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.glm = context.glm
133 self.context = context
135 # Dicts used to declare locking needs to mcpu
136 self.needed_locks = None
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.glm.list_owned(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.glm.list_owned(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.glm.is_owned(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.glm.list_owned(level):
660 if should_release(name):
665 assert len(lu.glm.list_owned(level)) == (len(retain) + len(release))
667 # Release just some locks
668 lu.glm.release(level, names=release)
670 assert frozenset(lu.glm.list_owned(level)) == frozenset(retain)
673 lu.glm.release(level)
675 assert not lu.glm.is_owned(level), "No locks should be owned"
678 def _RunPostHook(lu, node_name):
679 """Runs the post-hook for an opcode on a single node.
682 hm = lu.proc.hmclass(lu.rpc.call_hooks_runner, lu)
684 hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
686 # pylint: disable-msg=W0702
687 lu.LogWarning("Errors occurred running hooks on %s" % node_name)
690 def _CheckOutputFields(static, dynamic, selected):
691 """Checks whether all selected fields are valid.
693 @type static: L{utils.FieldSet}
694 @param static: static fields set
695 @type dynamic: L{utils.FieldSet}
696 @param dynamic: dynamic fields set
703 delta = f.NonMatching(selected)
705 raise errors.OpPrereqError("Unknown output fields selected: %s"
706 % ",".join(delta), errors.ECODE_INVAL)
709 def _CheckGlobalHvParams(params):
710 """Validates that given hypervisor params are not global ones.
712 This will ensure that instances don't get customised versions of
716 used_globals = constants.HVC_GLOBALS.intersection(params)
718 msg = ("The following hypervisor parameters are global and cannot"
719 " be customized at instance level, please modify them at"
720 " cluster level: %s" % utils.CommaJoin(used_globals))
721 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
724 def _CheckNodeOnline(lu, node, msg=None):
725 """Ensure that a given node is online.
727 @param lu: the LU on behalf of which we make the check
728 @param node: the node to check
729 @param msg: if passed, should be a message to replace the default one
730 @raise errors.OpPrereqError: if the node is offline
734 msg = "Can't use offline node"
735 if lu.cfg.GetNodeInfo(node).offline:
736 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
739 def _CheckNodeNotDrained(lu, node):
740 """Ensure that a given node is not drained.
742 @param lu: the LU on behalf of which we make the check
743 @param node: the node to check
744 @raise errors.OpPrereqError: if the node is drained
747 if lu.cfg.GetNodeInfo(node).drained:
748 raise errors.OpPrereqError("Can't use drained node %s" % node,
752 def _CheckNodeVmCapable(lu, node):
753 """Ensure that a given node is vm capable.
755 @param lu: the LU on behalf of which we make the check
756 @param node: the node to check
757 @raise errors.OpPrereqError: if the node is not vm capable
760 if not lu.cfg.GetNodeInfo(node).vm_capable:
761 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
765 def _CheckNodeHasOS(lu, node, os_name, force_variant):
766 """Ensure that a node supports a given OS.
768 @param lu: the LU on behalf of which we make the check
769 @param node: the node to check
770 @param os_name: the OS to query about
771 @param force_variant: whether to ignore variant errors
772 @raise errors.OpPrereqError: if the node is not supporting the OS
775 result = lu.rpc.call_os_get(node, os_name)
776 result.Raise("OS '%s' not in supported OS list for node %s" %
778 prereq=True, ecode=errors.ECODE_INVAL)
779 if not force_variant:
780 _CheckOSVariant(result.payload, os_name)
783 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
784 """Ensure that a node has the given secondary ip.
786 @type lu: L{LogicalUnit}
787 @param lu: the LU on behalf of which we make the check
789 @param node: the node to check
790 @type secondary_ip: string
791 @param secondary_ip: the ip to check
792 @type prereq: boolean
793 @param prereq: whether to throw a prerequisite or an execute error
794 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
795 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
798 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
799 result.Raise("Failure checking secondary ip on node %s" % node,
800 prereq=prereq, ecode=errors.ECODE_ENVIRON)
801 if not result.payload:
802 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
803 " please fix and re-run this command" % secondary_ip)
805 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
807 raise errors.OpExecError(msg)
810 def _GetClusterDomainSecret():
811 """Reads the cluster domain secret.
814 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
818 def _CheckInstanceDown(lu, instance, reason):
819 """Ensure that an instance is not running."""
820 if instance.admin_up:
821 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
822 (instance.name, reason), errors.ECODE_STATE)
824 pnode = instance.primary_node
825 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
826 ins_l.Raise("Can't contact node %s for instance information" % pnode,
827 prereq=True, ecode=errors.ECODE_ENVIRON)
829 if instance.name in ins_l.payload:
830 raise errors.OpPrereqError("Instance %s is running, %s" %
831 (instance.name, reason), errors.ECODE_STATE)
834 def _ExpandItemName(fn, name, kind):
835 """Expand an item name.
837 @param fn: the function to use for expansion
838 @param name: requested item name
839 @param kind: text description ('Node' or 'Instance')
840 @return: the resolved (full) name
841 @raise errors.OpPrereqError: if the item is not found
845 if full_name is None:
846 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
851 def _ExpandNodeName(cfg, name):
852 """Wrapper over L{_ExpandItemName} for nodes."""
853 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
856 def _ExpandInstanceName(cfg, name):
857 """Wrapper over L{_ExpandItemName} for instance."""
858 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
861 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
862 memory, vcpus, nics, disk_template, disks,
863 bep, hvp, hypervisor_name):
864 """Builds instance related env variables for hooks
866 This builds the hook environment from individual variables.
869 @param name: the name of the instance
870 @type primary_node: string
871 @param primary_node: the name of the instance's primary node
872 @type secondary_nodes: list
873 @param secondary_nodes: list of secondary nodes as strings
874 @type os_type: string
875 @param os_type: the name of the instance's OS
876 @type status: boolean
877 @param status: the should_run status of the instance
879 @param memory: the memory size of the instance
881 @param vcpus: the count of VCPUs the instance has
883 @param nics: list of tuples (ip, mac, mode, link) representing
884 the NICs the instance has
885 @type disk_template: string
886 @param disk_template: the disk template of the instance
888 @param disks: the list of (size, mode) pairs
890 @param bep: the backend parameters for the instance
892 @param hvp: the hypervisor parameters for the instance
893 @type hypervisor_name: string
894 @param hypervisor_name: the hypervisor for the instance
896 @return: the hook environment for this instance
905 "INSTANCE_NAME": name,
906 "INSTANCE_PRIMARY": primary_node,
907 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
908 "INSTANCE_OS_TYPE": os_type,
909 "INSTANCE_STATUS": str_status,
910 "INSTANCE_MEMORY": memory,
911 "INSTANCE_VCPUS": vcpus,
912 "INSTANCE_DISK_TEMPLATE": disk_template,
913 "INSTANCE_HYPERVISOR": hypervisor_name,
917 nic_count = len(nics)
918 for idx, (ip, mac, mode, link) in enumerate(nics):
921 env["INSTANCE_NIC%d_IP" % idx] = ip
922 env["INSTANCE_NIC%d_MAC" % idx] = mac
923 env["INSTANCE_NIC%d_MODE" % idx] = mode
924 env["INSTANCE_NIC%d_LINK" % idx] = link
925 if mode == constants.NIC_MODE_BRIDGED:
926 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
930 env["INSTANCE_NIC_COUNT"] = nic_count
933 disk_count = len(disks)
934 for idx, (size, mode) in enumerate(disks):
935 env["INSTANCE_DISK%d_SIZE" % idx] = size
936 env["INSTANCE_DISK%d_MODE" % idx] = mode
940 env["INSTANCE_DISK_COUNT"] = disk_count
942 for source, kind in [(bep, "BE"), (hvp, "HV")]:
943 for key, value in source.items():
944 env["INSTANCE_%s_%s" % (kind, key)] = value
949 def _NICListToTuple(lu, nics):
950 """Build a list of nic information tuples.
952 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
953 value in LUInstanceQueryData.
955 @type lu: L{LogicalUnit}
956 @param lu: the logical unit on whose behalf we execute
957 @type nics: list of L{objects.NIC}
958 @param nics: list of nics to convert to hooks tuples
962 cluster = lu.cfg.GetClusterInfo()
966 filled_params = cluster.SimpleFillNIC(nic.nicparams)
967 mode = filled_params[constants.NIC_MODE]
968 link = filled_params[constants.NIC_LINK]
969 hooks_nics.append((ip, mac, mode, link))
973 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
974 """Builds instance related env variables for hooks from an object.
976 @type lu: L{LogicalUnit}
977 @param lu: the logical unit on whose behalf we execute
978 @type instance: L{objects.Instance}
979 @param instance: the instance for which we should build the
982 @param override: dictionary with key/values that will override
985 @return: the hook environment dictionary
988 cluster = lu.cfg.GetClusterInfo()
989 bep = cluster.FillBE(instance)
990 hvp = cluster.FillHV(instance)
992 'name': instance.name,
993 'primary_node': instance.primary_node,
994 'secondary_nodes': instance.secondary_nodes,
995 'os_type': instance.os,
996 'status': instance.admin_up,
997 'memory': bep[constants.BE_MEMORY],
998 'vcpus': bep[constants.BE_VCPUS],
999 'nics': _NICListToTuple(lu, instance.nics),
1000 'disk_template': instance.disk_template,
1001 'disks': [(disk.size, disk.mode) for disk in instance.disks],
1004 'hypervisor_name': instance.hypervisor,
1007 args.update(override)
1008 return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
1011 def _AdjustCandidatePool(lu, exceptions):
1012 """Adjust the candidate pool after node operations.
1015 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1017 lu.LogInfo("Promoted nodes to master candidate role: %s",
1018 utils.CommaJoin(node.name for node in mod_list))
1019 for name in mod_list:
1020 lu.context.ReaddNode(name)
1021 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1023 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1027 def _DecideSelfPromotion(lu, exceptions=None):
1028 """Decide whether I should promote myself as a master candidate.
1031 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1032 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1033 # the new node will increase mc_max with one, so:
1034 mc_should = min(mc_should + 1, cp_size)
1035 return mc_now < mc_should
1038 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1039 """Check that the brigdes needed by a list of nics exist.
1042 cluster = lu.cfg.GetClusterInfo()
1043 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1044 brlist = [params[constants.NIC_LINK] for params in paramslist
1045 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1047 result = lu.rpc.call_bridges_exist(target_node, brlist)
1048 result.Raise("Error checking bridges on destination node '%s'" %
1049 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1052 def _CheckInstanceBridgesExist(lu, instance, node=None):
1053 """Check that the brigdes needed by an instance exist.
1057 node = instance.primary_node
1058 _CheckNicsBridgesExist(lu, instance.nics, node)
1061 def _CheckOSVariant(os_obj, name):
1062 """Check whether an OS name conforms to the os variants specification.
1064 @type os_obj: L{objects.OS}
1065 @param os_obj: OS object to check
1067 @param name: OS name passed by the user, to check for validity
1070 if not os_obj.supported_variants:
1072 variant = objects.OS.GetVariant(name)
1074 raise errors.OpPrereqError("OS name must include a variant",
1077 if variant not in os_obj.supported_variants:
1078 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1081 def _GetNodeInstancesInner(cfg, fn):
1082 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1085 def _GetNodeInstances(cfg, node_name):
1086 """Returns a list of all primary and secondary instances on a node.
1090 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1093 def _GetNodePrimaryInstances(cfg, node_name):
1094 """Returns primary instances on a node.
1097 return _GetNodeInstancesInner(cfg,
1098 lambda inst: node_name == inst.primary_node)
1101 def _GetNodeSecondaryInstances(cfg, node_name):
1102 """Returns secondary instances on a node.
1105 return _GetNodeInstancesInner(cfg,
1106 lambda inst: node_name in inst.secondary_nodes)
1109 def _GetStorageTypeArgs(cfg, storage_type):
1110 """Returns the arguments for a storage type.
1113 # Special case for file storage
1114 if storage_type == constants.ST_FILE:
1115 # storage.FileStorage wants a list of storage directories
1116 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1121 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1124 for dev in instance.disks:
1125 cfg.SetDiskID(dev, node_name)
1127 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1128 result.Raise("Failed to get disk status from node %s" % node_name,
1129 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1131 for idx, bdev_status in enumerate(result.payload):
1132 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1138 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1139 """Check the sanity of iallocator and node arguments and use the
1140 cluster-wide iallocator if appropriate.
1142 Check that at most one of (iallocator, node) is specified. If none is
1143 specified, then the LU's opcode's iallocator slot is filled with the
1144 cluster-wide default iallocator.
1146 @type iallocator_slot: string
1147 @param iallocator_slot: the name of the opcode iallocator slot
1148 @type node_slot: string
1149 @param node_slot: the name of the opcode target node slot
1152 node = getattr(lu.op, node_slot, None)
1153 iallocator = getattr(lu.op, iallocator_slot, None)
1155 if node is not None and iallocator is not None:
1156 raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1158 elif node is None and iallocator is None:
1159 default_iallocator = lu.cfg.GetDefaultIAllocator()
1160 if default_iallocator:
1161 setattr(lu.op, iallocator_slot, default_iallocator)
1163 raise errors.OpPrereqError("No iallocator or node given and no"
1164 " cluster-wide default iallocator found."
1165 " Please specify either an iallocator or a"
1166 " node, or set a cluster-wide default"
1170 class LUClusterPostInit(LogicalUnit):
1171 """Logical unit for running hooks after cluster initialization.
1174 HPATH = "cluster-init"
1175 HTYPE = constants.HTYPE_CLUSTER
1177 def BuildHooksEnv(self):
1182 "OP_TARGET": self.cfg.GetClusterName(),
1185 def BuildHooksNodes(self):
1186 """Build hooks nodes.
1189 return ([], [self.cfg.GetMasterNode()])
1191 def Exec(self, feedback_fn):
1198 class LUClusterDestroy(LogicalUnit):
1199 """Logical unit for destroying the cluster.
1202 HPATH = "cluster-destroy"
1203 HTYPE = constants.HTYPE_CLUSTER
1205 def BuildHooksEnv(self):
1210 "OP_TARGET": self.cfg.GetClusterName(),
1213 def BuildHooksNodes(self):
1214 """Build hooks nodes.
1219 def CheckPrereq(self):
1220 """Check prerequisites.
1222 This checks whether the cluster is empty.
1224 Any errors are signaled by raising errors.OpPrereqError.
1227 master = self.cfg.GetMasterNode()
1229 nodelist = self.cfg.GetNodeList()
1230 if len(nodelist) != 1 or nodelist[0] != master:
1231 raise errors.OpPrereqError("There are still %d node(s) in"
1232 " this cluster." % (len(nodelist) - 1),
1234 instancelist = self.cfg.GetInstanceList()
1236 raise errors.OpPrereqError("There are still %d instance(s) in"
1237 " this cluster." % len(instancelist),
1240 def Exec(self, feedback_fn):
1241 """Destroys the cluster.
1244 master = self.cfg.GetMasterNode()
1246 # Run post hooks on master node before it's removed
1247 _RunPostHook(self, master)
1249 result = self.rpc.call_node_stop_master(master, False)
1250 result.Raise("Could not disable the master role")
1255 def _VerifyCertificate(filename):
1256 """Verifies a certificate for LUClusterVerify.
1258 @type filename: string
1259 @param filename: Path to PEM file
1263 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1264 utils.ReadFile(filename))
1265 except Exception, err: # pylint: disable-msg=W0703
1266 return (LUClusterVerify.ETYPE_ERROR,
1267 "Failed to load X509 certificate %s: %s" % (filename, err))
1270 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1271 constants.SSL_CERT_EXPIRATION_ERROR)
1274 fnamemsg = "While verifying %s: %s" % (filename, msg)
1279 return (None, fnamemsg)
1280 elif errcode == utils.CERT_WARNING:
1281 return (LUClusterVerify.ETYPE_WARNING, fnamemsg)
1282 elif errcode == utils.CERT_ERROR:
1283 return (LUClusterVerify.ETYPE_ERROR, fnamemsg)
1285 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1288 class LUClusterVerify(LogicalUnit):
1289 """Verifies the cluster status.
1292 HPATH = "cluster-verify"
1293 HTYPE = constants.HTYPE_CLUSTER
1296 TCLUSTER = "cluster"
1298 TINSTANCE = "instance"
1300 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1301 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1302 ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK")
1303 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1304 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1305 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1306 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1307 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1308 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1309 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1310 ENODEDRBD = (TNODE, "ENODEDRBD")
1311 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1312 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1313 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1314 ENODEHV = (TNODE, "ENODEHV")
1315 ENODELVM = (TNODE, "ENODELVM")
1316 ENODEN1 = (TNODE, "ENODEN1")
1317 ENODENET = (TNODE, "ENODENET")
1318 ENODEOS = (TNODE, "ENODEOS")
1319 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1320 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1321 ENODERPC = (TNODE, "ENODERPC")
1322 ENODESSH = (TNODE, "ENODESSH")
1323 ENODEVERSION = (TNODE, "ENODEVERSION")
1324 ENODESETUP = (TNODE, "ENODESETUP")
1325 ENODETIME = (TNODE, "ENODETIME")
1326 ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1328 ETYPE_FIELD = "code"
1329 ETYPE_ERROR = "ERROR"
1330 ETYPE_WARNING = "WARNING"
1332 _HOOKS_INDENT_RE = re.compile("^", re.M)
1334 class NodeImage(object):
1335 """A class representing the logical and physical status of a node.
1338 @ivar name: the node name to which this object refers
1339 @ivar volumes: a structure as returned from
1340 L{ganeti.backend.GetVolumeList} (runtime)
1341 @ivar instances: a list of running instances (runtime)
1342 @ivar pinst: list of configured primary instances (config)
1343 @ivar sinst: list of configured secondary instances (config)
1344 @ivar sbp: dictionary of {primary-node: list of instances} for all
1345 instances for which this node is secondary (config)
1346 @ivar mfree: free memory, as reported by hypervisor (runtime)
1347 @ivar dfree: free disk, as reported by the node (runtime)
1348 @ivar offline: the offline status (config)
1349 @type rpc_fail: boolean
1350 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1351 not whether the individual keys were correct) (runtime)
1352 @type lvm_fail: boolean
1353 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1354 @type hyp_fail: boolean
1355 @ivar hyp_fail: whether the RPC call didn't return the instance list
1356 @type ghost: boolean
1357 @ivar ghost: whether this is a known node or not (config)
1358 @type os_fail: boolean
1359 @ivar os_fail: whether the RPC call didn't return valid OS data
1361 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1362 @type vm_capable: boolean
1363 @ivar vm_capable: whether the node can host instances
1366 def __init__(self, offline=False, name=None, vm_capable=True):
1375 self.offline = offline
1376 self.vm_capable = vm_capable
1377 self.rpc_fail = False
1378 self.lvm_fail = False
1379 self.hyp_fail = False
1381 self.os_fail = False
1384 def ExpandNames(self):
1385 self.needed_locks = {
1386 locking.LEVEL_NODE: locking.ALL_SET,
1387 locking.LEVEL_INSTANCE: locking.ALL_SET,
1389 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1391 def _Error(self, ecode, item, msg, *args, **kwargs):
1392 """Format an error message.
1394 Based on the opcode's error_codes parameter, either format a
1395 parseable error code, or a simpler error string.
1397 This must be called only from Exec and functions called from Exec.
1400 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1402 # first complete the msg
1405 # then format the whole message
1406 if self.op.error_codes:
1407 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1413 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1414 # and finally report it via the feedback_fn
1415 self._feedback_fn(" - %s" % msg)
1417 def _ErrorIf(self, cond, *args, **kwargs):
1418 """Log an error message if the passed condition is True.
