4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Module implementing the master-side code."""
24 # pylint: disable-msg=W0201,C0302
26 # W0201 since most LU attributes are defined in CheckPrereq or similar
29 # C0302: since we have waaaay to many lines in this module
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 import ganeti.masterd.instance # pylint: disable-msg=W0611
64 def _SupportsOob(cfg, node):
65 """Tells if node supports OOB.
67 @type cfg: L{config.ConfigWriter}
68 @param cfg: The cluster configuration
69 @type node: L{objects.Node}
71 @return: The OOB script if supported or an empty string otherwise
74 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
78 """Data container for LU results with jobs.
80 Instances of this class returned from L{LogicalUnit.Exec} will be recognized
81 by L{mcpu.Processor._ProcessResult}. The latter will then submit the jobs
82 contained in the C{jobs} attribute and include the job IDs in the opcode
86 def __init__(self, jobs, **kwargs):
87 """Initializes this class.
89 Additional return values can be specified as keyword arguments.
91 @type jobs: list of lists of L{opcode.OpCode}
92 @param jobs: A list of lists of opcode objects
99 class LogicalUnit(object):
100 """Logical Unit base class.
102 Subclasses must follow these rules:
103 - implement ExpandNames
104 - implement CheckPrereq (except when tasklets are used)
105 - implement Exec (except when tasklets are used)
106 - implement BuildHooksEnv
107 - implement BuildHooksNodes
108 - redefine HPATH and HTYPE
109 - optionally redefine their run requirements:
110 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
112 Note that all commands require root permissions.
114 @ivar dry_run_result: the value (if any) that will be returned to the caller
115 in dry-run mode (signalled by opcode dry_run parameter)
122 def __init__(self, processor, op, context, rpc):
123 """Constructor for LogicalUnit.
125 This needs to be overridden in derived classes in order to check op
129 self.proc = processor
131 self.cfg = context.cfg
132 self.context = context
134 # Dicts used to declare locking needs to mcpu
135 self.needed_locks = None
136 self.acquired_locks = {}
137 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
139 self.remove_locks = {}
140 # Used to force good behavior when calling helper functions
141 self.recalculate_locks = {}
143 self.Log = processor.Log # pylint: disable-msg=C0103
144 self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
145 self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
146 self.LogStep = processor.LogStep # pylint: disable-msg=C0103
147 # support for dry-run
148 self.dry_run_result = None
149 # support for generic debug attribute
150 if (not hasattr(self.op, "debug_level") or
151 not isinstance(self.op.debug_level, int)):
152 self.op.debug_level = 0
157 # Validate opcode parameters and set defaults
158 self.op.Validate(True)
160 self.CheckArguments()
162 def CheckArguments(self):
163 """Check syntactic validity for the opcode arguments.
165 This method is for doing a simple syntactic check and ensure
166 validity of opcode parameters, without any cluster-related
167 checks. While the same can be accomplished in ExpandNames and/or
168 CheckPrereq, doing these separate is better because:
170 - ExpandNames is left as as purely a lock-related function
171 - CheckPrereq is run after we have acquired locks (and possible
174 The function is allowed to change the self.op attribute so that
175 later methods can no longer worry about missing parameters.
180 def ExpandNames(self):
181 """Expand names for this LU.
183 This method is called before starting to execute the opcode, and it should
184 update all the parameters of the opcode to their canonical form (e.g. a
185 short node name must be fully expanded after this method has successfully
186 completed). This way locking, hooks, logging, etc. can work correctly.
188 LUs which implement this method must also populate the self.needed_locks
189 member, as a dict with lock levels as keys, and a list of needed lock names
192 - use an empty dict if you don't need any lock
193 - if you don't need any lock at a particular level omit that level
194 - don't put anything for the BGL level
195 - if you want all locks at a level use locking.ALL_SET as a value
197 If you need to share locks (rather than acquire them exclusively) at one
198 level you can modify self.share_locks, setting a true value (usually 1) for
199 that level. By default locks are not shared.
201 This function can also define a list of tasklets, which then will be
202 executed in order instead of the usual LU-level CheckPrereq and Exec
203 functions, if those are not defined by the LU.
207 # Acquire all nodes and one instance
208 self.needed_locks = {
209 locking.LEVEL_NODE: locking.ALL_SET,
210 locking.LEVEL_INSTANCE: ['instance1.example.com'],
212 # Acquire just two nodes
213 self.needed_locks = {
214 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
217 self.needed_locks = {} # No, you can't leave it to the default value None
220 # The implementation of this method is mandatory only if the new LU is
221 # concurrent, so that old LUs don't need to be changed all at the same
224 self.needed_locks = {} # Exclusive LUs don't need locks.
226 raise NotImplementedError
228 def DeclareLocks(self, level):
229 """Declare LU locking needs for a level
231 While most LUs can just declare their locking needs at ExpandNames time,
232 sometimes there's the need to calculate some locks after having acquired
233 the ones before. This function is called just before acquiring locks at a
234 particular level, but after acquiring the ones at lower levels, and permits
235 such calculations. It can be used to modify self.needed_locks, and by
236 default it does nothing.
238 This function is only called if you have something already set in
239 self.needed_locks for the level.
241 @param level: Locking level which is going to be locked
242 @type level: member of ganeti.locking.LEVELS
246 def CheckPrereq(self):
247 """Check prerequisites for this LU.
249 This method should check that the prerequisites for the execution
250 of this LU are fulfilled. It can do internode communication, but
251 it should be idempotent - no cluster or system changes are
254 The method should raise errors.OpPrereqError in case something is
255 not fulfilled. Its return value is ignored.
257 This method should also update all the parameters of the opcode to
258 their canonical form if it hasn't been done by ExpandNames before.
261 if self.tasklets is not None:
262 for (idx, tl) in enumerate(self.tasklets):
263 logging.debug("Checking prerequisites for tasklet %s/%s",
264 idx + 1, len(self.tasklets))
269 def Exec(self, feedback_fn):
272 This method should implement the actual work. It should raise
273 errors.OpExecError for failures that are somewhat dealt with in
277 if self.tasklets is not None:
278 for (idx, tl) in enumerate(self.tasklets):
279 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
282 raise NotImplementedError
284 def BuildHooksEnv(self):
285 """Build hooks environment for this LU.
288 @return: Dictionary containing the environment that will be used for
289 running the hooks for this LU. The keys of the dict must not be prefixed
290 with "GANETI_"--that'll be added by the hooks runner. The hooks runner
291 will extend the environment with additional variables. If no environment
292 should be defined, an empty dictionary should be returned (not C{None}).
293 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
297 raise NotImplementedError
299 def BuildHooksNodes(self):
300 """Build list of nodes to run LU's hooks.
302 @rtype: tuple; (list, list)
303 @return: Tuple containing a list of node names on which the hook
304 should run before the execution and a list of node names on which the
305 hook should run after the execution. No nodes should be returned as an
306 empty list (and not None).
307 @note: If the C{HPATH} attribute of the LU class is C{None}, this function
311 raise NotImplementedError
313 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
314 """Notify the LU about the results of its hooks.
316 This method is called every time a hooks phase is executed, and notifies
317 the Logical Unit about the hooks' result. The LU can then use it to alter
318 its result based on the hooks. By default the method does nothing and the
319 previous result is passed back unchanged but any LU can define it if it
320 wants to use the local cluster hook-scripts somehow.
322 @param phase: one of L{constants.HOOKS_PHASE_POST} or
323 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
324 @param hook_results: the results of the multi-node hooks rpc call
325 @param feedback_fn: function used send feedback back to the caller
326 @param lu_result: the previous Exec result this LU had, or None
328 @return: the new Exec result, based on the previous result
332 # API must be kept, thus we ignore the unused argument and could
333 # be a function warnings
334 # pylint: disable-msg=W0613,R0201
337 def _ExpandAndLockInstance(self):
338 """Helper function to expand and lock an instance.
340 Many LUs that work on an instance take its name in self.op.instance_name
341 and need to expand it and then declare the expanded name for locking. This
342 function does it, and then updates self.op.instance_name to the expanded
343 name. It also initializes needed_locks as a dict, if this hasn't been done
347 if self.needed_locks is None:
348 self.needed_locks = {}
350 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
351 "_ExpandAndLockInstance called with instance-level locks set"
352 self.op.instance_name = _ExpandInstanceName(self.cfg,
353 self.op.instance_name)
354 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
356 def _LockInstancesNodes(self, primary_only=False):
357 """Helper function to declare instances' nodes for locking.
359 This function should be called after locking one or more instances to lock
360 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
361 with all primary or secondary nodes for instances already locked and
362 present in self.needed_locks[locking.LEVEL_INSTANCE].
364 It should be called from DeclareLocks, and for safety only works if
365 self.recalculate_locks[locking.LEVEL_NODE] is set.
367 In the future it may grow parameters to just lock some instance's nodes, or
368 to just lock primaries or secondary nodes, if needed.
370 If should be called in DeclareLocks in a way similar to::
372 if level == locking.LEVEL_NODE:
373 self._LockInstancesNodes()
375 @type primary_only: boolean
376 @param primary_only: only lock primary nodes of locked instances
379 assert locking.LEVEL_NODE in self.recalculate_locks, \
380 "_LockInstancesNodes helper function called with no nodes to recalculate"
382 # TODO: check if we're really been called with the instance locks held
384 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
385 # future we might want to have different behaviors depending on the value
386 # of self.recalculate_locks[locking.LEVEL_NODE]
388 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
389 instance = self.context.cfg.GetInstanceInfo(instance_name)
390 wanted_nodes.append(instance.primary_node)
392 wanted_nodes.extend(instance.secondary_nodes)
394 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
395 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
396 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
397 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
399 del self.recalculate_locks[locking.LEVEL_NODE]
402 class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
403 """Simple LU which runs no hooks.
405 This LU is intended as a parent for other LogicalUnits which will
406 run no hooks, in order to reduce duplicate code.
412 def BuildHooksEnv(self):
413 """Empty BuildHooksEnv for NoHooksLu.
415 This just raises an error.
418 raise AssertionError("BuildHooksEnv called for NoHooksLUs")
420 def BuildHooksNodes(self):
421 """Empty BuildHooksNodes for NoHooksLU.
424 raise AssertionError("BuildHooksNodes called for NoHooksLU")
428 """Tasklet base class.
430 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
431 they can mix legacy code with tasklets. Locking needs to be done in the LU,
432 tasklets know nothing about locks.
434 Subclasses must follow these rules:
435 - Implement CheckPrereq
439 def __init__(self, lu):
446 def CheckPrereq(self):
447 """Check prerequisites for this tasklets.
449 This method should check whether the prerequisites for the execution of
450 this tasklet are fulfilled. It can do internode communication, but it
451 should be idempotent - no cluster or system changes are allowed.
453 The method should raise errors.OpPrereqError in case something is not
454 fulfilled. Its return value is ignored.
456 This method should also update all parameters to their canonical form if it
457 hasn't been done before.
462 def Exec(self, feedback_fn):
463 """Execute the tasklet.
465 This method should implement the actual work. It should raise
466 errors.OpExecError for failures that are somewhat dealt with in code, or
470 raise NotImplementedError
474 """Base for query utility classes.
477 #: Attribute holding field definitions
480 def __init__(self, filter_, fields, use_locking):
481 """Initializes this class.
484 self.use_locking = use_locking
486 self.query = query.Query(self.FIELDS, fields, filter_=filter_,
488 self.requested_data = self.query.RequestedData()
489 self.names = self.query.RequestedNames()
491 # Sort only if no names were requested
492 self.sort_by_name = not self.names
494 self.do_locking = None
497 def _GetNames(self, lu, all_names, lock_level):
498 """Helper function to determine names asked for in the query.
502 names = lu.acquired_locks[lock_level]
506 if self.wanted == locking.ALL_SET:
507 assert not self.names
508 # caller didn't specify names, so ordering is not important
509 return utils.NiceSort(names)
511 # caller specified names and we must keep the same order
513 assert not self.do_locking or lu.acquired_locks[lock_level]
515 missing = set(self.wanted).difference(names)
517 raise errors.OpExecError("Some items were removed before retrieving"
518 " their data: %s" % missing)
520 # Return expanded names
523 def ExpandNames(self, lu):
524 """Expand names for this query.
526 See L{LogicalUnit.ExpandNames}.
529 raise NotImplementedError()
531 def DeclareLocks(self, lu, level):
532 """Declare locks for this query.
534 See L{LogicalUnit.DeclareLocks}.
537 raise NotImplementedError()
539 def _GetQueryData(self, lu):
540 """Collects all data for this query.
542 @return: Query data object
545 raise NotImplementedError()
547 def NewStyleQuery(self, lu):
548 """Collect data and execute query.
551 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
552 sort_by_name=self.sort_by_name)
554 def OldStyleQuery(self, lu):
555 """Collect data and execute query.
558 return self.query.OldStyleQuery(self._GetQueryData(lu),
559 sort_by_name=self.sort_by_name)
562 def _GetWantedNodes(lu, nodes):
563 """Returns list of checked and expanded node names.
565 @type lu: L{LogicalUnit}
566 @param lu: the logical unit on whose behalf we execute
568 @param nodes: list of node names or None for all nodes
570 @return: the list of nodes, sorted
571 @raise errors.ProgrammerError: if the nodes parameter is wrong type
575 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
577 return utils.NiceSort(lu.cfg.GetNodeList())
580 def _GetWantedInstances(lu, instances):
581 """Returns list of checked and expanded instance names.
583 @type lu: L{LogicalUnit}
584 @param lu: the logical unit on whose behalf we execute
585 @type instances: list
586 @param instances: list of instance names or None for all instances
588 @return: the list of instances, sorted
589 @raise errors.OpPrereqError: if the instances parameter is wrong type
590 @raise errors.OpPrereqError: if any of the passed instances is not found
594 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
596 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
600 def _GetUpdatedParams(old_params, update_dict,
601 use_default=True, use_none=False):
602 """Return the new version of a parameter dictionary.
604 @type old_params: dict
605 @param old_params: old parameters
606 @type update_dict: dict
607 @param update_dict: dict containing new parameter values, or
608 constants.VALUE_DEFAULT to reset the parameter to its default
610 @param use_default: boolean
611 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
612 values as 'to be deleted' values
613 @param use_none: boolean
614 @type use_none: whether to recognise C{None} values as 'to be
617 @return: the new parameter dictionary
620 params_copy = copy.deepcopy(old_params)
621 for key, val in update_dict.iteritems():
622 if ((use_default and val == constants.VALUE_DEFAULT) or
623 (use_none and val is None)):
629 params_copy[key] = val
633 def _ReleaseLocks(lu, level, names=None, keep=None):
634 """Releases locks owned by an LU.
636 @type lu: L{LogicalUnit}
637 @param level: Lock level
638 @type names: list or None
639 @param names: Names of locks to release
640 @type keep: list or None
641 @param keep: Names of locks to retain
644 assert not (keep is not None and names is not None), \
645 "Only one of the 'names' and the 'keep' parameters can be given"
647 if names is not None:
648 should_release = names.__contains__
650 should_release = lambda name: name not in keep
652 should_release = None
658 # Determine which locks to release
659 for name in lu.acquired_locks[level]:
660 if should_release(name):
665 assert len(lu.acquired_locks[level]) == (len(retain) + len(release))
667 # Release just some locks
668 lu.context.glm.release(level, names=release)
669 lu.acquired_locks[level] = retain
671 assert frozenset(lu.context.glm.list_owned(level)) == frozenset(retain)
674 lu.context.glm.release(level)
675 del lu.acquired_locks[level]
677 assert not lu.context.glm.list_owned(level), "No locks should be owned"
680 def _RunPostHook(lu, node_name):
681 """Runs the post-hook for an opcode on a single node.
684 hm = lu.proc.hmclass(lu.rpc.call_hooks_runner, lu)
686 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
688 # pylint: disable-msg=W0702
689 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
692 def _CheckOutputFields(static, dynamic, selected):
693 """Checks whether all selected fields are valid.
695 @type static: L{utils.FieldSet}
696 @param static: static fields set
697 @type dynamic: L{utils.FieldSet}
698 @param dynamic: dynamic fields set
705 delta = f.NonMatching(selected)
707 raise errors.OpPrereqError("Unknown output fields selected: %s"
708 % ",".join(delta), errors.ECODE_INVAL)
711 def _CheckGlobalHvParams(params):
712 """Validates that given hypervisor params are not global ones.
714 This will ensure that instances don't get customised versions of
718 used_globals = constants.HVC_GLOBALS.intersection(params)
720 msg = ("The following hypervisor parameters are global and cannot"
721 " be customized at instance level, please modify them at"
722 " cluster level: %s" % utils.CommaJoin(used_globals))
723 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
726 def _CheckNodeOnline(lu, node, msg=None):
727 """Ensure that a given node is online.
729 @param lu: the LU on behalf of which we make the check
730 @param node: the node to check
731 @param msg: if passed, should be a message to replace the default one
732 @raise errors.OpPrereqError: if the node is offline
736 msg = "Can't use offline node"
737 if lu.cfg.GetNodeInfo(node).offline:
738 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
741 def _CheckNodeNotDrained(lu, node):
742 """Ensure that a given node is not drained.
744 @param lu: the LU on behalf of which we make the check
745 @param node: the node to check
746 @raise errors.OpPrereqError: if the node is drained
749 if lu.cfg.GetNodeInfo(node).drained:
750 raise errors.OpPrereqError("Can't use drained node %s" % node,
754 def _CheckNodeVmCapable(lu, node):
755 """Ensure that a given node is vm capable.
757 @param lu: the LU on behalf of which we make the check
758 @param node: the node to check
759 @raise errors.OpPrereqError: if the node is not vm capable
762 if not lu.cfg.GetNodeInfo(node).vm_capable:
763 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
767 def _CheckNodeHasOS(lu, node, os_name, force_variant):
768 """Ensure that a node supports a given OS.
770 @param lu: the LU on behalf of which we make the check
771 @param node: the node to check
772 @param os_name: the OS to query about
773 @param force_variant: whether to ignore variant errors
774 @raise errors.OpPrereqError: if the node is not supporting the OS
777 result = lu.rpc.call_os_get(node, os_name)
778 result.Raise("OS '%s' not in supported OS list for node %s" %
780 prereq=True, ecode=errors.ECODE_INVAL)
781 if not force_variant:
782 _CheckOSVariant(result.payload, os_name)
785 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
786 """Ensure that a node has the given secondary ip.
788 @type lu: L{LogicalUnit}
789 @param lu: the LU on behalf of which we make the check
791 @param node: the node to check
792 @type secondary_ip: string
793 @param secondary_ip: the ip to check
794 @type prereq: boolean
795 @param prereq: whether to throw a prerequisite or an execute error
796 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
797 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
800 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
801 result.Raise("Failure checking secondary ip on node %s" % node,
802 prereq=prereq, ecode=errors.ECODE_ENVIRON)
803 if not result.payload:
804 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
805 " please fix and re-run this command" % secondary_ip)
807 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
809 raise errors.OpExecError(msg)
812 def _GetClusterDomainSecret():
813 """Reads the cluster domain secret.
816 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
820 def _CheckInstanceDown(lu, instance, reason):
821 """Ensure that an instance is not running."""
822 if instance.admin_up:
823 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
824 (instance.name, reason), errors.ECODE_STATE)
826 pnode = instance.primary_node
827 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
828 ins_l.Raise("Can't contact node %s for instance information" % pnode,
829 prereq=True, ecode=errors.ECODE_ENVIRON)
831 if instance.name in ins_l.payload:
832 raise errors.OpPrereqError("Instance %s is running, %s" %
833 (instance.name, reason), errors.ECODE_STATE)
836 def _ExpandItemName(fn, name, kind):
837 """Expand an item name.
839 @param fn: the function to use for expansion
840 @param name: requested item name
841 @param kind: text description ('Node' or 'Instance')
842 @return: the resolved (full) name
843 @raise errors.OpPrereqError: if the item is not found
847 if full_name is None:
848 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
853 def _ExpandNodeName(cfg, name):
854 """Wrapper over L{_ExpandItemName} for nodes."""
855 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
858 def _ExpandInstanceName(cfg, name):
859 """Wrapper over L{_ExpandItemName} for instance."""
860 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
863 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
864 memory, vcpus, nics, disk_template, disks,
865 bep, hvp, hypervisor_name):
866 """Builds instance related env variables for hooks
868 This builds the hook environment from individual variables.
871 @param name: the name of the instance
872 @type primary_node: string
873 @param primary_node: the name of the instance's primary node
874 @type secondary_nodes: list
875 @param secondary_nodes: list of secondary nodes as strings
876 @type os_type: string
877 @param os_type: the name of the instance's OS
878 @type status: boolean
879 @param status: the should_run status of the instance
881 @param memory: the memory size of the instance
883 @param vcpus: the count of VCPUs the instance has
885 @param nics: list of tuples (ip, mac, mode, link) representing
886 the NICs the instance has
887 @type disk_template: string
888 @param disk_template: the disk template of the instance
890 @param disks: the list of (size, mode) pairs
892 @param bep: the backend parameters for the instance
894 @param hvp: the hypervisor parameters for the instance
895 @type hypervisor_name: string
896 @param hypervisor_name: the hypervisor for the instance
898 @return: the hook environment for this instance
907 "INSTANCE_NAME": name,
908 "INSTANCE_PRIMARY": primary_node,
909 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
910 "INSTANCE_OS_TYPE": os_type,
911 "INSTANCE_STATUS": str_status,
912 "INSTANCE_MEMORY": memory,
913 "INSTANCE_VCPUS": vcpus,
914 "INSTANCE_DISK_TEMPLATE": disk_template,
915 "INSTANCE_HYPERVISOR": hypervisor_name,
919 nic_count = len(nics)
920 for idx, (ip, mac, mode, link) in enumerate(nics):
923 env["INSTANCE_NIC%d_IP" % idx] = ip
924 env["INSTANCE_NIC%d_MAC" % idx] = mac
925 env["INSTANCE_NIC%d_MODE" % idx] = mode
926 env["INSTANCE_NIC%d_LINK" % idx] = link
927 if mode == constants.NIC_MODE_BRIDGED:
928 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
932 env["INSTANCE_NIC_COUNT"] = nic_count
935 disk_count = len(disks)
936 for idx, (size, mode) in enumerate(disks):
937 env["INSTANCE_DISK%d_SIZE" % idx] = size
938 env["INSTANCE_DISK%d_MODE" % idx] = mode
942 env["INSTANCE_DISK_COUNT"] = disk_count
944 for source, kind in [(bep, "BE"), (hvp, "HV")]:
945 for key, value in source.items():
946 env["INSTANCE_%s_%s" % (kind, key)] = value
951 def _NICListToTuple(lu, nics):
952 """Build a list of nic information tuples.
954 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
955 value in LUInstanceQueryData.
957 @type lu: L{LogicalUnit}
958 @param lu: the logical unit on whose behalf we execute
959 @type nics: list of L{objects.NIC}
960 @param nics: list of nics to convert to hooks tuples
964 cluster = lu.cfg.GetClusterInfo()
968 filled_params = cluster.SimpleFillNIC(nic.nicparams)
969 mode = filled_params[constants.NIC_MODE]
970 link = filled_params[constants.NIC_LINK]
971 hooks_nics.append((ip, mac, mode, link))
975 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
976 """Builds instance related env variables for hooks from an object.
978 @type lu: L{LogicalUnit}
979 @param lu: the logical unit on whose behalf we execute
980 @type instance: L{objects.Instance}
981 @param instance: the instance for which we should build the
984 @param override: dictionary with key/values that will override
987 @return: the hook environment dictionary
990 cluster = lu.cfg.GetClusterInfo()
991 bep = cluster.FillBE(instance)
992 hvp = cluster.FillHV(instance)
994 'name': instance.name,
995 'primary_node': instance.primary_node,
996 'secondary_nodes': instance.secondary_nodes,
997 'os_type': instance.os,
998 'status': instance.admin_up,
999 'memory': bep[constants.BE_MEMORY],
1000 'vcpus': bep[constants.BE_VCPUS],
1001 'nics': _NICListToTuple(lu, instance.nics),
1002 'disk_template': instance.disk_template,
1003 'disks': [(disk.size, disk.mode) for disk in instance.disks],
1006 'hypervisor_name': instance.hypervisor,
1009 args.update(override)
1010 return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
1013 def _AdjustCandidatePool(lu, exceptions):
1014 """Adjust the candidate pool after node operations.
1017 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1019 lu.LogInfo("Promoted nodes to master candidate role: %s",
1020 utils.CommaJoin(node.name for node in mod_list))
1021 for name in mod_list:
1022 lu.context.ReaddNode(name)
1023 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1025 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1029 def _DecideSelfPromotion(lu, exceptions=None):
1030 """Decide whether I should promote myself as a master candidate.
1033 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1034 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1035 # the new node will increase mc_max with one, so:
1036 mc_should = min(mc_should + 1, cp_size)
1037 return mc_now < mc_should
1040 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1041 """Check that the brigdes needed by a list of nics exist.
1044 cluster = lu.cfg.GetClusterInfo()
1045 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1046 brlist = [params[constants.NIC_LINK] for params in paramslist
1047 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1049 result = lu.rpc.call_bridges_exist(target_node, brlist)
1050 result.Raise("Error checking bridges on destination node '%s'" %
1051 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1054 def _CheckInstanceBridgesExist(lu, instance, node=None):
1055 """Check that the brigdes needed by an instance exist.
1059 node = instance.primary_node
1060 _CheckNicsBridgesExist(lu, instance.nics, node)
1063 def _CheckOSVariant(os_obj, name):
1064 """Check whether an OS name conforms to the os variants specification.
1066 @type os_obj: L{objects.OS}
1067 @param os_obj: OS object to check
1069 @param name: OS name passed by the user, to check for validity
1072 if not os_obj.supported_variants:
1074 variant = objects.OS.GetVariant(name)
1076 raise errors.OpPrereqError("OS name must include a variant",
1079 if variant not in os_obj.supported_variants:
1080 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1083 def _GetNodeInstancesInner(cfg, fn):
1084 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1087 def _GetNodeInstances(cfg, node_name):
1088 """Returns a list of all primary and secondary instances on a node.
1092 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1095 def _GetNodePrimaryInstances(cfg, node_name):
1096 """Returns primary instances on a node.
1099 return _GetNodeInstancesInner(cfg,
1100 lambda inst: node_name == inst.primary_node)
1103 def _GetNodeSecondaryInstances(cfg, node_name):
1104 """Returns secondary instances on a node.
1107 return _GetNodeInstancesInner(cfg,
1108 lambda inst: node_name in inst.secondary_nodes)
1111 def _GetStorageTypeArgs(cfg, storage_type):
1112 """Returns the arguments for a storage type.
1115 # Special case for file storage
1116 if storage_type == constants.ST_FILE:
1117 # storage.FileStorage wants a list of storage directories
1118 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1123 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1126 for dev in instance.disks:
1127 cfg.SetDiskID(dev, node_name)
1129 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1130 result.Raise("Failed to get disk status from node %s" % node_name,
1131 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1133 for idx, bdev_status in enumerate(result.payload):
1134 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1140 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1141 """Check the sanity of iallocator and node arguments and use the
1142 cluster-wide iallocator if appropriate.
1144 Check that at most one of (iallocator, node) is specified. If none is
1145 specified, then the LU's opcode's iallocator slot is filled with the
1146 cluster-wide default iallocator.
1148 @type iallocator_slot: string
1149 @param iallocator_slot: the name of the opcode iallocator slot
1150 @type node_slot: string
1151 @param node_slot: the name of the opcode target node slot
1154 node = getattr(lu.op, node_slot, None)
1155 iallocator = getattr(lu.op, iallocator_slot, None)
1157 if node is not None and iallocator is not None:
1158 raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1160 elif node is None and iallocator is None:
1161 default_iallocator = lu.cfg.GetDefaultIAllocator()
1162 if default_iallocator:
1163 setattr(lu.op, iallocator_slot, default_iallocator)
1165 raise errors.OpPrereqError("No iallocator or node given and no"
1166 " cluster-wide default iallocator found."
1167 " Please specify either an iallocator or a"
1168 " node, or set a cluster-wide default"
1172 class LUClusterPostInit(LogicalUnit):
1173 """Logical unit for running hooks after cluster initialization.
1176 HPATH = "cluster-init"
1177 HTYPE = constants.HTYPE_CLUSTER
1179 def BuildHooksEnv(self):
1184 "OP_TARGET": self.cfg.GetClusterName(),
1187 def BuildHooksNodes(self):
1188 """Build hooks nodes.
1191 return ([], [self.cfg.GetMasterNode()])
1193 def Exec(self, feedback_fn):
1200 class LUClusterDestroy(LogicalUnit):
1201 """Logical unit for destroying the cluster.
1204 HPATH = "cluster-destroy"
1205 HTYPE = constants.HTYPE_CLUSTER
1207 def BuildHooksEnv(self):
1212 "OP_TARGET": self.cfg.GetClusterName(),
1215 def BuildHooksNodes(self):
1216 """Build hooks nodes.
1221 def CheckPrereq(self):
1222 """Check prerequisites.
1224 This checks whether the cluster is empty.
1226 Any errors are signaled by raising errors.OpPrereqError.
1229 master = self.cfg.GetMasterNode()
1231 nodelist = self.cfg.GetNodeList()
1232 if len(nodelist) != 1 or nodelist[0] != master:
1233 raise errors.OpPrereqError("There are still %d node(s) in"
1234 " this cluster." % (len(nodelist) - 1),
1236 instancelist = self.cfg.GetInstanceList()
1238 raise errors.OpPrereqError("There are still %d instance(s) in"
1239 " this cluster." % len(instancelist),
1242 def Exec(self, feedback_fn):
1243 """Destroys the cluster.
1246 master = self.cfg.GetMasterNode()
1248 # Run post hooks on master node before it's removed
1249 _RunPostHook(self, master)
1251 result = self.rpc.call_node_stop_master(master, False)
1252 result.Raise("Could not disable the master role")
1257 def _VerifyCertificate(filename):
1258 """Verifies a certificate for LUClusterVerify.
1260 @type filename: string
1261 @param filename: Path to PEM file
1265 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1266 utils.ReadFile(filename))
1267 except Exception, err: # pylint: disable-msg=W0703
1268 return (LUClusterVerify.ETYPE_ERROR,
1269 "Failed to load X509 certificate %s: %s" % (filename, err))
1272 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1273 constants.SSL_CERT_EXPIRATION_ERROR)
1276 fnamemsg = "While verifying %s: %s" % (filename, msg)
1281 return (None, fnamemsg)
1282 elif errcode == utils.CERT_WARNING:
1283 return (LUClusterVerify.ETYPE_WARNING, fnamemsg)
1284 elif errcode == utils.CERT_ERROR:
1285 return (LUClusterVerify.ETYPE_ERROR, fnamemsg)
1287 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1290 class LUClusterVerify(LogicalUnit):
1291 """Verifies the cluster status.
1294 HPATH = "cluster-verify"
1295 HTYPE = constants.HTYPE_CLUSTER
1298 TCLUSTER = "cluster"
1300 TINSTANCE = "instance"
1302 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1303 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1304 ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK")
1305 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1306 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1307 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1308 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1309 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1310 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1311 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1312 ENODEDRBD = (TNODE, "ENODEDRBD")
1313 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1314 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1315 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1316 ENODEHV = (TNODE, "ENODEHV")
1317 ENODELVM = (TNODE, "ENODELVM")
1318 ENODEN1 = (TNODE, "ENODEN1")
1319 ENODENET = (TNODE, "ENODENET")
1320 ENODEOS = (TNODE, "ENODEOS")
1321 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1322 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1323 ENODERPC = (TNODE, "ENODERPC")
1324 ENODESSH = (TNODE, "ENODESSH")
1325 ENODEVERSION = (TNODE, "ENODEVERSION")
1326 ENODESETUP = (TNODE, "ENODESETUP")
1327 ENODETIME = (TNODE, "ENODETIME")
1328 ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1330 ETYPE_FIELD = "code"
1331 ETYPE_ERROR = "ERROR"
1332 ETYPE_WARNING = "WARNING"
1334 _HOOKS_INDENT_RE = re.compile("^", re.M)
1336 class NodeImage(object):
1337 """A class representing the logical and physical status of a node.
1340 @ivar name: the node name to which this object refers
1341 @ivar volumes: a structure as returned from
1342 L{ganeti.backend.GetVolumeList} (runtime)
1343 @ivar instances: a list of running instances (runtime)
1344 @ivar pinst: list of configured primary instances (config)
1345 @ivar sinst: list of configured secondary instances (config)
1346 @ivar sbp: dictionary of {primary-node: list of instances} for all
1347 instances for which this node is secondary (config)
1348 @ivar mfree: free memory, as reported by hypervisor (runtime)
1349 @ivar dfree: free disk, as reported by the node (runtime)
1350 @ivar offline: the offline status (config)
1351 @type rpc_fail: boolean
1352 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1353 not whether the individual keys were correct) (runtime)
1354 @type lvm_fail: boolean
1355 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1356 @type hyp_fail: boolean
1357 @ivar hyp_fail: whether the RPC call didn't return the instance list
1358 @type ghost: boolean
1359 @ivar ghost: whether this is a known node or not (config)
1360 @type os_fail: boolean
1361 @ivar os_fail: whether the RPC call didn't return valid OS data
1363 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1364 @type vm_capable: boolean
1365 @ivar vm_capable: whether the node can host instances
1368 def __init__(self, offline=False, name=None, vm_capable=True):
1377 self.offline = offline
1378 self.vm_capable = vm_capable
1379 self.rpc_fail = False
1380 self.lvm_fail = False
1381 self.hyp_fail = False
1383 self.os_fail = False
1386 def ExpandNames(self):
1387 self.needed_locks = {
1388 locking.LEVEL_NODE: locking.ALL_SET,
1389 locking.LEVEL_INSTANCE: locking.ALL_SET,
1391 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1393 def _Error(self, ecode, item, msg, *args, **kwargs):
1394 """Format an error message.
1396 Based on the opcode's error_codes parameter, either format a
1397 parseable error code, or a simpler error string.
1399 This must be called only from Exec and functions called from Exec.
1402 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1404 # first complete the msg
1407 # then format the whole message
1408 if self.op.error_codes:
1409 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1415 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1416 # and finally report it via the feedback_fn
1417 self._feedback_fn(" - %s" % msg)
1419 def _ErrorIf(self, cond, *args, **kwargs):
1420 """Log an error message if the passed condition is True.
1423 cond = bool(cond) or self.op.debug_simulate_errors
1425 self._Error(*args, **kwargs)
1426 # do not mark the operation as failed for WARN cases only
1427 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1428 self.bad = self.bad or cond
1430 def _VerifyNode(self, ninfo, nresult):
1431 """Perform some basic validation on data returned from a node.
1433 - check the result data structure is well formed and has all the
1435 - check ganeti version
1437 @type ninfo: L{objects.Node}
1438 @param ninfo: the node to check
1439 @param nresult: the results from the node
1441 @return: whether overall this call was successful (and we can expect
1442 reasonable values in the respose)
1446 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1448 # main result, nresult should be a non-empty dict
1449 test = not nresult or not isinstance(nresult, dict)
1450 _ErrorIf(test, self.ENODERPC, node,
1451 "unable to verify node: no data returned")
1455 # compares ganeti version
1456 local_version = constants.PROTOCOL_VERSION
1457 remote_version = nresult.get("version", None)
1458 test = not (remote_version and
1459 isinstance(remote_version, (list, tuple)) and
1460 len(remote_version) == 2)
1461 _ErrorIf(test, self.ENODERPC, node,
1462 "connection to node returned invalid data")
1466 test = local_version != remote_version[0]
1467 _ErrorIf(test, self.ENODEVERSION, node,
1468 "incompatible protocol versions: master %s,"
1469 " node %s", local_version, remote_version[0])
1473 # node seems compatible, we can actually try to look into its results
1475 # full package version
1476 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1477 self.ENODEVERSION, node,
1478 "software version mismatch: master %s, node %s",
1479 constants.RELEASE_VERSION, remote_version[1],
1480 code=self.ETYPE_WARNING)
1482 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1483 if ninfo.vm_capable and isinstance(hyp_result, dict):
1484 for hv_name, hv_result in hyp_result.iteritems():
1485 test = hv_result is not None
1486 _ErrorIf(test, self.ENODEHV, node,
1487 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1489 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1490 if ninfo.vm_capable and isinstance(hvp_result, list):
1491 for item, hv_name, hv_result in hvp_result:
1492 _ErrorIf(True, self.ENODEHV, node,
1493 "hypervisor %s parameter verify failure (source %s): %s",
1494 hv_name, item, hv_result)
1496 test = nresult.get(constants.NV_NODESETUP,
1497 ["Missing NODESETUP results"])
1498 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1503 def _VerifyNodeTime(self, ninfo, nresult,
1504 nvinfo_starttime, nvinfo_endtime):
1505 """Check the node time.
1507 @type ninfo: L{objects.Node}
1508 @param ninfo: the node to check
1509 @param nresult: the remote results for the node
1510 @param nvinfo_starttime: the start time of the RPC call
1511 @param nvinfo_endtime: the end time of the RPC call
1515 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1517 ntime = nresult.get(constants.NV_TIME, None)
1519 ntime_merged = utils.MergeTime(ntime)
1520 except (ValueError, TypeError):
1521 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1524 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1525 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1526 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1527 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1531 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1532 "Node time diverges by at least %s from master node time",
1535 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1536 """Check the node time.
1538 @type ninfo: L{objects.Node}
1539 @param ninfo: the node to check
1540 @param nresult: the remote results for the node
1541 @param vg_name: the configured VG name
1548 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1550 # checks vg existence and size > 20G
1551 vglist = nresult.get(constants.NV_VGLIST, None)
1553 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1555 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1556 constants.MIN_VG_SIZE)
1557 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1560 pvlist = nresult.get(constants.NV_PVLIST, None)
1561 test = pvlist is None
1562 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1564 # check that ':' is not present in PV names, since it's a
1565 # special character for lvcreate (denotes the range of PEs to
1567 for _, pvname, owner_vg in pvlist:
1568 test = ":" in pvname
1569 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1570 " '%s' of VG '%s'", pvname, owner_vg)
1572 def _VerifyNodeNetwork(self, ninfo, nresult):
1573 """Check the node time.
1575 @type ninfo: L{objects.Node}
1576 @param ninfo: the node to check
1577 @param nresult: the remote results for the node
1581 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1583 test = constants.NV_NODELIST not in nresult
1584 _ErrorIf(test, self.ENODESSH, node,
1585 "node hasn't returned node ssh connectivity data")
1587 if nresult[constants.NV_NODELIST]:
1588 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1589 _ErrorIf(True, self.ENODESSH, node,
1590 "ssh communication with node '%s': %s", a_node, a_msg)
1592 test = constants.NV_NODENETTEST not in nresult
1593 _ErrorIf(test, self.ENODENET, node,
1594 "node hasn't returned node tcp connectivity data")
1596 if nresult[constants.NV_NODENETTEST]:
1597 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1599 _ErrorIf(True, self.ENODENET, node,
1600 "tcp communication with node '%s': %s",
1601 anode, nresult[constants.NV_NODENETTEST][anode])
1603 test = constants.NV_MASTERIP not in nresult
1604 _ErrorIf(test, self.ENODENET, node,
1605 "node hasn't returned node master IP reachability data")
1607 if not nresult[constants.NV_MASTERIP]:
1608 if node == self.master_node:
1609 msg = "the master node cannot reach the master IP (not configured?)"
1611 msg = "cannot reach the master IP"
1612 _ErrorIf(True, self.ENODENET, node, msg)
1614 def _VerifyInstance(self, instance, instanceconfig, node_image,
1616 """Verify an instance.
1618 This function checks to see if the required block devices are
1619 available on the instance's node.
1622 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1623 node_current = instanceconfig.primary_node
1625 node_vol_should = {}
1626 instanceconfig.MapLVsByNode(node_vol_should)
1628 for node in node_vol_should:
1629 n_img = node_image[node]
1630 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1631 # ignore missing volumes on offline or broken nodes
1633 for volume in node_vol_should[node]:
1634 test = volume not in n_img.volumes
1635 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1636 "volume %s missing on node %s", volume, node)
1638 if instanceconfig.admin_up:
1639 pri_img = node_image[node_current]
1640 test = instance not in pri_img.instances and not pri_img.offline
1641 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1642 "instance not running on its primary node %s",
1645 for node, n_img in node_image.items():
1646 if node != node_current:
1647 test = instance in n_img.instances
1648 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1649 "instance should not run on node %s", node)
1651 diskdata = [(nname, success, status, idx)
1652 for (nname, disks) in diskstatus.items()
1653 for idx, (success, status) in enumerate(disks)]
1655 for nname, success, bdev_status, idx in diskdata:
1656 # the 'ghost node' construction in Exec() ensures that we have a
1658 snode = node_image[nname]
1659 bad_snode = snode.ghost or snode.offline
1660 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
1661 self.EINSTANCEFAULTYDISK, instance,
1662 "couldn't retrieve status for disk/%s on %s: %s",
1663 idx, nname, bdev_status)
1664 _ErrorIf((instanceconfig.admin_up and success and
1665 bdev_status.ldisk_status == constants.LDS_FAULTY),
1666 self.EINSTANCEFAULTYDISK, instance,
1667 "disk/%s on %s is faulty", idx, nname)
1669 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1670 """Verify if there are any unknown volumes in the cluster.
1672 The .os, .swap and backup volumes are ignored. All other volumes are
1673 reported as unknown.
1675 @type reserved: L{ganeti.utils.FieldSet}
1676 @param reserved: a FieldSet of reserved volume names
1679 for node, n_img in node_image.items():
1680 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1681 # skip non-healthy nodes
1683 for volume in n_img.volumes:
1684 test = ((node not in node_vol_should or
1685 volume not in node_vol_should[node]) and
1686 not reserved.Matches(volume))
1687 self._ErrorIf(test, self.ENODEORPHANLV, node,
1688 "volume %s is unknown", volume)
1690 def _VerifyOrphanInstances(self, instancelist, node_image):
1691 """Verify the list of running instances.
