks)
7105 self.disks = dict((idx, {}) for idx in range(len(instance.disks)))
7107 maxidx = max(self.disks.keys())
7108 if maxidx >= len(instance.disks):
7109 raise errors.OpPrereqError("Invalid disk index '%s'" % maxidx,
7112 if (self.op.nodes and
7113 sorted(self.disks.keys()) != range(len(instance.disks))):
7114 raise errors.OpPrereqError("Can't recreate disks partially and"
7115 " change the nodes at the same time",
7118 self.instance = instance
7120 def Exec(self, feedback_fn):
7121 """Recreate the disks.
7124 instance = self.instance
7126 assert (self.owned_locks(locking.LEVEL_NODE) ==
7127 self.owned_locks(locking.LEVEL_NODE_RES))
7130 mods = [] # keeps track of needed changes
7132 for idx, disk in enumerate(instance.disks):
7134 changes = self.disks[idx]
7136 # Disk should not be recreated
7140 # update secondaries for disks, if needed
7141 if self.op.nodes and disk.dev_type == constants.LD_DRBD8:
7142 # need to update the nodes and minors
7143 assert len(self.op.nodes) == 2
7144 assert len(disk.logical_id) == 6 # otherwise disk internals
7146 (_, _, old_port, _, _, old_secret) = disk.logical_id
7147 new_minors = self.cfg.AllocateDRBDMinor(self.op.nodes, instance.name)
7148 new_id = (self.op.nodes[0], self.op.nodes[1], old_port,
7149 new_minors[0], new_minors[1], old_secret)
7150 assert len(disk.logical_id) == len(new_id)
7154 mods.append((idx, new_id, changes))
7156 # now that we have passed all asserts above, we can apply the mods
7157 # in a single run (to avoid partial changes)
7158 for idx, new_id, changes in mods:
7159 disk = instance.disks[idx]
7160 if new_id is not None:
7161 assert disk.dev_type == constants.LD_DRBD8
7162 disk.logical_id = new_id
7164 disk.Update(size=changes.get(constants.IDISK_SIZE, None),
7165 mode=changes.get(constants.IDISK_MODE, None))
7167 # change primary node, if needed
7169 instance.primary_node = self.op.nodes[0]
7170 self.LogWarning("Changing the instance's nodes, you will have to"
7171 " remove any disks left on the older nodes manually")
7174 self.cfg.Update(instance, feedback_fn)
7176 _CreateDisks(self, instance, to_skip=to_skip)
7179 class LUInstanceRename(LogicalUnit):
7180 """Rename an instance.
7183 HPATH = "instance-rename"
7184 HTYPE = constants.HTYPE_INSTANCE
7186 def CheckArguments(self):
7190 if self.op.ip_check and not self.op.name_check:
7191 # TODO: make the ip check more flexible and not depend on the name check
7192 raise errors.OpPrereqError("IP address check requires a name check",
7195 def BuildHooksEnv(self):
7198 This runs on master, primary and secondary nodes of the instance.
7201 env = _BuildInstanceHookEnvByObject(self, self.instance)
7202 env["INSTANCE_NEW_NAME"] = self.op.new_name
7205 def BuildHooksNodes(self):
7206 """Build hooks nodes.
7209 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7212 def CheckPrereq(self):
7213 """Check prerequisites.
7215 This checks that the instance is in the cluster and is not running.
7218 self.op.instance_name = _ExpandInstanceName(self.cfg,
7219 self.op.instance_name)
7220 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7221 assert instance is not None
7222 _CheckNodeOnline(self, instance.primary_node)
7223 _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
7224 msg="cannot rename")
7225 self.instance = instance
7227 new_name = self.op.new_name
7228 if self.op.name_check:
7229 hostname = netutils.GetHostname(name=new_name)
7230 if hostname.name != new_name:
7231 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
7233 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
7234 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
7235 " same as given hostname '%s'") %
7236 (hostname.name, self.op.new_name),
7238 new_name = self.op.new_name = hostname.name
7239 if (self.op.ip_check and
7240 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
7241 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7242 (hostname.ip, new_name),
7243 errors.ECODE_NOTUNIQUE)
7245 instance_list = self.cfg.GetInstanceList()
7246 if new_name in instance_list and new_name != instance.name:
7247 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7248 new_name, errors.ECODE_EXISTS)
7250 def Exec(self, feedback_fn):
7251 """Rename the instance.
7254 inst = self.instance
7255 old_name = inst.name
7257 rename_file_storage = False
7258 if (inst.disk_template in constants.DTS_FILEBASED and
7259 self.op.new_name != inst.name):
7260 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
7261 rename_file_storage = True
7263 self.cfg.RenameInstance(inst.name, self.op.new_name)
7264 # Change the instance lock. This is definitely safe while we hold the BGL.
7265 # Otherwise the new lock would have to be added in acquired mode.
7267 self.glm.remove(locking.LEVEL_INSTANCE, old_name)
7268 self.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
7270 # re-read the instance from the configuration after rename
7271 inst = self.cfg.GetInstanceInfo(self.op.new_name)
7273 if rename_file_storage:
7274 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
7275 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
7276 old_file_storage_dir,
7277 new_file_storage_dir)
7278 result.Raise("Could not rename on node %s directory '%s' to '%s'"
7279 " (but the instance has been renamed in Ganeti)" %
7280 (inst.primary_node, old_file_storage_dir,
7281 new_file_storage_dir))
7283 _StartInstanceDisks(self, inst, None)
7285 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
7286 old_name, self.op.debug_level)
7287 msg = result.fail_msg
7289 msg = ("Could not run OS rename script for instance %s on node %s"
7290 " (but the instance has been renamed in Ganeti): %s" %
7291 (inst.name, inst.primary_node, msg))
7292 self.proc.LogWarning(msg)
7294 _ShutdownInstanceDisks(self, inst)
7299 class LUInstanceRemove(LogicalUnit):
7300 """Remove an instance.
7303 HPATH = "instance-remove"
7304 HTYPE = constants.HTYPE_INSTANCE
7307 def ExpandNames(self):
7308 self._ExpandAndLockInstance()
7309 self.needed_locks[locking.LEVEL_NODE] = []
7310 self.needed_locks[locking.LEVEL_NODE_RES] = []
7311 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7313 def DeclareLocks(self, level):
7314 if level == locking.LEVEL_NODE:
7315 self._LockInstancesNodes()
7316 elif level == locking.LEVEL_NODE_RES:
7318 self.needed_locks[locking.LEVEL_NODE_RES] = \
7319 self.needed_locks[locking.LEVEL_NODE][:]
7321 def BuildHooksEnv(self):
7324 This runs on master, primary and secondary nodes of the instance.
7327 env = _BuildInstanceHookEnvByObject(self, self.instance)
7328 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
7331 def BuildHooksNodes(self):
7332 """Build hooks nodes.
7335 nl = [self.cfg.GetMasterNode()]
7336 nl_post = list(self.instance.all_nodes) + nl
7337 return (nl, nl_post)
7339 def CheckPrereq(self):
7340 """Check prerequisites.
7342 This checks that the instance is in the cluster.
7345 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7346 assert self.instance is not None, \
7347 "Cannot retrieve locked instance %s" % self.op.instance_name
7349 def Exec(self, feedback_fn):
7350 """Remove the instance.
7353 instance = self.instance
7354 logging.info("Shutting down instance %s on node %s",
7355 instance.name, instance.primary_node)
7357 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
7358 self.op.shutdown_timeout)
7359 msg = result.fail_msg
7361 if self.op.ignore_failures:
7362 feedback_fn("Warning: can't shutdown instance: %s" % msg)
7364 raise errors.OpExecError("Could not shutdown instance %s on"
7366 (instance.name, instance.primary_node, msg))
7368 assert (self.owned_locks(locking.LEVEL_NODE) ==
7369 self.owned_locks(locking.LEVEL_NODE_RES))
7370 assert not (set(instance.all_nodes) -
7371 self.owned_locks(locking.LEVEL_NODE)), \
7372 "Not owning correct locks"
7374 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
7377 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
7378 """Utility function to remove an instance.
7381 logging.info("Removing block devices for instance %s", instance.name)
7383 if not _RemoveDisks(lu, instance, ignore_failures=ignore_failures):
7384 if not ignore_failures:
7385 raise errors.OpExecError("Can't remove instance's disks")
7386 feedback_fn("Warning: can't remove instance's disks")
7388 logging.info("Removing instance %s out of cluster config", instance.name)
7390 lu.cfg.RemoveInstance(instance.name)
7392 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
7393 "Instance lock removal conflict"
7395 # Remove lock for the instance
7396 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
7399 class LUInstanceQuery(NoHooksLU):
7400 """Logical unit for querying instances.
7403 # pylint: disable=W0142
7406 def CheckArguments(self):
7407 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
7408 self.op.output_fields, self.op.use_locking)
7410 def ExpandNames(self):
7411 self.iq.ExpandNames(self)
7413 def DeclareLocks(self, level):
7414 self.iq.DeclareLocks(self, level)
7416 def Exec(self, feedback_fn):
7417 return self.iq.OldStyleQuery(self)
7420 class LUInstanceFailover(LogicalUnit):
7421 """Failover an instance.
7424 HPATH = "instance-failover"
7425 HTYPE = constants.HTYPE_INSTANCE
7428 def CheckArguments(self):
7429 """Check the arguments.
7432 self.iallocator = getattr(self.op, "iallocator", None)
7433 self.target_node = getattr(self.op, "target_node", None)
7435 def ExpandNames(self):
7436 self._ExpandAndLockInstance()
7438 if self.op.target_node is not None:
7439 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7441 self.needed_locks[locking.LEVEL_NODE] = []
7442 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7444 self.needed_locks[locking.LEVEL_NODE_RES] = []
7445 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
7447 ignore_consistency = self.op.ignore_consistency
7448 shutdown_timeout = self.op.shutdown_timeout
7449 self._migrater = TLMigrateInstance(self, self.op.instance_name,
7452 ignore_consistency=ignore_consistency,
7453 shutdown_timeout=shutdown_timeout,
7454 ignore_ipolicy=self.op.ignore_ipolicy)
7455 self.tasklets = [self._migrater]
7457 def DeclareLocks(self, level):
7458 if level == locking.LEVEL_NODE:
7459 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
7460 if instance.disk_template in constants.DTS_EXT_MIRROR:
7461 if self.op.target_node is None:
7462 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7464 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
7465 self.op.target_node]
7466 del self.recalculate_locks[locking.LEVEL_NODE]
7468 self._LockInstancesNodes()
7469 elif level == locking.LEVEL_NODE_RES:
7471 self.needed_locks[locking.LEVEL_NODE_RES] = \
7472 self.needed_locks[locking.LEVEL_NODE][:]
7474 def BuildHooksEnv(self):
7477 This runs on master, primary and secondary nodes of the instance.
7480 instance = self._migrater.instance
7481 source_node = instance.primary_node
7482 target_node = self.op.target_node
7484 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
7485 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
7486 "OLD_PRIMARY": source_node,
7487 "NEW_PRIMARY": target_node,
7490 if instance.disk_template in constants.DTS_INT_MIRROR:
7491 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
7492 env["NEW_SECONDARY"] = source_node
7494 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
7496 env.update(_BuildInstanceHookEnvByObject(self, instance))
7500 def BuildHooksNodes(self):
7501 """Build hooks nodes.
7504 instance = self._migrater.instance
7505 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
7506 return (nl, nl + [instance.primary_node])
7509 class LUInstanceMigrate(LogicalUnit):
7510 """Migrate an instance.
7512 This is migration without shutting down, compared to the failover,
7513 which is done with shutdown.
7516 HPATH = "instance-migrate"
7517 HTYPE = constants.HTYPE_INSTANCE
7520 def ExpandNames(self):
7521 self._ExpandAndLockInstance()
7523 if self.op.target_node is not None:
7524 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7526 self.needed_locks[locking.LEVEL_NODE] = []
7527 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7529 self.needed_locks[locking.LEVEL_NODE] = []
7530 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7533 TLMigrateInstance(self, self.op.instance_name,
7534 cleanup=self.op.cleanup,
7536 fallback=self.op.allow_failover,
7537 allow_runtime_changes=self.op.allow_runtime_changes,
7538 ignore_ipolicy=self.op.ignore_ipolicy)
7539 self.tasklets = [self._migrater]
7541 def DeclareLocks(self, level):
7542 if level == locking.LEVEL_NODE:
7543 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
7544 if instance.disk_template in constants.DTS_EXT_MIRROR:
7545 if self.op.target_node is None:
7546 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7548 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
7549 self.op.target_node]
7550 del self.recalculate_locks[locking.LEVEL_NODE]
7552 self._LockInstancesNodes()
7553 elif level == locking.LEVEL_NODE_RES:
7555 self.needed_locks[locking.LEVEL_NODE_RES] = \
7556 self.needed_locks[locking.LEVEL_NODE][:]
7558 def BuildHooksEnv(self):
7561 This runs on master, primary and secondary nodes of the instance.
7564 instance = self._migrater.instance
7565 source_node = instance.primary_node
7566 target_node = self.op.target_node
7567 env = _BuildInstanceHookEnvByObject(self, instance)
7569 "MIGRATE_LIVE": self._migrater.live,
7570 "MIGRATE_CLEANUP": self.op.cleanup,
7571 "OLD_PRIMARY": source_node,
7572 "NEW_PRIMARY": target_node,
7573 "ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes,
7576 if instance.disk_template in constants.DTS_INT_MIRROR:
7577 env["OLD_SECONDARY"] = target_node
7578 env["NEW_SECONDARY"] = source_node
7580 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
7584 def BuildHooksNodes(self):
7585 """Build hooks nodes.
7588 instance = self._migrater.instance
7589 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
7590 return (nl, nl + [instance.primary_node])
7593 class LUInstanceMove(LogicalUnit):
7594 """Move an instance by data-copying.
7597 HPATH = "instance-move"
7598 HTYPE = constants.HTYPE_INSTANCE
7601 def ExpandNames(self):
7602 self._ExpandAndLockInstance()
7603 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
7604 self.op.target_node = target_node
7605 self.needed_locks[locking.LEVEL_NODE] = [target_node]
7606 self.needed_locks[locking.LEVEL_NODE_RES] = []
7607 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7609 def DeclareLocks(self, level):
7610 if level == locking.LEVEL_NODE:
7611 self._LockInstancesNodes(primary_only=True)
7612 elif level == locking.LEVEL_NODE_RES:
7614 self.needed_locks[locking.LEVEL_NODE_RES] = \
7615 self.needed_locks[locking.LEVEL_NODE][:]
7617 def BuildHooksEnv(self):
7620 This runs on master, primary and secondary nodes of the instance.
7624 "TARGET_NODE": self.op.target_node,
7625 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
7627 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7630 def BuildHooksNodes(self):
7631 """Build hooks nodes.
7635 self.cfg.GetMasterNode(),
7636 self.instance.primary_node,
7637 self.op.target_node,
7641 def CheckPrereq(self):
7642 """Check prerequisites.
7644 This checks that the instance is in the cluster.
7647 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7648 assert self.instance is not None, \
7649 "Cannot retrieve locked instance %s" % self.op.instance_name
7651 node = self.cfg.GetNodeInfo(self.op.target_node)
7652 assert node is not None, \
7653 "Cannot retrieve locked node %s" % self.op.target_node
7655 self.target_node = target_node = node.name
7657 if target_node == instance.primary_node:
7658 raise errors.OpPrereqError("Instance %s is already on the node %s" %
7659 (instance.name, target_node),
7662 bep = self.cfg.GetClusterInfo().FillBE(instance)
7664 for idx, dsk in enumerate(instance.disks):
7665 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
7666 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
7667 " cannot copy" % idx, errors.ECODE_STATE)
7669 _CheckNodeOnline(self, target_node)
7670 _CheckNodeNotDrained(self, target_node)
7671 _CheckNodeVmCapable(self, target_node)
7672 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(),
7673 self.cfg.GetNodeGroup(node.group))
7674 _CheckTargetNodeIPolicy(self, ipolicy, instance, node,
7675 ignore=self.op.ignore_ipolicy)
7677 if instance.admin_state == constants.ADMINST_UP:
7678 # check memory requirements on the secondary node
7679 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
7680 instance.name, bep[constants.BE_MAXMEM],
7681 instance.hypervisor)
7683 self.LogInfo("Not checking memory on the secondary node as"
7684 " instance will not be started")
7686 # check bridge existance
7687 _CheckInstanceBridgesExist(self, instance, node=target_node)
7689 def Exec(self, feedback_fn):
7690 """Move an instance.
7692 The move is done by shutting it down on its present node, copying
7693 the data over (slow) and starting it on the new node.
7696 instance = self.instance
7698 source_node = instance.primary_node
7699 target_node = self.target_node
7701 self.LogInfo("Shutting down instance %s on source node %s",
7702 instance.name, source_node)
7704 assert (self.owned_locks(locking.LEVEL_NODE) ==
7705 self.owned_locks(locking.LEVEL_NODE_RES))
7707 result = self.rpc.call_instance_shutdown(source_node, instance,
7708 self.op.shutdown_timeout)
7709 msg = result.fail_msg
7711 if self.op.ignore_consistency:
7712 self.proc.LogWarning("Could not shutdown instance %s on node %s."
7713 " Proceeding anyway. Please make sure node"
7714 " %s is down. Error details: %s",
7715 instance.name, source_node, source_node, msg)
7717 raise errors.OpExecError("Could not shutdown instance %s on"
7719 (instance.name, source_node, msg))
7721 # create the target disks
7723 _CreateDisks(self, instance, target_node=target_node)
7724 except errors.OpExecError:
7725 self.LogWarning("Device creation failed, reverting...")
7727 _RemoveDisks(self, instance, target_node=target_node)
7729 self.cfg.ReleaseDRBDMinors(instance.name)
7732 cluster_name = self.cfg.GetClusterInfo().cluster_name
7735 # activate, get path, copy the data over
7736 for idx, disk in enumerate(instance.disks):
7737 self.LogInfo("Copying data for disk %d", idx)
7738 result = self.rpc.call_blockdev_assemble(target_node, (disk, instance),
7739 instance.name, True, idx)
7741 self.LogWarning("Can't assemble newly created disk %d: %s",
7742 idx, result.fail_msg)
7743 errs.append(result.fail_msg)
7745 dev_path = result.payload
7746 result = self.rpc.call_blockdev_export(source_node, (disk, instance),
7747 target_node, dev_path,
7750 self.LogWarning("Can't copy data over for disk %d: %s",
7751 idx, result.fail_msg)
7752 errs.append(result.fail_msg)
7756 self.LogWarning("Some disks failed to copy, aborting")
7758 _RemoveDisks(self, instance, target_node=target_node)
7760 self.cfg.ReleaseDRBDMinors(instance.name)
7761 raise errors.OpExecError("Errors during disk copy: %s" %
7764 instance.primary_node = target_node
7765 self.cfg.Update(instance, feedback_fn)
7767 self.LogInfo("Removing the disks on the original node")
7768 _RemoveDisks(self, instance, target_node=source_node)
7770 # Only start the instance if it's marked as up
7771 if instance.admin_state == constants.ADMINST_UP:
7772 self.LogInfo("Starting instance %s on node %s",
7773 instance.name, target_node)
7775 disks_ok, _ = _AssembleInstanceDisks(self, instance,
7776 ignore_secondaries=True)
7778 _ShutdownInstanceDisks(self, instance)
7779 raise errors.OpExecError("Can't activate the instance's disks")
7781 result = self.rpc.call_instance_start(target_node,
7782 (instance, None, None), False)
7783 msg = result.fail_msg
7785 _ShutdownInstanceDisks(self, instance)
7786 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
7787 (instance.name, target_node, msg))
7790 class LUNodeMigrate(LogicalUnit):
7791 """Migrate all instances from a node.
7794 HPATH = "node-migrate"
7795 HTYPE = constants.HTYPE_NODE
7798 def CheckArguments(self):
7801 def ExpandNames(self):
7802 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
7804 self.share_locks = _ShareAll()
7805 self.needed_locks = {
7806 locking.LEVEL_NODE: [self.op.node_name],
7809 def BuildHooksEnv(self):
7812 This runs on the master, the primary and all the secondaries.
7816 "NODE_NAME": self.op.node_name,
7817 "ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes,
7820 def BuildHooksNodes(self):
7821 """Build hooks nodes.
7824 nl = [self.cfg.GetMasterNode()]
7827 def CheckPrereq(self):
7830 def Exec(self, feedback_fn):
7831 # Prepare jobs for migration instances
7832 allow_runtime_changes = self.op.allow_runtime_changes
7834 [opcodes.OpInstanceMigrate(instance_name=inst.name,
7837 iallocator=self.op.iallocator,
7838 target_node=self.op.target_node,
7839 allow_runtime_changes=allow_runtime_changes,
7840 ignore_ipolicy=self.op.ignore_ipolicy)]
7841 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name)
7844 # TODO: Run iallocator in this opcode and pass correct placement options to
7845 # OpInstanceMigrate. Since other jobs can modify the cluster between
7846 # running the iallocator and the actual migration, a good consistency model
7847 # will have to be found.
7849 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
7850 frozenset([self.op.node_name]))
7852 return ResultWithJobs(jobs)
7855 class TLMigrateInstance(Tasklet):
7856 """Tasklet class for instance migration.
7859 @ivar live: whether the migration will be done live or non-live;
7860 this variable is initalized only after CheckPrereq has run
7861 @type cleanup: boolean
7862 @ivar cleanup: Wheater we cleanup from a failed migration
7863 @type iallocator: string
7864 @ivar iallocator: The iallocator used to determine target_node
7865 @type target_node: string
7866 @ivar target_node: If given, the target_node to reallocate the instance to
7867 @type failover: boolean
7868 @ivar failover: Whether operation results in failover or migration
7869 @type fallback: boolean
7870 @ivar fallback: Whether fallback to failover is allowed if migration not
7872 @type ignore_consistency: boolean
7873 @ivar ignore_consistency: Wheter we should ignore consistency between source
7875 @type shutdown_timeout: int
7876 @ivar shutdown_timeout: In case of failover timeout of the shutdown
7877 @type ignore_ipolicy: bool
7878 @ivar ignore_ipolicy: If true, we can ignore instance policy when migrating
7883 _MIGRATION_POLL_INTERVAL = 1 # seconds
7884 _MIGRATION_FEEDBACK_INTERVAL = 10 # seconds
7886 def __init__(self, lu, instance_name, cleanup=False,
7887 failover=False, fallback=False,
7888 ignore_consistency=False,
7889 allow_runtime_changes=True,
7890 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT,
7891 ignore_ipolicy=False):
7892 """Initializes this class.
7895 Tasklet.__init__(self, lu)
7898 self.instance_name = instance_name
7899 self.cleanup = cleanup
7900 self.live = False # will be overridden later
7901 self.failover = failover
7902 self.fallback = fallback
7903 self.ignore_consistency = ignore_consistency
7904 self.shutdown_timeout = shutdown_timeout
7905 self.ignore_ipolicy = ignore_ipolicy
7906 self.allow_runtime_changes = allow_runtime_changes
7908 def CheckPrereq(self):
7909 """Check prerequisites.
7911 This checks that the instance is in the cluster.
7914 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
7915 instance = self.cfg.GetInstanceInfo(instance_name)
7916 assert instance is not None
7917 self.instance = instance
7918 cluster = self.cfg.GetClusterInfo()
7920 if (not self.cleanup and
7921 not instance.admin_state == constants.ADMINST_UP and
7922 not self.failover and self.fallback):
7923 self.lu.LogInfo("Instance is marked down or offline, fallback allowed,"
7924 " switching to failover")
7925 self.failover = True
7927 if instance.disk_template not in constants.DTS_MIRRORED:
7932 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
7933 " %s" % (instance.disk_template, text),
7936 if instance.disk_template in constants.DTS_EXT_MIRROR:
7937 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
7939 if self.lu.op.iallocator:
7940 self._RunAllocator()
7942 # We set set self.target_node as it is required by
7944 self.target_node = self.lu.op.target_node
7946 # Check that the target node is correct in terms of instance policy
7947 nodeinfo = self.cfg.GetNodeInfo(self.target_node)
7948 group_info = self.cfg.GetNodeGroup(nodeinfo.group)
7949 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
7950 _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
7951 ignore=self.ignore_ipolicy)
7953 # self.target_node is already populated, either directly or by the
7955 target_node = self.target_node
7956 if self.target_node == instance.primary_node:
7957 raise errors.OpPrereqError("Cannot migrate instance %s"
7958 " to its primary (%s)" %
7959 (instance.name, instance.primary_node))
7961 if len(self.lu.tasklets) == 1:
7962 # It is safe to release locks only when we're the only tasklet
7964 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
7965 keep=[instance.primary_node, self.target_node])
7968 secondary_nodes = instance.secondary_nodes
7969 if not secondary_nodes:
7970 raise errors.ConfigurationError("No secondary node but using"
7971 " %s disk template" %
7972 instance.disk_template)
7973 target_node = secondary_nodes[0]
7974 if self.lu.op.iallocator or (self.lu.op.target_node and
7975 self.lu.op.target_node != target_node):
7977 text = "failed over"
7980 raise errors.OpPrereqError("Instances with disk template %s cannot"
7981 " be %s to arbitrary nodes"
7982 " (neither an iallocator nor a target"
7983 " node can be passed)" %
7984 (instance.disk_template, text),
7986 nodeinfo = self.cfg.GetNodeInfo(target_node)
7987 group_info = self.cfg.GetNodeGroup(nodeinfo.group)
7988 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
7989 _CheckTargetNodeIPolicy(self.lu, ipolicy, instance, nodeinfo,
7990 ignore=self.ignore_ipolicy)
7992 i_be = cluster.FillBE(instance)
7994 # check memory requirements on the secondary node
7995 if (not self.cleanup and
7996 (not self.failover or instance.admin_state == constants.ADMINST_UP)):
7997 self.tgt_free_mem = _CheckNodeFreeMemory(self.lu, target_node,
7998 "migrating instance %s" %
8000 i_be[constants.BE_MINMEM],
8001 instance.hypervisor)
8003 self.lu.LogInfo("Not checking memory on the secondary node as"
8004 " instance will not be started")
8006 # check if failover must be forced instead of migration
8007 if (not self.cleanup and not self.failover and
8008 i_be[constants.BE_ALWAYS_FAILOVER]):
8010 self.lu.LogInfo("Instance configured to always failover; fallback"
8012 self.failover = True
8014 raise errors.OpPrereqError("This instance has been configured to"
8015 " always failover, please allow failover",
8018 # check bridge existance
8019 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
8021 if not self.cleanup:
8022 _CheckNodeNotDrained(self.lu, target_node)
8023 if not self.failover:
8024 result = self.rpc.call_instance_migratable(instance.primary_node,
8026 if result.fail_msg and self.fallback:
8027 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
8029 self.failover = True
8031 result.Raise("Can't migrate, please use failover",
8032 prereq=True, ecode=errors.ECODE_STATE)
8034 assert not (self.failover and self.cleanup)
8036 if not self.failover:
8037 if self.lu.op.live is not None and self.lu.op.mode is not None:
8038 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
8039 " parameters are accepted",
8041 if self.lu.op.live is not None:
8043 self.lu.op.mode = constants.HT_MIGRATION_LIVE
8045 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
8046 # reset the 'live' parameter to None so that repeated
8047 # invocations of CheckPrereq do not raise an exception
8048 self.lu.op.live = None
8049 elif self.lu.op.mode is None:
8050 # read the default value from the hypervisor
8051 i_hv = cluster.FillHV(self.instance, skip_globals=False)
8052 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
8054 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
8056 # Failover is never live
8059 if not (self.failover or self.cleanup):
8060 remote_info = self.rpc.call_instance_info(instance.primary_node,
8062 instance.hypervisor)
8063 remote_info.Raise("Error checking instance on node %s" %
8064 instance.primary_node)
8065 instance_running = bool(remote_info.payload)
8066 if instance_running:
8067 self.current_mem = int(remote_info.payload["memory"])
8069 def _RunAllocator(self):
8070 """Run the allocator based on input opcode.
8073 # FIXME: add a self.ignore_ipolicy option
8074 ial = IAllocator(self.cfg, self.rpc,
8075 mode=constants.IALLOCATOR_MODE_RELOC,
8076 name=self.instance_name,
8077 relocate_from=[self.instance.primary_node],
8080 ial.Run(self.lu.op.iallocator)
8083 raise errors.OpPrereqError("Can't compute nodes using"
8084 " iallocator '%s': %s" %
8085 (self.lu.op.iallocator, ial.info),
8087 if len(ial.result) != ial.required_nodes:
8088 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8089 " of nodes (%s), required %s" %
8090 (self.lu.op.iallocator, len(ial.result),
8091 ial.required_nodes), errors.ECODE_FAULT)
8092 self.target_node = ial.result[0]
8093 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
8094 self.instance_name, self.lu.op.iallocator,
8095 utils.CommaJoin(ial.result))
8097 def _WaitUntilSync(self):
8098 """Poll with custom rpc for disk sync.
8100 This uses our own step-based rpc call.
8103 self.feedback_fn("* wait until resync is done")
8107 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
8109 (self.instance.disks,
8112 for node, nres in result.items():
8113 nres.Raise("Cannot resync disks on node %s" % node)
8114 node_done, node_percent = nres.payload
8115 all_done = all_done and node_done
8116 if node_percent is not None:
8117 min_percent = min(min_percent, node_percent)
8119 if min_percent < 100:
8120 self.feedback_fn(" - progress: %.1f%%" % min_percent)
8123 def _EnsureSecondary(self, node):
8124 """Demote a node to secondary.
8127 self.feedback_fn("* switching node %s to secondary mode" % node)
8129 for dev in self.instance.disks:
8130 self.cfg.SetDiskID(dev, node)
8132 result = self.rpc.call_blockdev_close(node, self.instance.name,
8133 self.instance.disks)
8134 result.Raise("Cannot change disk to secondary on node %s" % node)
8136 def _GoStandalone(self):
8137 """Disconnect from the network.
8140 self.feedback_fn("* changing into standalone mode")
8141 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
8142 self.instance.disks)
8143 for node, nres in result.items():
8144 nres.Raise("Cannot disconnect disks node %s" % node)
8146 def _GoReconnect(self, multimaster):
8147 """Reconnect to the network.
8153 msg = "single-master"
8154 self.feedback_fn("* changing disks into %s mode" % msg)
8155 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
8156 (self.instance.disks, self.instance),
8157 self.instance.name, multimaster)
8158 for node, nres in result.items():
8159 nres.Raise("Cannot change disks config on node %s" % node)
8161 def _ExecCleanup(self):
8162 """Try to cleanup after a failed migration.