1421 cond = bool(cond) or self.op.debug_simulate_errors
1423 self._Error(*args, **kwargs)
1424 # do not mark the operation as failed for WARN cases only
1425 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1426 self.bad = self.bad or cond
1428 def _VerifyNode(self, ninfo, nresult):
1429 """Perform some basic validation on data returned from a node.
1431 - check the result data structure is well formed and has all the
1433 - check ganeti version
1435 @type ninfo: L{objects.Node}
1436 @param ninfo: the node to check
1437 @param nresult: the results from the node
1439 @return: whether overall this call was successful (and we can expect
1440 reasonable values in the respose)
1444 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1446 # main result, nresult should be a non-empty dict
1447 test = not nresult or not isinstance(nresult, dict)
1448 _ErrorIf(test, self.ENODERPC, node,
1449 "unable to verify node: no data returned")
1453 # compares ganeti version
1454 local_version = constants.PROTOCOL_VERSION
1455 remote_version = nresult.get("version", None)
1456 test = not (remote_version and
1457 isinstance(remote_version, (list, tuple)) and
1458 len(remote_version) == 2)
1459 _ErrorIf(test, self.ENODERPC, node,
1460 "connection to node returned invalid data")
1464 test = local_version != remote_version[0]
1465 _ErrorIf(test, self.ENODEVERSION, node,
1466 "incompatible protocol versions: master %s,"
1467 " node %s", local_version, remote_version[0])
1471 # node seems compatible, we can actually try to look into its results
1473 # full package version
1474 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1475 self.ENODEVERSION, node,
1476 "software version mismatch: master %s, node %s",
1477 constants.RELEASE_VERSION, remote_version[1],
1478 code=self.ETYPE_WARNING)
1480 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1481 if ninfo.vm_capable and isinstance(hyp_result, dict):
1482 for hv_name, hv_result in hyp_result.iteritems():
1483 test = hv_result is not None
1484 _ErrorIf(test, self.ENODEHV, node,
1485 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1487 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1488 if ninfo.vm_capable and isinstance(hvp_result, list):
1489 for item, hv_name, hv_result in hvp_result:
1490 _ErrorIf(True, self.ENODEHV, node,
1491 "hypervisor %s parameter verify failure (source %s): %s",
1492 hv_name, item, hv_result)
1494 test = nresult.get(constants.NV_NODESETUP,
1495 ["Missing NODESETUP results"])
1496 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1501 def _VerifyNodeTime(self, ninfo, nresult,
1502 nvinfo_starttime, nvinfo_endtime):
1503 """Check the node time.
1505 @type ninfo: L{objects.Node}
1506 @param ninfo: the node to check
1507 @param nresult: the remote results for the node
1508 @param nvinfo_starttime: the start time of the RPC call
1509 @param nvinfo_endtime: the end time of the RPC call
1513 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1515 ntime = nresult.get(constants.NV_TIME, None)
1517 ntime_merged = utils.MergeTime(ntime)
1518 except (ValueError, TypeError):
1519 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1522 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1523 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1524 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1525 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1529 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1530 "Node time diverges by at least %s from master node time",
1533 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1534 """Check the node time.
1536 @type ninfo: L{objects.Node}
1537 @param ninfo: the node to check
1538 @param nresult: the remote results for the node
1539 @param vg_name: the configured VG name
1546 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1548 # checks vg existence and size > 20G
1549 vglist = nresult.get(constants.NV_VGLIST, None)
1551 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1553 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1554 constants.MIN_VG_SIZE)
1555 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1558 pvlist = nresult.get(constants.NV_PVLIST, None)
1559 test = pvlist is None
1560 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1562 # check that ':' is not present in PV names, since it's a
1563 # special character for lvcreate (denotes the range of PEs to
1565 for _, pvname, owner_vg in pvlist:
1566 test = ":" in pvname
1567 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1568 " '%s' of VG '%s'", pvname, owner_vg)
1570 def _VerifyNodeNetwork(self, ninfo, nresult):
1571 """Check the node time.
1573 @type ninfo: L{objects.Node}
1574 @param ninfo: the node to check
1575 @param nresult: the remote results for the node
1579 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1581 test = constants.NV_NODELIST not in nresult
1582 _ErrorIf(test, self.ENODESSH, node,
1583 "node hasn't returned node ssh connectivity data")
1585 if nresult[constants.NV_NODELIST]:
1586 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1587 _ErrorIf(True, self.ENODESSH, node,
1588 "ssh communication with node '%s': %s", a_node, a_msg)
1590 test = constants.NV_NODENETTEST not in nresult
1591 _ErrorIf(test, self.ENODENET, node,
1592 "node hasn't returned node tcp connectivity data")
1594 if nresult[constants.NV_NODENETTEST]:
1595 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1597 _ErrorIf(True, self.ENODENET, node,
1598 "tcp communication with node '%s': %s",
1599 anode, nresult[constants.NV_NODENETTEST][anode])
1601 test = constants.NV_MASTERIP not in nresult
1602 _ErrorIf(test, self.ENODENET, node,
1603 "node hasn't returned node master IP reachability data")
1605 if not nresult[constants.NV_MASTERIP]:
1606 if node == self.master_node:
1607 msg = "the master node cannot reach the master IP (not configured?)"
1609 msg = "cannot reach the master IP"
1610 _ErrorIf(True, self.ENODENET, node, msg)
1612 def _VerifyInstance(self, instance, instanceconfig, node_image,
1614 """Verify an instance.
1616 This function checks to see if the required block devices are
1617 available on the instance's node.
1620 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1621 node_current = instanceconfig.primary_node
1623 node_vol_should = {}
1624 instanceconfig.MapLVsByNode(node_vol_should)
1626 for node in node_vol_should:
1627 n_img = node_image[node]
1628 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1629 # ignore missing volumes on offline or broken nodes
1631 for volume in node_vol_should[node]:
1632 test = volume not in n_img.volumes
1633 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1634 "volume %s missing on node %s", volume, node)
1636 if instanceconfig.admin_up:
1637 pri_img = node_image[node_current]
1638 test = instance not in pri_img.instances and not pri_img.offline
1639 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1640 "instance not running on its primary node %s",
1643 for node, n_img in node_image.items():
1644 if node != node_current:
1645 test = instance in n_img.instances
1646 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1647 "instance should not run on node %s", node)
1649 diskdata = [(nname, success, status, idx)
1650 for (nname, disks) in diskstatus.items()
1651 for idx, (success, status) in enumerate(disks)]
1653 for nname, success, bdev_status, idx in diskdata:
1654 # the 'ghost node' construction in Exec() ensures that we have a
1656 snode = node_image[nname]
1657 bad_snode = snode.ghost or snode.offline
1658 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
1659 self.EINSTANCEFAULTYDISK, instance,
1660 "couldn't retrieve status for disk/%s on %s: %s",
1661 idx, nname, bdev_status)
1662 _ErrorIf((instanceconfig.admin_up and success and
1663 bdev_status.ldisk_status == constants.LDS_FAULTY),
1664 self.EINSTANCEFAULTYDISK, instance,
1665 "disk/%s on %s is faulty", idx, nname)
1667 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1668 """Verify if there are any unknown volumes in the cluster.
1670 The .os, .swap and backup volumes are ignored. All other volumes are
1671 reported as unknown.
1673 @type reserved: L{ganeti.utils.FieldSet}
1674 @param reserved: a FieldSet of reserved volume names
1677 for node, n_img in node_image.items():
1678 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1679 # skip non-healthy nodes
1681 for volume in n_img.volumes:
1682 test = ((node not in node_vol_should or
1683 volume not in node_vol_should[node]) and
1684 not reserved.Matches(volume))
1685 self._ErrorIf(test, self.ENODEORPHANLV, node,
1686 "volume %s is unknown", volume)
1688 def _VerifyOrphanInstances(self, instancelist, node_image):
1689 """Verify the list of running instances.
1691 This checks what instances are running but unknown to the cluster.
1694 for node, n_img in node_image.items():
1695 for o_inst in n_img.instances:
1696 test = o_inst not in instancelist
1697 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1698 "instance %s on node %s should not exist", o_inst, node)
1700 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1701 """Verify N+1 Memory Resilience.
1703 Check that if one single node dies we can still start all the
1704 instances it was primary for.
1707 cluster_info = self.cfg.GetClusterInfo()
1708 for node, n_img in node_image.items():
1709 # This code checks that every node which is now listed as
1710 # secondary has enough memory to host all instances it is
1711 # supposed to should a single other node in the cluster fail.
1712 # FIXME: not ready for failover to an arbitrary node
1713 # FIXME: does not support file-backed instances
1714 # WARNING: we currently take into account down instances as well
1715 # as up ones, considering that even if they're down someone
1716 # might want to start them even in the event of a node failure.
1718 # we're skipping offline nodes from the N+1 warning, since
1719 # most likely we don't have good memory infromation from them;
1720 # we already list instances living on such nodes, and that's
1723 for prinode, instances in n_img.sbp.items():
1725 for instance in instances:
1726 bep = cluster_info.FillBE(instance_cfg[instance])
1727 if bep[constants.BE_AUTO_BALANCE]:
1728 needed_mem += bep[constants.BE_MEMORY]
1729 test = n_img.mfree < needed_mem
1730 self._ErrorIf(test, self.ENODEN1, node,
1731 "not enough memory to accomodate instance failovers"
1732 " should node %s fail (%dMiB needed, %dMiB available)",
1733 prinode, needed_mem, n_img.mfree)
1736 def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
1737 (files_all, files_all_opt, files_mc, files_vm)):
1738 """Verifies file checksums collected from all nodes.
1740 @param errorif: Callback for reporting errors
1741 @param nodeinfo: List of L{objects.Node} objects
1742 @param master_node: Name of master node
1743 @param all_nvinfo: RPC results
1746 node_names = frozenset(node.name for node in nodeinfo)
1748 assert master_node in node_names
1749 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
1750 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
1751 "Found file listed in more than one file list"
1753 # Define functions determining which nodes to consider for a file
1754 file2nodefn = dict([(filename, fn)
1755 for (files, fn) in [(files_all, None),
1756 (files_all_opt, None),
1757 (files_mc, lambda node: (node.master_candidate or
1758 node.name == master_node)),
1759 (files_vm, lambda node: node.vm_capable)]
1760 for filename in files])
1762 fileinfo = dict((filename, {}) for filename in file2nodefn.keys())
1764 for node in nodeinfo:
1765 nresult = all_nvinfo[node.name]
1767 if nresult.fail_msg or not nresult.payload:
1770 node_files = nresult.payload.get(constants.NV_FILELIST, None)
1772 test = not (node_files and isinstance(node_files, dict))
1773 errorif(test, cls.ENODEFILECHECK, node.name,
1774 "Node did not return file checksum data")
1778 for (filename, checksum) in node_files.items():
1779 # Check if the file should be considered for a node
1780 fn = file2nodefn[filename]
1781 if fn is None or fn(node):
1782 fileinfo[filename].setdefault(checksum, set()).add(node.name)
1784 for (filename, checksums) in fileinfo.items():
1785 assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
1787 # Nodes having the file
1788 with_file = frozenset(node_name
1789 for nodes in fileinfo[filename].values()
1790 for node_name in nodes)
1792 # Nodes missing file
1793 missing_file = node_names - with_file
1795 if filename in files_all_opt:
1797 errorif(missing_file and missing_file != node_names,
1798 cls.ECLUSTERFILECHECK, None,
1799 "File %s is optional, but it must exist on all or no nodes (not"
1801 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
1803 errorif(missing_file, cls.ECLUSTERFILECHECK, None,
1804 "File %s is missing from node(s) %s", filename,
1805 utils.CommaJoin(utils.NiceSort(missing_file)))
1807 # See if there are multiple versions of the file
1808 test = len(checksums) > 1
1810 variants = ["variant %s on %s" %
1811 (idx + 1, utils.CommaJoin(utils.NiceSort(nodes)))
1812 for (idx, (checksum, nodes)) in
1813 enumerate(sorted(checksums.items()))]
1817 errorif(test, cls.ECLUSTERFILECHECK, None,
1818 "File %s found with %s different checksums (%s)",
1819 filename, len(checksums), "; ".join(variants))
1821 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1823 """Verifies and the node DRBD status.
1825 @type ninfo: L{objects.Node}
1826 @param ninfo: the node to check
1827 @param nresult: the remote results for the node
1828 @param instanceinfo: the dict of instances
1829 @param drbd_helper: the configured DRBD usermode helper
1830 @param drbd_map: the DRBD map as returned by
1831 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1835 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1838 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1839 test = (helper_result == None)
1840 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1841 "no drbd usermode helper returned")
1843 status, payload = helper_result
1845 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1846 "drbd usermode helper check unsuccessful: %s", payload)
1847 test = status and (payload != drbd_helper)
1848 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1849 "wrong drbd usermode helper: %s", payload)
1851 # compute the DRBD minors
1853 for minor, instance in drbd_map[node].items():
1854 test = instance not in instanceinfo
1855 _ErrorIf(test, self.ECLUSTERCFG, None,
1856 "ghost instance '%s' in temporary DRBD map", instance)
1857 # ghost instance should not be running, but otherwise we
1858 # don't give double warnings (both ghost instance and
1859 # unallocated minor in use)
1861 node_drbd[minor] = (instance, False)
1863 instance = instanceinfo[instance]
1864 node_drbd[minor] = (instance.name, instance.admin_up)
1866 # and now check them
1867 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1868 test = not isinstance(used_minors, (tuple, list))
1869 _ErrorIf(test, self.ENODEDRBD, node,
1870 "cannot parse drbd status file: %s", str(used_minors))
1872 # we cannot check drbd status
1875 for minor, (iname, must_exist) in node_drbd.items():
1876 test = minor not in used_minors and must_exist
1877 _ErrorIf(test, self.ENODEDRBD, node,
1878 "drbd minor %d of instance %s is not active", minor, iname)
1879 for minor in used_minors:
1880 test = minor not in node_drbd
1881 _ErrorIf(test, self.ENODEDRBD, node,
1882 "unallocated drbd minor %d is in use", minor)
1884 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1885 """Builds the node OS structures.
1887 @type ninfo: L{objects.Node}
1888 @param ninfo: the node to check
1889 @param nresult: the remote results for the node
1890 @param nimg: the node image object
1894 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1896 remote_os = nresult.get(constants.NV_OSLIST, None)
1897 test = (not isinstance(remote_os, list) or
1898 not compat.all(isinstance(v, list) and len(v) == 7
1899 for v in remote_os))
1901 _ErrorIf(test, self.ENODEOS, node,
1902 "node hasn't returned valid OS data")
1911 for (name, os_path, status, diagnose,
1912 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1914 if name not in os_dict:
1917 # parameters is a list of lists instead of list of tuples due to
1918 # JSON lacking a real tuple type, fix it:
1919 parameters = [tuple(v) for v in parameters]
1920 os_dict[name].append((os_path, status, diagnose,
1921 set(variants), set(parameters), set(api_ver)))
1923 nimg.oslist = os_dict
1925 def _VerifyNodeOS(self, ninfo, nimg, base):
1926 """Verifies the node OS list.
1928 @type ninfo: L{objects.Node}
1929 @param ninfo: the node to check
1930 @param nimg: the node image object
1931 @param base: the 'template' node we match against (e.g. from the master)
1935 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1937 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1939 beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
1940 for os_name, os_data in nimg.oslist.items():
1941 assert os_data, "Empty OS status for OS %s?!" % os_name
1942 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1943 _ErrorIf(not f_status, self.ENODEOS, node,
1944 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1945 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1946 "OS '%s' has multiple entries (first one shadows the rest): %s",
1947 os_name, utils.CommaJoin([v[0] for v in os_data]))
1948 # this will catched in backend too
1949 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1950 and not f_var, self.ENODEOS, node,
1951 "OS %s with API at least %d does not declare any variant",
1952 os_name, constants.OS_API_V15)
1953 # comparisons with the 'base' image
1954 test = os_name not in base.oslist
1955 _ErrorIf(test, self.ENODEOS, node,
1956 "Extra OS %s not present on reference node (%s)",
1960 assert base.oslist[os_name], "Base node has empty OS status?"
1961 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1963 # base OS is invalid, skipping
1965 for kind, a, b in [("API version", f_api, b_api),
1966 ("variants list", f_var, b_var),
1967 ("parameters", beautify_params(f_param),
1968 beautify_params(b_param))]:
1969 _ErrorIf(a != b, self.ENODEOS, node,
1970 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
1971 kind, os_name, base.name,
1972 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
1974 # check any missing OSes
1975 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1976 _ErrorIf(missing, self.ENODEOS, node,
1977 "OSes present on reference node %s but missing on this node: %s",
1978 base.name, utils.CommaJoin(missing))
1980 def _VerifyOob(self, ninfo, nresult):
1981 """Verifies out of band functionality of a node.
1983 @type ninfo: L{objects.Node}
1984 @param ninfo: the node to check
1985 @param nresult: the remote results for the node
1989 # We just have to verify the paths on master and/or master candidates
1990 # as the oob helper is invoked on the master
1991 if ((ninfo.master_candidate or ninfo.master_capable) and
1992 constants.NV_OOB_PATHS in nresult):
1993 for path_result in nresult[constants.NV_OOB_PATHS]:
1994 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1996 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1997 """Verifies and updates the node volume data.
1999 This function will update a L{NodeImage}'s internal structures
2000 with data from the remote call.
2002 @type ninfo: L{objects.Node}
2003 @param ninfo: the node to check
2004 @param nresult: the remote results for the node
2005 @param nimg: the node image object
2006 @param vg_name: the configured VG name
2010 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2012 nimg.lvm_fail = True
2013 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
2016 elif isinstance(lvdata, basestring):
2017 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
2018 utils.SafeEncode(lvdata))
2019 elif not isinstance(lvdata, dict):
2020 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
2022 nimg.volumes = lvdata
2023 nimg.lvm_fail = False
2025 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
2026 """Verifies and updates the node instance list.
2028 If the listing was successful, then updates this node's instance
2029 list. Otherwise, it marks the RPC call as failed for the instance
2032 @type ninfo: L{objects.Node}
2033 @param ninfo: the node to check
2034 @param nresult: the remote results for the node
2035 @param nimg: the node image object
2038 idata = nresult.get(constants.NV_INSTANCELIST, None)
2039 test = not isinstance(idata, list)
2040 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
2041 " (instancelist): %s", utils.SafeEncode(str(idata)))
2043 nimg.hyp_fail = True
2045 nimg.instances = idata
2047 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
2048 """Verifies and computes a node information map
2050 @type ninfo: L{objects.Node}
2051 @param ninfo: the node to check
2052 @param nresult: the remote results for the node
2053 @param nimg: the node image object
2054 @param vg_name: the configured VG name
2058 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2060 # try to read free memory (from the hypervisor)
2061 hv_info = nresult.get(constants.NV_HVINFO, None)
2062 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2063 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
2066 nimg.mfree = int(hv_info["memory_free"])
2067 except (ValueError, TypeError):
2068 _ErrorIf(True, self.ENODERPC, node,
2069 "node returned invalid nodeinfo, check hypervisor")
2071 # FIXME: devise a free space model for file based instances as well
2072 if vg_name is not None:
2073 test = (constants.NV_VGLIST not in nresult or
2074 vg_name not in nresult[constants.NV_VGLIST])
2075 _ErrorIf(test, self.ENODELVM, node,
2076 "node didn't return data for the volume group '%s'"
2077 " - it is either missing or broken", vg_name)
2080 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2081 except (ValueError, TypeError):
2082 _ErrorIf(True, self.ENODERPC, node,
2083 "node returned invalid LVM info, check LVM status")
2085 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
2086 """Gets per-disk status information for all instances.