1693 This checks what instances are running but unknown to the cluster.
1696 for node, n_img in node_image.items():
1697 for o_inst in n_img.instances:
1698 test = o_inst not in instancelist
1699 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1700 "instance %s on node %s should not exist", o_inst, node)
1702 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1703 """Verify N+1 Memory Resilience.
1705 Check that if one single node dies we can still start all the
1706 instances it was primary for.
1709 cluster_info = self.cfg.GetClusterInfo()
1710 for node, n_img in node_image.items():
1711 # This code checks that every node which is now listed as
1712 # secondary has enough memory to host all instances it is
1713 # supposed to should a single other node in the cluster fail.
1714 # FIXME: not ready for failover to an arbitrary node
1715 # FIXME: does not support file-backed instances
1716 # WARNING: we currently take into account down instances as well
1717 # as up ones, considering that even if they're down someone
1718 # might want to start them even in the event of a node failure.
1720 # we're skipping offline nodes from the N+1 warning, since
1721 # most likely we don't have good memory infromation from them;
1722 # we already list instances living on such nodes, and that's
1725 for prinode, instances in n_img.sbp.items():
1727 for instance in instances:
1728 bep = cluster_info.FillBE(instance_cfg[instance])
1729 if bep[constants.BE_AUTO_BALANCE]:
1730 needed_mem += bep[constants.BE_MEMORY]
1731 test = n_img.mfree < needed_mem
1732 self._ErrorIf(test, self.ENODEN1, node,
1733 "not enough memory to accomodate instance failovers"
1734 " should node %s fail (%dMiB needed, %dMiB available)",
1735 prinode, needed_mem, n_img.mfree)
1738 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
1739 (files_all, files_all_opt, files_mc, files_vm)):
1740 """Verifies file checksums collected from all nodes.
1742 @param errorif: Callback for reporting errors
1743 @param nodeinfo: List of L{objects.Node} objects
1744 @param master_node: Name of master node
1745 @param all_nvinfo: RPC results
1748 node_names = frozenset(node.name for node in nodeinfo)
1750 assert master_node in node_names
1751 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
1752 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
1753 "Found file listed in more than one file list"
1755 # Define functions determining which nodes to consider for a file
1756 file2nodefn = dict([(filename, fn)
1757 for (files, fn) in [(files_all, None),
1758 (files_all_opt, None),
1759 (files_mc, lambda node: (node.master_candidate or
1760 node.name == master_node)),
1761 (files_vm, lambda node: node.vm_capable)]
1762 for filename in files])
1764 fileinfo = dict((filename, {}) for filename in file2nodefn.keys())
1766 for node in nodeinfo:
1767 nresult = all_nvinfo[node.name]
1769 if nresult.fail_msg or not nresult.payload:
1772 node_files = nresult.payload.get(constants.NV_FILELIST, None)
1774 test = not (node_files and isinstance(node_files, dict))
1775 errorif(test, cls.ENODEFILECHECK, node.name,
1776 "Node did not return file checksum data")
1780 for (filename, checksum) in node_files.items():
1781 # Check if the file should be considered for a node
1782 fn = file2nodefn[filename]
1783 if fn is None or fn(node):
1784 fileinfo[filename].setdefault(checksum, set()).add(node.name)
1786 for (filename, checksums) in fileinfo.items():
1787 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
1789 # Nodes having the file
1790 with_file = frozenset(node_name
1791 for nodes in fileinfo[filename].values()
1792 for node_name in nodes)
1794 # Nodes missing file
1795 missing_file = node_names - with_file
1797 if filename in files_all_opt:
1799 errorif(missing_file and missing_file != node_names,
1800 cls.ECLUSTERFILECHECK, None,
1801 "File %s is optional, but it must exist on all or no nodes (not"
1803 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
1805 errorif(missing_file, cls.ECLUSTERFILECHECK, None,
1806 "File %s is missing from node(s) %s", filename,
1807 utils.CommaJoin(utils.NiceSort(missing_file)))
1809 # See if there are multiple versions of the file
1810 test = len(checksums) > 1
1812 variants = ["variant %s on %s" %
1813 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
1814 for (idx, (checksum, nodes)) in
1815 enumerate(sorted(checksums.items()))]
1819 errorif(test, cls.ECLUSTERFILECHECK, None,
1820 "File %s found with %s different checksums (%s)",
1821 filename, len(checksums), "; ".join(variants))
1823 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1825 """Verifies and the node DRBD status.
1827 @type ninfo: L{objects.Node}
1828 @param ninfo: the node to check
1829 @param nresult: the remote results for the node
1830 @param instanceinfo: the dict of instances
1831 @param drbd_helper: the configured DRBD usermode helper
1832 @param drbd_map: the DRBD map as returned by
1833 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1837 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1840 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1841 test = (helper_result == None)
1842 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1843 "no drbd usermode helper returned")
1845 status, payload = helper_result
1847 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1848 "drbd usermode helper check unsuccessful: %s", payload)
1849 test = status and (payload != drbd_helper)
1850 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1851 "wrong drbd usermode helper: %s", payload)
1853 # compute the DRBD minors
1855 for minor, instance in drbd_map[node].items():
1856 test = instance not in instanceinfo
1857 _ErrorIf(test, self.ECLUSTERCFG, None,
1858 "ghost instance '%s' in temporary DRBD map", instance)
1859 # ghost instance should not be running, but otherwise we
1860 # don't give double warnings (both ghost instance and
1861 # unallocated minor in use)
1863 node_drbd[minor] = (instance, False)
1865 instance = instanceinfo[instance]
1866 node_drbd[minor] = (instance.name, instance.admin_up)
1868 # and now check them
1869 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1870 test = not isinstance(used_minors, (tuple, list))
1871 _ErrorIf(test, self.ENODEDRBD, node,
1872 "cannot parse drbd status file: %s", str(used_minors))
1874 # we cannot check drbd status
1877 for minor, (iname, must_exist) in node_drbd.items():
1878 test = minor not in used_minors and must_exist
1879 _ErrorIf(test, self.ENODEDRBD, node,
1880 "drbd minor %d of instance %s is not active", minor, iname)
1881 for minor in used_minors:
1882 test = minor not in node_drbd
1883 _ErrorIf(test, self.ENODEDRBD, node,
1884 "unallocated drbd minor %d is in use", minor)
1886 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1887 """Builds the node OS structures.
1889 @type ninfo: L{objects.Node}
1890 @param ninfo: the node to check
1891 @param nresult: the remote results for the node
1892 @param nimg: the node image object
1896 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1898 remote_os = nresult.get(constants.NV_OSLIST, None)
1899 test = (not isinstance(remote_os, list) or
1900 not compat.all(isinstance(v, list) and len(v) == 7
1901 for v in remote_os))
1903 _ErrorIf(test, self.ENODEOS, node,
1904 "node hasn't returned valid OS data")
1913 for (name, os_path, status, diagnose,
1914 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1916 if name not in os_dict:
1919 # parameters is a list of lists instead of list of tuples due to
1920 # JSON lacking a real tuple type, fix it:
1921 parameters = [tuple(v) for v in parameters]
1922 os_dict[name].append((os_path, status, diagnose,
1923 set(variants), set(parameters), set(api_ver)))
1925 nimg.oslist = os_dict
1927 def _VerifyNodeOS(self, ninfo, nimg, base):
1928 """Verifies the node OS list.
1930 @type ninfo: L{objects.Node}
1931 @param ninfo: the node to check
1932 @param nimg: the node image object
1933 @param base: the 'template' node we match against (e.g. from the master)
1937 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1939 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1941 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
1942 for os_name, os_data in nimg.oslist.items():
1943 assert os_data, "Empty OS status for OS %s?!" % os_name
1944 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1945 _ErrorIf(not f_status, self.ENODEOS, node,
1946 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1947 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1948 "OS '%s' has multiple entries (first one shadows the rest): %s",
1949 os_name, utils.CommaJoin([v[0] for v in os_data]))
1950 # this will catched in backend too
1951 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1952 and not f_var, self.ENODEOS, node,
1953 "OS %s with API at least %d does not declare any variant",
1954 os_name, constants.OS_API_V15)
1955 # comparisons with the 'base' image
1956 test = os_name not in base.oslist
1957 _ErrorIf(test, self.ENODEOS, node,
1958 "Extra OS %s not present on reference node (%s)",
1962 assert base.oslist[os_name], "Base node has empty OS status?"
1963 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1965 # base OS is invalid, skipping
1967 for kind, a, b in [("API version", f_api, b_api),
1968 ("variants list", f_var, b_var),
1969 ("parameters", beautify_params(f_param),
1970 beautify_params(b_param))]:
1971 _ErrorIf(a != b, self.ENODEOS, node,
1972 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
1973 kind, os_name, base.name,
1974 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
1976 # check any missing OSes
1977 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1978 _ErrorIf(missing, self.ENODEOS, node,
1979 "OSes present on reference node %s but missing on this node: %s",
1980 base.name, utils.CommaJoin(missing))
1982 def _VerifyOob(self, ninfo, nresult):
1983 """Verifies out of band functionality of a node.
1985 @type ninfo: L{objects.Node}
1986 @param ninfo: the node to check
1987 @param nresult: the remote results for the node
1991 # We just have to verify the paths on master and/or master candidates
1992 # as the oob helper is invoked on the master
1993 if ((ninfo.master_candidate or ninfo.master_capable) and
1994 constants.NV_OOB_PATHS in nresult):
1995 for path_result in nresult[constants.NV_OOB_PATHS]:
1996 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1998 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1999 """Verifies and updates the node volume data.
2001 This function will update a L{NodeImage}'s internal structures
2002 with data from the remote call.
2004 @type ninfo: L{objects.Node}
2005 @param ninfo: the node to check
2006 @param nresult: the remote results for the node
2007 @param nimg: the node image object
2008 @param vg_name: the configured VG name
2012 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2014 nimg.lvm_fail = True
2015 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2018 elif isinstance(lvdata, basestring):
2019 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
2020 utils.SafeEncode(lvdata))
2021 elif not isinstance(lvdata, dict):
2022 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
2024 nimg.volumes = lvdata
2025 nimg.lvm_fail = False
2027 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2028 """Verifies and updates the node instance list.
2030 If the listing was successful, then updates this node's instance
2031 list. Otherwise, it marks the RPC call as failed for the instance
2034 @type ninfo: L{objects.Node}
2035 @param ninfo: the node to check
2036 @param nresult: the remote results for the node
2037 @param nimg: the node image object
2040 idata = nresult.get(constants.NV_INSTANCELIST, None)
2041 test = not isinstance(idata, list)
2042 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
2043 " (instancelist): %s", utils.SafeEncode(str(idata)))
2045 nimg.hyp_fail = True
2047 nimg.instances = idata
2049 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2050 """Verifies and computes a node information map
2052 @type ninfo: L{objects.Node}
2053 @param ninfo: the node to check
2054 @param nresult: the remote results for the node
2055 @param nimg: the node image object
2056 @param vg_name: the configured VG name
2060 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2062 # try to read free memory (from the hypervisor)
2063 hv_info = nresult.get(constants.NV_HVINFO, None)
2064 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2065 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
2068 nimg.mfree = int(hv_info["memory_free"])
2069 except (ValueError, TypeError):
2070 _ErrorIf(True, self.ENODERPC, node,
2071 "node returned invalid nodeinfo, check hypervisor")
2073 # FIXME: devise a free space model for file based instances as well
2074 if vg_name is not None:
2075 test = (constants.NV_VGLIST not in nresult or
2076 vg_name not in nresult[constants.NV_VGLIST])
2077 _ErrorIf(test, self.ENODELVM, node,
2078 "node didn't return data for the volume group '%s'"
2079 " - it is either missing or broken", vg_name)
2082 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2083 except (ValueError, TypeError):
2084 _ErrorIf(True, self.ENODERPC, node,
2085 "node returned invalid LVM info, check LVM status")
2087 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2088 """Gets per-disk status information for all instances.
2090 @type nodelist: list of strings
2091 @param nodelist: Node names
2092 @type node_image: dict of (name, L{objects.Node})
2093 @param node_image: Node objects
2094 @type instanceinfo: dict of (name, L{objects.Instance})
2095 @param instanceinfo: Instance objects
2096 @rtype: {instance: {node: [(succes, payload)]}}
2097 @return: a dictionary of per-instance dictionaries with nodes as
2098 keys and disk information as values; the disk information is a
2099 list of tuples (success, payload)
2102 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2105 node_disks_devonly = {}
2106 diskless_instances = set()
2107 diskless = constants.DT_DISKLESS
2109 for nname in nodelist:
2110 node_instances = list(itertools.chain(node_image[nname].pinst,
2111 node_image[nname].sinst))
2112 diskless_instances.update(inst for inst in node_instances
2113 if instanceinfo[inst].disk_template == diskless)
2114 disks = [(inst, disk)
2115 for inst in node_instances
2116 for disk in instanceinfo[inst].disks]
2119 # No need to collect data
2122 node_disks[nname] = disks
2124 # Creating copies as SetDiskID below will modify the objects and that can
2125 # lead to incorrect data returned from nodes
2126 devonly = [dev.Copy() for (_, dev) in disks]
2129 self.cfg.SetDiskID(dev, nname)
2131 node_disks_devonly[nname] = devonly
2133 assert len(node_disks) == len(node_disks_devonly)
2135 # Collect data from all nodes with disks
2136 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2139 assert len(result) == len(node_disks)
2143 for (nname, nres) in result.items():
2144 disks = node_disks[nname]
2147 # No data from this node
2148 data = len(disks) * [(False, "node offline")]
2151 _ErrorIf(msg, self.ENODERPC, nname,
2152 "while getting disk information: %s", msg)
2154 # No data from this node
2155 data = len(disks) * [(False, msg)]
2158 for idx, i in enumerate(nres.payload):
2159 if isinstance(i, (tuple, list)) and len(i) == 2:
2162 logging.warning("Invalid result from node %s, entry %d: %s",
2164 data.append((False, "Invalid result from the remote node"))
2166 for ((inst, _), status) in zip(disks, data):
2167 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2169 # Add empty entries for diskless instances.
2170 for inst in diskless_instances:
2171 assert inst not in instdisk
2174 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2175 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2176 compat.all(isinstance(s, (tuple, list)) and
2177 len(s) == 2 for s in statuses)
2178 for inst, nnames in instdisk.items()
2179 for nname, statuses in nnames.items())
2180 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2184 def _VerifyHVP(self, hvp_data):
2185 """Verifies locally the syntax of the hypervisor parameters.
2188 for item, hv_name, hv_params in hvp_data:
2189 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2192 hv_class = hypervisor.GetHypervisor(hv_name)
2193 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2194 hv_class.CheckParameterSyntax(hv_params)
2195 except errors.GenericError, err:
2196 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2198 def BuildHooksEnv(self):
2201 Cluster-Verify hooks just ran in the post phase and their failure makes
2202 the output be logged in the verify output and the verification to fail.
2208 "CLUSTER_TAGS": " ".join(cfg.GetClusterInfo().GetTags())
2211 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
2212 for node in cfg.GetAllNodesInfo().values())
2216 def BuildHooksNodes(self):
2217 """Build hooks nodes.
2220 return ([], self.cfg.GetNodeList())
2222 def Exec(self, feedback_fn):
2223 """Verify integrity of cluster, performing various test on nodes.
2226 # This method has too many local variables. pylint: disable-msg=R0914
2228 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2229 verbose = self.op.verbose
2230 self._feedback_fn = feedback_fn
2231 feedback_fn("* Verifying global settings")
2232 for msg in self.cfg.VerifyConfig():
2233 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2235 # Check the cluster certificates
2236 for cert_filename in constants.ALL_CERT_FILES:
2237 (errcode, msg) = _VerifyCertificate(cert_filename)
2238 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2240 vg_name = self.cfg.GetVGName()
2241 drbd_helper = self.cfg.GetDRBDHelper()
2242 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2243 cluster = self.cfg.GetClusterInfo()
2244 nodelist = utils.NiceSort(self.cfg.GetNodeList())
2245 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2246 nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2247 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2248 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2249 for iname in instancelist)
2250 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2251 i_non_redundant = [] # Non redundant instances
2252 i_non_a_balanced = [] # Non auto-balanced instances
2253 n_offline = 0 # Count of offline nodes
2254 n_drained = 0 # Count of nodes being drained
2255 node_vol_should = {}
2257 # FIXME: verify OS list
2260 filemap = _ComputeAncillaryFiles(cluster, False)
2262 # do local checksums
2263 master_node = self.master_node = self.cfg.GetMasterNode()
2264 master_ip = self.cfg.GetMasterIP()
2266 # Compute the set of hypervisor parameters
2268 for hv_name in hypervisors:
2269 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2270 for os_name, os_hvp in cluster.os_hvp.items():
2271 for hv_name, hv_params in os_hvp.items():
2274 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2275 hvp_data.append(("os %s" % os_name, hv_name, full_params))
2276 # TODO: collapse identical parameter values in a single one
2277 for instance in instanceinfo.values():
2278 if not instance.hvparams:
2280 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2281 cluster.FillHV(instance)))
2282 # and verify them locally
2283 self._VerifyHVP(hvp_data)
2285 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2286 node_verify_param = {
2287 constants.NV_FILELIST:
2288 utils.UniqueSequence(filename
2289 for files in filemap
2290 for filename in files),
2291 constants.NV_NODELIST: [node.name for node in nodeinfo
2292 if not node.offline],
2293 constants.NV_HYPERVISOR: hypervisors,
2294 constants.NV_HVPARAMS: hvp_data,
2295 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2296 node.secondary_ip) for node in nodeinfo
2297 if not node.offline],
2298 constants.NV_INSTANCELIST: hypervisors,
2299 constants.NV_VERSION: None,
2300 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2301 constants.NV_NODESETUP: None,
2302 constants.NV_TIME: None,
2303 constants.NV_MASTERIP: (master_node, master_ip),
2304 constants.NV_OSLIST: None,
2305 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2308 if vg_name is not None:
2309 node_verify_param[constants.NV_VGLIST] = None
2310 node_verify_param[constants.NV_LVLIST] = vg_name
2311 node_verify_param[constants.NV_PVLIST] = [vg_name]
2312 node_verify_param[constants.NV_DRBDLIST] = None
2315 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2317 # Build our expected cluster state
2318 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2320 vm_capable=node.vm_capable))
2321 for node in nodeinfo)
2325 for node in nodeinfo:
2326 path = _SupportsOob(self.cfg, node)
2327 if path and path not in oob_paths:
2328 oob_paths.append(path)
2331 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2333 for instance in instancelist:
2334 inst_config = instanceinfo[instance]
2336 for nname in inst_config.all_nodes:
2337 if nname not in node_image:
2339 gnode = self.NodeImage(name=nname)
2341 node_image[nname] = gnode
2343 inst_config.MapLVsByNode(node_vol_should)
2345 pnode = inst_config.primary_node
2346 node_image[pnode].pinst.append(instance)
2348 for snode in inst_config.secondary_nodes:
2349 nimg = node_image[snode]
2350 nimg.sinst.append(instance)
2351 if pnode not in nimg.sbp:
2352 nimg.sbp[pnode] = []
2353 nimg.sbp[pnode].append(instance)
2355 # At this point, we have the in-memory data structures complete,
2356 # except for the runtime information, which we'll gather next
2358 # Due to the way our RPC system works, exact response times cannot be
2359 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2360 # time before and after executing the request, we can at least have a time
2362 nvinfo_starttime = time.time()
2363 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2364 self.cfg.GetClusterName())
2365 nvinfo_endtime = time.time()
2367 all_drbd_map = self.cfg.ComputeDRBDMap()
2369 feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2370 instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2372 feedback_fn("* Verifying configuration file consistency")
2373 self._VerifyFiles(_ErrorIf, nodeinfo, master_node, all_nvinfo, filemap)
2375 feedback_fn("* Verifying node status")
2379 for node_i in nodeinfo:
2381 nimg = node_image[node]
2385 feedback_fn("* Skipping offline node %s" % (node,))
2389 if node == master_node:
2391 elif node_i.master_candidate:
2392 ntype = "master candidate"
2393 elif node_i.drained:
2399 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2401 msg = all_nvinfo[node].fail_msg
2402 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2404 nimg.rpc_fail = True
2407 nresult = all_nvinfo[node].payload
2409 nimg.call_ok = self._VerifyNode(node_i, nresult)
2410 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2411 self._VerifyNodeNetwork(node_i, nresult)
2412 self._VerifyOob(node_i, nresult)
2415 self._VerifyNodeLVM(node_i, nresult, vg_name)
2416 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2419 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2420 self._UpdateNodeInstances(node_i, nresult, nimg)
2421 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2422 self._UpdateNodeOS(node_i, nresult, nimg)
2423 if not nimg.os_fail:
2424 if refos_img is None:
2426 self._VerifyNodeOS(node_i, nimg, refos_img)
2428 feedback_fn("* Verifying instance status")
2429 for instance in instancelist:
2431 feedback_fn("* Verifying instance %s" % instance)
2432 inst_config = instanceinfo[instance]
2433 self._VerifyInstance(instance, inst_config, node_image,
2435 inst_nodes_offline = []
2437 pnode = inst_config.primary_node
2438 pnode_img = node_image[pnode]
2439 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2440 self.ENODERPC, pnode, "instance %s, connection to"
2441 " primary node failed", instance)
2443 _ErrorIf(inst_config.admin_up and pnode_img.offline,
2444 self.EINSTANCEBADNODE, instance,
2445 "instance is marked as running and lives on offline node %s",
2446 inst_config.primary_node)
2448 # If the instance is non-redundant we cannot survive losing its primary
2449 # node, so we are not N+1 compliant. On the other hand we have no disk
2450 # templates with more than one secondary so that situation is not well
2452 # FIXME: does not support file-backed instances
2453 if not inst_config.secondary_nodes:
2454 i_non_redundant.append(instance)
2456 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2457 instance, "instance has multiple secondary nodes: %s",
2458 utils.CommaJoin(inst_config.secondary_nodes),
2459 code=self.ETYPE_WARNING)
2461 if inst_config.disk_template in constants.DTS_INT_MIRROR:
2462 pnode = inst_config.primary_node
2463 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2464 instance_groups = {}
2466 for node in instance_nodes:
2467 instance_groups.setdefault(nodeinfo_byname[node].group,
2471 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2472 # Sort so that we always list the primary node first.
2473 for group, nodes in sorted(instance_groups.items(),
2474 key=lambda (_, nodes): pnode in nodes,
2477 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2478 instance, "instance has primary and secondary nodes in"
2479 " different groups: %s", utils.CommaJoin(pretty_list),
2480 code=self.ETYPE_WARNING)
2482 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2483 i_non_a_balanced.append(instance)
2485 for snode in inst_config.secondary_nodes:
2486 s_img = node_image[snode]
2487 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2488 "instance %s, connection to secondary node failed", instance)
2491 inst_nodes_offline.append(snode)
2493 # warn that the instance lives on offline nodes
2494 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2495 "instance has offline secondary node(s) %s",
2496 utils.CommaJoin(inst_nodes_offline))
2497 # ... or ghost/non-vm_capable nodes
2498 for node in inst_config.all_nodes:
2499 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2500 "instance lives on ghost node %s", node)
2501 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2502 instance, "instance lives on non-vm_capable node %s", node)
2504 feedback_fn("* Verifying orphan volumes")
2505 reserved = utils.FieldSet(*cluster.reserved_lvs)
2506 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2508 feedback_fn("* Verifying orphan instances")
2509 self._VerifyOrphanInstances(instancelist, node_image)
2511 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2512 feedback_fn("* Verifying N+1 Memory redundancy")
2513 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2515 feedback_fn("* Other Notes")
2517 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2518 % len(i_non_redundant))
2520 if i_non_a_balanced:
2521 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2522 % len(i_non_a_balanced))
2525 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2528 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2532 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2533 """Analyze the post-hooks' result
2535 This method analyses the hook result, handles it, and sends some
2536 nicely-formatted feedback back to the user.
2538 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2539 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2540 @param hooks_results: the results of the multi-node hooks rpc call
2541 @param feedback_fn: function used send feedback back to the caller
2542 @param lu_result: previous Exec result
2543 @return: the new Exec result, based on the previous result
2547 # We only really run POST phase hooks, and are only interested in
2549 if phase == constants.HOOKS_PHASE_POST:
2550 # Used to change hooks' output to proper indentation
2551 feedback_fn("* Hooks Results")
2552 assert hooks_results, "invalid result from hooks"
2554 for node_name in hooks_results:
2555 res = hooks_results[node_name]
2557 test = msg and not res.offline
2558 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2559 "Communication failure in hooks execution: %s", msg)
2560 if res.offline or msg:
2561 # No need to investigate payload if node is offline or gave an error.
2562 # override manually lu_result here as _ErrorIf only
2563 # overrides self.bad
2566 for script, hkr, output in res.payload:
2567 test = hkr == constants.HKR_FAIL
2568 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2569 "Script %s failed, output:", script)
2571 output = self._HOOKS_INDENT_RE.sub(' ', output)
2572 feedback_fn("%s" % output)
2578 class LUClusterVerifyDisks(NoHooksLU):
2579 """Verifies the cluster disks status.
2584 def ExpandNames(self):
2585 self.needed_locks = {
2586 locking.LEVEL_NODE: locking.ALL_SET,
2587 locking.LEVEL_INSTANCE: locking.ALL_SET,
2589 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2591 def Exec(self, feedback_fn):
2592 """Verify integrity of cluster disks.
2594 @rtype: tuple of three items
2595 @return: a tuple of (dict of node-to-node_error, list of instances
2596 which need activate-disks, dict of instance: (node, volume) for
2600 result = res_nodes, res_instances, res_missing = {}, [], {}
2602 nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2603 instances = self.cfg.GetAllInstancesInfo().values()
2606 for inst in instances:
2608 if not inst.admin_up:
2610 inst.MapLVsByNode(inst_lvs)
2611 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2612 for node, vol_list in inst_lvs.iteritems():
2613 for vol in vol_list:
2614 nv_dict[(node, vol)] = inst
2619 node_lvs = self.rpc.call_lv_list(nodes, [])
2620 for node, node_res in node_lvs.items():
2621 if node_res.offline:
2623 msg = node_res.fail_msg
2625 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2626 res_nodes[node] = msg
2629 lvs = node_res.payload
2630 for lv_name, (_, _, lv_online) in lvs.items():
2631 inst = nv_dict.pop((node, lv_name), None)
2632 if (not lv_online and inst is not None
2633 and inst.name not in res_instances):
2634 res_instances.append(inst.name)
2636 # any leftover items in nv_dict are missing LVs, let's arrange the
2638 for key, inst in nv_dict.iteritems():
2639 if inst.name not in res_missing:
2640 res_missing[inst.name] = []
2641 res_missing[inst.name].append(key)
2646 class LUClusterRepairDiskSizes(NoHooksLU):
2647 """Verifies the cluster disks sizes.
2652 def ExpandNames(self):
2653 if self.op.instances:
2654 self.wanted_names = []
2655 for name in self.op.instances:
2656 full_name = _ExpandInstanceName(self.cfg, name)
2657 self.wanted_names.append(full_name)
2658 self.needed_locks = {
2659 locking.LEVEL_NODE: [],
2660 locking.LEVEL_INSTANCE: self.wanted_names,
2662 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2664 self.wanted_names = None
2665 self.needed_locks = {
2666 locking.LEVEL_NODE: locking.ALL_SET,
2667 locking.LEVEL_INSTANCE: locking.ALL_SET,
2669 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2671 def DeclareLocks(self, level):
2672 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2673 self._LockInstancesNodes(primary_only=True)
2675 def CheckPrereq(self):
2676 """Check prerequisites.
2678 This only checks the optional instance list against the existing names.
2681 if self.wanted_names is None:
2682 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2684 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2685 in self.wanted_names]
2687 def _EnsureChildSizes(self, disk):
2688 """Ensure children of the disk have the needed disk size.
2690 This is valid mainly for DRBD8 and fixes an issue where the
2691 children have smaller disk size.
2693 @param disk: an L{ganeti.objects.Disk} object
2696 if disk.dev_type == constants.LD_DRBD8:
2697 assert disk.children, "Empty children for DRBD8?"
2698 fchild = disk.children[0]
2699 mismatch = fchild.size < disk.size
2701 self.LogInfo("Child disk has size %d, parent %d, fixing",
2702 fchild.size, disk.size)
2703 fchild.size = disk.size
2705 # and we recurse on this child only, not on the metadev
2706 return self._EnsureChildSizes(fchild) or mismatch
2710 def Exec(self, feedback_fn):
2711 """Verify the size of cluster disks.
2714 # TODO: check child disks too
2715 # TODO: check differences in size between primary/secondary nodes
2717 for instance in self.wanted_instances:
2718 pnode = instance.primary_node
2719 if pnode not in per_node_disks:
2720 per_node_disks[pnode] = []
2721 for idx, disk in enumerate(instance.disks):
2722 per_node_disks[pnode].append((instance, idx, disk))
2725 for node, dskl in per_node_disks.items():
2726 newl = [v[2].Copy() for v in dskl]
2728 self.cfg.SetDiskID(dsk, node)
2729 result = self.rpc.call_blockdev_getsize(node, newl)
2731 self.LogWarning("Failure in blockdev_getsize call to node"
2732 " %s, ignoring", node)
2734 if len(result.payload) != len(dskl):
2735 logging.warning("Invalid result from node %s: len(dksl)=%d,"
2736 " result.payload=%s", node, len(dskl), result.payload)
2737 self.LogWarning("Invalid result from node %s, ignoring node results",
2740 for ((instance, idx, disk), size) in zip(dskl, result.payload):
2742 self.LogWarning("Disk %d of instance %s did not return size"
2743 " information, ignoring", idx, instance.name)
2745 if not isinstance(size, (int, long)):
2746 self.LogWarning("Disk %d of instance %s did not return valid"
2747 " size information, ignoring", idx, instance.name)
2750 if size != disk.size:
2751 self.LogInfo("Disk %d of instance %s has mismatched size,"
2752 " correcting: recorded %d, actual %d", idx,
2753 instance.name, disk.size, size)
2755 self.cfg.Update(instance, feedback_fn)
2756 changed.append((instance.name, idx, size))
2757 if self._EnsureChildSizes(disk):
2758 self.cfg.Update(instance, feedback_fn)
2759 changed.append((instance.name, idx, disk.size))
2763 class LUClusterRename(LogicalUnit):
2764 """Rename the cluster.
2767 HPATH = "cluster-rename"
2768 HTYPE = constants.HTYPE_CLUSTER
2770 def BuildHooksEnv(self):
2775 "OP_TARGET": self.cfg.GetClusterName(),
2776 "NEW_NAME": self.op.name,
2779 def BuildHooksNodes(self):
2780 """Build hooks nodes.
2783 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
2785 def CheckPrereq(self):
2786 """Verify that the passed name is a valid one.
2789 hostname = netutils.GetHostname(name=self.op.name,
2790 family=self.cfg.GetPrimaryIPFamily())
2792 new_name = hostname.name
2793 self.ip = new_ip = hostname.ip
2794 old_name = self.cfg.GetClusterName()
2795 old_ip = self.cfg.GetMasterIP()
2796 if new_name == old_name and new_ip == old_ip:
2797 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2798 " cluster has changed",
2800 if new_ip != old_ip:
2801 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2802 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2803 " reachable on the network" %
2804 new_ip, errors.ECODE_NOTUNIQUE)
2806 self.op.name = new_name
2808 def Exec(self, feedback_fn):
2809 """Rename the cluster.
2812 clustername = self.op.name
2815 # shutdown the master IP
2816 master = self.cfg.GetMasterNode()
2817 result = self.rpc.call_node_stop_master(master, False)
2818 result.Raise("Could not disable the master role")
2821 cluster = self.cfg.GetClusterInfo()
2822 cluster.cluster_name = clustername
2823 cluster.master_ip = ip
2824 self.cfg.Update(cluster, feedback_fn)
2826 # update the known hosts file
2827 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2828 node_list = self.cfg.GetOnlineNodeList()
2830 node_list.remove(master)
2833 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2835 result = self.rpc.call_node_start_master(master, False, False)
2836 msg = result.fail_msg
2838 self.LogWarning("Could not re-enable the master role on"
2839 " the master, please restart manually: %s", msg)
2844 class LUClusterSetParams(LogicalUnit):
2845 """Change the parameters of the cluster.
2848 HPATH = "cluster-modify"
2849 HTYPE = constants.HTYPE_CLUSTER
2852 def CheckArguments(self):
2856 if self.op.uid_pool:
2857 uidpool.CheckUidPool(self.op.uid_pool)
2859 if self.op.add_uids:
2860 uidpool.CheckUidPool(self.op.add_uids)
2862 if self.op.remove_uids:
2863 uidpool.CheckUidPool(self.op.remove_uids)
2865 def ExpandNames(self):
2866 # FIXME: in the future maybe other cluster params won't require checking on
2867 # all nodes to be modified.
2868 self.needed_locks = {
2869 locking.LEVEL_NODE: locking.ALL_SET,
2871 self.share_locks[locking.LEVEL_NODE] = 1
2873 def BuildHooksEnv(self):
2878 "OP_TARGET": self.cfg.GetClusterName(),
2879 "NEW_VG_NAME": self.op.vg_name,
2882 def BuildHooksNodes(self):
2883 """Build hooks nodes.
2886 mn = self.cfg.GetMasterNode()
2889 def CheckPrereq(self):
2890 """Check prerequisites.
2892 This checks whether the given params don't conflict and
2893 if the given volume group is valid.
2896 if self.op.vg_name is not None and not self.op.vg_name:
2897 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2898 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2899 " instances exist", errors.ECODE_INVAL)
2901 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2902 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2903 raise errors.OpPrereqError("Cannot disable drbd helper while"
2904 " drbd-based instances exist",
2907 node_list = self.acquired_locks[locking.LEVEL_NODE]
2909 # if vg_name not None, checks given volume group on all nodes
2911 vglist = self.rpc.call_vg_list(node_list)
2912 for node in node_list:
2913 msg = vglist[node].fail_msg
2915 # ignoring down node
2916 self.LogWarning("Error while gathering data on node %s"
2917 " (ignoring node): %s", node, msg)
2919 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2921 constants.MIN_VG_SIZE)
2923 raise errors.OpPrereqError("Error on node '%s': %s" %
2924 (node, vgstatus), errors.ECODE_ENVIRON)
2926 if self.op.drbd_helper:
2927 # checks given drbd helper on all nodes
2928 helpers = self.rpc.call_drbd_helper(node_list)
2929 for node in node_list:
2930 ninfo = self.cfg.GetNodeInfo(node)
2932 self.LogInfo("Not checking drbd helper on offline node %s", node)
2934 msg = helpers[node].fail_msg
2936 raise errors.OpPrereqError("Error checking drbd helper on node"
2937 " '%s': %s" % (node, msg),
2938 errors.ECODE_ENVIRON)
2939 node_helper = helpers[node].payload
2940 if node_helper != self.op.drbd_helper:
2941 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2942 (node, node_helper), errors.ECODE_ENVIRON)
2944 self.cluster = cluster = self.cfg.GetClusterInfo()
2945 # validate params changes
2946 if self.op.beparams:
2947 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2948 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2950 if self.op.ndparams:
2951 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2952 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2954 # TODO: we need a more general way to handle resetting
2955 # cluster-level parameters to default values
2956 if self.new_ndparams["oob_program"] == "":
2957 self.new_ndparams["oob_program"] = \
2958 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
2960 if self.op.nicparams:
2961 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2962 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2963 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2966 # check all instances for consistency
2967 for instance in self.cfg.GetAllInstancesInfo().values():
2968 for nic_idx, nic in enumerate(instance.nics):
2969 params_copy = copy.deepcopy(nic.nicparams)
2970 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2972 # check parameter syntax
2974 objects.NIC.CheckParameterSyntax(params_filled)
2975 except errors.ConfigurationError, err:
2976 nic_errors.append("Instance %s, nic/%d: %s" %
2977 (instance.name, nic_idx, err))
2979 # if we're moving instances to routed, check that they have an ip
2980 target_mode = params_filled[constants.NIC_MODE]
2981 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2982 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2983 (instance.name, nic_idx))
2985 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2986 "\n".join(nic_errors))
2988 # hypervisor list/parameters
2989 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2990 if self.op.hvparams:
2991 for hv_name, hv_dict in self.op.hvparams.items():
2992 if hv_name not in self.new_hvparams:
2993 self.new_hvparams[hv_name] = hv_dict
2995 self.new_hvparams[hv_name].update(hv_dict)
2997 # os hypervisor parameters
2998 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
3000 for os_name, hvs in self.op.os_hvp.items():
3001 if os_name not in self.new_os_hvp:
3002 self.new_os_hvp[os_name] = hvs
3004 for hv_name, hv_dict in hvs.items():
3005 if hv_name not in self.new_os_hvp[os_name]:
3006 self.new_os_hvp[os_name][hv_name] = hv_dict
3008 self.new_os_hvp[os_name][hv_name].update(hv_dict)
3011 self.new_osp = objects.FillDict(cluster.osparams, {})
3012 if self.op.osparams:
3013 for os_name, osp in self.op.osparams.items():
3014 if os_name not in self.new_osp:
3015 self.new_osp[os_name] = {}
3017 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
3020 if not self.new_osp[os_name]:
3021 # we removed all parameters
3022 del self.new_osp[os_name]
3024 # check the parameter validity (remote check)
3025 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
3026 os_name, self.new_osp[os_name])
3028 # changes to the hypervisor list
3029 if self.op.enabled_hypervisors is not None:
3030 self.hv_list = self.op.enabled_hypervisors
3031 for hv in self.hv_list:
3032 # if the hypervisor doesn't already exist in the cluster
3033 # hvparams, we initialize it to empty, and then (in both
3034 # cases) we make sure to fill the defaults, as we might not
3035 # have a complete defaults list if the hypervisor wasn't
3037 if hv not in new_hvp:
3039 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
3040 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
3042 self.hv_list = cluster.enabled_hypervisors
3044 if self.op.hvparams or self.op.enabled_hypervisors is not None:
3045 # either the enabled list has changed, or the parameters have, validate
3046 for hv_name, hv_params in self.new_hvparams.items():
3047 if ((self.op.hvparams and hv_name in self.op.hvparams) or
3048 (self.op.enabled_hypervisors and
3049 hv_name in self.op.enabled_hypervisors)):
3050 # either this is a new hypervisor, or its parameters have changed
3051 hv_class = hypervisor.GetHypervisor(hv_name)
3052 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3053 hv_class.CheckParameterSyntax(hv_params)
3054 _CheckHVParams(self, node_list, hv_name, hv_params)
3057 # no need to check any newly-enabled hypervisors, since the
3058 # defaults have already been checked in the above code-block
3059 for os_name, os_hvp in self.new_os_hvp.items():
3060 for hv_name, hv_params in os_hvp.items():
3061 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3062 # we need to fill in the new os_hvp on top of the actual hv_p
3063 cluster_defaults = self.new_hvparams.get(hv_name, {})
3064 new_osp = objects.FillDict(cluster_defaults, hv_params)
3065 hv_class = hypervisor.GetHypervisor(hv_name)
3066 hv_class.CheckParameterSyntax(new_osp)
3067 _CheckHVParams(self, node_list, hv_name, new_osp)
3069 if self.op.default_iallocator:
3070 alloc_script = utils.FindFile(self.op.default_iallocator,
3071 constants.IALLOCATOR_SEARCH_PATH,
3073 if alloc_script is None:
3074 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
3075 " specified" % self.op.default_iallocator,
3078 def Exec(self, feedback_fn):
3079 """Change the parameters of the cluster.
3082 if self.op.vg_name is not None:
3083 new_volume = self.op.vg_name
3086 if new_volume != self.cfg.GetVGName():
3087 self.cfg.SetVGName(new_volume)
3089 feedback_fn("Cluster LVM configuration already in desired"
3090 " state, not changing")
3091 if self.op.drbd_helper is not None:
3092 new_helper = self.op.drbd_helper
3095 if new_helper != self.cfg.GetDRBDHelper():
3096 self.cfg.SetDRBDHelper(new_helper)
3098 feedback_fn("Cluster DRBD helper already in desired state,"
3100 if self.op.hvparams:
3101 self.cluster.hvparams = self.new_hvparams
3103 self.cluster.os_hvp = self.new_os_hvp
3104 if self.op.enabled_hypervisors is not None:
3105 self.cluster.hvparams = self.new_hvparams
3106 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
3107 if self.op.beparams:
3108 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
3109 if self.op.nicparams:
3110 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
3111 if self.op.osparams:
3112 self.cluster.osparams = self.new_osp
3113 if self.op.ndparams:
3114 self.cluster.ndparams = self.new_ndparams
3116 if self.op.candidate_pool_size is not None:
3117 self.cluster.candidate_pool_size = self.op.candidate_pool_size
3118 # we need to update the pool size here, otherwise the save will fail
3119 _AdjustCandidatePool(self, [])
3121 if self.op.maintain_node_health is not None:
3122 self.cluster.maintain_node_health = self.op.maintain_node_health
3124 if self.op.prealloc_wipe_disks is not None:
3125 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
3127 if self.op.add_uids is not None:
3128 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
3130 if self.op.remove_uids is not None:
3131 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
3133 if self.op.uid_pool is not None:
3134 self.cluster.uid_pool = self.op.uid_pool
3136 if self.op.default_iallocator is not None:
3137 self.cluster.default_iallocator = self.op.default_iallocator
3139 if self.op.reserved_lvs is not None:
3140 self.cluster.reserved_lvs = self.op.reserved_lvs
3142 def helper_os(aname, mods, desc):
3144 lst = getattr(self.cluster, aname)
3145 for key, val in mods:
3146 if key == constants.DDM_ADD:
3148 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3151 elif key == constants.DDM_REMOVE:
3155 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3157 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3159 if self.op.hidden_os:
3160 helper_os("hidden_os", self.op.hidden_os, "hidden")
3162 if self.op.blacklisted_os:
3163 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3165 if self.op.master_netdev:
3166 master = self.cfg.GetMasterNode()
3167 feedback_fn("Shutting down master ip on the current netdev (%s)" %
3168 self.cluster.master_netdev)
3169 result = self.rpc.call_node_stop_master(master, False)
3170 result.Raise("Could not disable the master ip")
3171 feedback_fn("Changing master_netdev from %s to %s" %
3172 (self.cluster.master_netdev, self.op.master_netdev))
3173 self.cluster.master_netdev = self.op.master_netdev
3175 self.cfg.Update(self.cluster, feedback_fn)
3177 if self.op.master_netdev:
3178 feedback_fn("Starting the master ip on the new master netdev (%s)" %
3179 self.op.master_netdev)
3180 result = self.rpc.call_node_start_master(master, False, False)
3182 self.LogWarning("Could not re-enable the master ip on"
3183 " the master, please restart manually: %s",
3187 def _UploadHelper(lu, nodes, fname):
3188 """Helper for uploading a file and showing warnings.