8164 The cleanup is done by:
8165 - check that the instance is running only on one node
8166 (and update the config if needed)
8167 - change disks on its secondary node to secondary
8168 - wait until disks are fully synchronized
8169 - disconnect from the network
8170 - change disks into single-master mode
8171 - wait again until disks are fully synchronized
8174 instance = self.instance
8175 target_node = self.target_node
8176 source_node = self.source_node
8178 # check running on only one node
8179 self.feedback_fn("* checking where the instance actually runs"
8180 " (if this hangs, the hypervisor might be in"
8182 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
8183 for node, result in ins_l.items():
8184 result.Raise("Can't contact node %s" % node)
8186 runningon_source = instance.name in ins_l[source_node].payload
8187 runningon_target = instance.name in ins_l[target_node].payload
8189 if runningon_source and runningon_target:
8190 raise errors.OpExecError("Instance seems to be running on two nodes,"
8191 " or the hypervisor is confused; you will have"
8192 " to ensure manually that it runs only on one"
8193 " and restart this operation")
8195 if not (runningon_source or runningon_target):
8196 raise errors.OpExecError("Instance does not seem to be running at all;"
8197 " in this case it's safer to repair by"
8198 " running 'gnt-instance stop' to ensure disk"
8199 " shutdown, and then restarting it")
8201 if runningon_target:
8202 # the migration has actually succeeded, we need to update the config
8203 self.feedback_fn("* instance running on secondary node (%s),"
8204 " updating config" % target_node)
8205 instance.primary_node = target_node
8206 self.cfg.Update(instance, self.feedback_fn)
8207 demoted_node = source_node
8209 self.feedback_fn("* instance confirmed to be running on its"
8210 " primary node (%s)" % source_node)
8211 demoted_node = target_node
8213 if instance.disk_template in constants.DTS_INT_MIRROR:
8214 self._EnsureSecondary(demoted_node)
8216 self._WaitUntilSync()
8217 except errors.OpExecError:
8218 # we ignore here errors, since if the device is standalone, it
8219 # won't be able to sync
8221 self._GoStandalone()
8222 self._GoReconnect(False)
8223 self._WaitUntilSync()
8225 self.feedback_fn("* done")
8227 def _RevertDiskStatus(self):
8228 """Try to revert the disk status after a failed migration.
8231 target_node = self.target_node
8232 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
8236 self._EnsureSecondary(target_node)
8237 self._GoStandalone()
8238 self._GoReconnect(False)
8239 self._WaitUntilSync()
8240 except errors.OpExecError, err:
8241 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
8242 " please try to recover the instance manually;"
8243 " error '%s'" % str(err))
8245 def _AbortMigration(self):
8246 """Call the hypervisor code to abort a started migration.
8249 instance = self.instance
8250 target_node = self.target_node
8251 source_node = self.source_node
8252 migration_info = self.migration_info
8254 abort_result = self.rpc.call_instance_finalize_migration_dst(target_node,
8258 abort_msg = abort_result.fail_msg
8260 logging.error("Aborting migration failed on target node %s: %s",
8261 target_node, abort_msg)
8262 # Don't raise an exception here, as we stil have to try to revert the
8263 # disk status, even if this step failed.
8265 abort_result = self.rpc.call_instance_finalize_migration_src(source_node,
8266 instance, False, self.live)
8267 abort_msg = abort_result.fail_msg
8269 logging.error("Aborting migration failed on source node %s: %s",
8270 source_node, abort_msg)
8272 def _ExecMigration(self):
8273 """Migrate an instance.
8275 The migrate is done by:
8276 - change the disks into dual-master mode
8277 - wait until disks are fully synchronized again
8278 - migrate the instance
8279 - change disks on the new secondary node (the old primary) to secondary
8280 - wait until disks are fully synchronized
8281 - change disks into single-master mode
8284 instance = self.instance
8285 target_node = self.target_node
8286 source_node = self.source_node
8288 # Check for hypervisor version mismatch and warn the user.
8289 nodeinfo = self.rpc.call_node_info([source_node, target_node],
8290 None, [self.instance.hypervisor])
8291 for ninfo in nodeinfo.values():
8292 ninfo.Raise("Unable to retrieve node information from node '%s'" %
8294 (_, _, (src_info, )) = nodeinfo[source_node].payload
8295 (_, _, (dst_info, )) = nodeinfo[target_node].payload
8297 if ((constants.HV_NODEINFO_KEY_VERSION in src_info) and
8298 (constants.HV_NODEINFO_KEY_VERSION in dst_info)):
8299 src_version = src_info[constants.HV_NODEINFO_KEY_VERSION]
8300 dst_version = dst_info[constants.HV_NODEINFO_KEY_VERSION]
8301 if src_version != dst_version:
8302 self.feedback_fn("* warning: hypervisor version mismatch between"
8303 " source (%s) and target (%s) node" %
8304 (src_version, dst_version))
8306 self.feedback_fn("* checking disk consistency between source and target")
8307 for (idx, dev) in enumerate(instance.disks):
8308 if not _CheckDiskConsistency(self.lu, instance, dev, target_node, False):
8309 raise errors.OpExecError("Disk %s is degraded or not fully"
8310 " synchronized on target node,"
8311 " aborting migration" % idx)
8313 if self.current_mem > self.tgt_free_mem:
8314 if not self.allow_runtime_changes:
8315 raise errors.OpExecError("Memory ballooning not allowed and not enough"
8316 " free memory to fit instance %s on target"
8317 " node %s (have %dMB, need %dMB)" %
8318 (instance.name, target_node,
8319 self.tgt_free_mem, self.current_mem))
8320 self.feedback_fn("* setting instance memory to %s" % self.tgt_free_mem)
8321 rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
8324 rpcres.Raise("Cannot modify instance runtime memory")
8326 # First get the migration information from the remote node
8327 result = self.rpc.call_migration_info(source_node, instance)
8328 msg = result.fail_msg
8330 log_err = ("Failed fetching source migration information from %s: %s" %
8332 logging.error(log_err)
8333 raise errors.OpExecError(log_err)
8335 self.migration_info = migration_info = result.payload
8337 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
8338 # Then switch the disks to master/master mode
8339 self._EnsureSecondary(target_node)
8340 self._GoStandalone()
8341 self._GoReconnect(True)
8342 self._WaitUntilSync()
8344 self.feedback_fn("* preparing %s to accept the instance" % target_node)
8345 result = self.rpc.call_accept_instance(target_node,
8348 self.nodes_ip[target_node])
8350 msg = result.fail_msg
8352 logging.error("Instance pre-migration failed, trying to revert"
8353 " disk status: %s", msg)
8354 self.feedback_fn("Pre-migration failed, aborting")
8355 self._AbortMigration()
8356 self._RevertDiskStatus()
8357 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
8358 (instance.name, msg))
8360 self.feedback_fn("* migrating instance to %s" % target_node)
8361 result = self.rpc.call_instance_migrate(source_node, instance,
8362 self.nodes_ip[target_node],
8364 msg = result.fail_msg
8366 logging.error("Instance migration failed, trying to revert"
8367 " disk status: %s", msg)
8368 self.feedback_fn("Migration failed, aborting")
8369 self._AbortMigration()
8370 self._RevertDiskStatus()
8371 raise errors.OpExecError("Could not migrate instance %s: %s" %
8372 (instance.name, msg))
8374 self.feedback_fn("* starting memory transfer")
8375 last_feedback = time.time()
8377 result = self.rpc.call_instance_get_migration_status(source_node,
8379 msg = result.fail_msg
8380 ms = result.payload # MigrationStatus instance
8381 if msg or (ms.status in constants.HV_MIGRATION_FAILED_STATUSES):
8382 logging.error("Instance migration failed, trying to revert"
8383 " disk status: %s", msg)
8384 self.feedback_fn("Migration failed, aborting")
8385 self._AbortMigration()
8386 self._RevertDiskStatus()
8387 raise errors.OpExecError("Could not migrate instance %s: %s" %
8388 (instance.name, msg))
8390 if result.payload.status != constants.HV_MIGRATION_ACTIVE:
8391 self.feedback_fn("* memory transfer complete")
8394 if (utils.TimeoutExpired(last_feedback,
8395 self._MIGRATION_FEEDBACK_INTERVAL) and
8396 ms.transferred_ram is not None):
8397 mem_progress = 100 * float(ms.transferred_ram) / float(ms.total_ram)
8398 self.feedback_fn("* memory transfer progress: %.2f %%" % mem_progress)
8399 last_feedback = time.time()
8401 time.sleep(self._MIGRATION_POLL_INTERVAL)
8403 result = self.rpc.call_instance_finalize_migration_src(source_node,
8407 msg = result.fail_msg
8409 logging.error("Instance migration succeeded, but finalization failed"
8410 " on the source node: %s", msg)
8411 raise errors.OpExecError("Could not finalize instance migration: %s" %
8414 instance.primary_node = target_node
8416 # distribute new instance config to the other nodes
8417 self.cfg.Update(instance, self.feedback_fn)
8419 result = self.rpc.call_instance_finalize_migration_dst(target_node,
8423 msg = result.fail_msg
8425 logging.error("Instance migration succeeded, but finalization failed"
8426 " on the target node: %s", msg)
8427 raise errors.OpExecError("Could not finalize instance migration: %s" %
8430 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
8431 self._EnsureSecondary(source_node)
8432 self._WaitUntilSync()
8433 self._GoStandalone()
8434 self._GoReconnect(False)
8435 self._WaitUntilSync()
8437 # If the instance's disk template is `rbd' and there was a successful
8438 # migration, unmap the device from the source node.
8439 if self.instance.disk_template == constants.DT_RBD:
8440 disks = _ExpandCheckDisks(instance, instance.disks)
8441 self.feedback_fn("* unmapping instance's disks from %s" % source_node)
8443 result = self.rpc.call_blockdev_shutdown(source_node, disk)
8444 msg = result.fail_msg
8446 logging.error("Migration was successful, but couldn't unmap the"
8447 " block device %s on source node %s: %s",
8448 disk.iv_name, source_node, msg)
8449 logging.error("You need to unmap the device %s manually on %s",
8450 disk.iv_name, source_node)
8452 self.feedback_fn("* done")
8454 def _ExecFailover(self):
8455 """Failover an instance.
8457 The failover is done by shutting it down on its present node and
8458 starting it on the secondary.
8461 instance = self.instance
8462 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
8464 source_node = instance.primary_node
8465 target_node = self.target_node
8467 if instance.admin_state == constants.ADMINST_UP:
8468 self.feedback_fn("* checking disk consistency between source and target")
8469 for (idx, dev) in enumerate(instance.disks):
8470 # for drbd, these are drbd over lvm
8471 if not _CheckDiskConsistency(self.lu, instance, dev, target_node,
8473 if primary_node.offline:
8474 self.feedback_fn("Node %s is offline, ignoring degraded disk %s on"
8476 (primary_node.name, idx, target_node))
8477 elif not self.ignore_consistency:
8478 raise errors.OpExecError("Disk %s is degraded on target node,"
8479 " aborting failover" % idx)
8481 self.feedback_fn("* not checking disk consistency as instance is not"
8484 self.feedback_fn("* shutting down instance on source node")
8485 logging.info("Shutting down instance %s on node %s",
8486 instance.name, source_node)
8488 result = self.rpc.call_instance_shutdown(source_node, instance,
8489 self.shutdown_timeout)
8490 msg = result.fail_msg
8492 if self.ignore_consistency or primary_node.offline:
8493 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
8494 " proceeding anyway; please make sure node"
8495 " %s is down; error details: %s",
8496 instance.name, source_node, source_node, msg)
8498 raise errors.OpExecError("Could not shutdown instance %s on"
8500 (instance.name, source_node, msg))
8502 self.feedback_fn("* deactivating the instance's disks on source node")
8503 if not _ShutdownInstanceDisks(self.lu, instance, ignore_primary=True):
8504 raise errors.OpExecError("Can't shut down the instance's disks")
8506 instance.primary_node = target_node
8507 # distribute new instance config to the other nodes
8508 self.cfg.Update(instance, self.feedback_fn)
8510 # Only start the instance if it's marked as up
8511 if instance.admin_state == constants.ADMINST_UP:
8512 self.feedback_fn("* activating the instance's disks on target node %s" %
8514 logging.info("Starting instance %s on node %s",
8515 instance.name, target_node)
8517 disks_ok, _ = _AssembleInstanceDisks(self.lu, instance,
8518 ignore_secondaries=True)
8520 _ShutdownInstanceDisks(self.lu, instance)
8521 raise errors.OpExecError("Can't activate the instance's disks")
8523 self.feedback_fn("* starting the instance on the target node %s" %
8525 result = self.rpc.call_instance_start(target_node, (instance, None, None),
8527 msg = result.fail_msg
8529 _ShutdownInstanceDisks(self.lu, instance)
8530 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
8531 (instance.name, target_node, msg))
8533 def Exec(self, feedback_fn):
8534 """Perform the migration.
8537 self.feedback_fn = feedback_fn
8538 self.source_node = self.instance.primary_node
8540 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
8541 if self.instance.disk_template in constants.DTS_INT_MIRROR:
8542 self.target_node = self.instance.secondary_nodes[0]
8543 # Otherwise self.target_node has been populated either
8544 # directly, or through an iallocator.
8546 self.all_nodes = [self.source_node, self.target_node]
8547 self.nodes_ip = dict((name, node.secondary_ip) for (name, node)
8548 in self.cfg.GetMultiNodeInfo(self.all_nodes))
8551 feedback_fn("Failover instance %s" % self.instance.name)
8552 self._ExecFailover()
8554 feedback_fn("Migrating instance %s" % self.instance.name)
8557 return self._ExecCleanup()
8559 return self._ExecMigration()
8562 def _CreateBlockDev(lu, node, instance, device, force_create,
8564 """Create a tree of block devices on a given node.
8566 If this device type has to be created on secondaries, create it and
8569 If not, just recurse to children keeping the same 'force' value.
8571 @param lu: the lu on whose behalf we execute
8572 @param node: the node on which to create the device
8573 @type instance: L{objects.Instance}
8574 @param instance: the instance which owns the device
8575 @type device: L{objects.Disk}
8576 @param device: the device to create
8577 @type force_create: boolean
8578 @param force_create: whether to force creation of this device; this
8579 will be change to True whenever we find a device which has
8580 CreateOnSecondary() attribute
8581 @param info: the extra 'metadata' we should attach to the device
8582 (this will be represented as a LVM tag)
8583 @type force_open: boolean
8584 @param force_open: this parameter will be passes to the
8585 L{backend.BlockdevCreate} function where it specifies
8586 whether we run on primary or not, and it affects both
8587 the child assembly and the device own Open() execution
8590 if device.CreateOnSecondary():
8594 for child in device.children:
8595 _CreateBlockDev(lu, node, instance, child, force_create,
8598 if not force_create:
8601 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
8604 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
8605 """Create a single block device on a given node.
8607 This will not recurse over children of the device, so they must be
8610 @param lu: the lu on whose behalf we execute
8611 @param node: the node on which to create the device
8612 @type instance: L{objects.Instance}
8613 @param instance: the instance which owns the device
8614 @type device: L{objects.Disk}
8615 @param device: the device to create
8616 @param info: the extra 'metadata' we should attach to the device
8617 (this will be represented as a LVM tag)
8618 @type force_open: boolean
8619 @param force_open: this parameter will be passes to the
8620 L{backend.BlockdevCreate} function where it specifies
8621 whether we run on primary or not, and it affects both
8622 the child assembly and the device own Open() execution
8625 lu.cfg.SetDiskID(device, node)
8626 result = lu.rpc.call_blockdev_create(node, device, device.size,
8627 instance.name, force_open, info)
8628 result.Raise("Can't create block device %s on"
8629 " node %s for instance %s" % (device, node, instance.name))
8630 if device.physical_id is None:
8631 device.physical_id = result.payload
8634 def _GenerateUniqueNames(lu, exts):
8635 """Generate a suitable LV name.
8637 This will generate a logical volume name for the given instance.
8642 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
8643 results.append("%s%s" % (new_id, val))
8647 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
8648 iv_name, p_minor, s_minor, drbd_params, data_params,
8650 """Generate a drbd8 device complete with its children.
8653 assert len(vgnames) == len(names) == 2
8654 port = lu.cfg.AllocatePort()
8655 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
8657 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
8658 logical_id=(vgnames[0], names[0]),
8660 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,
8661 logical_id=(vgnames[1], names[1]),
8663 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
8664 logical_id=(primary, secondary, port,
8667 children=[dev_data, dev_meta],
8668 iv_name=iv_name, params=drbd_params)
8672 _DISK_TEMPLATE_NAME_PREFIX = {
8673 constants.DT_PLAIN: "",
8674 constants.DT_RBD: ".rbd",
8678 _DISK_TEMPLATE_DEVICE_TYPE = {
8679 constants.DT_PLAIN: constants.LD_LV,
8680 constants.DT_FILE: constants.LD_FILE,
8681 constants.DT_SHARED_FILE: constants.LD_FILE,
8682 constants.DT_BLOCK: constants.LD_BLOCKDEV,
8683 constants.DT_RBD: constants.LD_RBD,
8687 def _GenerateDiskTemplate(lu, template_name, instance_name, primary_node,
8688 secondary_nodes, disk_info, file_storage_dir, file_driver, base_index,
8689 feedback_fn, disk_params,
8690 _req_file_storage=opcodes.RequireFileStorage,
8691 _req_shr_file_storage=opcodes.RequireSharedFileStorage):
8692 """Generate the entire disk layout for a given template type.
8695 #TODO: compute space requirements
8697 vgname = lu.cfg.GetVGName()
8698 disk_count = len(disk_info)
8700 ld_params = _ComputeLDParams(template_name, disk_params)
8702 if template_name == constants.DT_DISKLESS:
8704 elif template_name == constants.DT_DRBD8:
8705 drbd_params, data_params, meta_params = ld_params
8706 if len(secondary_nodes) != 1:
8707 raise errors.ProgrammerError("Wrong template configuration")
8708 remote_node = secondary_nodes[0]
8709 minors = lu.cfg.AllocateDRBDMinor(
8710 [primary_node, remote_node] * len(disk_info), instance_name)
8713 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
8714 for i in range(disk_count)]):
8715 names.append(lv_prefix + "_data")
8716 names.append(lv_prefix + "_meta")
8717 for idx, disk in enumerate(disk_info):
8718 disk_index = idx + base_index
8719 drbd_default_metavg = drbd_params[constants.LDP_DEFAULT_METAVG]
8720 data_vg = disk.get(constants.IDISK_VG, vgname)
8721 meta_vg = disk.get(constants.IDISK_METAVG, drbd_default_metavg)
8722 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
8723 disk[constants.IDISK_SIZE],
8725 names[idx * 2:idx * 2 + 2],
8726 "disk/%d" % disk_index,
8727 minors[idx * 2], minors[idx * 2 + 1],
8728 drbd_params, data_params, meta_params)
8729 disk_dev.mode = disk[constants.IDISK_MODE]
8730 disks.append(disk_dev)
8733 raise errors.ProgrammerError("Wrong template configuration")
8735 if template_name == constants.DT_FILE:
8737 elif template_name == constants.DT_SHARED_FILE:
8738 _req_shr_file_storage()
8740 name_prefix = _DISK_TEMPLATE_NAME_PREFIX.get(template_name, None)
8741 if name_prefix is None:
8744 names = _GenerateUniqueNames(lu, ["%s.disk%s" %
8745 (name_prefix, base_index + i)
8746 for i in range(disk_count)])
8748 dev_type = _DISK_TEMPLATE_DEVICE_TYPE[template_name]
8750 if template_name == constants.DT_PLAIN:
8751 def logical_id_fn(idx, _, disk):
8752 vg = disk.get(constants.IDISK_VG, vgname)
8753 return (vg, names[idx])
8754 elif template_name in (constants.DT_FILE, constants.DT_SHARED_FILE):
8756 lambda _, disk_index, disk: (file_driver,
8757 "%s/disk%d" % (file_storage_dir,
8759 elif template_name == constants.DT_BLOCK:
8761 lambda idx, disk_index, disk: (constants.BLOCKDEV_DRIVER_MANUAL,
8762 disk[constants.IDISK_ADOPT])
8763 elif template_name == constants.DT_RBD:
8764 logical_id_fn = lambda idx, _, disk: ("rbd", names[idx])
8766 raise errors.ProgrammerError("Unknown disk template '%s'" % template_name)
8768 for idx, disk in enumerate(disk_info):
8769 disk_index = idx + base_index
8770 size = disk[constants.IDISK_SIZE]
8771 feedback_fn("* disk %s, size %s" %
8772 (disk_index, utils.FormatUnit(size, "h")))
8773 disks.append(objects.Disk(dev_type=dev_type, size=size,
8774 logical_id=logical_id_fn(idx, disk_index, disk),
8775 iv_name="disk/%d" % disk_index,
8776 mode=disk[constants.IDISK_MODE],
8777 params=ld_params[0]))
8782 def _GetInstanceInfoText(instance):
8783 """Compute that text that should be added to the disk's metadata.
8786 return "originstname+%s" % instance.name
8789 def _CalcEta(time_taken, written, total_size):
8790 """Calculates the ETA based on size written and total size.
8792 @param time_taken: The time taken so far
8793 @param written: amount written so far
8794 @param total_size: The total size of data to be written
8795 @return: The remaining time in seconds
8798 avg_time = time_taken / float(written)
8799 return (total_size - written) * avg_time
8802 def _WipeDisks(lu, instance):
8803 """Wipes instance disks.
8805 @type lu: L{LogicalUnit}
8806 @param lu: the logical unit on whose behalf we execute
8807 @type instance: L{objects.Instance}
8808 @param instance: the instance whose disks we should create
8809 @return: the success of the wipe
8812 node = instance.primary_node
8814 for device in instance.disks:
8815 lu.cfg.SetDiskID(device, node)
8817 logging.info("Pause sync of instance %s disks", instance.name)
8818 result = lu.rpc.call_blockdev_pause_resume_sync(node,
8819 (instance.disks, instance),
8822 for idx, success in enumerate(result.payload):
8824 logging.warn("pause-sync of instance %s for disks %d failed",
8828 for idx, device in enumerate(instance.disks):
8829 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
8830 # MAX_WIPE_CHUNK at max
8831 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
8832 constants.MIN_WIPE_CHUNK_PERCENT)
8833 # we _must_ make this an int, otherwise rounding errors will
8835 wipe_chunk_size = int(wipe_chunk_size)
8837 lu.LogInfo("* Wiping disk %d", idx)
8838 logging.info("Wiping disk %d for instance %s, node %s using"
8839 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
8844 start_time = time.time()
8846 while offset < size:
8847 wipe_size = min(wipe_chunk_size, size - offset)
8848 logging.debug("Wiping disk %d, offset %s, chunk %s",
8849 idx, offset, wipe_size)
8850 result = lu.rpc.call_blockdev_wipe(node, (device, instance), offset,
8852 result.Raise("Could not wipe disk %d at offset %d for size %d" %
8853 (idx, offset, wipe_size))
8856 if now - last_output >= 60:
8857 eta = _CalcEta(now - start_time, offset, size)
8858 lu.LogInfo(" - done: %.1f%% ETA: %s" %
8859 (offset / float(size) * 100, utils.FormatSeconds(eta)))
8862 logging.info("Resume sync of instance %s disks", instance.name)
8864 result = lu.rpc.call_blockdev_pause_resume_sync(node,
8865 (instance.disks, instance),
8868 for idx, success in enumerate(result.payload):
8870 lu.LogWarning("Resume sync of disk %d failed, please have a"
8871 " look at the status and troubleshoot the issue", idx)
8872 logging.warn("resume-sync of instance %s for disks %d failed",
8876 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
8877 """Create all disks for an instance.
8879 This abstracts away some work from AddInstance.
8881 @type lu: L{LogicalUnit}
8882 @param lu: the logical unit on whose behalf we execute
8883 @type instance: L{objects.Instance}
8884 @param instance: the instance whose disks we should create
8886 @param to_skip: list of indices to skip
8887 @type target_node: string
8888 @param target_node: if passed, overrides the target node for creation
8890 @return: the success of the creation
8893 info = _GetInstanceInfoText(instance)
8894 if target_node is None:
8895 pnode = instance.primary_node
8896 all_nodes = instance.all_nodes
8901 if instance.disk_template in constants.DTS_FILEBASED:
8902 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
8903 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
8905 result.Raise("Failed to create directory '%s' on"
8906 " node %s" % (file_storage_dir, pnode))
8908 # Note: this needs to be kept in sync with adding of disks in
8909 # LUInstanceSetParams
8910 for idx, device in enumerate(instance.disks):
8911 if to_skip and idx in to_skip:
8913 logging.info("Creating disk %s for instance '%s'", idx, instance.name)
8915 for node in all_nodes:
8916 f_create = node == pnode
8917 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
8920 def _RemoveDisks(lu, instance, target_node=None, ignore_failures=False):
8921 """Remove all disks for an instance.
8923 This abstracts away some work from `AddInstance()` and
8924 `RemoveInstance()`. Note that in case some of the devices couldn't
8925 be removed, the removal will continue with the other ones (compare
8926 with `_CreateDisks()`).
8928 @type lu: L{LogicalUnit}
8929 @param lu: the logical unit on whose behalf we execute
8930 @type instance: L{objects.Instance}
8931 @param instance: the instance whose disks we should remove
8932 @type target_node: string
8933 @param target_node: used to override the node on which to remove the disks
8935 @return: the success of the removal
8938 logging.info("Removing block devices for instance %s", instance.name)
8941 ports_to_release = set()
8942 for (idx, device) in enumerate(instance.disks):
8944 edata = [(target_node, device)]
8946 edata = device.ComputeNodeTree(instance.primary_node)
8947 for node, disk in edata:
8948 lu.cfg.SetDiskID(disk, node)
8949 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
8951 lu.LogWarning("Could not remove disk %s on node %s,"
8952 " continuing anyway: %s", idx, node, msg)
8955 # if this is a DRBD disk, return its port to the pool
8956 if device.dev_type in constants.LDS_DRBD:
8957 ports_to_release.add(device.logical_id[2])
8959 if all_result or ignore_failures:
8960 for port in ports_to_release:
8961 lu.cfg.AddTcpUdpPort(port)
8963 if instance.disk_template == constants.DT_FILE:
8964 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
8968 tgt = instance.primary_node
8969 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
8971 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
8972 file_storage_dir, instance.primary_node, result.fail_msg)
8978 def _ComputeDiskSizePerVG(disk_template, disks):
8979 """Compute disk size requirements in the volume group
8982 def _compute(disks, payload):
8983 """Universal algorithm.
8988 vgs[disk[constants.IDISK_VG]] = \
8989 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
8993 # Required free disk space as a function of disk and swap space
8995 constants.DT_DISKLESS: {},
8996 constants.DT_PLAIN: _compute(disks, 0),
8997 # 128 MB are added for drbd metadata for each disk
8998 constants.DT_DRBD8: _compute(disks, DRBD_META_SIZE),
8999 constants.DT_FILE: {},
9000 constants.DT_SHARED_FILE: {},
9003 if disk_template not in req_size_dict:
9004 raise errors.ProgrammerError("Disk template '%s' size requirement"
9005 " is unknown" % disk_template)
9007 return req_size_dict[disk_template]
9010 def _ComputeDiskSize(disk_template, disks):
9011 """Compute disk size requirements in the volume group
9014 # Required free disk space as a function of disk and swap space
9016 constants.DT_DISKLESS: None,
9017 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
9018 # 128 MB are added for drbd metadata for each disk
9020 sum(d[constants.IDISK_SIZE] + DRBD_META_SIZE for d in disks),
9021 constants.DT_FILE: None,
9022 constants.DT_SHARED_FILE: 0,
9023 constants.DT_BLOCK: 0,
9024 constants.DT_RBD: 0,
9027 if disk_template not in req_size_dict:
9028 raise errors.ProgrammerError("Disk template '%s' size requirement"
9029 " is unknown" % disk_template)
9031 return req_size_dict[disk_template]
9034 def _FilterVmNodes(lu, nodenames):
9035 """Filters out non-vm_capable nodes from a list.
9037 @type lu: L{LogicalUnit}
9038 @param lu: the logical unit for which we check
9039 @type nodenames: list
9040 @param nodenames: the list of nodes on which we should check
9042 @return: the list of vm-capable nodes
9045 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
9046 return [name for name in nodenames if name not in vm_nodes]
9049 def _CheckHVParams(lu, nodenames, hvname, hvparams):
9050 """Hypervisor parameter validation.
9052 This function abstract the hypervisor parameter validation to be
9053 used in both instance create and instance modify.
9055 @type lu: L{LogicalUnit}
9056 @param lu: the logical unit for which we check
9057 @type nodenames: list
9058 @param nodenames: the list of nodes on which we should check
9059 @type hvname: string
9060 @param hvname: the name of the hypervisor we should use
9061 @type hvparams: dict
9062 @param hvparams: the parameters which we need to check
9063 @raise errors.OpPrereqError: if the parameters are not valid
9066 nodenames = _FilterVmNodes(lu, nodenames)
9068 cluster = lu.cfg.GetClusterInfo()
9069 hvfull = objects.FillDict(cluster.hvparams.get(hvname, {}), hvparams)
9071 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames, hvname, hvfull)
9072 for node in nodenames:
9076 info.Raise("Hypervisor parameter validation failed on node %s" % node)
9079 def _CheckOSParams(lu, required, nodenames, osname, osparams):
9080 """OS parameters validation.
9082 @type lu: L{LogicalUnit}
9083 @param lu: the logical unit for which we check
9084 @type required: boolean
9085 @param required: whether the validation should fail if the OS is not
9087 @type nodenames: list
9088 @param nodenames: the list of nodes on which we should check
9089 @type osname: string
9090 @param osname: the name of the hypervisor we should use
9091 @type osparams: dict
9092 @param osparams: the parameters which we need to check
9093 @raise errors.OpPrereqError: if the parameters are not valid
9096 nodenames = _FilterVmNodes(lu, nodenames)
9097 result = lu.rpc.call_os_validate(nodenames, required, osname,
9098 [constants.OS_VALIDATE_PARAMETERS],
9100 for node, nres in result.items():
9101 # we don't check for offline cases since this should be run only
9102 # against the master node and/or an instance's nodes
9103 nres.Raise("OS Parameters validation failed on node %s" % node)
9104 if not nres.payload:
9105 lu.LogInfo("OS %s not found on node %s, validation skipped",
9109 class LUInstanceCreate(LogicalUnit):
9110 """Create an instance.