2088 @type nodelist: list of strings
2089 @param nodelist: Node names
2090 @type node_image: dict of (name, L{objects.Node})
2091 @param node_image: Node objects
2092 @type instanceinfo: dict of (name, L{objects.Instance})
2093 @param instanceinfo: Instance objects
2094 @rtype: {instance: {node: [(succes, payload)]}}
2095 @return: a dictionary of per-instance dictionaries with nodes as
2096 keys and disk information as values; the disk information is a
2097 list of tuples (success, payload)
2100 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2103 node_disks_devonly = {}
2104 diskless_instances = set()
2105 diskless = constants.DT_DISKLESS
2107 for nname in nodelist:
2108 node_instances = list(itertools.chain(node_image[nname].pinst,
2109 node_image[nname].sinst))
2110 diskless_instances.update(inst for inst in node_instances
2111 if instanceinfo[inst].disk_template == diskless)
2112 disks = [(inst, disk)
2113 for inst in node_instances
2114 for disk in instanceinfo[inst].disks]
2117 # No need to collect data
2120 node_disks[nname] = disks
2122 # Creating copies as SetDiskID below will modify the objects and that can
2123 # lead to incorrect data returned from nodes
2124 devonly = [dev.Copy() for (_, dev) in disks]
2127 self.cfg.SetDiskID(dev, nname)
2129 node_disks_devonly[nname] = devonly
2131 assert len(node_disks) == len(node_disks_devonly)
2133 # Collect data from all nodes with disks
2134 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2137 assert len(result) == len(node_disks)
2141 for (nname, nres) in result.items():
2142 disks = node_disks[nname]
2145 # No data from this node
2146 data = len(disks) * [(False, "node offline")]
2149 _ErrorIf(msg, self.ENODERPC, nname,
2150 "while getting disk information: %s", msg)
2152 # No data from this node
2153 data = len(disks) * [(False, msg)]
2156 for idx, i in enumerate(nres.payload):
2157 if isinstance(i, (tuple, list)) and len(i) == 2:
2160 logging.warning("Invalid result from node %s, entry %d: %s",
2162 data.append((False, "Invalid result from the remote node"))
2164 for ((inst, _), status) in zip(disks, data):
2165 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2167 # Add empty entries for diskless instances.
2168 for inst in diskless_instances:
2169 assert inst not in instdisk
2172 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2173 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2174 compat.all(isinstance(s, (tuple, list)) and
2175 len(s) == 2 for s in statuses)
2176 for inst, nnames in instdisk.items()
2177 for nname, statuses in nnames.items())
2178 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2182 def _VerifyHVP(self, hvp_data):
2183 """Verifies locally the syntax of the hypervisor parameters.
2186 for item, hv_name, hv_params in hvp_data:
2187 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2190 hv_class = hypervisor.GetHypervisor(hv_name)
2191 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2192 hv_class.CheckParameterSyntax(hv_params)
2193 except errors.GenericError, err:
2194 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2196 def BuildHooksEnv(self):
2199 Cluster-Verify hooks just ran in the post phase and their failure makes
2200 the output be logged in the verify output and the verification to fail.
2206 "CLUSTER_TAGS": " ".join(cfg.GetClusterInfo().GetTags())
2209 env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
2210 for node in cfg.GetAllNodesInfo().values())
2214 def BuildHooksNodes(self):
2215 """Build hooks nodes.
2218 return ([], self.cfg.GetNodeList())
2220 def Exec(self, feedback_fn):
2221 """Verify integrity of cluster, performing various test on nodes.
2224 # This method has too many local variables. pylint: disable-msg=R0914
2226 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2227 verbose = self.op.verbose
2228 self._feedback_fn = feedback_fn
2229 feedback_fn("* Verifying global settings")
2230 for msg in self.cfg.VerifyConfig():
2231 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2233 # Check the cluster certificates
2234 for cert_filename in constants.ALL_CERT_FILES:
2235 (errcode, msg) = _VerifyCertificate(cert_filename)
2236 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2238 vg_name = self.cfg.GetVGName()
2239 drbd_helper = self.cfg.GetDRBDHelper()
2240 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2241 cluster = self.cfg.GetClusterInfo()
2242 nodelist = utils.NiceSort(self.cfg.GetNodeList())
2243 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2244 nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2245 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2246 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2247 for iname in instancelist)
2248 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2249 i_non_redundant = [] # Non redundant instances
2250 i_non_a_balanced = [] # Non auto-balanced instances
2251 n_offline = 0 # Count of offline nodes
2252 n_drained = 0 # Count of nodes being drained
2253 node_vol_should = {}
2255 # FIXME: verify OS list
2258 filemap = _ComputeAncillaryFiles(cluster, False)
2260 # do local checksums
2261 master_node = self.master_node = self.cfg.GetMasterNode()
2262 master_ip = self.cfg.GetMasterIP()
2264 # Compute the set of hypervisor parameters
2266 for hv_name in hypervisors:
2267 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2268 for os_name, os_hvp in cluster.os_hvp.items():
2269 for hv_name, hv_params in os_hvp.items():
2272 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2273 hvp_data.append(("os %s" % os_name, hv_name, full_params))
2274 # TODO: collapse identical parameter values in a single one
2275 for instance in instanceinfo.values():
2276 if not instance.hvparams:
2278 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2279 cluster.FillHV(instance)))
2280 # and verify them locally
2281 self._VerifyHVP(hvp_data)
2283 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2284 node_verify_param = {
2285 constants.NV_FILELIST:
2286 utils.UniqueSequence(filename
2287 for files in filemap
2288 for filename in files),
2289 constants.NV_NODELIST: [node.name for node in nodeinfo
2290 if not node.offline],
2291 constants.NV_HYPERVISOR: hypervisors,
2292 constants.NV_HVPARAMS: hvp_data,
2293 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2294 node.secondary_ip) for node in nodeinfo
2295 if not node.offline],
2296 constants.NV_INSTANCELIST: hypervisors,
2297 constants.NV_VERSION: None,
2298 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2299 constants.NV_NODESETUP: None,
2300 constants.NV_TIME: None,
2301 constants.NV_MASTERIP: (master_node, master_ip),
2302 constants.NV_OSLIST: None,
2303 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2306 if vg_name is not None:
2307 node_verify_param[constants.NV_VGLIST] = None
2308 node_verify_param[constants.NV_LVLIST] = vg_name
2309 node_verify_param[constants.NV_PVLIST] = [vg_name]
2310 node_verify_param[constants.NV_DRBDLIST] = None
2313 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2315 # Build our expected cluster state
2316 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2318 vm_capable=node.vm_capable))
2319 for node in nodeinfo)
2323 for node in nodeinfo:
2324 path = _SupportsOob(self.cfg, node)
2325 if path and path not in oob_paths:
2326 oob_paths.append(path)
2329 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2331 for instance in instancelist:
2332 inst_config = instanceinfo[instance]
2334 for nname in inst_config.all_nodes:
2335 if nname not in node_image:
2337 gnode = self.NodeImage(name=nname)
2339 node_image[nname] = gnode
2341 inst_config.MapLVsByNode(node_vol_should)
2343 pnode = inst_config.primary_node
2344 node_image[pnode].pinst.append(instance)
2346 for snode in inst_config.secondary_nodes:
2347 nimg = node_image[snode]
2348 nimg.sinst.append(instance)
2349 if pnode not in nimg.sbp:
2350 nimg.sbp[pnode] = []
2351 nimg.sbp[pnode].append(instance)
2353 # At this point, we have the in-memory data structures complete,
2354 # except for the runtime information, which we'll gather next
2356 # Due to the way our RPC system works, exact response times cannot be
2357 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2358 # time before and after executing the request, we can at least have a time
2360 nvinfo_starttime = time.time()
2361 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2362 self.cfg.GetClusterName())
2363 nvinfo_endtime = time.time()
2365 all_drbd_map = self.cfg.ComputeDRBDMap()
2367 feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2368 instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2370 feedback_fn("* Verifying configuration file consistency")
2371 self._VerifyFiles(_ErrorIf, nodeinfo, master_node, all_nvinfo, filemap)
2373 feedback_fn("* Verifying node status")
2377 for node_i in nodeinfo:
2379 nimg = node_image[node]
2383 feedback_fn("* Skipping offline node %s" % (node,))
2387 if node == master_node:
2389 elif node_i.master_candidate:
2390 ntype = "master candidate"
2391 elif node_i.drained:
2397 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2399 msg = all_nvinfo[node].fail_msg
2400 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2402 nimg.rpc_fail = True
2405 nresult = all_nvinfo[node].payload
2407 nimg.call_ok = self._VerifyNode(node_i, nresult)
2408 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2409 self._VerifyNodeNetwork(node_i, nresult)
2410 self._VerifyOob(node_i, nresult)
2413 self._VerifyNodeLVM(node_i, nresult, vg_name)
2414 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2417 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2418 self._UpdateNodeInstances(node_i, nresult, nimg)
2419 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2420 self._UpdateNodeOS(node_i, nresult, nimg)
2421 if not nimg.os_fail:
2422 if refos_img is None:
2424 self._VerifyNodeOS(node_i, nimg, refos_img)
2426 feedback_fn("* Verifying instance status")
2427 for instance in instancelist:
2429 feedback_fn("* Verifying instance %s" % instance)
2430 inst_config = instanceinfo[instance]
2431 self._VerifyInstance(instance, inst_config, node_image,
2433 inst_nodes_offline = []
2435 pnode = inst_config.primary_node
2436 pnode_img = node_image[pnode]
2437 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2438 self.ENODERPC, pnode, "instance %s, connection to"
2439 " primary node failed", instance)
2441 _ErrorIf(inst_config.admin_up and pnode_img.offline,
2442 self.EINSTANCEBADNODE, instance,
2443 "instance is marked as running and lives on offline node %s",
2444 inst_config.primary_node)
2446 # If the instance is non-redundant we cannot survive losing its primary
2447 # node, so we are not N+1 compliant. On the other hand we have no disk
2448 # templates with more than one secondary so that situation is not well
2450 # FIXME: does not support file-backed instances
2451 if not inst_config.secondary_nodes:
2452 i_non_redundant.append(instance)
2454 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2455 instance, "instance has multiple secondary nodes: %s",
2456 utils.CommaJoin(inst_config.secondary_nodes),
2457 code=self.ETYPE_WARNING)
2459 if inst_config.disk_template in constants.DTS_INT_MIRROR:
2460 pnode = inst_config.primary_node
2461 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2462 instance_groups = {}
2464 for node in instance_nodes:
2465 instance_groups.setdefault(nodeinfo_byname[node].group,
2469 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2470 # Sort so that we always list the primary node first.
2471 for group, nodes in sorted(instance_groups.items(),
2472 key=lambda (_, nodes): pnode in nodes,
2475 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2476 instance, "instance has primary and secondary nodes in"
2477 " different groups: %s", utils.CommaJoin(pretty_list),
2478 code=self.ETYPE_WARNING)
2480 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2481 i_non_a_balanced.append(instance)
2483 for snode in inst_config.secondary_nodes:
2484 s_img = node_image[snode]
2485 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2486 "instance %s, connection to secondary node failed", instance)
2489 inst_nodes_offline.append(snode)
2491 # warn that the instance lives on offline nodes
2492 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2493 "instance has offline secondary node(s) %s",
2494 utils.CommaJoin(inst_nodes_offline))
2495 # ... or ghost/non-vm_capable nodes
2496 for node in inst_config.all_nodes:
2497 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2498 "instance lives on ghost node %s", node)
2499 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2500 instance, "instance lives on non-vm_capable node %s", node)
2502 feedback_fn("* Verifying orphan volumes")
2503 reserved = utils.FieldSet(*cluster.reserved_lvs)
2504 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2506 feedback_fn("* Verifying orphan instances")
2507 self._VerifyOrphanInstances(instancelist, node_image)
2509 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2510 feedback_fn("* Verifying N+1 Memory redundancy")
2511 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2513 feedback_fn("* Other Notes")
2515 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2516 % len(i_non_redundant))
2518 if i_non_a_balanced:
2519 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2520 % len(i_non_a_balanced))
2523 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2526 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2530 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2531 """Analyze the post-hooks' result
2533 This method analyses the hook result, handles it, and sends some
2534 nicely-formatted feedback back to the user.
2536 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2537 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2538 @param hooks_results: the results of the multi-node hooks rpc call
2539 @param feedback_fn: function used send feedback back to the caller
2540 @param lu_result: previous Exec result
2541 @return: the new Exec result, based on the previous result
2545 # We only really run POST phase hooks, and are only interested in
2547 if phase == constants.HOOKS_PHASE_POST:
2548 # Used to change hooks' output to proper indentation
2549 feedback_fn("* Hooks Results")
2550 assert hooks_results, "invalid result from hooks"
2552 for node_name in hooks_results:
2553 res = hooks_results[node_name]
2555 test = msg and not res.offline
2556 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2557 "Communication failure in hooks execution: %s", msg)
2558 if res.offline or msg:
2559 # No need to investigate payload if node is offline or gave an error.
2560 # override manually lu_result here as _ErrorIf only
2561 # overrides self.bad
2564 for script, hkr, output in res.payload:
2565 test = hkr == constants.HKR_FAIL
2566 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2567 "Script %s failed, output:", script)
2569 output = self._HOOKS_INDENT_RE.sub(' ', output)
2570 feedback_fn("%s" % output)
2576 class LUClusterVerifyDisks(NoHooksLU):
2577 """Verifies the cluster disks status.
2582 def ExpandNames(self):
2583 self.needed_locks = {
2584 locking.LEVEL_NODE: locking.ALL_SET,
2585 locking.LEVEL_INSTANCE: locking.ALL_SET,
2587 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2589 def Exec(self, feedback_fn):
2590 """Verify integrity of cluster disks.
2592 @rtype: tuple of three items
2593 @return: a tuple of (dict of node-to-node_error, list of instances
2594 which need activate-disks, dict of instance: (node, volume) for
2598 result = res_nodes, res_instances, res_missing = {}, [], {}
2600 nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2601 instances = self.cfg.GetAllInstancesInfo().values()
2604 for inst in instances:
2606 if not inst.admin_up:
2608 inst.MapLVsByNode(inst_lvs)
2609 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2610 for node, vol_list in inst_lvs.iteritems():
2611 for vol in vol_list:
2612 nv_dict[(node, vol)] = inst
2617 node_lvs = self.rpc.call_lv_list(nodes, [])
2618 for node, node_res in node_lvs.items():
2619 if node_res.offline:
2621 msg = node_res.fail_msg
2623 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2624 res_nodes[node] = msg
2627 lvs = node_res.payload
2628 for lv_name, (_, _, lv_online) in lvs.items():
2629 inst = nv_dict.pop((node, lv_name), None)
2630 if (not lv_online and inst is not None
2631 and inst.name not in res_instances):
2632 res_instances.append(inst.name)
2634 # any leftover items in nv_dict are missing LVs, let's arrange the
2636 for key, inst in nv_dict.iteritems():
2637 if inst.name not in res_missing:
2638 res_missing[inst.name] = []
2639 res_missing[inst.name].append(key)
2644 class LUClusterRepairDiskSizes(NoHooksLU):
2645 """Verifies the cluster disks sizes.
2650 def ExpandNames(self):
2651 if self.op.instances:
2652 self.wanted_names = []
2653 for name in self.op.instances:
2654 full_name = _ExpandInstanceName(self.cfg, name)
2655 self.wanted_names.append(full_name)
2656 self.needed_locks = {
2657 locking.LEVEL_NODE: [],
2658 locking.LEVEL_INSTANCE: self.wanted_names,
2660 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2662 self.wanted_names = None
2663 self.needed_locks = {
2664 locking.LEVEL_NODE: locking.ALL_SET,
2665 locking.LEVEL_INSTANCE: locking.ALL_SET,
2667 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2669 def DeclareLocks(self, level):
2670 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2671 self._LockInstancesNodes(primary_only=True)
2673 def CheckPrereq(self):
2674 """Check prerequisites.
2676 This only checks the optional instance list against the existing names.
2679 if self.wanted_names is None:
2680 self.wanted_names = self.glm.list_owned(locking.LEVEL_INSTANCE)
2682 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2683 in self.wanted_names]
2685 def _EnsureChildSizes(self, disk):
2686 """Ensure children of the disk have the needed disk size.
2688 This is valid mainly for DRBD8 and fixes an issue where the
2689 children have smaller disk size.
2691 @param disk: an L{ganeti.objects.Disk} object
2694 if disk.dev_type == constants.LD_DRBD8:
2695 assert disk.children, "Empty children for DRBD8?"
2696 fchild = disk.children[0]
2697 mismatch = fchild.size < disk.size
2699 self.LogInfo("Child disk has size %d, parent %d, fixing",
2700 fchild.size, disk.size)
2701 fchild.size = disk.size
2703 # and we recurse on this child only, not on the metadev
2704 return self._EnsureChildSizes(fchild) or mismatch
2708 def Exec(self, feedback_fn):
2709 """Verify the size of cluster disks.
2712 # TODO: check child disks too
2713 # TODO: check differences in size between primary/secondary nodes
2715 for instance in self.wanted_instances:
2716 pnode = instance.primary_node
2717 if pnode not in per_node_disks:
2718 per_node_disks[pnode] = []
2719 for idx, disk in enumerate(instance.disks):
2720 per_node_disks[pnode].append((instance, idx, disk))
2723 for node, dskl in per_node_disks.items():
2724 newl = [v[2].Copy() for v in dskl]
2726 self.cfg.SetDiskID(dsk, node)
2727 result = self.rpc.call_blockdev_getsize(node, newl)
2729 self.LogWarning("Failure in blockdev_getsize call to node"
2730 " %s, ignoring", node)
2732 if len(result.payload) != len(dskl):
2733 logging.warning("Invalid result from node %s: len(dksl)=%d,"
2734 " result.payload=%s", node, len(dskl), result.payload)
2735 self.LogWarning("Invalid result from node %s, ignoring node results",
2738 for ((instance, idx, disk), size) in zip(dskl, result.payload):
2740 self.LogWarning("Disk %d of instance %s did not return size"
2741 " information, ignoring", idx, instance.name)
2743 if not isinstance(size, (int, long)):
2744 self.LogWarning("Disk %d of instance %s did not return valid"
2745 " size information, ignoring", idx, instance.name)
2748 if size != disk.size:
2749 self.LogInfo("Disk %d of instance %s has mismatched size,"
2750 " correcting: recorded %d, actual %d", idx,
2751 instance.name, disk.size, size)
2753 self.cfg.Update(instance, feedback_fn)
2754 changed.append((instance.name, idx, size))
2755 if self._EnsureChildSizes(disk):
2756 self.cfg.Update(instance, feedback_fn)
2757 changed.append((instance.name, idx, disk.size))
2761 class LUClusterRename(LogicalUnit):
2762 """Rename the cluster.
2765 HPATH = "cluster-rename"
2766 HTYPE = constants.HTYPE_CLUSTER
2768 def BuildHooksEnv(self):
2773 "OP_TARGET": self.cfg.GetClusterName(),
2774 "NEW_NAME": self.op.name,
2777 def BuildHooksNodes(self):
2778 """Build hooks nodes.
2781 return ([self.cfg.GetMasterNode()], self.cfg.GetNodeList())
2783 def CheckPrereq(self):
2784 """Verify that the passed name is a valid one.
2787 hostname = netutils.GetHostname(name=self.op.name,
2788 family=self.cfg.GetPrimaryIPFamily())
2790 new_name = hostname.name
2791 self.ip = new_ip = hostname.ip
2792 old_name = self.cfg.GetClusterName()
2793 old_ip = self.cfg.GetMasterIP()
2794 if new_name == old_name and new_ip == old_ip:
2795 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2796 " cluster has changed",
2798 if new_ip != old_ip:
2799 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2800 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2801 " reachable on the network" %
2802 new_ip, errors.ECODE_NOTUNIQUE)
2804 self.op.name = new_name
2806 def Exec(self, feedback_fn):
2807 """Rename the cluster.