3191 if os.path.exists(fname):
3192 result = lu.rpc.call_upload_file(nodes, fname)
3193 for to_node, to_result in result.items():
3194 msg = to_result.fail_msg
3196 msg = ("Copy of file %s to node %s failed: %s" %
3197 (fname, to_node, msg))
3198 lu.proc.LogWarning(msg)
3201 def _ComputeAncillaryFiles(cluster, redist):
3202 """Compute files external to Ganeti which need to be consistent.
3204 @type redist: boolean
3205 @param redist: Whether to include files which need to be redistributed
3208 # Compute files for all nodes
3210 constants.SSH_KNOWN_HOSTS_FILE,
3211 constants.CONFD_HMAC_KEY,
3212 constants.CLUSTER_DOMAIN_SECRET_FILE,
3216 files_all.update(constants.ALL_CERT_FILES)
3217 files_all.update(ssconf.SimpleStore().GetFileList())
3219 if cluster.modify_etc_hosts:
3220 files_all.add(constants.ETC_HOSTS)
3222 # Files which must either exist on all nodes or on none
3223 files_all_opt = set([
3224 constants.RAPI_USERS_FILE,
3227 # Files which should only be on master candidates
3230 files_mc.add(constants.CLUSTER_CONF_FILE)
3232 # Files which should only be on VM-capable nodes
3233 files_vm = set(filename
3234 for hv_name in cluster.enabled_hypervisors
3235 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles())
3237 # Filenames must be unique
3238 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
3239 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
3240 "Found file listed in more than one file list"
3242 return (files_all, files_all_opt, files_mc, files_vm)
3245 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3246 """Distribute additional files which are part of the cluster configuration.
3248 ConfigWriter takes care of distributing the config and ssconf files, but
3249 there are more files which should be distributed to all nodes. This function
3250 makes sure those are copied.
3252 @param lu: calling logical unit
3253 @param additional_nodes: list of nodes not in the config to distribute to
3254 @type additional_vm: boolean
3255 @param additional_vm: whether the additional nodes are vm-capable or not
3258 # Gather target nodes
3259 cluster = lu.cfg.GetClusterInfo()
3260 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3262 online_nodes = lu.cfg.GetOnlineNodeList()
3263 vm_nodes = lu.cfg.GetVmCapableNodeList()
3265 if additional_nodes is not None:
3266 online_nodes.extend(additional_nodes)
3268 vm_nodes.extend(additional_nodes)
3270 # Never distribute to master node
3271 for nodelist in [online_nodes, vm_nodes]:
3272 if master_info.name in nodelist:
3273 nodelist.remove(master_info.name)
3276 (files_all, files_all_opt, files_mc, files_vm) = \
3277 _ComputeAncillaryFiles(cluster, True)
3279 # Never re-distribute configuration file from here
3280 assert not (constants.CLUSTER_CONF_FILE in files_all or
3281 constants.CLUSTER_CONF_FILE in files_vm)
3282 assert not files_mc, "Master candidates not handled in this function"
3285 (online_nodes, files_all),
3286 (online_nodes, files_all_opt),
3287 (vm_nodes, files_vm),
3291 for (node_list, files) in filemap:
3293 _UploadHelper(lu, node_list, fname)
3296 class LUClusterRedistConf(NoHooksLU):
3297 """Force the redistribution of cluster configuration.
3299 This is a very simple LU.
3304 def ExpandNames(self):
3305 self.needed_locks = {
3306 locking.LEVEL_NODE: locking.ALL_SET,
3308 self.share_locks[locking.LEVEL_NODE] = 1
3310 def Exec(self, feedback_fn):
3311 """Redistribute the configuration.
3314 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3315 _RedistributeAncillaryFiles(self)
3318 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3319 """Sleep and poll for an instance's disk to sync.
3322 if not instance.disks or disks is not None and not disks:
3325 disks = _ExpandCheckDisks(instance, disks)
3328 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3330 node = instance.primary_node
3333 lu.cfg.SetDiskID(dev, node)
3335 # TODO: Convert to utils.Retry
3338 degr_retries = 10 # in seconds, as we sleep 1 second each time
3342 cumul_degraded = False
3343 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3344 msg = rstats.fail_msg
3346 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3349 raise errors.RemoteError("Can't contact node %s for mirror data,"
3350 " aborting." % node)
3353 rstats = rstats.payload
3355 for i, mstat in enumerate(rstats):
3357 lu.LogWarning("Can't compute data for node %s/%s",
3358 node, disks[i].iv_name)
3361 cumul_degraded = (cumul_degraded or
3362 (mstat.is_degraded and mstat.sync_percent is None))
3363 if mstat.sync_percent is not None:
3365 if mstat.estimated_time is not None:
3366 rem_time = ("%s remaining (estimated)" %
3367 utils.FormatSeconds(mstat.estimated_time))
3368 max_time = mstat.estimated_time
3370 rem_time = "no time estimate"
3371 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3372 (disks[i].iv_name, mstat.sync_percent, rem_time))
3374 # if we're done but degraded, let's do a few small retries, to
3375 # make sure we see a stable and not transient situation; therefore
3376 # we force restart of the loop
3377 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3378 logging.info("Degraded disks found, %d retries left", degr_retries)
3386 time.sleep(min(60, max_time))
3389 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3390 return not cumul_degraded
3393 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3394 """Check that mirrors are not degraded.
3396 The ldisk parameter, if True, will change the test from the
3397 is_degraded attribute (which represents overall non-ok status for
3398 the device(s)) to the ldisk (representing the local storage status).
3401 lu.cfg.SetDiskID(dev, node)
3405 if on_primary or dev.AssembleOnSecondary():
3406 rstats = lu.rpc.call_blockdev_find(node, dev)
3407 msg = rstats.fail_msg
3409 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3411 elif not rstats.payload:
3412 lu.LogWarning("Can't find disk on node %s", node)
3416 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3418 result = result and not rstats.payload.is_degraded
3421 for child in dev.children:
3422 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3427 class LUOobCommand(NoHooksLU):
3428 """Logical unit for OOB handling.
3432 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
3434 def CheckPrereq(self):
3435 """Check prerequisites.
3438 - the node exists in the configuration
3441 Any errors are signaled by raising errors.OpPrereqError.
3445 self.master_node = self.cfg.GetMasterNode()
3447 assert self.op.power_delay >= 0.0
3449 if self.op.node_names:
3450 if (self.op.command in self._SKIP_MASTER and
3451 self.master_node in self.op.node_names):
3452 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
3453 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
3455 if master_oob_handler:
3456 additional_text = ("run '%s %s %s' if you want to operate on the"
3457 " master regardless") % (master_oob_handler,
3461 additional_text = "it does not support out-of-band operations"
3463 raise errors.OpPrereqError(("Operating on the master node %s is not"
3464 " allowed for %s; %s") %
3465 (self.master_node, self.op.command,
3466 additional_text), errors.ECODE_INVAL)
3468 self.op.node_names = self.cfg.GetNodeList()
3469 if self.op.command in self._SKIP_MASTER:
3470 self.op.node_names.remove(self.master_node)
3472 if self.op.command in self._SKIP_MASTER:
3473 assert self.master_node not in self.op.node_names
3475 for node_name in self.op.node_names:
3476 node = self.cfg.GetNodeInfo(node_name)
3479 raise errors.OpPrereqError("Node %s not found" % node_name,
3482 self.nodes.append(node)
3484 if (not self.op.ignore_status and
3485 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
3486 raise errors.OpPrereqError(("Cannot power off node %s because it is"
3487 " not marked offline") % node_name,
3490 def ExpandNames(self):
3491 """Gather locks we need.
3494 if self.op.node_names:
3495 self.op.node_names = [_ExpandNodeName(self.cfg, name)
3496 for name in self.op.node_names]
3497 lock_names = self.op.node_names
3499 lock_names = locking.ALL_SET
3501 self.needed_locks = {
3502 locking.LEVEL_NODE: lock_names,
3505 def Exec(self, feedback_fn):
3506 """Execute OOB and return result if we expect any.
3509 master_node = self.master_node
3512 for idx, node in enumerate(self.nodes):
3513 node_entry = [(constants.RS_NORMAL, node.name)]
3514 ret.append(node_entry)
3516 oob_program = _SupportsOob(self.cfg, node)
3519 node_entry.append((constants.RS_UNAVAIL, None))
3522 logging.info("Executing out-of-band command '%s' using '%s' on %s",
3523 self.op.command, oob_program, node.name)
3524 result = self.rpc.call_run_oob(master_node, oob_program,
3525 self.op.command, node.name,
3529 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
3530 node.name, result.fail_msg)
3531 node_entry.append((constants.RS_NODATA, None))
3534 self._CheckPayload(result)
3535 except errors.OpExecError, err:
3536 self.LogWarning("Payload returned by node '%s' is not valid: %s",
3538 node_entry.append((constants.RS_NODATA, None))
3540 if self.op.command == constants.OOB_HEALTH:
3541 # For health we should log important events
3542 for item, status in result.payload:
3543 if status in [constants.OOB_STATUS_WARNING,
3544 constants.OOB_STATUS_CRITICAL]:
3545 self.LogWarning("Item '%s' on node '%s' has status '%s'",
3546 item, node.name, status)
3548 if self.op.command == constants.OOB_POWER_ON:
3550 elif self.op.command == constants.OOB_POWER_OFF:
3551 node.powered = False
3552 elif self.op.command == constants.OOB_POWER_STATUS:
3553 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3554 if powered != node.powered:
3555 logging.warning(("Recorded power state (%s) of node '%s' does not"
3556 " match actual power state (%s)"), node.powered,
3559 # For configuration changing commands we should update the node
3560 if self.op.command in (constants.OOB_POWER_ON,
3561 constants.OOB_POWER_OFF):
3562 self.cfg.Update(node, feedback_fn)
3564 node_entry.append((constants.RS_NORMAL, result.payload))
3566 if (self.op.command == constants.OOB_POWER_ON and
3567 idx < len(self.nodes) - 1):
3568 time.sleep(self.op.power_delay)
3572 def _CheckPayload(self, result):
3573 """Checks if the payload is valid.
3575 @param result: RPC result
3576 @raises errors.OpExecError: If payload is not valid
3580 if self.op.command == constants.OOB_HEALTH:
3581 if not isinstance(result.payload, list):
3582 errs.append("command 'health' is expected to return a list but got %s" %
3583 type(result.payload))
3585 for item, status in result.payload:
3586 if status not in constants.OOB_STATUSES:
3587 errs.append("health item '%s' has invalid status '%s'" %
3590 if self.op.command == constants.OOB_POWER_STATUS:
3591 if not isinstance(result.payload, dict):
3592 errs.append("power-status is expected to return a dict but got %s" %
3593 type(result.payload))
3595 if self.op.command in [
3596 constants.OOB_POWER_ON,
3597 constants.OOB_POWER_OFF,
3598 constants.OOB_POWER_CYCLE,
3600 if result.payload is not None:
3601 errs.append("%s is expected to not return payload but got '%s'" %
3602 (self.op.command, result.payload))
3605 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3606 utils.CommaJoin(errs))
3608 class _OsQuery(_QueryBase):
3609 FIELDS = query.OS_FIELDS
3611 def ExpandNames(self, lu):
3612 # Lock all nodes in shared mode
3613 # Temporary removal of locks, should be reverted later
3614 # TODO: reintroduce locks when they are lighter-weight
3615 lu.needed_locks = {}
3616 #self.share_locks[locking.LEVEL_NODE] = 1
3617 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3619 # The following variables interact with _QueryBase._GetNames
3621 self.wanted = self.names
3623 self.wanted = locking.ALL_SET
3625 self.do_locking = self.use_locking
3627 def DeclareLocks(self, lu, level):
3631 def _DiagnoseByOS(rlist):
3632 """Remaps a per-node return list into an a per-os per-node dictionary
3634 @param rlist: a map with node names as keys and OS objects as values
3637 @return: a dictionary with osnames as keys and as value another
3638 map, with nodes as keys and tuples of (path, status, diagnose,
3639 variants, parameters, api_versions) as values, eg::
3641 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3642 (/srv/..., False, "invalid api")],
3643 "node2": [(/srv/..., True, "", [], [])]}
3648 # we build here the list of nodes that didn't fail the RPC (at RPC
3649 # level), so that nodes with a non-responding node daemon don't
3650 # make all OSes invalid
3651 good_nodes = [node_name for node_name in rlist
3652 if not rlist[node_name].fail_msg]
3653 for node_name, nr in rlist.items():
3654 if nr.fail_msg or not nr.payload:
3656 for (name, path, status, diagnose, variants,
3657 params, api_versions) in nr.payload:
3658 if name not in all_os:
3659 # build a list of nodes for this os containing empty lists
3660 # for each node in node_list
3662 for nname in good_nodes:
3663 all_os[name][nname] = []
3664 # convert params from [name, help] to (name, help)
3665 params = [tuple(v) for v in params]
3666 all_os[name][node_name].append((path, status, diagnose,
3667 variants, params, api_versions))
3670 def _GetQueryData(self, lu):
3671 """Computes the list of nodes and their attributes.
3674 # Locking is not used
3675 assert not (lu.acquired_locks or self.do_locking or self.use_locking)
3677 valid_nodes = [node.name
3678 for node in lu.cfg.GetAllNodesInfo().values()
3679 if not node.offline and node.vm_capable]
3680 pol = self._DiagnoseByOS(lu.rpc.call_os_diagnose(valid_nodes))
3681 cluster = lu.cfg.GetClusterInfo()
3685 for (os_name, os_data) in pol.items():
3686 info = query.OsInfo(name=os_name, valid=True, node_status=os_data,
3687 hidden=(os_name in cluster.hidden_os),
3688 blacklisted=(os_name in cluster.blacklisted_os))
3692 api_versions = set()
3694 for idx, osl in enumerate(os_data.values()):
3695 info.valid = bool(info.valid and osl and osl[0][1])
3699 (node_variants, node_params, node_api) = osl[0][3:6]
3702 variants.update(node_variants)
3703 parameters.update(node_params)
3704 api_versions.update(node_api)
3706 # Filter out inconsistent values
3707 variants.intersection_update(node_variants)
3708 parameters.intersection_update(node_params)
3709 api_versions.intersection_update(node_api)
3711 info.variants = list(variants)
3712 info.parameters = list(parameters)
3713 info.api_versions = list(api_versions)
3715 data[os_name] = info
3717 # Prepare data in requested order
3718 return [data[name] for name in self._GetNames(lu, pol.keys(), None)
3722 class LUOsDiagnose(NoHooksLU):
3723 """Logical unit for OS diagnose/query.
3729 def _BuildFilter(fields, names):
3730 """Builds a filter for querying OSes.
3733 name_filter = qlang.MakeSimpleFilter("name", names)
3735 # Legacy behaviour: Hide hidden, blacklisted or invalid OSes if the
3736 # respective field is not requested
3737 status_filter = [[qlang.OP_NOT, [qlang.OP_TRUE, fname]]
3738 for fname in ["hidden", "blacklisted"]
3739 if fname not in fields]
3740 if "valid" not in fields:
3741 status_filter.append([qlang.OP_TRUE, "valid"])
3744 status_filter.insert(0, qlang.OP_AND)
3746 status_filter = None
3748 if name_filter and status_filter:
3749 return [qlang.OP_AND, name_filter, status_filter]
3753 return status_filter
3755 def CheckArguments(self):
3756 self.oq = _OsQuery(self._BuildFilter(self.op.output_fields, self.op.names),
3757 self.op.output_fields, False)
3759 def ExpandNames(self):
3760 self.oq.ExpandNames(self)
3762 def Exec(self, feedback_fn):
3763 return self.oq.OldStyleQuery(self)
3766 class LUNodeRemove(LogicalUnit):
3767 """Logical unit for removing a node.
3770 HPATH = "node-remove"
3771 HTYPE = constants.HTYPE_NODE
3773 def BuildHooksEnv(self):
3776 This doesn't run on the target node in the pre phase as a failed
3777 node would then be impossible to remove.
3781 "OP_TARGET": self.op.node_name,
3782 "NODE_NAME": self.op.node_name,
3785 def BuildHooksNodes(self):
3786 """Build hooks nodes.
3789 all_nodes = self.cfg.GetNodeList()
3791 all_nodes.remove(self.op.node_name)
3793 logging.warning("Node '%s', which is about to be removed, was not found"
3794 " in the list of all nodes", self.op.node_name)
3795 return (all_nodes, all_nodes)
3797 def CheckPrereq(self):
3798 """Check prerequisites.
3801 - the node exists in the configuration
3802 - it does not have primary or secondary instances
3803 - it's not the master
3805 Any errors are signaled by raising errors.OpPrereqError.
3808 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3809 node = self.cfg.GetNodeInfo(self.op.node_name)
3810 assert node is not None
3812 instance_list = self.cfg.GetInstanceList()
3814 masternode = self.cfg.GetMasterNode()
3815 if node.name == masternode:
3816 raise errors.OpPrereqError("Node is the master node, failover to another"
3817 " node is required", errors.ECODE_INVAL)
3819 for instance_name in instance_list:
3820 instance = self.cfg.GetInstanceInfo(instance_name)
3821 if node.name in instance.all_nodes:
3822 raise errors.OpPrereqError("Instance %s is still running on the node,"
3823 " please remove first" % instance_name,
3825 self.op.node_name = node.name
3828 def Exec(self, feedback_fn):
3829 """Removes the node from the cluster.
3833 logging.info("Stopping the node daemon and removing configs from node %s",
3836 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3838 # Promote nodes to master candidate as needed
3839 _AdjustCandidatePool(self, exceptions=[node.name])
3840 self.context.RemoveNode(node.name)
3842 # Run post hooks on the node before it's removed
3843 _RunPostHook(self, node.name)
3845 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3846 msg = result.fail_msg
3848 self.LogWarning("Errors encountered on the remote node while leaving"
3849 " the cluster: %s", msg)
3851 # Remove node from our /etc/hosts
3852 if self.cfg.GetClusterInfo().modify_etc_hosts:
3853 master_node = self.cfg.GetMasterNode()
3854 result = self.rpc.call_etc_hosts_modify(master_node,
3855 constants.ETC_HOSTS_REMOVE,
3857 result.Raise("Can't update hosts file with new host data")
3858 _RedistributeAncillaryFiles(self)
3861 class _NodeQuery(_QueryBase):
3862 FIELDS = query.NODE_FIELDS
3864 def ExpandNames(self, lu):
3865 lu.needed_locks = {}
3866 lu.share_locks[locking.LEVEL_NODE] = 1
3869 self.wanted = _GetWantedNodes(lu, self.names)
3871 self.wanted = locking.ALL_SET
3873 self.do_locking = (self.use_locking and
3874 query.NQ_LIVE in self.requested_data)
3877 # if we don't request only static fields, we need to lock the nodes
3878 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3880 def DeclareLocks(self, lu, level):
3883 def _GetQueryData(self, lu):
3884 """Computes the list of nodes and their attributes.
3887 all_info = lu.cfg.GetAllNodesInfo()
3889 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3891 # Gather data as requested
3892 if query.NQ_LIVE in self.requested_data:
3893 # filter out non-vm_capable nodes
3894 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
3896 node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
3897 lu.cfg.GetHypervisorType())
3898 live_data = dict((name, nresult.payload)
3899 for (name, nresult) in node_data.items()
3900 if not nresult.fail_msg and nresult.payload)
3904 if query.NQ_INST in self.requested_data:
3905 node_to_primary = dict([(name, set()) for name in nodenames])
3906 node_to_secondary = dict([(name, set()) for name in nodenames])
3908 inst_data = lu.cfg.GetAllInstancesInfo()
3910 for inst in inst_data.values():
3911 if inst.primary_node in node_to_primary:
3912 node_to_primary[inst.primary_node].add(inst.name)
3913 for secnode in inst.secondary_nodes:
3914 if secnode in node_to_secondary:
3915 node_to_secondary[secnode].add(inst.name)
3917 node_to_primary = None
3918 node_to_secondary = None
3920 if query.NQ_OOB in self.requested_data:
3921 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3922 for name, node in all_info.iteritems())
3926 if query.NQ_GROUP in self.requested_data:
3927 groups = lu.cfg.GetAllNodeGroupsInfo()
3931 return query.NodeQueryData([all_info[name] for name in nodenames],
3932 live_data, lu.cfg.GetMasterNode(),
3933 node_to_primary, node_to_secondary, groups,
3934 oob_support, lu.cfg.GetClusterInfo())
3937 class LUNodeQuery(NoHooksLU):
3938 """Logical unit for querying nodes.
3941 # pylint: disable-msg=W0142
3944 def CheckArguments(self):
3945 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
3946 self.op.output_fields, self.op.use_locking)
3948 def ExpandNames(self):
3949 self.nq.ExpandNames(self)
3951 def Exec(self, feedback_fn):
3952 return self.nq.OldStyleQuery(self)
3955 class LUNodeQueryvols(NoHooksLU):
3956 """Logical unit for getting volumes on node(s).
3960 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3961 _FIELDS_STATIC = utils.FieldSet("node")
3963 def CheckArguments(self):
3964 _CheckOutputFields(static=self._FIELDS_STATIC,
3965 dynamic=self._FIELDS_DYNAMIC,
3966 selected=self.op.output_fields)
3968 def ExpandNames(self):
3969 self.needed_locks = {}
3970 self.share_locks[locking.LEVEL_NODE] = 1
3971 if not self.op.nodes:
3972 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3974 self.needed_locks[locking.LEVEL_NODE] = \
3975 _GetWantedNodes(self, self.op.nodes)
3977 def Exec(self, feedback_fn):
3978 """Computes the list of nodes and their attributes.
3981 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3982 volumes = self.rpc.call_node_volumes(nodenames)
3984 ilist = [self.cfg.GetInstanceInfo(iname) for iname
3985 in self.cfg.GetInstanceList()]
3987 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3990 for node in nodenames:
3991 nresult = volumes[node]
3994 msg = nresult.fail_msg
3996 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3999 node_vols = nresult.payload[:]
4000 node_vols.sort(key=lambda vol: vol['dev'])
4002 for vol in node_vols:
4004 for field in self.op.output_fields:
4007 elif field == "phys":
4011 elif field == "name":
4013 elif field == "size":
4014 val = int(float(vol['size']))
4015 elif field == "instance":
4017 if node not in lv_by_node[inst]:
4019 if vol['name'] in lv_by_node[inst][node]:
4025 raise errors.ParameterError(field)
4026 node_output.append(str(val))
4028 output.append(node_output)
4033 class LUNodeQueryStorage(NoHooksLU):
4034 """Logical unit for getting information on storage units on node(s).
4037 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
4040 def CheckArguments(self):
4041 _CheckOutputFields(static=self._FIELDS_STATIC,
4042 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
4043 selected=self.op.output_fields)
4045 def ExpandNames(self):
4046 self.needed_locks = {}
4047 self.share_locks[locking.LEVEL_NODE] = 1
4050 self.needed_locks[locking.LEVEL_NODE] = \
4051 _GetWantedNodes(self, self.op.nodes)
4053 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4055 def Exec(self, feedback_fn):
4056 """Computes the list of nodes and their attributes.
4059 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
4061 # Always get name to sort by
4062 if constants.SF_NAME in self.op.output_fields:
4063 fields = self.op.output_fields[:]
4065 fields = [constants.SF_NAME] + self.op.output_fields
4067 # Never ask for node or type as it's only known to the LU
4068 for extra in [constants.SF_NODE, constants.SF_TYPE]:
4069 while extra in fields:
4070 fields.remove(extra)
4072 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
4073 name_idx = field_idx[constants.SF_NAME]
4075 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4076 data = self.rpc.call_storage_list(self.nodes,
4077 self.op.storage_type, st_args,
4078 self.op.name, fields)
4082 for node in utils.NiceSort(self.nodes):
4083 nresult = data[node]
4087 msg = nresult.fail_msg
4089 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
4092 rows = dict([(row[name_idx], row) for row in nresult.payload])
4094 for name in utils.NiceSort(rows.keys()):
4099 for field in self.op.output_fields:
4100 if field == constants.SF_NODE:
4102 elif field == constants.SF_TYPE:
4103 val = self.op.storage_type
4104 elif field in field_idx:
4105 val = row[field_idx[field]]
4107 raise errors.ParameterError(field)
4116 class _InstanceQuery(_QueryBase):
4117 FIELDS = query.INSTANCE_FIELDS
4119 def ExpandNames(self, lu):
4120 lu.needed_locks = {}
4121 lu.share_locks[locking.LEVEL_INSTANCE] = 1
4122 lu.share_locks[locking.LEVEL_NODE] = 1
4125 self.wanted = _GetWantedInstances(lu, self.names)
4127 self.wanted = locking.ALL_SET
4129 self.do_locking = (self.use_locking and
4130 query.IQ_LIVE in self.requested_data)
4132 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4133 lu.needed_locks[locking.LEVEL_NODE] = []
4134 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4136 def DeclareLocks(self, lu, level):
4137 if level == locking.LEVEL_NODE and self.do_locking:
4138 lu._LockInstancesNodes() # pylint: disable-msg=W0212
4140 def _GetQueryData(self, lu):
4141 """Computes the list of instances and their attributes.
4144 cluster = lu.cfg.GetClusterInfo()
4145 all_info = lu.cfg.GetAllInstancesInfo()
4147 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
4149 instance_list = [all_info[name] for name in instance_names]
4150 nodes = frozenset(itertools.chain(*(inst.all_nodes
4151 for inst in instance_list)))
4152 hv_list = list(set([inst.hypervisor for inst in instance_list]))
4155 wrongnode_inst = set()
4157 # Gather data as requested
4158 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
4160 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
4162 result = node_data[name]
4164 # offline nodes will be in both lists
4165 assert result.fail_msg
4166 offline_nodes.append(name)
4168 bad_nodes.append(name)
4169 elif result.payload:
4170 for inst in result.payload:
4171 if inst in all_info:
4172 if all_info[inst].primary_node == name:
4173 live_data.update(result.payload)
4175 wrongnode_inst.add(inst)
4177 # orphan instance; we don't list it here as we don't
4178 # handle this case yet in the output of instance listing
4179 logging.warning("Orphan instance '%s' found on node %s",
4181 # else no instance is alive
4185 if query.IQ_DISKUSAGE in self.requested_data:
4186 disk_usage = dict((inst.name,
4187 _ComputeDiskSize(inst.disk_template,
4188 [{constants.IDISK_SIZE: disk.size}
4189 for disk in inst.disks]))
4190 for inst in instance_list)
4194 if query.IQ_CONSOLE in self.requested_data:
4196 for inst in instance_list:
4197 if inst.name in live_data:
4198 # Instance is running
4199 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
4201 consinfo[inst.name] = None
4202 assert set(consinfo.keys()) == set(instance_names)
4206 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
4207 disk_usage, offline_nodes, bad_nodes,
4208 live_data, wrongnode_inst, consinfo)
4211 class LUQuery(NoHooksLU):
4212 """Query for resources/items of a certain kind.
4215 # pylint: disable-msg=W0142
4218 def CheckArguments(self):
4219 qcls = _GetQueryImplementation(self.op.what)
4221 self.impl = qcls(self.op.filter, self.op.fields, False)
4223 def ExpandNames(self):
4224 self.impl.ExpandNames(self)
4226 def DeclareLocks(self, level):
4227 self.impl.DeclareLocks(self, level)
4229 def Exec(self, feedback_fn):
4230 return self.impl.NewStyleQuery(self)
4233 class LUQueryFields(NoHooksLU):
4234 """Query for resources/items of a certain kind.
4237 # pylint: disable-msg=W0142
4240 def CheckArguments(self):
4241 self.qcls = _GetQueryImplementation(self.op.what)
4243 def ExpandNames(self):
4244 self.needed_locks = {}
4246 def Exec(self, feedback_fn):
4247 return query.QueryFields(self.qcls.FIELDS, self.op.fields)
4250 class LUNodeModifyStorage(NoHooksLU):
4251 """Logical unit for modifying a storage volume on a node.
4256 def CheckArguments(self):
4257 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4259 storage_type = self.op.storage_type
4262 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4264 raise errors.OpPrereqError("Storage units of type '%s' can not be"
4265 " modified" % storage_type,
4268 diff = set(self.op.changes.keys()) - modifiable
4270 raise errors.OpPrereqError("The following fields can not be modified for"
4271 " storage units of type '%s': %r" %
4272 (storage_type, list(diff)),
4275 def ExpandNames(self):
4276 self.needed_locks = {
4277 locking.LEVEL_NODE: self.op.node_name,
4280 def Exec(self, feedback_fn):
4281 """Computes the list of nodes and their attributes.
4284 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4285 result = self.rpc.call_storage_modify(self.op.node_name,
4286 self.op.storage_type, st_args,
4287 self.op.name, self.op.changes)
4288 result.Raise("Failed to modify storage unit '%s' on %s" %
4289 (self.op.name, self.op.node_name))
4292 class LUNodeAdd(LogicalUnit):
4293 """Logical unit for adding node to the cluster.
4297 HTYPE = constants.HTYPE_NODE
4298 _NFLAGS = ["master_capable", "vm_capable"]
4300 def CheckArguments(self):
4301 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4302 # validate/normalize the node name
4303 self.hostname = netutils.GetHostname(name=self.op.node_name,
4304 family=self.primary_ip_family)
4305 self.op.node_name = self.hostname.name
4307 if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
4308 raise errors.OpPrereqError("Cannot readd the master node",
4311 if self.op.readd and self.op.group:
4312 raise errors.OpPrereqError("Cannot pass a node group when a node is"
4313 " being readded", errors.ECODE_INVAL)
4315 def BuildHooksEnv(self):
4318 This will run on all nodes before, and on all nodes + the new node after.
4322 "OP_TARGET": self.op.node_name,
4323 "NODE_NAME": self.op.node_name,
4324 "NODE_PIP": self.op.primary_ip,
4325 "NODE_SIP": self.op.secondary_ip,
4326 "MASTER_CAPABLE": str(self.op.master_capable),
4327 "VM_CAPABLE": str(self.op.vm_capable),
4330 def BuildHooksNodes(self):
4331 """Build hooks nodes.
4334 # Exclude added node
4335 pre_nodes = list(set(self.cfg.GetNodeList()) - set([self.op.node_name]))
4336 post_nodes = pre_nodes + [self.op.node_name, ]
4338 return (pre_nodes, post_nodes)
4340 def CheckPrereq(self):
4341 """Check prerequisites.
4344 - the new node is not already in the config
4346 - its parameters (single/dual homed) matches the cluster
4348 Any errors are signaled by raising errors.OpPrereqError.
4352 hostname = self.hostname
4353 node = hostname.name
4354 primary_ip = self.op.primary_ip = hostname.ip
4355 if self.op.secondary_ip is None:
4356 if self.primary_ip_family == netutils.IP6Address.family:
4357 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4358 " IPv4 address must be given as secondary",
4360 self.op.secondary_ip = primary_ip
4362 secondary_ip = self.op.secondary_ip
4363 if not netutils.IP4Address.IsValid(secondary_ip):
4364 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4365 " address" % secondary_ip, errors.ECODE_INVAL)
4367 node_list = cfg.GetNodeList()
4368 if not self.op.readd and node in node_list:
4369 raise errors.OpPrereqError("Node %s is already in the configuration" %
4370 node, errors.ECODE_EXISTS)
4371 elif self.op.readd and node not in node_list:
4372 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4375 self.changed_primary_ip = False
4377 for existing_node_name in node_list:
4378 existing_node = cfg.GetNodeInfo(existing_node_name)
4380 if self.op.readd and node == existing_node_name:
4381 if existing_node.secondary_ip != secondary_ip:
4382 raise errors.OpPrereqError("Readded node doesn't have the same IP"
4383 " address configuration as before",
4385 if existing_node.primary_ip != primary_ip:
4386 self.changed_primary_ip = True
4390 if (existing_node.primary_ip == primary_ip or
4391 existing_node.secondary_ip == primary_ip or
4392 existing_node.primary_ip == secondary_ip or
4393 existing_node.secondary_ip == secondary_ip):
4394 raise errors.OpPrereqError("New node ip address(es) conflict with"
4395 " existing node %s" % existing_node.name,
4396 errors.ECODE_NOTUNIQUE)
4398 # After this 'if' block, None is no longer a valid value for the
4399 # _capable op attributes
4401 old_node = self.cfg.GetNodeInfo(node)
4402 assert old_node is not None, "Can't retrieve locked node %s" % node
4403 for attr in self._NFLAGS:
4404 if getattr(self.op, attr) is None:
4405 setattr(self.op, attr, getattr(old_node, attr))
4407 for attr in self._NFLAGS:
4408 if getattr(self.op, attr) is None:
4409 setattr(self.op, attr, True)
4411 if self.op.readd and not self.op.vm_capable:
4412 pri, sec = cfg.GetNodeInstances(node)
4414 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4415 " flag set to false, but it already holds"
4416 " instances" % node,
4419 # check that the type of the node (single versus dual homed) is the
4420 # same as for the master
4421 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4422 master_singlehomed = myself.secondary_ip == myself.primary_ip
4423 newbie_singlehomed = secondary_ip == primary_ip
4424 if master_singlehomed != newbie_singlehomed:
4425 if master_singlehomed:
4426 raise errors.OpPrereqError("The master has no secondary ip but the"
4427 " new node has one",
4430 raise errors.OpPrereqError("The master has a secondary ip but the"
4431 " new node doesn't have one",
4434 # checks reachability
4435 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
4436 raise errors.OpPrereqError("Node not reachable by ping",
4437 errors.ECODE_ENVIRON)
4439 if not newbie_singlehomed:
4440 # check reachability from my secondary ip to newbie's secondary ip
4441 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
4442 source=myself.secondary_ip):
4443 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4444 " based ping to node daemon port",
4445 errors.ECODE_ENVIRON)
4452 if self.op.master_capable:
4453 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
4455 self.master_candidate = False
4458 self.new_node = old_node
4460 node_group = cfg.LookupNodeGroup(self.op.group)
4461 self.new_node = objects.Node(name=node,
4462 primary_ip=primary_ip,
4463 secondary_ip=secondary_ip,
4464 master_candidate=self.master_candidate,
4465 offline=False, drained=False,
4468 if self.op.ndparams:
4469 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
4471 def Exec(self, feedback_fn):
4472 """Adds the new node to the cluster.
4475 new_node = self.new_node
4476 node = new_node.name
4478 # We adding a new node so we assume it's powered
4479 new_node.powered = True
4481 # for re-adds, reset the offline/drained/master-candidate flags;
4482 # we need to reset here, otherwise offline would prevent RPC calls
4483 # later in the procedure; this also means that if the re-add
4484 # fails, we are left with a non-offlined, broken node
4486 new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
4487 self.LogInfo("Readding a node, the offline/drained flags were reset")
4488 # if we demote the node, we do cleanup later in the procedure
4489 new_node.master_candidate = self.master_candidate
4490 if self.changed_primary_ip:
4491 new_node.primary_ip = self.op.primary_ip
4493 # copy the master/vm_capable flags
4494 for attr in self._NFLAGS:
4495 setattr(new_node, attr, getattr(self.op, attr))
4497 # notify the user about any possible mc promotion
4498 if new_node.master_candidate:
4499 self.LogInfo("Node will be a master candidate")
4501 if self.op.ndparams:
4502 new_node.ndparams = self.op.ndparams
4504 new_node.ndparams = {}
4506 # check connectivity
4507 result = self.rpc.call_version([node])[node]
4508 result.Raise("Can't get version information from node %s" % node)
4509 if constants.PROTOCOL_VERSION == result.payload:
4510 logging.info("Communication to node %s fine, sw version %s match",
4511 node, result.payload)
4513 raise errors.OpExecError("Version mismatch master version %s,"
4514 " node version %s" %
4515 (constants.PROTOCOL_VERSION, result.payload))
4517 # Add node to our /etc/hosts, and add key to known_hosts
4518 if self.cfg.GetClusterInfo().modify_etc_hosts:
4519 master_node = self.cfg.GetMasterNode()
4520 result = self.rpc.call_etc_hosts_modify(master_node,
4521 constants.ETC_HOSTS_ADD,
4524 result.Raise("Can't update hosts file with new host data")
4526 if new_node.secondary_ip != new_node.primary_ip:
4527 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
4530 node_verify_list = [self.cfg.GetMasterNode()]
4531 node_verify_param = {
4532 constants.NV_NODELIST: [node],
4533 # TODO: do a node-net-test as well?
4536 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
4537 self.cfg.GetClusterName())
4538 for verifier in node_verify_list:
4539 result[verifier].Raise("Cannot communicate with node %s" % verifier)
4540 nl_payload = result[verifier].payload[constants.NV_NODELIST]
4542 for failed in nl_payload:
4543 feedback_fn("ssh/hostname verification failed"
4544 " (checking from %s): %s" %
4545 (verifier, nl_payload[failed]))
4546 raise errors.OpExecError("ssh/hostname verification failed")
4549 _RedistributeAncillaryFiles(self)
4550 self.context.ReaddNode(new_node)
4551 # make sure we redistribute the config
4552 self.cfg.Update(new_node, feedback_fn)
4553 # and make sure the new node will not have old files around
4554 if not new_node.master_candidate:
4555 result = self.rpc.call_node_demote_from_mc(new_node.name)
4556 msg = result.fail_msg
4558 self.LogWarning("Node failed to demote itself from master"
4559 " candidate status: %s" % msg)
4561 _RedistributeAncillaryFiles(self, additional_nodes=[node],
4562 additional_vm=self.op.vm_capable)
4563 self.context.AddNode(new_node, self.proc.GetECId())
4566 class LUNodeSetParams(LogicalUnit):
4567 """Modifies the parameters of a node.
4569 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
4570 to the node role (as _ROLE_*)
4571 @cvar _R2F: a dictionary from node role to tuples of flags
4572 @cvar _FLAGS: a list of attribute names corresponding to the flags
4575 HPATH = "node-modify"
4576 HTYPE = constants.HTYPE_NODE
4578 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
4580 (True, False, False): _ROLE_CANDIDATE,
4581 (False, True, False): _ROLE_DRAINED,
4582 (False, False, True): _ROLE_OFFLINE,
4583 (False, False, False): _ROLE_REGULAR,
4585 _R2F = dict((v, k) for k, v in _F2R.items())
4586 _FLAGS = ["master_candidate", "drained", "offline"]
4588 def CheckArguments(self):
4589 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4590 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
4591 self.op.master_capable, self.op.vm_capable,
4592 self.op.secondary_ip, self.op.ndparams]
4593 if all_mods.count(None) == len(all_mods):
4594 raise errors.OpPrereqError("Please pass at least one modification",
4596 if all_mods.count(True) > 1:
4597 raise errors.OpPrereqError("Can't set the node into more than one"
4598 " state at the same time",
4601 # Boolean value that tells us whether we might be demoting from MC
4602 self.might_demote = (self.op.master_candidate == False or
4603 self.op.offline == True or
4604 self.op.drained == True or
4605 self.op.master_capable == False)
4607 if self.op.secondary_ip:
4608 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
4609 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4610 " address" % self.op.secondary_ip,
4613 self.lock_all = self.op.auto_promote and self.might_demote
4614 self.lock_instances = self.op.secondary_ip is not None
4616 def ExpandNames(self):
4618 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
4620 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
4622 if self.lock_instances:
4623 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4625 def DeclareLocks(self, level):
4626 # If we have locked all instances, before waiting to lock nodes, release
4627 # all the ones living on nodes unrelated to the current operation.
4628 if level == locking.LEVEL_NODE and self.lock_instances:
4629 self.affected_instances = []
4630 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4633 # Build list of instances to release
4634 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
4635 instance = self.context.cfg.GetInstanceInfo(instance_name)
4636 if (instance.disk_template in constants.DTS_INT_MIRROR and
4637 self.op.node_name in instance.all_nodes):
4638 instances_keep.append(instance_name)
4639 self.affected_instances.append(instance)
4641 _ReleaseLocks(self, locking.LEVEL_INSTANCE, keep=instances_keep)
4643 assert (set(self.acquired_locks.get(locking.LEVEL_INSTANCE, [])) ==
4644 set(instances_keep))
4646 def BuildHooksEnv(self):
4649 This runs on the master node.
4653 "OP_TARGET": self.op.node_name,
4654 "MASTER_CANDIDATE": str(self.op.master_candidate),
4655 "OFFLINE": str(self.op.offline),
4656 "DRAINED": str(self.op.drained),
4657 "MASTER_CAPABLE": str(self.op.master_capable),
4658 "VM_CAPABLE": str(self.op.vm_capable),
4661 def BuildHooksNodes(self):
4662 """Build hooks nodes.
4665 nl = [self.cfg.GetMasterNode(), self.op.node_name]
4668 def CheckPrereq(self):
4669 """Check prerequisites.