9113 HPATH = "instance-add"
9114 HTYPE = constants.HTYPE_INSTANCE
9117 def CheckArguments(self):
9121 # do not require name_check to ease forward/backward compatibility
9123 if self.op.no_install and self.op.start:
9124 self.LogInfo("No-installation mode selected, disabling startup")
9125 self.op.start = False
9126 # validate/normalize the instance name
9127 self.op.instance_name = \
9128 netutils.Hostname.GetNormalizedName(self.op.instance_name)
9130 if self.op.ip_check and not self.op.name_check:
9131 # TODO: make the ip check more flexible and not depend on the name check
9132 raise errors.OpPrereqError("Cannot do IP address check without a name"
9133 " check", errors.ECODE_INVAL)
9135 # check nics' parameter names
9136 for nic in self.op.nics:
9137 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
9139 # check disks. parameter names and consistent adopt/no-adopt strategy
9140 has_adopt = has_no_adopt = False
9141 for disk in self.op.disks:
9142 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
9143 if constants.IDISK_ADOPT in disk:
9147 if has_adopt and has_no_adopt:
9148 raise errors.OpPrereqError("Either all disks are adopted or none is",
9151 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
9152 raise errors.OpPrereqError("Disk adoption is not supported for the"
9153 " '%s' disk template" %
9154 self.op.disk_template,
9156 if self.op.iallocator is not None:
9157 raise errors.OpPrereqError("Disk adoption not allowed with an"
9158 " iallocator script", errors.ECODE_INVAL)
9159 if self.op.mode == constants.INSTANCE_IMPORT:
9160 raise errors.OpPrereqError("Disk adoption not allowed for"
9161 " instance import", errors.ECODE_INVAL)
9163 if self.op.disk_template in constants.DTS_MUST_ADOPT:
9164 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
9165 " but no 'adopt' parameter given" %
9166 self.op.disk_template,
9169 self.adopt_disks = has_adopt
9171 # instance name verification
9172 if self.op.name_check:
9173 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
9174 self.op.instance_name = self.hostname1.name
9175 # used in CheckPrereq for ip ping check
9176 self.check_ip = self.hostname1.ip
9178 self.check_ip = None
9180 # file storage checks
9181 if (self.op.file_driver and
9182 not self.op.file_driver in constants.FILE_DRIVER):
9183 raise errors.OpPrereqError("Invalid file driver name '%s'" %
9184 self.op.file_driver, errors.ECODE_INVAL)
9186 if self.op.disk_template == constants.DT_FILE:
9187 opcodes.RequireFileStorage()
9188 elif self.op.disk_template == constants.DT_SHARED_FILE:
9189 opcodes.RequireSharedFileStorage()
9191 ### Node/iallocator related checks
9192 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
9194 if self.op.pnode is not None:
9195 if self.op.disk_template in constants.DTS_INT_MIRROR:
9196 if self.op.snode is None:
9197 raise errors.OpPrereqError("The networked disk templates need"
9198 " a mirror node", errors.ECODE_INVAL)
9200 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
9202 self.op.snode = None
9204 self._cds = _GetClusterDomainSecret()
9206 if self.op.mode == constants.INSTANCE_IMPORT:
9207 # On import force_variant must be True, because if we forced it at
9208 # initial install, our only chance when importing it back is that it
9210 self.op.force_variant = True
9212 if self.op.no_install:
9213 self.LogInfo("No-installation mode has no effect during import")
9215 elif self.op.mode == constants.INSTANCE_CREATE:
9216 if self.op.os_type is None:
9217 raise errors.OpPrereqError("No guest OS specified",
9219 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
9220 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
9221 " installation" % self.op.os_type,
9223 if self.op.disk_template is None:
9224 raise errors.OpPrereqError("No disk template specified",
9227 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
9228 # Check handshake to ensure both clusters have the same domain secret
9229 src_handshake = self.op.source_handshake
9230 if not src_handshake:
9231 raise errors.OpPrereqError("Missing source handshake",
9234 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
9237 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
9240 # Load and check source CA
9241 self.source_x509_ca_pem = self.op.source_x509_ca
9242 if not self.source_x509_ca_pem:
9243 raise errors.OpPrereqError("Missing source X509 CA",
9247 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
9249 except OpenSSL.crypto.Error, err:
9250 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
9251 (err, ), errors.ECODE_INVAL)
9253 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9254 if errcode is not None:
9255 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
9258 self.source_x509_ca = cert
9260 src_instance_name = self.op.source_instance_name
9261 if not src_instance_name:
9262 raise errors.OpPrereqError("Missing source instance name",
9265 self.source_instance_name = \
9266 netutils.GetHostname(name=src_instance_name).name
9269 raise errors.OpPrereqError("Invalid instance creation mode %r" %
9270 self.op.mode, errors.ECODE_INVAL)
9272 def ExpandNames(self):
9273 """ExpandNames for CreateInstance.
9275 Figure out the right locks for instance creation.
9278 self.needed_locks = {}
9280 instance_name = self.op.instance_name
9281 # this is just a preventive check, but someone might still add this
9282 # instance in the meantime, and creation will fail at lock-add time
9283 if instance_name in self.cfg.GetInstanceList():
9284 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
9285 instance_name, errors.ECODE_EXISTS)
9287 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
9289 if self.op.iallocator:
9290 # TODO: Find a solution to not lock all nodes in the cluster, e.g. by
9291 # specifying a group on instance creation and then selecting nodes from
9293 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9294 self.needed_locks[locking.LEVEL_NODE_RES] = locking.ALL_SET
9296 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
9297 nodelist = [self.op.pnode]
9298 if self.op.snode is not None:
9299 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
9300 nodelist.append(self.op.snode)
9301 self.needed_locks[locking.LEVEL_NODE] = nodelist
9302 # Lock resources of instance's primary and secondary nodes (copy to
9303 # prevent accidential modification)
9304 self.needed_locks[locking.LEVEL_NODE_RES] = list(nodelist)
9306 # in case of import lock the source node too
9307 if self.op.mode == constants.INSTANCE_IMPORT:
9308 src_node = self.op.src_node
9309 src_path = self.op.src_path
9311 if src_path is None:
9312 self.op.src_path = src_path = self.op.instance_name
9314 if src_node is None:
9315 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9316 self.op.src_node = None
9317 if os.path.isabs(src_path):
9318 raise errors.OpPrereqError("Importing an instance from a path"
9319 " requires a source node option",
9322 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
9323 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
9324 self.needed_locks[locking.LEVEL_NODE].append(src_node)
9325 if not os.path.isabs(src_path):
9326 self.op.src_path = src_path = \
9327 utils.PathJoin(constants.EXPORT_DIR, src_path)
9329 def _RunAllocator(self):
9330 """Run the allocator based on input opcode.
9333 nics = [n.ToDict() for n in self.nics]
9334 ial = IAllocator(self.cfg, self.rpc,
9335 mode=constants.IALLOCATOR_MODE_ALLOC,
9336 name=self.op.instance_name,
9337 disk_template=self.op.disk_template,
9340 vcpus=self.be_full[constants.BE_VCPUS],
9341 memory=self.be_full[constants.BE_MAXMEM],
9342 spindle_use=self.be_full[constants.BE_SPINDLE_USE],
9345 hypervisor=self.op.hypervisor,
9348 ial.Run(self.op.iallocator)
9351 raise errors.OpPrereqError("Can't compute nodes using"
9352 " iallocator '%s': %s" %
9353 (self.op.iallocator, ial.info),
9355 if len(ial.result) != ial.required_nodes:
9356 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
9357 " of nodes (%s), required %s" %
9358 (self.op.iallocator, len(ial.result),
9359 ial.required_nodes), errors.ECODE_FAULT)
9360 self.op.pnode = ial.result[0]
9361 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
9362 self.op.instance_name, self.op.iallocator,
9363 utils.CommaJoin(ial.result))
9364 if ial.required_nodes == 2:
9365 self.op.snode = ial.result[1]
9367 def BuildHooksEnv(self):
9370 This runs on master, primary and secondary nodes of the instance.
9374 "ADD_MODE": self.op.mode,
9376 if self.op.mode == constants.INSTANCE_IMPORT:
9377 env["SRC_NODE"] = self.op.src_node
9378 env["SRC_PATH"] = self.op.src_path
9379 env["SRC_IMAGES"] = self.src_images
9381 env.update(_BuildInstanceHookEnv(
9382 name=self.op.instance_name,
9383 primary_node=self.op.pnode,
9384 secondary_nodes=self.secondaries,
9385 status=self.op.start,
9386 os_type=self.op.os_type,
9387 minmem=self.be_full[constants.BE_MINMEM],
9388 maxmem=self.be_full[constants.BE_MAXMEM],
9389 vcpus=self.be_full[constants.BE_VCPUS],
9390 nics=_NICListToTuple(self, self.nics),
9391 disk_template=self.op.disk_template,
9392 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
9393 for d in self.disks],
9396 hypervisor_name=self.op.hypervisor,
9402 def BuildHooksNodes(self):
9403 """Build hooks nodes.
9406 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
9409 def _ReadExportInfo(self):
9410 """Reads the export information from disk.
9412 It will override the opcode source node and path with the actual
9413 information, if these two were not specified before.
9415 @return: the export information
9418 assert self.op.mode == constants.INSTANCE_IMPORT
9420 src_node = self.op.src_node
9421 src_path = self.op.src_path
9423 if src_node is None:
9424 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
9425 exp_list = self.rpc.call_export_list(locked_nodes)
9427 for node in exp_list:
9428 if exp_list[node].fail_msg:
9430 if src_path in exp_list[node].payload:
9432 self.op.src_node = src_node = node
9433 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
9437 raise errors.OpPrereqError("No export found for relative path %s" %
9438 src_path, errors.ECODE_INVAL)
9440 _CheckNodeOnline(self, src_node)
9441 result = self.rpc.call_export_info(src_node, src_path)
9442 result.Raise("No export or invalid export found in dir %s" % src_path)
9444 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
9445 if not export_info.has_section(constants.INISECT_EXP):
9446 raise errors.ProgrammerError("Corrupted export config",
9447 errors.ECODE_ENVIRON)
9449 ei_version = export_info.get(constants.INISECT_EXP, "version")
9450 if (int(ei_version) != constants.EXPORT_VERSION):
9451 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
9452 (ei_version, constants.EXPORT_VERSION),
9453 errors.ECODE_ENVIRON)
9456 def _ReadExportParams(self, einfo):
9457 """Use export parameters as defaults.
9459 In case the opcode doesn't specify (as in override) some instance
9460 parameters, then try to use them from the export information, if
9464 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
9466 if self.op.disk_template is None:
9467 if einfo.has_option(constants.INISECT_INS, "disk_template"):
9468 self.op.disk_template = einfo.get(constants.INISECT_INS,
9470 if self.op.disk_template not in constants.DISK_TEMPLATES:
9471 raise errors.OpPrereqError("Disk template specified in configuration"
9472 " file is not one of the allowed values:"
9473 " %s" % " ".join(constants.DISK_TEMPLATES))
9475 raise errors.OpPrereqError("No disk template specified and the export"
9476 " is missing the disk_template information",
9479 if not self.op.disks:
9481 # TODO: import the disk iv_name too
9482 for idx in range(constants.MAX_DISKS):
9483 if einfo.has_option(constants.INISECT_INS, "disk%d_size" % idx):
9484 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
9485 disks.append({constants.IDISK_SIZE: disk_sz})
9486 self.op.disks = disks
9487 if not disks and self.op.disk_template != constants.DT_DISKLESS:
9488 raise errors.OpPrereqError("No disk info specified and the export"
9489 " is missing the disk information",
9492 if not self.op.nics:
9494 for idx in range(constants.MAX_NICS):
9495 if einfo.has_option(constants.INISECT_INS, "nic%d_mac" % idx):
9497 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
9498 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
9505 if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):
9506 self.op.tags = einfo.get(constants.INISECT_INS, "tags").split()
9508 if (self.op.hypervisor is None and
9509 einfo.has_option(constants.INISECT_INS, "hypervisor")):
9510 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
9512 if einfo.has_section(constants.INISECT_HYP):
9513 # use the export parameters but do not override the ones
9514 # specified by the user
9515 for name, value in einfo.items(constants.INISECT_HYP):
9516 if name not in self.op.hvparams:
9517 self.op.hvparams[name] = value
9519 if einfo.has_section(constants.INISECT_BEP):
9520 # use the parameters, without overriding
9521 for name, value in einfo.items(constants.INISECT_BEP):
9522 if name not in self.op.beparams:
9523 self.op.beparams[name] = value
9524 # Compatibility for the old "memory" be param
9525 if name == constants.BE_MEMORY:
9526 if constants.BE_MAXMEM not in self.op.beparams:
9527 self.op.beparams[constants.BE_MAXMEM] = value
9528 if constants.BE_MINMEM not in self.op.beparams:
9529 self.op.beparams[constants.BE_MINMEM] = value
9531 # try to read the parameters old style, from the main section
9532 for name in constants.BES_PARAMETERS:
9533 if (name not in self.op.beparams and
9534 einfo.has_option(constants.INISECT_INS, name)):
9535 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
9537 if einfo.has_section(constants.INISECT_OSP):
9538 # use the parameters, without overriding
9539 for name, value in einfo.items(constants.INISECT_OSP):
9540 if name not in self.op.osparams:
9541 self.op.osparams[name] = value
9543 def _RevertToDefaults(self, cluster):
9544 """Revert the instance parameters to the default values.
9548 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
9549 for name in self.op.hvparams.keys():
9550 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
9551 del self.op.hvparams[name]
9553 be_defs = cluster.SimpleFillBE({})
9554 for name in self.op.beparams.keys():
9555 if name in be_defs and be_defs[name] == self.op.beparams[name]:
9556 del self.op.beparams[name]
9558 nic_defs = cluster.SimpleFillNIC({})
9559 for nic in self.op.nics:
9560 for name in constants.NICS_PARAMETERS:
9561 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
9564 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
9565 for name in self.op.osparams.keys():
9566 if name in os_defs and os_defs[name] == self.op.osparams[name]:
9567 del self.op.osparams[name]
9569 def _CalculateFileStorageDir(self):
9570 """Calculate final instance file storage dir.
9573 # file storage dir calculation/check
9574 self.instance_file_storage_dir = None
9575 if self.op.disk_template in constants.DTS_FILEBASED:
9576 # build the full file storage dir path
9579 if self.op.disk_template == constants.DT_SHARED_FILE:
9580 get_fsd_fn = self.cfg.GetSharedFileStorageDir
9582 get_fsd_fn = self.cfg.GetFileStorageDir
9584 cfg_storagedir = get_fsd_fn()
9585 if not cfg_storagedir:
9586 raise errors.OpPrereqError("Cluster file storage dir not defined")
9587 joinargs.append(cfg_storagedir)
9589 if self.op.file_storage_dir is not None:
9590 joinargs.append(self.op.file_storage_dir)
9592 joinargs.append(self.op.instance_name)
9594 # pylint: disable=W0142
9595 self.instance_file_storage_dir = utils.PathJoin(*joinargs)
9597 def CheckPrereq(self): # pylint: disable=R0914
9598 """Check prerequisites.
9601 self._CalculateFileStorageDir()
9603 if self.op.mode == constants.INSTANCE_IMPORT:
9604 export_info = self._ReadExportInfo()
9605 self._ReadExportParams(export_info)
9607 if (not self.cfg.GetVGName() and
9608 self.op.disk_template not in constants.DTS_NOT_LVM):
9609 raise errors.OpPrereqError("Cluster does not support lvm-based"
9610 " instances", errors.ECODE_STATE)
9612 if (self.op.hypervisor is None or
9613 self.op.hypervisor == constants.VALUE_AUTO):
9614 self.op.hypervisor = self.cfg.GetHypervisorType()
9616 cluster = self.cfg.GetClusterInfo()
9617 enabled_hvs = cluster.enabled_hypervisors
9618 if self.op.hypervisor not in enabled_hvs:
9619 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
9620 " cluster (%s)" % (self.op.hypervisor,
9621 ",".join(enabled_hvs)),
9624 # Check tag validity
9625 for tag in self.op.tags:
9626 objects.TaggableObject.ValidateTag(tag)
9628 # check hypervisor parameter syntax (locally)
9629 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
9630 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
9632 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
9633 hv_type.CheckParameterSyntax(filled_hvp)
9634 self.hv_full = filled_hvp
9635 # check that we don't specify global parameters on an instance
9636 _CheckGlobalHvParams(self.op.hvparams)
9638 # fill and remember the beparams dict
9639 default_beparams = cluster.beparams[constants.PP_DEFAULT]
9640 for param, value in self.op.beparams.iteritems():
9641 if value == constants.VALUE_AUTO:
9642 self.op.beparams[param] = default_beparams[param]
9643 objects.UpgradeBeParams(self.op.beparams)
9644 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
9645 self.be_full = cluster.SimpleFillBE(self.op.beparams)
9647 # build os parameters
9648 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
9650 # now that hvp/bep are in final format, let's reset to defaults,
9652 if self.op.identify_defaults:
9653 self._RevertToDefaults(cluster)
9657 for idx, nic in enumerate(self.op.nics):
9658 nic_mode_req = nic.get(constants.INIC_MODE, None)
9659 nic_mode = nic_mode_req
9660 if nic_mode is None or nic_mode == constants.VALUE_AUTO:
9661 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
9663 # in routed mode, for the first nic, the default ip is 'auto'
9664 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
9665 default_ip_mode = constants.VALUE_AUTO
9667 default_ip_mode = constants.VALUE_NONE
9669 # ip validity checks
9670 ip = nic.get(constants.INIC_IP, default_ip_mode)
9671 if ip is None or ip.lower() == constants.VALUE_NONE:
9673 elif ip.lower() == constants.VALUE_AUTO:
9674 if not self.op.name_check:
9675 raise errors.OpPrereqError("IP address set to auto but name checks"
9676 " have been skipped",
9678 nic_ip = self.hostname1.ip
9680 if not netutils.IPAddress.IsValid(ip):
9681 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
9685 # TODO: check the ip address for uniqueness
9686 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
9687 raise errors.OpPrereqError("Routed nic mode requires an ip address",
9690 # MAC address verification
9691 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
9692 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9693 mac = utils.NormalizeAndValidateMac(mac)
9696 self.cfg.ReserveMAC(mac, self.proc.GetECId())
9697 except errors.ReservationError:
9698 raise errors.OpPrereqError("MAC address %s already in use"
9699 " in cluster" % mac,
9700 errors.ECODE_NOTUNIQUE)
9702 # Build nic parameters
9703 link = nic.get(constants.INIC_LINK, None)
9704 if link == constants.VALUE_AUTO:
9705 link = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_LINK]
9708 nicparams[constants.NIC_MODE] = nic_mode
9710 nicparams[constants.NIC_LINK] = link
9712 check_params = cluster.SimpleFillNIC(nicparams)
9713 objects.NIC.CheckParameterSyntax(check_params)
9714 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
9716 # disk checks/pre-build
9717 default_vg = self.cfg.GetVGName()
9719 for disk in self.op.disks:
9720 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
9721 if mode not in constants.DISK_ACCESS_SET:
9722 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
9723 mode, errors.ECODE_INVAL)
9724 size = disk.get(constants.IDISK_SIZE, None)
9726 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
9729 except (TypeError, ValueError):
9730 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
9733 data_vg = disk.get(constants.IDISK_VG, default_vg)
9735 constants.IDISK_SIZE: size,
9736 constants.IDISK_MODE: mode,
9737 constants.IDISK_VG: data_vg,
9739 if constants.IDISK_METAVG in disk:
9740 new_disk[constants.IDISK_METAVG] = disk[constants.IDISK_METAVG]
9741 if constants.IDISK_ADOPT in disk:
9742 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
9743 self.disks.append(new_disk)
9745 if self.op.mode == constants.INSTANCE_IMPORT:
9747 for idx in range(len(self.disks)):
9748 option = "disk%d_dump" % idx
9749 if export_info.has_option(constants.INISECT_INS, option):
9750 # FIXME: are the old os-es, disk sizes, etc. useful?
9751 export_name = export_info.get(constants.INISECT_INS, option)
9752 image = utils.PathJoin(self.op.src_path, export_name)
9753 disk_images.append(image)
9755 disk_images.append(False)
9757 self.src_images = disk_images
9759 old_name = export_info.get(constants.INISECT_INS, "name")
9760 if self.op.instance_name == old_name:
9761 for idx, nic in enumerate(self.nics):
9762 if nic.mac == constants.VALUE_AUTO:
9763 nic_mac_ini = "nic%d_mac" % idx
9764 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
9766 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
9768 # ip ping checks (we use the same ip that was resolved in ExpandNames)
9769 if self.op.ip_check:
9770 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
9771 raise errors.OpPrereqError("IP %s of instance %s already in use" %
9772 (self.check_ip, self.op.instance_name),
9773 errors.ECODE_NOTUNIQUE)
9775 #### mac address generation
9776 # By generating here the mac address both the allocator and the hooks get
9777 # the real final mac address rather than the 'auto' or 'generate' value.
9778 # There is a race condition between the generation and the instance object
9779 # creation, which means that we know the mac is valid now, but we're not
9780 # sure it will be when we actually add the instance. If things go bad
9781 # adding the instance will abort because of a duplicate mac, and the
9782 # creation job will fail.
9783 for nic in self.nics:
9784 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9785 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
9789 if self.op.iallocator is not None:
9790 self._RunAllocator()
9792 # Release all unneeded node locks
9793 _ReleaseLocks(self, locking.LEVEL_NODE,
9794 keep=filter(None, [self.op.pnode, self.op.snode,
9796 _ReleaseLocks(self, locking.LEVEL_NODE_RES,
9797 keep=filter(None, [self.op.pnode, self.op.snode,
9800 #### node related checks
9802 # check primary node
9803 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
9804 assert self.pnode is not None, \
9805 "Cannot retrieve locked node %s" % self.op.pnode
9807 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
9808 pnode.name, errors.ECODE_STATE)
9810 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
9811 pnode.name, errors.ECODE_STATE)
9812 if not pnode.vm_capable:
9813 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
9814 " '%s'" % pnode.name, errors.ECODE_STATE)
9816 self.secondaries = []
9818 # mirror node verification
9819 if self.op.disk_template in constants.DTS_INT_MIRROR:
9820 if self.op.snode == pnode.name:
9821 raise errors.OpPrereqError("The secondary node cannot be the"
9822 " primary node", errors.ECODE_INVAL)
9823 _CheckNodeOnline(self, self.op.snode)
9824 _CheckNodeNotDrained(self, self.op.snode)
9825 _CheckNodeVmCapable(self, self.op.snode)
9826 self.secondaries.append(self.op.snode)
9828 snode = self.cfg.GetNodeInfo(self.op.snode)
9829 if pnode.group != snode.group:
9830 self.LogWarning("The primary and secondary nodes are in two"
9831 " different node groups; the disk parameters"
9832 " from the first disk's node group will be"
9835 nodenames = [pnode.name] + self.secondaries
9837 # Verify instance specs
9838 spindle_use = self.be_full.get(constants.BE_SPINDLE_USE, None)
9840 constants.ISPEC_MEM_SIZE: self.be_full.get(constants.BE_MAXMEM, None),
9841 constants.ISPEC_CPU_COUNT: self.be_full.get(constants.BE_VCPUS, None),
9842 constants.ISPEC_DISK_COUNT: len(self.disks),
9843 constants.ISPEC_DISK_SIZE: [disk["size"] for disk in self.disks],
9844 constants.ISPEC_NIC_COUNT: len(self.nics),
9845 constants.ISPEC_SPINDLE_USE: spindle_use,
9848 group_info = self.cfg.GetNodeGroup(pnode.group)
9849 ipolicy = _CalculateGroupIPolicy(cluster, group_info)
9850 res = _ComputeIPolicyInstanceSpecViolation(ipolicy, ispec)
9851 if not self.op.ignore_ipolicy and res:
9852 raise errors.OpPrereqError(("Instance allocation to group %s violates"
9853 " policy: %s") % (pnode.group,
9854 utils.CommaJoin(res)),
9857 # disk parameters (not customizable at instance or node level)
9858 # just use the primary node parameters, ignoring the secondary.
9859 self.diskparams = group_info.diskparams
9861 if not self.adopt_disks:
9862 if self.op.disk_template == constants.DT_RBD:
9863 # _CheckRADOSFreeSpace() is just a placeholder.
9864 # Any function that checks prerequisites can be placed here.
9865 # Check if there is enough space on the RADOS cluster.
9866 _CheckRADOSFreeSpace()
9868 # Check lv size requirements, if not adopting
9869 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
9870 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
9872 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
9873 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
9874 disk[constants.IDISK_ADOPT])
9875 for disk in self.disks])
9876 if len(all_lvs) != len(self.disks):
9877 raise errors.OpPrereqError("Duplicate volume names given for adoption",
9879 for lv_name in all_lvs:
9881 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
9882 # to ReserveLV uses the same syntax
9883 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
9884 except errors.ReservationError:
9885 raise errors.OpPrereqError("LV named %s used by another instance" %
9886 lv_name, errors.ECODE_NOTUNIQUE)
9888 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
9889 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
9891 node_lvs = self.rpc.call_lv_list([pnode.name],
9892 vg_names.payload.keys())[pnode.name]
9893 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
9894 node_lvs = node_lvs.payload
9896 delta = all_lvs.difference(node_lvs.keys())
9898 raise errors.OpPrereqError("Missing logical volume(s): %s" %
9899 utils.CommaJoin(delta),
9901 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
9903 raise errors.OpPrereqError("Online logical volumes found, cannot"
9904 " adopt: %s" % utils.CommaJoin(online_lvs),
9906 # update the size of disk based on what is found
9907 for dsk in self.disks:
9908 dsk[constants.IDISK_SIZE] = \
9909 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
9910 dsk[constants.IDISK_ADOPT])][0]))
9912 elif self.op.disk_template == constants.DT_BLOCK:
9913 # Normalize and de-duplicate device paths
9914 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
9915 for disk in self.disks])
9916 if len(all_disks) != len(self.disks):
9917 raise errors.OpPrereqError("Duplicate disk names given for adoption",
9919 baddisks = [d for d in all_disks
9920 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
9922 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
9923 " cannot be adopted" %
9924 (", ".join(baddisks),
9925 constants.ADOPTABLE_BLOCKDEV_ROOT),
9928 node_disks = self.rpc.call_bdev_sizes([pnode.name],
9929 list(all_disks))[pnode.name]
9930 node_disks.Raise("Cannot get block device information from node %s" %
9932 node_disks = node_disks.payload
9933 delta = all_disks.difference(node_disks.keys())
9935 raise errors.OpPrereqError("Missing block device(s): %s" %
9936 utils.CommaJoin(delta),
9938 for dsk in self.disks:
9939 dsk[constants.IDISK_SIZE] = \
9940 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
9942 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
9944 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
9945 # check OS parameters (remotely)
9946 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
9948 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
9950 # memory check on primary node
9951 #TODO(dynmem): use MINMEM for checking
9953 _CheckNodeFreeMemory(self, self.pnode.name,
9954 "creating instance %s" % self.op.instance_name,
9955 self.be_full[constants.BE_MAXMEM],
9958 self.dry_run_result = list(nodenames)
9960 def Exec(self, feedback_fn):
9961 """Create and add the instance to the cluster.
9964 instance = self.op.instance_name
9965 pnode_name = self.pnode.name
9967 assert not (self.owned_locks(locking.LEVEL_NODE_RES) -
9968 self.owned_locks(locking.LEVEL_NODE)), \
9969 "Node locks differ from node resource locks"
9971 ht_kind = self.op.hypervisor
9972 if ht_kind in constants.HTS_REQ_PORT:
9973 network_port = self.cfg.AllocatePort()
9977 disks = _GenerateDiskTemplate(self,
9978 self.op.disk_template,
9979 instance, pnode_name,
9982 self.instance_file_storage_dir,
9983 self.op.file_driver,
9988 iobj = objects.Instance(name=instance, os=self.op.os_type,
9989 primary_node=pnode_name,
9990 nics=self.nics, disks=disks,
9991 disk_template=self.op.disk_template,
9992 admin_state=constants.ADMINST_DOWN,
9993 network_port=network_port,
9994 beparams=self.op.beparams,
9995 hvparams=self.op.hvparams,
9996 hypervisor=self.op.hypervisor,
9997 osparams=self.op.osparams,
10001 for tag in self.op.tags:
10004 if self.adopt_disks:
10005 if self.op.disk_template == constants.DT_PLAIN:
10006 # rename LVs to the newly-generated names; we need to construct
10007 # 'fake' LV disks with the old data, plus the new unique_id
10008 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
10010 for t_dsk, a_dsk in zip(tmp_disks, self.disks):
10011 rename_to.append(t_dsk.logical_id)
10012 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
10013 self.cfg.SetDiskID(t_dsk, pnode_name)
10014 result = self.rpc.call_blockdev_rename(pnode_name,
10015 zip(tmp_disks, rename_to))
10016 result.Raise("Failed to rename adoped LVs")
10018 feedback_fn("* creating instance disks...")
10020 _CreateDisks(self, iobj)
10021 except errors.OpExecError:
10022 self.LogWarning("Device creation failed, reverting...")
10024 _RemoveDisks(self, iobj)
10026 self.cfg.ReleaseDRBDMinors(instance)
10029 feedback_fn("adding instance %s to cluster config" % instance)
10031 self.cfg.AddInstance(iobj, self.proc.GetECId())
10033 # Declare that we don't want to remove the instance lock anymore, as we've
10034 # added the instance to the config
10035 del self.remove_locks[locking.LEVEL_INSTANCE]
10037 if self.op.mode == constants.INSTANCE_IMPORT:
10038 # Release unused nodes
10039 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
10041 # Release all nodes
10042 _ReleaseLocks(self, locking.LEVEL_NODE)
10045 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
10046 feedback_fn("* wiping instance disks...")
10048 _WipeDisks(self, iobj)
10049 except errors.OpExecError, err:
10050 logging.exception("Wiping disks failed")
10051 self.LogWarning("Wiping instance disks failed (%s)", err)
10055 # Something is already wrong with the disks, don't do anything else
10057 elif self.op.wait_for_sync:
10058 disk_abort = not _WaitForSync(self, iobj)
10059 elif iobj.disk_template in constants.DTS_INT_MIRROR:
10060 # make sure the disks are not degraded (still sync-ing is ok)
10061 feedback_fn("* checking mirrors status")
10062 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
10067 _RemoveDisks(self, iobj)
10068 self.cfg.RemoveInstance(iobj.name)
10069 # Make sure the instance lock gets removed
10070 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
10071 raise errors.OpExecError("There are some degraded disks for"
10074 # Release all node resource locks
10075 _ReleaseLocks(self, locking.LEVEL_NODE_RES)
10077 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
10078 if self.op.mode == constants.INSTANCE_CREATE:
10079 if not self.op.no_install:
10080 pause_sync = (iobj.disk_template in constants.DTS_INT_MIRROR and
10081 not self.op.wait_for_sync)
10083 feedback_fn("* pausing disk sync to install instance OS")
10084 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
10087 for idx, success in enumerate(result.payload):
10089 logging.warn("pause-sync of instance %s for disk %d failed",
10092 feedback_fn("* running the instance OS create scripts...")
10093 # FIXME: pass debug option from opcode to backend
10095 self.rpc.call_instance_os_add(pnode_name, (iobj, None), False,
10096 self.op.debug_level)
10098 feedback_fn("* resuming disk sync")
10099 result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
10102 for idx, success in enumerate(result.payload):
10104 logging.warn("resume-sync of instance %s for disk %d failed",
10107 os_add_result.Raise("Could not add os for instance %s"
10108 " on node %s" % (instance, pnode_name))
10110 elif self.op.mode == constants.INSTANCE_IMPORT:
10111 feedback_fn("* running the instance OS import scripts...")