2810 clustername = self.op.name
2813 # shutdown the master IP
2814 master = self.cfg.GetMasterNode()
2815 result = self.rpc.call_node_stop_master(master, False)
2816 result.Raise("Could not disable the master role")
2819 cluster = self.cfg.GetClusterInfo()
2820 cluster.cluster_name = clustername
2821 cluster.master_ip = ip
2822 self.cfg.Update(cluster, feedback_fn)
2824 # update the known hosts file
2825 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2826 node_list = self.cfg.GetOnlineNodeList()
2828 node_list.remove(master)
2831 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2833 result = self.rpc.call_node_start_master(master, False, False)
2834 msg = result.fail_msg
2836 self.LogWarning("Could not re-enable the master role on"
2837 " the master, please restart manually: %s", msg)
2842 class LUClusterSetParams(LogicalUnit):
2843 """Change the parameters of the cluster.
2846 HPATH = "cluster-modify"
2847 HTYPE = constants.HTYPE_CLUSTER
2850 def CheckArguments(self):
2854 if self.op.uid_pool:
2855 uidpool.CheckUidPool(self.op.uid_pool)
2857 if self.op.add_uids:
2858 uidpool.CheckUidPool(self.op.add_uids)
2860 if self.op.remove_uids:
2861 uidpool.CheckUidPool(self.op.remove_uids)
2863 def ExpandNames(self):
2864 # FIXME: in the future maybe other cluster params won't require checking on
2865 # all nodes to be modified.
2866 self.needed_locks = {
2867 locking.LEVEL_NODE: locking.ALL_SET,
2869 self.share_locks[locking.LEVEL_NODE] = 1
2871 def BuildHooksEnv(self):
2876 "OP_TARGET": self.cfg.GetClusterName(),
2877 "NEW_VG_NAME": self.op.vg_name,
2880 def BuildHooksNodes(self):
2881 """Build hooks nodes.
2884 mn = self.cfg.GetMasterNode()
2887 def CheckPrereq(self):
2888 """Check prerequisites.
2890 This checks whether the given params don't conflict and
2891 if the given volume group is valid.
2894 if self.op.vg_name is not None and not self.op.vg_name:
2895 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2896 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2897 " instances exist", errors.ECODE_INVAL)
2899 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2900 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2901 raise errors.OpPrereqError("Cannot disable drbd helper while"
2902 " drbd-based instances exist",
2905 node_list = self.glm.list_owned(locking.LEVEL_NODE)
2907 # if vg_name not None, checks given volume group on all nodes
2909 vglist = self.rpc.call_vg_list(node_list)
2910 for node in node_list:
2911 msg = vglist[node].fail_msg
2913 # ignoring down node
2914 self.LogWarning("Error while gathering data on node %s"
2915 " (ignoring node): %s", node, msg)
2917 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2919 constants.MIN_VG_SIZE)
2921 raise errors.OpPrereqError("Error on node '%s': %s" %
2922 (node, vgstatus), errors.ECODE_ENVIRON)
2924 if self.op.drbd_helper:
2925 # checks given drbd helper on all nodes
2926 helpers = self.rpc.call_drbd_helper(node_list)
2927 for node in node_list:
2928 ninfo = self.cfg.GetNodeInfo(node)
2930 self.LogInfo("Not checking drbd helper on offline node %s", node)
2932 msg = helpers[node].fail_msg
2934 raise errors.OpPrereqError("Error checking drbd helper on node"
2935 " '%s': %s" % (node, msg),
2936 errors.ECODE_ENVIRON)
2937 node_helper = helpers[node].payload
2938 if node_helper != self.op.drbd_helper:
2939 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2940 (node, node_helper), errors.ECODE_ENVIRON)
2942 self.cluster = cluster = self.cfg.GetClusterInfo()
2943 # validate params changes
2944 if self.op.beparams:
2945 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2946 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2948 if self.op.ndparams:
2949 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2950 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2952 # TODO: we need a more general way to handle resetting
2953 # cluster-level parameters to default values
2954 if self.new_ndparams["oob_program"] == "":
2955 self.new_ndparams["oob_program"] = \
2956 constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
2958 if self.op.nicparams:
2959 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2960 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2961 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2964 # check all instances for consistency
2965 for instance in self.cfg.GetAllInstancesInfo().values():
2966 for nic_idx, nic in enumerate(instance.nics):
2967 params_copy = copy.deepcopy(nic.nicparams)
2968 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2970 # check parameter syntax
2972 objects.NIC.CheckParameterSyntax(params_filled)
2973 except errors.ConfigurationError, err:
2974 nic_errors.append("Instance %s, nic/%d: %s" %
2975 (instance.name, nic_idx, err))
2977 # if we're moving instances to routed, check that they have an ip
2978 target_mode = params_filled[constants.NIC_MODE]
2979 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2980 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2981 (instance.name, nic_idx))
2983 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2984 "\n".join(nic_errors))
2986 # hypervisor list/parameters
2987 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2988 if self.op.hvparams:
2989 for hv_name, hv_dict in self.op.hvparams.items():
2990 if hv_name not in self.new_hvparams:
2991 self.new_hvparams[hv_name] = hv_dict
2993 self.new_hvparams[hv_name].update(hv_dict)
2995 # os hypervisor parameters
2996 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2998 for os_name, hvs in self.op.os_hvp.items():
2999 if os_name not in self.new_os_hvp:
3000 self.new_os_hvp[os_name] = hvs
3002 for hv_name, hv_dict in hvs.items():
3003 if hv_name not in self.new_os_hvp[os_name]:
3004 self.new_os_hvp[os_name][hv_name] = hv_dict
3006 self.new_os_hvp[os_name][hv_name].update(hv_dict)
3009 self.new_osp = objects.FillDict(cluster.osparams, {})
3010 if self.op.osparams:
3011 for os_name, osp in self.op.osparams.items():
3012 if os_name not in self.new_osp:
3013 self.new_osp[os_name] = {}
3015 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
3018 if not self.new_osp[os_name]:
3019 # we removed all parameters
3020 del self.new_osp[os_name]
3022 # check the parameter validity (remote check)
3023 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
3024 os_name, self.new_osp[os_name])
3026 # changes to the hypervisor list
3027 if self.op.enabled_hypervisors is not None:
3028 self.hv_list = self.op.enabled_hypervisors
3029 for hv in self.hv_list:
3030 # if the hypervisor doesn't already exist in the cluster
3031 # hvparams, we initialize it to empty, and then (in both
3032 # cases) we make sure to fill the defaults, as we might not
3033 # have a complete defaults list if the hypervisor wasn't
3035 if hv not in new_hvp:
3037 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
3038 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
3040 self.hv_list = cluster.enabled_hypervisors
3042 if self.op.hvparams or self.op.enabled_hypervisors is not None:
3043 # either the enabled list has changed, or the parameters have, validate
3044 for hv_name, hv_params in self.new_hvparams.items():
3045 if ((self.op.hvparams and hv_name in self.op.hvparams) or
3046 (self.op.enabled_hypervisors and
3047 hv_name in self.op.enabled_hypervisors)):
3048 # either this is a new hypervisor, or its parameters have changed
3049 hv_class = hypervisor.GetHypervisor(hv_name)
3050 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3051 hv_class.CheckParameterSyntax(hv_params)
3052 _CheckHVParams(self, node_list, hv_name, hv_params)
3055 # no need to check any newly-enabled hypervisors, since the
3056 # defaults have already been checked in the above code-block
3057 for os_name, os_hvp in self.new_os_hvp.items():
3058 for hv_name, hv_params in os_hvp.items():
3059 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
3060 # we need to fill in the new os_hvp on top of the actual hv_p
3061 cluster_defaults = self.new_hvparams.get(hv_name, {})
3062 new_osp = objects.FillDict(cluster_defaults, hv_params)
3063 hv_class = hypervisor.GetHypervisor(hv_name)
3064 hv_class.CheckParameterSyntax(new_osp)
3065 _CheckHVParams(self, node_list, hv_name, new_osp)
3067 if self.op.default_iallocator:
3068 alloc_script = utils.FindFile(self.op.default_iallocator,
3069 constants.IALLOCATOR_SEARCH_PATH,
3071 if alloc_script is None:
3072 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
3073 " specified" % self.op.default_iallocator,
3076 def Exec(self, feedback_fn):
3077 """Change the parameters of the cluster.
3080 if self.op.vg_name is not None:
3081 new_volume = self.op.vg_name
3084 if new_volume != self.cfg.GetVGName():
3085 self.cfg.SetVGName(new_volume)
3087 feedback_fn("Cluster LVM configuration already in desired"
3088 " state, not changing")
3089 if self.op.drbd_helper is not None:
3090 new_helper = self.op.drbd_helper
3093 if new_helper != self.cfg.GetDRBDHelper():
3094 self.cfg.SetDRBDHelper(new_helper)
3096 feedback_fn("Cluster DRBD helper already in desired state,"
3098 if self.op.hvparams:
3099 self.cluster.hvparams = self.new_hvparams
3101 self.cluster.os_hvp = self.new_os_hvp
3102 if self.op.enabled_hypervisors is not None:
3103 self.cluster.hvparams = self.new_hvparams
3104 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
3105 if self.op.beparams:
3106 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
3107 if self.op.nicparams:
3108 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
3109 if self.op.osparams:
3110 self.cluster.osparams = self.new_osp
3111 if self.op.ndparams:
3112 self.cluster.ndparams = self.new_ndparams
3114 if self.op.candidate_pool_size is not None:
3115 self.cluster.candidate_pool_size = self.op.candidate_pool_size
3116 # we need to update the pool size here, otherwise the save will fail
3117 _AdjustCandidatePool(self, [])
3119 if self.op.maintain_node_health is not None:
3120 self.cluster.maintain_node_health = self.op.maintain_node_health
3122 if self.op.prealloc_wipe_disks is not None:
3123 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
3125 if self.op.add_uids is not None:
3126 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
3128 if self.op.remove_uids is not None:
3129 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
3131 if self.op.uid_pool is not None:
3132 self.cluster.uid_pool = self.op.uid_pool
3134 if self.op.default_iallocator is not None:
3135 self.cluster.default_iallocator = self.op.default_iallocator
3137 if self.op.reserved_lvs is not None:
3138 self.cluster.reserved_lvs = self.op.reserved_lvs
3140 def helper_os(aname, mods, desc):
3142 lst = getattr(self.cluster, aname)
3143 for key, val in mods:
3144 if key == constants.DDM_ADD:
3146 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3149 elif key == constants.DDM_REMOVE:
3153 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3155 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3157 if self.op.hidden_os:
3158 helper_os("hidden_os", self.op.hidden_os, "hidden")
3160 if self.op.blacklisted_os:
3161 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3163 if self.op.master_netdev:
3164 master = self.cfg.GetMasterNode()
3165 feedback_fn("Shutting down master ip on the current netdev (%s)" %
3166 self.cluster.master_netdev)
3167 result = self.rpc.call_node_stop_master(master, False)
3168 result.Raise("Could not disable the master ip")
3169 feedback_fn("Changing master_netdev from %s to %s" %
3170 (self.cluster.master_netdev, self.op.master_netdev))
3171 self.cluster.master_netdev = self.op.master_netdev
3173 self.cfg.Update(self.cluster, feedback_fn)
3175 if self.op.master_netdev:
3176 feedback_fn("Starting the master ip on the new master netdev (%s)" %
3177 self.op.master_netdev)
3178 result = self.rpc.call_node_start_master(master, False, False)
3180 self.LogWarning("Could not re-enable the master ip on"
3181 " the master, please restart manually: %s",
3185 def _UploadHelper(lu, nodes, fname):
3186 """Helper for uploading a file and showing warnings.
3189 if os.path.exists(fname):
3190 result = lu.rpc.call_upload_file(nodes, fname)
3191 for to_node, to_result in result.items():
3192 msg = to_result.fail_msg
3194 msg = ("Copy of file %s to node %s failed: %s" %
3195 (fname, to_node, msg))
3196 lu.proc.LogWarning(msg)
3199 def _ComputeAncillaryFiles(cluster, redist):
3200 """Compute files external to Ganeti which need to be consistent.
3202 @type redist: boolean
3203 @param redist: Whether to include files which need to be redistributed
3206 # Compute files for all nodes
3208 constants.SSH_KNOWN_HOSTS_FILE,
3209 constants.CONFD_HMAC_KEY,
3210 constants.CLUSTER_DOMAIN_SECRET_FILE,
3214 files_all.update(constants.ALL_CERT_FILES)
3215 files_all.update(ssconf.SimpleStore().GetFileList())
3217 if cluster.modify_etc_hosts:
3218 files_all.add(constants.ETC_HOSTS)
3220 # Files which must either exist on all nodes or on none
3221 files_all_opt = set([
3222 constants.RAPI_USERS_FILE,
3225 # Files which should only be on master candidates
3228 files_mc.add(constants.CLUSTER_CONF_FILE)
3230 # Files which should only be on VM-capable nodes
3231 files_vm = set(filename
3232 for hv_name in cluster.enabled_hypervisors
3233 for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles())
3235 # Filenames must be unique
3236 assert (len(files_all | files_all_opt | files_mc | files_vm) ==
3237 sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
3238 "Found file listed in more than one file list"
3240 return (files_all, files_all_opt, files_mc, files_vm)
3243 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3244 """Distribute additional files which are part of the cluster configuration.
3246 ConfigWriter takes care of distributing the config and ssconf files, but
3247 there are more files which should be distributed to all nodes. This function
3248 makes sure those are copied.
3250 @param lu: calling logical unit
3251 @param additional_nodes: list of nodes not in the config to distribute to
3252 @type additional_vm: boolean
3253 @param additional_vm: whether the additional nodes are vm-capable or not
3256 # Gather target nodes
3257 cluster = lu.cfg.GetClusterInfo()
3258 master_info = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3260 online_nodes = lu.cfg.GetOnlineNodeList()
3261 vm_nodes = lu.cfg.GetVmCapableNodeList()
3263 if additional_nodes is not None:
3264 online_nodes.extend(additional_nodes)
3266 vm_nodes.extend(additional_nodes)
3268 # Never distribute to master node
3269 for nodelist in [online_nodes, vm_nodes]:
3270 if master_info.name in nodelist:
3271 nodelist.remove(master_info.name)
3274 (files_all, files_all_opt, files_mc, files_vm) = \
3275 _ComputeAncillaryFiles(cluster, True)
3277 # Never re-distribute configuration file from here
3278 assert not (constants.CLUSTER_CONF_FILE in files_all or
3279 constants.CLUSTER_CONF_FILE in files_vm)
3280 assert not files_mc, "Master candidates not handled in this function"
3283 (online_nodes, files_all),
3284 (online_nodes, files_all_opt),
3285 (vm_nodes, files_vm),
3289 for (node_list, files) in filemap:
3291 _UploadHelper(lu, node_list, fname)
3294 class LUClusterRedistConf(NoHooksLU):
3295 """Force the redistribution of cluster configuration.
3297 This is a very simple LU.
3302 def ExpandNames(self):
3303 self.needed_locks = {
3304 locking.LEVEL_NODE: locking.ALL_SET,
3306 self.share_locks[locking.LEVEL_NODE] = 1
3308 def Exec(self, feedback_fn):
3309 """Redistribute the configuration.
3312 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3313 _RedistributeAncillaryFiles(self)
3316 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3317 """Sleep and poll for an instance's disk to sync.
3320 if not instance.disks or disks is not None and not disks:
3323 disks = _ExpandCheckDisks(instance, disks)
3326 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3328 node = instance.primary_node
3331 lu.cfg.SetDiskID(dev, node)
3333 # TODO: Convert to utils.Retry
3336 degr_retries = 10 # in seconds, as we sleep 1 second each time
3340 cumul_degraded = False
3341 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3342 msg = rstats.fail_msg
3344 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3347 raise errors.RemoteError("Can't contact node %s for mirror data,"
3348 " aborting." % node)
3351 rstats = rstats.payload
3353 for i, mstat in enumerate(rstats):
3355 lu.LogWarning("Can't compute data for node %s/%s",
3356 node, disks[i].iv_name)
3359 cumul_degraded = (cumul_degraded or
3360 (mstat.is_degraded and mstat.sync_percent is None))
3361 if mstat.sync_percent is not None:
3363 if mstat.estimated_time is not None:
3364 rem_time = ("%s remaining (estimated)" %
3365 utils.FormatSeconds(mstat.estimated_time))
3366 max_time = mstat.estimated_time
3368 rem_time = "no time estimate"
3369 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3370 (disks[i].iv_name, mstat.sync_percent, rem_time))
3372 # if we're done but degraded, let's do a few small retries, to
3373 # make sure we see a stable and not transient situation; therefore
3374 # we force restart of the loop
3375 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3376 logging.info("Degraded disks found, %d retries left", degr_retries)
3384 time.sleep(min(60, max_time))
3387 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3388 return not cumul_degraded
3391 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3392 """Check that mirrors are not degraded.
3394 The ldisk parameter, if True, will change the test from the
3395 is_degraded attribute (which represents overall non-ok status for
3396 the device(s)) to the ldisk (representing the local storage status).
3399 lu.cfg.SetDiskID(dev, node)
3403 if on_primary or dev.AssembleOnSecondary():
3404 rstats = lu.rpc.call_blockdev_find(node, dev)
3405 msg = rstats.fail_msg
3407 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3409 elif not rstats.payload:
3410 lu.LogWarning("Can't find disk on node %s", node)
3414 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3416 result = result and not rstats.payload.is_degraded
3419 for child in dev.children:
3420 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3425 class LUOobCommand(NoHooksLU):
3426 """Logical unit for OOB handling.
3430 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
3432 def CheckPrereq(self):
3433 """Check prerequisites.
3436 - the node exists in the configuration
3439 Any errors are signaled by raising errors.OpPrereqError.
3443 self.master_node = self.cfg.GetMasterNode()
3445 assert self.op.power_delay >= 0.0
3447 if self.op.node_names:
3448 if (self.op.command in self._SKIP_MASTER and
3449 self.master_node in self.op.node_names):
3450 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
3451 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
3453 if master_oob_handler:
3454 additional_text = ("run '%s %s %s' if you want to operate on the"
3455 " master regardless") % (master_oob_handler,
3459 additional_text = "it does not support out-of-band operations"
3461 raise errors.OpPrereqError(("Operating on the master node %s is not"
3462 " allowed for %s; %s") %
3463 (self.master_node, self.op.command,
3464 additional_text), errors.ECODE_INVAL)
3466 self.op.node_names = self.cfg.GetNodeList()
3467 if self.op.command in self._SKIP_MASTER:
3468 self.op.node_names.remove(self.master_node)
3470 if self.op.command in self._SKIP_MASTER:
3471 assert self.master_node not in self.op.node_names
3473 for node_name in self.op.node_names:
3474 node = self.cfg.GetNodeInfo(node_name)
3477 raise errors.OpPrereqError("Node %s not found" % node_name,
3480 self.nodes.append(node)
3482 if (not self.op.ignore_status and
3483 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
3484 raise errors.OpPrereqError(("Cannot power off node %s because it is"
3485 " not marked offline") % node_name,
3488 def ExpandNames(self):
3489 """Gather locks we need.
3492 if self.op.node_names:
3493 self.op.node_names = [_ExpandNodeName(self.cfg, name)
3494 for name in self.op.node_names]
3495 lock_names = self.op.node_names
3497 lock_names = locking.ALL_SET
3499 self.needed_locks = {
3500 locking.LEVEL_NODE: lock_names,
3503 def Exec(self, feedback_fn):
3504 """Execute OOB and return result if we expect any.