4671 This only checks the instance list against the existing names.
4674 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
4676 if (self.op.master_candidate is not None or
4677 self.op.drained is not None or
4678 self.op.offline is not None):
4679 # we can't change the master's node flags
4680 if self.op.node_name == self.cfg.GetMasterNode():
4681 raise errors.OpPrereqError("The master role can be changed"
4682 " only via master-failover",
4685 if self.op.master_candidate and not node.master_capable:
4686 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
4687 " it a master candidate" % node.name,
4690 if self.op.vm_capable == False:
4691 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
4693 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
4694 " the vm_capable flag" % node.name,
4697 if node.master_candidate and self.might_demote and not self.lock_all:
4698 assert not self.op.auto_promote, "auto_promote set but lock_all not"
4699 # check if after removing the current node, we're missing master
4701 (mc_remaining, mc_should, _) = \
4702 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
4703 if mc_remaining < mc_should:
4704 raise errors.OpPrereqError("Not enough master candidates, please"
4705 " pass auto promote option to allow"
4706 " promotion", errors.ECODE_STATE)
4708 self.old_flags = old_flags = (node.master_candidate,
4709 node.drained, node.offline)
4710 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
4711 self.old_role = old_role = self._F2R[old_flags]
4713 # Check for ineffective changes
4714 for attr in self._FLAGS:
4715 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
4716 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
4717 setattr(self.op, attr, None)
4719 # Past this point, any flag change to False means a transition
4720 # away from the respective state, as only real changes are kept
4722 # TODO: We might query the real power state if it supports OOB
4723 if _SupportsOob(self.cfg, node):
4724 if self.op.offline is False and not (node.powered or
4725 self.op.powered == True):
4726 raise errors.OpPrereqError(("Node %s needs to be turned on before its"
4727 " offline status can be reset") %
4729 elif self.op.powered is not None:
4730 raise errors.OpPrereqError(("Unable to change powered state for node %s"
4731 " as it does not support out-of-band"
4732 " handling") % self.op.node_name)
4734 # If we're being deofflined/drained, we'll MC ourself if needed
4735 if (self.op.drained == False or self.op.offline == False or
4736 (self.op.master_capable and not node.master_capable)):
4737 if _DecideSelfPromotion(self):
4738 self.op.master_candidate = True
4739 self.LogInfo("Auto-promoting node to master candidate")
4741 # If we're no longer master capable, we'll demote ourselves from MC
4742 if self.op.master_capable == False and node.master_candidate:
4743 self.LogInfo("Demoting from master candidate")
4744 self.op.master_candidate = False
4747 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
4748 if self.op.master_candidate:
4749 new_role = self._ROLE_CANDIDATE
4750 elif self.op.drained:
4751 new_role = self._ROLE_DRAINED
4752 elif self.op.offline:
4753 new_role = self._ROLE_OFFLINE
4754 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
4755 # False is still in new flags, which means we're un-setting (the
4757 new_role = self._ROLE_REGULAR
4758 else: # no new flags, nothing, keep old role
4761 self.new_role = new_role
4763 if old_role == self._ROLE_OFFLINE and new_role != old_role:
4764 # Trying to transition out of offline status
4765 result = self.rpc.call_version([node.name])[node.name]
4767 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
4768 " to report its version: %s" %
4769 (node.name, result.fail_msg),
4772 self.LogWarning("Transitioning node from offline to online state"
4773 " without using re-add. Please make sure the node"
4776 if self.op.secondary_ip:
4777 # Ok even without locking, because this can't be changed by any LU
4778 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
4779 master_singlehomed = master.secondary_ip == master.primary_ip
4780 if master_singlehomed and self.op.secondary_ip:
4781 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
4782 " homed cluster", errors.ECODE_INVAL)
4785 if self.affected_instances:
4786 raise errors.OpPrereqError("Cannot change secondary ip: offline"
4787 " node has instances (%s) configured"
4788 " to use it" % self.affected_instances)
4790 # On online nodes, check that no instances are running, and that
4791 # the node has the new ip and we can reach it.
4792 for instance in self.affected_instances:
4793 _CheckInstanceDown(self, instance, "cannot change secondary ip")
4795 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
4796 if master.name != node.name:
4797 # check reachability from master secondary ip to new secondary ip
4798 if not netutils.TcpPing(self.op.secondary_ip,
4799 constants.DEFAULT_NODED_PORT,
4800 source=master.secondary_ip):
4801 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4802 " based ping to node daemon port",
4803 errors.ECODE_ENVIRON)
4805 if self.op.ndparams:
4806 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
4807 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
4808 self.new_ndparams = new_ndparams
4810 def Exec(self, feedback_fn):
4815 old_role = self.old_role
4816 new_role = self.new_role
4820 if self.op.ndparams:
4821 node.ndparams = self.new_ndparams
4823 if self.op.powered is not None:
4824 node.powered = self.op.powered
4826 for attr in ["master_capable", "vm_capable"]:
4827 val = getattr(self.op, attr)
4829 setattr(node, attr, val)
4830 result.append((attr, str(val)))
4832 if new_role != old_role:
4833 # Tell the node to demote itself, if no longer MC and not offline
4834 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
4835 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
4837 self.LogWarning("Node failed to demote itself: %s", msg)
4839 new_flags = self._R2F[new_role]
4840 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
4842 result.append((desc, str(nf)))
4843 (node.master_candidate, node.drained, node.offline) = new_flags
4845 # we locked all nodes, we adjust the CP before updating this node
4847 _AdjustCandidatePool(self, [node.name])
4849 if self.op.secondary_ip:
4850 node.secondary_ip = self.op.secondary_ip
4851 result.append(("secondary_ip", self.op.secondary_ip))
4853 # this will trigger configuration file update, if needed
4854 self.cfg.Update(node, feedback_fn)
4856 # this will trigger job queue propagation or cleanup if the mc
4858 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
4859 self.context.ReaddNode(node)
4864 class LUNodePowercycle(NoHooksLU):
4865 """Powercycles a node.
4870 def CheckArguments(self):
4871 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4872 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4873 raise errors.OpPrereqError("The node is the master and the force"
4874 " parameter was not set",
4877 def ExpandNames(self):
4878 """Locking for PowercycleNode.
4880 This is a last-resort option and shouldn't block on other
4881 jobs. Therefore, we grab no locks.
4884 self.needed_locks = {}
4886 def Exec(self, feedback_fn):
4890 result = self.rpc.call_node_powercycle(self.op.node_name,
4891 self.cfg.GetHypervisorType())
4892 result.Raise("Failed to schedule the reboot")
4893 return result.payload
4896 class LUClusterQuery(NoHooksLU):
4897 """Query cluster configuration.
4902 def ExpandNames(self):
4903 self.needed_locks = {}
4905 def Exec(self, feedback_fn):
4906 """Return cluster config.
4909 cluster = self.cfg.GetClusterInfo()
4912 # Filter just for enabled hypervisors
4913 for os_name, hv_dict in cluster.os_hvp.items():
4914 os_hvp[os_name] = {}
4915 for hv_name, hv_params in hv_dict.items():
4916 if hv_name in cluster.enabled_hypervisors:
4917 os_hvp[os_name][hv_name] = hv_params
4919 # Convert ip_family to ip_version
4920 primary_ip_version = constants.IP4_VERSION
4921 if cluster.primary_ip_family == netutils.IP6Address.family:
4922 primary_ip_version = constants.IP6_VERSION
4925 "software_version": constants.RELEASE_VERSION,
4926 "protocol_version": constants.PROTOCOL_VERSION,
4927 "config_version": constants.CONFIG_VERSION,
4928 "os_api_version": max(constants.OS_API_VERSIONS),
4929 "export_version": constants.EXPORT_VERSION,
4930 "architecture": (platform.architecture()[0], platform.machine()),
4931 "name": cluster.cluster_name,
4932 "master": cluster.master_node,
4933 "default_hypervisor": cluster.enabled_hypervisors[0],
4934 "enabled_hypervisors": cluster.enabled_hypervisors,
4935 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4936 for hypervisor_name in cluster.enabled_hypervisors]),
4938 "beparams": cluster.beparams,
4939 "osparams": cluster.osparams,
4940 "nicparams": cluster.nicparams,
4941 "ndparams": cluster.ndparams,
4942 "candidate_pool_size": cluster.candidate_pool_size,
4943 "master_netdev": cluster.master_netdev,
4944 "volume_group_name": cluster.volume_group_name,
4945 "drbd_usermode_helper": cluster.drbd_usermode_helper,
4946 "file_storage_dir": cluster.file_storage_dir,
4947 "shared_file_storage_dir": cluster.shared_file_storage_dir,
4948 "maintain_node_health": cluster.maintain_node_health,
4949 "ctime": cluster.ctime,
4950 "mtime": cluster.mtime,
4951 "uuid": cluster.uuid,
4952 "tags": list(cluster.GetTags()),
4953 "uid_pool": cluster.uid_pool,
4954 "default_iallocator": cluster.default_iallocator,
4955 "reserved_lvs": cluster.reserved_lvs,
4956 "primary_ip_version": primary_ip_version,
4957 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4958 "hidden_os": cluster.hidden_os,
4959 "blacklisted_os": cluster.blacklisted_os,
4965 class LUClusterConfigQuery(NoHooksLU):
4966 """Return configuration values.
4970 _FIELDS_DYNAMIC = utils.FieldSet()
4971 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4972 "watcher_pause", "volume_group_name")
4974 def CheckArguments(self):
4975 _CheckOutputFields(static=self._FIELDS_STATIC,
4976 dynamic=self._FIELDS_DYNAMIC,
4977 selected=self.op.output_fields)
4979 def ExpandNames(self):
4980 self.needed_locks = {}
4982 def Exec(self, feedback_fn):
4983 """Dump a representation of the cluster config to the standard output.
4987 for field in self.op.output_fields:
4988 if field == "cluster_name":
4989 entry = self.cfg.GetClusterName()
4990 elif field == "master_node":
4991 entry = self.cfg.GetMasterNode()
4992 elif field == "drain_flag":
4993 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4994 elif field == "watcher_pause":
4995 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4996 elif field == "volume_group_name":
4997 entry = self.cfg.GetVGName()
4999 raise errors.ParameterError(field)
5000 values.append(entry)
5004 class LUInstanceActivateDisks(NoHooksLU):
5005 """Bring up an instance's disks.
5010 def ExpandNames(self):
5011 self._ExpandAndLockInstance()
5012 self.needed_locks[locking.LEVEL_NODE] = []
5013 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5015 def DeclareLocks(self, level):
5016 if level == locking.LEVEL_NODE:
5017 self._LockInstancesNodes()
5019 def CheckPrereq(self):
5020 """Check prerequisites.
5022 This checks that the instance is in the cluster.
5025 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5026 assert self.instance is not None, \
5027 "Cannot retrieve locked instance %s" % self.op.instance_name
5028 _CheckNodeOnline(self, self.instance.primary_node)
5030 def Exec(self, feedback_fn):
5031 """Activate the disks.
5034 disks_ok, disks_info = \
5035 _AssembleInstanceDisks(self, self.instance,
5036 ignore_size=self.op.ignore_size)
5038 raise errors.OpExecError("Cannot activate block devices")
5043 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
5045 """Prepare the block devices for an instance.
5047 This sets up the block devices on all nodes.
5049 @type lu: L{LogicalUnit}
5050 @param lu: the logical unit on whose behalf we execute
5051 @type instance: L{objects.Instance}
5052 @param instance: the instance for whose disks we assemble
5053 @type disks: list of L{objects.Disk} or None
5054 @param disks: which disks to assemble (or all, if None)
5055 @type ignore_secondaries: boolean
5056 @param ignore_secondaries: if true, errors on secondary nodes
5057 won't result in an error return from the function
5058 @type ignore_size: boolean
5059 @param ignore_size: if true, the current known size of the disk
5060 will not be used during the disk activation, useful for cases
5061 when the size is wrong
5062 @return: False if the operation failed, otherwise a list of
5063 (host, instance_visible_name, node_visible_name)
5064 with the mapping from node devices to instance devices
5069 iname = instance.name
5070 disks = _ExpandCheckDisks(instance, disks)
5072 # With the two passes mechanism we try to reduce the window of
5073 # opportunity for the race condition of switching DRBD to primary
5074 # before handshaking occured, but we do not eliminate it
5076 # The proper fix would be to wait (with some limits) until the
5077 # connection has been made and drbd transitions from WFConnection
5078 # into any other network-connected state (Connected, SyncTarget,
5081 # 1st pass, assemble on all nodes in secondary mode
5082 for idx, inst_disk in enumerate(disks):
5083 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
5085 node_disk = node_disk.Copy()
5086 node_disk.UnsetSize()
5087 lu.cfg.SetDiskID(node_disk, node)
5088 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
5089 msg = result.fail_msg
5091 lu.proc.LogWarning("Could not prepare block device %s on node %s"
5092 " (is_primary=False, pass=1): %s",
5093 inst_disk.iv_name, node, msg)
5094 if not ignore_secondaries:
5097 # FIXME: race condition on drbd migration to primary
5099 # 2nd pass, do only the primary node
5100 for idx, inst_disk in enumerate(disks):
5103 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
5104 if node != instance.primary_node:
5107 node_disk = node_disk.Copy()
5108 node_disk.UnsetSize()
5109 lu.cfg.SetDiskID(node_disk, node)
5110 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
5111 msg = result.fail_msg
5113 lu.proc.LogWarning("Could not prepare block device %s on node %s"
5114 " (is_primary=True, pass=2): %s",
5115 inst_disk.iv_name, node, msg)
5118 dev_path = result.payload
5120 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
5122 # leave the disks configured for the primary node
5123 # this is a workaround that would be fixed better by
5124 # improving the logical/physical id handling
5126 lu.cfg.SetDiskID(disk, instance.primary_node)
5128 return disks_ok, device_info
5131 def _StartInstanceDisks(lu, instance, force):
5132 """Start the disks of an instance.
5135 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
5136 ignore_secondaries=force)
5138 _ShutdownInstanceDisks(lu, instance)
5139 if force is not None and not force:
5140 lu.proc.LogWarning("", hint="If the message above refers to a"
5142 " you can retry the operation using '--force'.")
5143 raise errors.OpExecError("Disk consistency error")
5146 class LUInstanceDeactivateDisks(NoHooksLU):
5147 """Shutdown an instance's disks.
5152 def ExpandNames(self):
5153 self._ExpandAndLockInstance()
5154 self.needed_locks[locking.LEVEL_NODE] = []
5155 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5157 def DeclareLocks(self, level):
5158 if level == locking.LEVEL_NODE:
5159 self._LockInstancesNodes()
5161 def CheckPrereq(self):
5162 """Check prerequisites.
5164 This checks that the instance is in the cluster.
5167 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5168 assert self.instance is not None, \
5169 "Cannot retrieve locked instance %s" % self.op.instance_name
5171 def Exec(self, feedback_fn):
5172 """Deactivate the disks
5175 instance = self.instance
5177 _ShutdownInstanceDisks(self, instance)
5179 _SafeShutdownInstanceDisks(self, instance)
5182 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
5183 """Shutdown block devices of an instance.
5185 This function checks if an instance is running, before calling
5186 _ShutdownInstanceDisks.
5189 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
5190 _ShutdownInstanceDisks(lu, instance, disks=disks)
5193 def _ExpandCheckDisks(instance, disks):
5194 """Return the instance disks selected by the disks list
5196 @type disks: list of L{objects.Disk} or None
5197 @param disks: selected disks
5198 @rtype: list of L{objects.Disk}
5199 @return: selected instance disks to act on
5203 return instance.disks
5205 if not set(disks).issubset(instance.disks):
5206 raise errors.ProgrammerError("Can only act on disks belonging to the"
5211 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
5212 """Shutdown block devices of an instance.
5214 This does the shutdown on all nodes of the instance.
5216 If the ignore_primary is false, errors on the primary node are
5221 disks = _ExpandCheckDisks(instance, disks)
5224 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
5225 lu.cfg.SetDiskID(top_disk, node)
5226 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
5227 msg = result.fail_msg
5229 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
5230 disk.iv_name, node, msg)
5231 if ((node == instance.primary_node and not ignore_primary) or
5232 (node != instance.primary_node and not result.offline)):
5237 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
5238 """Checks if a node has enough free memory.
5240 This function check if a given node has the needed amount of free
5241 memory. In case the node has less memory or we cannot get the
5242 information from the node, this function raise an OpPrereqError
5245 @type lu: C{LogicalUnit}
5246 @param lu: a logical unit from which we get configuration data
5248 @param node: the node to check
5249 @type reason: C{str}
5250 @param reason: string to use in the error message
5251 @type requested: C{int}
5252 @param requested: the amount of memory in MiB to check for
5253 @type hypervisor_name: C{str}
5254 @param hypervisor_name: the hypervisor to ask for memory stats
5255 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
5256 we cannot check the node
5259 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
5260 nodeinfo[node].Raise("Can't get data from node %s" % node,
5261 prereq=True, ecode=errors.ECODE_ENVIRON)
5262 free_mem = nodeinfo[node].payload.get('memory_free', None)
5263 if not isinstance(free_mem, int):
5264 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
5265 " was '%s'" % (node, free_mem),
5266 errors.ECODE_ENVIRON)
5267 if requested > free_mem:
5268 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
5269 " needed %s MiB, available %s MiB" %
5270 (node, reason, requested, free_mem),
5274 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5275 """Checks if nodes have enough free disk space in the all VGs.
5277 This function check if all given nodes have the needed amount of
5278 free disk. In case any node has less disk or we cannot get the
5279 information from the node, this function raise an OpPrereqError
5282 @type lu: C{LogicalUnit}
5283 @param lu: a logical unit from which we get configuration data
5284 @type nodenames: C{list}
5285 @param nodenames: the list of node names to check
5286 @type req_sizes: C{dict}
5287 @param req_sizes: the hash of vg and corresponding amount of disk in
5289 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5290 or we cannot check the node
5293 for vg, req_size in req_sizes.items():
5294 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5297 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5298 """Checks if nodes have enough free disk space in the specified VG.
5300 This function check if all given nodes have the needed amount of
5301 free disk. In case any node has less disk or we cannot get the
5302 information from the node, this function raise an OpPrereqError
5305 @type lu: C{LogicalUnit}
5306 @param lu: a logical unit from which we get configuration data
5307 @type nodenames: C{list}
5308 @param nodenames: the list of node names to check
5310 @param vg: the volume group to check
5311 @type requested: C{int}
5312 @param requested: the amount of disk in MiB to check for
5313 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5314 or we cannot check the node
5317 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
5318 for node in nodenames:
5319 info = nodeinfo[node]
5320 info.Raise("Cannot get current information from node %s" % node,
5321 prereq=True, ecode=errors.ECODE_ENVIRON)
5322 vg_free = info.payload.get("vg_free", None)
5323 if not isinstance(vg_free, int):
5324 raise errors.OpPrereqError("Can't compute free disk space on node"
5325 " %s for vg %s, result was '%s'" %
5326 (node, vg, vg_free), errors.ECODE_ENVIRON)
5327 if requested > vg_free:
5328 raise errors.OpPrereqError("Not enough disk space on target node %s"
5329 " vg %s: required %d MiB, available %d MiB" %
5330 (node, vg, requested, vg_free),
5334 class LUInstanceStartup(LogicalUnit):
5335 """Starts an instance.
5338 HPATH = "instance-start"
5339 HTYPE = constants.HTYPE_INSTANCE
5342 def CheckArguments(self):
5344 if self.op.beparams:
5345 # fill the beparams dict
5346 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5348 def ExpandNames(self):
5349 self._ExpandAndLockInstance()
5351 def BuildHooksEnv(self):
5354 This runs on master, primary and secondary nodes of the instance.
5358 "FORCE": self.op.force,
5361 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5365 def BuildHooksNodes(self):
5366 """Build hooks nodes.
5369 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5372 def CheckPrereq(self):
5373 """Check prerequisites.
5375 This checks that the instance is in the cluster.
5378 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5379 assert self.instance is not None, \
5380 "Cannot retrieve locked instance %s" % self.op.instance_name
5383 if self.op.hvparams:
5384 # check hypervisor parameter syntax (locally)
5385 cluster = self.cfg.GetClusterInfo()
5386 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5387 filled_hvp = cluster.FillHV(instance)
5388 filled_hvp.update(self.op.hvparams)
5389 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5390 hv_type.CheckParameterSyntax(filled_hvp)
5391 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5393 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5395 if self.primary_offline and self.op.ignore_offline_nodes:
5396 self.proc.LogWarning("Ignoring offline primary node")
5398 if self.op.hvparams or self.op.beparams:
5399 self.proc.LogWarning("Overridden parameters are ignored")
5401 _CheckNodeOnline(self, instance.primary_node)
5403 bep = self.cfg.GetClusterInfo().FillBE(instance)
5405 # check bridges existence
5406 _CheckInstanceBridgesExist(self, instance)
5408 remote_info = self.rpc.call_instance_info(instance.primary_node,
5410 instance.hypervisor)
5411 remote_info.Raise("Error checking node %s" % instance.primary_node,
5412 prereq=True, ecode=errors.ECODE_ENVIRON)
5413 if not remote_info.payload: # not running already
5414 _CheckNodeFreeMemory(self, instance.primary_node,
5415 "starting instance %s" % instance.name,
5416 bep[constants.BE_MEMORY], instance.hypervisor)
5418 def Exec(self, feedback_fn):
5419 """Start the instance.
5422 instance = self.instance
5423 force = self.op.force
5425 self.cfg.MarkInstanceUp(instance.name)
5427 if self.primary_offline:
5428 assert self.op.ignore_offline_nodes
5429 self.proc.LogInfo("Primary node offline, marked instance as started")
5431 node_current = instance.primary_node
5433 _StartInstanceDisks(self, instance, force)
5435 result = self.rpc.call_instance_start(node_current, instance,
5436 self.op.hvparams, self.op.beparams)
5437 msg = result.fail_msg
5439 _ShutdownInstanceDisks(self, instance)
5440 raise errors.OpExecError("Could not start instance: %s" % msg)
5443 class LUInstanceReboot(LogicalUnit):
5444 """Reboot an instance.
5447 HPATH = "instance-reboot"
5448 HTYPE = constants.HTYPE_INSTANCE
5451 def ExpandNames(self):
5452 self._ExpandAndLockInstance()
5454 def BuildHooksEnv(self):
5457 This runs on master, primary and secondary nodes of the instance.
5461 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
5462 "REBOOT_TYPE": self.op.reboot_type,
5463 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5466 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5470 def BuildHooksNodes(self):
5471 """Build hooks nodes.
5474 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5477 def CheckPrereq(self):
5478 """Check prerequisites.
5480 This checks that the instance is in the cluster.
5483 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5484 assert self.instance is not None, \
5485 "Cannot retrieve locked instance %s" % self.op.instance_name
5487 _CheckNodeOnline(self, instance.primary_node)
5489 # check bridges existence
5490 _CheckInstanceBridgesExist(self, instance)
5492 def Exec(self, feedback_fn):
5493 """Reboot the instance.
5496 instance = self.instance
5497 ignore_secondaries = self.op.ignore_secondaries
5498 reboot_type = self.op.reboot_type
5500 remote_info = self.rpc.call_instance_info(instance.primary_node,
5502 instance.hypervisor)
5503 remote_info.Raise("Error checking node %s" % instance.primary_node)
5504 instance_running = bool(remote_info.payload)
5506 node_current = instance.primary_node
5508 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
5509 constants.INSTANCE_REBOOT_HARD]:
5510 for disk in instance.disks:
5511 self.cfg.SetDiskID(disk, node_current)
5512 result = self.rpc.call_instance_reboot(node_current, instance,
5514 self.op.shutdown_timeout)
5515 result.Raise("Could not reboot instance")
5517 if instance_running:
5518 result = self.rpc.call_instance_shutdown(node_current, instance,
5519 self.op.shutdown_timeout)
5520 result.Raise("Could not shutdown instance for full reboot")
5521 _ShutdownInstanceDisks(self, instance)
5523 self.LogInfo("Instance %s was already stopped, starting now",
5525 _StartInstanceDisks(self, instance, ignore_secondaries)
5526 result = self.rpc.call_instance_start(node_current, instance, None, None)
5527 msg = result.fail_msg
5529 _ShutdownInstanceDisks(self, instance)
5530 raise errors.OpExecError("Could not start instance for"
5531 " full reboot: %s" % msg)
5533 self.cfg.MarkInstanceUp(instance.name)
5536 class LUInstanceShutdown(LogicalUnit):
5537 """Shutdown an instance.
5540 HPATH = "instance-stop"
5541 HTYPE = constants.HTYPE_INSTANCE
5544 def ExpandNames(self):
5545 self._ExpandAndLockInstance()
5547 def BuildHooksEnv(self):
5550 This runs on master, primary and secondary nodes of the instance.
5553 env = _BuildInstanceHookEnvByObject(self, self.instance)
5554 env["TIMEOUT"] = self.op.timeout
5557 def BuildHooksNodes(self):
5558 """Build hooks nodes.
5561 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5564 def CheckPrereq(self):
5565 """Check prerequisites.
5567 This checks that the instance is in the cluster.
5570 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5571 assert self.instance is not None, \
5572 "Cannot retrieve locked instance %s" % self.op.instance_name
5574 self.primary_offline = \
5575 self.cfg.GetNodeInfo(self.instance.primary_node).offline
5577 if self.primary_offline and self.op.ignore_offline_nodes:
5578 self.proc.LogWarning("Ignoring offline primary node")
5580 _CheckNodeOnline(self, self.instance.primary_node)
5582 def Exec(self, feedback_fn):
5583 """Shutdown the instance.
5586 instance = self.instance
5587 node_current = instance.primary_node
5588 timeout = self.op.timeout
5590 self.cfg.MarkInstanceDown(instance.name)
5592 if self.primary_offline:
5593 assert self.op.ignore_offline_nodes
5594 self.proc.LogInfo("Primary node offline, marked instance as stopped")
5596 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
5597 msg = result.fail_msg
5599 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
5601 _ShutdownInstanceDisks(self, instance)
5604 class LUInstanceReinstall(LogicalUnit):
5605 """Reinstall an instance.
5608 HPATH = "instance-reinstall"
5609 HTYPE = constants.HTYPE_INSTANCE
5612 def ExpandNames(self):
5613 self._ExpandAndLockInstance()
5615 def BuildHooksEnv(self):
5618 This runs on master, primary and secondary nodes of the instance.
5621 return _BuildInstanceHookEnvByObject(self, self.instance)
5623 def BuildHooksNodes(self):
5624 """Build hooks nodes.
5627 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5630 def CheckPrereq(self):
5631 """Check prerequisites.
5633 This checks that the instance is in the cluster and is not running.
5636 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5637 assert instance is not None, \
5638 "Cannot retrieve locked instance %s" % self.op.instance_name
5639 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
5640 " offline, cannot reinstall")
5641 for node in instance.secondary_nodes:
5642 _CheckNodeOnline(self, node, "Instance secondary node offline,"
5643 " cannot reinstall")
5645 if instance.disk_template == constants.DT_DISKLESS:
5646 raise errors.OpPrereqError("Instance '%s' has no disks" %
5647 self.op.instance_name,
5649 _CheckInstanceDown(self, instance, "cannot reinstall")
5651 if self.op.os_type is not None:
5653 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
5654 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
5655 instance_os = self.op.os_type
5657 instance_os = instance.os
5659 nodelist = list(instance.all_nodes)
5661 if self.op.osparams:
5662 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
5663 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
5664 self.os_inst = i_osdict # the new dict (without defaults)
5668 self.instance = instance
5670 def Exec(self, feedback_fn):
5671 """Reinstall the instance.
5674 inst = self.instance
5676 if self.op.os_type is not None:
5677 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
5678 inst.os = self.op.os_type
5679 # Write to configuration
5680 self.cfg.Update(inst, feedback_fn)
5682 _StartInstanceDisks(self, inst, None)
5684 feedback_fn("Running the instance OS create scripts...")
5685 # FIXME: pass debug option from opcode to backend
5686 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
5687 self.op.debug_level,
5688 osparams=self.os_inst)
5689 result.Raise("Could not install OS for instance %s on node %s" %
5690 (inst.name, inst.primary_node))
5692 _ShutdownInstanceDisks(self, inst)
5695 class LUInstanceRecreateDisks(LogicalUnit):
5696 """Recreate an instance's missing disks.
5699 HPATH = "instance-recreate-disks"
5700 HTYPE = constants.HTYPE_INSTANCE
5703 def ExpandNames(self):
5704 self._ExpandAndLockInstance()
5706 def BuildHooksEnv(self):
5709 This runs on master, primary and secondary nodes of the instance.
5712 return _BuildInstanceHookEnvByObject(self, self.instance)
5714 def BuildHooksNodes(self):
5715 """Build hooks nodes.
5718 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5721 def CheckPrereq(self):
5722 """Check prerequisites.
5724 This checks that the instance is in the cluster and is not running.
5727 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5728 assert instance is not None, \
5729 "Cannot retrieve locked instance %s" % self.op.instance_name
5730 _CheckNodeOnline(self, instance.primary_node)
5732 if instance.disk_template == constants.DT_DISKLESS:
5733 raise errors.OpPrereqError("Instance '%s' has no disks" %
5734 self.op.instance_name, errors.ECODE_INVAL)
5735 _CheckInstanceDown(self, instance, "cannot recreate disks")
5737 if not self.op.disks:
5738 self.op.disks = range(len(instance.disks))
5740 for idx in self.op.disks:
5741 if idx >= len(instance.disks):
5742 raise errors.OpPrereqError("Invalid disk index '%s'" % idx,
5745 self.instance = instance
5747 def Exec(self, feedback_fn):
5748 """Recreate the disks.
5752 for idx, _ in enumerate(self.instance.disks):
5753 if idx not in self.op.disks: # disk idx has not been passed in
5757 _CreateDisks(self, self.instance, to_skip=to_skip)
5760 class LUInstanceRename(LogicalUnit):
5761 """Rename an instance.
5764 HPATH = "instance-rename"
5765 HTYPE = constants.HTYPE_INSTANCE
5767 def CheckArguments(self):
5771 if self.op.ip_check and not self.op.name_check:
5772 # TODO: make the ip check more flexible and not depend on the name check
5773 raise errors.OpPrereqError("IP address check requires a name check",
5776 def BuildHooksEnv(self):
5779 This runs on master, primary and secondary nodes of the instance.
5782 env = _BuildInstanceHookEnvByObject(self, self.instance)
5783 env["INSTANCE_NEW_NAME"] = self.op.new_name
5786 def BuildHooksNodes(self):
5787 """Build hooks nodes.
5790 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5793 def CheckPrereq(self):
5794 """Check prerequisites.
5796 This checks that the instance is in the cluster and is not running.
5799 self.op.instance_name = _ExpandInstanceName(self.cfg,
5800 self.op.instance_name)
5801 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5802 assert instance is not None
5803 _CheckNodeOnline(self, instance.primary_node)
5804 _CheckInstanceDown(self, instance, "cannot rename")
5805 self.instance = instance
5807 new_name = self.op.new_name
5808 if self.op.name_check:
5809 hostname = netutils.GetHostname(name=new_name)
5810 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
5812 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
5813 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
5814 " same as given hostname '%s'") %
5815 (hostname.name, self.op.new_name),
5817 new_name = self.op.new_name = hostname.name
5818 if (self.op.ip_check and
5819 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
5820 raise errors.OpPrereqError("IP %s of instance %s already in use" %
5821 (hostname.ip, new_name),
5822 errors.ECODE_NOTUNIQUE)
5824 instance_list = self.cfg.GetInstanceList()
5825 if new_name in instance_list and new_name != instance.name:
5826 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5827 new_name, errors.ECODE_EXISTS)
5829 def Exec(self, feedback_fn):
5830 """Rename the instance.
5833 inst = self.instance
5834 old_name = inst.name
5836 rename_file_storage = False
5837 if (inst.disk_template in (constants.DT_FILE, constants.DT_SHARED_FILE) and
5838 self.op.new_name != inst.name):
5839 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5840 rename_file_storage = True
5842 self.cfg.RenameInstance(inst.name, self.op.new_name)
5843 # Change the instance lock. This is definitely safe while we hold the BGL.
5844 # Otherwise the new lock would have to be added in acquired mode.
5846 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5847 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5849 # re-read the instance from the configuration after rename
5850 inst = self.cfg.GetInstanceInfo(self.op.new_name)
5852 if rename_file_storage:
5853 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5854 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5855 old_file_storage_dir,
5856 new_file_storage_dir)
5857 result.Raise("Could not rename on node %s directory '%s' to '%s'"
5858 " (but the instance has been renamed in Ganeti)" %
5859 (inst.primary_node, old_file_storage_dir,
5860 new_file_storage_dir))
5862 _StartInstanceDisks(self, inst, None)
5864 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5865 old_name, self.op.debug_level)
5866 msg = result.fail_msg
5868 msg = ("Could not run OS rename script for instance %s on node %s"
5869 " (but the instance has been renamed in Ganeti): %s" %
5870 (inst.name, inst.primary_node, msg))
5871 self.proc.LogWarning(msg)
5873 _ShutdownInstanceDisks(self, inst)
5878 class LUInstanceRemove(LogicalUnit):
5879 """Remove an instance.
5882 HPATH = "instance-remove"
5883 HTYPE = constants.HTYPE_INSTANCE
5886 def ExpandNames(self):
5887 self._ExpandAndLockInstance()
5888 self.needed_locks[locking.LEVEL_NODE] = []
5889 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5891 def DeclareLocks(self, level):
5892 if level == locking.LEVEL_NODE:
5893 self._LockInstancesNodes()
5895 def BuildHooksEnv(self):
5898 This runs on master, primary and secondary nodes of the instance.
5901 env = _BuildInstanceHookEnvByObject(self, self.instance)
5902 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5905 def BuildHooksNodes(self):
5906 """Build hooks nodes.
5909 nl = [self.cfg.GetMasterNode()]
5910 nl_post = list(self.instance.all_nodes) + nl
5911 return (nl, nl_post)
5913 def CheckPrereq(self):
5914 """Check prerequisites.
5916 This checks that the instance is in the cluster.
5919 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5920 assert self.instance is not None, \
5921 "Cannot retrieve locked instance %s" % self.op.instance_name
5923 def Exec(self, feedback_fn):
5924 """Remove the instance.
5927 instance = self.instance
5928 logging.info("Shutting down instance %s on node %s",
5929 instance.name, instance.primary_node)
5931 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5932 self.op.shutdown_timeout)
5933 msg = result.fail_msg
5935 if self.op.ignore_failures:
5936 feedback_fn("Warning: can't shutdown instance: %s" % msg)
5938 raise errors.OpExecError("Could not shutdown instance %s on"
5940 (instance.name, instance.primary_node, msg))
5942 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5945 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5946 """Utility function to remove an instance.
5949 logging.info("Removing block devices for instance %s", instance.name)
5951 if not _RemoveDisks(lu, instance):
5952 if not ignore_failures:
5953 raise errors.OpExecError("Can't remove instance's disks")
5954 feedback_fn("Warning: can't remove instance's disks")
5956 logging.info("Removing instance %s out of cluster config", instance.name)
5958 lu.cfg.RemoveInstance(instance.name)
5960 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5961 "Instance lock removal conflict"
5963 # Remove lock for the instance
5964 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5967 class LUInstanceQuery(NoHooksLU):
5968 """Logical unit for querying instances.
5971 # pylint: disable-msg=W0142
5974 def CheckArguments(self):
5975 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
5976 self.op.output_fields, self.op.use_locking)
5978 def ExpandNames(self):
5979 self.iq.ExpandNames(self)
5981 def DeclareLocks(self, level):
5982 self.iq.DeclareLocks(self, level)
5984 def Exec(self, feedback_fn):
5985 return self.iq.OldStyleQuery(self)
5988 class LUInstanceFailover(LogicalUnit):
5989 """Failover an instance.
5992 HPATH = "instance-failover"
5993 HTYPE = constants.HTYPE_INSTANCE
5996 def CheckArguments(self):
5997 """Check the arguments.
6000 self.iallocator = getattr(self.op, "iallocator", None)
6001 self.target_node = getattr(self.op, "target_node", None)
6003 def ExpandNames(self):
6004 self._ExpandAndLockInstance()
6006 if self.op.target_node is not None:
6007 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6009 self.needed_locks[locking.LEVEL_NODE] = []
6010 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6012 ignore_consistency = self.op.ignore_consistency
6013 shutdown_timeout = self.op.shutdown_timeout
6014 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6016 iallocator=self.op.iallocator,
6017 target_node=self.op.target_node,
6019 ignore_consistency=ignore_consistency,
6020 shutdown_timeout=shutdown_timeout)
6021 self.tasklets = [self._migrater]
6023 def DeclareLocks(self, level):
6024 if level == locking.LEVEL_NODE:
6025 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6026 if instance.disk_template in constants.DTS_EXT_MIRROR:
6027 if self.op.target_node is None:
6028 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6030 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6031 self.op.target_node]
6032 del self.recalculate_locks[locking.LEVEL_NODE]
6034 self._LockInstancesNodes()
6036 def BuildHooksEnv(self):
6039 This runs on master, primary and secondary nodes of the instance.
6042 instance = self._migrater.instance
6043 source_node = instance.primary_node
6044 target_node = self._migrater.target_node
6046 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
6047 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6048 "OLD_PRIMARY": source_node,
6049 "NEW_PRIMARY": target_node,
6052 if instance.disk_template in constants.DTS_INT_MIRROR:
6053 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
6054 env["NEW_SECONDARY"] = source_node
6056 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
6058 env.update(_BuildInstanceHookEnvByObject(self, instance))
6062 def BuildHooksNodes(self):
6063 """Build hooks nodes.
6066 instance = self._migrater.instance
6067 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6068 return (nl, nl + [instance.primary_node])
6071 class LUInstanceMigrate(LogicalUnit):
6072 """Migrate an instance.
6074 This is migration without shutting down, compared to the failover,
6075 which is done with shutdown.
6078 HPATH = "instance-migrate"
6079 HTYPE = constants.HTYPE_INSTANCE
6082 def ExpandNames(self):
6083 self._ExpandAndLockInstance()
6085 if self.op.target_node is not None:
6086 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6088 self.needed_locks[locking.LEVEL_NODE] = []
6089 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6091 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6092 cleanup=self.op.cleanup,
6094 fallback=self.op.allow_failover)
6095 self.tasklets = [self._migrater]
6097 def DeclareLocks(self, level):
6098 if level == locking.LEVEL_NODE:
6099 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6100 if instance.disk_template in constants.DTS_EXT_MIRROR:
6101 if self.op.target_node is None:
6102 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6104 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6105 self.op.target_node]
6106 del self.recalculate_locks[locking.LEVEL_NODE]
6108 self._LockInstancesNodes()
6110 def BuildHooksEnv(self):
6113 This runs on master, primary and secondary nodes of the instance.
6116 instance = self._migrater.instance
6117 source_node = instance.primary_node
6118 target_node = self._migrater.target_node
6119 env = _BuildInstanceHookEnvByObject(self, instance)
6121 "MIGRATE_LIVE": self._migrater.live,
6122 "MIGRATE_CLEANUP": self.op.cleanup,
6123 "OLD_PRIMARY": source_node,
6124 "NEW_PRIMARY": target_node,
6127 if instance.disk_template in constants.DTS_INT_MIRROR:
6128 env["OLD_SECONDARY"] = target_node
6129 env["NEW_SECONDARY"] = source_node
6131 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
6135 def BuildHooksNodes(self):
6136 """Build hooks nodes.
6139 instance = self._migrater.instance
6140 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6141 return (nl, nl + [instance.primary_node])
6144 class LUInstanceMove(LogicalUnit):
6145 """Move an instance by data-copying.
6148 HPATH = "instance-move"
6149 HTYPE = constants.HTYPE_INSTANCE
6152 def ExpandNames(self):
6153 self._ExpandAndLockInstance()
6154 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6155 self.op.target_node = target_node
6156 self.needed_locks[locking.LEVEL_NODE] = [target_node]
6157 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6159 def DeclareLocks(self, level):
6160 if level == locking.LEVEL_NODE:
6161 self._LockInstancesNodes(primary_only=True)
6163 def BuildHooksEnv(self):
6166 This runs on master, primary and secondary nodes of the instance.
6170 "TARGET_NODE": self.op.target_node,
6171 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6173 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6176 def BuildHooksNodes(self):
6177 """Build hooks nodes.
6181 self.cfg.GetMasterNode(),
6182 self.instance.primary_node,
6183 self.op.target_node,
6187 def CheckPrereq(self):
6188 """Check prerequisites.
6190 This checks that the instance is in the cluster.
6193 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6194 assert self.instance is not None, \
6195 "Cannot retrieve locked instance %s" % self.op.instance_name
6197 node = self.cfg.GetNodeInfo(self.op.target_node)
6198 assert node is not None, \
6199 "Cannot retrieve locked node %s" % self.op.target_node
6201 self.target_node = target_node = node.name
6203 if target_node == instance.primary_node:
6204 raise errors.OpPrereqError("Instance %s is already on the node %s" %
6205 (instance.name, target_node),
6208 bep = self.cfg.GetClusterInfo().FillBE(instance)
6210 for idx, dsk in enumerate(instance.disks):
6211 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
6212 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
6213 " cannot copy" % idx, errors.ECODE_STATE)
6215 _CheckNodeOnline(self, target_node)
6216 _CheckNodeNotDrained(self, target_node)
6217 _CheckNodeVmCapable(self, target_node)
6219 if instance.admin_up:
6220 # check memory requirements on the secondary node
6221 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
6222 instance.name, bep[constants.BE_MEMORY],
6223 instance.hypervisor)
6225 self.LogInfo("Not checking memory on the secondary node as"
6226 " instance will not be started")
6228 # check bridge existance
6229 _CheckInstanceBridgesExist(self, instance, node=target_node)
6231 def Exec(self, feedback_fn):
6232 """Move an instance.
6234 The move is done by shutting it down on its present node, copying
6235 the data over (slow) and starting it on the new node.
6238 instance = self.instance
6240 source_node = instance.primary_node
6241 target_node = self.target_node
6243 self.LogInfo("Shutting down instance %s on source node %s",
6244 instance.name, source_node)
6246 result = self.rpc.call_instance_shutdown(source_node, instance,
6247 self.op.shutdown_timeout)
6248 msg = result.fail_msg
6250 if self.op.ignore_consistency:
6251 self.proc.LogWarning("Could not shutdown instance %s on node %s."