10115 for idx, image in enumerate(self.src_images):
10119 # FIXME: pass debug option from opcode to backend
10120 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
10121 constants.IEIO_FILE, (image, ),
10122 constants.IEIO_SCRIPT,
10123 (iobj.disks[idx], idx),
10125 transfers.append(dt)
10128 masterd.instance.TransferInstanceData(self, feedback_fn,
10129 self.op.src_node, pnode_name,
10130 self.pnode.secondary_ip,
10132 if not compat.all(import_result):
10133 self.LogWarning("Some disks for instance %s on node %s were not"
10134 " imported successfully" % (instance, pnode_name))
10136 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
10137 feedback_fn("* preparing remote import...")
10138 # The source cluster will stop the instance before attempting to make a
10139 # connection. In some cases stopping an instance can take a long time,
10140 # hence the shutdown timeout is added to the connection timeout.
10141 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
10142 self.op.source_shutdown_timeout)
10143 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
10145 assert iobj.primary_node == self.pnode.name
10147 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
10148 self.source_x509_ca,
10149 self._cds, timeouts)
10150 if not compat.all(disk_results):
10151 # TODO: Should the instance still be started, even if some disks
10152 # failed to import (valid for local imports, too)?
10153 self.LogWarning("Some disks for instance %s on node %s were not"
10154 " imported successfully" % (instance, pnode_name))
10156 # Run rename script on newly imported instance
10157 assert iobj.name == instance
10158 feedback_fn("Running rename script for %s" % instance)
10159 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
10160 self.source_instance_name,
10161 self.op.debug_level)
10162 if result.fail_msg:
10163 self.LogWarning("Failed to run rename script for %s on node"
10164 " %s: %s" % (instance, pnode_name, result.fail_msg))
10167 # also checked in the prereq part
10168 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
10171 assert not self.owned_locks(locking.LEVEL_NODE_RES)
10174 iobj.admin_state = constants.ADMINST_UP
10175 self.cfg.Update(iobj, feedback_fn)
10176 logging.info("Starting instance %s on node %s", instance, pnode_name)
10177 feedback_fn("* starting instance...")
10178 result = self.rpc.call_instance_start(pnode_name, (iobj, None, None),
10180 result.Raise("Could not start instance")
10182 return list(iobj.all_nodes)
10185 def _CheckRADOSFreeSpace():
10186 """Compute disk size requirements inside the RADOS cluster.
10189 # For the RADOS cluster we assume there is always enough space.
10193 class LUInstanceConsole(NoHooksLU):
10194 """Connect to an instance's console.
10196 This is somewhat special in that it returns the command line that
10197 you need to run on the master node in order to connect to the
10203 def ExpandNames(self):
10204 self.share_locks = _ShareAll()
10205 self._ExpandAndLockInstance()
10207 def CheckPrereq(self):
10208 """Check prerequisites.
10210 This checks that the instance is in the cluster.
10213 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
10214 assert self.instance is not None, \
10215 "Cannot retrieve locked instance %s" % self.op.instance_name
10216 _CheckNodeOnline(self, self.instance.primary_node)
10218 def Exec(self, feedback_fn):
10219 """Connect to the console of an instance
10222 instance = self.instance
10223 node = instance.primary_node
10225 node_insts = self.rpc.call_instance_list([node],
10226 [instance.hypervisor])[node]
10227 node_insts.Raise("Can't get node information from %s" % node)
10229 if instance.name not in node_insts.payload:
10230 if instance.admin_state == constants.ADMINST_UP:
10231 state = constants.INSTST_ERRORDOWN
10232 elif instance.admin_state == constants.ADMINST_DOWN:
10233 state = constants.INSTST_ADMINDOWN
10235 state = constants.INSTST_ADMINOFFLINE
10236 raise errors.OpExecError("Instance %s is not running (state %s)" %
10237 (instance.name, state))
10239 logging.debug("Connecting to console of %s on %s", instance.name, node)
10241 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
10244 def _GetInstanceConsole(cluster, instance):
10245 """Returns console information for an instance.
10247 @type cluster: L{objects.Cluster}
10248 @type instance: L{objects.Instance}
10252 hyper = hypervisor.GetHypervisor(instance.hypervisor)
10253 # beparams and hvparams are passed separately, to avoid editing the
10254 # instance and then saving the defaults in the instance itself.
10255 hvparams = cluster.FillHV(instance)
10256 beparams = cluster.FillBE(instance)
10257 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
10259 assert console.instance == instance.name
10260 assert console.Validate()
10262 return console.ToDict()
10265 class LUInstanceReplaceDisks(LogicalUnit):
10266 """Replace the disks of an instance.
10269 HPATH = "mirrors-replace"
10270 HTYPE = constants.HTYPE_INSTANCE
10273 def CheckArguments(self):
10274 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
10275 self.op.iallocator)
10277 def ExpandNames(self):
10278 self._ExpandAndLockInstance()
10280 assert locking.LEVEL_NODE not in self.needed_locks
10281 assert locking.LEVEL_NODE_RES not in self.needed_locks
10282 assert locking.LEVEL_NODEGROUP not in self.needed_locks
10284 assert self.op.iallocator is None or self.op.remote_node is None, \
10285 "Conflicting options"
10287 if self.op.remote_node is not None:
10288 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
10290 # Warning: do not remove the locking of the new secondary here
10291 # unless DRBD8.AddChildren is changed to work in parallel;
10292 # currently it doesn't since parallel invocations of
10293 # FindUnusedMinor will conflict
10294 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
10295 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
10297 self.needed_locks[locking.LEVEL_NODE] = []
10298 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
10300 if self.op.iallocator is not None:
10301 # iallocator will select a new node in the same group
10302 self.needed_locks[locking.LEVEL_NODEGROUP] = []
10304 self.needed_locks[locking.LEVEL_NODE_RES] = []
10306 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
10307 self.op.iallocator, self.op.remote_node,
10308 self.op.disks, False, self.op.early_release,
10309 self.op.ignore_ipolicy)
10311 self.tasklets = [self.replacer]
10313 def DeclareLocks(self, level):
10314 if level == locking.LEVEL_NODEGROUP:
10315 assert self.op.remote_node is None
10316 assert self.op.iallocator is not None
10317 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
10319 self.share_locks[locking.LEVEL_NODEGROUP] = 1
10320 # Lock all groups used by instance optimistically; this requires going
10321 # via the node before it's locked, requiring verification later on
10322 self.needed_locks[locking.LEVEL_NODEGROUP] = \
10323 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
10325 elif level == locking.LEVEL_NODE:
10326 if self.op.iallocator is not None:
10327 assert self.op.remote_node is None
10328 assert not self.needed_locks[locking.LEVEL_NODE]
10330 # Lock member nodes of all locked groups
10331 self.needed_locks[locking.LEVEL_NODE] = [node_name
10332 for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
10333 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
10335 self._LockInstancesNodes()
10336 elif level == locking.LEVEL_NODE_RES:
10338 self.needed_locks[locking.LEVEL_NODE_RES] = \
10339 self.needed_locks[locking.LEVEL_NODE]
10341 def BuildHooksEnv(self):
10342 """Build hooks env.
10344 This runs on the master, the primary and all the secondaries.
10347 instance = self.replacer.instance
10349 "MODE": self.op.mode,
10350 "NEW_SECONDARY": self.op.remote_node,
10351 "OLD_SECONDARY": instance.secondary_nodes[0],
10353 env.update(_BuildInstanceHookEnvByObject(self, instance))
10356 def BuildHooksNodes(self):
10357 """Build hooks nodes.
10360 instance = self.replacer.instance
10362 self.cfg.GetMasterNode(),
10363 instance.primary_node,
10365 if self.op.remote_node is not None:
10366 nl.append(self.op.remote_node)
10369 def CheckPrereq(self):
10370 """Check prerequisites.
10373 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
10374 self.op.iallocator is None)
10376 # Verify if node group locks are still correct
10377 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
10379 _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
10381 return LogicalUnit.CheckPrereq(self)
10384 class TLReplaceDisks(Tasklet):
10385 """Replaces disks for an instance.
10387 Note: Locking is not within the scope of this class.
10390 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
10391 disks, delay_iallocator, early_release, ignore_ipolicy):
10392 """Initializes this class.
10395 Tasklet.__init__(self, lu)
10398 self.instance_name = instance_name
10400 self.iallocator_name = iallocator_name
10401 self.remote_node = remote_node
10403 self.delay_iallocator = delay_iallocator
10404 self.early_release = early_release
10405 self.ignore_ipolicy = ignore_ipolicy
10408 self.instance = None
10409 self.new_node = None
10410 self.target_node = None
10411 self.other_node = None
10412 self.remote_node_info = None
10413 self.node_secondary_ip = None
10416 def CheckArguments(mode, remote_node, iallocator):
10417 """Helper function for users of this class.
10420 # check for valid parameter combination
10421 if mode == constants.REPLACE_DISK_CHG:
10422 if remote_node is None and iallocator is None:
10423 raise errors.OpPrereqError("When changing the secondary either an"
10424 " iallocator script must be used or the"
10425 " new node given", errors.ECODE_INVAL)
10427 if remote_node is not None and iallocator is not None:
10428 raise errors.OpPrereqError("Give either the iallocator or the new"
10429 " secondary, not both", errors.ECODE_INVAL)
10431 elif remote_node is not None or iallocator is not None:
10432 # Not replacing the secondary
10433 raise errors.OpPrereqError("The iallocator and new node options can"
10434 " only be used when changing the"
10435 " secondary node", errors.ECODE_INVAL)
10438 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
10439 """Compute a new secondary node using an IAllocator.
10442 ial = IAllocator(lu.cfg, lu.rpc,
10443 mode=constants.IALLOCATOR_MODE_RELOC,
10444 name=instance_name,
10445 relocate_from=list(relocate_from))
10447 ial.Run(iallocator_name)
10449 if not ial.success:
10450 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
10451 " %s" % (iallocator_name, ial.info),
10452 errors.ECODE_NORES)
10454 if len(ial.result) != ial.required_nodes:
10455 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
10456 " of nodes (%s), required %s" %
10458 len(ial.result), ial.required_nodes),
10459 errors.ECODE_FAULT)
10461 remote_node_name = ial.result[0]
10463 lu.LogInfo("Selected new secondary for instance '%s': %s",
10464 instance_name, remote_node_name)
10466 return remote_node_name
10468 def _FindFaultyDisks(self, node_name):
10469 """Wrapper for L{_FindFaultyInstanceDisks}.
10472 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
10475 def _CheckDisksActivated(self, instance):
10476 """Checks if the instance disks are activated.
10478 @param instance: The instance to check disks
10479 @return: True if they are activated, False otherwise
10482 nodes = instance.all_nodes
10484 for idx, dev in enumerate(instance.disks):
10486 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
10487 self.cfg.SetDiskID(dev, node)
10489 result = self.rpc.call_blockdev_find(node, dev)
10493 elif result.fail_msg or not result.payload:
10498 def CheckPrereq(self):
10499 """Check prerequisites.
10501 This checks that the instance is in the cluster.
10504 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
10505 assert instance is not None, \
10506 "Cannot retrieve locked instance %s" % self.instance_name
10508 if instance.disk_template != constants.DT_DRBD8:
10509 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
10510 " instances", errors.ECODE_INVAL)
10512 if len(instance.secondary_nodes) != 1:
10513 raise errors.OpPrereqError("The instance has a strange layout,"
10514 " expected one secondary but found %d" %
10515 len(instance.secondary_nodes),
10516 errors.ECODE_FAULT)
10518 if not self.delay_iallocator:
10519 self._CheckPrereq2()
10521 def _CheckPrereq2(self):
10522 """Check prerequisites, second part.
10524 This function should always be part of CheckPrereq. It was separated and is
10525 now called from Exec because during node evacuation iallocator was only
10526 called with an unmodified cluster model, not taking planned changes into
10530 instance = self.instance
10531 secondary_node = instance.secondary_nodes[0]
10533 if self.iallocator_name is None:
10534 remote_node = self.remote_node
10536 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
10537 instance.name, instance.secondary_nodes)
10539 if remote_node is None:
10540 self.remote_node_info = None
10542 assert remote_node in self.lu.owned_locks(locking.LEVEL_NODE), \
10543 "Remote node '%s' is not locked" % remote_node
10545 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
10546 assert self.remote_node_info is not None, \
10547 "Cannot retrieve locked node %s" % remote_node
10549 if remote_node == self.instance.primary_node:
10550 raise errors.OpPrereqError("The specified node is the primary node of"
10551 " the instance", errors.ECODE_INVAL)
10553 if remote_node == secondary_node:
10554 raise errors.OpPrereqError("The specified node is already the"
10555 " secondary node of the instance",
10556 errors.ECODE_INVAL)
10558 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
10559 constants.REPLACE_DISK_CHG):
10560 raise errors.OpPrereqError("Cannot specify disks to be replaced",
10561 errors.ECODE_INVAL)
10563 if self.mode == constants.REPLACE_DISK_AUTO:
10564 if not self._CheckDisksActivated(instance):
10565 raise errors.OpPrereqError("Please run activate-disks on instance %s"
10566 " first" % self.instance_name,
10567 errors.ECODE_STATE)
10568 faulty_primary = self._FindFaultyDisks(instance.primary_node)
10569 faulty_secondary = self._FindFaultyDisks(secondary_node)
10571 if faulty_primary and faulty_secondary:
10572 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
10573 " one node and can not be repaired"
10574 " automatically" % self.instance_name,
10575 errors.ECODE_STATE)
10578 self.disks = faulty_primary
10579 self.target_node = instance.primary_node
10580 self.other_node = secondary_node
10581 check_nodes = [self.target_node, self.other_node]
10582 elif faulty_secondary:
10583 self.disks = faulty_secondary
10584 self.target_node = secondary_node
10585 self.other_node = instance.primary_node
10586 check_nodes = [self.target_node, self.other_node]
10592 # Non-automatic modes
10593 if self.mode == constants.REPLACE_DISK_PRI:
10594 self.target_node = instance.primary_node
10595 self.other_node = secondary_node
10596 check_nodes = [self.target_node, self.other_node]
10598 elif self.mode == constants.REPLACE_DISK_SEC:
10599 self.target_node = secondary_node
10600 self.other_node = instance.primary_node
10601 check_nodes = [self.target_node, self.other_node]
10603 elif self.mode == constants.REPLACE_DISK_CHG:
10604 self.new_node = remote_node
10605 self.other_node = instance.primary_node
10606 self.target_node = secondary_node
10607 check_nodes = [self.new_node, self.other_node]
10609 _CheckNodeNotDrained(self.lu, remote_node)
10610 _CheckNodeVmCapable(self.lu, remote_node)
10612 old_node_info = self.cfg.GetNodeInfo(secondary_node)
10613 assert old_node_info is not None
10614 if old_node_info.offline and not self.early_release:
10615 # doesn't make sense to delay the release
10616 self.early_release = True
10617 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
10618 " early-release mode", secondary_node)
10621 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
10624 # If not specified all disks should be replaced
10626 self.disks = range(len(self.instance.disks))
10628 # TODO: This is ugly, but right now we can't distinguish between internal
10629 # submitted opcode and external one. We should fix that.
10630 if self.remote_node_info:
10631 # We change the node, lets verify it still meets instance policy
10632 new_group_info = self.cfg.GetNodeGroup(self.remote_node_info.group)
10633 ipolicy = _CalculateGroupIPolicy(self.cfg.GetClusterInfo(),
10635 _CheckTargetNodeIPolicy(self, ipolicy, instance, self.remote_node_info,
10636 ignore=self.ignore_ipolicy)
10638 # TODO: compute disk parameters
10639 primary_node_info = self.cfg.GetNodeInfo(instance.primary_node)
10640 secondary_node_info = self.cfg.GetNodeInfo(secondary_node)
10641 if primary_node_info.group != secondary_node_info.group:
10642 self.lu.LogInfo("The instance primary and secondary nodes are in two"
10643 " different node groups; the disk parameters of the"
10644 " primary node's group will be applied.")
10646 self.diskparams = self.cfg.GetNodeGroup(primary_node_info.group).diskparams
10648 for node in check_nodes:
10649 _CheckNodeOnline(self.lu, node)
10651 touched_nodes = frozenset(node_name for node_name in [self.new_node,
10654 if node_name is not None)
10656 # Release unneeded node and node resource locks
10657 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
10658 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES, keep=touched_nodes)
10660 # Release any owned node group
10661 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
10662 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
10664 # Check whether disks are valid
10665 for disk_idx in self.disks:
10666 instance.FindDisk(disk_idx)
10668 # Get secondary node IP addresses
10669 self.node_secondary_ip = dict((name, node.secondary_ip) for (name, node)
10670 in self.cfg.GetMultiNodeInfo(touched_nodes))
10672 def Exec(self, feedback_fn):
10673 """Execute disk replacement.
10675 This dispatches the disk replacement to the appropriate handler.
10678 if self.delay_iallocator:
10679 self._CheckPrereq2()
10682 # Verify owned locks before starting operation
10683 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
10684 assert set(owned_nodes) == set(self.node_secondary_ip), \
10685 ("Incorrect node locks, owning %s, expected %s" %
10686 (owned_nodes, self.node_secondary_ip.keys()))
10687 assert (self.lu.owned_locks(locking.LEVEL_NODE) ==
10688 self.lu.owned_locks(locking.LEVEL_NODE_RES))
10690 owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
10691 assert list(owned_instances) == [self.instance_name], \
10692 "Instance '%s' not locked" % self.instance_name
10694 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
10695 "Should not own any node group lock at this point"
10698 feedback_fn("No disks need replacement")
10701 feedback_fn("Replacing disk(s) %s for %s" %
10702 (utils.CommaJoin(self.disks), self.instance.name))
10704 activate_disks = (self.instance.admin_state != constants.ADMINST_UP)
10706 # Activate the instance disks if we're replacing them on a down instance
10708 _StartInstanceDisks(self.lu, self.instance, True)
10711 # Should we replace the secondary node?
10712 if self.new_node is not None:
10713 fn = self._ExecDrbd8Secondary
10715 fn = self._ExecDrbd8DiskOnly
10717 result = fn(feedback_fn)
10719 # Deactivate the instance disks if we're replacing them on a
10722 _SafeShutdownInstanceDisks(self.lu, self.instance)
10724 assert not self.lu.owned_locks(locking.LEVEL_NODE)
10727 # Verify owned locks
10728 owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE_RES)
10729 nodes = frozenset(self.node_secondary_ip)
10730 assert ((self.early_release and not owned_nodes) or
10731 (not self.early_release and not (set(owned_nodes) - nodes))), \
10732 ("Not owning the correct locks, early_release=%s, owned=%r,"
10733 " nodes=%r" % (self.early_release, owned_nodes, nodes))
10737 def _CheckVolumeGroup(self, nodes):
10738 self.lu.LogInfo("Checking volume groups")
10740 vgname = self.cfg.GetVGName()
10742 # Make sure volume group exists on all involved nodes
10743 results = self.rpc.call_vg_list(nodes)
10745 raise errors.OpExecError("Can't list volume groups on the nodes")
10748 res = results[node]
10749 res.Raise("Error checking node %s" % node)
10750 if vgname not in res.payload:
10751 raise errors.OpExecError("Volume group '%s' not found on node %s" %
10754 def _CheckDisksExistence(self, nodes):
10755 # Check disk existence
10756 for idx, dev in enumerate(self.instance.disks):
10757 if idx not in self.disks:
10761 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
10762 self.cfg.SetDiskID(dev, node)
10764 result = self.rpc.call_blockdev_find(node, dev)
10766 msg = result.fail_msg
10767 if msg or not result.payload:
10769 msg = "disk not found"
10770 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
10773 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
10774 for idx, dev in enumerate(self.instance.disks):
10775 if idx not in self.disks:
10778 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
10781 if not _CheckDiskConsistency(self.lu, self.instance, dev, node_name,
10782 on_primary, ldisk=ldisk):
10783 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
10784 " replace disks for instance %s" %
10785 (node_name, self.instance.name))
10787 def _CreateNewStorage(self, node_name):
10788 """Create new storage on the primary or secondary node.
10790 This is only used for same-node replaces, not for changing the
10791 secondary node, hence we don't want to modify the existing disk.
10796 for idx, dev in enumerate(self.instance.disks):
10797 if idx not in self.disks:
10800 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
10802 self.cfg.SetDiskID(dev, node_name)
10804 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
10805 names = _GenerateUniqueNames(self.lu, lv_names)
10807 _, data_p, meta_p = _ComputeLDParams(constants.DT_DRBD8, self.diskparams)
10809 vg_data = dev.children[0].logical_id[0]
10810 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
10811 logical_id=(vg_data, names[0]), params=data_p)
10812 vg_meta = dev.children[1].logical_id[0]
10813 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,
10814 logical_id=(vg_meta, names[1]), params=meta_p)
10816 new_lvs = [lv_data, lv_meta]
10817 old_lvs = [child.Copy() for child in dev.children]
10818 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
10820 # we pass force_create=True to force the LVM creation
10821 for new_lv in new_lvs:
10822 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
10823 _GetInstanceInfoText(self.instance), False)
10827 def _CheckDevices(self, node_name, iv_names):
10828 for name, (dev, _, _) in iv_names.iteritems():
10829 self.cfg.SetDiskID(dev, node_name)
10831 result = self.rpc.call_blockdev_find(node_name, dev)
10833 msg = result.fail_msg
10834 if msg or not result.payload:
10836 msg = "disk not found"
10837 raise errors.OpExecError("Can't find DRBD device %s: %s" %
10840 if result.payload.is_degraded:
10841 raise errors.OpExecError("DRBD device %s is degraded!" % name)
10843 def _RemoveOldStorage(self, node_name, iv_names):
10844 for name, (_, old_lvs, _) in iv_names.iteritems():
10845 self.lu.LogInfo("Remove logical volumes for %s" % name)
10848 self.cfg.SetDiskID(lv, node_name)
10850 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
10852 self.lu.LogWarning("Can't remove old LV: %s" % msg,
10853 hint="remove unused LVs manually")
10855 def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
10856 """Replace a disk on the primary or secondary for DRBD 8.
10858 The algorithm for replace is quite complicated:
10860 1. for each disk to be replaced:
10862 1. create new LVs on the target node with unique names
10863 1. detach old LVs from the drbd device
10864 1. rename old LVs to name_replaced.<time_t>
10865 1. rename new LVs to old LVs
10866 1. attach the new LVs (with the old names now) to the drbd device
10868 1. wait for sync across all devices
10870 1. for each modified disk:
10872 1. remove old LVs (which have the name name_replaces.<time_t>)
10874 Failures are not very well handled.
10879 # Step: check device activation
10880 self.lu.LogStep(1, steps_total, "Check device existence")
10881 self._CheckDisksExistence([self.other_node, self.target_node])
10882 self._CheckVolumeGroup([self.target_node, self.other_node])
10884 # Step: check other node consistency
10885 self.lu.LogStep(2, steps_total, "Check peer consistency")
10886 self._CheckDisksConsistency(self.other_node,
10887 self.other_node == self.instance.primary_node,
10890 # Step: create new storage
10891 self.lu.LogStep(3, steps_total, "Allocate new storage")
10892 iv_names = self._CreateNewStorage(self.target_node)
10894 # Step: for each lv, detach+rename*2+attach
10895 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
10896 for dev, old_lvs, new_lvs in iv_names.itervalues():
10897 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
10899 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
10901 result.Raise("Can't detach drbd from local storage on node"
10902 " %s for device %s" % (self.target_node, dev.iv_name))
10904 #cfg.Update(instance)
10906 # ok, we created the new LVs, so now we know we have the needed
10907 # storage; as such, we proceed on the target node to rename
10908 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
10909 # using the assumption that logical_id == physical_id (which in
10910 # turn is the unique_id on that node)
10912 # FIXME(iustin): use a better name for the replaced LVs
10913 temp_suffix = int(time.time())
10914 ren_fn = lambda d, suff: (d.physical_id[0],
10915 d.physical_id[1] + "_replaced-%s" % suff)
10917 # Build the rename list based on what LVs exist on the node
10918 rename_old_to_new = []
10919 for to_ren in old_lvs:
10920 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
10921 if not result.fail_msg and result.payload:
10923 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
10925 self.lu.LogInfo("Renaming the old LVs on the target node")
10926 result = self.rpc.call_blockdev_rename(self.target_node,
10928 result.Raise("Can't rename old LVs on node %s" % self.target_node)
10930 # Now we rename the new LVs to the old LVs
10931 self.lu.LogInfo("Renaming the new LVs on the target node")
10932 rename_new_to_old = [(new, old.physical_id)
10933 for old, new in zip(old_lvs, new_lvs)]
10934 result = self.rpc.call_blockdev_rename(self.target_node,
10936 result.Raise("Can't rename new LVs on node %s" % self.target_node)
10938 # Intermediate steps of in memory modifications
10939 for old, new in zip(old_lvs, new_lvs):
10940 new.logical_id = old.logical_id
10941 self.cfg.SetDiskID(new, self.target_node)
10943 # We need to modify old_lvs so that removal later removes the
10944 # right LVs, not the newly added ones; note that old_lvs is a
10946 for disk in old_lvs:
10947 disk.logical_id = ren_fn(disk, temp_suffix)
10948 self.cfg.SetDiskID(disk, self.target_node)
10950 # Now that the new lvs have the old name, we can add them to the device
10951 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
10952 result = self.rpc.call_blockdev_addchildren(self.target_node,
10953 (dev, self.instance),
10954 (new_lvs, self.instance))
10955 msg = result.fail_msg
10957 for new_lv in new_lvs:
10958 msg2 = self.rpc.call_blockdev_remove(self.target_node,
10961 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
10962 hint=("cleanup manually the unused logical"
10964 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
10966 cstep = itertools.count(5)
10968 if self.early_release:
10969 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10970 self._RemoveOldStorage(self.target_node, iv_names)
10971 # TODO: Check if releasing locks early still makes sense
10972 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
10974 # Release all resource locks except those used by the instance
10975 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
10976 keep=self.node_secondary_ip.keys())
10978 # Release all node locks while waiting for sync
10979 _ReleaseLocks(self.lu, locking.LEVEL_NODE)
10981 # TODO: Can the instance lock be downgraded here? Take the optional disk
10982 # shutdown in the caller into consideration.
10985 # This can fail as the old devices are degraded and _WaitForSync
10986 # does a combined result over all disks, so we don't check its return value
10987 self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
10988 _WaitForSync(self.lu, self.instance)
10990 # Check all devices manually
10991 self._CheckDevices(self.instance.primary_node, iv_names)
10993 # Step: remove old storage
10994 if not self.early_release:
10995 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
10996 self._RemoveOldStorage(self.target_node, iv_names)
10998 def _ExecDrbd8Secondary(self, feedback_fn):
10999 """Replace the secondary node for DRBD 8.
11001 The algorithm for replace is quite complicated:
11002 - for all disks of the instance:
11003 - create new LVs on the new node with same names
11004 - shutdown the drbd device on the old secondary
11005 - disconnect the drbd network on the primary
11006 - create the drbd device on the new secondary
11007 - network attach the drbd on the primary, using an artifice:
11008 the drbd code for Attach() will connect to the network if it
11009 finds a device which is connected to the good local disks but
11010 not network enabled
11011 - wait for sync across all devices
11012 - remove all disks from the old secondary
11014 Failures are not very well handled.
11019 pnode = self.instance.primary_node
11021 # Step: check device activation
11022 self.lu.LogStep(1, steps_total, "Check device existence")
11023 self._CheckDisksExistence([self.instance.primary_node])
11024 self._CheckVolumeGroup([self.instance.primary_node])
11026 # Step: check other node consistency
11027 self.lu.LogStep(2, steps_total, "Check peer consistency")
11028 self._CheckDisksConsistency(self.instance.primary_node, True, True)
11030 # Step: create new storage
11031 self.lu.LogStep(3, steps_total, "Allocate new storage")
11032 for idx, dev in enumerate(self.instance.disks):
11033 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
11034 (self.new_node, idx))
11035 # we pass force_create=True to force LVM creation
11036 for new_lv in dev.children:
11037 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
11038 _GetInstanceInfoText(self.instance), False)
11040 # Step 4: dbrd minors and drbd setups changes
11041 # after this, we must manually remove the drbd minors on both the
11042 # error and the success paths
11043 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
11044 minors = self.cfg.AllocateDRBDMinor([self.new_node
11045 for dev in self.instance.disks],
11046 self.instance.name)
11047 logging.debug("Allocated minors %r", minors)
11050 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
11051 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
11052 (self.new_node, idx))
11053 # create new devices on new_node; note that we create two IDs:
11054 # one without port, so the drbd will be activated without
11055 # networking information on the new node at this stage, and one
11056 # with network, for the latter activation in step 4
11057 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
11058 if self.instance.primary_node == o_node1:
11061 assert self.instance.primary_node == o_node2, "Three-node instance?"
11064 new_alone_id = (self.instance.primary_node, self.new_node, None,
11065 p_minor, new_minor, o_secret)
11066 new_net_id = (self.instance.primary_node, self.new_node, o_port,
11067 p_minor, new_minor, o_secret)
11069 iv_names[idx] = (dev, dev.children, new_net_id)
11070 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
11072 drbd_params, _, _ = _ComputeLDParams(constants.DT_DRBD8, self.diskparams)
11073 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
11074 logical_id=new_alone_id,
11075 children=dev.children,
11077 params=drbd_params)
11079 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
11080 _GetInstanceInfoText(self.instance), False)
11081 except errors.GenericError:
11082 self.cfg.ReleaseDRBDMinors(self.instance.name)
11085 # We have new devices, shutdown the drbd on the old secondary
11086 for idx, dev in enumerate(self.instance.disks):
11087 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
11088 self.cfg.SetDiskID(dev, self.target_node)
11089 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
11091 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
11092 "node: %s" % (idx, msg),
11093 hint=("Please cleanup this device manually as"
11094 " soon as possible"))
11096 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
11097 result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
11098 self.instance.disks)[pnode]
11100 msg = result.fail_msg
11102 # detaches didn't succeed (unlikely)
11103 self.cfg.ReleaseDRBDMinors(self.instance.name)
11104 raise errors.OpExecError("Can't detach the disks from the network on"
11105 " old node: %s" % (msg,))
11107 # if we managed to detach at least one, we update all the disks of
11108 # the instance to point to the new secondary
11109 self.lu.LogInfo("Updating instance configuration")
11110 for dev, _, new_logical_id in iv_names.itervalues():
11111 dev.logical_id = new_logical_id
11112 self.cfg.SetDiskID(dev, self.instance.primary_node)
11114 self.cfg.Update(self.instance, feedback_fn)
11116 # Release all node locks (the configuration has been updated)
11117 _ReleaseLocks(self.lu, locking.LEVEL_NODE)
11119 # and now perform the drbd attach
11120 self.lu.LogInfo("Attaching primary drbds to new secondary"
11121 " (standalone => connected)")
11122 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
11124 self.node_secondary_ip,
11125 (self.instance.disks, self.instance),
11126 self.instance.name,
11128 for to_node, to_result in result.items():
11129 msg = to_result.fail_msg
11131 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
11133 hint=("please do a gnt-instance info to see the"
11134 " status of disks"))
11136 cstep = itertools.count(5)
11138 if self.early_release:
11139 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
11140 self._RemoveOldStorage(self.target_node, iv_names)
11141 # TODO: Check if releasing locks early still makes sense
11142 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
11144 # Release all resource locks except those used by the instance
11145 _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
11146 keep=self.node_secondary_ip.keys())
11148 # TODO: Can the instance lock be downgraded here? Take the optional disk
11149 # shutdown in the caller into consideration.