3507 master_node = self.master_node
3510 for idx, node in enumerate(self.nodes):
3511 node_entry = [(constants.RS_NORMAL, node.name)]
3512 ret.append(node_entry)
3514 oob_program = _SupportsOob(self.cfg, node)
3517 node_entry.append((constants.RS_UNAVAIL, None))
3520 logging.info("Executing out-of-band command '%s' using '%s' on %s",
3521 self.op.command, oob_program, node.name)
3522 result = self.rpc.call_run_oob(master_node, oob_program,
3523 self.op.command, node.name,
3527 self.LogWarning("Out-of-band RPC failed on node '%s': %s",
3528 node.name, result.fail_msg)
3529 node_entry.append((constants.RS_NODATA, None))
3532 self._CheckPayload(result)
3533 except errors.OpExecError, err:
3534 self.LogWarning("Payload returned by node '%s' is not valid: %s",
3536 node_entry.append((constants.RS_NODATA, None))
3538 if self.op.command == constants.OOB_HEALTH:
3539 # For health we should log important events
3540 for item, status in result.payload:
3541 if status in [constants.OOB_STATUS_WARNING,
3542 constants.OOB_STATUS_CRITICAL]:
3543 self.LogWarning("Item '%s' on node '%s' has status '%s'",
3544 item, node.name, status)
3546 if self.op.command == constants.OOB_POWER_ON:
3548 elif self.op.command == constants.OOB_POWER_OFF:
3549 node.powered = False
3550 elif self.op.command == constants.OOB_POWER_STATUS:
3551 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3552 if powered != node.powered:
3553 logging.warning(("Recorded power state (%s) of node '%s' does not"
3554 " match actual power state (%s)"), node.powered,
3557 # For configuration changing commands we should update the node
3558 if self.op.command in (constants.OOB_POWER_ON,
3559 constants.OOB_POWER_OFF):
3560 self.cfg.Update(node, feedback_fn)
3562 node_entry.append((constants.RS_NORMAL, result.payload))
3564 if (self.op.command == constants.OOB_POWER_ON and
3565 idx < len(self.nodes) - 1):
3566 time.sleep(self.op.power_delay)
3570 def _CheckPayload(self, result):
3571 """Checks if the payload is valid.
3573 @param result: RPC result
3574 @raises errors.OpExecError: If payload is not valid
3578 if self.op.command == constants.OOB_HEALTH:
3579 if not isinstance(result.payload, list):
3580 errs.append("command 'health' is expected to return a list but got %s" %
3581 type(result.payload))
3583 for item, status in result.payload:
3584 if status not in constants.OOB_STATUSES:
3585 errs.append("health item '%s' has invalid status '%s'" %
3588 if self.op.command == constants.OOB_POWER_STATUS:
3589 if not isinstance(result.payload, dict):
3590 errs.append("power-status is expected to return a dict but got %s" %
3591 type(result.payload))
3593 if self.op.command in [
3594 constants.OOB_POWER_ON,
3595 constants.OOB_POWER_OFF,
3596 constants.OOB_POWER_CYCLE,
3598 if result.payload is not None:
3599 errs.append("%s is expected to not return payload but got '%s'" %
3600 (self.op.command, result.payload))
3603 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3604 utils.CommaJoin(errs))
3606 class _OsQuery(_QueryBase):
3607 FIELDS = query.OS_FIELDS
3609 def ExpandNames(self, lu):
3610 # Lock all nodes in shared mode
3611 # Temporary removal of locks, should be reverted later
3612 # TODO: reintroduce locks when they are lighter-weight
3613 lu.needed_locks = {}
3614 #self.share_locks[locking.LEVEL_NODE] = 1
3615 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3617 # The following variables interact with _QueryBase._GetNames
3619 self.wanted = self.names
3621 self.wanted = locking.ALL_SET
3623 self.do_locking = self.use_locking
3625 def DeclareLocks(self, lu, level):
3629 def _DiagnoseByOS(rlist):
3630 """Remaps a per-node return list into an a per-os per-node dictionary
3632 @param rlist: a map with node names as keys and OS objects as values
3635 @return: a dictionary with osnames as keys and as value another
3636 map, with nodes as keys and tuples of (path, status, diagnose,
3637 variants, parameters, api_versions) as values, eg::
3639 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3640 (/srv/..., False, "invalid api")],
3641 "node2": [(/srv/..., True, "", [], [])]}
3646 # we build here the list of nodes that didn't fail the RPC (at RPC
3647 # level), so that nodes with a non-responding node daemon don't
3648 # make all OSes invalid
3649 good_nodes = [node_name for node_name in rlist
3650 if not rlist[node_name].fail_msg]
3651 for node_name, nr in rlist.items():
3652 if nr.fail_msg or not nr.payload:
3654 for (name, path, status, diagnose, variants,
3655 params, api_versions) in nr.payload:
3656 if name not in all_os:
3657 # build a list of nodes for this os containing empty lists
3658 # for each node in node_list
3660 for nname in good_nodes:
3661 all_os[name][nname] = []
3662 # convert params from [name, help] to (name, help)
3663 params = [tuple(v) for v in params]
3664 all_os[name][node_name].append((path, status, diagnose,
3665 variants, params, api_versions))
3668 def _GetQueryData(self, lu):
3669 """Computes the list of nodes and their attributes.
3672 # Locking is not used
3673 assert not (compat.any(lu.glm.is_owned(level)
3674 for level in locking.LEVELS) or
3675 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.glm.list_owned(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.glm.list_owned(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.glm.list_owned(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.glm.list_owned(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.glm.remove(locking.LEVEL_INSTANCE, old_name)
5847 self.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,
6017 ignore_consistency=ignore_consistency,
6018 shutdown_timeout=shutdown_timeout)
6019 self.tasklets = [self._migrater]
6021 def DeclareLocks(self, level):
6022 if level == locking.LEVEL_NODE:
6023 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6024 if instance.disk_template in constants.DTS_EXT_MIRROR:
6025 if self.op.target_node is None:
6026 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6028 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6029 self.op.target_node]
6030 del self.recalculate_locks[locking.LEVEL_NODE]
6032 self._LockInstancesNodes()
6034 def BuildHooksEnv(self):
6037 This runs on master, primary and secondary nodes of the instance.
6040 instance = self._migrater.instance
6041 source_node = instance.primary_node
6042 target_node = self.op.target_node
6044 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
6045 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6046 "OLD_PRIMARY": source_node,
6047 "NEW_PRIMARY": target_node,
6050 if instance.disk_template in constants.DTS_INT_MIRROR:
6051 env["OLD_SECONDARY"] = instance.secondary_nodes[0]
6052 env["NEW_SECONDARY"] = source_node
6054 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = ""
6056 env.update(_BuildInstanceHookEnvByObject(self, instance))
6060 def BuildHooksNodes(self):
6061 """Build hooks nodes.
6064 instance = self._migrater.instance
6065 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6066 return (nl, nl + [instance.primary_node])
6069 class LUInstanceMigrate(LogicalUnit):
6070 """Migrate an instance.
6072 This is migration without shutting down, compared to the failover,
6073 which is done with shutdown.
6076 HPATH = "instance-migrate"
6077 HTYPE = constants.HTYPE_INSTANCE
6080 def ExpandNames(self):
6081 self._ExpandAndLockInstance()
6083 if self.op.target_node is not None:
6084 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6086 self.needed_locks[locking.LEVEL_NODE] = []
6087 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6089 self._migrater = TLMigrateInstance(self, self.op.instance_name,
6090 cleanup=self.op.cleanup,
6092 fallback=self.op.allow_failover)
6093 self.tasklets = [self._migrater]
6095 def DeclareLocks(self, level):
6096 if level == locking.LEVEL_NODE:
6097 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
6098 if instance.disk_template in constants.DTS_EXT_MIRROR:
6099 if self.op.target_node is None:
6100 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6102 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
6103 self.op.target_node]
6104 del self.recalculate_locks[locking.LEVEL_NODE]
6106 self._LockInstancesNodes()
6108 def BuildHooksEnv(self):
6111 This runs on master, primary and secondary nodes of the instance.
6114 instance = self._migrater.instance
6115 source_node = instance.primary_node
6116 target_node = self.op.target_node
6117 env = _BuildInstanceHookEnvByObject(self, instance)
6119 "MIGRATE_LIVE": self._migrater.live,
6120 "MIGRATE_CLEANUP": self.op.cleanup,
6121 "OLD_PRIMARY": source_node,
6122 "NEW_PRIMARY": target_node,
6125 if instance.disk_template in constants.DTS_INT_MIRROR:
6126 env["OLD_SECONDARY"] = target_node
6127 env["NEW_SECONDARY"] = source_node
6129 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
6133 def BuildHooksNodes(self):
6134 """Build hooks nodes.
6137 instance = self._migrater.instance
6138 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
6139 return (nl, nl + [instance.primary_node])
6142 class LUInstanceMove(LogicalUnit):
6143 """Move an instance by data-copying.
6146 HPATH = "instance-move"
6147 HTYPE = constants.HTYPE_INSTANCE
6150 def ExpandNames(self):
6151 self._ExpandAndLockInstance()
6152 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
6153 self.op.target_node = target_node
6154 self.needed_locks[locking.LEVEL_NODE] = [target_node]
6155 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6157 def DeclareLocks(self, level):
6158 if level == locking.LEVEL_NODE:
6159 self._LockInstancesNodes(primary_only=True)
6161 def BuildHooksEnv(self):
6164 This runs on master, primary and secondary nodes of the instance.
6168 "TARGET_NODE": self.op.target_node,
6169 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
6171 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6174 def BuildHooksNodes(self):
6175 """Build hooks nodes.
6179 self.cfg.GetMasterNode(),
6180 self.instance.primary_node,
6181 self.op.target_node,
6185 def CheckPrereq(self):
6186 """Check prerequisites.
6188 This checks that the instance is in the cluster.
6191 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6192 assert self.instance is not None, \
6193 "Cannot retrieve locked instance %s" % self.op.instance_name
6195 node = self.cfg.GetNodeInfo(self.op.target_node)
6196 assert node is not None, \
6197 "Cannot retrieve locked node %s" % self.op.target_node
6199 self.target_node = target_node = node.name
6201 if target_node == instance.primary_node:
6202 raise errors.OpPrereqError("Instance %s is already on the node %s" %
6203 (instance.name, target_node),
6206 bep = self.cfg.GetClusterInfo().FillBE(instance)
6208 for idx, dsk in enumerate(instance.disks):
6209 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
6210 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
6211 " cannot copy" % idx, errors.ECODE_STATE)
6213 _CheckNodeOnline(self, target_node)
6214 _CheckNodeNotDrained(self, target_node)
6215 _CheckNodeVmCapable(self, target_node)
6217 if instance.admin_up:
6218 # check memory requirements on the secondary node
6219 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
6220 instance.name, bep[constants.BE_MEMORY],
6221 instance.hypervisor)
6223 self.LogInfo("Not checking memory on the secondary node as"
6224 " instance will not be started")
6226 # check bridge existance
6227 _CheckInstanceBridgesExist(self, instance, node=target_node)
6229 def Exec(self, feedback_fn):
6230 """Move an instance.
6232 The move is done by shutting it down on its present node, copying
6233 the data over (slow) and starting it on the new node.
6236 instance = self.instance
6238 source_node = instance.primary_node
6239 target_node = self.target_node
6241 self.LogInfo("Shutting down instance %s on source node %s",
6242 instance.name, source_node)
6244 result = self.rpc.call_instance_shutdown(source_node, instance,
6245 self.op.shutdown_timeout)
6246 msg = result.fail_msg
6248 if self.op.ignore_consistency:
6249 self.proc.LogWarning("Could not shutdown instance %s on node %s."
6250 " Proceeding anyway. Please make sure node"
6251 " %s is down. Error details: %s",
6252 instance.name, source_node, source_node, msg)
6254 raise errors.OpExecError("Could not shutdown instance %s on"
6256 (instance.name, source_node, msg))
6258 # create the target disks
6260 _CreateDisks(self, instance, target_node=target_node)
6261 except errors.OpExecError:
6262 self.LogWarning("Device creation failed, reverting...")
6264 _RemoveDisks(self, instance, target_node=target_node)
6266 self.cfg.ReleaseDRBDMinors(instance.name)
6269 cluster_name = self.cfg.GetClusterInfo().cluster_name
6272 # activate, get path, copy the data over
6273 for idx, disk in enumerate(instance.disks):
6274 self.LogInfo("Copying data for disk %d", idx)
6275 result = self.rpc.call_blockdev_assemble(target_node, disk,
6276 instance.name, True, idx)
6278 self.LogWarning("Can't assemble newly created disk %d: %s",
6279 idx, result.fail_msg)
6280 errs.append(result.fail_msg)
6282 dev_path = result.payload
6283 result = self.rpc.call_blockdev_export(source_node, disk,
6284 target_node, dev_path,
6287 self.LogWarning("Can't copy data over for disk %d: %s",
6288 idx, result.fail_msg)
6289 errs.append(result.fail_msg)
6293 self.LogWarning("Some disks failed to copy, aborting")
6295 _RemoveDisks(self, instance, target_node=target_node)
6297 self.cfg.ReleaseDRBDMinors(instance.name)
6298 raise errors.OpExecError("Errors during disk copy: %s" %
6301 instance.primary_node = target_node
6302 self.cfg.Update(instance, feedback_fn)
6304 self.LogInfo("Removing the disks on the original node")
6305 _RemoveDisks(self, instance, target_node=source_node)
6307 # Only start the instance if it's marked as up
6308 if instance.admin_up:
6309 self.LogInfo("Starting instance %s on node %s",
6310 instance.name, target_node)
6312 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6313 ignore_secondaries=True)
6315 _ShutdownInstanceDisks(self, instance)
6316 raise errors.OpExecError("Can't activate the instance's disks")
6318 result = self.rpc.call_instance_start(target_node, instance, None, None)
6319 msg = result.fail_msg
6321 _ShutdownInstanceDisks(self, instance)
6322 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6323 (instance.name, target_node, msg))
6326 class LUNodeMigrate(LogicalUnit):
6327 """Migrate all instances from a node.
6330 HPATH = "node-migrate"
6331 HTYPE = constants.HTYPE_NODE
6334 def CheckArguments(self):
6335 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
6337 def ExpandNames(self):
6338 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6340 self.needed_locks = {}
6342 # Create tasklets for migrating instances for all instances on this node
6346 self.lock_all_nodes = False
6348 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
6349 logging.debug("Migrating instance %s", inst.name)
6350 names.append(inst.name)
6352 tasklets.append(TLMigrateInstance(self, inst.name, cleanup=False))
6354 if inst.disk_template in constants.DTS_EXT_MIRROR:
6355 # We need to lock all nodes, as the iallocator will choose the
6356 # destination nodes afterwards
6357 self.lock_all_nodes = True
6359 self.tasklets = tasklets
6361 # Declare node locks
6362 if self.lock_all_nodes:
6363 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6365 self.needed_locks[locking.LEVEL_NODE] = [self.op.node_name]
6366 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6368 # Declare instance locks
6369 self.needed_locks[locking.LEVEL_INSTANCE] = names
6371 def DeclareLocks(self, level):
6372 if level == locking.LEVEL_NODE and not self.lock_all_nodes:
6373 self._LockInstancesNodes()
6375 def BuildHooksEnv(self):
6378 This runs on the master, the primary and all the secondaries.
6382 "NODE_NAME": self.op.node_name,
6385 def BuildHooksNodes(self):
6386 """Build hooks nodes.
6389 nl = [self.cfg.GetMasterNode()]
6393 class TLMigrateInstance(Tasklet):
6394 """Tasklet class for instance migration.
6397 @ivar live: whether the migration will be done live or non-live;
6398 this variable is initalized only after CheckPrereq has run
6399 @type cleanup: boolean
6400 @ivar cleanup: Wheater we cleanup from a failed migration
6401 @type iallocator: string
6402 @ivar iallocator: The iallocator used to determine target_node
6403 @type target_node: string
6404 @ivar target_node: If given, the target_node to reallocate the instance to
6405 @type failover: boolean
6406 @ivar failover: Whether operation results in failover or migration
6407 @type fallback: boolean
6408 @ivar fallback: Whether fallback to failover is allowed if migration not
6410 @type ignore_consistency: boolean
6411 @ivar ignore_consistency: Wheter we should ignore consistency between source
6413 @type shutdown_timeout: int
6414 @ivar shutdown_timeout: In case of failover timeout of the shutdown
6417 def __init__(self, lu, instance_name, cleanup=False,
6418 failover=False, fallback=False,
6419 ignore_consistency=False,
6420 shutdown_timeout=constants.DEFAULT_SHUTDOWN_TIMEOUT):
6421 """Initializes this class.
6424 Tasklet.__init__(self, lu)
6427 self.instance_name = instance_name
6428 self.cleanup = cleanup
6429 self.live = False # will be overridden later
6430 self.failover = failover
6431 self.fallback = fallback
6432 self.ignore_consistency = ignore_consistency
6433 self.shutdown_timeout = shutdown_timeout
6435 def CheckPrereq(self):
6436 """Check prerequisites.
6438 This checks that the instance is in the cluster.
6441 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
6442 instance = self.cfg.GetInstanceInfo(instance_name)
6443 assert instance is not None
6444 self.instance = instance
6446 if (not self.cleanup and not instance.admin_up and not self.failover and
6448 self.lu.LogInfo("Instance is marked down, fallback allowed, switching"
6450 self.failover = True
6452 if instance.disk_template not in constants.DTS_MIRRORED:
6457 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
6458 " %s" % (instance.disk_template, text),
6461 if instance.disk_template in constants.DTS_EXT_MIRROR:
6462 _CheckIAllocatorOrNode(self.lu, "iallocator", "target_node")
6464 if self.lu.op.iallocator:
6465 self._RunAllocator()
6467 # We set set self.target_node as it is required by
6469 self.target_node = self.lu.op.target_node
6471 # self.target_node is already populated, either directly or by the
6473 target_node = self.target_node
6475 if len(self.lu.tasklets) == 1:
6476 # It is safe to release locks only when we're the only tasklet
6478 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
6479 keep=[instance.primary_node, self.target_node])
6482 secondary_nodes = instance.secondary_nodes
6483 if not secondary_nodes:
6484 raise errors.ConfigurationError("No secondary node but using"
6485 " %s disk template" %
6486 instance.disk_template)
6487 target_node = secondary_nodes[0]
6488 if self.lu.op.iallocator or (self.lu.op.target_node and
6489 self.lu.op.target_node != target_node):
6491 text = "failed over"
6494 raise errors.OpPrereqError("Instances with disk template %s cannot"
6495 " be %s to arbitrary nodes"
6496 " (neither an iallocator nor a target"
6497 " node can be passed)" %
6498 (instance.disk_template, text),
6501 i_be = self.cfg.GetClusterInfo().FillBE(instance)
6503 # check memory requirements on the secondary node
6504 if not self.failover or instance.admin_up:
6505 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
6506 instance.name, i_be[constants.BE_MEMORY],
6507 instance.hypervisor)
6509 self.lu.LogInfo("Not checking memory on the secondary node as"
6510 " instance will not be started")
6512 # check bridge existance
6513 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
6515 if not self.cleanup:
6516 _CheckNodeNotDrained(self.lu, target_node)
6517 if not self.failover:
6518 result = self.rpc.call_instance_migratable(instance.primary_node,
6520 if result.fail_msg and self.fallback:
6521 self.lu.LogInfo("Can't migrate, instance offline, fallback to"
6523 self.failover = True
6525 result.Raise("Can't migrate, please use failover",
6526 prereq=True, ecode=errors.ECODE_STATE)
6528 assert not (self.failover and self.cleanup)
6530 if not self.failover:
6531 if self.lu.op.live is not None and self.lu.op.mode is not None:
6532 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
6533 " parameters are accepted",
6535 if self.lu.op.live is not None:
6537 self.lu.op.mode = constants.HT_MIGRATION_LIVE
6539 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
6540 # reset the 'live' parameter to None so that repeated
6541 # invocations of CheckPrereq do not raise an exception
6542 self.lu.op.live = None
6543 elif self.lu.op.mode is None:
6544 # read the default value from the hypervisor
6545 i_hv = self.cfg.GetClusterInfo().FillHV(self.instance,
6547 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
6549 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
6551 # Failover is never live
6554 def _RunAllocator(self):
6555 """Run the allocator based on input opcode.