6252 " Proceeding anyway. Please make sure node"
6253 " %s is down. Error details: %s",
6254 instance.name, source_node, source_node, msg)
6256 raise errors.OpExecError("Could not shutdown instance %s on"
6258 (instance.name, source_node, msg))
6260 # create the target disks
6262 _CreateDisks(self, instance, target_node=target_node)
6263 except errors.OpExecError:
6264 self.LogWarning("Device creation failed, reverting...")
6266 _RemoveDisks(self, instance, target_node=target_node)
6268 self.cfg.ReleaseDRBDMinors(instance.name)
6271 cluster_name = self.cfg.GetClusterInfo().cluster_name
6274 # activate, get path, copy the data over
6275 for idx, disk in enumerate(instance.disks):
6276 self.LogInfo("Copying data for disk %d", idx)
6277 result = self.rpc.call_blockdev_assemble(target_node, disk,
6278 instance.name, True, idx)
6280 self.LogWarning("Can't assemble newly created disk %d: %s",
6281 idx, result.fail_msg)
6282 errs.append(result.fail_msg)
6284 dev_path = result.payload
6285 result = self.rpc.call_blockdev_export(source_node, disk,
6286 target_node, dev_path,
6289 self.LogWarning("Can't copy data over for disk %d: %s",
6290 idx, result.fail_msg)
6291 errs.append(result.fail_msg)
6295 self.LogWarning("Some disks failed to copy, aborting")
6297 _RemoveDisks(self, instance, target_node=target_node)
6299 self.cfg.ReleaseDRBDMinors(instance.name)
6300 raise errors.OpExecError("Errors during disk copy: %s" %
6303 instance.primary_node = target_node
6304 self.cfg.Update(instance, feedback_fn)
6306 self.LogInfo("Removing the disks on the original node")
6307 _RemoveDisks(self, instance, target_node=source_node)
6309 # Only start the instance if it's marked as up
6310 if instance.admin_up:
6311 self.LogInfo("Starting instance %s on node %s",
6312 instance.name, target_node)
6314 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6315 ignore_secondaries=True)
6317 _ShutdownInstanceDisks(self, instance)
6318 raise errors.OpExecError("Can't activate the instance's disks")
6320 result = self.rpc.call_instance_start(target_node, instance, None, None)
6321 msg = result.fail_msg
6323 _ShutdownInstanceDisks(self, instance)
6324 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6325 (instance.name, target_node, msg))
6328 class LUNodeMigrate(LogicalUnit):
6329 """Migrate all instances from a node.
6332 HPATH = "node-migrate"
6333 HTYPE = constants.HTYPE_NODE
6336 def CheckArguments(self):
6337 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
6339 def ExpandNames(self):
6340 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6342 self.needed_locks = {}
6344 # Create tasklets for migrating instances for all instances on this node
6348 self.lock_all_nodes = False
6350 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
6351 logging.debug("Migrating instance %s", inst.name)
6352 names.append(inst.name)
6354 tasklets.append(TLMigrateInstance(self, inst.name, cleanup=False))
6356 if inst.disk_template in constants.DTS_EXT_MIRROR:
6357 # We need to lock all nodes, as the iallocator will choose the
6358 # destination nodes afterwards
6359 self.lock_all_nodes = True
6361 self.tasklets = tasklets
6363 # Declare node locks
6364 if self.lock_all_nodes:
6365 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6367 self.needed_locks[locking.LEVEL_NODE] = [self.op.node_name]
6368 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6370 # Declare instance locks
6371 self.needed_locks[locking.LEVEL_INSTANCE] = names
6373 def DeclareLocks(self, level):
6374 if level == locking.LEVEL_NODE and not self.lock_all_nodes:
6375 self._LockInstancesNodes()
6377 def BuildHooksEnv(self):
6380 This runs on the master, the primary and all the secondaries.
6384 "NODE_NAME": self.op.node_name,
6387 def BuildHooksNodes(self):
6388 """Build hooks nodes.
6391 nl = [self.cfg.GetMasterNode()]
6395 class TLMigrateInstance(Tasklet):
6396 """Tasklet class for instance migration.
6399 @ivar live: whether the migration will be done live or non-live;
6400 this variable is initalized only after CheckPrereq has run
6401 @type cleanup: boolean
6402 @ivar cleanup: Wheater we cleanup from a failed migration
6403 @type iallocator: string
6404 @ivar iallocator: The iallocator used to determine target_node
6405 @type target_node: string
6406 @ivar target_node: If given, the target_node to reallocate the instance to
6407 @type failover: boolean
6408 @ivar failover: Whether operation results in failover or migration
6409 @type fallback: boolean
6410 @ivar fallback: Whether fallback to failover is allowed if migration not
6412 @type ignore_consistency: boolean
6413 @ivar ignore_consistency: Wheter we should ignore consistency between source
6415 @type shutdown_timeout: int
6416 @ivar shutdown_timeout: In case of failover timeout of the shutdown
6419 def __init__(self, lu, instance_name, cleanup=False,
6420 failover=False, fallback=False,
6421 ignore_consistency=False,
6422 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT):
6423 """Initializes this class.
6426 Tasklet.__init__(self, lu)
6429 self.instance_name = instance_name
6430 self.cleanup = cleanup
6431 self.live = False # will be overridden later
6432 self.failover = failover
6433 self.fallback = fallback
6434 self.ignore_consistency = ignore_consistency
6435 self.shutdown_timeout = shutdown_timeout
6437 def CheckPrereq(self):
6438 """Check prerequisites.
6440 This checks that the instance is in the cluster.
6443 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
6444 instance = self.cfg.GetInstanceInfo(instance_name)
6445 assert instance is not None
6446 self.instance = instance
6448 if (not self.cleanup and not instance.admin_up and not self.failover and
6450 self.lu.LogInfo("Instance is marked down, fallback allowed, switching"
6452 self.failover = True
6454 if instance.disk_template not in constants.DTS_MIRRORED:
6459 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
6460 " %s" % (instance.disk_template, text),
6463 if instance.disk_template in constants.DTS_EXT_MIRROR:
6464 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
6466 if self.lu.op.iallocator:
6467 self._RunAllocator()
6469 # We set set self.target_node as it is required by
6471 self.target_node = self.lu.op.target_node
6473 # self.target_node is already populated, either directly or by the
6475 target_node = self.target_node
6477 if len(self.lu.tasklets) == 1:
6478 # It is safe to release locks only when we're the only tasklet in the LU
6479 _ReleaseLocks(self, locking.LEVEL_NODE,
6480 keep=[instance.primary_node, self.target_node])
6483 secondary_nodes = instance.secondary_nodes
6484 if not secondary_nodes:
6485 raise errors.ConfigurationError("No secondary node but using"
6486 " %s disk template" %
6487 instance.disk_template)
6488 target_node = secondary_nodes[0]
6489 if self.lu.op.iallocator or (self.target_node and
6490 self.target_node != target_node):
6492 text = "failed over"
6495 raise errors.OpPrereqError("Instances with disk template %s cannot"
6496 " be %s to arbitrary nodes"
6497 " (neither an iallocator nor a target"
6498 " node can be passed)" %
6499 (instance.disk_template, text),
6502 i_be = self.cfg.GetClusterInfo().FillBE(instance)
6504 # check memory requirements on the secondary node
6505 if not self.failover or instance.admin_up:
6506 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
6507 instance.name, i_be[constants.BE_MEMORY],
6508 instance.hypervisor)
6510 self.lu.LogInfo("Not checking memory on the secondary node as"
6511 " instance will not be started")
6513 # check bridge existance
6514 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
6516 if not self.cleanup:
6517 _CheckNodeNotDrained(self.lu, target_node)
6518 if not self.failover:
6519 result = self.rpc.call_instance_migratable(instance.primary_node,
6521 if result.fail_msg and self.fallback:
6522 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
6524 self.failover = True
6526 result.Raise("Can't migrate, please use failover",
6527 prereq=True, ecode=errors.ECODE_STATE)
6529 assert not (self.failover and self.cleanup)
6531 if not self.failover:
6532 if self.lu.op.live is not None and self.lu.op.mode is not None:
6533 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
6534 " parameters are accepted",
6536 if self.lu.op.live is not None:
6538 self.lu.op.mode = constants.HT_MIGRATION_LIVE
6540 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
6541 # reset the 'live' parameter to None so that repeated
6542 # invocations of CheckPrereq do not raise an exception
6543 self.lu.op.live = None
6544 elif self.lu.op.mode is None:
6545 # read the default value from the hypervisor
6546 i_hv = self.cfg.GetClusterInfo().FillHV(self.instance,
6548 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
6550 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
6552 # Failover is never live
6555 def _RunAllocator(self):
6556 """Run the allocator based on input opcode.
6559 ial = IAllocator(self.cfg, self.rpc,
6560 mode=constants.IALLOCATOR_MODE_RELOC,
6561 name=self.instance_name,
6562 # TODO See why hail breaks with a single node below
6563 relocate_from=[self.instance.primary_node,
6564 self.instance.primary_node],
6567 ial.Run(self.lu.op.iallocator)
6570 raise errors.OpPrereqError("Can't compute nodes using"
6571 " iallocator '%s': %s" %
6572 (self.lu.op.iallocator, ial.info),
6574 if len(ial.result) != ial.required_nodes:
6575 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6576 " of nodes (%s), required %s" %
6577 (self.lu.op.iallocator, len(ial.result),
6578 ial.required_nodes), errors.ECODE_FAULT)
6579 self.target_node = ial.result[0]
6580 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6581 self.instance_name, self.lu.op.iallocator,
6582 utils.CommaJoin(ial.result))
6584 def _WaitUntilSync(self):
6585 """Poll with custom rpc for disk sync.
6587 This uses our own step-based rpc call.
6590 self.feedback_fn("* wait until resync is done")
6594 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
6596 self.instance.disks)
6598 for node, nres in result.items():
6599 nres.Raise("Cannot resync disks on node %s" % node)
6600 node_done, node_percent = nres.payload
6601 all_done = all_done and node_done
6602 if node_percent is not None:
6603 min_percent = min(min_percent, node_percent)
6605 if min_percent < 100:
6606 self.feedback_fn(" - progress: %.1f%%" % min_percent)
6609 def _EnsureSecondary(self, node):
6610 """Demote a node to secondary.
6613 self.feedback_fn("* switching node %s to secondary mode" % node)
6615 for dev in self.instance.disks:
6616 self.cfg.SetDiskID(dev, node)
6618 result = self.rpc.call_blockdev_close(node, self.instance.name,
6619 self.instance.disks)
6620 result.Raise("Cannot change disk to secondary on node %s" % node)
6622 def _GoStandalone(self):
6623 """Disconnect from the network.
6626 self.feedback_fn("* changing into standalone mode")
6627 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
6628 self.instance.disks)
6629 for node, nres in result.items():
6630 nres.Raise("Cannot disconnect disks node %s" % node)
6632 def _GoReconnect(self, multimaster):
6633 """Reconnect to the network.
6639 msg = "single-master"
6640 self.feedback_fn("* changing disks into %s mode" % msg)
6641 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
6642 self.instance.disks,
6643 self.instance.name, multimaster)
6644 for node, nres in result.items():
6645 nres.Raise("Cannot change disks config on node %s" % node)
6647 def _ExecCleanup(self):
6648 """Try to cleanup after a failed migration.
6650 The cleanup is done by:
6651 - check that the instance is running only on one node
6652 (and update the config if needed)
6653 - change disks on its secondary node to secondary
6654 - wait until disks are fully synchronized
6655 - disconnect from the network
6656 - change disks into single-master mode
6657 - wait again until disks are fully synchronized
6660 instance = self.instance
6661 target_node = self.target_node
6662 source_node = self.source_node
6664 # check running on only one node
6665 self.feedback_fn("* checking where the instance actually runs"
6666 " (if this hangs, the hypervisor might be in"
6668 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6669 for node, result in ins_l.items():
6670 result.Raise("Can't contact node %s" % node)
6672 runningon_source = instance.name in ins_l[source_node].payload
6673 runningon_target = instance.name in ins_l[target_node].payload
6675 if runningon_source and runningon_target:
6676 raise errors.OpExecError("Instance seems to be running on two nodes,"
6677 " or the hypervisor is confused; you will have"
6678 " to ensure manually that it runs only on one"
6679 " and restart this operation")
6681 if not (runningon_source or runningon_target):
6682 raise errors.OpExecError("Instance does not seem to be running at all;"
6683 " in this case it's safer to repair by"
6684 " running 'gnt-instance stop' to ensure disk"
6685 " shutdown, and then restarting it")
6687 if runningon_target:
6688 # the migration has actually succeeded, we need to update the config
6689 self.feedback_fn("* instance running on secondary node (%s),"
6690 " updating config" % target_node)
6691 instance.primary_node = target_node
6692 self.cfg.Update(instance, self.feedback_fn)
6693 demoted_node = source_node
6695 self.feedback_fn("* instance confirmed to be running on its"
6696 " primary node (%s)" % source_node)
6697 demoted_node = target_node
6699 if instance.disk_template in constants.DTS_INT_MIRROR:
6700 self._EnsureSecondary(demoted_node)
6702 self._WaitUntilSync()
6703 except errors.OpExecError:
6704 # we ignore here errors, since if the device is standalone, it
6705 # won't be able to sync
6707 self._GoStandalone()
6708 self._GoReconnect(False)
6709 self._WaitUntilSync()
6711 self.feedback_fn("* done")
6713 def _RevertDiskStatus(self):
6714 """Try to revert the disk status after a failed migration.
6717 target_node = self.target_node
6718 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
6722 self._EnsureSecondary(target_node)
6723 self._GoStandalone()
6724 self._GoReconnect(False)
6725 self._WaitUntilSync()
6726 except errors.OpExecError, err:
6727 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
6728 " please try to recover the instance manually;"
6729 " error '%s'" % str(err))
6731 def _AbortMigration(self):
6732 """Call the hypervisor code to abort a started migration.
6735 instance = self.instance
6736 target_node = self.target_node
6737 migration_info = self.migration_info
6739 abort_result = self.rpc.call_finalize_migration(target_node,
6743 abort_msg = abort_result.fail_msg
6745 logging.error("Aborting migration failed on target node %s: %s",
6746 target_node, abort_msg)
6747 # Don't raise an exception here, as we stil have to try to revert the
6748 # disk status, even if this step failed.
6750 def _ExecMigration(self):
6751 """Migrate an instance.
6753 The migrate is done by:
6754 - change the disks into dual-master mode
6755 - wait until disks are fully synchronized again
6756 - migrate the instance
6757 - change disks on the new secondary node (the old primary) to secondary
6758 - wait until disks are fully synchronized
6759 - change disks into single-master mode
6762 instance = self.instance
6763 target_node = self.target_node
6764 source_node = self.source_node
6766 self.feedback_fn("* checking disk consistency between source and target")
6767 for dev in instance.disks:
6768 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6769 raise errors.OpExecError("Disk %s is degraded or not fully"
6770 " synchronized on target node,"
6771 " aborting migration" % dev.iv_name)
6773 # First get the migration information from the remote node
6774 result = self.rpc.call_migration_info(source_node, instance)
6775 msg = result.fail_msg
6777 log_err = ("Failed fetching source migration information from %s: %s" %
6779 logging.error(log_err)
6780 raise errors.OpExecError(log_err)
6782 self.migration_info = migration_info = result.payload
6784 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
6785 # Then switch the disks to master/master mode
6786 self._EnsureSecondary(target_node)
6787 self._GoStandalone()
6788 self._GoReconnect(True)
6789 self._WaitUntilSync()
6791 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6792 result = self.rpc.call_accept_instance(target_node,
6795 self.nodes_ip[target_node])
6797 msg = result.fail_msg
6799 logging.error("Instance pre-migration failed, trying to revert"
6800 " disk status: %s", msg)
6801 self.feedback_fn("Pre-migration failed, aborting")
6802 self._AbortMigration()
6803 self._RevertDiskStatus()
6804 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6805 (instance.name, msg))
6807 self.feedback_fn("* migrating instance to %s" % target_node)
6808 result = self.rpc.call_instance_migrate(source_node, instance,
6809 self.nodes_ip[target_node],
6811 msg = result.fail_msg
6813 logging.error("Instance migration failed, trying to revert"
6814 " disk status: %s", msg)
6815 self.feedback_fn("Migration failed, aborting")
6816 self._AbortMigration()
6817 self._RevertDiskStatus()
6818 raise errors.OpExecError("Could not migrate instance %s: %s" %
6819 (instance.name, msg))
6821 instance.primary_node = target_node
6822 # distribute new instance config to the other nodes
6823 self.cfg.Update(instance, self.feedback_fn)
6825 result = self.rpc.call_finalize_migration(target_node,
6829 msg = result.fail_msg
6831 logging.error("Instance migration succeeded, but finalization failed:"
6833 raise errors.OpExecError("Could not finalize instance migration: %s" %
6836 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
6837 self._EnsureSecondary(source_node)
6838 self._WaitUntilSync()
6839 self._GoStandalone()
6840 self._GoReconnect(False)
6841 self._WaitUntilSync()
6843 self.feedback_fn("* done")
6845 def _ExecFailover(self):
6846 """Failover an instance.
6848 The failover is done by shutting it down on its present node and
6849 starting it on the secondary.
6852 instance = self.instance
6853 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
6855 source_node = instance.primary_node
6856 target_node = self.target_node
6858 if instance.admin_up:
6859 self.feedback_fn("* checking disk consistency between source and target")
6860 for dev in instance.disks:
6861 # for drbd, these are drbd over lvm
6862 if not _CheckDiskConsistency(self, dev, target_node, False):
6863 if not self.ignore_consistency:
6864 raise errors.OpExecError("Disk %s is degraded on target node,"
6865 " aborting failover" % dev.iv_name)
6867 self.feedback_fn("* not checking disk consistency as instance is not"
6870 self.feedback_fn("* shutting down instance on source node")
6871 logging.info("Shutting down instance %s on node %s",
6872 instance.name, source_node)
6874 result = self.rpc.call_instance_shutdown(source_node, instance,
6875 self.shutdown_timeout)
6876 msg = result.fail_msg
6878 if self.ignore_consistency or primary_node.offline:
6879 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
6880 " proceeding anyway; please make sure node"
6881 " %s is down; error details: %s",
6882 instance.name, source_node, source_node, msg)
6884 raise errors.OpExecError("Could not shutdown instance %s on"
6886 (instance.name, source_node, msg))
6888 self.feedback_fn("* deactivating the instance's disks on source node")
6889 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
6890 raise errors.OpExecError("Can't shut down the instance's disks.")
6892 instance.primary_node = target_node
6893 # distribute new instance config to the other nodes
6894 self.cfg.Update(instance, self.feedback_fn)
6896 # Only start the instance if it's marked as up
6897 if instance.admin_up:
6898 self.feedback_fn("* activating the instance's disks on target node")
6899 logging.info("Starting instance %s on node %s",
6900 instance.name, target_node)
6902 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6903 ignore_secondaries=True)
6905 _ShutdownInstanceDisks(self, instance)
6906 raise errors.OpExecError("Can't activate the instance's disks")
6908 self.feedback_fn("* starting the instance on the target node")
6909 result = self.rpc.call_instance_start(target_node, instance, None, None)
6910 msg = result.fail_msg
6912 _ShutdownInstanceDisks(self, instance)
6913 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6914 (instance.name, target_node, msg))
6916 def Exec(self, feedback_fn):
6917 """Perform the migration.
6920 self.feedback_fn = feedback_fn
6921 self.source_node = self.instance.primary_node
6923 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
6924 if self.instance.disk_template in constants.DTS_INT_MIRROR:
6925 self.target_node = self.instance.secondary_nodes[0]
6926 # Otherwise self.target_node has been populated either
6927 # directly, or through an iallocator.
6929 self.all_nodes = [self.source_node, self.target_node]
6931 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6932 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6936 feedback_fn("Failover instance %s" % self.instance.name)
6937 self._ExecFailover()
6939 feedback_fn("Migrating instance %s" % self.instance.name)
6942 return self._ExecCleanup()
6944 return self._ExecMigration()
6947 def _CreateBlockDev(lu, node, instance, device, force_create,
6949 """Create a tree of block devices on a given node.
6951 If this device type has to be created on secondaries, create it and
6954 If not, just recurse to children keeping the same 'force' value.
6956 @param lu: the lu on whose behalf we execute
6957 @param node: the node on which to create the device
6958 @type instance: L{objects.Instance}
6959 @param instance: the instance which owns the device
6960 @type device: L{objects.Disk}
6961 @param device: the device to create
6962 @type force_create: boolean
6963 @param force_create: whether to force creation of this device; this
6964 will be change to True whenever we find a device which has
6965 CreateOnSecondary() attribute
6966 @param info: the extra 'metadata' we should attach to the device
6967 (this will be represented as a LVM tag)
6968 @type force_open: boolean
6969 @param force_open: this parameter will be passes to the
6970 L{backend.BlockdevCreate} function where it specifies
6971 whether we run on primary or not, and it affects both
6972 the child assembly and the device own Open() execution
6975 if device.CreateOnSecondary():
6979 for child in device.children:
6980 _CreateBlockDev(lu, node, instance, child, force_create,
6983 if not force_create:
6986 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6989 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6990 """Create a single block device on a given node.
6992 This will not recurse over children of the device, so they must be
6995 @param lu: the lu on whose behalf we execute
6996 @param node: the node on which to create the device
6997 @type instance: L{objects.Instance}
6998 @param instance: the instance which owns the device
6999 @type device: L{objects.Disk}
7000 @param device: the device to create
7001 @param info: the extra 'metadata' we should attach to the device
7002 (this will be represented as a LVM tag)
7003 @type force_open: boolean
7004 @param force_open: this parameter will be passes to the
7005 L{backend.BlockdevCreate} function where it specifies
7006 whether we run on primary or not, and it affects both
7007 the child assembly and the device own Open() execution
7010 lu.cfg.SetDiskID(device, node)
7011 result = lu.rpc.call_blockdev_create(node, device, device.size,
7012 instance.name, force_open, info)
7013 result.Raise("Can't create block device %s on"
7014 " node %s for instance %s" % (device, node, instance.name))
7015 if device.physical_id is None:
7016 device.physical_id = result.payload
7019 def _GenerateUniqueNames(lu, exts):
7020 """Generate a suitable LV name.
7022 This will generate a logical volume name for the given instance.
7027 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
7028 results.append("%s%s" % (new_id, val))
7032 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
7033 iv_name, p_minor, s_minor):
7034 """Generate a drbd8 device complete with its children.
7037 assert len(vgnames) == len(names) == 2
7038 port = lu.cfg.AllocatePort()
7039 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
7040 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
7041 logical_id=(vgnames[0], names[0]))
7042 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7043 logical_id=(vgnames[1], names[1]))
7044 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
7045 logical_id=(primary, secondary, port,
7048 children=[dev_data, dev_meta],
7053 def _GenerateDiskTemplate(lu, template_name,
7054 instance_name, primary_node,
7055 secondary_nodes, disk_info,
7056 file_storage_dir, file_driver,
7057 base_index, feedback_fn):
7058 """Generate the entire disk layout for a given template type.
7061 #TODO: compute space requirements
7063 vgname = lu.cfg.GetVGName()
7064 disk_count = len(disk_info)
7066 if template_name == constants.DT_DISKLESS:
7068 elif template_name == constants.DT_PLAIN:
7069 if len(secondary_nodes) != 0:
7070 raise errors.ProgrammerError("Wrong template configuration")
7072 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7073 for i in range(disk_count)])
7074 for idx, disk in enumerate(disk_info):
7075 disk_index = idx + base_index
7076 vg = disk.get(constants.IDISK_VG, vgname)
7077 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
7078 disk_dev = objects.Disk(dev_type=constants.LD_LV,
7079 size=disk[constants.IDISK_SIZE],
7080 logical_id=(vg, names[idx]),
7081 iv_name="disk/%d" % disk_index,
7082 mode=disk[constants.IDISK_MODE])
7083 disks.append(disk_dev)
7084 elif template_name == constants.DT_DRBD8:
7085 if len(secondary_nodes) != 1:
7086 raise errors.ProgrammerError("Wrong template configuration")
7087 remote_node = secondary_nodes[0]
7088 minors = lu.cfg.AllocateDRBDMinor(
7089 [primary_node, remote_node] * len(disk_info), instance_name)
7092 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7093 for i in range(disk_count)]):
7094 names.append(lv_prefix + "_data")
7095 names.append(lv_prefix + "_meta")
7096 for idx, disk in enumerate(disk_info):
7097 disk_index = idx + base_index
7098 data_vg = disk.get(constants.IDISK_VG, vgname)
7099 meta_vg = disk.get(constants.IDISK_METAVG, data_vg)
7100 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
7101 disk[constants.IDISK_SIZE],
7103 names[idx * 2:idx * 2 + 2],
7104 "disk/%d" % disk_index,
7105 minors[idx * 2], minors[idx * 2 + 1])
7106 disk_dev.mode = disk[constants.IDISK_MODE]
7107 disks.append(disk_dev)
7108 elif template_name == constants.DT_FILE:
7109 if len(secondary_nodes) != 0:
7110 raise errors.ProgrammerError("Wrong template configuration")
7112 opcodes.RequireFileStorage()
7114 for idx, disk in enumerate(disk_info):
7115 disk_index = idx + base_index
7116 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7117 size=disk[constants.IDISK_SIZE],
7118 iv_name="disk/%d" % disk_index,
7119 logical_id=(file_driver,
7120 "%s/disk%d" % (file_storage_dir,
7122 mode=disk[constants.IDISK_MODE])
7123 disks.append(disk_dev)
7124 elif template_name == constants.DT_SHARED_FILE:
7125 if len(secondary_nodes) != 0:
7126 raise errors.ProgrammerError("Wrong template configuration")
7128 opcodes.RequireSharedFileStorage()
7130 for idx, disk in enumerate(disk_info):
7131 disk_index = idx + base_index
7132 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7133 size=disk[constants.IDISK_SIZE],
7134 iv_name="disk/%d" % disk_index,
7135 logical_id=(file_driver,
7136 "%s/disk%d" % (file_storage_dir,
7138 mode=disk[constants.IDISK_MODE])
7139 disks.append(disk_dev)
7140 elif template_name == constants.DT_BLOCK:
7141 if len(secondary_nodes) != 0:
7142 raise errors.ProgrammerError("Wrong template configuration")
7144 for idx, disk in enumerate(disk_info):
7145 disk_index = idx + base_index
7146 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV,
7147 size=disk[constants.IDISK_SIZE],
7148 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL,
7149 disk[constants.IDISK_ADOPT]),
7150 iv_name="disk/%d" % disk_index,
7151 mode=disk[constants.IDISK_MODE])
7152 disks.append(disk_dev)
7155 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
7159 def _GetInstanceInfoText(instance):
7160 """Compute that text that should be added to the disk's metadata.
7163 return "originstname+%s" % instance.name
7166 def _CalcEta(time_taken, written, total_size):
7167 """Calculates the ETA based on size written and total size.
7169 @param time_taken: The time taken so far
7170 @param written: amount written so far
7171 @param total_size: The total size of data to be written
7172 @return: The remaining time in seconds
7175 avg_time = time_taken / float(written)
7176 return (total_size - written) * avg_time
7179 def _WipeDisks(lu, instance):
7180 """Wipes instance disks.
7182 @type lu: L{LogicalUnit}
7183 @param lu: the logical unit on whose behalf we execute
7184 @type instance: L{objects.Instance}
7185 @param instance: the instance whose disks we should create
7186 @return: the success of the wipe
7189 node = instance.primary_node
7191 for device in instance.disks:
7192 lu.cfg.SetDiskID(device, node)
7194 logging.info("Pause sync of instance %s disks", instance.name)
7195 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
7197 for idx, success in enumerate(result.payload):
7199 logging.warn("pause-sync of instance %s for disks %d failed",
7203 for idx, device in enumerate(instance.disks):
7204 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
7205 # MAX_WIPE_CHUNK at max
7206 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
7207 constants.MIN_WIPE_CHUNK_PERCENT)
7208 # we _must_ make this an int, otherwise rounding errors will
7210 wipe_chunk_size = int(wipe_chunk_size)
7212 lu.LogInfo("* Wiping disk %d", idx)
7213 logging.info("Wiping disk %d for instance %s, node %s using"
7214 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
7219 start_time = time.time()
7221 while offset < size:
7222 wipe_size = min(wipe_chunk_size, size - offset)
7223 logging.debug("Wiping disk %d, offset %s, chunk %s",
7224 idx, offset, wipe_size)
7225 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
7226 result.Raise("Could not wipe disk %d at offset %d for size %d" %
7227 (idx, offset, wipe_size))
7230 if now - last_output >= 60:
7231 eta = _CalcEta(now - start_time, offset, size)
7232 lu.LogInfo(" - done: %.1f%% ETA: %s" %
7233 (offset / float(size) * 100, utils.FormatSeconds(eta)))
7236 logging.info("Resume sync of instance %s disks", instance.name)
7238 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
7240 for idx, success in enumerate(result.payload):
7242 lu.LogWarning("Resume sync of disk %d failed, please have a"
7243 " look at the status and troubleshoot the issue", idx)
7244 logging.warn("resume-sync of instance %s for disks %d failed",
7248 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
7249 """Create all disks for an instance.
7251 This abstracts away some work from AddInstance.
7253 @type lu: L{LogicalUnit}
7254 @param lu: the logical unit on whose behalf we execute
7255 @type instance: L{objects.Instance}
7256 @param instance: the instance whose disks we should create
7258 @param to_skip: list of indices to skip
7259 @type target_node: string
7260 @param target_node: if passed, overrides the target node for creation
7262 @return: the success of the creation
7265 info = _GetInstanceInfoText(instance)
7266 if target_node is None:
7267 pnode = instance.primary_node
7268 all_nodes = instance.all_nodes
7273 if instance.disk_template in (constants.DT_FILE, constants.DT_SHARED_FILE):
7274 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7275 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
7277 result.Raise("Failed to create directory '%s' on"
7278 " node %s" % (file_storage_dir, pnode))
7280 # Note: this needs to be kept in sync with adding of disks in
7281 # LUInstanceSetParams
7282 for idx, device in enumerate(instance.disks):
7283 if to_skip and idx in to_skip:
7285 logging.info("Creating volume %s for instance %s",
7286 device.iv_name, instance.name)
7288 for node in all_nodes:
7289 f_create = node == pnode
7290 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
7293 def _RemoveDisks(lu, instance, target_node=None):
7294 """Remove all disks for an instance.
7296 This abstracts away some work from `AddInstance()` and
7297 `RemoveInstance()`. Note that in case some of the devices couldn't
7298 be removed, the removal will continue with the other ones (compare
7299 with `_CreateDisks()`).
7301 @type lu: L{LogicalUnit}
7302 @param lu: the logical unit on whose behalf we execute
7303 @type instance: L{objects.Instance}
7304 @param instance: the instance whose disks we should remove
7305 @type target_node: string
7306 @param target_node: used to override the node on which to remove the disks
7308 @return: the success of the removal
7311 logging.info("Removing block devices for instance %s", instance.name)
7314 for device in instance.disks:
7316 edata = [(target_node, device)]
7318 edata = device.ComputeNodeTree(instance.primary_node)
7319 for node, disk in edata:
7320 lu.cfg.SetDiskID(disk, node)
7321 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
7323 lu.LogWarning("Could not remove block device %s on node %s,"
7324 " continuing anyway: %s", device.iv_name, node, msg)
7327 if instance.disk_template == constants.DT_FILE:
7328 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7332 tgt = instance.primary_node
7333 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
7335 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
7336 file_storage_dir, instance.primary_node, result.fail_msg)
7342 def _ComputeDiskSizePerVG(disk_template, disks):
7343 """Compute disk size requirements in the volume group
7346 def _compute(disks, payload):
7347 """Universal algorithm.
7352 vgs[disk[constants.IDISK_VG]] = \
7353 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
7357 # Required free disk space as a function of disk and swap space
7359 constants.DT_DISKLESS: {},
7360 constants.DT_PLAIN: _compute(disks, 0),
7361 # 128 MB are added for drbd metadata for each disk
7362 constants.DT_DRBD8: _compute(disks, 128),
7363 constants.DT_FILE: {},
7364 constants.DT_SHARED_FILE: {},
7367 if disk_template not in req_size_dict:
7368 raise errors.ProgrammerError("Disk template '%s' size requirement"
7369 " is unknown" % disk_template)
7371 return req_size_dict[disk_template]
7374 def _ComputeDiskSize(disk_template, disks):
7375 """Compute disk size requirements in the volume group
7378 # Required free disk space as a function of disk and swap space
7380 constants.DT_DISKLESS: None,
7381 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
7382 # 128 MB are added for drbd metadata for each disk
7383 constants.DT_DRBD8: sum(d[constants.IDISK_SIZE] + 128 for d in disks),
7384 constants.DT_FILE: None,
7385 constants.DT_SHARED_FILE: 0,
7386 constants.DT_BLOCK: 0,
7389 if disk_template not in req_size_dict:
7390 raise errors.ProgrammerError("Disk template '%s' size requirement"
7391 " is unknown" % disk_template)
7393 return req_size_dict[disk_template]
7396 def _FilterVmNodes(lu, nodenames):
7397 """Filters out non-vm_capable nodes from a list.
7399 @type lu: L{LogicalUnit}
7400 @param lu: the logical unit for which we check
7401 @type nodenames: list
7402 @param nodenames: the list of nodes on which we should check
7404 @return: the list of vm-capable nodes
7407 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
7408 return [name for name in nodenames if name not in vm_nodes]
7411 def _CheckHVParams(lu, nodenames, hvname, hvparams):
7412 """Hypervisor parameter validation.
7414 This function abstract the hypervisor parameter validation to be
7415 used in both instance create and instance modify.
7417 @type lu: L{LogicalUnit}
7418 @param lu: the logical unit for which we check
7419 @type nodenames: list
7420 @param nodenames: the list of nodes on which we should check
7421 @type hvname: string
7422 @param hvname: the name of the hypervisor we should use
7423 @type hvparams: dict
7424 @param hvparams: the parameters which we need to check
7425 @raise errors.OpPrereqError: if the parameters are not valid
7428 nodenames = _FilterVmNodes(lu, nodenames)
7429 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
7432 for node in nodenames:
7436 info.Raise("Hypervisor parameter validation failed on node %s" % node)
7439 def _CheckOSParams(lu, required, nodenames, osname, osparams):
7440 """OS parameters validation.
7442 @type lu: L{LogicalUnit}
7443 @param lu: the logical unit for which we check
7444 @type required: boolean
7445 @param required: whether the validation should fail if the OS is not
7447 @type nodenames: list
7448 @param nodenames: the list of nodes on which we should check
7449 @type osname: string
7450 @param osname: the name of the hypervisor we should use
7451 @type osparams: dict
7452 @param osparams: the parameters which we need to check
7453 @raise errors.OpPrereqError: if the parameters are not valid
7456 nodenames = _FilterVmNodes(lu, nodenames)
7457 result = lu.rpc.call_os_validate(required, nodenames, osname,
7458 [constants.OS_VALIDATE_PARAMETERS],
7460 for node, nres in result.items():
7461 # we don't check for offline cases since this should be run only
7462 # against the master node and/or an instance's nodes
7463 nres.Raise("OS Parameters validation failed on node %s" % node)
7464 if not nres.payload:
7465 lu.LogInfo("OS %s not found on node %s, validation skipped",
7469 class LUInstanceCreate(LogicalUnit):
7470 """Create an instance.
7473 HPATH = "instance-add"
7474 HTYPE = constants.HTYPE_INSTANCE
7477 def CheckArguments(self):
7481 # do not require name_check to ease forward/backward compatibility
7483 if self.op.no_install and self.op.start:
7484 self.LogInfo("No-installation mode selected, disabling startup")
7485 self.op.start = False
7486 # validate/normalize the instance name
7487 self.op.instance_name = \
7488 netutils.Hostname.GetNormalizedName(self.op.instance_name)
7490 if self.op.ip_check and not self.op.name_check:
7491 # TODO: make the ip check more flexible and not depend on the name check
7492 raise errors.OpPrereqError("Cannot do IP address check without a name"
7493 " check", errors.ECODE_INVAL)
7495 # check nics' parameter names
7496 for nic in self.op.nics:
7497 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
7499 # check disks. parameter names and consistent adopt/no-adopt strategy
7500 has_adopt = has_no_adopt = False
7501 for disk in self.op.disks:
7502 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
7503 if constants.IDISK_ADOPT in disk:
7507 if has_adopt and has_no_adopt:
7508 raise errors.OpPrereqError("Either all disks are adopted or none is",
7511 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
7512 raise errors.OpPrereqError("Disk adoption is not supported for the"
7513 " '%s' disk template" %
7514 self.op.disk_template,
7516 if self.op.iallocator is not None:
7517 raise errors.OpPrereqError("Disk adoption not allowed with an"
7518 " iallocator script", errors.ECODE_INVAL)
7519 if self.op.mode == constants.INSTANCE_IMPORT:
7520 raise errors.OpPrereqError("Disk adoption not allowed for"
7521 " instance import", errors.ECODE_INVAL)
7523 if self.op.disk_template in constants.DTS_MUST_ADOPT:
7524 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
7525 " but no 'adopt' parameter given" %
7526 self.op.disk_template,
7529 self.adopt_disks = has_adopt
7531 # instance name verification
7532 if self.op.name_check:
7533 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
7534 self.op.instance_name = self.hostname1.name
7535 # used in CheckPrereq for ip ping check
7536 self.check_ip = self.hostname1.ip
7538 self.check_ip = None
7540 # file storage checks
7541 if (self.op.file_driver and
7542 not self.op.file_driver in constants.FILE_DRIVER):
7543 raise errors.OpPrereqError("Invalid file driver name '%s'" %
7544 self.op.file_driver, errors.ECODE_INVAL)
7546 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
7547 raise errors.OpPrereqError("File storage directory path not absolute",
7550 ### Node/iallocator related checks
7551 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
7553 if self.op.pnode is not None:
7554 if self.op.disk_template in constants.DTS_INT_MIRROR:
7555 if self.op.snode is None:
7556 raise errors.OpPrereqError("The networked disk templates need"
7557 " a mirror node", errors.ECODE_INVAL)
7559 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
7561 self.op.snode = None
7563 self._cds = _GetClusterDomainSecret()
7565 if self.op.mode == constants.INSTANCE_IMPORT:
7566 # On import force_variant must be True, because if we forced it at
7567 # initial install, our only chance when importing it back is that it
7569 self.op.force_variant = True
7571 if self.op.no_install:
7572 self.LogInfo("No-installation mode has no effect during import")
7574 elif self.op.mode == constants.INSTANCE_CREATE:
7575 if self.op.os_type is None:
7576 raise errors.OpPrereqError("No guest OS specified",
7578 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
7579 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
7580 " installation" % self.op.os_type,
7582 if self.op.disk_template is None:
7583 raise errors.OpPrereqError("No disk template specified",
7586 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7587 # Check handshake to ensure both clusters have the same domain secret
7588 src_handshake = self.op.source_handshake
7589 if not src_handshake:
7590 raise errors.OpPrereqError("Missing source handshake",
7593 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
7596 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
7599 # Load and check source CA
7600 self.source_x509_ca_pem = self.op.source_x509_ca
7601 if not self.source_x509_ca_pem:
7602 raise errors.OpPrereqError("Missing source X509 CA",
7606 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
7608 except OpenSSL.crypto.Error, err:
7609 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
7610 (err, ), errors.ECODE_INVAL)
7612 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
7613 if errcode is not None:
7614 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
7617 self.source_x509_ca = cert
7619 src_instance_name = self.op.source_instance_name
7620 if not src_instance_name:
7621 raise errors.OpPrereqError("Missing source instance name",
7624 self.source_instance_name = \
7625 netutils.GetHostname(name=src_instance_name).name
7628 raise errors.OpPrereqError("Invalid instance creation mode %r" %
7629 self.op.mode, errors.ECODE_INVAL)
7631 def ExpandNames(self):
7632 """ExpandNames for CreateInstance.
7634 Figure out the right locks for instance creation.
7637 self.needed_locks = {}
7639 instance_name = self.op.instance_name
7640 # this is just a preventive check, but someone might still add this
7641 # instance in the meantime, and creation will fail at lock-add time
7642 if instance_name in self.cfg.GetInstanceList():
7643 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7644 instance_name, errors.ECODE_EXISTS)
7646 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
7648 if self.op.iallocator:
7649 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7651 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
7652 nodelist = [self.op.pnode]
7653 if self.op.snode is not None:
7654 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
7655 nodelist.append(self.op.snode)
7656 self.needed_locks[locking.LEVEL_NODE] = nodelist
7658 # in case of import lock the source node too
7659 if self.op.mode == constants.INSTANCE_IMPORT:
7660 src_node = self.op.src_node
7661 src_path = self.op.src_path
7663 if src_path is None:
7664 self.op.src_path = src_path = self.op.instance_name
7666 if src_node is None:
7667 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7668 self.op.src_node = None
7669 if os.path.isabs(src_path):
7670 raise errors.OpPrereqError("Importing an instance from an absolute"
7671 " path requires a source node option",
7674 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
7675 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
7676 self.needed_locks[locking.LEVEL_NODE].append(src_node)
7677 if not os.path.isabs(src_path):
7678 self.op.src_path = src_path = \
7679 utils.PathJoin(constants.EXPORT_DIR, src_path)
7681 def _RunAllocator(self):
7682 """Run the allocator based on input opcode.