11152 # This can fail as the old devices are degraded and _WaitForSync
11153 # does a combined result over all disks, so we don't check its return value
11154 self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
11155 _WaitForSync(self.lu, self.instance)
11157 # Check all devices manually
11158 self._CheckDevices(self.instance.primary_node, iv_names)
11160 # Step: remove old storage
11161 if not self.early_release:
11162 self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
11163 self._RemoveOldStorage(self.target_node, iv_names)
11166 class LURepairNodeStorage(NoHooksLU):
11167 """Repairs the volume group on a node.
11172 def CheckArguments(self):
11173 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
11175 storage_type = self.op.storage_type
11177 if (constants.SO_FIX_CONSISTENCY not in
11178 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
11179 raise errors.OpPrereqError("Storage units of type '%s' can not be"
11180 " repaired" % storage_type,
11181 errors.ECODE_INVAL)
11183 def ExpandNames(self):
11184 self.needed_locks = {
11185 locking.LEVEL_NODE: [self.op.node_name],
11188 def _CheckFaultyDisks(self, instance, node_name):
11189 """Ensure faulty disks abort the opcode or at least warn."""
11191 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
11193 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
11194 " node '%s'" % (instance.name, node_name),
11195 errors.ECODE_STATE)
11196 except errors.OpPrereqError, err:
11197 if self.op.ignore_consistency:
11198 self.proc.LogWarning(str(err.args[0]))
11202 def CheckPrereq(self):
11203 """Check prerequisites.
11206 # Check whether any instance on this node has faulty disks
11207 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
11208 if inst.admin_state != constants.ADMINST_UP:
11210 check_nodes = set(inst.all_nodes)
11211 check_nodes.discard(self.op.node_name)
11212 for inst_node_name in check_nodes:
11213 self._CheckFaultyDisks(inst, inst_node_name)
11215 def Exec(self, feedback_fn):
11216 feedback_fn("Repairing storage unit '%s' on %s ..." %
11217 (self.op.name, self.op.node_name))
11219 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
11220 result = self.rpc.call_storage_execute(self.op.node_name,
11221 self.op.storage_type, st_args,
11223 constants.SO_FIX_CONSISTENCY)
11224 result.Raise("Failed to repair storage unit '%s' on %s" %
11225 (self.op.name, self.op.node_name))
11228 class LUNodeEvacuate(NoHooksLU):
11229 """Evacuates instances off a list of nodes.
11234 _MODE2IALLOCATOR = {
11235 constants.NODE_EVAC_PRI: constants.IALLOCATOR_NEVAC_PRI,
11236 constants.NODE_EVAC_SEC: constants.IALLOCATOR_NEVAC_SEC,
11237 constants.NODE_EVAC_ALL: constants.IALLOCATOR_NEVAC_ALL,
11239 assert frozenset(_MODE2IALLOCATOR.keys()) == constants.NODE_EVAC_MODES
11240 assert (frozenset(_MODE2IALLOCATOR.values()) ==
11241 constants.IALLOCATOR_NEVAC_MODES)
11243 def CheckArguments(self):
11244 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
11246 def ExpandNames(self):
11247 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
11249 if self.op.remote_node is not None:
11250 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
11251 assert self.op.remote_node
11253 if self.op.remote_node == self.op.node_name:
11254 raise errors.OpPrereqError("Can not use evacuated node as a new"
11255 " secondary node", errors.ECODE_INVAL)
11257 if self.op.mode != constants.NODE_EVAC_SEC:
11258 raise errors.OpPrereqError("Without the use of an iallocator only"
11259 " secondary instances can be evacuated",
11260 errors.ECODE_INVAL)
11263 self.share_locks = _ShareAll()
11264 self.needed_locks = {
11265 locking.LEVEL_INSTANCE: [],
11266 locking.LEVEL_NODEGROUP: [],
11267 locking.LEVEL_NODE: [],
11270 # Determine nodes (via group) optimistically, needs verification once locks
11271 # have been acquired
11272 self.lock_nodes = self._DetermineNodes()
11274 def _DetermineNodes(self):
11275 """Gets the list of nodes to operate on.
11278 if self.op.remote_node is None:
11279 # Iallocator will choose any node(s) in the same group
11280 group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_name])
11282 group_nodes = frozenset([self.op.remote_node])
11284 # Determine nodes to be locked
11285 return set([self.op.node_name]) | group_nodes
11287 def _DetermineInstances(self):
11288 """Builds list of instances to operate on.
11291 assert self.op.mode in constants.NODE_EVAC_MODES
11293 if self.op.mode == constants.NODE_EVAC_PRI:
11294 # Primary instances only
11295 inst_fn = _GetNodePrimaryInstances
11296 assert self.op.remote_node is None, \
11297 "Evacuating primary instances requires iallocator"
11298 elif self.op.mode == constants.NODE_EVAC_SEC:
11299 # Secondary instances only
11300 inst_fn = _GetNodeSecondaryInstances
11303 assert self.op.mode == constants.NODE_EVAC_ALL
11304 inst_fn = _GetNodeInstances
11305 # TODO: In 2.6, change the iallocator interface to take an evacuation mode
11307 raise errors.OpPrereqError("Due to an issue with the iallocator"
11308 " interface it is not possible to evacuate"
11309 " all instances at once; specify explicitly"
11310 " whether to evacuate primary or secondary"
11312 errors.ECODE_INVAL)
11314 return inst_fn(self.cfg, self.op.node_name)
11316 def DeclareLocks(self, level):
11317 if level == locking.LEVEL_INSTANCE:
11318 # Lock instances optimistically, needs verification once node and group
11319 # locks have been acquired
11320 self.needed_locks[locking.LEVEL_INSTANCE] = \
11321 set(i.name for i in self._DetermineInstances())
11323 elif level == locking.LEVEL_NODEGROUP:
11324 # Lock node groups for all potential target nodes optimistically, needs
11325 # verification once nodes have been acquired
11326 self.needed_locks[locking.LEVEL_NODEGROUP] = \
11327 self.cfg.GetNodeGroupsFromNodes(self.lock_nodes)
11329 elif level == locking.LEVEL_NODE:
11330 self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes
11332 def CheckPrereq(self):
11334 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
11335 owned_nodes = self.owned_locks(locking.LEVEL_NODE)
11336 owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
11338 need_nodes = self._DetermineNodes()
11340 if not owned_nodes.issuperset(need_nodes):
11341 raise errors.OpPrereqError("Nodes in same group as '%s' changed since"
11342 " locks were acquired, current nodes are"
11343 " are '%s', used to be '%s'; retry the"
11345 (self.op.node_name,
11346 utils.CommaJoin(need_nodes),
11347 utils.CommaJoin(owned_nodes)),
11348 errors.ECODE_STATE)
11350 wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
11351 if owned_groups != wanted_groups:
11352 raise errors.OpExecError("Node groups changed since locks were acquired,"
11353 " current groups are '%s', used to be '%s';"
11354 " retry the operation" %
11355 (utils.CommaJoin(wanted_groups),
11356 utils.CommaJoin(owned_groups)))
11358 # Determine affected instances
11359 self.instances = self._DetermineInstances()
11360 self.instance_names = [i.name for i in self.instances]
11362 if set(self.instance_names) != owned_instances:
11363 raise errors.OpExecError("Instances on node '%s' changed since locks"
11364 " were acquired, current instances are '%s',"
11365 " used to be '%s'; retry the operation" %
11366 (self.op.node_name,
11367 utils.CommaJoin(self.instance_names),
11368 utils.CommaJoin(owned_instances)))
11370 if self.instance_names:
11371 self.LogInfo("Evacuating instances from node '%s': %s",
11373 utils.CommaJoin(utils.NiceSort(self.instance_names)))
11375 self.LogInfo("No instances to evacuate from node '%s'",
11378 if self.op.remote_node is not None:
11379 for i in self.instances:
11380 if i.primary_node == self.op.remote_node:
11381 raise errors.OpPrereqError("Node %s is the primary node of"
11382 " instance %s, cannot use it as"
11384 (self.op.remote_node, i.name),
11385 errors.ECODE_INVAL)
11387 def Exec(self, feedback_fn):
11388 assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None)
11390 if not self.instance_names:
11391 # No instances to evacuate
11394 elif self.op.iallocator is not None:
11395 # TODO: Implement relocation to other group
11396 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_NODE_EVAC,
11397 evac_mode=self._MODE2IALLOCATOR[self.op.mode],
11398 instances=list(self.instance_names))
11400 ial.Run(self.op.iallocator)
11402 if not ial.success:
11403 raise errors.OpPrereqError("Can't compute node evacuation using"
11404 " iallocator '%s': %s" %
11405 (self.op.iallocator, ial.info),
11406 errors.ECODE_NORES)
11408 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, True)
11410 elif self.op.remote_node is not None:
11411 assert self.op.mode == constants.NODE_EVAC_SEC
11413 [opcodes.OpInstanceReplaceDisks(instance_name=instance_name,
11414 remote_node=self.op.remote_node,
11416 mode=constants.REPLACE_DISK_CHG,
11417 early_release=self.op.early_release)]
11418 for instance_name in self.instance_names
11422 raise errors.ProgrammerError("No iallocator or remote node")
11424 return ResultWithJobs(jobs)
11427 def _SetOpEarlyRelease(early_release, op):
11428 """Sets C{early_release} flag on opcodes if available.
11432 op.early_release = early_release
11433 except AttributeError:
11434 assert not isinstance(op, opcodes.OpInstanceReplaceDisks)
11439 def _NodeEvacDest(use_nodes, group, nodes):
11440 """Returns group or nodes depending on caller's choice.
11444 return utils.CommaJoin(nodes)
11449 def _LoadNodeEvacResult(lu, alloc_result, early_release, use_nodes):
11450 """Unpacks the result of change-group and node-evacuate iallocator requests.
11452 Iallocator modes L{constants.IALLOCATOR_MODE_NODE_EVAC} and
11453 L{constants.IALLOCATOR_MODE_CHG_GROUP}.
11455 @type lu: L{LogicalUnit}
11456 @param lu: Logical unit instance
11457 @type alloc_result: tuple/list
11458 @param alloc_result: Result from iallocator
11459 @type early_release: bool
11460 @param early_release: Whether to release locks early if possible
11461 @type use_nodes: bool
11462 @param use_nodes: Whether to display node names instead of groups
11465 (moved, failed, jobs) = alloc_result
11468 failreason = utils.CommaJoin("%s (%s)" % (name, reason)
11469 for (name, reason) in failed)
11470 lu.LogWarning("Unable to evacuate instances %s", failreason)
11471 raise errors.OpExecError("Unable to evacuate instances %s" % failreason)
11474 lu.LogInfo("Instances to be moved: %s",
11475 utils.CommaJoin("%s (to %s)" %
11476 (name, _NodeEvacDest(use_nodes, group, nodes))
11477 for (name, group, nodes) in moved))
11479 return [map(compat.partial(_SetOpEarlyRelease, early_release),
11480 map(opcodes.OpCode.LoadOpCode, ops))
11484 class LUInstanceGrowDisk(LogicalUnit):
11485 """Grow a disk of an instance.
11488 HPATH = "disk-grow"
11489 HTYPE = constants.HTYPE_INSTANCE
11492 def ExpandNames(self):
11493 self._ExpandAndLockInstance()
11494 self.needed_locks[locking.LEVEL_NODE] = []
11495 self.needed_locks[locking.LEVEL_NODE_RES] = []
11496 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11497 self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
11499 def DeclareLocks(self, level):
11500 if level == locking.LEVEL_NODE:
11501 self._LockInstancesNodes()
11502 elif level == locking.LEVEL_NODE_RES:
11504 self.needed_locks[locking.LEVEL_NODE_RES] = \
11505 self.needed_locks[locking.LEVEL_NODE][:]
11507 def BuildHooksEnv(self):
11508 """Build hooks env.
11510 This runs on the master, the primary and all the secondaries.
11514 "DISK": self.op.disk,
11515 "AMOUNT": self.op.amount,
11516 "ABSOLUTE": self.op.absolute,
11518 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
11521 def BuildHooksNodes(self):
11522 """Build hooks nodes.
11525 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
11528 def CheckPrereq(self):
11529 """Check prerequisites.
11531 This checks that the instance is in the cluster.
11534 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
11535 assert instance is not None, \
11536 "Cannot retrieve locked instance %s" % self.op.instance_name
11537 nodenames = list(instance.all_nodes)
11538 for node in nodenames:
11539 _CheckNodeOnline(self, node)
11541 self.instance = instance
11543 if instance.disk_template not in constants.DTS_GROWABLE:
11544 raise errors.OpPrereqError("Instance's disk layout does not support"
11545 " growing", errors.ECODE_INVAL)
11547 self.disk = instance.FindDisk(self.op.disk)
11549 if self.op.absolute:
11550 self.target = self.op.amount
11551 self.delta = self.target - self.disk.size
11553 raise errors.OpPrereqError("Requested size (%s) is smaller than "
11554 "current disk size (%s)" %
11555 (utils.FormatUnit(self.target, "h"),
11556 utils.FormatUnit(self.disk.size, "h")),
11557 errors.ECODE_STATE)
11559 self.delta = self.op.amount
11560 self.target = self.disk.size + self.delta
11562 raise errors.OpPrereqError("Requested increment (%s) is negative" %
11563 utils.FormatUnit(self.delta, "h"),
11564 errors.ECODE_INVAL)
11566 if instance.disk_template not in (constants.DT_FILE,
11567 constants.DT_SHARED_FILE,
11569 # TODO: check the free disk space for file, when that feature will be
11571 _CheckNodesFreeDiskPerVG(self, nodenames,
11572 self.disk.ComputeGrowth(self.delta))
11574 def Exec(self, feedback_fn):
11575 """Execute disk grow.
11578 instance = self.instance
11581 assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
11582 assert (self.owned_locks(locking.LEVEL_NODE) ==
11583 self.owned_locks(locking.LEVEL_NODE_RES))
11585 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
11587 raise errors.OpExecError("Cannot activate block device to grow")
11589 feedback_fn("Growing disk %s of instance '%s' by %s to %s" %
11590 (self.op.disk, instance.name,
11591 utils.FormatUnit(self.delta, "h"),
11592 utils.FormatUnit(self.target, "h")))
11594 # First run all grow ops in dry-run mode
11595 for node in instance.all_nodes:
11596 self.cfg.SetDiskID(disk, node)
11597 result = self.rpc.call_blockdev_grow(node, (disk, instance), self.delta,
11599 result.Raise("Grow request failed to node %s" % node)
11601 # We know that (as far as we can test) operations across different
11602 # nodes will succeed, time to run it for real
11603 for node in instance.all_nodes:
11604 self.cfg.SetDiskID(disk, node)
11605 result = self.rpc.call_blockdev_grow(node, (disk, instance), self.delta,
11607 result.Raise("Grow request failed to node %s" % node)
11609 # TODO: Rewrite code to work properly
11610 # DRBD goes into sync mode for a short amount of time after executing the
11611 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
11612 # calling "resize" in sync mode fails. Sleeping for a short amount of
11613 # time is a work-around.
11616 disk.RecordGrow(self.delta)
11617 self.cfg.Update(instance, feedback_fn)
11619 # Changes have been recorded, release node lock
11620 _ReleaseLocks(self, locking.LEVEL_NODE)
11622 # Downgrade lock while waiting for sync
11623 self.glm.downgrade(locking.LEVEL_INSTANCE)
11625 if self.op.wait_for_sync:
11626 disk_abort = not _WaitForSync(self, instance, disks=[disk])
11628 self.proc.LogWarning("Disk sync-ing has not returned a good"
11629 " status; please check the instance")
11630 if instance.admin_state != constants.ADMINST_UP:
11631 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
11632 elif instance.admin_state != constants.ADMINST_UP:
11633 self.proc.LogWarning("Not shutting down the disk even if the instance is"
11634 " not supposed to be running because no wait for"
11635 " sync mode was requested")
11637 assert self.owned_locks(locking.LEVEL_NODE_RES)
11638 assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
11641 class LUInstanceQueryData(NoHooksLU):
11642 """Query runtime instance data.
11647 def ExpandNames(self):
11648 self.needed_locks = {}
11650 # Use locking if requested or when non-static information is wanted
11651 if not (self.op.static or self.op.use_locking):
11652 self.LogWarning("Non-static data requested, locks need to be acquired")
11653 self.op.use_locking = True
11655 if self.op.instances or not self.op.use_locking:
11656 # Expand instance names right here
11657 self.wanted_names = _GetWantedInstances(self, self.op.instances)
11659 # Will use acquired locks
11660 self.wanted_names = None
11662 if self.op.use_locking:
11663 self.share_locks = _ShareAll()
11665 if self.wanted_names is None:
11666 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
11668 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
11670 self.needed_locks[locking.LEVEL_NODEGROUP] = []
11671 self.needed_locks[locking.LEVEL_NODE] = []
11672 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
11674 def DeclareLocks(self, level):
11675 if self.op.use_locking:
11676 if level == locking.LEVEL_NODEGROUP:
11677 owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
11679 # Lock all groups used by instances optimistically; this requires going
11680 # via the node before it's locked, requiring verification later on
11681 self.needed_locks[locking.LEVEL_NODEGROUP] = \
11682 frozenset(group_uuid
11683 for instance_name in owned_instances
11685 self.cfg.GetInstanceNodeGroups(instance_name))
11687 elif level == locking.LEVEL_NODE:
11688 self._LockInstancesNodes()
11690 def CheckPrereq(self):
11691 """Check prerequisites.
11693 This only checks the optional instance list against the existing names.
11696 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
11697 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
11698 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
11700 if self.wanted_names is None:
11701 assert self.op.use_locking, "Locking was not used"
11702 self.wanted_names = owned_instances
11704 instances = dict(self.cfg.GetMultiInstanceInfo(self.wanted_names))
11706 if self.op.use_locking:
11707 _CheckInstancesNodeGroups(self.cfg, instances, owned_groups, owned_nodes,
11710 assert not (owned_instances or owned_groups or owned_nodes)
11712 self.wanted_instances = instances.values()
11714 def _ComputeBlockdevStatus(self, node, instance, dev):
11715 """Returns the status of a block device
11718 if self.op.static or not node:
11721 self.cfg.SetDiskID(dev, node)
11723 result = self.rpc.call_blockdev_find(node, dev)
11727 result.Raise("Can't compute disk status for %s" % instance.name)
11729 status = result.payload
11733 return (status.dev_path, status.major, status.minor,
11734 status.sync_percent, status.estimated_time,
11735 status.is_degraded, status.ldisk_status)
11737 def _ComputeDiskStatus(self, instance, snode, dev):
11738 """Compute block device status.
11741 if dev.dev_type in constants.LDS_DRBD:
11742 # we change the snode then (otherwise we use the one passed in)
11743 if dev.logical_id[0] == instance.primary_node:
11744 snode = dev.logical_id[1]
11746 snode = dev.logical_id[0]
11748 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
11750 dev_sstatus = self._ComputeBlockdevStatus(snode, instance, dev)
11753 dev_children = map(compat.partial(self._ComputeDiskStatus,
11760 "iv_name": dev.iv_name,
11761 "dev_type": dev.dev_type,
11762 "logical_id": dev.logical_id,
11763 "physical_id": dev.physical_id,
11764 "pstatus": dev_pstatus,
11765 "sstatus": dev_sstatus,
11766 "children": dev_children,
11771 def Exec(self, feedback_fn):
11772 """Gather and return data"""
11775 cluster = self.cfg.GetClusterInfo()
11777 node_names = itertools.chain(*(i.all_nodes for i in self.wanted_instances))
11778 nodes = dict(self.cfg.GetMultiNodeInfo(node_names))
11780 groups = dict(self.cfg.GetMultiNodeGroupInfo(node.group
11781 for node in nodes.values()))
11783 group2name_fn = lambda uuid: groups[uuid].name
11785 for instance in self.wanted_instances:
11786 pnode = nodes[instance.primary_node]
11788 if self.op.static or pnode.offline:
11789 remote_state = None
11791 self.LogWarning("Primary node %s is marked offline, returning static"
11792 " information only for instance %s" %
11793 (pnode.name, instance.name))
11795 remote_info = self.rpc.call_instance_info(instance.primary_node,
11797 instance.hypervisor)
11798 remote_info.Raise("Error checking node %s" % instance.primary_node)
11799 remote_info = remote_info.payload
11800 if remote_info and "state" in remote_info:
11801 remote_state = "up"
11803 if instance.admin_state == constants.ADMINST_UP:
11804 remote_state = "down"
11806 remote_state = instance.admin_state
11808 disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
11811 snodes_group_uuids = [nodes[snode_name].group
11812 for snode_name in instance.secondary_nodes]
11814 result[instance.name] = {
11815 "name": instance.name,
11816 "config_state": instance.admin_state,
11817 "run_state": remote_state,
11818 "pnode": instance.primary_node,
11819 "pnode_group_uuid": pnode.group,
11820 "pnode_group_name": group2name_fn(pnode.group),
11821 "snodes": instance.secondary_nodes,
11822 "snodes_group_uuids": snodes_group_uuids,
11823 "snodes_group_names": map(group2name_fn, snodes_group_uuids),
11825 # this happens to be the same format used for hooks
11826 "nics": _NICListToTuple(self, instance.nics),
11827 "disk_template": instance.disk_template,
11829 "hypervisor": instance.hypervisor,
11830 "network_port": instance.network_port,
11831 "hv_instance": instance.hvparams,
11832 "hv_actual": cluster.FillHV(instance, skip_globals=True),
11833 "be_instance": instance.beparams,
11834 "be_actual": cluster.FillBE(instance),
11835 "os_instance": instance.osparams,
11836 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
11837 "serial_no": instance.serial_no,
11838 "mtime": instance.mtime,
11839 "ctime": instance.ctime,
11840 "uuid": instance.uuid,
11846 def PrepareContainerMods(mods, private_fn):
11847 """Prepares a list of container modifications by adding a private data field.
11849 @type mods: list of tuples; (operation, index, parameters)
11850 @param mods: List of modifications
11851 @type private_fn: callable or None
11852 @param private_fn: Callable for constructing a private data field for a
11857 if private_fn is None:
11862 return [(op, idx, params, fn()) for (op, idx, params) in mods]
11865 #: Type description for changes as returned by L{ApplyContainerMods}'s
11867 _TApplyContModsCbChanges = \
11868 ht.TMaybeListOf(ht.TAnd(ht.TIsLength(2), ht.TItems([
11869 ht.TNonEmptyString,
11874 def ApplyContainerMods(kind, container, chgdesc, mods,
11875 create_fn, modify_fn, remove_fn):
11876 """Applies descriptions in C{mods} to C{container}.
11879 @param kind: One-word item description
11880 @type container: list
11881 @param container: Container to modify
11882 @type chgdesc: None or list
11883 @param chgdesc: List of applied changes
11885 @param mods: Modifications as returned by L{PrepareContainerMods}
11886 @type create_fn: callable
11887 @param create_fn: Callback for creating a new item (L{constants.DDM_ADD});
11888 receives absolute item index, parameters and private data object as added
11889 by L{PrepareContainerMods}, returns tuple containing new item and changes
11891 @type modify_fn: callable
11892 @param modify_fn: Callback for modifying an existing item
11893 (L{constants.DDM_MODIFY}); receives absolute item index, item, parameters
11894 and private data object as added by L{PrepareContainerMods}, returns
11896 @type remove_fn: callable
11897 @param remove_fn: Callback on removing item; receives absolute item index,
11898 item and private data object as added by L{PrepareContainerMods}
11901 for (op, idx, params, private) in mods:
11904 absidx = len(container) - 1
11906 raise IndexError("Not accepting negative indices other than -1")
11907 elif idx > len(container):
11908 raise IndexError("Got %s index %s, but there are only %s" %
11909 (kind, idx, len(container)))
11915 if op == constants.DDM_ADD:
11916 # Calculate where item will be added
11918 addidx = len(container)
11922 if create_fn is None:
11925 (item, changes) = create_fn(addidx, params, private)
11928 container.append(item)
11931 assert idx <= len(container)
11932 # list.insert does so before the specified index
11933 container.insert(idx, item)
11935 # Retrieve existing item
11937 item = container[absidx]
11939 raise IndexError("Invalid %s index %s" % (kind, idx))
11941 if op == constants.DDM_REMOVE:
11944 if remove_fn is not None:
11945 remove_fn(absidx, item, private)
11947 changes = [("%s/%s" % (kind, absidx), "remove")]
11949 assert container[absidx] == item
11950 del container[absidx]
11951 elif op == constants.DDM_MODIFY:
11952 if modify_fn is not None:
11953 changes = modify_fn(absidx, item, params, private)
11955 raise errors.ProgrammerError("Unhandled operation '%s'" % op)
11957 assert _TApplyContModsCbChanges(changes)
11959 if not (chgdesc is None or changes is None):
11960 chgdesc.extend(changes)
11963 def _UpdateIvNames(base_index, disks):
11964 """Updates the C{iv_name} attribute of disks.
11966 @type disks: list of L{objects.Disk}
11969 for (idx, disk) in enumerate(disks):
11970 disk.iv_name = "disk/%s" % (base_index + idx, )
11973 class _InstNicModPrivate:
11974 """Data structure for network interface modifications.
11976 Used by L{LUInstanceSetParams}.
11979 def __init__(self):
11984 class LUInstanceSetParams(LogicalUnit):
11985 """Modifies an instances's parameters.
11988 HPATH = "instance-modify"
11989 HTYPE = constants.HTYPE_INSTANCE
11993 def _UpgradeDiskNicMods(kind, mods, verify_fn):
11994 assert ht.TList(mods)
11995 assert not mods or len(mods[0]) in (2, 3)
11997 if mods and len(mods[0]) == 2:
12001 for op, params in mods:
12002 if op in (constants.DDM_ADD, constants.DDM_REMOVE):
12003 result.append((op, -1, params))
12007 raise errors.OpPrereqError("Only one %s add or remove operation is"
12008 " supported at a time" % kind,
12009 errors.ECODE_INVAL)
12011 result.append((constants.DDM_MODIFY, op, params))
12013 assert verify_fn(result)
12020 def _CheckMods(kind, mods, key_types, item_fn):
12021 """Ensures requested disk/NIC modifications are valid.
12024 for (op, _, params) in mods:
12025 assert ht.TDict(params)
12027 utils.ForceDictType(params, key_types)
12029 if op == constants.DDM_REMOVE:
12031 raise errors.OpPrereqError("No settings should be passed when"
12032 " removing a %s" % kind,
12033 errors.ECODE_INVAL)
12034 elif op in (constants.DDM_ADD, constants.DDM_MODIFY):
12035 item_fn(op, params)
12037 raise errors.ProgrammerError("Unhandled operation '%s'" % op)
12040 def _VerifyDiskModification(op, params):
12041 """Verifies a disk modification.
12044 if op == constants.DDM_ADD:
12045 mode = params.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
12046 if mode not in constants.DISK_ACCESS_SET:
12047 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
12048 errors.ECODE_INVAL)
12050 size = params.get(constants.IDISK_SIZE, None)
12052 raise errors.OpPrereqError("Required disk parameter '%s' missing" %
12053 constants.IDISK_SIZE, errors.ECODE_INVAL)
12057 except (TypeError, ValueError), err:
12058 raise errors.OpPrereqError("Invalid disk size parameter: %s" % err,
12059 errors.ECODE_INVAL)
12061 params[constants.IDISK_SIZE] = size
12063 elif op == constants.DDM_MODIFY and constants.IDISK_SIZE in params:
12064 raise errors.OpPrereqError("Disk size change not possible, use"
12065 " grow-disk", errors.ECODE_INVAL)
12068 def _VerifyNicModification(op, params):
12069 """Verifies a network interface modification.
12072 if op in (constants.DDM_ADD, constants.DDM_MODIFY):
12073 ip = params.get(constants.INIC_IP, None)
12076 elif ip.lower() == constants.VALUE_NONE:
12077 params[constants.INIC_IP] = None
12078 elif not netutils.IPAddress.IsValid(ip):
12079 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
12080 errors.ECODE_INVAL)
12082 bridge = params.get("bridge", None)
12083 link = params.get(constants.INIC_LINK, None)
12084 if bridge and link:
12085 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
12086 " at the same time", errors.ECODE_INVAL)
12087 elif bridge and bridge.lower() == constants.VALUE_NONE:
12088 params["bridge"] = None
12089 elif link and link.lower() == constants.VALUE_NONE:
12090 params[constants.INIC_LINK] = None
12092 if op == constants.DDM_ADD:
12093 macaddr = params.get(constants.INIC_MAC, None)
12094 if macaddr is None:
12095 params[constants.INIC_MAC] = constants.VALUE_AUTO
12097 if constants.INIC_MAC in params:
12098 macaddr = params[constants.INIC_MAC]
12099 if macaddr not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
12100 macaddr = utils.NormalizeAndValidateMac(macaddr)
12102 if op == constants.DDM_MODIFY and macaddr == constants.VALUE_AUTO:
12103 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
12104 " modifying an existing NIC",
12105 errors.ECODE_INVAL)
12107 def CheckArguments(self):
12108 if not (self.op.nics or self.op.disks or self.op.disk_template or
12109 self.op.hvparams or self.op.beparams or self.op.os_name or
12110 self.op.offline is not None or self.op.runtime_mem):
12111 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
12113 if self.op.hvparams:
12114 _CheckGlobalHvParams(self.op.hvparams)
12117 self._UpgradeDiskNicMods("disk", self.op.disks,
12118 opcodes.OpInstanceSetParams.TestDiskModifications)
12120 self._UpgradeDiskNicMods("NIC", self.op.nics,
12121 opcodes.OpInstanceSetParams.TestNicModifications)
12123 # Check disk modifications
12124 self._CheckMods("disk", self.op.disks, constants.IDISK_PARAMS_TYPES,
12125 self._VerifyDiskModification)
12127 if self.op.disks and self.op.disk_template is not None:
12128 raise errors.OpPrereqError("Disk template conversion and other disk"
12129 " changes not supported at the same time",
12130 errors.ECODE_INVAL)
12132 if (self.op.disk_template and
12133 self.op.disk_template in constants.DTS_INT_MIRROR and
12134 self.op.remote_node is None):
12135 raise errors.OpPrereqError("Changing the disk template to a mirrored"
12136 " one requires specifying a secondary node",
12137 errors.ECODE_INVAL)
12139 # Check NIC modifications
12140 self._CheckMods("NIC", self.op.nics, constants.INIC_PARAMS_TYPES,
12141 self._VerifyNicModification)
12143 def ExpandNames(self):
12144 self._ExpandAndLockInstance()
12145 # Can't even acquire node locks in shared mode as upcoming changes in
12146 # Ganeti 2.6 will start to modify the node object on disk conversion
12147 self.needed_locks[locking.LEVEL_NODE] = []
12148 self.needed_locks[locking.LEVEL_NODE_RES] = []
12149 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
12151 def DeclareLocks(self, level):
12152 # TODO: Acquire group lock in shared mode (disk parameters)
12153 if level == locking.LEVEL_NODE:
12154 self._LockInstancesNodes()
12155 if self.op.disk_template and self.op.remote_node:
12156 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
12157 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
12158 elif level == locking.LEVEL_NODE_RES and self.op.disk_template:
12160 self.needed_locks[locking.LEVEL_NODE_RES] = \
12161 self.needed_locks[locking.LEVEL_NODE][:]
12163 def BuildHooksEnv(self):
12164 """Build hooks env.