6558 ial = IAllocator(self.cfg, self.rpc,
6559 mode=constants.IALLOCATOR_MODE_RELOC,
6560 name=self.instance_name,
6561 # TODO See why hail breaks with a single node below
6562 relocate_from=[self.instance.primary_node,
6563 self.instance.primary_node],
6566 ial.Run(self.lu.op.iallocator)
6569 raise errors.OpPrereqError("Can't compute nodes using"
6570 " iallocator '%s': %s" %
6571 (self.lu.op.iallocator, ial.info),
6573 if len(ial.result) != ial.required_nodes:
6574 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6575 " of nodes (%s), required %s" %
6576 (self.lu.op.iallocator, len(ial.result),
6577 ial.required_nodes), errors.ECODE_FAULT)
6578 self.target_node = ial.result[0]
6579 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6580 self.instance_name, self.lu.op.iallocator,
6581 utils.CommaJoin(ial.result))
6583 def _WaitUntilSync(self):
6584 """Poll with custom rpc for disk sync.
6586 This uses our own step-based rpc call.
6589 self.feedback_fn("* wait until resync is done")
6593 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
6595 self.instance.disks)
6597 for node, nres in result.items():
6598 nres.Raise("Cannot resync disks on node %s" % node)
6599 node_done, node_percent = nres.payload
6600 all_done = all_done and node_done
6601 if node_percent is not None:
6602 min_percent = min(min_percent, node_percent)
6604 if min_percent < 100:
6605 self.feedback_fn(" - progress: %.1f%%" % min_percent)
6608 def _EnsureSecondary(self, node):
6609 """Demote a node to secondary.
6612 self.feedback_fn("* switching node %s to secondary mode" % node)
6614 for dev in self.instance.disks:
6615 self.cfg.SetDiskID(dev, node)
6617 result = self.rpc.call_blockdev_close(node, self.instance.name,
6618 self.instance.disks)
6619 result.Raise("Cannot change disk to secondary on node %s" % node)
6621 def _GoStandalone(self):
6622 """Disconnect from the network.
6625 self.feedback_fn("* changing into standalone mode")
6626 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
6627 self.instance.disks)
6628 for node, nres in result.items():
6629 nres.Raise("Cannot disconnect disks node %s" % node)
6631 def _GoReconnect(self, multimaster):
6632 """Reconnect to the network.
6638 msg = "single-master"
6639 self.feedback_fn("* changing disks into %s mode" % msg)
6640 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
6641 self.instance.disks,
6642 self.instance.name, multimaster)
6643 for node, nres in result.items():
6644 nres.Raise("Cannot change disks config on node %s" % node)
6646 def _ExecCleanup(self):
6647 """Try to cleanup after a failed migration.
6649 The cleanup is done by:
6650 - check that the instance is running only on one node
6651 (and update the config if needed)
6652 - change disks on its secondary node to secondary
6653 - wait until disks are fully synchronized
6654 - disconnect from the network
6655 - change disks into single-master mode
6656 - wait again until disks are fully synchronized
6659 instance = self.instance
6660 target_node = self.target_node
6661 source_node = self.source_node
6663 # check running on only one node
6664 self.feedback_fn("* checking where the instance actually runs"
6665 " (if this hangs, the hypervisor might be in"
6667 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6668 for node, result in ins_l.items():
6669 result.Raise("Can't contact node %s" % node)
6671 runningon_source = instance.name in ins_l[source_node].payload
6672 runningon_target = instance.name in ins_l[target_node].payload
6674 if runningon_source and runningon_target:
6675 raise errors.OpExecError("Instance seems to be running on two nodes,"
6676 " or the hypervisor is confused; you will have"
6677 " to ensure manually that it runs only on one"
6678 " and restart this operation")
6680 if not (runningon_source or runningon_target):
6681 raise errors.OpExecError("Instance does not seem to be running at all;"
6682 " in this case it's safer to repair by"
6683 " running 'gnt-instance stop' to ensure disk"
6684 " shutdown, and then restarting it")
6686 if runningon_target:
6687 # the migration has actually succeeded, we need to update the config
6688 self.feedback_fn("* instance running on secondary node (%s),"
6689 " updating config" % target_node)
6690 instance.primary_node = target_node
6691 self.cfg.Update(instance, self.feedback_fn)
6692 demoted_node = source_node
6694 self.feedback_fn("* instance confirmed to be running on its"
6695 " primary node (%s)" % source_node)
6696 demoted_node = target_node
6698 if instance.disk_template in constants.DTS_INT_MIRROR:
6699 self._EnsureSecondary(demoted_node)
6701 self._WaitUntilSync()
6702 except errors.OpExecError:
6703 # we ignore here errors, since if the device is standalone, it
6704 # won't be able to sync
6706 self._GoStandalone()
6707 self._GoReconnect(False)
6708 self._WaitUntilSync()
6710 self.feedback_fn("* done")
6712 def _RevertDiskStatus(self):
6713 """Try to revert the disk status after a failed migration.
6716 target_node = self.target_node
6717 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
6721 self._EnsureSecondary(target_node)
6722 self._GoStandalone()
6723 self._GoReconnect(False)
6724 self._WaitUntilSync()
6725 except errors.OpExecError, err:
6726 self.lu.LogWarning("Migration failed and I can't reconnect the drives,"
6727 " please try to recover the instance manually;"
6728 " error '%s'" % str(err))
6730 def _AbortMigration(self):
6731 """Call the hypervisor code to abort a started migration.
6734 instance = self.instance
6735 target_node = self.target_node
6736 migration_info = self.migration_info
6738 abort_result = self.rpc.call_finalize_migration(target_node,
6742 abort_msg = abort_result.fail_msg
6744 logging.error("Aborting migration failed on target node %s: %s",
6745 target_node, abort_msg)
6746 # Don't raise an exception here, as we stil have to try to revert the
6747 # disk status, even if this step failed.
6749 def _ExecMigration(self):
6750 """Migrate an instance.
6752 The migrate is done by:
6753 - change the disks into dual-master mode
6754 - wait until disks are fully synchronized again
6755 - migrate the instance
6756 - change disks on the new secondary node (the old primary) to secondary
6757 - wait until disks are fully synchronized
6758 - change disks into single-master mode
6761 instance = self.instance
6762 target_node = self.target_node
6763 source_node = self.source_node
6765 self.feedback_fn("* checking disk consistency between source and target")
6766 for dev in instance.disks:
6767 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6768 raise errors.OpExecError("Disk %s is degraded or not fully"
6769 " synchronized on target node,"
6770 " aborting migration" % dev.iv_name)
6772 # First get the migration information from the remote node
6773 result = self.rpc.call_migration_info(source_node, instance)
6774 msg = result.fail_msg
6776 log_err = ("Failed fetching source migration information from %s: %s" %
6778 logging.error(log_err)
6779 raise errors.OpExecError(log_err)
6781 self.migration_info = migration_info = result.payload
6783 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
6784 # Then switch the disks to master/master mode
6785 self._EnsureSecondary(target_node)
6786 self._GoStandalone()
6787 self._GoReconnect(True)
6788 self._WaitUntilSync()
6790 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6791 result = self.rpc.call_accept_instance(target_node,
6794 self.nodes_ip[target_node])
6796 msg = result.fail_msg
6798 logging.error("Instance pre-migration failed, trying to revert"
6799 " disk status: %s", msg)
6800 self.feedback_fn("Pre-migration failed, aborting")
6801 self._AbortMigration()
6802 self._RevertDiskStatus()
6803 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6804 (instance.name, msg))
6806 self.feedback_fn("* migrating instance to %s" % target_node)
6807 result = self.rpc.call_instance_migrate(source_node, instance,
6808 self.nodes_ip[target_node],
6810 msg = result.fail_msg
6812 logging.error("Instance migration failed, trying to revert"
6813 " disk status: %s", msg)
6814 self.feedback_fn("Migration failed, aborting")
6815 self._AbortMigration()
6816 self._RevertDiskStatus()
6817 raise errors.OpExecError("Could not migrate instance %s: %s" %
6818 (instance.name, msg))
6820 instance.primary_node = target_node
6821 # distribute new instance config to the other nodes
6822 self.cfg.Update(instance, self.feedback_fn)
6824 result = self.rpc.call_finalize_migration(target_node,
6828 msg = result.fail_msg
6830 logging.error("Instance migration succeeded, but finalization failed:"
6832 raise errors.OpExecError("Could not finalize instance migration: %s" %
6835 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
6836 self._EnsureSecondary(source_node)
6837 self._WaitUntilSync()
6838 self._GoStandalone()
6839 self._GoReconnect(False)
6840 self._WaitUntilSync()
6842 self.feedback_fn("* done")
6844 def _ExecFailover(self):
6845 """Failover an instance.
6847 The failover is done by shutting it down on its present node and
6848 starting it on the secondary.
6851 instance = self.instance
6852 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
6854 source_node = instance.primary_node
6855 target_node = self.target_node
6857 if instance.admin_up:
6858 self.feedback_fn("* checking disk consistency between source and target")
6859 for dev in instance.disks:
6860 # for drbd, these are drbd over lvm
6861 if not _CheckDiskConsistency(self, dev, target_node, False):
6862 if not self.ignore_consistency:
6863 raise errors.OpExecError("Disk %s is degraded on target node,"
6864 " aborting failover" % dev.iv_name)
6866 self.feedback_fn("* not checking disk consistency as instance is not"
6869 self.feedback_fn("* shutting down instance on source node")
6870 logging.info("Shutting down instance %s on node %s",
6871 instance.name, source_node)
6873 result = self.rpc.call_instance_shutdown(source_node, instance,
6874 self.shutdown_timeout)
6875 msg = result.fail_msg
6877 if self.ignore_consistency or primary_node.offline:
6878 self.lu.LogWarning("Could not shutdown instance %s on node %s,"
6879 " proceeding anyway; please make sure node"
6880 " %s is down; error details: %s",
6881 instance.name, source_node, source_node, msg)
6883 raise errors.OpExecError("Could not shutdown instance %s on"
6885 (instance.name, source_node, msg))
6887 self.feedback_fn("* deactivating the instance's disks on source node")
6888 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
6889 raise errors.OpExecError("Can't shut down the instance's disks.")
6891 instance.primary_node = target_node
6892 # distribute new instance config to the other nodes
6893 self.cfg.Update(instance, self.feedback_fn)
6895 # Only start the instance if it's marked as up
6896 if instance.admin_up:
6897 self.feedback_fn("* activating the instance's disks on target node")
6898 logging.info("Starting instance %s on node %s",
6899 instance.name, target_node)
6901 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6902 ignore_secondaries=True)
6904 _ShutdownInstanceDisks(self, instance)
6905 raise errors.OpExecError("Can't activate the instance's disks")
6907 self.feedback_fn("* starting the instance on the target node")
6908 result = self.rpc.call_instance_start(target_node, instance, None, None)
6909 msg = result.fail_msg
6911 _ShutdownInstanceDisks(self, instance)
6912 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6913 (instance.name, target_node, msg))
6915 def Exec(self, feedback_fn):
6916 """Perform the migration.
6919 self.feedback_fn = feedback_fn
6920 self.source_node = self.instance.primary_node
6922 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
6923 if self.instance.disk_template in constants.DTS_INT_MIRROR:
6924 self.target_node = self.instance.secondary_nodes[0]
6925 # Otherwise self.target_node has been populated either
6926 # directly, or through an iallocator.
6928 self.all_nodes = [self.source_node, self.target_node]
6930 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6931 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6935 feedback_fn("Failover instance %s" % self.instance.name)
6936 self._ExecFailover()
6938 feedback_fn("Migrating instance %s" % self.instance.name)
6941 return self._ExecCleanup()
6943 return self._ExecMigration()
6946 def _CreateBlockDev(lu, node, instance, device, force_create,
6948 """Create a tree of block devices on a given node.
6950 If this device type has to be created on secondaries, create it and
6953 If not, just recurse to children keeping the same 'force' value.
6955 @param lu: the lu on whose behalf we execute
6956 @param node: the node on which to create the device
6957 @type instance: L{objects.Instance}
6958 @param instance: the instance which owns the device
6959 @type device: L{objects.Disk}
6960 @param device: the device to create
6961 @type force_create: boolean
6962 @param force_create: whether to force creation of this device; this
6963 will be change to True whenever we find a device which has
6964 CreateOnSecondary() attribute
6965 @param info: the extra 'metadata' we should attach to the device
6966 (this will be represented as a LVM tag)
6967 @type force_open: boolean
6968 @param force_open: this parameter will be passes to the
6969 L{backend.BlockdevCreate} function where it specifies
6970 whether we run on primary or not, and it affects both
6971 the child assembly and the device own Open() execution
6974 if device.CreateOnSecondary():
6978 for child in device.children:
6979 _CreateBlockDev(lu, node, instance, child, force_create,
6982 if not force_create:
6985 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6988 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6989 """Create a single block device on a given node.
6991 This will not recurse over children of the device, so they must be
6994 @param lu: the lu on whose behalf we execute
6995 @param node: the node on which to create the device
6996 @type instance: L{objects.Instance}
6997 @param instance: the instance which owns the device
6998 @type device: L{objects.Disk}
6999 @param device: the device to create
7000 @param info: the extra 'metadata' we should attach to the device
7001 (this will be represented as a LVM tag)
7002 @type force_open: boolean
7003 @param force_open: this parameter will be passes to the
7004 L{backend.BlockdevCreate} function where it specifies
7005 whether we run on primary or not, and it affects both
7006 the child assembly and the device own Open() execution
7009 lu.cfg.SetDiskID(device, node)
7010 result = lu.rpc.call_blockdev_create(node, device, device.size,
7011 instance.name, force_open, info)
7012 result.Raise("Can't create block device %s on"
7013 " node %s for instance %s" % (device, node, instance.name))
7014 if device.physical_id is None:
7015 device.physical_id = result.payload
7018 def _GenerateUniqueNames(lu, exts):
7019 """Generate a suitable LV name.
7021 This will generate a logical volume name for the given instance.
7026 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
7027 results.append("%s%s" % (new_id, val))
7031 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
7032 iv_name, p_minor, s_minor):
7033 """Generate a drbd8 device complete with its children.
7036 assert len(vgnames) == len(names) == 2
7037 port = lu.cfg.AllocatePort()
7038 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
7039 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
7040 logical_id=(vgnames[0], names[0]))
7041 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7042 logical_id=(vgnames[1], names[1]))
7043 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
7044 logical_id=(primary, secondary, port,
7047 children=[dev_data, dev_meta],
7052 def _GenerateDiskTemplate(lu, template_name,
7053 instance_name, primary_node,
7054 secondary_nodes, disk_info,
7055 file_storage_dir, file_driver,
7056 base_index, feedback_fn):
7057 """Generate the entire disk layout for a given template type.
7060 #TODO: compute space requirements
7062 vgname = lu.cfg.GetVGName()
7063 disk_count = len(disk_info)
7065 if template_name == constants.DT_DISKLESS:
7067 elif template_name == constants.DT_PLAIN:
7068 if len(secondary_nodes) != 0:
7069 raise errors.ProgrammerError("Wrong template configuration")
7071 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7072 for i in range(disk_count)])
7073 for idx, disk in enumerate(disk_info):
7074 disk_index = idx + base_index
7075 vg = disk.get(constants.IDISK_VG, vgname)
7076 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
7077 disk_dev = objects.Disk(dev_type=constants.LD_LV,
7078 size=disk[constants.IDISK_SIZE],
7079 logical_id=(vg, names[idx]),
7080 iv_name="disk/%d" % disk_index,
7081 mode=disk[constants.IDISK_MODE])
7082 disks.append(disk_dev)
7083 elif template_name == constants.DT_DRBD8:
7084 if len(secondary_nodes) != 1:
7085 raise errors.ProgrammerError("Wrong template configuration")
7086 remote_node = secondary_nodes[0]
7087 minors = lu.cfg.AllocateDRBDMinor(
7088 [primary_node, remote_node] * len(disk_info), instance_name)
7091 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
7092 for i in range(disk_count)]):
7093 names.append(lv_prefix + "_data")
7094 names.append(lv_prefix + "_meta")
7095 for idx, disk in enumerate(disk_info):
7096 disk_index = idx + base_index
7097 data_vg = disk.get(constants.IDISK_VG, vgname)
7098 meta_vg = disk.get(constants.IDISK_METAVG, data_vg)
7099 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
7100 disk[constants.IDISK_SIZE],
7102 names[idx * 2:idx * 2 + 2],
7103 "disk/%d" % disk_index,
7104 minors[idx * 2], minors[idx * 2 + 1])
7105 disk_dev.mode = disk[constants.IDISK_MODE]
7106 disks.append(disk_dev)
7107 elif template_name == constants.DT_FILE:
7108 if len(secondary_nodes) != 0:
7109 raise errors.ProgrammerError("Wrong template configuration")
7111 opcodes.RequireFileStorage()
7113 for idx, disk in enumerate(disk_info):
7114 disk_index = idx + base_index
7115 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7116 size=disk[constants.IDISK_SIZE],
7117 iv_name="disk/%d" % disk_index,
7118 logical_id=(file_driver,
7119 "%s/disk%d" % (file_storage_dir,
7121 mode=disk[constants.IDISK_MODE])
7122 disks.append(disk_dev)
7123 elif template_name == constants.DT_SHARED_FILE:
7124 if len(secondary_nodes) != 0:
7125 raise errors.ProgrammerError("Wrong template configuration")
7127 opcodes.RequireSharedFileStorage()
7129 for idx, disk in enumerate(disk_info):
7130 disk_index = idx + base_index
7131 disk_dev = objects.Disk(dev_type=constants.LD_FILE,
7132 size=disk[constants.IDISK_SIZE],
7133 iv_name="disk/%d" % disk_index,
7134 logical_id=(file_driver,
7135 "%s/disk%d" % (file_storage_dir,
7137 mode=disk[constants.IDISK_MODE])
7138 disks.append(disk_dev)
7139 elif template_name == constants.DT_BLOCK:
7140 if len(secondary_nodes) != 0:
7141 raise errors.ProgrammerError("Wrong template configuration")
7143 for idx, disk in enumerate(disk_info):
7144 disk_index = idx + base_index
7145 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV,
7146 size=disk[constants.IDISK_SIZE],
7147 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL,
7148 disk[constants.IDISK_ADOPT]),
7149 iv_name="disk/%d" % disk_index,
7150 mode=disk[constants.IDISK_MODE])
7151 disks.append(disk_dev)
7154 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
7158 def _GetInstanceInfoText(instance):
7159 """Compute that text that should be added to the disk's metadata.
7162 return "originstname+%s" % instance.name
7165 def _CalcEta(time_taken, written, total_size):
7166 """Calculates the ETA based on size written and total size.
7168 @param time_taken: The time taken so far
7169 @param written: amount written so far
7170 @param total_size: The total size of data to be written
7171 @return: The remaining time in seconds
7174 avg_time = time_taken / float(written)
7175 return (total_size - written) * avg_time
7178 def _WipeDisks(lu, instance):
7179 """Wipes instance disks.