7685 nics = [n.ToDict() for n in self.nics]
7686 ial = IAllocator(self.cfg, self.rpc,
7687 mode=constants.IALLOCATOR_MODE_ALLOC,
7688 name=self.op.instance_name,
7689 disk_template=self.op.disk_template,
7692 vcpus=self.be_full[constants.BE_VCPUS],
7693 mem_size=self.be_full[constants.BE_MEMORY],
7696 hypervisor=self.op.hypervisor,
7699 ial.Run(self.op.iallocator)
7702 raise errors.OpPrereqError("Can't compute nodes using"
7703 " iallocator '%s': %s" %
7704 (self.op.iallocator, ial.info),
7706 if len(ial.result) != ial.required_nodes:
7707 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7708 " of nodes (%s), required %s" %
7709 (self.op.iallocator, len(ial.result),
7710 ial.required_nodes), errors.ECODE_FAULT)
7711 self.op.pnode = ial.result[0]
7712 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7713 self.op.instance_name, self.op.iallocator,
7714 utils.CommaJoin(ial.result))
7715 if ial.required_nodes == 2:
7716 self.op.snode = ial.result[1]
7718 def BuildHooksEnv(self):
7721 This runs on master, primary and secondary nodes of the instance.
7725 "ADD_MODE": self.op.mode,
7727 if self.op.mode == constants.INSTANCE_IMPORT:
7728 env["SRC_NODE"] = self.op.src_node
7729 env["SRC_PATH"] = self.op.src_path
7730 env["SRC_IMAGES"] = self.src_images
7732 env.update(_BuildInstanceHookEnv(
7733 name=self.op.instance_name,
7734 primary_node=self.op.pnode,
7735 secondary_nodes=self.secondaries,
7736 status=self.op.start,
7737 os_type=self.op.os_type,
7738 memory=self.be_full[constants.BE_MEMORY],
7739 vcpus=self.be_full[constants.BE_VCPUS],
7740 nics=_NICListToTuple(self, self.nics),
7741 disk_template=self.op.disk_template,
7742 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
7743 for d in self.disks],
7746 hypervisor_name=self.op.hypervisor,
7751 def BuildHooksNodes(self):
7752 """Build hooks nodes.
7755 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
7758 def _ReadExportInfo(self):
7759 """Reads the export information from disk.
7761 It will override the opcode source node and path with the actual
7762 information, if these two were not specified before.
7764 @return: the export information
7767 assert self.op.mode == constants.INSTANCE_IMPORT
7769 src_node = self.op.src_node
7770 src_path = self.op.src_path
7772 if src_node is None:
7773 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
7774 exp_list = self.rpc.call_export_list(locked_nodes)
7776 for node in exp_list:
7777 if exp_list[node].fail_msg:
7779 if src_path in exp_list[node].payload:
7781 self.op.src_node = src_node = node
7782 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
7786 raise errors.OpPrereqError("No export found for relative path %s" %
7787 src_path, errors.ECODE_INVAL)
7789 _CheckNodeOnline(self, src_node)
7790 result = self.rpc.call_export_info(src_node, src_path)
7791 result.Raise("No export or invalid export found in dir %s" % src_path)
7793 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
7794 if not export_info.has_section(constants.INISECT_EXP):
7795 raise errors.ProgrammerError("Corrupted export config",
7796 errors.ECODE_ENVIRON)
7798 ei_version = export_info.get(constants.INISECT_EXP, "version")
7799 if (int(ei_version) != constants.EXPORT_VERSION):
7800 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
7801 (ei_version, constants.EXPORT_VERSION),
7802 errors.ECODE_ENVIRON)
7805 def _ReadExportParams(self, einfo):
7806 """Use export parameters as defaults.
7808 In case the opcode doesn't specify (as in override) some instance
7809 parameters, then try to use them from the export information, if
7813 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
7815 if self.op.disk_template is None:
7816 if einfo.has_option(constants.INISECT_INS, "disk_template"):
7817 self.op.disk_template = einfo.get(constants.INISECT_INS,
7820 raise errors.OpPrereqError("No disk template specified and the export"
7821 " is missing the disk_template information",
7824 if not self.op.disks:
7825 if einfo.has_option(constants.INISECT_INS, "disk_count"):
7827 # TODO: import the disk iv_name too
7828 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
7829 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
7830 disks.append({constants.IDISK_SIZE: disk_sz})
7831 self.op.disks = disks
7833 raise errors.OpPrereqError("No disk info specified and the export"
7834 " is missing the disk information",
7837 if (not self.op.nics and
7838 einfo.has_option(constants.INISECT_INS, "nic_count")):
7840 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
7842 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
7843 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
7848 if (self.op.hypervisor is None and
7849 einfo.has_option(constants.INISECT_INS, "hypervisor")):
7850 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
7851 if einfo.has_section(constants.INISECT_HYP):
7852 # use the export parameters but do not override the ones
7853 # specified by the user
7854 for name, value in einfo.items(constants.INISECT_HYP):
7855 if name not in self.op.hvparams:
7856 self.op.hvparams[name] = value
7858 if einfo.has_section(constants.INISECT_BEP):
7859 # use the parameters, without overriding
7860 for name, value in einfo.items(constants.INISECT_BEP):
7861 if name not in self.op.beparams:
7862 self.op.beparams[name] = value
7864 # try to read the parameters old style, from the main section
7865 for name in constants.BES_PARAMETERS:
7866 if (name not in self.op.beparams and
7867 einfo.has_option(constants.INISECT_INS, name)):
7868 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
7870 if einfo.has_section(constants.INISECT_OSP):
7871 # use the parameters, without overriding
7872 for name, value in einfo.items(constants.INISECT_OSP):
7873 if name not in self.op.osparams:
7874 self.op.osparams[name] = value
7876 def _RevertToDefaults(self, cluster):
7877 """Revert the instance parameters to the default values.
7881 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7882 for name in self.op.hvparams.keys():
7883 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7884 del self.op.hvparams[name]
7886 be_defs = cluster.SimpleFillBE({})
7887 for name in self.op.beparams.keys():
7888 if name in be_defs and be_defs[name] == self.op.beparams[name]:
7889 del self.op.beparams[name]
7891 nic_defs = cluster.SimpleFillNIC({})
7892 for nic in self.op.nics:
7893 for name in constants.NICS_PARAMETERS:
7894 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7897 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7898 for name in self.op.osparams.keys():
7899 if name in os_defs and os_defs[name] == self.op.osparams[name]:
7900 del self.op.osparams[name]
7902 def CheckPrereq(self):
7903 """Check prerequisites.
7906 if self.op.mode == constants.INSTANCE_IMPORT:
7907 export_info = self._ReadExportInfo()
7908 self._ReadExportParams(export_info)
7910 if (not self.cfg.GetVGName() and
7911 self.op.disk_template not in constants.DTS_NOT_LVM):
7912 raise errors.OpPrereqError("Cluster does not support lvm-based"
7913 " instances", errors.ECODE_STATE)
7915 if self.op.hypervisor is None:
7916 self.op.hypervisor = self.cfg.GetHypervisorType()
7918 cluster = self.cfg.GetClusterInfo()
7919 enabled_hvs = cluster.enabled_hypervisors
7920 if self.op.hypervisor not in enabled_hvs:
7921 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7922 " cluster (%s)" % (self.op.hypervisor,
7923 ",".join(enabled_hvs)),
7926 # check hypervisor parameter syntax (locally)
7927 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7928 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7930 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7931 hv_type.CheckParameterSyntax(filled_hvp)
7932 self.hv_full = filled_hvp
7933 # check that we don't specify global parameters on an instance
7934 _CheckGlobalHvParams(self.op.hvparams)
7936 # fill and remember the beparams dict
7937 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7938 self.be_full = cluster.SimpleFillBE(self.op.beparams)
7940 # build os parameters
7941 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7943 # now that hvp/bep are in final format, let's reset to defaults,
7945 if self.op.identify_defaults:
7946 self._RevertToDefaults(cluster)
7950 for idx, nic in enumerate(self.op.nics):
7951 nic_mode_req = nic.get(constants.INIC_MODE, None)
7952 nic_mode = nic_mode_req
7953 if nic_mode is None:
7954 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7956 # in routed mode, for the first nic, the default ip is 'auto'
7957 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7958 default_ip_mode = constants.VALUE_AUTO
7960 default_ip_mode = constants.VALUE_NONE
7962 # ip validity checks
7963 ip = nic.get(constants.INIC_IP, default_ip_mode)
7964 if ip is None or ip.lower() == constants.VALUE_NONE:
7966 elif ip.lower() == constants.VALUE_AUTO:
7967 if not self.op.name_check:
7968 raise errors.OpPrereqError("IP address set to auto but name checks"
7969 " have been skipped",
7971 nic_ip = self.hostname1.ip
7973 if not netutils.IPAddress.IsValid(ip):
7974 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7978 # TODO: check the ip address for uniqueness
7979 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7980 raise errors.OpPrereqError("Routed nic mode requires an ip address",
7983 # MAC address verification
7984 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
7985 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7986 mac = utils.NormalizeAndValidateMac(mac)
7989 self.cfg.ReserveMAC(mac, self.proc.GetECId())
7990 except errors.ReservationError:
7991 raise errors.OpPrereqError("MAC address %s already in use"
7992 " in cluster" % mac,
7993 errors.ECODE_NOTUNIQUE)
7995 # Build nic parameters
7996 link = nic.get(constants.INIC_LINK, None)
7999 nicparams[constants.NIC_MODE] = nic_mode_req
8001 nicparams[constants.NIC_LINK] = link
8003 check_params = cluster.SimpleFillNIC(nicparams)
8004 objects.NIC.CheckParameterSyntax(check_params)
8005 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
8007 # disk checks/pre-build
8008 default_vg = self.cfg.GetVGName()
8010 for disk in self.op.disks:
8011 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
8012 if mode not in constants.DISK_ACCESS_SET:
8013 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
8014 mode, errors.ECODE_INVAL)
8015 size = disk.get(constants.IDISK_SIZE, None)
8017 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
8020 except (TypeError, ValueError):
8021 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
8024 data_vg = disk.get(constants.IDISK_VG, default_vg)
8026 constants.IDISK_SIZE: size,
8027 constants.IDISK_MODE: mode,
8028 constants.IDISK_VG: data_vg,
8029 constants.IDISK_METAVG: disk.get(constants.IDISK_METAVG, data_vg),
8031 if constants.IDISK_ADOPT in disk:
8032 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
8033 self.disks.append(new_disk)
8035 if self.op.mode == constants.INSTANCE_IMPORT:
8037 # Check that the new instance doesn't have less disks than the export
8038 instance_disks = len(self.disks)
8039 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
8040 if instance_disks < export_disks:
8041 raise errors.OpPrereqError("Not enough disks to import."
8042 " (instance: %d, export: %d)" %
8043 (instance_disks, export_disks),
8047 for idx in range(export_disks):
8048 option = 'disk%d_dump' % idx
8049 if export_info.has_option(constants.INISECT_INS, option):
8050 # FIXME: are the old os-es, disk sizes, etc. useful?
8051 export_name = export_info.get(constants.INISECT_INS, option)
8052 image = utils.PathJoin(self.op.src_path, export_name)
8053 disk_images.append(image)
8055 disk_images.append(False)
8057 self.src_images = disk_images
8059 old_name = export_info.get(constants.INISECT_INS, 'name')
8061 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
8062 except (TypeError, ValueError), err:
8063 raise errors.OpPrereqError("Invalid export file, nic_count is not"
8064 " an integer: %s" % str(err),
8066 if self.op.instance_name == old_name:
8067 for idx, nic in enumerate(self.nics):
8068 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
8069 nic_mac_ini = 'nic%d_mac' % idx
8070 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
8072 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
8074 # ip ping checks (we use the same ip that was resolved in ExpandNames)
8075 if self.op.ip_check:
8076 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
8077 raise errors.OpPrereqError("IP %s of instance %s already in use" %
8078 (self.check_ip, self.op.instance_name),
8079 errors.ECODE_NOTUNIQUE)
8081 #### mac address generation
8082 # By generating here the mac address both the allocator and the hooks get
8083 # the real final mac address rather than the 'auto' or 'generate' value.
8084 # There is a race condition between the generation and the instance object
8085 # creation, which means that we know the mac is valid now, but we're not
8086 # sure it will be when we actually add the instance. If things go bad
8087 # adding the instance will abort because of a duplicate mac, and the
8088 # creation job will fail.
8089 for nic in self.nics:
8090 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8091 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
8095 if self.op.iallocator is not None:
8096 self._RunAllocator()
8098 #### node related checks
8100 # check primary node
8101 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
8102 assert self.pnode is not None, \
8103 "Cannot retrieve locked node %s" % self.op.pnode
8105 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
8106 pnode.name, errors.ECODE_STATE)
8108 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
8109 pnode.name, errors.ECODE_STATE)
8110 if not pnode.vm_capable:
8111 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
8112 " '%s'" % pnode.name, errors.ECODE_STATE)
8114 self.secondaries = []
8116 # mirror node verification
8117 if self.op.disk_template in constants.DTS_INT_MIRROR:
8118 if self.op.snode == pnode.name:
8119 raise errors.OpPrereqError("The secondary node cannot be the"
8120 " primary node", errors.ECODE_INVAL)
8121 _CheckNodeOnline(self, self.op.snode)
8122 _CheckNodeNotDrained(self, self.op.snode)
8123 _CheckNodeVmCapable(self, self.op.snode)
8124 self.secondaries.append(self.op.snode)
8126 nodenames = [pnode.name] + self.secondaries
8128 if not self.adopt_disks:
8129 # Check lv size requirements, if not adopting
8130 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
8131 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
8133 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
8134 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
8135 disk[constants.IDISK_ADOPT])
8136 for disk in self.disks])
8137 if len(all_lvs) != len(self.disks):
8138 raise errors.OpPrereqError("Duplicate volume names given for adoption",
8140 for lv_name in all_lvs:
8142 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
8143 # to ReserveLV uses the same syntax
8144 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
8145 except errors.ReservationError:
8146 raise errors.OpPrereqError("LV named %s used by another instance" %
8147 lv_name, errors.ECODE_NOTUNIQUE)
8149 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
8150 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
8152 node_lvs = self.rpc.call_lv_list([pnode.name],
8153 vg_names.payload.keys())[pnode.name]
8154 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
8155 node_lvs = node_lvs.payload
8157 delta = all_lvs.difference(node_lvs.keys())
8159 raise errors.OpPrereqError("Missing logical volume(s): %s" %
8160 utils.CommaJoin(delta),
8162 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
8164 raise errors.OpPrereqError("Online logical volumes found, cannot"
8165 " adopt: %s" % utils.CommaJoin(online_lvs),
8167 # update the size of disk based on what is found
8168 for dsk in self.disks:
8169 dsk[constants.IDISK_SIZE] = \
8170 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
8171 dsk[constants.IDISK_ADOPT])][0]))
8173 elif self.op.disk_template == constants.DT_BLOCK:
8174 # Normalize and de-duplicate device paths
8175 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
8176 for disk in self.disks])
8177 if len(all_disks) != len(self.disks):
8178 raise errors.OpPrereqError("Duplicate disk names given for adoption",
8180 baddisks = [d for d in all_disks
8181 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
8183 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
8184 " cannot be adopted" %
8185 (", ".join(baddisks),
8186 constants.ADOPTABLE_BLOCKDEV_ROOT),
8189 node_disks = self.rpc.call_bdev_sizes([pnode.name],
8190 list(all_disks))[pnode.name]
8191 node_disks.Raise("Cannot get block device information from node %s" %
8193 node_disks = node_disks.payload
8194 delta = all_disks.difference(node_disks.keys())
8196 raise errors.OpPrereqError("Missing block device(s): %s" %
8197 utils.CommaJoin(delta),
8199 for dsk in self.disks:
8200 dsk[constants.IDISK_SIZE] = \
8201 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
8203 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
8205 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
8206 # check OS parameters (remotely)
8207 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
8209 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
8211 # memory check on primary node
8213 _CheckNodeFreeMemory(self, self.pnode.name,
8214 "creating instance %s" % self.op.instance_name,
8215 self.be_full[constants.BE_MEMORY],
8218 self.dry_run_result = list(nodenames)
8220 def Exec(self, feedback_fn):
8221 """Create and add the instance to the cluster.
8224 instance = self.op.instance_name
8225 pnode_name = self.pnode.name
8227 ht_kind = self.op.hypervisor
8228 if ht_kind in constants.HTS_REQ_PORT:
8229 network_port = self.cfg.AllocatePort()
8233 if constants.ENABLE_FILE_STORAGE or constants.ENABLE_SHARED_FILE_STORAGE:
8234 # this is needed because os.path.join does not accept None arguments
8235 if self.op.file_storage_dir is None:
8236 string_file_storage_dir = ""
8238 string_file_storage_dir = self.op.file_storage_dir
8240 # build the full file storage dir path
8241 if self.op.disk_template == constants.DT_SHARED_FILE:
8242 get_fsd_fn = self.cfg.GetSharedFileStorageDir
8244 get_fsd_fn = self.cfg.GetFileStorageDir
8246 file_storage_dir = utils.PathJoin(get_fsd_fn(),
8247 string_file_storage_dir, instance)
8249 file_storage_dir = ""
8251 disks = _GenerateDiskTemplate(self,
8252 self.op.disk_template,
8253 instance, pnode_name,
8257 self.op.file_driver,
8261 iobj = objects.Instance(name=instance, os=self.op.os_type,
8262 primary_node=pnode_name,
8263 nics=self.nics, disks=disks,
8264 disk_template=self.op.disk_template,
8266 network_port=network_port,
8267 beparams=self.op.beparams,
8268 hvparams=self.op.hvparams,
8269 hypervisor=self.op.hypervisor,
8270 osparams=self.op.osparams,
8273 if self.adopt_disks:
8274 if self.op.disk_template == constants.DT_PLAIN:
8275 # rename LVs to the newly-generated names; we need to construct
8276 # 'fake' LV disks with the old data, plus the new unique_id
8277 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
8279 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
8280 rename_to.append(t_dsk.logical_id)
8281 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
8282 self.cfg.SetDiskID(t_dsk, pnode_name)
8283 result = self.rpc.call_blockdev_rename(pnode_name,
8284 zip(tmp_disks, rename_to))
8285 result.Raise("Failed to rename adoped LVs")
8287 feedback_fn("* creating instance disks...")
8289 _CreateDisks(self, iobj)
8290 except errors.OpExecError:
8291 self.LogWarning("Device creation failed, reverting...")
8293 _RemoveDisks(self, iobj)
8295 self.cfg.ReleaseDRBDMinors(instance)
8298 feedback_fn("adding instance %s to cluster config" % instance)
8300 self.cfg.AddInstance(iobj, self.proc.GetECId())
8302 # Declare that we don't want to remove the instance lock anymore, as we've
8303 # added the instance to the config
8304 del self.remove_locks[locking.LEVEL_INSTANCE]
8306 if self.op.mode == constants.INSTANCE_IMPORT:
8307 # Release unused nodes
8308 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
8311 _ReleaseLocks(self, locking.LEVEL_NODE)
8314 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
8315 feedback_fn("* wiping instance disks...")
8317 _WipeDisks(self, iobj)
8318 except errors.OpExecError, err:
8319 logging.exception("Wiping disks failed")
8320 self.LogWarning("Wiping instance disks failed (%s)", err)
8324 # Something is already wrong with the disks, don't do anything else
8326 elif self.op.wait_for_sync:
8327 disk_abort = not _WaitForSync(self, iobj)
8328 elif iobj.disk_template in constants.DTS_INT_MIRROR:
8329 # make sure the disks are not degraded (still sync-ing is ok)
8331 feedback_fn("* checking mirrors status")
8332 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
8337 _RemoveDisks(self, iobj)
8338 self.cfg.RemoveInstance(iobj.name)
8339 # Make sure the instance lock gets removed
8340 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
8341 raise errors.OpExecError("There are some degraded disks for"
8344 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
8345 if self.op.mode == constants.INSTANCE_CREATE:
8346 if not self.op.no_install:
8347 feedback_fn("* running the instance OS create scripts...")
8348 # FIXME: pass debug option from opcode to backend
8349 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
8350 self.op.debug_level)
8351 result.Raise("Could not add os for instance %s"
8352 " on node %s" % (instance, pnode_name))
8354 elif self.op.mode == constants.INSTANCE_IMPORT:
8355 feedback_fn("* running the instance OS import scripts...")
8359 for idx, image in enumerate(self.src_images):
8363 # FIXME: pass debug option from opcode to backend
8364 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
8365 constants.IEIO_FILE, (image, ),
8366 constants.IEIO_SCRIPT,
8367 (iobj.disks[idx], idx),
8369 transfers.append(dt)
8372 masterd.instance.TransferInstanceData(self, feedback_fn,
8373 self.op.src_node, pnode_name,
8374 self.pnode.secondary_ip,
8376 if not compat.all(import_result):
8377 self.LogWarning("Some disks for instance %s on node %s were not"
8378 " imported successfully" % (instance, pnode_name))
8380 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
8381 feedback_fn("* preparing remote import...")
8382 # The source cluster will stop the instance before attempting to make a
8383 # connection. In some cases stopping an instance can take a long time,
8384 # hence the shutdown timeout is added to the connection timeout.
8385 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
8386 self.op.source_shutdown_timeout)
8387 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
8389 assert iobj.primary_node == self.pnode.name
8391 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
8392 self.source_x509_ca,
8393 self._cds, timeouts)
8394 if not compat.all(disk_results):
8395 # TODO: Should the instance still be started, even if some disks
8396 # failed to import (valid for local imports, too)?
8397 self.LogWarning("Some disks for instance %s on node %s were not"
8398 " imported successfully" % (instance, pnode_name))
8400 # Run rename script on newly imported instance
8401 assert iobj.name == instance
8402 feedback_fn("Running rename script for %s" % instance)
8403 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
8404 self.source_instance_name,
8405 self.op.debug_level)
8407 self.LogWarning("Failed to run rename script for %s on node"
8408 " %s: %s" % (instance, pnode_name, result.fail_msg))
8411 # also checked in the prereq part
8412 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
8416 iobj.admin_up = True
8417 self.cfg.Update(iobj, feedback_fn)
8418 logging.info("Starting instance %s on node %s", instance, pnode_name)
8419 feedback_fn("* starting instance...")
8420 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
8421 result.Raise("Could not start instance")
8423 return list(iobj.all_nodes)
8426 class LUInstanceConsole(NoHooksLU):
8427 """Connect to an instance's console.
8429 This is somewhat special in that it returns the command line that
8430 you need to run on the master node in order to connect to the
8436 def ExpandNames(self):
8437 self._ExpandAndLockInstance()
8439 def CheckPrereq(self):
8440 """Check prerequisites.
8442 This checks that the instance is in the cluster.
8445 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8446 assert self.instance is not None, \
8447 "Cannot retrieve locked instance %s" % self.op.instance_name
8448 _CheckNodeOnline(self, self.instance.primary_node)
8450 def Exec(self, feedback_fn):
8451 """Connect to the console of an instance
8454 instance = self.instance
8455 node = instance.primary_node
8457 node_insts = self.rpc.call_instance_list([node],
8458 [instance.hypervisor])[node]
8459 node_insts.Raise("Can't get node information from %s" % node)
8461 if instance.name not in node_insts.payload:
8462 if instance.admin_up:
8463 state = constants.INSTST_ERRORDOWN
8465 state = constants.INSTST_ADMINDOWN
8466 raise errors.OpExecError("Instance %s is not running (state %s)" %
8467 (instance.name, state))
8469 logging.debug("Connecting to console of %s on %s", instance.name, node)
8471 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
8474 def _GetInstanceConsole(cluster, instance):
8475 """Returns console information for an instance.
8477 @type cluster: L{objects.Cluster}
8478 @type instance: L{objects.Instance}
8482 hyper = hypervisor.GetHypervisor(instance.hypervisor)
8483 # beparams and hvparams are passed separately, to avoid editing the
8484 # instance and then saving the defaults in the instance itself.
8485 hvparams = cluster.FillHV(instance)
8486 beparams = cluster.FillBE(instance)
8487 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
8489 assert console.instance == instance.name
8490 assert console.Validate()
8492 return console.ToDict()
8495 class LUInstanceReplaceDisks(LogicalUnit):
8496 """Replace the disks of an instance.
8499 HPATH = "mirrors-replace"
8500 HTYPE = constants.HTYPE_INSTANCE
8503 def CheckArguments(self):
8504 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
8507 def ExpandNames(self):
8508 self._ExpandAndLockInstance()
8510 assert locking.LEVEL_NODE not in self.needed_locks
8511 assert locking.LEVEL_NODEGROUP not in self.needed_locks
8513 assert self.op.iallocator is None or self.op.remote_node is None, \
8514 "Conflicting options"
8516 if self.op.remote_node is not None:
8517 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8519 # Warning: do not remove the locking of the new secondary here
8520 # unless DRBD8.AddChildren is changed to work in parallel;
8521 # currently it doesn't since parallel invocations of
8522 # FindUnusedMinor will conflict
8523 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
8524 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
8526 self.needed_locks[locking.LEVEL_NODE] = []
8527 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8529 if self.op.iallocator is not None:
8530 # iallocator will select a new node in the same group
8531 self.needed_locks[locking.LEVEL_NODEGROUP] = []
8533 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
8534 self.op.iallocator, self.op.remote_node,
8535 self.op.disks, False, self.op.early_release)
8537 self.tasklets = [self.replacer]
8539 def DeclareLocks(self, level):
8540 if level == locking.LEVEL_NODEGROUP:
8541 assert self.op.remote_node is None
8542 assert self.op.iallocator is not None
8543 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
8545 self.share_locks[locking.LEVEL_NODEGROUP] = 1
8546 self.needed_locks[locking.LEVEL_NODEGROUP] = \
8547 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
8549 elif level == locking.LEVEL_NODE:
8550 if self.op.iallocator is not None:
8551 assert self.op.remote_node is None
8552 assert not self.needed_locks[locking.LEVEL_NODE]
8554 # Lock member nodes of all locked groups
8555 self.needed_locks[locking.LEVEL_NODE] = [node_name
8556 for group_uuid in self.acquired_locks[locking.LEVEL_NODEGROUP]
8557 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
8559 self._LockInstancesNodes()
8561 def BuildHooksEnv(self):
8564 This runs on the master, the primary and all the secondaries.
8567 instance = self.replacer.instance
8569 "MODE": self.op.mode,
8570 "NEW_SECONDARY": self.op.remote_node,
8571 "OLD_SECONDARY": instance.secondary_nodes[0],
8573 env.update(_BuildInstanceHookEnvByObject(self, instance))
8576 def BuildHooksNodes(self):
8577 """Build hooks nodes.
8580 instance = self.replacer.instance
8582 self.cfg.GetMasterNode(),
8583 instance.primary_node,
8585 if self.op.remote_node is not None:
8586 nl.append(self.op.remote_node)
8589 def CheckPrereq(self):
8590 """Check prerequisites.
8593 assert (locking.LEVEL_NODEGROUP in self.acquired_locks or
8594 self.op.iallocator is None)
8596 if locking.LEVEL_NODEGROUP in self.acquired_locks:
8597 groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
8598 prevgroups = self.acquired_locks[locking.LEVEL_NODEGROUP]
8599 if prevgroups != groups:
8600 raise errors.OpExecError("Node groups used by instance '%s' changed"
8601 " since lock was acquired, current list is %r,"
8602 " used to be '%s'" %
8603 (self.op.instance_name,
8604 utils.CommaJoin(groups),
8605 utils.CommaJoin(prevgroups)))
8607 return LogicalUnit.CheckPrereq(self)
8610 class TLReplaceDisks(Tasklet):
8611 """Replaces disks for an instance.
8613 Note: Locking is not within the scope of this class.
8616 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
8617 disks, delay_iallocator, early_release):
8618 """Initializes this class.
8621 Tasklet.__init__(self, lu)
8624 self.instance_name = instance_name
8626 self.iallocator_name = iallocator_name
8627 self.remote_node = remote_node
8629 self.delay_iallocator = delay_iallocator
8630 self.early_release = early_release
8633 self.instance = None
8634 self.new_node = None
8635 self.target_node = None
8636 self.other_node = None
8637 self.remote_node_info = None
8638 self.node_secondary_ip = None
8641 def CheckArguments(mode, remote_node, iallocator):
8642 """Helper function for users of this class.
8645 # check for valid parameter combination
8646 if mode == constants.REPLACE_DISK_CHG:
8647 if remote_node is None and iallocator is None:
8648 raise errors.OpPrereqError("When changing the secondary either an"
8649 " iallocator script must be used or the"
8650 " new node given", errors.ECODE_INVAL)
8652 if remote_node is not None and iallocator is not None:
8653 raise errors.OpPrereqError("Give either the iallocator or the new"
8654 " secondary, not both", errors.ECODE_INVAL)
8656 elif remote_node is not None or iallocator is not None:
8657 # Not replacing the secondary
8658 raise errors.OpPrereqError("The iallocator and new node options can"
8659 " only be used when changing the"
8660 " secondary node", errors.ECODE_INVAL)
8663 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
8664 """Compute a new secondary node using an IAllocator.
8667 ial = IAllocator(lu.cfg, lu.rpc,
8668 mode=constants.IALLOCATOR_MODE_RELOC,
8670 relocate_from=relocate_from)
8672 ial.Run(iallocator_name)
8675 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
8676 " %s" % (iallocator_name, ial.info),
8679 if len(ial.result) != ial.required_nodes:
8680 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8681 " of nodes (%s), required %s" %
8683 len(ial.result), ial.required_nodes),
8686 remote_node_name = ial.result[0]
8688 lu.LogInfo("Selected new secondary for instance '%s': %s",
8689 instance_name, remote_node_name)
8691 return remote_node_name
8693 def _FindFaultyDisks(self, node_name):
8694 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
8697 def _CheckDisksActivated(self, instance):
8698 """Checks if the instance disks are activated.
8700 @param instance: The instance to check disks
8701 @return: True if they are activated, False otherwise
8704 nodes = instance.all_nodes
8706 for idx, dev in enumerate(instance.disks):
8708 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
8709 self.cfg.SetDiskID(dev, node)
8711 result = self.rpc.call_blockdev_find(node, dev)
8715 elif result.fail_msg or not result.payload:
8720 def CheckPrereq(self):
8721 """Check prerequisites.
8723 This checks that the instance is in the cluster.
8726 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
8727 assert instance is not None, \
8728 "Cannot retrieve locked instance %s" % self.instance_name
8730 if instance.disk_template != constants.DT_DRBD8:
8731 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
8732 " instances", errors.ECODE_INVAL)
8734 if len(instance.secondary_nodes) != 1:
8735 raise errors.OpPrereqError("The instance has a strange layout,"
8736 " expected one secondary but found %d" %
8737 len(instance.secondary_nodes),
8740 if not self.delay_iallocator:
8741 self._CheckPrereq2()
8743 def _CheckPrereq2(self):
8744 """Check prerequisites, second part.
8746 This function should always be part of CheckPrereq. It was separated and is
8747 now called from Exec because during node evacuation iallocator was only
8748 called with an unmodified cluster model, not taking planned changes into
8752 instance = self.instance
8753 secondary_node = instance.secondary_nodes[0]
8755 if self.iallocator_name is None:
8756 remote_node = self.remote_node
8758 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
8759 instance.name, instance.secondary_nodes)
8761 if remote_node is None:
8762 self.remote_node_info = None
8764 assert remote_node in self.lu.acquired_locks[locking.LEVEL_NODE], \
8765 "Remote node '%s' is not locked" % remote_node
8767 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
8768 assert self.remote_node_info is not None, \
8769 "Cannot retrieve locked node %s" % remote_node
8771 if remote_node == self.instance.primary_node:
8772 raise errors.OpPrereqError("The specified node is the primary node of"
8773 " the instance", errors.ECODE_INVAL)
8775 if remote_node == secondary_node:
8776 raise errors.OpPrereqError("The specified node is already the"
8777 " secondary node of the instance",
8780 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
8781 constants.REPLACE_DISK_CHG):
8782 raise errors.OpPrereqError("Cannot specify disks to be replaced",
8785 if self.mode == constants.REPLACE_DISK_AUTO:
8786 if not self._CheckDisksActivated(instance):
8787 raise errors.OpPrereqError("Please run activate-disks on instance %s"
8788 " first" % self.instance_name,
8790 faulty_primary = self._FindFaultyDisks(instance.primary_node)
8791 faulty_secondary = self._FindFaultyDisks(secondary_node)
8793 if faulty_primary and faulty_secondary:
8794 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
8795 " one node and can not be repaired"
8796 " automatically" % self.instance_name,
8800 self.disks = faulty_primary
8801 self.target_node = instance.primary_node
8802 self.other_node = secondary_node
8803 check_nodes = [self.target_node, self.other_node]
8804 elif faulty_secondary:
8805 self.disks = faulty_secondary
8806 self.target_node = secondary_node
8807 self.other_node = instance.primary_node
8808 check_nodes = [self.target_node, self.other_node]
8814 # Non-automatic modes
8815 if self.mode == constants.REPLACE_DISK_PRI:
8816 self.target_node = instance.primary_node
8817 self.other_node = secondary_node
8818 check_nodes = [self.target_node, self.other_node]
8820 elif self.mode == constants.REPLACE_DISK_SEC:
8821 self.target_node = secondary_node
8822 self.other_node = instance.primary_node
8823 check_nodes = [self.target_node, self.other_node]
8825 elif self.mode == constants.REPLACE_DISK_CHG:
8826 self.new_node = remote_node
8827 self.other_node = instance.primary_node
8828 self.target_node = secondary_node
8829 check_nodes = [self.new_node, self.other_node]
8831 _CheckNodeNotDrained(self.lu, remote_node)
8832 _CheckNodeVmCapable(self.lu, remote_node)
8834 old_node_info = self.cfg.GetNodeInfo(secondary_node)
8835 assert old_node_info is not None
8836 if old_node_info.offline and not self.early_release:
8837 # doesn't make sense to delay the release
8838 self.early_release = True
8839 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
8840 " early-release mode", secondary_node)
8843 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
8846 # If not specified all disks should be replaced
8848 self.disks = range(len(self.instance.disks))
8850 for node in check_nodes:
8851 _CheckNodeOnline(self.lu, node)
8853 touched_nodes = frozenset(node_name for node_name in [self.new_node,
8856 if node_name is not None)
8858 # Release unneeded node locks
8859 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
8861 # Release any owned node group
8862 if self.lu.context.glm.is_owned(locking.LEVEL_NODEGROUP):
8863 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
8865 # Check whether disks are valid
8866 for disk_idx in self.disks:
8867 instance.FindDisk(disk_idx)
8869 # Get secondary node IP addresses
8870 self.node_secondary_ip = \
8871 dict((node_name, self.cfg.GetNodeInfo(node_name).secondary_ip)
8872 for node_name in touched_nodes)
8874 def Exec(self, feedback_fn):
8875 """Execute disk replacement.
8877 This dispatches the disk replacement to the appropriate handler.
8880 if self.delay_iallocator:
8881 self._CheckPrereq2()
8884 # Verify owned locks before starting operation
8885 owned_locks = self.lu.context.glm.list_owned(locking.LEVEL_NODE)
8886 assert set(owned_locks) == set(self.node_secondary_ip), \
8887 ("Incorrect node locks, owning %s, expected %s" %
8888 (owned_locks, self.node_secondary_ip.keys()))
8890 owned_locks = self.lu.context.glm.list_owned(locking.LEVEL_INSTANCE)
8891 assert list(owned_locks) == [self.instance_name], \
8892 "Instance '%s' not locked" % self.instance_name
8894 assert not self.lu.context.glm.is_owned(locking.LEVEL_NODEGROUP), \
8895 "Should not own any node group lock at this point"
8898 feedback_fn("No disks need replacement")
8901 feedback_fn("Replacing disk(s) %s for %s" %
8902 (utils.CommaJoin(self.disks), self.instance.name))
8904 activate_disks = (not self.instance.admin_up)
8906 # Activate the instance disks if we're replacing them on a down instance
8908 _StartInstanceDisks(self.lu, self.instance, True)
8911 # Should we replace the secondary node?
8912 if self.new_node is not None:
8913 fn = self._ExecDrbd8Secondary
8915 fn = self._ExecDrbd8DiskOnly
8917 result = fn(feedback_fn)
8919 # Deactivate the instance disks if we're replacing them on a
8922 _SafeShutdownInstanceDisks(self.lu, self.instance)
8925 # Verify owned locks
8926 owned_locks = self.lu.context.glm.list_owned(locking.LEVEL_NODE)
8927 nodes = frozenset(self.node_secondary_ip)
8928 assert ((self.early_release and not owned_locks) or
8929 (not self.early_release and not (set(owned_locks) - nodes))), \
8930 ("Not owning the correct locks, early_release=%s, owned=%r,"
8931 " nodes=%r" % (self.early_release, owned_locks, nodes))
8935 def _CheckVolumeGroup(self, nodes):
8936 self.lu.LogInfo("Checking volume groups")
8938 vgname = self.cfg.GetVGName()
8940 # Make sure volume group exists on all involved nodes
8941 results = self.rpc.call_vg_list(nodes)
8943 raise errors.OpExecError("Can't list volume groups on the nodes")
8947 res.Raise("Error checking node %s" % node)
8948 if vgname not in res.payload:
8949 raise errors.OpExecError("Volume group '%s' not found on node %s" %
8952 def _CheckDisksExistence(self, nodes):
8953 # Check disk existence
8954 for idx, dev in enumerate(self.instance.disks):
8955 if idx not in self.disks:
8959 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
8960 self.cfg.SetDiskID(dev, node)
8962 result = self.rpc.call_blockdev_find(node, dev)
8964 msg = result.fail_msg
8965 if msg or not result.payload:
8967 msg = "disk not found"
8968 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
8971 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
8972 for idx, dev in enumerate(self.instance.disks):
8973 if idx not in self.disks:
8976 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
8979 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
8981 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
8982 " replace disks for instance %s" %
8983 (node_name, self.instance.name))
8985 def _CreateNewStorage(self, node_name):
8988 for idx, dev in enumerate(self.instance.disks):
8989 if idx not in self.disks:
8992 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
8994 self.cfg.SetDiskID(dev, node_name)
8996 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
8997 names = _GenerateUniqueNames(self.lu, lv_names)
8999 vg_data = dev.children[0].logical_id[0]
9000 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
9001 logical_id=(vg_data, names[0]))
9002 vg_meta = dev.children[1].logical_id[0]
9003 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
9004 logical_id=(vg_meta, names[1]))
9006 new_lvs = [lv_data, lv_meta]
9007 old_lvs = dev.children
9008 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
9010 # we pass force_create=True to force the LVM creation
9011 for new_lv in new_lvs:
9012 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
9013 _GetInstanceInfoText(self.instance), False)
9017 def _CheckDevices(self, node_name, iv_names):
9018 for name, (dev, _, _) in iv_names.iteritems():
9019 self.cfg.SetDiskID(dev, node_name)
9021 result = self.rpc.call_blockdev_find(node_name, dev)
9023 msg = result.fail_msg
9024 if msg or not result.payload:
9026 msg = "disk not found"
9027 raise errors.OpExecError("Can't find DRBD device %s: %s" %
9030 if result.payload.is_degraded:
9031 raise errors.OpExecError("DRBD device %s is degraded!" % name)
9033 def _RemoveOldStorage(self, node_name, iv_names):
9034 for name, (_, old_lvs, _) in iv_names.iteritems():
9035 self.lu.LogInfo("Remove logical volumes for %s" % name)
9038 self.cfg.SetDiskID(lv, node_name)
9040 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
9042 self.lu.LogWarning("Can't remove old LV: %s" % msg,
9043 hint="remove unused LVs manually")
9045 def _ExecDrbd8DiskOnly(self, feedback_fn):
9046 """Replace a disk on the primary or secondary for DRBD 8.
9048 The algorithm for replace is quite complicated:
9050 1. for each disk to be replaced:
9052 1. create new LVs on the target node with unique names
9053 1. detach old LVs from the drbd device
9054 1. rename old LVs to name_replaced.<time_t>
9055 1. rename new LVs to old LVs
9056 1. attach the new LVs (with the old names now) to the drbd device
9058 1. wait for sync across all devices
9060 1. for each modified disk:
9062 1. remove old LVs (which have the name name_replaces.<time_t>)
9064 Failures are not very well handled.
9069 # Step: check device activation
9070 self.lu.LogStep(1, steps_total, "Check device existence")
9071 self._CheckDisksExistence([self.other_node, self.target_node])
9072 self._CheckVolumeGroup([self.target_node, self.other_node])
9074 # Step: check other node consistency
9075 self.lu.LogStep(2, steps_total, "Check peer consistency")
9076 self._CheckDisksConsistency(self.other_node,
9077 self.other_node == self.instance.primary_node,
9080 # Step: create new storage
9081 self.lu.LogStep(3, steps_total, "Allocate new storage")
9082 iv_names = self._CreateNewStorage(self.target_node)
9084 # Step: for each lv, detach+rename*2+attach
9085 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9086 for dev, old_lvs, new_lvs in iv_names.itervalues():
9087 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
9089 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
9091 result.Raise("Can't detach drbd from local storage on node"
9092 " %s for device %s" % (self.target_node, dev.iv_name))
9094 #cfg.Update(instance)
9096 # ok, we created the new LVs, so now we know we have the needed
9097 # storage; as such, we proceed on the target node to rename
9098 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
9099 # using the assumption that logical_id == physical_id (which in
9100 # turn is the unique_id on that node)
9102 # FIXME(iustin): use a better name for the replaced LVs
9103 temp_suffix = int(time.time())
9104 ren_fn = lambda d, suff: (d.physical_id[0],
9105 d.physical_id[1] + "_replaced-%s" % suff)
9107 # Build the rename list based on what LVs exist on the node
9108 rename_old_to_new = []
9109 for to_ren in old_lvs:
9110 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
9111 if not result.fail_msg and result.payload:
9113 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
9115 self.lu.LogInfo("Renaming the old LVs on the target node")
9116 result = self.rpc.call_blockdev_rename(self.target_node,
9118 result.Raise("Can't rename old LVs on node %s" % self.target_node)
9120 # Now we rename the new LVs to the old LVs
9121 self.lu.LogInfo("Renaming the new LVs on the target node")
9122 rename_new_to_old = [(new, old.physical_id)
9123 for old, new in zip(old_lvs, new_lvs)]
9124 result = self.rpc.call_blockdev_rename(self.target_node,
9126 result.Raise("Can't rename new LVs on node %s" % self.target_node)
9128 for old, new in zip(old_lvs, new_lvs):
9129 new.logical_id = old.logical_id
9130 self.cfg.SetDiskID(new, self.target_node)
9132 for disk in old_lvs:
9133 disk.logical_id = ren_fn(disk, temp_suffix)
9134 self.cfg.SetDiskID(disk, self.target_node)
9136 # Now that the new lvs have the old name, we can add them to the device
9137 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
9138 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
9140 msg = result.fail_msg
9142 for new_lv in new_lvs:
9143 msg2 = self.rpc.call_blockdev_remove(self.target_node,
9146 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
9147 hint=("cleanup manually the unused logical"
9149 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
9151 dev.children = new_lvs
9153 self.cfg.Update(self.instance, feedback_fn)
9156 if self.early_release:
9157 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9159 self._RemoveOldStorage(self.target_node, iv_names)
9160 # WARNING: we release both node locks here, do not do other RPCs
9161 # than WaitForSync to the primary node
9162 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9163 names=[self.target_node, self.other_node])
9166 # This can fail as the old devices are degraded and _WaitForSync
9167 # does a combined result over all disks, so we don't check its return value
9168 self.lu.LogStep(cstep, steps_total, "Sync devices")
9170 _WaitForSync(self.lu, self.instance)
9172 # Check all devices manually
9173 self._CheckDevices(self.instance.primary_node, iv_names)
9175 # Step: remove old storage
9176 if not self.early_release:
9177 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9179 self._RemoveOldStorage(self.target_node, iv_names)
9181 def _ExecDrbd8Secondary(self, feedback_fn):
9182 """Replace the secondary node for DRBD 8.