12166 This runs on the master, primary and secondaries.
12170 if constants.BE_MINMEM in self.be_new:
12171 args["minmem"] = self.be_new[constants.BE_MINMEM]
12172 if constants.BE_MAXMEM in self.be_new:
12173 args["maxmem"] = self.be_new[constants.BE_MAXMEM]
12174 if constants.BE_VCPUS in self.be_new:
12175 args["vcpus"] = self.be_new[constants.BE_VCPUS]
12176 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
12177 # information at all.
12179 if self._new_nics is not None:
12182 for nic in self._new_nics:
12183 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
12184 mode = nicparams[constants.NIC_MODE]
12185 link = nicparams[constants.NIC_LINK]
12186 nics.append((nic.ip, nic.mac, mode, link))
12188 args["nics"] = nics
12190 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
12191 if self.op.disk_template:
12192 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
12193 if self.op.runtime_mem:
12194 env["RUNTIME_MEMORY"] = self.op.runtime_mem
12198 def BuildHooksNodes(self):
12199 """Build hooks nodes.
12202 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
12205 def _PrepareNicModification(self, params, private, old_ip, old_params,
12207 update_params_dict = dict([(key, params[key])
12208 for key in constants.NICS_PARAMETERS
12211 if "bridge" in params:
12212 update_params_dict[constants.NIC_LINK] = params["bridge"]
12214 new_params = _GetUpdatedParams(old_params, update_params_dict)
12215 utils.ForceDictType(new_params, constants.NICS_PARAMETER_TYPES)
12217 new_filled_params = cluster.SimpleFillNIC(new_params)
12218 objects.NIC.CheckParameterSyntax(new_filled_params)
12220 new_mode = new_filled_params[constants.NIC_MODE]
12221 if new_mode == constants.NIC_MODE_BRIDGED:
12222 bridge = new_filled_params[constants.NIC_LINK]
12223 msg = self.rpc.call_bridges_exist(pnode, [bridge]).fail_msg
12225 msg = "Error checking bridges on node '%s': %s" % (pnode, msg)
12227 self.warn.append(msg)
12229 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
12231 elif new_mode == constants.NIC_MODE_ROUTED:
12232 ip = params.get(constants.INIC_IP, old_ip)
12234 raise errors.OpPrereqError("Cannot set the NIC IP address to None"
12235 " on a routed NIC", errors.ECODE_INVAL)
12237 if constants.INIC_MAC in params:
12238 mac = params[constants.INIC_MAC]
12240 raise errors.OpPrereqError("Cannot unset the NIC MAC address",
12241 errors.ECODE_INVAL)
12242 elif mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
12243 # otherwise generate the MAC address
12244 params[constants.INIC_MAC] = \
12245 self.cfg.GenerateMAC(self.proc.GetECId())
12247 # or validate/reserve the current one
12249 self.cfg.ReserveMAC(mac, self.proc.GetECId())
12250 except errors.ReservationError:
12251 raise errors.OpPrereqError("MAC address '%s' already in use"
12252 " in cluster" % mac,
12253 errors.ECODE_NOTUNIQUE)
12255 private.params = new_params
12256 private.filled = new_filled_params
12258 return (None, None)
12260 def CheckPrereq(self):
12261 """Check prerequisites.
12263 This only checks the instance list against the existing names.
12266 # checking the new params on the primary/secondary nodes
12268 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
12269 cluster = self.cluster = self.cfg.GetClusterInfo()
12270 assert self.instance is not None, \
12271 "Cannot retrieve locked instance %s" % self.op.instance_name
12272 pnode = instance.primary_node
12273 nodelist = list(instance.all_nodes)
12274 pnode_info = self.cfg.GetNodeInfo(pnode)
12275 self.diskparams = self.cfg.GetNodeGroup(pnode_info.group).diskparams
12277 # Prepare disk/NIC modifications
12278 self.diskmod = PrepareContainerMods(self.op.disks, None)
12279 self.nicmod = PrepareContainerMods(self.op.nics, _InstNicModPrivate)
12282 if self.op.os_name and not self.op.force:
12283 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
12284 self.op.force_variant)
12285 instance_os = self.op.os_name
12287 instance_os = instance.os
12289 assert not (self.op.disk_template and self.op.disks), \
12290 "Can't modify disk template and apply disk changes at the same time"
12292 if self.op.disk_template:
12293 if instance.disk_template == self.op.disk_template:
12294 raise errors.OpPrereqError("Instance already has disk template %s" %
12295 instance.disk_template, errors.ECODE_INVAL)
12297 if (instance.disk_template,
12298 self.op.disk_template) not in self._DISK_CONVERSIONS:
12299 raise errors.OpPrereqError("Unsupported disk template conversion from"
12300 " %s to %s" % (instance.disk_template,
12301 self.op.disk_template),
12302 errors.ECODE_INVAL)
12303 _CheckInstanceState(self, instance, INSTANCE_DOWN,
12304 msg="cannot change disk template")
12305 if self.op.disk_template in constants.DTS_INT_MIRROR:
12306 if self.op.remote_node == pnode:
12307 raise errors.OpPrereqError("Given new secondary node %s is the same"
12308 " as the primary node of the instance" %
12309 self.op.remote_node, errors.ECODE_STATE)
12310 _CheckNodeOnline(self, self.op.remote_node)
12311 _CheckNodeNotDrained(self, self.op.remote_node)
12312 # FIXME: here we assume that the old instance type is DT_PLAIN
12313 assert instance.disk_template == constants.DT_PLAIN
12314 disks = [{constants.IDISK_SIZE: d.size,
12315 constants.IDISK_VG: d.logical_id[0]}
12316 for d in instance.disks]
12317 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
12318 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
12320 snode_info = self.cfg.GetNodeInfo(self.op.remote_node)
12321 snode_group = self.cfg.GetNodeGroup(snode_info.group)
12322 ipolicy = _CalculateGroupIPolicy(cluster, snode_group)
12323 _CheckTargetNodeIPolicy(self, ipolicy, instance, snode_info,
12324 ignore=self.op.ignore_ipolicy)
12325 if pnode_info.group != snode_info.group:
12326 self.LogWarning("The primary and secondary nodes are in two"
12327 " different node groups; the disk parameters"
12328 " from the first disk's node group will be"
12331 # hvparams processing
12332 if self.op.hvparams:
12333 hv_type = instance.hypervisor
12334 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
12335 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
12336 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
12339 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
12340 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
12341 self.hv_proposed = self.hv_new = hv_new # the new actual values
12342 self.hv_inst = i_hvdict # the new dict (without defaults)
12344 self.hv_proposed = cluster.SimpleFillHV(instance.hypervisor, instance.os,
12346 self.hv_new = self.hv_inst = {}
12348 # beparams processing
12349 if self.op.beparams:
12350 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
12352 objects.UpgradeBeParams(i_bedict)
12353 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
12354 be_new = cluster.SimpleFillBE(i_bedict)
12355 self.be_proposed = self.be_new = be_new # the new actual values
12356 self.be_inst = i_bedict # the new dict (without defaults)
12358 self.be_new = self.be_inst = {}
12359 self.be_proposed = cluster.SimpleFillBE(instance.beparams)
12360 be_old = cluster.FillBE(instance)
12362 # CPU param validation -- checking every time a paramtere is
12363 # changed to cover all cases where either CPU mask or vcpus have
12365 if (constants.BE_VCPUS in self.be_proposed and
12366 constants.HV_CPU_MASK in self.hv_proposed):
12368 utils.ParseMultiCpuMask(self.hv_proposed[constants.HV_CPU_MASK])
12369 # Verify mask is consistent with number of vCPUs. Can skip this
12370 # test if only 1 entry in the CPU mask, which means same mask
12371 # is applied to all vCPUs.
12372 if (len(cpu_list) > 1 and
12373 len(cpu_list) != self.be_proposed[constants.BE_VCPUS]):
12374 raise errors.OpPrereqError("Number of vCPUs [%d] does not match the"
12376 (self.be_proposed[constants.BE_VCPUS],
12377 self.hv_proposed[constants.HV_CPU_MASK]),
12378 errors.ECODE_INVAL)
12380 # Only perform this test if a new CPU mask is given
12381 if constants.HV_CPU_MASK in self.hv_new:
12382 # Calculate the largest CPU number requested
12383 max_requested_cpu = max(map(max, cpu_list))
12384 # Check that all of the instance's nodes have enough physical CPUs to
12385 # satisfy the requested CPU mask
12386 _CheckNodesPhysicalCPUs(self, instance.all_nodes,
12387 max_requested_cpu + 1, instance.hypervisor)
12389 # osparams processing
12390 if self.op.osparams:
12391 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
12392 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
12393 self.os_inst = i_osdict # the new dict (without defaults)
12399 #TODO(dynmem): do the appropriate check involving MINMEM
12400 if (constants.BE_MAXMEM in self.op.beparams and not self.op.force and
12401 be_new[constants.BE_MAXMEM] > be_old[constants.BE_MAXMEM]):
12402 mem_check_list = [pnode]
12403 if be_new[constants.BE_AUTO_BALANCE]:
12404 # either we changed auto_balance to yes or it was from before
12405 mem_check_list.extend(instance.secondary_nodes)
12406 instance_info = self.rpc.call_instance_info(pnode, instance.name,
12407 instance.hypervisor)
12408 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
12409 [instance.hypervisor])
12410 pninfo = nodeinfo[pnode]
12411 msg = pninfo.fail_msg
12413 # Assume the primary node is unreachable and go ahead
12414 self.warn.append("Can't get info from primary node %s: %s" %
12417 (_, _, (pnhvinfo, )) = pninfo.payload
12418 if not isinstance(pnhvinfo.get("memory_free", None), int):
12419 self.warn.append("Node data from primary node %s doesn't contain"
12420 " free memory information" % pnode)
12421 elif instance_info.fail_msg:
12422 self.warn.append("Can't get instance runtime information: %s" %
12423 instance_info.fail_msg)
12425 if instance_info.payload:
12426 current_mem = int(instance_info.payload["memory"])
12428 # Assume instance not running
12429 # (there is a slight race condition here, but it's not very
12430 # probable, and we have no other way to check)
12431 # TODO: Describe race condition
12433 #TODO(dynmem): do the appropriate check involving MINMEM
12434 miss_mem = (be_new[constants.BE_MAXMEM] - current_mem -
12435 pnhvinfo["memory_free"])
12437 raise errors.OpPrereqError("This change will prevent the instance"
12438 " from starting, due to %d MB of memory"
12439 " missing on its primary node" %
12441 errors.ECODE_NORES)
12443 if be_new[constants.BE_AUTO_BALANCE]:
12444 for node, nres in nodeinfo.items():
12445 if node not in instance.secondary_nodes:
12447 nres.Raise("Can't get info from secondary node %s" % node,
12448 prereq=True, ecode=errors.ECODE_STATE)
12449 (_, _, (nhvinfo, )) = nres.payload
12450 if not isinstance(nhvinfo.get("memory_free", None), int):
12451 raise errors.OpPrereqError("Secondary node %s didn't return free"
12452 " memory information" % node,
12453 errors.ECODE_STATE)
12454 #TODO(dynmem): do the appropriate check involving MINMEM
12455 elif be_new[constants.BE_MAXMEM] > nhvinfo["memory_free"]:
12456 raise errors.OpPrereqError("This change will prevent the instance"
12457 " from failover to its secondary node"
12458 " %s, due to not enough memory" % node,
12459 errors.ECODE_STATE)
12461 if self.op.runtime_mem:
12462 remote_info = self.rpc.call_instance_info(instance.primary_node,
12464 instance.hypervisor)
12465 remote_info.Raise("Error checking node %s" % instance.primary_node)
12466 if not remote_info.payload: # not running already
12467 raise errors.OpPrereqError("Instance %s is not running" % instance.name,
12468 errors.ECODE_STATE)
12470 current_memory = remote_info.payload["memory"]
12471 if (not self.op.force and
12472 (self.op.runtime_mem > self.be_proposed[constants.BE_MAXMEM] or
12473 self.op.runtime_mem < self.be_proposed[constants.BE_MINMEM])):
12474 raise errors.OpPrereqError("Instance %s must have memory between %d"
12475 " and %d MB of memory unless --force is"
12476 " given" % (instance.name,
12477 self.be_proposed[constants.BE_MINMEM],
12478 self.be_proposed[constants.BE_MAXMEM]),
12479 errors.ECODE_INVAL)
12481 if self.op.runtime_mem > current_memory:
12482 _CheckNodeFreeMemory(self, instance.primary_node,
12483 "ballooning memory for instance %s" %
12485 self.op.memory - current_memory,
12486 instance.hypervisor)
12488 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
12489 raise errors.OpPrereqError("Disk operations not supported for"
12490 " diskless instances",
12491 errors.ECODE_INVAL)
12493 def _PrepareNicCreate(_, params, private):
12494 return self._PrepareNicModification(params, private, None, {},
12497 def _PrepareNicMod(_, nic, params, private):
12498 return self._PrepareNicModification(params, private, nic.ip,
12499 nic.nicparams, cluster, pnode)
12501 # Verify NIC changes (operating on copy)
12502 nics = instance.nics[:]
12503 ApplyContainerMods("NIC", nics, None, self.nicmod,
12504 _PrepareNicCreate, _PrepareNicMod, None)
12505 if len(nics) > constants.MAX_NICS:
12506 raise errors.OpPrereqError("Instance has too many network interfaces"
12507 " (%d), cannot add more" % constants.MAX_NICS,
12508 errors.ECODE_STATE)
12510 # Verify disk changes (operating on a copy)
12511 disks = instance.disks[:]
12512 ApplyContainerMods("disk", disks, None, self.diskmod, None, None, None)
12513 if len(disks) > constants.MAX_DISKS:
12514 raise errors.OpPrereqError("Instance has too many disks (%d), cannot add"
12515 " more" % constants.MAX_DISKS,
12516 errors.ECODE_STATE)
12518 if self.op.offline is not None:
12519 if self.op.offline:
12520 msg = "can't change to offline"
12522 msg = "can't change to online"
12523 _CheckInstanceState(self, instance, CAN_CHANGE_INSTANCE_OFFLINE, msg=msg)
12525 # Pre-compute NIC changes (necessary to use result in hooks)
12526 self._nic_chgdesc = []
12528 # Operate on copies as this is still in prereq
12529 nics = [nic.Copy() for nic in instance.nics]
12530 ApplyContainerMods("NIC", nics, self._nic_chgdesc, self.nicmod,
12531 self._CreateNewNic, self._ApplyNicMods, None)
12532 self._new_nics = nics
12534 self._new_nics = None
12536 def _ConvertPlainToDrbd(self, feedback_fn):
12537 """Converts an instance from plain to drbd.
12540 feedback_fn("Converting template to drbd")
12541 instance = self.instance
12542 pnode = instance.primary_node
12543 snode = self.op.remote_node
12545 assert instance.disk_template == constants.DT_PLAIN
12547 # create a fake disk info for _GenerateDiskTemplate
12548 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
12549 constants.IDISK_VG: d.logical_id[0]}
12550 for d in instance.disks]
12551 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
12552 instance.name, pnode, [snode],
12553 disk_info, None, None, 0, feedback_fn,
12555 info = _GetInstanceInfoText(instance)
12556 feedback_fn("Creating additional volumes...")
12557 # first, create the missing data and meta devices
12558 for disk in new_disks:
12559 # unfortunately this is... not too nice
12560 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
12562 for child in disk.children:
12563 _CreateSingleBlockDev(self, snode, instance, child, info, True)
12564 # at this stage, all new LVs have been created, we can rename the
12566 feedback_fn("Renaming original volumes...")
12567 rename_list = [(o, n.children[0].logical_id)
12568 for (o, n) in zip(instance.disks, new_disks)]
12569 result = self.rpc.call_blockdev_rename(pnode, rename_list)
12570 result.Raise("Failed to rename original LVs")
12572 feedback_fn("Initializing DRBD devices...")
12573 # all child devices are in place, we can now create the DRBD devices
12574 for disk in new_disks:
12575 for node in [pnode, snode]:
12576 f_create = node == pnode
12577 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
12579 # at this point, the instance has been modified
12580 instance.disk_template = constants.DT_DRBD8
12581 instance.disks = new_disks
12582 self.cfg.Update(instance, feedback_fn)
12584 # Release node locks while waiting for sync
12585 _ReleaseLocks(self, locking.LEVEL_NODE)
12587 # disks are created, waiting for sync
12588 disk_abort = not _WaitForSync(self, instance,
12589 oneshot=not self.op.wait_for_sync)
12591 raise errors.OpExecError("There are some degraded disks for"
12592 " this instance, please cleanup manually")
12594 # Node resource locks will be released by caller
12596 def _ConvertDrbdToPlain(self, feedback_fn):
12597 """Converts an instance from drbd to plain.
12600 instance = self.instance
12602 assert len(instance.secondary_nodes) == 1
12603 assert instance.disk_template == constants.DT_DRBD8
12605 pnode = instance.primary_node
12606 snode = instance.secondary_nodes[0]
12607 feedback_fn("Converting template to plain")
12609 old_disks = instance.disks
12610 new_disks = [d.children[0] for d in old_disks]
12612 # copy over size and mode
12613 for parent, child in zip(old_disks, new_disks):
12614 child.size = parent.size
12615 child.mode = parent.mode
12617 # this is a DRBD disk, return its port to the pool
12618 # NOTE: this must be done right before the call to cfg.Update!
12619 for disk in old_disks:
12620 tcp_port = disk.logical_id[2]
12621 self.cfg.AddTcpUdpPort(tcp_port)
12623 # update instance structure
12624 instance.disks = new_disks
12625 instance.disk_template = constants.DT_PLAIN
12626 self.cfg.Update(instance, feedback_fn)
12628 # Release locks in case removing disks takes a while
12629 _ReleaseLocks(self, locking.LEVEL_NODE)
12631 feedback_fn("Removing volumes on the secondary node...")
12632 for disk in old_disks:
12633 self.cfg.SetDiskID(disk, snode)
12634 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
12636 self.LogWarning("Could not remove block device %s on node %s,"
12637 " continuing anyway: %s", disk.iv_name, snode, msg)
12639 feedback_fn("Removing unneeded volumes on the primary node...")
12640 for idx, disk in enumerate(old_disks):
12641 meta = disk.children[1]
12642 self.cfg.SetDiskID(meta, pnode)
12643 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
12645 self.LogWarning("Could not remove metadata for disk %d on node %s,"
12646 " continuing anyway: %s", idx, pnode, msg)
12648 def _CreateNewDisk(self, idx, params, _):
12649 """Creates a new disk.
12652 instance = self.instance
12655 if instance.disk_template in constants.DTS_FILEBASED:
12656 (file_driver, file_path) = instance.disks[0].logical_id
12657 file_path = os.path.dirname(file_path)
12659 file_driver = file_path = None
12662 _GenerateDiskTemplate(self, instance.disk_template, instance.name,
12663 instance.primary_node, instance.secondary_nodes,
12664 [params], file_path, file_driver, idx,
12665 self.Log, self.diskparams)[0]
12667 info = _GetInstanceInfoText(instance)
12669 logging.info("Creating volume %s for instance %s",
12670 disk.iv_name, instance.name)
12671 # Note: this needs to be kept in sync with _CreateDisks
12673 for node in instance.all_nodes:
12674 f_create = (node == instance.primary_node)
12676 _CreateBlockDev(self, node, instance, disk, f_create, info, f_create)
12677 except errors.OpExecError, err:
12678 self.LogWarning("Failed to create volume %s (%s) on node '%s': %s",
12679 disk.iv_name, disk, node, err)
12682 ("disk/%d" % idx, "add:size=%s,mode=%s" % (disk.size, disk.mode)),
12686 def _ModifyDisk(idx, disk, params, _):
12687 """Modifies a disk.
12690 disk.mode = params[constants.IDISK_MODE]
12693 ("disk.mode/%d" % idx, disk.mode),
12696 def _RemoveDisk(self, idx, root, _):
12700 for node, disk in root.ComputeNodeTree(self.instance.primary_node):
12701 self.cfg.SetDiskID(disk, node)
12702 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
12704 self.LogWarning("Could not remove disk/%d on node '%s': %s,"
12705 " continuing anyway", idx, node, msg)
12707 # if this is a DRBD disk, return its port to the pool
12708 if root.dev_type in constants.LDS_DRBD:
12709 self.cfg.AddTcpUdpPort(root.logical_id[2])
12712 def _CreateNewNic(idx, params, private):
12713 """Creates data structure for a new network interface.
12716 mac = params[constants.INIC_MAC]
12717 ip = params.get(constants.INIC_IP, None)
12718 nicparams = private.params
12720 return (objects.NIC(mac=mac, ip=ip, nicparams=nicparams), [
12722 "add:mac=%s,ip=%s,mode=%s,link=%s" %
12723 (mac, ip, private.filled[constants.NIC_MODE],
12724 private.filled[constants.NIC_LINK])),
12728 def _ApplyNicMods(idx, nic, params, private):
12729 """Modifies a network interface.
12734 for key in [constants.INIC_MAC, constants.INIC_IP]:
12736 changes.append(("nic.%s/%d" % (key, idx), params[key]))
12737 setattr(nic, key, params[key])
12740 nic.nicparams = private.params
12742 for (key, val) in params.items():
12743 changes.append(("nic.%s/%d" % (key, idx), val))
12747 def Exec(self, feedback_fn):
12748 """Modifies an instance.
12750 All parameters take effect only at the next restart of the instance.
12753 # Process here the warnings from CheckPrereq, as we don't have a
12754 # feedback_fn there.
12755 # TODO: Replace with self.LogWarning
12756 for warn in self.warn:
12757 feedback_fn("WARNING: %s" % warn)
12759 assert ((self.op.disk_template is None) ^
12760 bool(self.owned_locks(locking.LEVEL_NODE_RES))), \
12761 "Not owning any node resource locks"
12764 instance = self.instance
12767 if self.op.runtime_mem:
12768 rpcres = self.rpc.call_instance_balloon_memory(instance.primary_node,
12770 self.op.runtime_mem)
12771 rpcres.Raise("Cannot modify instance runtime memory")
12772 result.append(("runtime_memory", self.op.runtime_mem))
12774 # Apply disk changes
12775 ApplyContainerMods("disk", instance.disks, result, self.diskmod,
12776 self._CreateNewDisk, self._ModifyDisk, self._RemoveDisk)
12777 _UpdateIvNames(0, instance.disks)
12779 if self.op.disk_template:
12781 check_nodes = set(instance.all_nodes)
12782 if self.op.remote_node:
12783 check_nodes.add(self.op.remote_node)
12784 for level in [locking.LEVEL_NODE, locking.LEVEL_NODE_RES]:
12785 owned = self.owned_locks(level)
12786 assert not (check_nodes - owned), \
12787 ("Not owning the correct locks, owning %r, expected at least %r" %
12788 (owned, check_nodes))
12790 r_shut = _ShutdownInstanceDisks(self, instance)
12792 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
12793 " proceed with disk template conversion")
12794 mode = (instance.disk_template, self.op.disk_template)
12796 self._DISK_CONVERSIONS[mode](self, feedback_fn)
12798 self.cfg.ReleaseDRBDMinors(instance.name)
12800 result.append(("disk_template", self.op.disk_template))
12802 assert instance.disk_template == self.op.disk_template, \
12803 ("Expected disk template '%s', found '%s'" %
12804 (self.op.disk_template, instance.disk_template))
12806 # Release node and resource locks if there are any (they might already have
12807 # been released during disk conversion)
12808 _ReleaseLocks(self, locking.LEVEL_NODE)
12809 _ReleaseLocks(self, locking.LEVEL_NODE_RES)
12811 # Apply NIC changes
12812 if self._new_nics is not None:
12813 instance.nics = self._new_nics
12814 result.extend(self._nic_chgdesc)
12817 if self.op.hvparams:
12818 instance.hvparams = self.hv_inst
12819 for key, val in self.op.hvparams.iteritems():
12820 result.append(("hv/%s" % key, val))
12823 if self.op.beparams:
12824 instance.beparams = self.be_inst
12825 for key, val in self.op.beparams.iteritems():
12826 result.append(("be/%s" % key, val))
12829 if self.op.os_name:
12830 instance.os = self.op.os_name
12833 if self.op.osparams:
12834 instance.osparams = self.os_inst
12835 for key, val in self.op.osparams.iteritems():
12836 result.append(("os/%s" % key, val))
12838 if self.op.offline is None:
12841 elif self.op.offline:
12842 # Mark instance as offline
12843 self.cfg.MarkInstanceOffline(instance.name)
12844 result.append(("admin_state", constants.ADMINST_OFFLINE))
12846 # Mark instance as online, but stopped
12847 self.cfg.MarkInstanceDown(instance.name)
12848 result.append(("admin_state", constants.ADMINST_DOWN))
12850 self.cfg.Update(instance, feedback_fn)
12852 assert not (self.owned_locks(locking.LEVEL_NODE_RES) or
12853 self.owned_locks(locking.LEVEL_NODE)), \
12854 "All node locks should have been released by now"
12858 _DISK_CONVERSIONS = {
12859 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
12860 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
12864 class LUInstanceChangeGroup(LogicalUnit):
12865 HPATH = "instance-change-group"
12866 HTYPE = constants.HTYPE_INSTANCE
12869 def ExpandNames(self):
12870 self.share_locks = _ShareAll()
12871 self.needed_locks = {
12872 locking.LEVEL_NODEGROUP: [],
12873 locking.LEVEL_NODE: [],
12876 self._ExpandAndLockInstance()
12878 if self.op.target_groups:
12879 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
12880 self.op.target_groups)
12882 self.req_target_uuids = None
12884 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
12886 def DeclareLocks(self, level):
12887 if level == locking.LEVEL_NODEGROUP:
12888 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
12890 if self.req_target_uuids:
12891 lock_groups = set(self.req_target_uuids)
12893 # Lock all groups used by instance optimistically; this requires going
12894 # via the node before it's locked, requiring verification later on
12895 instance_groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
12896 lock_groups.update(instance_groups)
12898 # No target groups, need to lock all of them
12899 lock_groups = locking.ALL_SET
12901 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
12903 elif level == locking.LEVEL_NODE:
12904 if self.req_target_uuids:
12905 # Lock all nodes used by instances
12906 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
12907 self._LockInstancesNodes()
12909 # Lock all nodes in all potential target groups
12910 lock_groups = (frozenset(self.owned_locks(locking.LEVEL_NODEGROUP)) -
12911 self.cfg.GetInstanceNodeGroups(self.op.instance_name))
12912 member_nodes = [node_name
12913 for group in lock_groups
12914 for node_name in self.cfg.GetNodeGroup(group).members]
12915 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
12917 # Lock all nodes as all groups are potential targets
12918 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
12920 def CheckPrereq(self):
12921 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
12922 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
12923 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
12925 assert (self.req_target_uuids is None or
12926 owned_groups.issuperset(self.req_target_uuids))
12927 assert owned_instances == set([self.op.instance_name])
12929 # Get instance information
12930 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
12932 # Check if node groups for locked instance are still correct
12933 assert owned_nodes.issuperset(self.instance.all_nodes), \
12934 ("Instance %s's nodes changed while we kept the lock" %
12935 self.op.instance_name)
12937 inst_groups = _CheckInstanceNodeGroups(self.cfg, self.op.instance_name,
12940 if self.req_target_uuids:
12941 # User requested specific target groups
12942 self.target_uuids = frozenset(self.req_target_uuids)
12944 # All groups except those used by the instance are potential targets
12945 self.target_uuids = owned_groups - inst_groups
12947 conflicting_groups = self.target_uuids & inst_groups
12948 if conflicting_groups:
12949 raise errors.OpPrereqError("Can't use group(s) '%s' as targets, they are"
12950 " used by the instance '%s'" %
12951 (utils.CommaJoin(conflicting_groups),
12952 self.op.instance_name),
12953 errors.ECODE_INVAL)
12955 if not self.target_uuids:
12956 raise errors.OpPrereqError("There are no possible target groups",
12957 errors.ECODE_INVAL)
12959 def BuildHooksEnv(self):
12960 """Build hooks env.
12963 assert self.target_uuids
12966 "TARGET_GROUPS": " ".join(self.target_uuids),
12969 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
12973 def BuildHooksNodes(self):
12974 """Build hooks nodes.
12977 mn = self.cfg.GetMasterNode()
12978 return ([mn], [mn])
12980 def Exec(self, feedback_fn):
12981 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
12983 assert instances == [self.op.instance_name], "Instance not locked"
12985 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
12986 instances=instances, target_groups=list(self.target_uuids))
12988 ial.Run(self.op.iallocator)
12990 if not ial.success:
12991 raise errors.OpPrereqError("Can't compute solution for changing group of"
12992 " instance '%s' using iallocator '%s': %s" %
12993 (self.op.instance_name, self.op.iallocator,
12995 errors.ECODE_NORES)
12997 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
12999 self.LogInfo("Iallocator returned %s job(s) for changing group of"
13000 " instance '%s'", len(jobs), self.op.instance_name)
13002 return ResultWithJobs(jobs)
13005 class LUBackupQuery(NoHooksLU):
13006 """Query the exports list
13011 def CheckArguments(self):
13012 self.expq = _ExportQuery(qlang.MakeSimpleFilter("node", self.op.nodes),
13013 ["node", "export"], self.op.use_locking)
13015 def ExpandNames(self):
13016 self.expq.ExpandNames(self)
13018 def DeclareLocks(self, level):
13019 self.expq.DeclareLocks(self, level)
13021 def Exec(self, feedback_fn):
13024 for (node, expname) in self.expq.OldStyleQuery(self):
13025 if expname is None:
13026 result[node] = False
13028 result.setdefault(node, []).append(expname)
13033 class _ExportQuery(_QueryBase):
13034 FIELDS = query.EXPORT_FIELDS
13036 #: The node name is not a unique key for this query
13037 SORT_FIELD = "node"
13039 def ExpandNames(self, lu):
13040 lu.needed_locks = {}
13042 # The following variables interact with _QueryBase._GetNames
13044 self.wanted = _GetWantedNodes(lu, self.names)
13046 self.wanted = locking.ALL_SET
13048 self.do_locking = self.use_locking
13050 if self.do_locking:
13051 lu.share_locks = _ShareAll()
13052 lu.needed_locks = {
13053 locking.LEVEL_NODE: self.wanted,
13056 def DeclareLocks(self, lu, level):
13059 def _GetQueryData(self, lu):
13060 """Computes the list of nodes and their attributes.