7181 @type lu: L{LogicalUnit}
7182 @param lu: the logical unit on whose behalf we execute
7183 @type instance: L{objects.Instance}
7184 @param instance: the instance whose disks we should create
7185 @return: the success of the wipe
7188 node = instance.primary_node
7190 for device in instance.disks:
7191 lu.cfg.SetDiskID(device, node)
7193 logging.info("Pause sync of instance %s disks", instance.name)
7194 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
7196 for idx, success in enumerate(result.payload):
7198 logging.warn("pause-sync of instance %s for disks %d failed",
7202 for idx, device in enumerate(instance.disks):
7203 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
7204 # MAX_WIPE_CHUNK at max
7205 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
7206 constants.MIN_WIPE_CHUNK_PERCENT)
7207 # we _must_ make this an int, otherwise rounding errors will
7209 wipe_chunk_size = int(wipe_chunk_size)
7211 lu.LogInfo("* Wiping disk %d", idx)
7212 logging.info("Wiping disk %d for instance %s, node %s using"
7213 " chunk size %s", idx, instance.name, node, wipe_chunk_size)
7218 start_time = time.time()
7220 while offset < size:
7221 wipe_size = min(wipe_chunk_size, size - offset)
7222 logging.debug("Wiping disk %d, offset %s, chunk %s",
7223 idx, offset, wipe_size)
7224 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
7225 result.Raise("Could not wipe disk %d at offset %d for size %d" %
7226 (idx, offset, wipe_size))
7229 if now - last_output >= 60:
7230 eta = _CalcEta(now - start_time, offset, size)
7231 lu.LogInfo(" - done: %.1f%% ETA: %s" %
7232 (offset / float(size) * 100, utils.FormatSeconds(eta)))
7235 logging.info("Resume sync of instance %s disks", instance.name)
7237 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
7239 for idx, success in enumerate(result.payload):
7241 lu.LogWarning("Resume sync of disk %d failed, please have a"
7242 " look at the status and troubleshoot the issue", idx)
7243 logging.warn("resume-sync of instance %s for disks %d failed",
7247 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
7248 """Create all disks for an instance.
7250 This abstracts away some work from AddInstance.
7252 @type lu: L{LogicalUnit}
7253 @param lu: the logical unit on whose behalf we execute
7254 @type instance: L{objects.Instance}
7255 @param instance: the instance whose disks we should create
7257 @param to_skip: list of indices to skip
7258 @type target_node: string
7259 @param target_node: if passed, overrides the target node for creation
7261 @return: the success of the creation
7264 info = _GetInstanceInfoText(instance)
7265 if target_node is None:
7266 pnode = instance.primary_node
7267 all_nodes = instance.all_nodes
7272 if instance.disk_template in (constants.DT_FILE, constants.DT_SHARED_FILE):
7273 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7274 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
7276 result.Raise("Failed to create directory '%s' on"
7277 " node %s" % (file_storage_dir, pnode))
7279 # Note: this needs to be kept in sync with adding of disks in
7280 # LUInstanceSetParams
7281 for idx, device in enumerate(instance.disks):
7282 if to_skip and idx in to_skip:
7284 logging.info("Creating volume %s for instance %s",
7285 device.iv_name, instance.name)
7287 for node in all_nodes:
7288 f_create = node == pnode
7289 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
7292 def _RemoveDisks(lu, instance, target_node=None):
7293 """Remove all disks for an instance.
7295 This abstracts away some work from `AddInstance()` and
7296 `RemoveInstance()`. Note that in case some of the devices couldn't
7297 be removed, the removal will continue with the other ones (compare
7298 with `_CreateDisks()`).
7300 @type lu: L{LogicalUnit}
7301 @param lu: the logical unit on whose behalf we execute
7302 @type instance: L{objects.Instance}
7303 @param instance: the instance whose disks we should remove
7304 @type target_node: string
7305 @param target_node: used to override the node on which to remove the disks
7307 @return: the success of the removal
7310 logging.info("Removing block devices for instance %s", instance.name)
7313 for device in instance.disks:
7315 edata = [(target_node, device)]
7317 edata = device.ComputeNodeTree(instance.primary_node)
7318 for node, disk in edata:
7319 lu.cfg.SetDiskID(disk, node)
7320 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
7322 lu.LogWarning("Could not remove block device %s on node %s,"
7323 " continuing anyway: %s", device.iv_name, node, msg)
7326 if instance.disk_template == constants.DT_FILE:
7327 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
7331 tgt = instance.primary_node
7332 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
7334 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
7335 file_storage_dir, instance.primary_node, result.fail_msg)
7341 def _ComputeDiskSizePerVG(disk_template, disks):
7342 """Compute disk size requirements in the volume group
7345 def _compute(disks, payload):
7346 """Universal algorithm.
7351 vgs[disk[constants.IDISK_VG]] = \
7352 vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
7356 # Required free disk space as a function of disk and swap space
7358 constants.DT_DISKLESS: {},
7359 constants.DT_PLAIN: _compute(disks, 0),
7360 # 128 MB are added for drbd metadata for each disk
7361 constants.DT_DRBD8: _compute(disks, 128),
7362 constants.DT_FILE: {},
7363 constants.DT_SHARED_FILE: {},
7366 if disk_template not in req_size_dict:
7367 raise errors.ProgrammerError("Disk template '%s' size requirement"
7368 " is unknown" % disk_template)
7370 return req_size_dict[disk_template]
7373 def _ComputeDiskSize(disk_template, disks):
7374 """Compute disk size requirements in the volume group
7377 # Required free disk space as a function of disk and swap space
7379 constants.DT_DISKLESS: None,
7380 constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
7381 # 128 MB are added for drbd metadata for each disk
7382 constants.DT_DRBD8: sum(d[constants.IDISK_SIZE] + 128 for d in disks),
7383 constants.DT_FILE: None,
7384 constants.DT_SHARED_FILE: 0,
7385 constants.DT_BLOCK: 0,
7388 if disk_template not in req_size_dict:
7389 raise errors.ProgrammerError("Disk template '%s' size requirement"
7390 " is unknown" % disk_template)
7392 return req_size_dict[disk_template]
7395 def _FilterVmNodes(lu, nodenames):
7396 """Filters out non-vm_capable nodes from a list.
7398 @type lu: L{LogicalUnit}
7399 @param lu: the logical unit for which we check
7400 @type nodenames: list
7401 @param nodenames: the list of nodes on which we should check
7403 @return: the list of vm-capable nodes
7406 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
7407 return [name for name in nodenames if name not in vm_nodes]
7410 def _CheckHVParams(lu, nodenames, hvname, hvparams):
7411 """Hypervisor parameter validation.
7413 This function abstract the hypervisor parameter validation to be
7414 used in both instance create and instance modify.
7416 @type lu: L{LogicalUnit}
7417 @param lu: the logical unit for which we check
7418 @type nodenames: list
7419 @param nodenames: the list of nodes on which we should check
7420 @type hvname: string
7421 @param hvname: the name of the hypervisor we should use
7422 @type hvparams: dict
7423 @param hvparams: the parameters which we need to check
7424 @raise errors.OpPrereqError: if the parameters are not valid
7427 nodenames = _FilterVmNodes(lu, nodenames)
7428 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
7431 for node in nodenames:
7435 info.Raise("Hypervisor parameter validation failed on node %s" % node)
7438 def _CheckOSParams(lu, required, nodenames, osname, osparams):
7439 """OS parameters validation.
7441 @type lu: L{LogicalUnit}
7442 @param lu: the logical unit for which we check
7443 @type required: boolean
7444 @param required: whether the validation should fail if the OS is not
7446 @type nodenames: list
7447 @param nodenames: the list of nodes on which we should check
7448 @type osname: string
7449 @param osname: the name of the hypervisor we should use
7450 @type osparams: dict
7451 @param osparams: the parameters which we need to check
7452 @raise errors.OpPrereqError: if the parameters are not valid
7455 nodenames = _FilterVmNodes(lu, nodenames)
7456 result = lu.rpc.call_os_validate(required, nodenames, osname,
7457 [constants.OS_VALIDATE_PARAMETERS],
7459 for node, nres in result.items():
7460 # we don't check for offline cases since this should be run only
7461 # against the master node and/or an instance's nodes
7462 nres.Raise("OS Parameters validation failed on node %s" % node)
7463 if not nres.payload:
7464 lu.LogInfo("OS %s not found on node %s, validation skipped",
7468 class LUInstanceCreate(LogicalUnit):
7469 """Create an instance.
7472 HPATH = "instance-add"
7473 HTYPE = constants.HTYPE_INSTANCE
7476 def CheckArguments(self):
7480 # do not require name_check to ease forward/backward compatibility
7482 if self.op.no_install and self.op.start:
7483 self.LogInfo("No-installation mode selected, disabling startup")
7484 self.op.start = False
7485 # validate/normalize the instance name
7486 self.op.instance_name = \
7487 netutils.Hostname.GetNormalizedName(self.op.instance_name)
7489 if self.op.ip_check and not self.op.name_check:
7490 # TODO: make the ip check more flexible and not depend on the name check
7491 raise errors.OpPrereqError("Cannot do IP address check without a name"
7492 " check", errors.ECODE_INVAL)
7494 # check nics' parameter names
7495 for nic in self.op.nics:
7496 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
7498 # check disks. parameter names and consistent adopt/no-adopt strategy
7499 has_adopt = has_no_adopt = False
7500 for disk in self.op.disks:
7501 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
7502 if constants.IDISK_ADOPT in disk:
7506 if has_adopt and has_no_adopt:
7507 raise errors.OpPrereqError("Either all disks are adopted or none is",
7510 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
7511 raise errors.OpPrereqError("Disk adoption is not supported for the"
7512 " '%s' disk template" %
7513 self.op.disk_template,
7515 if self.op.iallocator is not None:
7516 raise errors.OpPrereqError("Disk adoption not allowed with an"
7517 " iallocator script", errors.ECODE_INVAL)
7518 if self.op.mode == constants.INSTANCE_IMPORT:
7519 raise errors.OpPrereqError("Disk adoption not allowed for"
7520 " instance import", errors.ECODE_INVAL)
7522 if self.op.disk_template in constants.DTS_MUST_ADOPT:
7523 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
7524 " but no 'adopt' parameter given" %
7525 self.op.disk_template,
7528 self.adopt_disks = has_adopt
7530 # instance name verification
7531 if self.op.name_check:
7532 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
7533 self.op.instance_name = self.hostname1.name
7534 # used in CheckPrereq for ip ping check
7535 self.check_ip = self.hostname1.ip
7537 self.check_ip = None
7539 # file storage checks
7540 if (self.op.file_driver and
7541 not self.op.file_driver in constants.FILE_DRIVER):
7542 raise errors.OpPrereqError("Invalid file driver name '%s'" %
7543 self.op.file_driver, errors.ECODE_INVAL)
7545 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
7546 raise errors.OpPrereqError("File storage directory path not absolute",
7549 ### Node/iallocator related checks
7550 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
7552 if self.op.pnode is not None:
7553 if self.op.disk_template in constants.DTS_INT_MIRROR:
7554 if self.op.snode is None:
7555 raise errors.OpPrereqError("The networked disk templates need"
7556 " a mirror node", errors.ECODE_INVAL)
7558 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
7560 self.op.snode = None
7562 self._cds = _GetClusterDomainSecret()
7564 if self.op.mode == constants.INSTANCE_IMPORT:
7565 # On import force_variant must be True, because if we forced it at
7566 # initial install, our only chance when importing it back is that it
7568 self.op.force_variant = True
7570 if self.op.no_install:
7571 self.LogInfo("No-installation mode has no effect during import")
7573 elif self.op.mode == constants.INSTANCE_CREATE:
7574 if self.op.os_type is None:
7575 raise errors.OpPrereqError("No guest OS specified",
7577 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
7578 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
7579 " installation" % self.op.os_type,
7581 if self.op.disk_template is None:
7582 raise errors.OpPrereqError("No disk template specified",
7585 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7586 # Check handshake to ensure both clusters have the same domain secret
7587 src_handshake = self.op.source_handshake
7588 if not src_handshake:
7589 raise errors.OpPrereqError("Missing source handshake",
7592 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
7595 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
7598 # Load and check source CA
7599 self.source_x509_ca_pem = self.op.source_x509_ca
7600 if not self.source_x509_ca_pem:
7601 raise errors.OpPrereqError("Missing source X509 CA",
7605 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
7607 except OpenSSL.crypto.Error, err:
7608 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
7609 (err, ), errors.ECODE_INVAL)
7611 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
7612 if errcode is not None:
7613 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
7616 self.source_x509_ca = cert
7618 src_instance_name = self.op.source_instance_name
7619 if not src_instance_name:
7620 raise errors.OpPrereqError("Missing source instance name",
7623 self.source_instance_name = \
7624 netutils.GetHostname(name=src_instance_name).name
7627 raise errors.OpPrereqError("Invalid instance creation mode %r" %
7628 self.op.mode, errors.ECODE_INVAL)
7630 def ExpandNames(self):
7631 """ExpandNames for CreateInstance.
7633 Figure out the right locks for instance creation.
7636 self.needed_locks = {}
7638 instance_name = self.op.instance_name
7639 # this is just a preventive check, but someone might still add this
7640 # instance in the meantime, and creation will fail at lock-add time
7641 if instance_name in self.cfg.GetInstanceList():
7642 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7643 instance_name, errors.ECODE_EXISTS)
7645 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
7647 if self.op.iallocator:
7648 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7650 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
7651 nodelist = [self.op.pnode]
7652 if self.op.snode is not None:
7653 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
7654 nodelist.append(self.op.snode)
7655 self.needed_locks[locking.LEVEL_NODE] = nodelist
7657 # in case of import lock the source node too
7658 if self.op.mode == constants.INSTANCE_IMPORT:
7659 src_node = self.op.src_node
7660 src_path = self.op.src_path
7662 if src_path is None:
7663 self.op.src_path = src_path = self.op.instance_name
7665 if src_node is None:
7666 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7667 self.op.src_node = None
7668 if os.path.isabs(src_path):
7669 raise errors.OpPrereqError("Importing an instance from an absolute"
7670 " path requires a source node option",
7673 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
7674 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
7675 self.needed_locks[locking.LEVEL_NODE].append(src_node)
7676 if not os.path.isabs(src_path):
7677 self.op.src_path = src_path = \
7678 utils.PathJoin(constants.EXPORT_DIR, src_path)
7680 def _RunAllocator(self):
7681 """Run the allocator based on input opcode.
7684 nics = [n.ToDict() for n in self.nics]
7685 ial = IAllocator(self.cfg, self.rpc,
7686 mode=constants.IALLOCATOR_MODE_ALLOC,
7687 name=self.op.instance_name,
7688 disk_template=self.op.disk_template,
7691 vcpus=self.be_full[constants.BE_VCPUS],
7692 mem_size=self.be_full[constants.BE_MEMORY],
7695 hypervisor=self.op.hypervisor,
7698 ial.Run(self.op.iallocator)
7701 raise errors.OpPrereqError("Can't compute nodes using"
7702 " iallocator '%s': %s" %
7703 (self.op.iallocator, ial.info),
7705 if len(ial.result) != ial.required_nodes:
7706 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7707 " of nodes (%s), required %s" %
7708 (self.op.iallocator, len(ial.result),
7709 ial.required_nodes), errors.ECODE_FAULT)
7710 self.op.pnode = ial.result[0]
7711 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7712 self.op.instance_name, self.op.iallocator,
7713 utils.CommaJoin(ial.result))
7714 if ial.required_nodes == 2:
7715 self.op.snode = ial.result[1]
7717 def BuildHooksEnv(self):
7720 This runs on master, primary and secondary nodes of the instance.
7724 "ADD_MODE": self.op.mode,
7726 if self.op.mode == constants.INSTANCE_IMPORT:
7727 env["SRC_NODE"] = self.op.src_node
7728 env["SRC_PATH"] = self.op.src_path
7729 env["SRC_IMAGES"] = self.src_images
7731 env.update(_BuildInstanceHookEnv(
7732 name=self.op.instance_name,
7733 primary_node=self.op.pnode,
7734 secondary_nodes=self.secondaries,
7735 status=self.op.start,
7736 os_type=self.op.os_type,
7737 memory=self.be_full[constants.BE_MEMORY],
7738 vcpus=self.be_full[constants.BE_VCPUS],
7739 nics=_NICListToTuple(self, self.nics),
7740 disk_template=self.op.disk_template,
7741 disks=[(d[constants.IDISK_SIZE], d[constants.IDISK_MODE])
7742 for d in self.disks],
7745 hypervisor_name=self.op.hypervisor,
7750 def BuildHooksNodes(self):
7751 """Build hooks nodes.
7754 nl = [self.cfg.GetMasterNode(), self.op.pnode] + self.secondaries
7757 def _ReadExportInfo(self):
7758 """Reads the export information from disk.
7760 It will override the opcode source node and path with the actual
7761 information, if these two were not specified before.
7763 @return: the export information
7766 assert self.op.mode == constants.INSTANCE_IMPORT
7768 src_node = self.op.src_node
7769 src_path = self.op.src_path
7771 if src_node is None:
7772 locked_nodes = self.glm.list_owned(locking.LEVEL_NODE)
7773 exp_list = self.rpc.call_export_list(locked_nodes)
7775 for node in exp_list:
7776 if exp_list[node].fail_msg:
7778 if src_path in exp_list[node].payload:
7780 self.op.src_node = src_node = node
7781 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
7785 raise errors.OpPrereqError("No export found for relative path %s" %
7786 src_path, errors.ECODE_INVAL)
7788 _CheckNodeOnline(self, src_node)
7789 result = self.rpc.call_export_info(src_node, src_path)
7790 result.Raise("No export or invalid export found in dir %s" % src_path)
7792 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
7793 if not export_info.has_section(constants.INISECT_EXP):
7794 raise errors.ProgrammerError("Corrupted export config",
7795 errors.ECODE_ENVIRON)
7797 ei_version = export_info.get(constants.INISECT_EXP, "version")
7798 if (int(ei_version) != constants.EXPORT_VERSION):
7799 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
7800 (ei_version, constants.EXPORT_VERSION),
7801 errors.ECODE_ENVIRON)
7804 def _ReadExportParams(self, einfo):
7805 """Use export parameters as defaults.
7807 In case the opcode doesn't specify (as in override) some instance
7808 parameters, then try to use them from the export information, if
7812 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
7814 if self.op.disk_template is None:
7815 if einfo.has_option(constants.INISECT_INS, "disk_template"):
7816 self.op.disk_template = einfo.get(constants.INISECT_INS,
7819 raise errors.OpPrereqError("No disk template specified and the export"
7820 " is missing the disk_template information",
7823 if not self.op.disks:
7824 if einfo.has_option(constants.INISECT_INS, "disk_count"):
7826 # TODO: import the disk iv_name too
7827 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
7828 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
7829 disks.append({constants.IDISK_SIZE: disk_sz})
7830 self.op.disks = disks
7832 raise errors.OpPrereqError("No disk info specified and the export"
7833 " is missing the disk information",
7836 if (not self.op.nics and
7837 einfo.has_option(constants.INISECT_INS, "nic_count")):
7839 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
7841 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
7842 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
7847 if (self.op.hypervisor is None and
7848 einfo.has_option(constants.INISECT_INS, "hypervisor")):
7849 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
7850 if einfo.has_section(constants.INISECT_HYP):
7851 # use the export parameters but do not override the ones
7852 # specified by the user
7853 for name, value in einfo.items(constants.INISECT_HYP):
7854 if name not in self.op.hvparams:
7855 self.op.hvparams[name] = value
7857 if einfo.has_section(constants.INISECT_BEP):
7858 # use the parameters, without overriding
7859 for name, value in einfo.items(constants.INISECT_BEP):
7860 if name not in self.op.beparams:
7861 self.op.beparams[name] = value
7863 # try to read the parameters old style, from the main section
7864 for name in constants.BES_PARAMETERS:
7865 if (name not in self.op.beparams and
7866 einfo.has_option(constants.INISECT_INS, name)):
7867 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
7869 if einfo.has_section(constants.INISECT_OSP):
7870 # use the parameters, without overriding
7871 for name, value in einfo.items(constants.INISECT_OSP):
7872 if name not in self.op.osparams:
7873 self.op.osparams[name] = value
7875 def _RevertToDefaults(self, cluster):
7876 """Revert the instance parameters to the default values.