9184 The algorithm for replace is quite complicated:
9185 - for all disks of the instance:
9186 - create new LVs on the new node with same names
9187 - shutdown the drbd device on the old secondary
9188 - disconnect the drbd network on the primary
9189 - create the drbd device on the new secondary
9190 - network attach the drbd on the primary, using an artifice:
9191 the drbd code for Attach() will connect to the network if it
9192 finds a device which is connected to the good local disks but
9194 - wait for sync across all devices
9195 - remove all disks from the old secondary
9197 Failures are not very well handled.
9202 # Step: check device activation
9203 self.lu.LogStep(1, steps_total, "Check device existence")
9204 self._CheckDisksExistence([self.instance.primary_node])
9205 self._CheckVolumeGroup([self.instance.primary_node])
9207 # Step: check other node consistency
9208 self.lu.LogStep(2, steps_total, "Check peer consistency")
9209 self._CheckDisksConsistency(self.instance.primary_node, True, True)
9211 # Step: create new storage
9212 self.lu.LogStep(3, steps_total, "Allocate new storage")
9213 for idx, dev in enumerate(self.instance.disks):
9214 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
9215 (self.new_node, idx))
9216 # we pass force_create=True to force LVM creation
9217 for new_lv in dev.children:
9218 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
9219 _GetInstanceInfoText(self.instance), False)
9221 # Step 4: dbrd minors and drbd setups changes
9222 # after this, we must manually remove the drbd minors on both the
9223 # error and the success paths
9224 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9225 minors = self.cfg.AllocateDRBDMinor([self.new_node
9226 for dev in self.instance.disks],
9228 logging.debug("Allocated minors %r", minors)
9231 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
9232 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
9233 (self.new_node, idx))
9234 # create new devices on new_node; note that we create two IDs:
9235 # one without port, so the drbd will be activated without
9236 # networking information on the new node at this stage, and one
9237 # with network, for the latter activation in step 4
9238 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
9239 if self.instance.primary_node == o_node1:
9242 assert self.instance.primary_node == o_node2, "Three-node instance?"
9245 new_alone_id = (self.instance.primary_node, self.new_node, None,
9246 p_minor, new_minor, o_secret)
9247 new_net_id = (self.instance.primary_node, self.new_node, o_port,
9248 p_minor, new_minor, o_secret)
9250 iv_names[idx] = (dev, dev.children, new_net_id)
9251 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
9253 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
9254 logical_id=new_alone_id,
9255 children=dev.children,
9258 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
9259 _GetInstanceInfoText(self.instance), False)
9260 except errors.GenericError:
9261 self.cfg.ReleaseDRBDMinors(self.instance.name)
9264 # We have new devices, shutdown the drbd on the old secondary
9265 for idx, dev in enumerate(self.instance.disks):
9266 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
9267 self.cfg.SetDiskID(dev, self.target_node)
9268 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
9270 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
9271 "node: %s" % (idx, msg),
9272 hint=("Please cleanup this device manually as"
9273 " soon as possible"))
9275 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
9276 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
9277 self.node_secondary_ip,
9278 self.instance.disks)\
9279 [self.instance.primary_node]
9281 msg = result.fail_msg
9283 # detaches didn't succeed (unlikely)
9284 self.cfg.ReleaseDRBDMinors(self.instance.name)
9285 raise errors.OpExecError("Can't detach the disks from the network on"
9286 " old node: %s" % (msg,))
9288 # if we managed to detach at least one, we update all the disks of
9289 # the instance to point to the new secondary
9290 self.lu.LogInfo("Updating instance configuration")
9291 for dev, _, new_logical_id in iv_names.itervalues():
9292 dev.logical_id = new_logical_id
9293 self.cfg.SetDiskID(dev, self.instance.primary_node)
9295 self.cfg.Update(self.instance, feedback_fn)
9297 # and now perform the drbd attach
9298 self.lu.LogInfo("Attaching primary drbds to new secondary"
9299 " (standalone => connected)")
9300 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
9302 self.node_secondary_ip,
9303 self.instance.disks,
9306 for to_node, to_result in result.items():
9307 msg = to_result.fail_msg
9309 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
9311 hint=("please do a gnt-instance info to see the"
9312 " status of disks"))
9314 if self.early_release:
9315 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9317 self._RemoveOldStorage(self.target_node, iv_names)
9318 # WARNING: we release all node locks here, do not do other RPCs
9319 # than WaitForSync to the primary node
9320 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9321 names=[self.instance.primary_node,
9326 # This can fail as the old devices are degraded and _WaitForSync
9327 # does a combined result over all disks, so we don't check its return value
9328 self.lu.LogStep(cstep, steps_total, "Sync devices")
9330 _WaitForSync(self.lu, self.instance)
9332 # Check all devices manually
9333 self._CheckDevices(self.instance.primary_node, iv_names)
9335 # Step: remove old storage
9336 if not self.early_release:
9337 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9338 self._RemoveOldStorage(self.target_node, iv_names)
9341 class LURepairNodeStorage(NoHooksLU):
9342 """Repairs the volume group on a node.
9347 def CheckArguments(self):
9348 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
9350 storage_type = self.op.storage_type
9352 if (constants.SO_FIX_CONSISTENCY not in
9353 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
9354 raise errors.OpPrereqError("Storage units of type '%s' can not be"
9355 " repaired" % storage_type,
9358 def ExpandNames(self):
9359 self.needed_locks = {
9360 locking.LEVEL_NODE: [self.op.node_name],
9363 def _CheckFaultyDisks(self, instance, node_name):
9364 """Ensure faulty disks abort the opcode or at least warn."""
9366 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
9368 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
9369 " node '%s'" % (instance.name, node_name),
9371 except errors.OpPrereqError, err:
9372 if self.op.ignore_consistency:
9373 self.proc.LogWarning(str(err.args[0]))
9377 def CheckPrereq(self):
9378 """Check prerequisites.
9381 # Check whether any instance on this node has faulty disks
9382 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
9383 if not inst.admin_up:
9385 check_nodes = set(inst.all_nodes)
9386 check_nodes.discard(self.op.node_name)
9387 for inst_node_name in check_nodes:
9388 self._CheckFaultyDisks(inst, inst_node_name)
9390 def Exec(self, feedback_fn):
9391 feedback_fn("Repairing storage unit '%s' on %s ..." %
9392 (self.op.name, self.op.node_name))
9394 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
9395 result = self.rpc.call_storage_execute(self.op.node_name,
9396 self.op.storage_type, st_args,
9398 constants.SO_FIX_CONSISTENCY)
9399 result.Raise("Failed to repair storage unit '%s' on %s" %
9400 (self.op.name, self.op.node_name))
9403 class LUNodeEvacStrategy(NoHooksLU):
9404 """Computes the node evacuation strategy.
9409 def CheckArguments(self):
9410 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
9412 def ExpandNames(self):
9413 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
9414 self.needed_locks = locks = {}
9415 if self.op.remote_node is None:
9416 locks[locking.LEVEL_NODE] = locking.ALL_SET
9418 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9419 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
9421 def Exec(self, feedback_fn):
9422 if self.op.remote_node is not None:
9424 for node in self.op.nodes:
9425 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
9428 if i.primary_node == self.op.remote_node:
9429 raise errors.OpPrereqError("Node %s is the primary node of"
9430 " instance %s, cannot use it as"
9432 (self.op.remote_node, i.name),
9434 result.append([i.name, self.op.remote_node])
9436 ial = IAllocator(self.cfg, self.rpc,
9437 mode=constants.IALLOCATOR_MODE_MEVAC,
9438 evac_nodes=self.op.nodes)
9439 ial.Run(self.op.iallocator, validate=True)
9441 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
9447 class LUInstanceGrowDisk(LogicalUnit):
9448 """Grow a disk of an instance.
9452 HTYPE = constants.HTYPE_INSTANCE
9455 def ExpandNames(self):
9456 self._ExpandAndLockInstance()
9457 self.needed_locks[locking.LEVEL_NODE] = []
9458 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9460 def DeclareLocks(self, level):
9461 if level == locking.LEVEL_NODE:
9462 self._LockInstancesNodes()
9464 def BuildHooksEnv(self):
9467 This runs on the master, the primary and all the secondaries.
9471 "DISK": self.op.disk,
9472 "AMOUNT": self.op.amount,
9474 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9477 def BuildHooksNodes(self):
9478 """Build hooks nodes.
9481 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9484 def CheckPrereq(self):
9485 """Check prerequisites.
9487 This checks that the instance is in the cluster.
9490 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9491 assert instance is not None, \
9492 "Cannot retrieve locked instance %s" % self.op.instance_name
9493 nodenames = list(instance.all_nodes)
9494 for node in nodenames:
9495 _CheckNodeOnline(self, node)
9497 self.instance = instance
9499 if instance.disk_template not in constants.DTS_GROWABLE:
9500 raise errors.OpPrereqError("Instance's disk layout does not support"
9501 " growing", errors.ECODE_INVAL)
9503 self.disk = instance.FindDisk(self.op.disk)
9505 if instance.disk_template not in (constants.DT_FILE,
9506 constants.DT_SHARED_FILE):
9507 # TODO: check the free disk space for file, when that feature will be
9509 _CheckNodesFreeDiskPerVG(self, nodenames,
9510 self.disk.ComputeGrowth(self.op.amount))
9512 def Exec(self, feedback_fn):
9513 """Execute disk grow.
9516 instance = self.instance
9519 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
9521 raise errors.OpExecError("Cannot activate block device to grow")
9523 for node in instance.all_nodes:
9524 self.cfg.SetDiskID(disk, node)
9525 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
9526 result.Raise("Grow request failed to node %s" % node)
9528 # TODO: Rewrite code to work properly
9529 # DRBD goes into sync mode for a short amount of time after executing the
9530 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
9531 # calling "resize" in sync mode fails. Sleeping for a short amount of
9532 # time is a work-around.
9535 disk.RecordGrow(self.op.amount)
9536 self.cfg.Update(instance, feedback_fn)
9537 if self.op.wait_for_sync:
9538 disk_abort = not _WaitForSync(self, instance, disks=[disk])
9540 self.proc.LogWarning("Disk sync-ing has not returned a good"
9541 " status; please check the instance")
9542 if not instance.admin_up:
9543 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
9544 elif not instance.admin_up:
9545 self.proc.LogWarning("Not shutting down the disk even if the instance is"
9546 " not supposed to be running because no wait for"
9547 " sync mode was requested")
9550 class LUInstanceQueryData(NoHooksLU):
9551 """Query runtime instance data.
9556 def ExpandNames(self):
9557 self.needed_locks = {}
9559 # Use locking if requested or when non-static information is wanted
9560 if not (self.op.static or self.op.use_locking):
9561 self.LogWarning("Non-static data requested, locks need to be acquired")
9562 self.op.use_locking = True
9564 if self.op.instances or not self.op.use_locking:
9565 # Expand instance names right here
9566 self.wanted_names = _GetWantedInstances(self, self.op.instances)
9568 # Will use acquired locks
9569 self.wanted_names = None
9571 if self.op.use_locking:
9572 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9574 if self.wanted_names is None:
9575 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
9577 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
9579 self.needed_locks[locking.LEVEL_NODE] = []
9580 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9581 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9583 def DeclareLocks(self, level):
9584 if self.op.use_locking and level == locking.LEVEL_NODE:
9585 self._LockInstancesNodes()
9587 def CheckPrereq(self):
9588 """Check prerequisites.
9590 This only checks the optional instance list against the existing names.
9593 if self.wanted_names is None:
9594 assert self.op.use_locking, "Locking was not used"
9595 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
9597 self.wanted_instances = [self.cfg.GetInstanceInfo(name)
9598 for name in self.wanted_names]
9600 def _ComputeBlockdevStatus(self, node, instance_name, dev):
9601 """Returns the status of a block device
9604 if self.op.static or not node:
9607 self.cfg.SetDiskID(dev, node)
9609 result = self.rpc.call_blockdev_find(node, dev)
9613 result.Raise("Can't compute disk status for %s" % instance_name)
9615 status = result.payload
9619 return (status.dev_path, status.major, status.minor,
9620 status.sync_percent, status.estimated_time,
9621 status.is_degraded, status.ldisk_status)
9623 def _ComputeDiskStatus(self, instance, snode, dev):
9624 """Compute block device status.
9627 if dev.dev_type in constants.LDS_DRBD:
9628 # we change the snode then (otherwise we use the one passed in)
9629 if dev.logical_id[0] == instance.primary_node:
9630 snode = dev.logical_id[1]
9632 snode = dev.logical_id[0]
9634 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
9636 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
9639 dev_children = [self._ComputeDiskStatus(instance, snode, child)
9640 for child in dev.children]
9645 "iv_name": dev.iv_name,
9646 "dev_type": dev.dev_type,
9647 "logical_id": dev.logical_id,
9648 "physical_id": dev.physical_id,
9649 "pstatus": dev_pstatus,
9650 "sstatus": dev_sstatus,
9651 "children": dev_children,
9656 def Exec(self, feedback_fn):
9657 """Gather and return data"""
9660 cluster = self.cfg.GetClusterInfo()
9662 for instance in self.wanted_instances:
9663 if not self.op.static:
9664 remote_info = self.rpc.call_instance_info(instance.primary_node,
9666 instance.hypervisor)
9667 remote_info.Raise("Error checking node %s" % instance.primary_node)
9668 remote_info = remote_info.payload
9669 if remote_info and "state" in remote_info:
9672 remote_state = "down"
9675 if instance.admin_up:
9678 config_state = "down"
9680 disks = [self._ComputeDiskStatus(instance, None, device)
9681 for device in instance.disks]
9683 result[instance.name] = {
9684 "name": instance.name,
9685 "config_state": config_state,
9686 "run_state": remote_state,
9687 "pnode": instance.primary_node,
9688 "snodes": instance.secondary_nodes,
9690 # this happens to be the same format used for hooks
9691 "nics": _NICListToTuple(self, instance.nics),
9692 "disk_template": instance.disk_template,
9694 "hypervisor": instance.hypervisor,
9695 "network_port": instance.network_port,
9696 "hv_instance": instance.hvparams,
9697 "hv_actual": cluster.FillHV(instance, skip_globals=True),
9698 "be_instance": instance.beparams,
9699 "be_actual": cluster.FillBE(instance),
9700 "os_instance": instance.osparams,
9701 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
9702 "serial_no": instance.serial_no,
9703 "mtime": instance.mtime,
9704 "ctime": instance.ctime,
9705 "uuid": instance.uuid,
9711 class LUInstanceSetParams(LogicalUnit):
9712 """Modifies an instances's parameters.
9715 HPATH = "instance-modify"
9716 HTYPE = constants.HTYPE_INSTANCE
9719 def CheckArguments(self):
9720 if not (self.op.nics or self.op.disks or self.op.disk_template or
9721 self.op.hvparams or self.op.beparams or self.op.os_name):
9722 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
9724 if self.op.hvparams:
9725 _CheckGlobalHvParams(self.op.hvparams)
9729 for disk_op, disk_dict in self.op.disks:
9730 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
9731 if disk_op == constants.DDM_REMOVE:
9734 elif disk_op == constants.DDM_ADD:
9737 if not isinstance(disk_op, int):
9738 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
9739 if not isinstance(disk_dict, dict):
9740 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
9741 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9743 if disk_op == constants.DDM_ADD:
9744 mode = disk_dict.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
9745 if mode not in constants.DISK_ACCESS_SET:
9746 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
9748 size = disk_dict.get(constants.IDISK_SIZE, None)
9750 raise errors.OpPrereqError("Required disk parameter size missing",
9754 except (TypeError, ValueError), err:
9755 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
9756 str(err), errors.ECODE_INVAL)
9757 disk_dict[constants.IDISK_SIZE] = size
9759 # modification of disk
9760 if constants.IDISK_SIZE in disk_dict:
9761 raise errors.OpPrereqError("Disk size change not possible, use"
9762 " grow-disk", errors.ECODE_INVAL)
9764 if disk_addremove > 1:
9765 raise errors.OpPrereqError("Only one disk add or remove operation"
9766 " supported at a time", errors.ECODE_INVAL)
9768 if self.op.disks and self.op.disk_template is not None:
9769 raise errors.OpPrereqError("Disk template conversion and other disk"
9770 " changes not supported at the same time",
9773 if (self.op.disk_template and
9774 self.op.disk_template in constants.DTS_INT_MIRROR and
9775 self.op.remote_node is None):
9776 raise errors.OpPrereqError("Changing the disk template to a mirrored"
9777 " one requires specifying a secondary node",
9782 for nic_op, nic_dict in self.op.nics:
9783 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
9784 if nic_op == constants.DDM_REMOVE:
9787 elif nic_op == constants.DDM_ADD:
9790 if not isinstance(nic_op, int):
9791 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
9792 if not isinstance(nic_dict, dict):
9793 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
9794 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9796 # nic_dict should be a dict
9797 nic_ip = nic_dict.get(constants.INIC_IP, None)
9798 if nic_ip is not None:
9799 if nic_ip.lower() == constants.VALUE_NONE:
9800 nic_dict[constants.INIC_IP] = None
9802 if not netutils.IPAddress.IsValid(nic_ip):
9803 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
9806 nic_bridge = nic_dict.get('bridge', None)
9807 nic_link = nic_dict.get(constants.INIC_LINK, None)
9808 if nic_bridge and nic_link:
9809 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
9810 " at the same time", errors.ECODE_INVAL)
9811 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
9812 nic_dict['bridge'] = None
9813 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
9814 nic_dict[constants.INIC_LINK] = None
9816 if nic_op == constants.DDM_ADD:
9817 nic_mac = nic_dict.get(constants.INIC_MAC, None)
9819 nic_dict[constants.INIC_MAC] = constants.VALUE_AUTO
9821 if constants.INIC_MAC in nic_dict:
9822 nic_mac = nic_dict[constants.INIC_MAC]
9823 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9824 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
9826 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
9827 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
9828 " modifying an existing nic",
9831 if nic_addremove > 1:
9832 raise errors.OpPrereqError("Only one NIC add or remove operation"
9833 " supported at a time", errors.ECODE_INVAL)
9835 def ExpandNames(self):
9836 self._ExpandAndLockInstance()
9837 self.needed_locks[locking.LEVEL_NODE] = []
9838 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9840 def DeclareLocks(self, level):
9841 if level == locking.LEVEL_NODE:
9842 self._LockInstancesNodes()
9843 if self.op.disk_template and self.op.remote_node:
9844 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9845 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
9847 def BuildHooksEnv(self):
9850 This runs on the master, primary and secondaries.
9854 if constants.BE_MEMORY in self.be_new:
9855 args['memory'] = self.be_new[constants.BE_MEMORY]
9856 if constants.BE_VCPUS in self.be_new:
9857 args['vcpus'] = self.be_new[constants.BE_VCPUS]
9858 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
9859 # information at all.
9862 nic_override = dict(self.op.nics)
9863 for idx, nic in enumerate(self.instance.nics):
9864 if idx in nic_override:
9865 this_nic_override = nic_override[idx]
9867 this_nic_override = {}
9868 if constants.INIC_IP in this_nic_override:
9869 ip = this_nic_override[constants.INIC_IP]
9872 if constants.INIC_MAC in this_nic_override:
9873 mac = this_nic_override[constants.INIC_MAC]
9876 if idx in self.nic_pnew:
9877 nicparams = self.nic_pnew[idx]
9879 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
9880 mode = nicparams[constants.NIC_MODE]
9881 link = nicparams[constants.NIC_LINK]
9882 args['nics'].append((ip, mac, mode, link))
9883 if constants.DDM_ADD in nic_override:
9884 ip = nic_override[constants.DDM_ADD].get(constants.INIC_IP, None)
9885 mac = nic_override[constants.DDM_ADD][constants.INIC_MAC]
9886 nicparams = self.nic_pnew[constants.DDM_ADD]
9887 mode = nicparams[constants.NIC_MODE]
9888 link = nicparams[constants.NIC_LINK]
9889 args['nics'].append((ip, mac, mode, link))
9890 elif constants.DDM_REMOVE in nic_override:
9891 del args['nics'][-1]
9893 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
9894 if self.op.disk_template:
9895 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
9899 def BuildHooksNodes(self):
9900 """Build hooks nodes.
9903 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9906 def CheckPrereq(self):
9907 """Check prerequisites.
9909 This only checks the instance list against the existing names.
9912 # checking the new params on the primary/secondary nodes
9914 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9915 cluster = self.cluster = self.cfg.GetClusterInfo()
9916 assert self.instance is not None, \
9917 "Cannot retrieve locked instance %s" % self.op.instance_name
9918 pnode = instance.primary_node
9919 nodelist = list(instance.all_nodes)
9922 if self.op.os_name and not self.op.force:
9923 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
9924 self.op.force_variant)
9925 instance_os = self.op.os_name
9927 instance_os = instance.os
9929 if self.op.disk_template:
9930 if instance.disk_template == self.op.disk_template:
9931 raise errors.OpPrereqError("Instance already has disk template %s" %
9932 instance.disk_template, errors.ECODE_INVAL)
9934 if (instance.disk_template,
9935 self.op.disk_template) not in self._DISK_CONVERSIONS:
9936 raise errors.OpPrereqError("Unsupported disk template conversion from"
9937 " %s to %s" % (instance.disk_template,
9938 self.op.disk_template),
9940 _CheckInstanceDown(self, instance, "cannot change disk template")
9941 if self.op.disk_template in constants.DTS_INT_MIRROR:
9942 if self.op.remote_node == pnode:
9943 raise errors.OpPrereqError("Given new secondary node %s is the same"
9944 " as the primary node of the instance" %
9945 self.op.remote_node, errors.ECODE_STATE)
9946 _CheckNodeOnline(self, self.op.remote_node)
9947 _CheckNodeNotDrained(self, self.op.remote_node)
9948 # FIXME: here we assume that the old instance type is DT_PLAIN
9949 assert instance.disk_template == constants.DT_PLAIN
9950 disks = [{constants.IDISK_SIZE: d.size,
9951 constants.IDISK_VG: d.logical_id[0]}
9952 for d in instance.disks]
9953 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
9954 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
9956 # hvparams processing
9957 if self.op.hvparams:
9958 hv_type = instance.hypervisor
9959 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
9960 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
9961 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
9964 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
9965 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
9966 self.hv_new = hv_new # the new actual values
9967 self.hv_inst = i_hvdict # the new dict (without defaults)
9969 self.hv_new = self.hv_inst = {}
9971 # beparams processing
9972 if self.op.beparams:
9973 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
9975 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
9976 be_new = cluster.SimpleFillBE(i_bedict)
9977 self.be_new = be_new # the new actual values
9978 self.be_inst = i_bedict # the new dict (without defaults)
9980 self.be_new = self.be_inst = {}
9982 # osparams processing
9983 if self.op.osparams:
9984 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
9985 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
9986 self.os_inst = i_osdict # the new dict (without defaults)
9992 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
9993 mem_check_list = [pnode]
9994 if be_new[constants.BE_AUTO_BALANCE]:
9995 # either we changed auto_balance to yes or it was from before
9996 mem_check_list.extend(instance.secondary_nodes)
9997 instance_info = self.rpc.call_instance_info(pnode, instance.name,
9998 instance.hypervisor)
9999 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
10000 instance.hypervisor)
10001 pninfo = nodeinfo[pnode]
10002 msg = pninfo.fail_msg
10004 # Assume the primary node is unreachable and go ahead
10005 self.warn.append("Can't get info from primary node %s: %s" %
10007 elif not isinstance(pninfo.payload.get('memory_free', None), int):
10008 self.warn.append("Node data from primary node %s doesn't contain"
10009 " free memory information" % pnode)
10010 elif instance_info.fail_msg:
10011 self.warn.append("Can't get instance runtime information: %s" %
10012 instance_info.fail_msg)
10014 if instance_info.payload:
10015 current_mem = int(instance_info.payload['memory'])
10017 # Assume instance not running
10018 # (there is a slight race condition here, but it's not very probable,
10019 # and we have no other way to check)
10021 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
10022 pninfo.payload['memory_free'])
10024 raise errors.OpPrereqError("This change will prevent the instance"
10025 " from starting, due to %d MB of memory"
10026 " missing on its primary node" % miss_mem,
10027 errors.ECODE_NORES)
10029 if be_new[constants.BE_AUTO_BALANCE]:
10030 for node, nres in nodeinfo.items():
10031 if node not in instance.secondary_nodes:
10033 msg = nres.fail_msg
10035 self.warn.append("Can't get info from secondary node %s: %s" %
10037 elif not isinstance(nres.payload.get('memory_free', None), int):
10038 self.warn.append("Secondary node %s didn't return free"
10039 " memory information" % node)
10040 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
10041 self.warn.append("Not enough memory to failover instance to"
10042 " secondary node %s" % node)
10046 self.nic_pinst = {}
10047 for nic_op, nic_dict in self.op.nics:
10048 if nic_op == constants.DDM_REMOVE:
10049 if not instance.nics:
10050 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
10051 errors.ECODE_INVAL)
10053 if nic_op != constants.DDM_ADD:
10055 if not instance.nics:
10056 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
10057 " no NICs" % nic_op,
10058 errors.ECODE_INVAL)
10059 if nic_op < 0 or nic_op >= len(instance.nics):
10060 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
10062 (nic_op, len(instance.nics) - 1),
10063 errors.ECODE_INVAL)
10064 old_nic_params = instance.nics[nic_op].nicparams
10065 old_nic_ip = instance.nics[nic_op].ip
10067 old_nic_params = {}
10070 update_params_dict = dict([(key, nic_dict[key])
10071 for key in constants.NICS_PARAMETERS
10072 if key in nic_dict])
10074 if 'bridge' in nic_dict:
10075 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
10077 new_nic_params = _GetUpdatedParams(old_nic_params,
10078 update_params_dict)
10079 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
10080 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
10081 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
10082 self.nic_pinst[nic_op] = new_nic_params
10083 self.nic_pnew[nic_op] = new_filled_nic_params
10084 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
10086 if new_nic_mode == constants.NIC_MODE_BRIDGED:
10087 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
10088 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
10090 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
10092 self.warn.append(msg)
10094 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
10095 if new_nic_mode == constants.NIC_MODE_ROUTED:
10096 if constants.INIC_IP in nic_dict:
10097 nic_ip = nic_dict[constants.INIC_IP]
10099 nic_ip = old_nic_ip
10101 raise errors.OpPrereqError('Cannot set the nic ip to None'
10102 ' on a routed nic', errors.ECODE_INVAL)
10103 if constants.INIC_MAC in nic_dict:
10104 nic_mac = nic_dict[constants.INIC_MAC]
10105 if nic_mac is None:
10106 raise errors.OpPrereqError('Cannot set the nic mac to None',
10107 errors.ECODE_INVAL)
10108 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
10109 # otherwise generate the mac
10110 nic_dict[constants.INIC_MAC] = \
10111 self.cfg.GenerateMAC(self.proc.GetECId())
10113 # or validate/reserve the current one
10115 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
10116 except errors.ReservationError:
10117 raise errors.OpPrereqError("MAC address %s already in use"
10118 " in cluster" % nic_mac,
10119 errors.ECODE_NOTUNIQUE)
10122 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
10123 raise errors.OpPrereqError("Disk operations not supported for"
10124 " diskless instances",
10125 errors.ECODE_INVAL)
10126 for disk_op, _ in self.op.disks:
10127 if disk_op == constants.DDM_REMOVE:
10128 if len(instance.disks) == 1:
10129 raise errors.OpPrereqError("Cannot remove the last disk of"
10130 " an instance", errors.ECODE_INVAL)
10131 _CheckInstanceDown(self, instance, "cannot remove disks")
10133 if (disk_op == constants.DDM_ADD and
10134 len(instance.disks) >= constants.MAX_DISKS):
10135 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
10136 " add more" % constants.MAX_DISKS,
10137 errors.ECODE_STATE)
10138 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
10140 if disk_op < 0 or disk_op >= len(instance.disks):
10141 raise errors.OpPrereqError("Invalid disk index %s, valid values"
10143 (disk_op, len(instance.disks)),
10144 errors.ECODE_INVAL)
10148 def _ConvertPlainToDrbd(self, feedback_fn):
10149 """Converts an instance from plain to drbd.
10152 feedback_fn("Converting template to drbd")
10153 instance = self.instance
10154 pnode = instance.primary_node
10155 snode = self.op.remote_node
10157 # create a fake disk info for _GenerateDiskTemplate
10158 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
10159 constants.IDISK_VG: d.logical_id[0]}
10160 for d in instance.disks]
10161 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
10162 instance.name, pnode, [snode],
10163 disk_info, None, None, 0, feedback_fn)
10164 info = _GetInstanceInfoText(instance)
10165 feedback_fn("Creating aditional volumes...")
10166 # first, create the missing data and meta devices
10167 for disk in new_disks:
10168 # unfortunately this is... not too nice
10169 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
10171 for child in disk.children:
10172 _CreateSingleBlockDev(self, snode, instance, child, info, True)
10173 # at this stage, all new LVs have been created, we can rename the
10175 feedback_fn("Renaming original volumes...")
10176 rename_list = [(o, n.children[0].logical_id)
10177 for (o, n) in zip(instance.disks, new_disks)]
10178 result = self.rpc.call_blockdev_rename(pnode, rename_list)
10179 result.Raise("Failed to rename original LVs")
10181 feedback_fn("Initializing DRBD devices...")
10182 # all child devices are in place, we can now create the DRBD devices
10183 for disk in new_disks:
10184 for node in [pnode, snode]:
10185 f_create = node == pnode
10186 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
10188 # at this point, the instance has been modified
10189 instance.disk_template = constants.DT_DRBD8
10190 instance.disks = new_disks
10191 self.cfg.Update(instance, feedback_fn)
10193 # disks are created, waiting for sync
10194 disk_abort = not _WaitForSync(self, instance)
10196 raise errors.OpExecError("There are some degraded disks for"
10197 " this instance, please cleanup manually")
10199 def _ConvertDrbdToPlain(self, feedback_fn):
10200 """Converts an instance from drbd to plain.
10203 instance = self.instance
10204 assert len(instance.secondary_nodes) == 1
10205 pnode = instance.primary_node
10206 snode = instance.secondary_nodes[0]
10207 feedback_fn("Converting template to plain")
10209 old_disks = instance.disks
10210 new_disks = [d.children[0] for d in old_disks]
10212 # copy over size and mode
10213 for parent, child in zip(old_disks, new_disks):
10214 child.size = parent.size
10215 child.mode = parent.mode
10217 # update instance structure
10218 instance.disks = new_disks
10219 instance.disk_template = constants.DT_PLAIN
10220 self.cfg.Update(instance, feedback_fn)
10222 feedback_fn("Removing volumes on the secondary node...")
10223 for disk in old_disks:
10224 self.cfg.SetDiskID(disk, snode)
10225 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
10227 self.LogWarning("Could not remove block device %s on node %s,"
10228 " continuing anyway: %s", disk.iv_name, snode, msg)
10230 feedback_fn("Removing unneeded volumes on the primary node...")
10231 for idx, disk in enumerate(old_disks):
10232 meta = disk.children[1]
10233 self.cfg.SetDiskID(meta, pnode)
10234 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
10236 self.LogWarning("Could not remove metadata for disk %d on node %s,"
10237 " continuing anyway: %s", idx, pnode, msg)
10239 def Exec(self, feedback_fn):
10240 """Modifies an instance.
10242 All parameters take effect only at the next restart of the instance.
10245 # Process here the warnings from CheckPrereq, as we don't have a
10246 # feedback_fn there.
10247 for warn in self.warn:
10248 feedback_fn("WARNING: %s" % warn)
10251 instance = self.instance
10253 for disk_op, disk_dict in self.op.disks:
10254 if disk_op == constants.DDM_REMOVE:
10255 # remove the last disk
10256 device = instance.disks.pop()
10257 device_idx = len(instance.disks)
10258 for node, disk in device.ComputeNodeTree(instance.primary_node):
10259 self.cfg.SetDiskID(disk, node)
10260 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
10262 self.LogWarning("Could not remove disk/%d on node %s: %s,"
10263 " continuing anyway", device_idx, node, msg)
10264 result.append(("disk/%d" % device_idx, "remove"))
10265 elif disk_op == constants.DDM_ADD:
10267 if instance.disk_template in (constants.DT_FILE,
10268 constants.DT_SHARED_FILE):
10269 file_driver, file_path = instance.disks[0].logical_id
10270 file_path = os.path.dirname(file_path)
10272 file_driver = file_path = None
10273 disk_idx_base = len(instance.disks)
10274 new_disk = _GenerateDiskTemplate(self,
10275 instance.disk_template,
10276 instance.name, instance.primary_node,
10277 instance.secondary_nodes,
10281 disk_idx_base, feedback_fn)[0]
10282 instance.disks.append(new_disk)
10283 info = _GetInstanceInfoText(instance)
10285 logging.info("Creating volume %s for instance %s",
10286 new_disk.iv_name, instance.name)
10287 # Note: this needs to be kept in sync with _CreateDisks
10289 for node in instance.all_nodes:
10290 f_create = node == instance.primary_node
10292 _CreateBlockDev(self, node, instance, new_disk,
10293 f_create, info, f_create)
10294 except errors.OpExecError, err:
10295 self.LogWarning("Failed to create volume %s (%s) on"
10297 new_disk.iv_name, new_disk, node, err)
10298 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
10299 (new_disk.size, new_disk.mode)))
10301 # change a given disk
10302 instance.disks[disk_op].mode = disk_dict[constants.IDISK_MODE]
10303 result.append(("disk.mode/%d" % disk_op,
10304 disk_dict[constants.IDISK_MODE]))
10306 if self.op.disk_template:
10307 r_shut = _ShutdownInstanceDisks(self, instance)
10309 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
10310 " proceed with disk template conversion")
10311 mode = (instance.disk_template, self.op.disk_template)
10313 self._DISK_CONVERSIONS[mode](self, feedback_fn)
10315 self.cfg.ReleaseDRBDMinors(instance.name)
10317 result.append(("disk_template", self.op.disk_template))
10320 for nic_op, nic_dict in self.op.nics:
10321 if nic_op == constants.DDM_REMOVE:
10322 # remove the last nic
10323 del instance.nics[-1]
10324 result.append(("nic.%d" % len(instance.nics), "remove"))
10325 elif nic_op == constants.DDM_ADD:
10326 # mac and bridge should be set, by now
10327 mac = nic_dict[constants.INIC_MAC]
10328 ip = nic_dict.get(constants.INIC_IP, None)
10329 nicparams = self.nic_pinst[constants.DDM_ADD]
10330 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
10331 instance.nics.append(new_nic)
10332 result.append(("nic.%d" % (len(instance.nics) - 1),
10333 "add:mac=%s,ip=%s,mode=%s,link=%s" %
10334 (new_nic.mac, new_nic.ip,
10335 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
10336 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
10339 for key in (constants.INIC_MAC, constants.INIC_IP):
10340 if key in nic_dict:
10341 setattr(instance.nics[nic_op], key, nic_dict[key])
10342 if nic_op in self.nic_pinst:
10343 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
10344 for key, val in nic_dict.iteritems():
10345 result.append(("nic.%s/%d" % (key, nic_op), val))
10348 if self.op.hvparams:
10349 instance.hvparams = self.hv_inst
10350 for key, val in self.op.hvparams.iteritems():
10351 result.append(("hv/%s" % key, val))
10354 if self.op.beparams:
10355 instance.beparams = self.be_inst
10356 for key, val in self.op.beparams.iteritems():
10357 result.append(("be/%s" % key, val))
10360 if self.op.os_name:
10361 instance.os = self.op.os_name
10364 if self.op.osparams:
10365 instance.osparams = self.os_inst
10366 for key, val in self.op.osparams.iteritems():
10367 result.append(("os/%s" % key, val))
10369 self.cfg.Update(instance, feedback_fn)
10373 _DISK_CONVERSIONS = {
10374 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
10375 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
10379 class LUBackupQuery(NoHooksLU):
10380 """Query the exports list
10385 def ExpandNames(self):
10386 self.needed_locks = {}
10387 self.share_locks[locking.LEVEL_NODE] = 1
10388 if not self.op.nodes:
10389 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10391 self.needed_locks[locking.LEVEL_NODE] = \
10392 _GetWantedNodes(self, self.op.nodes)
10394 def Exec(self, feedback_fn):
10395 """Compute the list of all the exported system images.
10398 @return: a dictionary with the structure node->(export-list)
10399 where export-list is a list of the instances exported on
10403 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
10404 rpcresult = self.rpc.call_export_list(self.nodes)
10406 for node in rpcresult:
10407 if rpcresult[node].fail_msg:
10408 result[node] = False
10410 result[node] = rpcresult[node].payload
10415 class LUBackupPrepare(NoHooksLU):
10416 """Prepares an instance for an export and returns useful information.
10421 def ExpandNames(self):
10422 self._ExpandAndLockInstance()
10424 def CheckPrereq(self):
10425 """Check prerequisites.
10428 instance_name = self.op.instance_name
10430 self.instance = self.cfg.GetInstanceInfo(instance_name)
10431 assert self.instance is not None, \
10432 "Cannot retrieve locked instance %s" % self.op.instance_name
10433 _CheckNodeOnline(self, self.instance.primary_node)
10435 self._cds = _GetClusterDomainSecret()
10437 def Exec(self, feedback_fn):
10438 """Prepares an instance for an export.
10441 instance = self.instance
10443 if self.op.mode == constants.EXPORT_MODE_REMOTE:
10444 salt = utils.GenerateSecret(8)
10446 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
10447 result = self.rpc.call_x509_cert_create(instance.primary_node,
10448 constants.RIE_CERT_VALIDITY)
10449 result.Raise("Can't create X509 key and certificate on %s" % result.node)
10451 (name, cert_pem) = result.payload
10453 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
10457 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
10458 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
10460 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
10466 class LUBackupExport(LogicalUnit):
10467 """Export an instance to an image in the cluster.
10470 HPATH = "instance-export"
10471 HTYPE = constants.HTYPE_INSTANCE
10474 def CheckArguments(self):
10475 """Check the arguments.
10478 self.x509_key_name = self.op.x509_key_name
10479 self.dest_x509_ca_pem = self.op.destination_x509_ca
10481 if self.op.mode == constants.EXPORT_MODE_REMOTE:
10482 if not self.x509_key_name:
10483 raise errors.OpPrereqError("Missing X509 key name for encryption",
10484 errors.ECODE_INVAL)
10486 if not self.dest_x509_ca_pem:
10487 raise errors.OpPrereqError("Missing destination X509 CA",
10488 errors.ECODE_INVAL)
10490 def ExpandNames(self):
10491 self._ExpandAndLockInstance()
10493 # Lock all nodes for local exports
10494 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10495 # FIXME: lock only instance primary and destination node
10497 # Sad but true, for now we have do lock all nodes, as we don't know where
10498 # the previous export might be, and in this LU we search for it and
10499 # remove it from its current node. In the future we could fix this by:
10500 # - making a tasklet to search (share-lock all), then create the
10501 # new one, then one to remove, after
10502 # - removing the removal operation altogether
10503 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10505 def DeclareLocks(self, level):
10506 """Last minute lock declaration."""
10507 # All nodes are locked anyway, so nothing to do here.
10509 def BuildHooksEnv(self):
10510 """Build hooks env.
10512 This will run on the master, primary node and target node.
10516 "EXPORT_MODE": self.op.mode,
10517 "EXPORT_NODE": self.op.target_node,
10518 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
10519 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
10520 # TODO: Generic function for boolean env variables
10521 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
10524 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
10528 def BuildHooksNodes(self):
10529 """Build hooks nodes.
10532 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
10534 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10535 nl.append(self.op.target_node)
10539 def CheckPrereq(self):
10540 """Check prerequisites.