13063 # Locking is not used
13065 assert not (compat.any(lu.glm.is_owned(level)
13066 for level in locking.LEVELS
13067 if level != locking.LEVEL_CLUSTER) or
13068 self.do_locking or self.use_locking)
13070 nodes = self._GetNames(lu, lu.cfg.GetNodeList(), locking.LEVEL_NODE)
13074 for (node, nres) in lu.rpc.call_export_list(nodes).items():
13076 result.append((node, None))
13078 result.extend((node, expname) for expname in nres.payload)
13083 class LUBackupPrepare(NoHooksLU):
13084 """Prepares an instance for an export and returns useful information.
13089 def ExpandNames(self):
13090 self._ExpandAndLockInstance()
13092 def CheckPrereq(self):
13093 """Check prerequisites.
13096 instance_name = self.op.instance_name
13098 self.instance = self.cfg.GetInstanceInfo(instance_name)
13099 assert self.instance is not None, \
13100 "Cannot retrieve locked instance %s" % self.op.instance_name
13101 _CheckNodeOnline(self, self.instance.primary_node)
13103 self._cds = _GetClusterDomainSecret()
13105 def Exec(self, feedback_fn):
13106 """Prepares an instance for an export.
13109 instance = self.instance
13111 if self.op.mode == constants.EXPORT_MODE_REMOTE:
13112 salt = utils.GenerateSecret(8)
13114 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
13115 result = self.rpc.call_x509_cert_create(instance.primary_node,
13116 constants.RIE_CERT_VALIDITY)
13117 result.Raise("Can't create X509 key and certificate on %s" % result.node)
13119 (name, cert_pem) = result.payload
13121 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
13125 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
13126 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
13128 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
13134 class LUBackupExport(LogicalUnit):
13135 """Export an instance to an image in the cluster.
13138 HPATH = "instance-export"
13139 HTYPE = constants.HTYPE_INSTANCE
13142 def CheckArguments(self):
13143 """Check the arguments.
13146 self.x509_key_name = self.op.x509_key_name
13147 self.dest_x509_ca_pem = self.op.destination_x509_ca
13149 if self.op.mode == constants.EXPORT_MODE_REMOTE:
13150 if not self.x509_key_name:
13151 raise errors.OpPrereqError("Missing X509 key name for encryption",
13152 errors.ECODE_INVAL)
13154 if not self.dest_x509_ca_pem:
13155 raise errors.OpPrereqError("Missing destination X509 CA",
13156 errors.ECODE_INVAL)
13158 def ExpandNames(self):
13159 self._ExpandAndLockInstance()
13161 # Lock all nodes for local exports
13162 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13163 # FIXME: lock only instance primary and destination node
13165 # Sad but true, for now we have do lock all nodes, as we don't know where
13166 # the previous export might be, and in this LU we search for it and
13167 # remove it from its current node. In the future we could fix this by:
13168 # - making a tasklet to search (share-lock all), then create the
13169 # new one, then one to remove, after
13170 # - removing the removal operation altogether
13171 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
13173 def DeclareLocks(self, level):
13174 """Last minute lock declaration."""
13175 # All nodes are locked anyway, so nothing to do here.
13177 def BuildHooksEnv(self):
13178 """Build hooks env.
13180 This will run on the master, primary node and target node.
13184 "EXPORT_MODE": self.op.mode,
13185 "EXPORT_NODE": self.op.target_node,
13186 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
13187 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
13188 # TODO: Generic function for boolean env variables
13189 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
13192 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
13196 def BuildHooksNodes(self):
13197 """Build hooks nodes.
13200 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
13202 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13203 nl.append(self.op.target_node)
13207 def CheckPrereq(self):
13208 """Check prerequisites.
13210 This checks that the instance and node names are valid.
13213 instance_name = self.op.instance_name
13215 self.instance = self.cfg.GetInstanceInfo(instance_name)
13216 assert self.instance is not None, \
13217 "Cannot retrieve locked instance %s" % self.op.instance_name
13218 _CheckNodeOnline(self, self.instance.primary_node)
13220 if (self.op.remove_instance and
13221 self.instance.admin_state == constants.ADMINST_UP and
13222 not self.op.shutdown):
13223 raise errors.OpPrereqError("Can not remove instance without shutting it"
13226 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13227 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
13228 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
13229 assert self.dst_node is not None
13231 _CheckNodeOnline(self, self.dst_node.name)
13232 _CheckNodeNotDrained(self, self.dst_node.name)
13235 self.dest_disk_info = None
13236 self.dest_x509_ca = None
13238 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
13239 self.dst_node = None
13241 if len(self.op.target_node) != len(self.instance.disks):
13242 raise errors.OpPrereqError(("Received destination information for %s"
13243 " disks, but instance %s has %s disks") %
13244 (len(self.op.target_node), instance_name,
13245 len(self.instance.disks)),
13246 errors.ECODE_INVAL)
13248 cds = _GetClusterDomainSecret()
13250 # Check X509 key name
13252 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
13253 except (TypeError, ValueError), err:
13254 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
13256 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
13257 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
13258 errors.ECODE_INVAL)
13260 # Load and verify CA
13262 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
13263 except OpenSSL.crypto.Error, err:
13264 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
13265 (err, ), errors.ECODE_INVAL)
13267 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
13268 if errcode is not None:
13269 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
13270 (msg, ), errors.ECODE_INVAL)
13272 self.dest_x509_ca = cert
13274 # Verify target information
13276 for idx, disk_data in enumerate(self.op.target_node):
13278 (host, port, magic) = \
13279 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
13280 except errors.GenericError, err:
13281 raise errors.OpPrereqError("Target info for disk %s: %s" %
13282 (idx, err), errors.ECODE_INVAL)
13284 disk_info.append((host, port, magic))
13286 assert len(disk_info) == len(self.op.target_node)
13287 self.dest_disk_info = disk_info
13290 raise errors.ProgrammerError("Unhandled export mode %r" %
13293 # instance disk type verification
13294 # TODO: Implement export support for file-based disks
13295 for disk in self.instance.disks:
13296 if disk.dev_type == constants.LD_FILE:
13297 raise errors.OpPrereqError("Export not supported for instances with"
13298 " file-based disks", errors.ECODE_INVAL)
13300 def _CleanupExports(self, feedback_fn):
13301 """Removes exports of current instance from all other nodes.
13303 If an instance in a cluster with nodes A..D was exported to node C, its
13304 exports will be removed from the nodes A, B and D.
13307 assert self.op.mode != constants.EXPORT_MODE_REMOTE
13309 nodelist = self.cfg.GetNodeList()
13310 nodelist.remove(self.dst_node.name)
13312 # on one-node clusters nodelist will be empty after the removal
13313 # if we proceed the backup would be removed because OpBackupQuery
13314 # substitutes an empty list with the full cluster node list.
13315 iname = self.instance.name
13317 feedback_fn("Removing old exports for instance %s" % iname)
13318 exportlist = self.rpc.call_export_list(nodelist)
13319 for node in exportlist:
13320 if exportlist[node].fail_msg:
13322 if iname in exportlist[node].payload:
13323 msg = self.rpc.call_export_remove(node, iname).fail_msg
13325 self.LogWarning("Could not remove older export for instance %s"
13326 " on node %s: %s", iname, node, msg)
13328 def Exec(self, feedback_fn):
13329 """Export an instance to an image in the cluster.
13332 assert self.op.mode in constants.EXPORT_MODES
13334 instance = self.instance
13335 src_node = instance.primary_node
13337 if self.op.shutdown:
13338 # shutdown the instance, but not the disks
13339 feedback_fn("Shutting down instance %s" % instance.name)
13340 result = self.rpc.call_instance_shutdown(src_node, instance,
13341 self.op.shutdown_timeout)
13342 # TODO: Maybe ignore failures if ignore_remove_failures is set
13343 result.Raise("Could not shutdown instance %s on"
13344 " node %s" % (instance.name, src_node))
13346 # set the disks ID correctly since call_instance_start needs the
13347 # correct drbd minor to create the symlinks
13348 for disk in instance.disks:
13349 self.cfg.SetDiskID(disk, src_node)
13351 activate_disks = (instance.admin_state != constants.ADMINST_UP)
13354 # Activate the instance disks if we'exporting a stopped instance
13355 feedback_fn("Activating disks for %s" % instance.name)
13356 _StartInstanceDisks(self, instance, None)
13359 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
13362 helper.CreateSnapshots()
13364 if (self.op.shutdown and
13365 instance.admin_state == constants.ADMINST_UP and
13366 not self.op.remove_instance):
13367 assert not activate_disks
13368 feedback_fn("Starting instance %s" % instance.name)
13369 result = self.rpc.call_instance_start(src_node,
13370 (instance, None, None), False)
13371 msg = result.fail_msg
13373 feedback_fn("Failed to start instance: %s" % msg)
13374 _ShutdownInstanceDisks(self, instance)
13375 raise errors.OpExecError("Could not start instance: %s" % msg)
13377 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13378 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
13379 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
13380 connect_timeout = constants.RIE_CONNECT_TIMEOUT
13381 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
13383 (key_name, _, _) = self.x509_key_name
13386 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
13389 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
13390 key_name, dest_ca_pem,
13395 # Check for backwards compatibility
13396 assert len(dresults) == len(instance.disks)
13397 assert compat.all(isinstance(i, bool) for i in dresults), \
13398 "Not all results are boolean: %r" % dresults
13402 feedback_fn("Deactivating disks for %s" % instance.name)
13403 _ShutdownInstanceDisks(self, instance)
13405 if not (compat.all(dresults) and fin_resu):
13408 failures.append("export finalization")
13409 if not compat.all(dresults):
13410 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
13412 failures.append("disk export: disk(s) %s" % fdsk)
13414 raise errors.OpExecError("Export failed, errors in %s" %
13415 utils.CommaJoin(failures))
13417 # At this point, the export was successful, we can cleanup/finish
13419 # Remove instance if requested
13420 if self.op.remove_instance:
13421 feedback_fn("Removing instance %s" % instance.name)
13422 _RemoveInstance(self, feedback_fn, instance,
13423 self.op.ignore_remove_failures)
13425 if self.op.mode == constants.EXPORT_MODE_LOCAL:
13426 self._CleanupExports(feedback_fn)
13428 return fin_resu, dresults
13431 class LUBackupRemove(NoHooksLU):
13432 """Remove exports related to the named instance.
13437 def ExpandNames(self):
13438 self.needed_locks = {}
13439 # We need all nodes to be locked in order for RemoveExport to work, but we
13440 # don't need to lock the instance itself, as nothing will happen to it (and
13441 # we can remove exports also for a removed instance)
13442 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
13444 def Exec(self, feedback_fn):
13445 """Remove any export.
13448 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
13449 # If the instance was not found we'll try with the name that was passed in.
13450 # This will only work if it was an FQDN, though.
13452 if not instance_name:
13454 instance_name = self.op.instance_name
13456 locked_nodes = self.owned_locks(locking.LEVEL_NODE)
13457 exportlist = self.rpc.call_export_list(locked_nodes)
13459 for node in exportlist:
13460 msg = exportlist[node].fail_msg
13462 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
13464 if instance_name in exportlist[node].payload:
13466 result = self.rpc.call_export_remove(node, instance_name)
13467 msg = result.fail_msg
13469 logging.error("Could not remove export for instance %s"
13470 " on node %s: %s", instance_name, node, msg)
13472 if fqdn_warn and not found:
13473 feedback_fn("Export not found. If trying to remove an export belonging"
13474 " to a deleted instance please use its Fully Qualified"
13478 class LUGroupAdd(LogicalUnit):
13479 """Logical unit for creating node groups.
13482 HPATH = "group-add"
13483 HTYPE = constants.HTYPE_GROUP
13486 def ExpandNames(self):
13487 # We need the new group's UUID here so that we can create and acquire the
13488 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
13489 # that it should not check whether the UUID exists in the configuration.
13490 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
13491 self.needed_locks = {}
13492 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
13494 def CheckPrereq(self):
13495 """Check prerequisites.
13497 This checks that the given group name is not an existing node group
13502 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13503 except errors.OpPrereqError:
13506 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
13507 " node group (UUID: %s)" %
13508 (self.op.group_name, existing_uuid),
13509 errors.ECODE_EXISTS)
13511 if self.op.ndparams:
13512 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
13514 if self.op.hv_state:
13515 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state, None)
13517 self.new_hv_state = None
13519 if self.op.disk_state:
13520 self.new_disk_state = _MergeAndVerifyDiskState(self.op.disk_state, None)
13522 self.new_disk_state = None
13524 if self.op.diskparams:
13525 for templ in constants.DISK_TEMPLATES:
13526 if templ not in self.op.diskparams:
13527 self.op.diskparams[templ] = {}
13528 utils.ForceDictType(self.op.diskparams[templ], constants.DISK_DT_TYPES)
13530 self.op.diskparams = self.cfg.GetClusterInfo().diskparams
13532 if self.op.ipolicy:
13533 cluster = self.cfg.GetClusterInfo()
13534 full_ipolicy = cluster.SimpleFillIPolicy(self.op.ipolicy)
13536 objects.InstancePolicy.CheckParameterSyntax(full_ipolicy)
13537 except errors.ConfigurationError, err:
13538 raise errors.OpPrereqError("Invalid instance policy: %s" % err,
13539 errors.ECODE_INVAL)
13541 def BuildHooksEnv(self):
13542 """Build hooks env.
13546 "GROUP_NAME": self.op.group_name,
13549 def BuildHooksNodes(self):
13550 """Build hooks nodes.
13553 mn = self.cfg.GetMasterNode()
13554 return ([mn], [mn])
13556 def Exec(self, feedback_fn):
13557 """Add the node group to the cluster.
13560 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
13561 uuid=self.group_uuid,
13562 alloc_policy=self.op.alloc_policy,
13563 ndparams=self.op.ndparams,
13564 diskparams=self.op.diskparams,
13565 ipolicy=self.op.ipolicy,
13566 hv_state_static=self.new_hv_state,
13567 disk_state_static=self.new_disk_state)
13569 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
13570 del self.remove_locks[locking.LEVEL_NODEGROUP]
13573 class LUGroupAssignNodes(NoHooksLU):
13574 """Logical unit for assigning nodes to groups.
13579 def ExpandNames(self):
13580 # These raise errors.OpPrereqError on their own:
13581 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13582 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
13584 # We want to lock all the affected nodes and groups. We have readily
13585 # available the list of nodes, and the *destination* group. To gather the
13586 # list of "source" groups, we need to fetch node information later on.
13587 self.needed_locks = {
13588 locking.LEVEL_NODEGROUP: set([self.group_uuid]),
13589 locking.LEVEL_NODE: self.op.nodes,
13592 def DeclareLocks(self, level):
13593 if level == locking.LEVEL_NODEGROUP:
13594 assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1
13596 # Try to get all affected nodes' groups without having the group or node
13597 # lock yet. Needs verification later in the code flow.
13598 groups = self.cfg.GetNodeGroupsFromNodes(self.op.nodes)
13600 self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)
13602 def CheckPrereq(self):
13603 """Check prerequisites.
13606 assert self.needed_locks[locking.LEVEL_NODEGROUP]
13607 assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==
13608 frozenset(self.op.nodes))
13610 expected_locks = (set([self.group_uuid]) |
13611 self.cfg.GetNodeGroupsFromNodes(self.op.nodes))
13612 actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP)
13613 if actual_locks != expected_locks:
13614 raise errors.OpExecError("Nodes changed groups since locks were acquired,"
13615 " current groups are '%s', used to be '%s'" %
13616 (utils.CommaJoin(expected_locks),
13617 utils.CommaJoin(actual_locks)))
13619 self.node_data = self.cfg.GetAllNodesInfo()
13620 self.group = self.cfg.GetNodeGroup(self.group_uuid)
13621 instance_data = self.cfg.GetAllInstancesInfo()
13623 if self.group is None:
13624 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13625 (self.op.group_name, self.group_uuid))
13627 (new_splits, previous_splits) = \
13628 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
13629 for node in self.op.nodes],
13630 self.node_data, instance_data)
13633 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
13635 if not self.op.force:
13636 raise errors.OpExecError("The following instances get split by this"
13637 " change and --force was not given: %s" %
13640 self.LogWarning("This operation will split the following instances: %s",
13643 if previous_splits:
13644 self.LogWarning("In addition, these already-split instances continue"
13645 " to be split across groups: %s",
13646 utils.CommaJoin(utils.NiceSort(previous_splits)))
13648 def Exec(self, feedback_fn):
13649 """Assign nodes to a new group.
13652 mods = [(node_name, self.group_uuid) for node_name in self.op.nodes]
13654 self.cfg.AssignGroupNodes(mods)
13657 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
13658 """Check for split instances after a node assignment.
13660 This method considers a series of node assignments as an atomic operation,
13661 and returns information about split instances after applying the set of
13664 In particular, it returns information about newly split instances, and
13665 instances that were already split, and remain so after the change.
13667 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
13670 @type changes: list of (node_name, new_group_uuid) pairs.
13671 @param changes: list of node assignments to consider.
13672 @param node_data: a dict with data for all nodes
13673 @param instance_data: a dict with all instances to consider
13674 @rtype: a two-tuple
13675 @return: a list of instances that were previously okay and result split as a
13676 consequence of this change, and a list of instances that were previously
13677 split and this change does not fix.
13680 changed_nodes = dict((node, group) for node, group in changes
13681 if node_data[node].group != group)
13683 all_split_instances = set()
13684 previously_split_instances = set()
13686 def InstanceNodes(instance):
13687 return [instance.primary_node] + list(instance.secondary_nodes)
13689 for inst in instance_data.values():
13690 if inst.disk_template not in constants.DTS_INT_MIRROR:
13693 instance_nodes = InstanceNodes(inst)
13695 if len(set(node_data[node].group for node in instance_nodes)) > 1:
13696 previously_split_instances.add(inst.name)
13698 if len(set(changed_nodes.get(node, node_data[node].group)
13699 for node in instance_nodes)) > 1:
13700 all_split_instances.add(inst.name)
13702 return (list(all_split_instances - previously_split_instances),
13703 list(previously_split_instances & all_split_instances))
13706 class _GroupQuery(_QueryBase):
13707 FIELDS = query.GROUP_FIELDS
13709 def ExpandNames(self, lu):
13710 lu.needed_locks = {}
13712 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
13713 self._cluster = lu.cfg.GetClusterInfo()
13714 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
13717 self.wanted = [name_to_uuid[name]
13718 for name in utils.NiceSort(name_to_uuid.keys())]
13720 # Accept names to be either names or UUIDs.
13723 all_uuid = frozenset(self._all_groups.keys())
13725 for name in self.names:
13726 if name in all_uuid:
13727 self.wanted.append(name)
13728 elif name in name_to_uuid:
13729 self.wanted.append(name_to_uuid[name])
13731 missing.append(name)
13734 raise errors.OpPrereqError("Some groups do not exist: %s" %
13735 utils.CommaJoin(missing),
13736 errors.ECODE_NOENT)
13738 def DeclareLocks(self, lu, level):
13741 def _GetQueryData(self, lu):
13742 """Computes the list of node groups and their attributes.
13745 do_nodes = query.GQ_NODE in self.requested_data
13746 do_instances = query.GQ_INST in self.requested_data
13748 group_to_nodes = None
13749 group_to_instances = None
13751 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
13752 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
13753 # latter GetAllInstancesInfo() is not enough, for we have to go through
13754 # instance->node. Hence, we will need to process nodes even if we only need
13755 # instance information.
13756 if do_nodes or do_instances:
13757 all_nodes = lu.cfg.GetAllNodesInfo()
13758 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
13761 for node in all_nodes.values():
13762 if node.group in group_to_nodes:
13763 group_to_nodes[node.group].append(node.name)
13764 node_to_group[node.name] = node.group
13767 all_instances = lu.cfg.GetAllInstancesInfo()
13768 group_to_instances = dict((uuid, []) for uuid in self.wanted)
13770 for instance in all_instances.values():
13771 node = instance.primary_node
13772 if node in node_to_group:
13773 group_to_instances[node_to_group[node]].append(instance.name)
13776 # Do not pass on node information if it was not requested.
13777 group_to_nodes = None
13779 return query.GroupQueryData(self._cluster,
13780 [self._all_groups[uuid]
13781 for uuid in self.wanted],
13782 group_to_nodes, group_to_instances)
13785 class LUGroupQuery(NoHooksLU):
13786 """Logical unit for querying node groups.
13791 def CheckArguments(self):
13792 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
13793 self.op.output_fields, False)
13795 def ExpandNames(self):
13796 self.gq.ExpandNames(self)
13798 def DeclareLocks(self, level):
13799 self.gq.DeclareLocks(self, level)
13801 def Exec(self, feedback_fn):
13802 return self.gq.OldStyleQuery(self)
13805 class LUGroupSetParams(LogicalUnit):
13806 """Modifies the parameters of a node group.
13809 HPATH = "group-modify"
13810 HTYPE = constants.HTYPE_GROUP
13813 def CheckArguments(self):
13816 self.op.diskparams,
13817 self.op.alloc_policy,
13819 self.op.disk_state,
13823 if all_changes.count(None) == len(all_changes):
13824 raise errors.OpPrereqError("Please pass at least one modification",
13825 errors.ECODE_INVAL)
13827 def ExpandNames(self):
13828 # This raises errors.OpPrereqError on its own:
13829 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13831 self.needed_locks = {
13832 locking.LEVEL_INSTANCE: [],
13833 locking.LEVEL_NODEGROUP: [self.group_uuid],
13836 self.share_locks[locking.LEVEL_INSTANCE] = 1
13838 def DeclareLocks(self, level):
13839 if level == locking.LEVEL_INSTANCE:
13840 assert not self.needed_locks[locking.LEVEL_INSTANCE]
13842 # Lock instances optimistically, needs verification once group lock has
13844 self.needed_locks[locking.LEVEL_INSTANCE] = \
13845 self.cfg.GetNodeGroupInstances(self.group_uuid)
13847 def CheckPrereq(self):
13848 """Check prerequisites.
13851 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
13853 # Check if locked instances are still correct
13854 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
13856 self.group = self.cfg.GetNodeGroup(self.group_uuid)
13857 cluster = self.cfg.GetClusterInfo()
13859 if self.group is None:
13860 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
13861 (self.op.group_name, self.group_uuid))
13863 if self.op.ndparams:
13864 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
13865 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
13866 self.new_ndparams = new_ndparams
13868 if self.op.diskparams:
13869 self.new_diskparams = dict()
13870 for templ in constants.DISK_TEMPLATES:
13871 if templ not in self.op.diskparams:
13872 self.op.diskparams[templ] = {}
13873 new_templ_params = _GetUpdatedParams(self.group.diskparams[templ],
13874 self.op.diskparams[templ])
13875 utils.ForceDictType(new_templ_params, constants.DISK_DT_TYPES)
13876 self.new_diskparams[templ] = new_templ_params
13878 if self.op.hv_state:
13879 self.new_hv_state = _MergeAndVerifyHvState(self.op.hv_state,
13880 self.group.hv_state_static)
13882 if self.op.disk_state:
13883 self.new_disk_state = \
13884 _MergeAndVerifyDiskState(self.op.disk_state,
13885 self.group.disk_state_static)
13887 if self.op.ipolicy:
13888 self.new_ipolicy = _GetUpdatedIPolicy(self.group.ipolicy,
13892 new_ipolicy = cluster.SimpleFillIPolicy(self.new_ipolicy)
13893 inst_filter = lambda inst: inst.name in owned_instances
13894 instances = self.cfg.GetInstancesInfoByFilter(inst_filter).values()
13896 _ComputeNewInstanceViolations(_CalculateGroupIPolicy(cluster,
13898 new_ipolicy, instances)
13901 self.LogWarning("After the ipolicy change the following instances"
13902 " violate them: %s",
13903 utils.CommaJoin(violations))
13905 def BuildHooksEnv(self):
13906 """Build hooks env.
13910 "GROUP_NAME": self.op.group_name,
13911 "NEW_ALLOC_POLICY": self.op.alloc_policy,
13914 def BuildHooksNodes(self):
13915 """Build hooks nodes.
13918 mn = self.cfg.GetMasterNode()
13919 return ([mn], [mn])
13921 def Exec(self, feedback_fn):
13922 """Modifies the node group.
13927 if self.op.ndparams:
13928 self.group.ndparams = self.new_ndparams
13929 result.append(("ndparams", str(self.group.ndparams)))
13931 if self.op.diskparams:
13932 self.group.diskparams = self.new_diskparams
13933 result.append(("diskparams", str(self.group.diskparams)))
13935 if self.op.alloc_policy:
13936 self.group.alloc_policy = self.op.alloc_policy
13938 if self.op.hv_state:
13939 self.group.hv_state_static = self.new_hv_state
13941 if self.op.disk_state:
13942 self.group.disk_state_static = self.new_disk_state
13944 if self.op.ipolicy:
13945 self.group.ipolicy = self.new_ipolicy
13947 self.cfg.Update(self.group, feedback_fn)
13951 class LUGroupRemove(LogicalUnit):
13952 HPATH = "group-remove"
13953 HTYPE = constants.HTYPE_GROUP
13956 def ExpandNames(self):
13957 # This will raises errors.OpPrereqError on its own:
13958 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
13959 self.needed_locks = {
13960 locking.LEVEL_NODEGROUP: [self.group_uuid],
13963 def CheckPrereq(self):
13964 """Check prerequisites.
13966 This checks that the given group name exists as a node group, that is
13967 empty (i.e., contains no nodes), and that is not the last group of the
13971 # Verify that the group is empty.
13972 group_nodes = [node.name
13973 for node in self.cfg.GetAllNodesInfo().values()
13974 if node.group == self.group_uuid]
13977 raise errors.OpPrereqError("Group '%s' not empty, has the following"
13979 (self.op.group_name,
13980 utils.CommaJoin(utils.NiceSort(group_nodes))),
13981 errors.ECODE_STATE)
13983 # Verify the cluster would not be left group-less.
13984 if len(self.cfg.GetNodeGroupList()) == 1:
13985 raise errors.OpPrereqError("Group '%s' is the only group,"
13986 " cannot be removed" %
13987 self.op.group_name,
13988 errors.ECODE_STATE)
13990 def BuildHooksEnv(self):
13991 """Build hooks env.
13995 "GROUP_NAME": self.op.group_name,
13998 def BuildHooksNodes(self):
13999 """Build hooks nodes.
14002 mn = self.cfg.GetMasterNode()
14003 return ([mn], [mn])
14005 def Exec(self, feedback_fn):
14006 """Remove the node group.
14010 self.cfg.RemoveNodeGroup(self.group_uuid)
14011 except errors.ConfigurationError:
14012 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
14013 (self.op.group_name, self.group_uuid))
14015 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
14018 class LUGroupRename(LogicalUnit):
14019 HPATH = "group-rename"
14020 HTYPE = constants.HTYPE_GROUP
14023 def ExpandNames(self):
14024 # This raises errors.OpPrereqError on its own:
14025 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
14027 self.needed_locks = {
14028 locking.LEVEL_NODEGROUP: [self.group_uuid],
14031 def CheckPrereq(self):
14032 """Check prerequisites.
14034 Ensures requested new name is not yet used.
14038 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
14039 except errors.OpPrereqError:
14042 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
14043 " node group (UUID: %s)" %
14044 (self.op.new_name, new_name_uuid),
14045 errors.ECODE_EXISTS)
14047 def BuildHooksEnv(self):
14048 """Build hooks env.
14052 "OLD_NAME": self.op.group_name,
14053 "NEW_NAME": self.op.new_name,
14056 def BuildHooksNodes(self):
14057 """Build hooks nodes.
14060 mn = self.cfg.GetMasterNode()
14062 all_nodes = self.cfg.GetAllNodesInfo()
14063 all_nodes.pop(mn, None)
14066 run_nodes.extend(node.name for node in all_nodes.values()
14067 if node.group == self.group_uuid)
14069 return (run_nodes, run_nodes)
14071 def Exec(self, feedback_fn):
14072 """Rename the node group.
14075 group = self.cfg.GetNodeGroup(self.group_uuid)
14078 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
14079 (self.op.group_name, self.group_uuid))
14081 group.name = self.op.new_name
14082 self.cfg.Update(group, feedback_fn)
14084 return self.op.new_name
14087 class LUGroupEvacuate(LogicalUnit):
14088 HPATH = "group-evacuate"
14089 HTYPE = constants.HTYPE_GROUP
14092 def ExpandNames(self):
14093 # This raises errors.OpPrereqError on its own:
14094 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
14096 if self.op.target_groups:
14097 self.req_target_uuids = map(self.cfg.LookupNodeGroup,
14098 self.op.target_groups)
14100 self.req_target_uuids = []
14102 if self.group_uuid in self.req_target_uuids:
14103 raise errors.OpPrereqError("Group to be evacuated (%s) can not be used"
14104 " as a target group (targets are %s)" %
14106 utils.CommaJoin(self.req_target_uuids)),
14107 errors.ECODE_INVAL)
14109 self.op.iallocator = _GetDefaultIAllocator(self.cfg, self.op.iallocator)
14111 self.share_locks = _ShareAll()
14112 self.needed_locks = {
14113 locking.LEVEL_INSTANCE: [],
14114 locking.LEVEL_NODEGROUP: [],
14115 locking.LEVEL_NODE: [],
14118 def DeclareLocks(self, level):
14119 if level == locking.LEVEL_INSTANCE:
14120 assert not self.needed_locks[locking.LEVEL_INSTANCE]
14122 # Lock instances optimistically, needs verification once node and group
14123 # locks have been acquired
14124 self.needed_locks[locking.LEVEL_INSTANCE] = \
14125 self.cfg.GetNodeGroupInstances(self.group_uuid)
14127 elif level == locking.LEVEL_NODEGROUP:
14128 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
14130 if self.req_target_uuids:
14131 lock_groups = set([self.group_uuid] + self.req_target_uuids)
14133 # Lock all groups used by instances optimistically; this requires going
14134 # via the node before it's locked, requiring verification later on
14135 lock_groups.update(group_uuid
14136 for instance_name in
14137 self.owned_locks(locking.LEVEL_INSTANCE)
14139 self.cfg.GetInstanceNodeGroups(instance_name))
14141 # No target groups, need to lock all of them
14142 lock_groups = locking.ALL_SET
14144 self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups
14146 elif level == locking.LEVEL_NODE:
14147 # This will only lock the nodes in the group to be evacuated which
14148 # contain actual instances
14149 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
14150 self._LockInstancesNodes()
14152 # Lock all nodes in group to be evacuated and target groups
14153 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
14154 assert self.group_uuid in owned_groups
14155 member_nodes = [node_name
14156 for group in owned_groups
14157 for node_name in self.cfg.GetNodeGroup(group).members]
14158 self.needed_locks[locking.LEVEL_NODE].extend(member_nodes)
14160 def CheckPrereq(self):
14161 owned_instances = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))
14162 owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))
14163 owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE))
14165 assert owned_groups.issuperset(self.req_target_uuids)
14166 assert self.group_uuid in owned_groups
14168 # Check if locked instances are still correct
14169 _CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instances)
14171 # Get instance information
14172 self.instances = dict(self.cfg.GetMultiInstanceInfo(owned_instances))
14174 # Check if node groups for locked instances are still correct
14175 _CheckInstancesNodeGroups(self.cfg, self.instances,
14176 owned_groups, owned_nodes, self.group_uuid)
14178 if self.req_target_uuids:
14179 # User requested specific target groups
14180 self.target_uuids = self.req_target_uuids
14182 # All groups except the one to be evacuated are potential targets
14183 self.target_uuids = [group_uuid for group_uuid in owned_groups
14184 if group_uuid != self.group_uuid]
14186 if not self.target_uuids:
14187 raise errors.OpPrereqError("There are no possible target groups",
14188 errors.ECODE_INVAL)
14190 def BuildHooksEnv(self):
14191 """Build hooks env.