10542 This checks that the instance and node names are valid.
10545 instance_name = self.op.instance_name
10547 self.instance = self.cfg.GetInstanceInfo(instance_name)
10548 assert self.instance is not None, \
10549 "Cannot retrieve locked instance %s" % self.op.instance_name
10550 _CheckNodeOnline(self, self.instance.primary_node)
10552 if (self.op.remove_instance and self.instance.admin_up and
10553 not self.op.shutdown):
10554 raise errors.OpPrereqError("Can not remove instance without shutting it"
10557 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10558 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
10559 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
10560 assert self.dst_node is not None
10562 _CheckNodeOnline(self, self.dst_node.name)
10563 _CheckNodeNotDrained(self, self.dst_node.name)
10566 self.dest_disk_info = None
10567 self.dest_x509_ca = None
10569 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
10570 self.dst_node = None
10572 if len(self.op.target_node) != len(self.instance.disks):
10573 raise errors.OpPrereqError(("Received destination information for %s"
10574 " disks, but instance %s has %s disks") %
10575 (len(self.op.target_node), instance_name,
10576 len(self.instance.disks)),
10577 errors.ECODE_INVAL)
10579 cds = _GetClusterDomainSecret()
10581 # Check X509 key name
10583 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
10584 except (TypeError, ValueError), err:
10585 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
10587 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
10588 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
10589 errors.ECODE_INVAL)
10591 # Load and verify CA
10593 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
10594 except OpenSSL.crypto.Error, err:
10595 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
10596 (err, ), errors.ECODE_INVAL)
10598 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
10599 if errcode is not None:
10600 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
10601 (msg, ), errors.ECODE_INVAL)
10603 self.dest_x509_ca = cert
10605 # Verify target information
10607 for idx, disk_data in enumerate(self.op.target_node):
10609 (host, port, magic) = \
10610 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
10611 except errors.GenericError, err:
10612 raise errors.OpPrereqError("Target info for disk %s: %s" %
10613 (idx, err), errors.ECODE_INVAL)
10615 disk_info.append((host, port, magic))
10617 assert len(disk_info) == len(self.op.target_node)
10618 self.dest_disk_info = disk_info
10621 raise errors.ProgrammerError("Unhandled export mode %r" %
10624 # instance disk type verification
10625 # TODO: Implement export support for file-based disks
10626 for disk in self.instance.disks:
10627 if disk.dev_type == constants.LD_FILE:
10628 raise errors.OpPrereqError("Export not supported for instances with"
10629 " file-based disks", errors.ECODE_INVAL)
10631 def _CleanupExports(self, feedback_fn):
10632 """Removes exports of current instance from all other nodes.
10634 If an instance in a cluster with nodes A..D was exported to node C, its
10635 exports will be removed from the nodes A, B and D.
10638 assert self.op.mode != constants.EXPORT_MODE_REMOTE
10640 nodelist = self.cfg.GetNodeList()
10641 nodelist.remove(self.dst_node.name)
10643 # on one-node clusters nodelist will be empty after the removal
10644 # if we proceed the backup would be removed because OpBackupQuery
10645 # substitutes an empty list with the full cluster node list.
10646 iname = self.instance.name
10648 feedback_fn("Removing old exports for instance %s" % iname)
10649 exportlist = self.rpc.call_export_list(nodelist)
10650 for node in exportlist:
10651 if exportlist[node].fail_msg:
10653 if iname in exportlist[node].payload:
10654 msg = self.rpc.call_export_remove(node, iname).fail_msg
10656 self.LogWarning("Could not remove older export for instance %s"
10657 " on node %s: %s", iname, node, msg)
10659 def Exec(self, feedback_fn):
10660 """Export an instance to an image in the cluster.
10663 assert self.op.mode in constants.EXPORT_MODES
10665 instance = self.instance
10666 src_node = instance.primary_node
10668 if self.op.shutdown:
10669 # shutdown the instance, but not the disks
10670 feedback_fn("Shutting down instance %s" % instance.name)
10671 result = self.rpc.call_instance_shutdown(src_node, instance,
10672 self.op.shutdown_timeout)
10673 # TODO: Maybe ignore failures if ignore_remove_failures is set
10674 result.Raise("Could not shutdown instance %s on"
10675 " node %s" % (instance.name, src_node))
10677 # set the disks ID correctly since call_instance_start needs the
10678 # correct drbd minor to create the symlinks
10679 for disk in instance.disks:
10680 self.cfg.SetDiskID(disk, src_node)
10682 activate_disks = (not instance.admin_up)
10685 # Activate the instance disks if we'exporting a stopped instance
10686 feedback_fn("Activating disks for %s" % instance.name)
10687 _StartInstanceDisks(self, instance, None)
10690 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
10693 helper.CreateSnapshots()
10695 if (self.op.shutdown and instance.admin_up and
10696 not self.op.remove_instance):
10697 assert not activate_disks
10698 feedback_fn("Starting instance %s" % instance.name)
10699 result = self.rpc.call_instance_start(src_node, instance, None, None)
10700 msg = result.fail_msg
10702 feedback_fn("Failed to start instance: %s" % msg)
10703 _ShutdownInstanceDisks(self, instance)
10704 raise errors.OpExecError("Could not start instance: %s" % msg)
10706 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10707 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
10708 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
10709 connect_timeout = constants.RIE_CONNECT_TIMEOUT
10710 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
10712 (key_name, _, _) = self.x509_key_name
10715 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
10718 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
10719 key_name, dest_ca_pem,
10724 # Check for backwards compatibility
10725 assert len(dresults) == len(instance.disks)
10726 assert compat.all(isinstance(i, bool) for i in dresults), \
10727 "Not all results are boolean: %r" % dresults
10731 feedback_fn("Deactivating disks for %s" % instance.name)
10732 _ShutdownInstanceDisks(self, instance)
10734 if not (compat.all(dresults) and fin_resu):
10737 failures.append("export finalization")
10738 if not compat.all(dresults):
10739 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
10741 failures.append("disk export: disk(s) %s" % fdsk)
10743 raise errors.OpExecError("Export failed, errors in %s" %
10744 utils.CommaJoin(failures))
10746 # At this point, the export was successful, we can cleanup/finish
10748 # Remove instance if requested
10749 if self.op.remove_instance:
10750 feedback_fn("Removing instance %s" % instance.name)
10751 _RemoveInstance(self, feedback_fn, instance,
10752 self.op.ignore_remove_failures)
10754 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10755 self._CleanupExports(feedback_fn)
10757 return fin_resu, dresults
10760 class LUBackupRemove(NoHooksLU):
10761 """Remove exports related to the named instance.
10766 def ExpandNames(self):
10767 self.needed_locks = {}
10768 # We need all nodes to be locked in order for RemoveExport to work, but we
10769 # don't need to lock the instance itself, as nothing will happen to it (and
10770 # we can remove exports also for a removed instance)
10771 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10773 def Exec(self, feedback_fn):
10774 """Remove any export.
10777 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
10778 # If the instance was not found we'll try with the name that was passed in.
10779 # This will only work if it was an FQDN, though.
10781 if not instance_name:
10783 instance_name = self.op.instance_name
10785 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
10786 exportlist = self.rpc.call_export_list(locked_nodes)
10788 for node in exportlist:
10789 msg = exportlist[node].fail_msg
10791 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
10793 if instance_name in exportlist[node].payload:
10795 result = self.rpc.call_export_remove(node, instance_name)
10796 msg = result.fail_msg
10798 logging.error("Could not remove export for instance %s"
10799 " on node %s: %s", instance_name, node, msg)
10801 if fqdn_warn and not found:
10802 feedback_fn("Export not found. If trying to remove an export belonging"
10803 " to a deleted instance please use its Fully Qualified"
10807 class LUGroupAdd(LogicalUnit):
10808 """Logical unit for creating node groups.
10811 HPATH = "group-add"
10812 HTYPE = constants.HTYPE_GROUP
10815 def ExpandNames(self):
10816 # We need the new group's UUID here so that we can create and acquire the
10817 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
10818 # that it should not check whether the UUID exists in the configuration.
10819 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
10820 self.needed_locks = {}
10821 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10823 def CheckPrereq(self):
10824 """Check prerequisites.
10826 This checks that the given group name is not an existing node group
10831 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10832 except errors.OpPrereqError:
10835 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
10836 " node group (UUID: %s)" %
10837 (self.op.group_name, existing_uuid),
10838 errors.ECODE_EXISTS)
10840 if self.op.ndparams:
10841 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10843 def BuildHooksEnv(self):
10844 """Build hooks env.
10848 "GROUP_NAME": self.op.group_name,
10851 def BuildHooksNodes(self):
10852 """Build hooks nodes.
10855 mn = self.cfg.GetMasterNode()
10856 return ([mn], [mn])
10858 def Exec(self, feedback_fn):
10859 """Add the node group to the cluster.
10862 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
10863 uuid=self.group_uuid,
10864 alloc_policy=self.op.alloc_policy,
10865 ndparams=self.op.ndparams)
10867 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
10868 del self.remove_locks[locking.LEVEL_NODEGROUP]
10871 class LUGroupAssignNodes(NoHooksLU):
10872 """Logical unit for assigning nodes to groups.
10877 def ExpandNames(self):
10878 # These raise errors.OpPrereqError on their own:
10879 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10880 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
10882 # We want to lock all the affected nodes and groups. We have readily
10883 # available the list of nodes, and the *destination* group. To gather the
10884 # list of "source" groups, we need to fetch node information.
10885 self.node_data = self.cfg.GetAllNodesInfo()
10886 affected_groups = set(self.node_data[node].group for node in self.op.nodes)
10887 affected_groups.add(self.group_uuid)
10889 self.needed_locks = {
10890 locking.LEVEL_NODEGROUP: list(affected_groups),
10891 locking.LEVEL_NODE: self.op.nodes,
10894 def CheckPrereq(self):
10895 """Check prerequisites.
10898 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10899 instance_data = self.cfg.GetAllInstancesInfo()
10901 if self.group is None:
10902 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10903 (self.op.group_name, self.group_uuid))
10905 (new_splits, previous_splits) = \
10906 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
10907 for node in self.op.nodes],
10908 self.node_data, instance_data)
10911 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
10913 if not self.op.force:
10914 raise errors.OpExecError("The following instances get split by this"
10915 " change and --force was not given: %s" %
10918 self.LogWarning("This operation will split the following instances: %s",
10921 if previous_splits:
10922 self.LogWarning("In addition, these already-split instances continue"
10923 " to be split across groups: %s",
10924 utils.CommaJoin(utils.NiceSort(previous_splits)))
10926 def Exec(self, feedback_fn):
10927 """Assign nodes to a new group.
10930 for node in self.op.nodes:
10931 self.node_data[node].group = self.group_uuid
10933 self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
10936 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
10937 """Check for split instances after a node assignment.
10939 This method considers a series of node assignments as an atomic operation,
10940 and returns information about split instances after applying the set of
10943 In particular, it returns information about newly split instances, and
10944 instances that were already split, and remain so after the change.
10946 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
10949 @type changes: list of (node_name, new_group_uuid) pairs.
10950 @param changes: list of node assignments to consider.
10951 @param node_data: a dict with data for all nodes
10952 @param instance_data: a dict with all instances to consider
10953 @rtype: a two-tuple
10954 @return: a list of instances that were previously okay and result split as a
10955 consequence of this change, and a list of instances that were previously
10956 split and this change does not fix.
10959 changed_nodes = dict((node, group) for node, group in changes
10960 if node_data[node].group != group)
10962 all_split_instances = set()
10963 previously_split_instances = set()
10965 def InstanceNodes(instance):
10966 return [instance.primary_node] + list(instance.secondary_nodes)
10968 for inst in instance_data.values():
10969 if inst.disk_template not in constants.DTS_INT_MIRROR:
10972 instance_nodes = InstanceNodes(inst)
10974 if len(set(node_data[node].group for node in instance_nodes)) > 1:
10975 previously_split_instances.add(inst.name)
10977 if len(set(changed_nodes.get(node, node_data[node].group)
10978 for node in instance_nodes)) > 1:
10979 all_split_instances.add(inst.name)
10981 return (list(all_split_instances - previously_split_instances),
10982 list(previously_split_instances & all_split_instances))
10985 class _GroupQuery(_QueryBase):
10986 FIELDS = query.GROUP_FIELDS
10988 def ExpandNames(self, lu):
10989 lu.needed_locks = {}
10991 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
10992 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
10995 self.wanted = [name_to_uuid[name]
10996 for name in utils.NiceSort(name_to_uuid.keys())]
10998 # Accept names to be either names or UUIDs.
11001 all_uuid = frozenset(self._all_groups.keys())
11003 for name in self.names:
11004 if name in all_uuid:
11005 self.wanted.append(name)
11006 elif name in name_to_uuid:
11007 self.wanted.append(name_to_uuid[name])
11009 missing.append(name)
11012 raise errors.OpPrereqError("Some groups do not exist: %s" %
11013 utils.CommaJoin(missing),
11014 errors.ECODE_NOENT)
11016 def DeclareLocks(self, lu, level):
11019 def _GetQueryData(self, lu):
11020 """Computes the list of node groups and their attributes.
11023 do_nodes = query.GQ_NODE in self.requested_data
11024 do_instances = query.GQ_INST in self.requested_data
11026 group_to_nodes = None
11027 group_to_instances = None
11029 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
11030 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
11031 # latter GetAllInstancesInfo() is not enough, for we have to go through
11032 # instance->node. Hence, we will need to process nodes even if we only need
11033 # instance information.
11034 if do_nodes or do_instances:
11035 all_nodes = lu.cfg.GetAllNodesInfo()
11036 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
11039 for node in all_nodes.values():
11040 if node.group in group_to_nodes:
11041 group_to_nodes[node.group].append(node.name)
11042 node_to_group[node.name] = node.group
11045 all_instances = lu.cfg.GetAllInstancesInfo()
11046 group_to_instances = dict((uuid, []) for uuid in self.wanted)
11048 for instance in all_instances.values():
11049 node = instance.primary_node
11050 if node in node_to_group:
11051 group_to_instances[node_to_group[node]].append(instance.name)
11054 # Do not pass on node information if it was not requested.
11055 group_to_nodes = None
11057 return query.GroupQueryData([self._all_groups[uuid]
11058 for uuid in self.wanted],
11059 group_to_nodes, group_to_instances)
11062 class LUGroupQuery(NoHooksLU):
11063 """Logical unit for querying node groups.
11068 def CheckArguments(self):
11069 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
11070 self.op.output_fields, False)
11072 def ExpandNames(self):
11073 self.gq.ExpandNames(self)
11075 def Exec(self, feedback_fn):
11076 return self.gq.OldStyleQuery(self)
11079 class LUGroupSetParams(LogicalUnit):
11080 """Modifies the parameters of a node group.
11083 HPATH = "group-modify"
11084 HTYPE = constants.HTYPE_GROUP
11087 def CheckArguments(self):
11090 self.op.alloc_policy,
11093 if all_changes.count(None) == len(all_changes):
11094 raise errors.OpPrereqError("Please pass at least one modification",
11095 errors.ECODE_INVAL)
11097 def ExpandNames(self):
11098 # This raises errors.OpPrereqError on its own:
11099 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11101 self.needed_locks = {
11102 locking.LEVEL_NODEGROUP: [self.group_uuid],
11105 def CheckPrereq(self):
11106 """Check prerequisites.
11109 self.group = self.cfg.GetNodeGroup(self.group_uuid)
11111 if self.group is None:
11112 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
11113 (self.op.group_name, self.group_uuid))
11115 if self.op.ndparams:
11116 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
11117 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
11118 self.new_ndparams = new_ndparams
11120 def BuildHooksEnv(self):
11121 """Build hooks env.
11125 "GROUP_NAME": self.op.group_name,
11126 "NEW_ALLOC_POLICY": self.op.alloc_policy,
11129 def BuildHooksNodes(self):
11130 """Build hooks nodes.
11133 mn = self.cfg.GetMasterNode()
11134 return ([mn], [mn])
11136 def Exec(self, feedback_fn):
11137 """Modifies the node group.
11142 if self.op.ndparams:
11143 self.group.ndparams = self.new_ndparams
11144 result.append(("ndparams", str(self.group.ndparams)))
11146 if self.op.alloc_policy:
11147 self.group.alloc_policy = self.op.alloc_policy
11149 self.cfg.Update(self.group, feedback_fn)
11154 class LUGroupRemove(LogicalUnit):
11155 HPATH = "group-remove"
11156 HTYPE = constants.HTYPE_GROUP
11159 def ExpandNames(self):
11160 # This will raises errors.OpPrereqError on its own:
11161 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11162 self.needed_locks = {
11163 locking.LEVEL_NODEGROUP: [self.group_uuid],
11166 def CheckPrereq(self):
11167 """Check prerequisites.
11169 This checks that the given group name exists as a node group, that is
11170 empty (i.e., contains no nodes), and that is not the last group of the
11174 # Verify that the group is empty.
11175 group_nodes = [node.name
11176 for node in self.cfg.GetAllNodesInfo().values()
11177 if node.group == self.group_uuid]
11180 raise errors.OpPrereqError("Group '%s' not empty, has the following"
11182 (self.op.group_name,
11183 utils.CommaJoin(utils.NiceSort(group_nodes))),
11184 errors.ECODE_STATE)
11186 # Verify the cluster would not be left group-less.
11187 if len(self.cfg.GetNodeGroupList()) == 1:
11188 raise errors.OpPrereqError("Group '%s' is the only group,"
11189 " cannot be removed" %
11190 self.op.group_name,
11191 errors.ECODE_STATE)
11193 def BuildHooksEnv(self):
11194 """Build hooks env.
11198 "GROUP_NAME": self.op.group_name,
11201 def BuildHooksNodes(self):
11202 """Build hooks nodes.
11205 mn = self.cfg.GetMasterNode()
11206 return ([mn], [mn])
11208 def Exec(self, feedback_fn):
11209 """Remove the node group.
11213 self.cfg.RemoveNodeGroup(self.group_uuid)
11214 except errors.ConfigurationError:
11215 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
11216 (self.op.group_name, self.group_uuid))
11218 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
11221 class LUGroupRename(LogicalUnit):
11222 HPATH = "group-rename"
11223 HTYPE = constants.HTYPE_GROUP
11226 def ExpandNames(self):
11227 # This raises errors.OpPrereqError on its own:
11228 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11230 self.needed_locks = {
11231 locking.LEVEL_NODEGROUP: [self.group_uuid],
11234 def CheckPrereq(self):
11235 """Check prerequisites.
11237 Ensures requested new name is not yet used.
11241 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
11242 except errors.OpPrereqError:
11245 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
11246 " node group (UUID: %s)" %
11247 (self.op.new_name, new_name_uuid),
11248 errors.ECODE_EXISTS)
11250 def BuildHooksEnv(self):
11251 """Build hooks env.
11255 "OLD_NAME": self.op.group_name,
11256 "NEW_NAME": self.op.new_name,
11259 def BuildHooksNodes(self):
11260 """Build hooks nodes.
11263 mn = self.cfg.GetMasterNode()
11265 all_nodes = self.cfg.GetAllNodesInfo()
11266 all_nodes.pop(mn, None)
11269 run_nodes.extend(node.name for node in all_nodes.values()
11270 if node.group == self.group_uuid)
11272 return (run_nodes, run_nodes)
11274 def Exec(self, feedback_fn):
11275 """Rename the node group.
11278 group = self.cfg.GetNodeGroup(self.group_uuid)
11281 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
11282 (self.op.group_name, self.group_uuid))
11284 group.name = self.op.new_name
11285 self.cfg.Update(group, feedback_fn)
11287 return self.op.new_name
11290 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
11291 """Generic tags LU.
11293 This is an abstract class which is the parent of all the other tags LUs.
11296 def ExpandNames(self):
11297 self.group_uuid = None
11298 self.needed_locks = {}
11299 if self.op.kind == constants.TAG_NODE:
11300 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
11301 self.needed_locks[locking.LEVEL_NODE] = self.op.name
11302 elif self.op.kind == constants.TAG_INSTANCE:
11303 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
11304 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
11305 elif self.op.kind == constants.TAG_NODEGROUP:
11306 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
11308 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
11309 # not possible to acquire the BGL based on opcode parameters)
11311 def CheckPrereq(self):
11312 """Check prerequisites.
11315 if self.op.kind == constants.TAG_CLUSTER:
11316 self.target = self.cfg.GetClusterInfo()
11317 elif self.op.kind == constants.TAG_NODE:
11318 self.target = self.cfg.GetNodeInfo(self.op.name)
11319 elif self.op.kind == constants.TAG_INSTANCE:
11320 self.target = self.cfg.GetInstanceInfo(self.op.name)
11321 elif self.op.kind == constants.TAG_NODEGROUP:
11322 self.target = self.cfg.GetNodeGroup(self.group_uuid)
11324 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
11325 str(self.op.kind), errors.ECODE_INVAL)
11328 class LUTagsGet(TagsLU):
11329 """Returns the tags of a given object.
11334 def ExpandNames(self):
11335 TagsLU.ExpandNames(self)
11337 # Share locks as this is only a read operation
11338 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
11340 def Exec(self, feedback_fn):
11341 """Returns the tag list.
11344 return list(self.target.GetTags())
11347 class LUTagsSearch(NoHooksLU):
11348 """Searches the tags for a given pattern.
11353 def ExpandNames(self):
11354 self.needed_locks = {}
11356 def CheckPrereq(self):
11357 """Check prerequisites.
11359 This checks the pattern passed for validity by compiling it.
11363 self.re = re.compile(self.op.pattern)
11364 except re.error, err:
11365 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
11366 (self.op.pattern, err), errors.ECODE_INVAL)
11368 def Exec(self, feedback_fn):
11369 """Returns the tag list.
11373 tgts = [("/cluster", cfg.GetClusterInfo())]
11374 ilist = cfg.GetAllInstancesInfo().values()
11375 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
11376 nlist = cfg.GetAllNodesInfo().values()
11377 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
11378 tgts.extend(("/nodegroup/%s" % n.name, n)
11379 for n in cfg.GetAllNodeGroupsInfo().values())
11381 for path, target in tgts:
11382 for tag in target.GetTags():
11383 if self.re.search(tag):
11384 results.append((path, tag))
11388 class LUTagsSet(TagsLU):
11389 """Sets a tag on a given object.
11394 def CheckPrereq(self):
11395 """Check prerequisites.
11397 This checks the type and length of the tag name and value.
11400 TagsLU.CheckPrereq(self)
11401 for tag in self.op.tags:
11402 objects.TaggableObject.ValidateTag(tag)
11404 def Exec(self, feedback_fn):
11409 for tag in self.op.tags:
11410 self.target.AddTag(tag)
11411 except errors.TagError, err:
11412 raise errors.OpExecError("Error while setting tag: %s" % str(err))
11413 self.cfg.Update(self.target, feedback_fn)
11416 class LUTagsDel(TagsLU):
11417 """Delete a list of tags from a given object.
11422 def CheckPrereq(self):
11423 """Check prerequisites.
11425 This checks that we have the given tag.
11428 TagsLU.CheckPrereq(self)
11429 for tag in self.op.tags:
11430 objects.TaggableObject.ValidateTag(tag)
11431 del_tags = frozenset(self.op.tags)
11432 cur_tags = self.target.GetTags()
11434 diff_tags = del_tags - cur_tags
11436 diff_names = ("'%s'" % i for i in sorted(diff_tags))
11437 raise errors.OpPrereqError("Tag(s) %s not found" %
11438 (utils.CommaJoin(diff_names), ),
11439 errors.ECODE_NOENT)
11441 def Exec(self, feedback_fn):
11442 """Remove the tag from the object.
11445 for tag in self.op.tags:
11446 self.target.RemoveTag(tag)
11447 self.cfg.Update(self.target, feedback_fn)
11450 class LUTestDelay(NoHooksLU):
11451 """Sleep for a specified amount of time.
11453 This LU sleeps on the master and/or nodes for a specified amount of
11459 def ExpandNames(self):
11460 """Expand names and set required locks.
11462 This expands the node list, if any.
11465 self.needed_locks = {}
11466 if self.op.on_nodes:
11467 # _GetWantedNodes can be used here, but is not always appropriate to use
11468 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
11469 # more information.
11470 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
11471 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
11473 def _TestDelay(self):
11474 """Do the actual sleep.
11477 if self.op.on_master:
11478 if not utils.TestDelay(self.op.duration):
11479 raise errors.OpExecError("Error during master delay test")
11480 if self.op.on_nodes:
11481 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
11482 for node, node_result in result.items():
11483 node_result.Raise("Failure during rpc call to node %s" % node)
11485 def Exec(self, feedback_fn):
11486 """Execute the test delay opcode, with the wanted repetitions.
11489 if self.op.repeat == 0:
11492 top_value = self.op.repeat - 1
11493 for i in range(self.op.repeat):
11494 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
11498 class LUTestJqueue(NoHooksLU):
11499 """Utility LU to test some aspects of the job queue.
11504 # Must be lower than default timeout for WaitForJobChange to see whether it
11505 # notices changed jobs
11506 _CLIENT_CONNECT_TIMEOUT = 20.0
11507 _CLIENT_CONFIRM_TIMEOUT = 60.0
11510 def _NotifyUsingSocket(cls, cb, errcls):
11511 """Opens a Unix socket and waits for another program to connect.
11514 @param cb: Callback to send socket name to client
11515 @type errcls: class
11516 @param errcls: Exception class to use for errors
11519 # Using a temporary directory as there's no easy way to create temporary
11520 # sockets without writing a custom loop around tempfile.mktemp and
11522 tmpdir = tempfile.mkdtemp()
11524 tmpsock = utils.PathJoin(tmpdir, "sock")
11526 logging.debug("Creating temporary socket at %s", tmpsock)
11527 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
11532 # Send details to client
11535 # Wait for client to connect before continuing
11536 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
11538 (conn, _) = sock.accept()
11539 except socket.error, err:
11540 raise errcls("Client didn't connect in time (%s)" % err)
11544 # Remove as soon as client is connected
11545 shutil.rmtree(tmpdir)
11547 # Wait for client to close
11550 # pylint: disable-msg=E1101
11551 # Instance of '_socketobject' has no ... member
11552 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
11554 except socket.error, err:
11555 raise errcls("Client failed to confirm notification (%s)" % err)
11559 def _SendNotification(self, test, arg, sockname):
11560 """Sends a notification to the client.
11563 @param test: Test name
11564 @param arg: Test argument (depends on test)
11565 @type sockname: string
11566 @param sockname: Socket path
11569 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
11571 def _Notify(self, prereq, test, arg):
11572 """Notifies the client of a test.
11575 @param prereq: Whether this is a prereq-phase test
11577 @param test: Test name
11578 @param arg: Test argument (depends on test)
11582 errcls = errors.OpPrereqError
11584 errcls = errors.OpExecError
11586 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
11590 def CheckArguments(self):
11591 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
11592 self.expandnames_calls = 0
11594 def ExpandNames(self):
11595 checkargs_calls = getattr(self, "checkargs_calls", 0)
11596 if checkargs_calls < 1:
11597 raise errors.ProgrammerError("CheckArguments was not called")
11599 self.expandnames_calls += 1
11601 if self.op.notify_waitlock:
11602 self._Notify(True, constants.JQT_EXPANDNAMES, None)
11604 self.LogInfo("Expanding names")
11606 # Get lock on master node (just to get a lock, not for a particular reason)
11607 self.needed_locks = {
11608 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
11611 def Exec(self, feedback_fn):
11612 if self.expandnames_calls < 1:
11613 raise errors.ProgrammerError("ExpandNames was not called")
11615 if self.op.notify_exec:
11616 self._Notify(False, constants.JQT_EXEC, None)
11618 self.LogInfo("Executing")
11620 if self.op.log_messages:
11621 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
11622 for idx, msg in enumerate(self.op.log_messages):
11623 self.LogInfo("Sending log message %s", idx + 1)
11624 feedback_fn(constants.JQT_MSGPREFIX + msg)
11625 # Report how many test messages have been sent
11626 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
11629 raise errors.OpExecError("Opcode failure was requested")
11634 class IAllocator(object):
11635 """IAllocator framework.
11637 An IAllocator instance has three sets of attributes:
11638 - cfg that is needed to query the cluster
11639 - input data (all members of the _KEYS class attribute are required)
11640 - four buffer attributes (in|out_data|text), that represent the
11641 input (to the external script) in text and data structure format,
11642 and the output from it, again in two formats
11643 - the result variables from the script (success, info, nodes) for
11647 # pylint: disable-msg=R0902
11648 # lots of instance attributes
11650 "name", "mem_size", "disks", "disk_template",
11651 "os", "tags", "nics", "vcpus", "hypervisor",
11654 "name", "relocate_from",
11660 def __init__(self, cfg, rpc, mode, **kwargs):
11663 # init buffer variables
11664 self.in_text = self.out_text = self.in_data = self.out_data = None
11665 # init all input fields so that pylint is happy
11667 self.mem_size = self.disks = self.disk_template = None
11668 self.os = self.tags = self.nics = self.vcpus = None
11669 self.hypervisor = None
11670 self.relocate_from = None
11672 self.evac_nodes = None
11674 self.required_nodes = None
11675 # init result fields
11676 self.success = self.info = self.result = None
11677 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
11678 keyset = self._ALLO_KEYS
11679 fn = self._AddNewInstance
11680 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
11681 keyset = self._RELO_KEYS
11682 fn = self._AddRelocateInstance
11683 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
11684 keyset = self._EVAC_KEYS
11685 fn = self._AddEvacuateNodes
11687 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
11688 " IAllocator" % self.mode)
11690 if key not in keyset:
11691 raise errors.ProgrammerError("Invalid input parameter '%s' to"
11692 " IAllocator" % key)
11693 setattr(self, key, kwargs[key])
11696 if key not in kwargs:
11697 raise errors.ProgrammerError("Missing input parameter '%s' to"
11698 " IAllocator" % key)
11699 self._BuildInputData(fn)
11701 def _ComputeClusterData(self):
11702 """Compute the generic allocator input data.
11704 This is the data that is independent of the actual operation.
11708 cluster_info = cfg.GetClusterInfo()
11711 "version": constants.IALLOCATOR_VERSION,
11712 "cluster_name": cfg.GetClusterName(),
11713 "cluster_tags": list(cluster_info.GetTags()),
11714 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
11715 # we don't have job IDs
11717 ninfo = cfg.GetAllNodesInfo()
11718 iinfo = cfg.GetAllInstancesInfo().values()
11719 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
11722 node_list = [n.name for n in ninfo.values() if n.vm_capable]
11724 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
11725 hypervisor_name = self.hypervisor
11726 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
11727 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
11728 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
11729 hypervisor_name = cluster_info.enabled_hypervisors[0]
11731 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
11734 self.rpc.call_all_instances_info(node_list,
11735 cluster_info.enabled_hypervisors)
11737 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
11739 config_ndata = self._ComputeBasicNodeData(ninfo)
11740 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
11741 i_list, config_ndata)
11742 assert len(data["nodes"]) == len(ninfo), \
11743 "Incomplete node data computed"
11745 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
11747 self.in_data = data
11750 def _ComputeNodeGroupData(cfg):
11751 """Compute node groups data.
11755 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
11757 "name": gdata.name,
11758 "alloc_policy": gdata.alloc_policy,
11763 def _ComputeBasicNodeData(node_cfg):
11764 """Compute global node data.
11767 @returns: a dict of name: (node dict, node config)
11771 for ninfo in node_cfg.values():
11772 # fill in static (config-based) values
11774 "tags": list(ninfo.GetTags()),
11775 "primary_ip": ninfo.primary_ip,
11776 "secondary_ip": ninfo.secondary_ip,
11777 "offline": ninfo.offline,
11778 "drained": ninfo.drained,
11779 "master_candidate": ninfo.master_candidate,
11780 "group": ninfo.group,
11781 "master_capable": ninfo.master_capable,
11782 "vm_capable": ninfo.vm_capable,
11785 node_results[ninfo.name] = pnr
11787 return node_results
11790 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
11792 """Compute global node data.
11794 @param node_results: the basic node structures as filled from the config
11797 # make a copy of the current dict
11798 node_results = dict(node_results)
11799 for nname, nresult in node_data.items():
11800 assert nname in node_results, "Missing basic data for node %s" % nname
11801 ninfo = node_cfg[nname]
11803 if not (ninfo.offline or ninfo.drained):
11804 nresult.Raise("Can't get data for node %s" % nname)
11805 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
11807 remote_info = nresult.payload
11809 for attr in ['memory_total', 'memory_free', 'memory_dom0',
11810 'vg_size', 'vg_free', 'cpu_total']:
11811 if attr not in remote_info:
11812 raise errors.OpExecError("Node '%s' didn't return attribute"
11813 " '%s'" % (nname, attr))
11814 if not isinstance(remote_info[attr], int):
11815 raise errors.OpExecError("Node '%s' returned invalid value"
11817 (nname, attr, remote_info[attr]))
11818 # compute memory used by primary instances
11819 i_p_mem = i_p_up_mem = 0
11820 for iinfo, beinfo in i_list:
11821 if iinfo.primary_node == nname:
11822 i_p_mem += beinfo[constants.BE_MEMORY]
11823 if iinfo.name not in node_iinfo[nname].payload:
11826 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
11827 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
11828 remote_info['memory_free'] -= max(0, i_mem_diff)
11831 i_p_up_mem += beinfo[constants.BE_MEMORY]
11833 # compute memory used by instances
11835 "total_memory": remote_info['memory_total'],
11836 "reserved_memory": remote_info['memory_dom0'],
11837 "free_memory": remote_info['memory_free'],
11838 "total_disk": remote_info['vg_size'],
11839 "free_disk": remote_info['vg_free'],
11840 "total_cpus": remote_info['cpu_total'],
11841 "i_pri_memory": i_p_mem,
11842 "i_pri_up_memory": i_p_up_mem,
11844 pnr_dyn.update(node_results[nname])
11845 node_results[nname] = pnr_dyn
11847 return node_results
11850 def _ComputeInstanceData(cluster_info, i_list):
11851 """Compute global instance data.
11855 for iinfo, beinfo in i_list:
11857 for nic in iinfo.nics:
11858 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
11859 nic_dict = {"mac": nic.mac,
11861 "mode": filled_params[constants.NIC_MODE],
11862 "link": filled_params[constants.NIC_LINK],
11864 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
11865 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
11866 nic_data.append(nic_dict)
11868 "tags": list(iinfo.GetTags()),
11869 "admin_up": iinfo.admin_up,
11870 "vcpus": beinfo[constants.BE_VCPUS],
11871 "memory": beinfo[constants.BE_MEMORY],
11873 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
11875 "disks": [{constants.IDISK_SIZE: dsk.size,
11876 constants.IDISK_MODE: dsk.mode}
11877 for dsk in iinfo.disks],
11878 "disk_template": iinfo.disk_template,
11879 "hypervisor": iinfo.hypervisor,
11881 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
11883 instance_data[iinfo.name] = pir
11885 return instance_data
11887 def _AddNewInstance(self):
11888 """Add new instance data to allocator structure.
11890 This in combination with _AllocatorGetClusterData will create the
11891 correct structure needed as input for the allocator.
11893 The checks for the completeness of the opcode must have already been
11897 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
11899 if self.disk_template in constants.DTS_INT_MIRROR:
11900 self.required_nodes = 2
11902 self.required_nodes = 1
11905 "disk_template": self.disk_template,
11908 "vcpus": self.vcpus,
11909 "memory": self.mem_size,
11910 "disks": self.disks,
11911 "disk_space_total": disk_space,
11913 "required_nodes": self.required_nodes,
11917 def _AddRelocateInstance(self):
11918 """Add relocate instance data to allocator structure.
11920 This in combination with _IAllocatorGetClusterData will create the
11921 correct structure needed as input for the allocator.
11923 The checks for the completeness of the opcode must have already been
11927 instance = self.cfg.GetInstanceInfo(self.name)
11928 if instance is None:
11929 raise errors.ProgrammerError("Unknown instance '%s' passed to"
11930 " IAllocator" % self.name)
11932 if instance.disk_template not in constants.DTS_MIRRORED:
11933 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
11934 errors.ECODE_INVAL)
11936 if instance.disk_template in constants.DTS_INT_MIRROR and \
11937 len(instance.secondary_nodes) != 1:
11938 raise errors.OpPrereqError("Instance has not exactly one secondary node",
11939 errors.ECODE_STATE)
11941 self.required_nodes = 1
11942 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
11943 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
11947 "disk_space_total": disk_space,
11948 "required_nodes": self.required_nodes,
11949 "relocate_from": self.relocate_from,
11953 def _AddEvacuateNodes(self):
11954 """Add evacuate nodes data to allocator structure.
11958 "evac_nodes": self.evac_nodes
11962 def _BuildInputData(self, fn):
11963 """Build input data structures.
11966 self._ComputeClusterData()
11969 request["type"] = self.mode
11970 self.in_data["request"] = request
11972 self.in_text = serializer.Dump(self.in_data)
11974 def Run(self, name, validate=True, call_fn=None):
11975 """Run an instance allocator and return the results.
11978 if call_fn is None:
11979 call_fn = self.rpc.call_iallocator_runner
11981 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
11982 result.Raise("Failure while running the iallocator script")
11984 self.out_text = result.payload
11986 self._ValidateResult()
11988 def _ValidateResult(self):
11989 """Process the allocator results.
11991 This will process and if successful save the result in
11992 self.out_data and the other parameters.
11996 rdict = serializer.Load(self.out_text)
11997 except Exception, err:
11998 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
12000 if not isinstance(rdict, dict):
12001 raise errors.OpExecError("Can't parse iallocator results: not a dict")
12003 # TODO: remove backwards compatiblity in later versions
12004 if "nodes" in rdict and "result" not in rdict:
12005 rdict["result"] = rdict["nodes"]
12008 for key in "success", "info", "result":
12009 if key not in rdict:
12010 raise errors.OpExecError("Can't parse iallocator results:"
12011 " missing key '%s'" % key)
12012 setattr(self, key, rdict[key])
12014 if not isinstance(rdict["result"], list):
12015 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
12018 if self.mode == constants.IALLOCATOR_MODE_RELOC:
12019 assert self.relocate_from is not None
12020 assert self.required_nodes == 1
12022 node2group = dict((name, ndata["group"])
12023 for (name, ndata) in self.in_data["nodes"].items())
12025 fn = compat.partial(self._NodesToGroups, node2group,
12026 self.in_data["nodegroups"])
12028 request_groups = fn(self.relocate_from)
12029 result_groups = fn(rdict["result"])
12031 if result_groups != request_groups:
12032 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
12033 " differ from original groups (%s)" %
12034 (utils.CommaJoin(result_groups),
12035 utils.CommaJoin(request_groups)))
12037 self.out_data = rdict
12040 def _NodesToGroups(node2group, groups, nodes):
12041 """Returns a list of unique group names for a list of nodes.
12043 @type node2group: dict
12044 @param node2group: Map from node name to group UUID
12046 @param groups: Group information
12048 @param nodes: Node names
12055 group_uuid = node2group[node]
12057 # Ignore unknown node
12061 group = groups[group_uuid]
12063 # Can't find group, let's use UUID
12064 group_name = group_uuid
12066 group_name = group["name"]
12068 result.add(group_name)
12070 return sorted(result)
12073 class LUTestAllocator(NoHooksLU):
12074 """Run allocator tests.
12076 This LU runs the allocator tests
12079 def CheckPrereq(self):
12080 """Check prerequisites.
12082 This checks the opcode parameters depending on the director and mode test.
12085 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
12086 for attr in ["mem_size", "disks", "disk_template",
12087 "os", "tags", "nics", "vcpus"]:
12088 if not hasattr(self.op, attr):
12089 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
12090 attr, errors.ECODE_INVAL)
12091 iname = self.cfg.ExpandInstanceName(self.op.name)
12092 if iname is not None:
12093 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
12094 iname, errors.ECODE_EXISTS)
12095 if not isinstance(self.op.nics, list):
12096 raise errors.OpPrereqError("Invalid parameter 'nics'",
12097 errors.ECODE_INVAL)
12098 if not isinstance(self.op.disks, list):
12099 raise errors.OpPrereqError("Invalid parameter 'disks'",
12100 errors.ECODE_INVAL)
12101 for row in self.op.disks:
12102 if (not isinstance(row, dict) or
12103 "size" not in row or
12104 not isinstance(row["size"], int) or
12105 "mode" not in row or
12106 row["mode"] not in ['r', 'w']):
12107 raise errors.OpPrereqError("Invalid contents of the 'disks'"
12108 " parameter", errors.ECODE_INVAL)
12109 if self.op.hypervisor is None:
12110 self.op.hypervisor = self.cfg.GetHypervisorType()
12111 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
12112 fname = _ExpandInstanceName(self.cfg, self.op.name)
12113 self.op.name = fname
12114 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
12115 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
12116 if not hasattr(self.op, "evac_nodes"):
12117 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
12118 " opcode input", errors.ECODE_INVAL)
12120 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
12121 self.op.mode, errors.ECODE_INVAL)
12123 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
12124 if self.op.allocator is None:
12125 raise errors.OpPrereqError("Missing allocator name",
12126 errors.ECODE_INVAL)
12127 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
12128 raise errors.OpPrereqError("Wrong allocator test '%s'" %
12129 self.op.direction, errors.ECODE_INVAL)
12131 def Exec(self, feedback_fn):
12132 """Run the allocator test.
12135 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
12136 ial = IAllocator(self.cfg, self.rpc,
12139 mem_size=self.op.mem_size,
12140 disks=self.op.disks,
12141 disk_template=self.op.disk_template,
12145 vcpus=self.op.vcpus,
12146 hypervisor=self.op.hypervisor,
12148 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
12149 ial = IAllocator(self.cfg, self.rpc,
12152 relocate_from=list(self.relocate_from),
12154 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
12155 ial = IAllocator(self.cfg, self.rpc,
12157 evac_nodes=self.op.evac_nodes)
12159 raise errors.ProgrammerError("Uncatched mode %s in"
12160 " LUTestAllocator.Exec", self.op.mode)
12162 if self.op.direction == constants.IALLOCATOR_DIR_IN:
12163 result = ial.in_text
12165 ial.Run(self.op.allocator, validate=False)
12166 result = ial.out_text
12170 #: Query type implementations
12172 constants.QR_INSTANCE: _InstanceQuery,
12173 constants.QR_NODE: _NodeQuery,
12174 constants.QR_GROUP: _GroupQuery,
12175 constants.QR_OS: _OsQuery,
12178 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
12181 def _GetQueryImplementation(name):
12182 """Returns the implemtnation for a query type.
12184 @param name: Query type, must be one of L{constants.QR_VIA_OP}
12188 return _QUERY_IMPL[name]
12190 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
12191 errors.ECODE_INVAL)