14195 "GROUP_NAME": self.op.group_name,
14196 "TARGET_GROUPS": " ".join(self.target_uuids),
14199 def BuildHooksNodes(self):
14200 """Build hooks nodes.
14203 mn = self.cfg.GetMasterNode()
14205 assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
14207 run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members
14209 return (run_nodes, run_nodes)
14211 def Exec(self, feedback_fn):
14212 instances = list(self.owned_locks(locking.LEVEL_INSTANCE))
14214 assert self.group_uuid not in self.target_uuids
14216 ial = IAllocator(self.cfg, self.rpc, constants.IALLOCATOR_MODE_CHG_GROUP,
14217 instances=instances, target_groups=self.target_uuids)
14219 ial.Run(self.op.iallocator)
14221 if not ial.success:
14222 raise errors.OpPrereqError("Can't compute group evacuation using"
14223 " iallocator '%s': %s" %
14224 (self.op.iallocator, ial.info),
14225 errors.ECODE_NORES)
14227 jobs = _LoadNodeEvacResult(self, ial.result, self.op.early_release, False)
14229 self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s",
14230 len(jobs), self.op.group_name)
14232 return ResultWithJobs(jobs)
14235 class TagsLU(NoHooksLU): # pylint: disable=W0223
14236 """Generic tags LU.
14238 This is an abstract class which is the parent of all the other tags LUs.
14241 def ExpandNames(self):
14242 self.group_uuid = None
14243 self.needed_locks = {}
14245 if self.op.kind == constants.TAG_NODE:
14246 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
14247 lock_level = locking.LEVEL_NODE
14248 lock_name = self.op.name
14249 elif self.op.kind == constants.TAG_INSTANCE:
14250 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
14251 lock_level = locking.LEVEL_INSTANCE
14252 lock_name = self.op.name
14253 elif self.op.kind == constants.TAG_NODEGROUP:
14254 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
14255 lock_level = locking.LEVEL_NODEGROUP
14256 lock_name = self.group_uuid
14261 if lock_level and getattr(self.op, "use_locking", True):
14262 self.needed_locks[lock_level] = lock_name
14264 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
14265 # not possible to acquire the BGL based on opcode parameters)
14267 def CheckPrereq(self):
14268 """Check prerequisites.
14271 if self.op.kind == constants.TAG_CLUSTER:
14272 self.target = self.cfg.GetClusterInfo()
14273 elif self.op.kind == constants.TAG_NODE:
14274 self.target = self.cfg.GetNodeInfo(self.op.name)
14275 elif self.op.kind == constants.TAG_INSTANCE:
14276 self.target = self.cfg.GetInstanceInfo(self.op.name)
14277 elif self.op.kind == constants.TAG_NODEGROUP:
14278 self.target = self.cfg.GetNodeGroup(self.group_uuid)
14280 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
14281 str(self.op.kind), errors.ECODE_INVAL)
14284 class LUTagsGet(TagsLU):
14285 """Returns the tags of a given object.
14290 def ExpandNames(self):
14291 TagsLU.ExpandNames(self)
14293 # Share locks as this is only a read operation
14294 self.share_locks = _ShareAll()
14296 def Exec(self, feedback_fn):
14297 """Returns the tag list.
14300 return list(self.target.GetTags())
14303 class LUTagsSearch(NoHooksLU):
14304 """Searches the tags for a given pattern.
14309 def ExpandNames(self):
14310 self.needed_locks = {}
14312 def CheckPrereq(self):
14313 """Check prerequisites.
14315 This checks the pattern passed for validity by compiling it.
14319 self.re = re.compile(self.op.pattern)
14320 except re.error, err:
14321 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
14322 (self.op.pattern, err), errors.ECODE_INVAL)
14324 def Exec(self, feedback_fn):
14325 """Returns the tag list.
14329 tgts = [("/cluster", cfg.GetClusterInfo())]
14330 ilist = cfg.GetAllInstancesInfo().values()
14331 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
14332 nlist = cfg.GetAllNodesInfo().values()
14333 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
14334 tgts.extend(("/nodegroup/%s" % n.name, n)
14335 for n in cfg.GetAllNodeGroupsInfo().values())
14337 for path, target in tgts:
14338 for tag in target.GetTags():
14339 if self.re.search(tag):
14340 results.append((path, tag))
14344 class LUTagsSet(TagsLU):
14345 """Sets a tag on a given object.
14350 def CheckPrereq(self):
14351 """Check prerequisites.
14353 This checks the type and length of the tag name and value.
14356 TagsLU.CheckPrereq(self)
14357 for tag in self.op.tags:
14358 objects.TaggableObject.ValidateTag(tag)
14360 def Exec(self, feedback_fn):
14365 for tag in self.op.tags:
14366 self.target.AddTag(tag)
14367 except errors.TagError, err:
14368 raise errors.OpExecError("Error while setting tag: %s" % str(err))
14369 self.cfg.Update(self.target, feedback_fn)
14372 class LUTagsDel(TagsLU):
14373 """Delete a list of tags from a given object.
14378 def CheckPrereq(self):
14379 """Check prerequisites.
14381 This checks that we have the given tag.
14384 TagsLU.CheckPrereq(self)
14385 for tag in self.op.tags:
14386 objects.TaggableObject.ValidateTag(tag)
14387 del_tags = frozenset(self.op.tags)
14388 cur_tags = self.target.GetTags()
14390 diff_tags = del_tags - cur_tags
14392 diff_names = ("'%s'" % i for i in sorted(diff_tags))
14393 raise errors.OpPrereqError("Tag(s) %s not found" %
14394 (utils.CommaJoin(diff_names), ),
14395 errors.ECODE_NOENT)
14397 def Exec(self, feedback_fn):
14398 """Remove the tag from the object.
14401 for tag in self.op.tags:
14402 self.target.RemoveTag(tag)
14403 self.cfg.Update(self.target, feedback_fn)
14406 class LUTestDelay(NoHooksLU):
14407 """Sleep for a specified amount of time.
14409 This LU sleeps on the master and/or nodes for a specified amount of
14415 def ExpandNames(self):
14416 """Expand names and set required locks.
14418 This expands the node list, if any.
14421 self.needed_locks = {}
14422 if self.op.on_nodes:
14423 # _GetWantedNodes can be used here, but is not always appropriate to use
14424 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
14425 # more information.
14426 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
14427 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
14429 def _TestDelay(self):
14430 """Do the actual sleep.
14433 if self.op.on_master:
14434 if not utils.TestDelay(self.op.duration):
14435 raise errors.OpExecError("Error during master delay test")
14436 if self.op.on_nodes:
14437 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
14438 for node, node_result in result.items():
14439 node_result.Raise("Failure during rpc call to node %s" % node)
14441 def Exec(self, feedback_fn):
14442 """Execute the test delay opcode, with the wanted repetitions.
14445 if self.op.repeat == 0:
14448 top_value = self.op.repeat - 1
14449 for i in range(self.op.repeat):
14450 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
14454 class LUTestJqueue(NoHooksLU):
14455 """Utility LU to test some aspects of the job queue.
14460 # Must be lower than default timeout for WaitForJobChange to see whether it
14461 # notices changed jobs
14462 _CLIENT_CONNECT_TIMEOUT = 20.0
14463 _CLIENT_CONFIRM_TIMEOUT = 60.0
14466 def _NotifyUsingSocket(cls, cb, errcls):
14467 """Opens a Unix socket and waits for another program to connect.
14470 @param cb: Callback to send socket name to client
14471 @type errcls: class
14472 @param errcls: Exception class to use for errors
14475 # Using a temporary directory as there's no easy way to create temporary
14476 # sockets without writing a custom loop around tempfile.mktemp and
14478 tmpdir = tempfile.mkdtemp()
14480 tmpsock = utils.PathJoin(tmpdir, "sock")
14482 logging.debug("Creating temporary socket at %s", tmpsock)
14483 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
14488 # Send details to client
14491 # Wait for client to connect before continuing
14492 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
14494 (conn, _) = sock.accept()
14495 except socket.error, err:
14496 raise errcls("Client didn't connect in time (%s)" % err)
14500 # Remove as soon as client is connected
14501 shutil.rmtree(tmpdir)
14503 # Wait for client to close
14506 # pylint: disable=E1101
14507 # Instance of '_socketobject' has no ... member
14508 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
14510 except socket.error, err:
14511 raise errcls("Client failed to confirm notification (%s)" % err)
14515 def _SendNotification(self, test, arg, sockname):
14516 """Sends a notification to the client.
14519 @param test: Test name
14520 @param arg: Test argument (depends on test)
14521 @type sockname: string
14522 @param sockname: Socket path
14525 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
14527 def _Notify(self, prereq, test, arg):
14528 """Notifies the client of a test.
14531 @param prereq: Whether this is a prereq-phase test
14533 @param test: Test name
14534 @param arg: Test argument (depends on test)
14538 errcls = errors.OpPrereqError
14540 errcls = errors.OpExecError
14542 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
14546 def CheckArguments(self):
14547 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
14548 self.expandnames_calls = 0
14550 def ExpandNames(self):
14551 checkargs_calls = getattr(self, "checkargs_calls", 0)
14552 if checkargs_calls < 1:
14553 raise errors.ProgrammerError("CheckArguments was not called")
14555 self.expandnames_calls += 1
14557 if self.op.notify_waitlock:
14558 self._Notify(True, constants.JQT_EXPANDNAMES, None)
14560 self.LogInfo("Expanding names")
14562 # Get lock on master node (just to get a lock, not for a particular reason)
14563 self.needed_locks = {
14564 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
14567 def Exec(self, feedback_fn):
14568 if self.expandnames_calls < 1:
14569 raise errors.ProgrammerError("ExpandNames was not called")
14571 if self.op.notify_exec:
14572 self._Notify(False, constants.JQT_EXEC, None)
14574 self.LogInfo("Executing")
14576 if self.op.log_messages:
14577 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
14578 for idx, msg in enumerate(self.op.log_messages):
14579 self.LogInfo("Sending log message %s", idx + 1)
14580 feedback_fn(constants.JQT_MSGPREFIX + msg)
14581 # Report how many test messages have been sent
14582 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
14585 raise errors.OpExecError("Opcode failure was requested")
14590 class IAllocator(object):
14591 """IAllocator framework.
14593 An IAllocator instance has three sets of attributes:
14594 - cfg that is needed to query the cluster
14595 - input data (all members of the _KEYS class attribute are required)
14596 - four buffer attributes (in|out_data|text), that represent the
14597 input (to the external script) in text and data structure format,
14598 and the output from it, again in two formats
14599 - the result variables from the script (success, info, nodes) for
14603 # pylint: disable=R0902
14604 # lots of instance attributes
14606 def __init__(self, cfg, rpc_runner, mode, **kwargs):
14608 self.rpc = rpc_runner
14609 # init buffer variables
14610 self.in_text = self.out_text = self.in_data = self.out_data = None
14611 # init all input fields so that pylint is happy
14613 self.memory = self.disks = self.disk_template = self.spindle_use = None
14614 self.os = self.tags = self.nics = self.vcpus = None
14615 self.hypervisor = None
14616 self.relocate_from = None
14618 self.instances = None
14619 self.evac_mode = None
14620 self.target_groups = []
14622 self.required_nodes = None
14623 # init result fields
14624 self.success = self.info = self.result = None
14627 (fn, keydata, self._result_check) = self._MODE_DATA[self.mode]
14629 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
14630 " IAllocator" % self.mode)
14632 keyset = [n for (n, _) in keydata]
14635 if key not in keyset:
14636 raise errors.ProgrammerError("Invalid input parameter '%s' to"
14637 " IAllocator" % key)
14638 setattr(self, key, kwargs[key])
14641 if key not in kwargs:
14642 raise errors.ProgrammerError("Missing input parameter '%s' to"
14643 " IAllocator" % key)
14644 self._BuildInputData(compat.partial(fn, self), keydata)
14646 def _ComputeClusterData(self):
14647 """Compute the generic allocator input data.
14649 This is the data that is independent of the actual operation.
14653 cluster_info = cfg.GetClusterInfo()
14656 "version": constants.IALLOCATOR_VERSION,
14657 "cluster_name": cfg.GetClusterName(),
14658 "cluster_tags": list(cluster_info.GetTags()),
14659 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
14660 "ipolicy": cluster_info.ipolicy,
14662 ninfo = cfg.GetAllNodesInfo()
14663 iinfo = cfg.GetAllInstancesInfo().values()
14664 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
14667 node_list = [n.name for n in ninfo.values() if n.vm_capable]
14669 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
14670 hypervisor_name = self.hypervisor
14671 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
14672 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
14674 hypervisor_name = cluster_info.primary_hypervisor
14676 node_data = self.rpc.call_node_info(node_list, [cfg.GetVGName()],
14679 self.rpc.call_all_instances_info(node_list,
14680 cluster_info.enabled_hypervisors)
14682 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
14684 config_ndata = self._ComputeBasicNodeData(cfg, ninfo)
14685 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
14686 i_list, config_ndata)
14687 assert len(data["nodes"]) == len(ninfo), \
14688 "Incomplete node data computed"
14690 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
14692 self.in_data = data
14695 def _ComputeNodeGroupData(cfg):
14696 """Compute node groups data.
14699 cluster = cfg.GetClusterInfo()
14700 ng = dict((guuid, {
14701 "name": gdata.name,
14702 "alloc_policy": gdata.alloc_policy,
14703 "ipolicy": _CalculateGroupIPolicy(cluster, gdata),
14705 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items())
14710 def _ComputeBasicNodeData(cfg, node_cfg):
14711 """Compute global node data.
14714 @returns: a dict of name: (node dict, node config)
14717 # fill in static (config-based) values
14718 node_results = dict((ninfo.name, {
14719 "tags": list(ninfo.GetTags()),
14720 "primary_ip": ninfo.primary_ip,
14721 "secondary_ip": ninfo.secondary_ip,
14722 "offline": ninfo.offline,
14723 "drained": ninfo.drained,
14724 "master_candidate": ninfo.master_candidate,
14725 "group": ninfo.group,
14726 "master_capable": ninfo.master_capable,
14727 "vm_capable": ninfo.vm_capable,
14728 "ndparams": cfg.GetNdParams(ninfo),
14730 for ninfo in node_cfg.values())
14732 return node_results
14735 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
14737 """Compute global node data.
14739 @param node_results: the basic node structures as filled from the config
14742 #TODO(dynmem): compute the right data on MAX and MIN memory
14743 # make a copy of the current dict
14744 node_results = dict(node_results)
14745 for nname, nresult in node_data.items():
14746 assert nname in node_results, "Missing basic data for node %s" % nname
14747 ninfo = node_cfg[nname]
14749 if not (ninfo.offline or ninfo.drained):
14750 nresult.Raise("Can't get data for node %s" % nname)
14751 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
14753 remote_info = _MakeLegacyNodeInfo(nresult.payload)
14755 for attr in ["memory_total", "memory_free", "memory_dom0",
14756 "vg_size", "vg_free", "cpu_total"]:
14757 if attr not in remote_info:
14758 raise errors.OpExecError("Node '%s' didn't return attribute"
14759 " '%s'" % (nname, attr))
14760 if not isinstance(remote_info[attr], int):
14761 raise errors.OpExecError("Node '%s' returned invalid value"
14763 (nname, attr, remote_info[attr]))
14764 # compute memory used by primary instances
14765 i_p_mem = i_p_up_mem = 0
14766 for iinfo, beinfo in i_list:
14767 if iinfo.primary_node == nname:
14768 i_p_mem += beinfo[constants.BE_MAXMEM]
14769 if iinfo.name not in node_iinfo[nname].payload:
14772 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]["memory"])
14773 i_mem_diff = beinfo[constants.BE_MAXMEM] - i_used_mem
14774 remote_info["memory_free"] -= max(0, i_mem_diff)
14776 if iinfo.admin_state == constants.ADMINST_UP:
14777 i_p_up_mem += beinfo[constants.BE_MAXMEM]
14779 # compute memory used by instances
14781 "total_memory": remote_info["memory_total"],
14782 "reserved_memory": remote_info["memory_dom0"],
14783 "free_memory": remote_info["memory_free"],
14784 "total_disk": remote_info["vg_size"],
14785 "free_disk": remote_info["vg_free"],
14786 "total_cpus": remote_info["cpu_total"],
14787 "i_pri_memory": i_p_mem,
14788 "i_pri_up_memory": i_p_up_mem,
14790 pnr_dyn.update(node_results[nname])
14791 node_results[nname] = pnr_dyn
14793 return node_results
14796 def _ComputeInstanceData(cluster_info, i_list):
14797 """Compute global instance data.
14801 for iinfo, beinfo in i_list:
14803 for nic in iinfo.nics:
14804 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
14808 "mode": filled_params[constants.NIC_MODE],
14809 "link": filled_params[constants.NIC_LINK],
14811 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
14812 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
14813 nic_data.append(nic_dict)
14815 "tags": list(iinfo.GetTags()),
14816 "admin_state": iinfo.admin_state,
14817 "vcpus": beinfo[constants.BE_VCPUS],
14818 "memory": beinfo[constants.BE_MAXMEM],
14819 "spindle_use": beinfo[constants.BE_SPINDLE_USE],
14821 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
14823 "disks": [{constants.IDISK_SIZE: dsk.size,
14824 constants.IDISK_MODE: dsk.mode}
14825 for dsk in iinfo.disks],
14826 "disk_template": iinfo.disk_template,
14827 "hypervisor": iinfo.hypervisor,
14829 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
14831 instance_data[iinfo.name] = pir
14833 return instance_data
14835 def _AddNewInstance(self):
14836 """Add new instance data to allocator structure.
14838 This in combination with _AllocatorGetClusterData will create the
14839 correct structure needed as input for the allocator.
14841 The checks for the completeness of the opcode must have already been
14845 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
14847 if self.disk_template in constants.DTS_INT_MIRROR:
14848 self.required_nodes = 2
14850 self.required_nodes = 1
14854 "disk_template": self.disk_template,
14857 "vcpus": self.vcpus,
14858 "memory": self.memory,
14859 "spindle_use": self.spindle_use,
14860 "disks": self.disks,
14861 "disk_space_total": disk_space,
14863 "required_nodes": self.required_nodes,
14864 "hypervisor": self.hypervisor,
14869 def _AddRelocateInstance(self):
14870 """Add relocate instance data to allocator structure.
14872 This in combination with _IAllocatorGetClusterData will create the
14873 correct structure needed as input for the allocator.
14875 The checks for the completeness of the opcode must have already been
14879 instance = self.cfg.GetInstanceInfo(self.name)
14880 if instance is None:
14881 raise errors.ProgrammerError("Unknown instance '%s' passed to"
14882 " IAllocator" % self.name)
14884 if instance.disk_template not in constants.DTS_MIRRORED:
14885 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
14886 errors.ECODE_INVAL)
14888 if instance.disk_template in constants.DTS_INT_MIRROR and \
14889 len(instance.secondary_nodes) != 1:
14890 raise errors.OpPrereqError("Instance has not exactly one secondary node",
14891 errors.ECODE_STATE)
14893 self.required_nodes = 1
14894 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
14895 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
14899 "disk_space_total": disk_space,
14900 "required_nodes": self.required_nodes,
14901 "relocate_from": self.relocate_from,
14905 def _AddNodeEvacuate(self):
14906 """Get data for node-evacuate requests.
14910 "instances": self.instances,
14911 "evac_mode": self.evac_mode,
14914 def _AddChangeGroup(self):
14915 """Get data for node-evacuate requests.
14919 "instances": self.instances,
14920 "target_groups": self.target_groups,
14923 def _BuildInputData(self, fn, keydata):
14924 """Build input data structures.
14927 self._ComputeClusterData()
14930 request["type"] = self.mode
14931 for keyname, keytype in keydata:
14932 if keyname not in request:
14933 raise errors.ProgrammerError("Request parameter %s is missing" %
14935 val = request[keyname]
14936 if not keytype(val):
14937 raise errors.ProgrammerError("Request parameter %s doesn't pass"
14938 " validation, value %s, expected"
14939 " type %s" % (keyname, val, keytype))
14940 self.in_data["request"] = request
14942 self.in_text = serializer.Dump(self.in_data)
14944 _STRING_LIST = ht.TListOf(ht.TString)
14945 _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
14946 # pylint: disable=E1101
14947 # Class '...' has no 'OP_ID' member
14948 "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
14949 opcodes.OpInstanceMigrate.OP_ID,
14950 opcodes.OpInstanceReplaceDisks.OP_ID])
14954 ht.TListOf(ht.TAnd(ht.TIsLength(3),
14955 ht.TItems([ht.TNonEmptyString,
14956 ht.TNonEmptyString,
14957 ht.TListOf(ht.TNonEmptyString),
14960 ht.TListOf(ht.TAnd(ht.TIsLength(2),
14961 ht.TItems([ht.TNonEmptyString,
14964 _NEVAC_RESULT = ht.TAnd(ht.TIsLength(3),
14965 ht.TItems([_NEVAC_MOVED, _NEVAC_FAILED, _JOB_LIST]))
14968 constants.IALLOCATOR_MODE_ALLOC:
14971 ("name", ht.TString),
14972 ("memory", ht.TInt),
14973 ("spindle_use", ht.TInt),
14974 ("disks", ht.TListOf(ht.TDict)),
14975 ("disk_template", ht.TString),
14976 ("os", ht.TString),
14977 ("tags", _STRING_LIST),
14978 ("nics", ht.TListOf(ht.TDict)),
14979 ("vcpus", ht.TInt),
14980 ("hypervisor", ht.TString),
14982 constants.IALLOCATOR_MODE_RELOC:
14983 (_AddRelocateInstance,
14984 [("name", ht.TString), ("relocate_from", _STRING_LIST)],
14986 constants.IALLOCATOR_MODE_NODE_EVAC:
14987 (_AddNodeEvacuate, [
14988 ("instances", _STRING_LIST),
14989 ("evac_mode", ht.TElemOf(constants.IALLOCATOR_NEVAC_MODES)),
14991 constants.IALLOCATOR_MODE_CHG_GROUP:
14992 (_AddChangeGroup, [
14993 ("instances", _STRING_LIST),
14994 ("target_groups", _STRING_LIST),
14998 def Run(self, name, validate=True, call_fn=None):
14999 """Run an instance allocator and return the results.
15002 if call_fn is None:
15003 call_fn = self.rpc.call_iallocator_runner
15005 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
15006 result.Raise("Failure while running the iallocator script")
15008 self.out_text = result.payload
15010 self._ValidateResult()
15012 def _ValidateResult(self):
15013 """Process the allocator results.
15015 This will process and if successful save the result in
15016 self.out_data and the other parameters.
15020 rdict = serializer.Load(self.out_text)
15021 except Exception, err:
15022 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
15024 if not isinstance(rdict, dict):
15025 raise errors.OpExecError("Can't parse iallocator results: not a dict")
15027 # TODO: remove backwards compatiblity in later versions
15028 if "nodes" in rdict and "result" not in rdict:
15029 rdict["result"] = rdict["nodes"]
15032 for key in "success", "info", "result":
15033 if key not in rdict:
15034 raise errors.OpExecError("Can't parse iallocator results:"
15035 " missing key '%s'" % key)
15036 setattr(self, key, rdict[key])
15038 if not self._result_check(self.result):
15039 raise errors.OpExecError("Iallocator returned invalid result,"
15040 " expected %s, got %s" %
15041 (self._result_check, self.result),
15042 errors.ECODE_INVAL)
15044 if self.mode == constants.IALLOCATOR_MODE_RELOC:
15045 assert self.relocate_from is not None
15046 assert self.required_nodes == 1
15048 node2group = dict((name, ndata["group"])
15049 for (name, ndata) in self.in_data["nodes"].items())
15051 fn = compat.partial(self._NodesToGroups, node2group,
15052 self.in_data["nodegroups"])
15054 instance = self.cfg.GetInstanceInfo(self.name)
15055 request_groups = fn(self.relocate_from + [instance.primary_node])
15056 result_groups = fn(rdict["result"] + [instance.primary_node])
15058 if self.success and not set(result_groups).issubset(request_groups):
15059 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
15060 " differ from original groups (%s)" %
15061 (utils.CommaJoin(result_groups),
15062 utils.CommaJoin(request_groups)))
15064 elif self.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
15065 assert self.evac_mode in constants.IALLOCATOR_NEVAC_MODES
15067 self.out_data = rdict
15070 def _NodesToGroups(node2group, groups, nodes):
15071 """Returns a list of unique group names for a list of nodes.
15073 @type node2group: dict
15074 @param node2group: Map from node name to group UUID
15076 @param groups: Group information
15078 @param nodes: Node names
15085 group_uuid = node2group[node]
15087 # Ignore unknown node
15091 group = groups[group_uuid]
15093 # Can't find group, let's use UUID
15094 group_name = group_uuid
15096 group_name = group["name"]
15098 result.add(group_name)
15100 return sorted(result)
15103 class LUTestAllocator(NoHooksLU):
15104 """Run allocator tests.
15106 This LU runs the allocator tests
15109 def CheckPrereq(self):
15110 """Check prerequisites.
15112 This checks the opcode parameters depending on the director and mode test.
15115 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
15116 for attr in ["memory", "disks", "disk_template",
15117 "os", "tags", "nics", "vcpus"]:
15118 if not hasattr(self.op, attr):
15119 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
15120 attr, errors.ECODE_INVAL)
15121 iname = self.cfg.ExpandInstanceName(self.op.name)
15122 if iname is not None:
15123 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
15124 iname, errors.ECODE_EXISTS)
15125 if not isinstance(self.op.nics, list):
15126 raise errors.OpPrereqError("Invalid parameter 'nics'",
15127 errors.ECODE_INVAL)
15128 if not isinstance(self.op.disks, list):
15129 raise errors.OpPrereqError("Invalid parameter 'disks'",
15130 errors.ECODE_INVAL)
15131 for row in self.op.disks:
15132 if (not isinstance(row, dict) or
15133 constants.IDISK_SIZE not in row or
15134 not isinstance(row[constants.IDISK_SIZE], int) or
15135 constants.IDISK_MODE not in row or
15136 row[constants.IDISK_MODE] not in constants.DISK_ACCESS_SET):
15137 raise errors.OpPrereqError("Invalid contents of the 'disks'"
15138 " parameter", errors.ECODE_INVAL)
15139 if self.op.hypervisor is None:
15140 self.op.hypervisor = self.cfg.GetHypervisorType()
15141 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
15142 fname = _ExpandInstanceName(self.cfg, self.op.name)
15143 self.op.name = fname
15144 self.relocate_from = \
15145 list(self.cfg.GetInstanceInfo(fname).secondary_nodes)
15146 elif self.op.mode in (constants.IALLOCATOR_MODE_CHG_GROUP,
15147 constants.IALLOCATOR_MODE_NODE_EVAC):
15148 if not self.op.instances:
15149 raise errors.OpPrereqError("Missing instances", errors.ECODE_INVAL)
15150 self.op.instances = _GetWantedInstances(self, self.op.instances)
15152 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
15153 self.op.mode, errors.ECODE_INVAL)
15155 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
15156 if self.op.allocator is None:
15157 raise errors.OpPrereqError("Missing allocator name",
15158 errors.ECODE_INVAL)
15159 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
15160 raise errors.OpPrereqError("Wrong allocator test '%s'" %
15161 self.op.direction, errors.ECODE_INVAL)
15163 def Exec(self, feedback_fn):
15164 """Run the allocator test.
15167 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
15168 ial = IAllocator(self.cfg, self.rpc,
15171 memory=self.op.memory,
15172 disks=self.op.disks,
15173 disk_template=self.op.disk_template,
15177 vcpus=self.op.vcpus,
15178 hypervisor=self.op.hypervisor,
15180 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
15181 ial = IAllocator(self.cfg, self.rpc,
15184 relocate_from=list(self.relocate_from),
15186 elif self.op.mode == constants.IALLOCATOR_MODE_CHG_GROUP:
15187 ial = IAllocator(self.cfg, self.rpc,
15189 instances=self.op.instances,
15190 target_groups=self.op.target_groups)
15191 elif self.op.mode == constants.IALLOCATOR_MODE_NODE_EVAC:
15192 ial = IAllocator(self.cfg, self.rpc,
15194 instances=self.op.instances,
15195 evac_mode=self.op.evac_mode)
15197 raise errors.ProgrammerError("Uncatched mode %s in"
15198 " LUTestAllocator.Exec", self.op.mode)
15200 if self.op.direction == constants.IALLOCATOR_DIR_IN:
15201 result = ial.in_text
15203 ial.Run(self.op.allocator, validate=False)
15204 result = ial.out_text
15208 #: Query type implementations
15210 constants.QR_CLUSTER: _ClusterQuery,
15211 constants.QR_INSTANCE: _InstanceQuery,
15212 constants.QR_NODE: _NodeQuery,
15213 constants.QR_GROUP: _GroupQuery,
15214 constants.QR_OS: _OsQuery,
15215 constants.QR_EXPORT: _ExportQuery,
15218 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
15221 def _GetQueryImplementation(name):
15222 """Returns the implemtnation for a query type.
15224 @param name: Query type, must be one of L{constants.QR_VIA_OP}
15228 return _QUERY_IMPL[name]
15230 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
15231 errors.ECODE_INVAL)