7880 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7881 for name in self.op.hvparams.keys():
7882 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7883 del self.op.hvparams[name]
7885 be_defs = cluster.SimpleFillBE({})
7886 for name in self.op.beparams.keys():
7887 if name in be_defs and be_defs[name] == self.op.beparams[name]:
7888 del self.op.beparams[name]
7890 nic_defs = cluster.SimpleFillNIC({})
7891 for nic in self.op.nics:
7892 for name in constants.NICS_PARAMETERS:
7893 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7896 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7897 for name in self.op.osparams.keys():
7898 if name in os_defs and os_defs[name] == self.op.osparams[name]:
7899 del self.op.osparams[name]
7901 def CheckPrereq(self):
7902 """Check prerequisites.
7905 if self.op.mode == constants.INSTANCE_IMPORT:
7906 export_info = self._ReadExportInfo()
7907 self._ReadExportParams(export_info)
7909 if (not self.cfg.GetVGName() and
7910 self.op.disk_template not in constants.DTS_NOT_LVM):
7911 raise errors.OpPrereqError("Cluster does not support lvm-based"
7912 " instances", errors.ECODE_STATE)
7914 if self.op.hypervisor is None:
7915 self.op.hypervisor = self.cfg.GetHypervisorType()
7917 cluster = self.cfg.GetClusterInfo()
7918 enabled_hvs = cluster.enabled_hypervisors
7919 if self.op.hypervisor not in enabled_hvs:
7920 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7921 " cluster (%s)" % (self.op.hypervisor,
7922 ",".join(enabled_hvs)),
7925 # check hypervisor parameter syntax (locally)
7926 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7927 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7929 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7930 hv_type.CheckParameterSyntax(filled_hvp)
7931 self.hv_full = filled_hvp
7932 # check that we don't specify global parameters on an instance
7933 _CheckGlobalHvParams(self.op.hvparams)
7935 # fill and remember the beparams dict
7936 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7937 self.be_full = cluster.SimpleFillBE(self.op.beparams)
7939 # build os parameters
7940 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7942 # now that hvp/bep are in final format, let's reset to defaults,
7944 if self.op.identify_defaults:
7945 self._RevertToDefaults(cluster)
7949 for idx, nic in enumerate(self.op.nics):
7950 nic_mode_req = nic.get(constants.INIC_MODE, None)
7951 nic_mode = nic_mode_req
7952 if nic_mode is None:
7953 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7955 # in routed mode, for the first nic, the default ip is 'auto'
7956 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7957 default_ip_mode = constants.VALUE_AUTO
7959 default_ip_mode = constants.VALUE_NONE
7961 # ip validity checks
7962 ip = nic.get(constants.INIC_IP, default_ip_mode)
7963 if ip is None or ip.lower() == constants.VALUE_NONE:
7965 elif ip.lower() == constants.VALUE_AUTO:
7966 if not self.op.name_check:
7967 raise errors.OpPrereqError("IP address set to auto but name checks"
7968 " have been skipped",
7970 nic_ip = self.hostname1.ip
7972 if not netutils.IPAddress.IsValid(ip):
7973 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7977 # TODO: check the ip address for uniqueness
7978 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7979 raise errors.OpPrereqError("Routed nic mode requires an ip address",
7982 # MAC address verification
7983 mac = nic.get(constants.INIC_MAC, constants.VALUE_AUTO)
7984 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7985 mac = utils.NormalizeAndValidateMac(mac)
7988 self.cfg.ReserveMAC(mac, self.proc.GetECId())
7989 except errors.ReservationError:
7990 raise errors.OpPrereqError("MAC address %s already in use"
7991 " in cluster" % mac,
7992 errors.ECODE_NOTUNIQUE)
7994 # Build nic parameters
7995 link = nic.get(constants.INIC_LINK, None)
7998 nicparams[constants.NIC_MODE] = nic_mode_req
8000 nicparams[constants.NIC_LINK] = link
8002 check_params = cluster.SimpleFillNIC(nicparams)
8003 objects.NIC.CheckParameterSyntax(check_params)
8004 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
8006 # disk checks/pre-build
8007 default_vg = self.cfg.GetVGName()
8009 for disk in self.op.disks:
8010 mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
8011 if mode not in constants.DISK_ACCESS_SET:
8012 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
8013 mode, errors.ECODE_INVAL)
8014 size = disk.get(constants.IDISK_SIZE, None)
8016 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
8019 except (TypeError, ValueError):
8020 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
8023 data_vg = disk.get(constants.IDISK_VG, default_vg)
8025 constants.IDISK_SIZE: size,
8026 constants.IDISK_MODE: mode,
8027 constants.IDISK_VG: data_vg,
8028 constants.IDISK_METAVG: disk.get(constants.IDISK_METAVG, data_vg),
8030 if constants.IDISK_ADOPT in disk:
8031 new_disk[constants.IDISK_ADOPT] = disk[constants.IDISK_ADOPT]
8032 self.disks.append(new_disk)
8034 if self.op.mode == constants.INSTANCE_IMPORT:
8036 # Check that the new instance doesn't have less disks than the export
8037 instance_disks = len(self.disks)
8038 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
8039 if instance_disks < export_disks:
8040 raise errors.OpPrereqError("Not enough disks to import."
8041 " (instance: %d, export: %d)" %
8042 (instance_disks, export_disks),
8046 for idx in range(export_disks):
8047 option = 'disk%d_dump' % idx
8048 if export_info.has_option(constants.INISECT_INS, option):
8049 # FIXME: are the old os-es, disk sizes, etc. useful?
8050 export_name = export_info.get(constants.INISECT_INS, option)
8051 image = utils.PathJoin(self.op.src_path, export_name)
8052 disk_images.append(image)
8054 disk_images.append(False)
8056 self.src_images = disk_images
8058 old_name = export_info.get(constants.INISECT_INS, 'name')
8060 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
8061 except (TypeError, ValueError), err:
8062 raise errors.OpPrereqError("Invalid export file, nic_count is not"
8063 " an integer: %s" % str(err),
8065 if self.op.instance_name == old_name:
8066 for idx, nic in enumerate(self.nics):
8067 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
8068 nic_mac_ini = 'nic%d_mac' % idx
8069 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
8071 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
8073 # ip ping checks (we use the same ip that was resolved in ExpandNames)
8074 if self.op.ip_check:
8075 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
8076 raise errors.OpPrereqError("IP %s of instance %s already in use" %
8077 (self.check_ip, self.op.instance_name),
8078 errors.ECODE_NOTUNIQUE)
8080 #### mac address generation
8081 # By generating here the mac address both the allocator and the hooks get
8082 # the real final mac address rather than the 'auto' or 'generate' value.
8083 # There is a race condition between the generation and the instance object
8084 # creation, which means that we know the mac is valid now, but we're not
8085 # sure it will be when we actually add the instance. If things go bad
8086 # adding the instance will abort because of a duplicate mac, and the
8087 # creation job will fail.
8088 for nic in self.nics:
8089 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8090 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
8094 if self.op.iallocator is not None:
8095 self._RunAllocator()
8097 #### node related checks
8099 # check primary node
8100 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
8101 assert self.pnode is not None, \
8102 "Cannot retrieve locked node %s" % self.op.pnode
8104 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
8105 pnode.name, errors.ECODE_STATE)
8107 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
8108 pnode.name, errors.ECODE_STATE)
8109 if not pnode.vm_capable:
8110 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
8111 " '%s'" % pnode.name, errors.ECODE_STATE)
8113 self.secondaries = []
8115 # mirror node verification
8116 if self.op.disk_template in constants.DTS_INT_MIRROR:
8117 if self.op.snode == pnode.name:
8118 raise errors.OpPrereqError("The secondary node cannot be the"
8119 " primary node", errors.ECODE_INVAL)
8120 _CheckNodeOnline(self, self.op.snode)
8121 _CheckNodeNotDrained(self, self.op.snode)
8122 _CheckNodeVmCapable(self, self.op.snode)
8123 self.secondaries.append(self.op.snode)
8125 nodenames = [pnode.name] + self.secondaries
8127 if not self.adopt_disks:
8128 # Check lv size requirements, if not adopting
8129 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
8130 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
8132 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
8133 all_lvs = set(["%s/%s" % (disk[constants.IDISK_VG],
8134 disk[constants.IDISK_ADOPT])
8135 for disk in self.disks])
8136 if len(all_lvs) != len(self.disks):
8137 raise errors.OpPrereqError("Duplicate volume names given for adoption",
8139 for lv_name in all_lvs:
8141 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
8142 # to ReserveLV uses the same syntax
8143 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
8144 except errors.ReservationError:
8145 raise errors.OpPrereqError("LV named %s used by another instance" %
8146 lv_name, errors.ECODE_NOTUNIQUE)
8148 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
8149 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
8151 node_lvs = self.rpc.call_lv_list([pnode.name],
8152 vg_names.payload.keys())[pnode.name]
8153 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
8154 node_lvs = node_lvs.payload
8156 delta = all_lvs.difference(node_lvs.keys())
8158 raise errors.OpPrereqError("Missing logical volume(s): %s" %
8159 utils.CommaJoin(delta),
8161 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
8163 raise errors.OpPrereqError("Online logical volumes found, cannot"
8164 " adopt: %s" % utils.CommaJoin(online_lvs),
8166 # update the size of disk based on what is found
8167 for dsk in self.disks:
8168 dsk[constants.IDISK_SIZE] = \
8169 int(float(node_lvs["%s/%s" % (dsk[constants.IDISK_VG],
8170 dsk[constants.IDISK_ADOPT])][0]))
8172 elif self.op.disk_template == constants.DT_BLOCK:
8173 # Normalize and de-duplicate device paths
8174 all_disks = set([os.path.abspath(disk[constants.IDISK_ADOPT])
8175 for disk in self.disks])
8176 if len(all_disks) != len(self.disks):
8177 raise errors.OpPrereqError("Duplicate disk names given for adoption",
8179 baddisks = [d for d in all_disks
8180 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
8182 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
8183 " cannot be adopted" %
8184 (", ".join(baddisks),
8185 constants.ADOPTABLE_BLOCKDEV_ROOT),
8188 node_disks = self.rpc.call_bdev_sizes([pnode.name],
8189 list(all_disks))[pnode.name]
8190 node_disks.Raise("Cannot get block device information from node %s" %
8192 node_disks = node_disks.payload
8193 delta = all_disks.difference(node_disks.keys())
8195 raise errors.OpPrereqError("Missing block device(s): %s" %
8196 utils.CommaJoin(delta),
8198 for dsk in self.disks:
8199 dsk[constants.IDISK_SIZE] = \
8200 int(float(node_disks[dsk[constants.IDISK_ADOPT]]))
8202 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
8204 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
8205 # check OS parameters (remotely)
8206 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
8208 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
8210 # memory check on primary node
8212 _CheckNodeFreeMemory(self, self.pnode.name,
8213 "creating instance %s" % self.op.instance_name,
8214 self.be_full[constants.BE_MEMORY],
8217 self.dry_run_result = list(nodenames)
8219 def Exec(self, feedback_fn):
8220 """Create and add the instance to the cluster.
8223 instance = self.op.instance_name
8224 pnode_name = self.pnode.name
8226 ht_kind = self.op.hypervisor
8227 if ht_kind in constants.HTS_REQ_PORT:
8228 network_port = self.cfg.AllocatePort()
8232 if constants.ENABLE_FILE_STORAGE or constants.ENABLE_SHARED_FILE_STORAGE:
8233 # this is needed because os.path.join does not accept None arguments
8234 if self.op.file_storage_dir is None:
8235 string_file_storage_dir = ""
8237 string_file_storage_dir = self.op.file_storage_dir
8239 # build the full file storage dir path
8240 if self.op.disk_template == constants.DT_SHARED_FILE:
8241 get_fsd_fn = self.cfg.GetSharedFileStorageDir
8243 get_fsd_fn = self.cfg.GetFileStorageDir
8245 file_storage_dir = utils.PathJoin(get_fsd_fn(),
8246 string_file_storage_dir, instance)
8248 file_storage_dir = ""
8250 disks = _GenerateDiskTemplate(self,
8251 self.op.disk_template,
8252 instance, pnode_name,
8256 self.op.file_driver,
8260 iobj = objects.Instance(name=instance, os=self.op.os_type,
8261 primary_node=pnode_name,
8262 nics=self.nics, disks=disks,
8263 disk_template=self.op.disk_template,
8265 network_port=network_port,
8266 beparams=self.op.beparams,
8267 hvparams=self.op.hvparams,
8268 hypervisor=self.op.hypervisor,
8269 osparams=self.op.osparams,
8272 if self.adopt_disks:
8273 if self.op.disk_template == constants.DT_PLAIN:
8274 # rename LVs to the newly-generated names; we need to construct
8275 # 'fake' LV disks with the old data, plus the new unique_id
8276 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
8278 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
8279 rename_to.append(t_dsk.logical_id)
8280 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
8281 self.cfg.SetDiskID(t_dsk, pnode_name)
8282 result = self.rpc.call_blockdev_rename(pnode_name,
8283 zip(tmp_disks, rename_to))
8284 result.Raise("Failed to rename adoped LVs")
8286 feedback_fn("* creating instance disks...")
8288 _CreateDisks(self, iobj)
8289 except errors.OpExecError:
8290 self.LogWarning("Device creation failed, reverting...")
8292 _RemoveDisks(self, iobj)
8294 self.cfg.ReleaseDRBDMinors(instance)
8297 feedback_fn("adding instance %s to cluster config" % instance)
8299 self.cfg.AddInstance(iobj, self.proc.GetECId())
8301 # Declare that we don't want to remove the instance lock anymore, as we've
8302 # added the instance to the config
8303 del self.remove_locks[locking.LEVEL_INSTANCE]
8305 if self.op.mode == constants.INSTANCE_IMPORT:
8306 # Release unused nodes
8307 _ReleaseLocks(self, locking.LEVEL_NODE, keep=[self.op.src_node])
8310 _ReleaseLocks(self, locking.LEVEL_NODE)
8313 if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
8314 feedback_fn("* wiping instance disks...")
8316 _WipeDisks(self, iobj)
8317 except errors.OpExecError, err:
8318 logging.exception("Wiping disks failed")
8319 self.LogWarning("Wiping instance disks failed (%s)", err)
8323 # Something is already wrong with the disks, don't do anything else
8325 elif self.op.wait_for_sync:
8326 disk_abort = not _WaitForSync(self, iobj)
8327 elif iobj.disk_template in constants.DTS_INT_MIRROR:
8328 # make sure the disks are not degraded (still sync-ing is ok)
8330 feedback_fn("* checking mirrors status")
8331 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
8336 _RemoveDisks(self, iobj)
8337 self.cfg.RemoveInstance(iobj.name)
8338 # Make sure the instance lock gets removed
8339 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
8340 raise errors.OpExecError("There are some degraded disks for"
8343 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
8344 if self.op.mode == constants.INSTANCE_CREATE:
8345 if not self.op.no_install:
8346 feedback_fn("* running the instance OS create scripts...")
8347 # FIXME: pass debug option from opcode to backend
8348 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
8349 self.op.debug_level)
8350 result.Raise("Could not add os for instance %s"
8351 " on node %s" % (instance, pnode_name))
8353 elif self.op.mode == constants.INSTANCE_IMPORT:
8354 feedback_fn("* running the instance OS import scripts...")
8358 for idx, image in enumerate(self.src_images):
8362 # FIXME: pass debug option from opcode to backend
8363 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
8364 constants.IEIO_FILE, (image, ),
8365 constants.IEIO_SCRIPT,
8366 (iobj.disks[idx], idx),
8368 transfers.append(dt)
8371 masterd.instance.TransferInstanceData(self, feedback_fn,
8372 self.op.src_node, pnode_name,
8373 self.pnode.secondary_ip,
8375 if not compat.all(import_result):
8376 self.LogWarning("Some disks for instance %s on node %s were not"
8377 " imported successfully" % (instance, pnode_name))
8379 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
8380 feedback_fn("* preparing remote import...")
8381 # The source cluster will stop the instance before attempting to make a
8382 # connection. In some cases stopping an instance can take a long time,
8383 # hence the shutdown timeout is added to the connection timeout.
8384 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
8385 self.op.source_shutdown_timeout)
8386 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
8388 assert iobj.primary_node == self.pnode.name
8390 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
8391 self.source_x509_ca,
8392 self._cds, timeouts)
8393 if not compat.all(disk_results):
8394 # TODO: Should the instance still be started, even if some disks
8395 # failed to import (valid for local imports, too)?
8396 self.LogWarning("Some disks for instance %s on node %s were not"
8397 " imported successfully" % (instance, pnode_name))
8399 # Run rename script on newly imported instance
8400 assert iobj.name == instance
8401 feedback_fn("Running rename script for %s" % instance)
8402 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
8403 self.source_instance_name,
8404 self.op.debug_level)
8406 self.LogWarning("Failed to run rename script for %s on node"
8407 " %s: %s" % (instance, pnode_name, result.fail_msg))
8410 # also checked in the prereq part
8411 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
8415 iobj.admin_up = True
8416 self.cfg.Update(iobj, feedback_fn)
8417 logging.info("Starting instance %s on node %s", instance, pnode_name)
8418 feedback_fn("* starting instance...")
8419 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
8420 result.Raise("Could not start instance")
8422 return list(iobj.all_nodes)
8425 class LUInstanceConsole(NoHooksLU):
8426 """Connect to an instance's console.
8428 This is somewhat special in that it returns the command line that
8429 you need to run on the master node in order to connect to the
8435 def ExpandNames(self):
8436 self._ExpandAndLockInstance()
8438 def CheckPrereq(self):
8439 """Check prerequisites.
8441 This checks that the instance is in the cluster.
8444 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8445 assert self.instance is not None, \
8446 "Cannot retrieve locked instance %s" % self.op.instance_name
8447 _CheckNodeOnline(self, self.instance.primary_node)
8449 def Exec(self, feedback_fn):
8450 """Connect to the console of an instance
8453 instance = self.instance
8454 node = instance.primary_node
8456 node_insts = self.rpc.call_instance_list([node],
8457 [instance.hypervisor])[node]
8458 node_insts.Raise("Can't get node information from %s" % node)
8460 if instance.name not in node_insts.payload:
8461 if instance.admin_up:
8462 state = constants.INSTST_ERRORDOWN
8464 state = constants.INSTST_ADMINDOWN
8465 raise errors.OpExecError("Instance %s is not running (state %s)" %
8466 (instance.name, state))
8468 logging.debug("Connecting to console of %s on %s", instance.name, node)
8470 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
8473 def _GetInstanceConsole(cluster, instance):
8474 """Returns console information for an instance.
8476 @type cluster: L{objects.Cluster}
8477 @type instance: L{objects.Instance}
8481 hyper = hypervisor.GetHypervisor(instance.hypervisor)
8482 # beparams and hvparams are passed separately, to avoid editing the
8483 # instance and then saving the defaults in the instance itself.
8484 hvparams = cluster.FillHV(instance)
8485 beparams = cluster.FillBE(instance)
8486 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
8488 assert console.instance == instance.name
8489 assert console.Validate()
8491 return console.ToDict()
8494 class LUInstanceReplaceDisks(LogicalUnit):
8495 """Replace the disks of an instance.
8498 HPATH = "mirrors-replace"
8499 HTYPE = constants.HTYPE_INSTANCE
8502 def CheckArguments(self):
8503 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
8506 def ExpandNames(self):
8507 self._ExpandAndLockInstance()
8509 assert locking.LEVEL_NODE not in self.needed_locks
8510 assert locking.LEVEL_NODEGROUP not in self.needed_locks
8512 assert self.op.iallocator is None or self.op.remote_node is None, \
8513 "Conflicting options"
8515 if self.op.remote_node is not None:
8516 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8518 # Warning: do not remove the locking of the new secondary here
8519 # unless DRBD8.AddChildren is changed to work in parallel;
8520 # currently it doesn't since parallel invocations of
8521 # FindUnusedMinor will conflict
8522 self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
8523 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
8525 self.needed_locks[locking.LEVEL_NODE] = []
8526 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8528 if self.op.iallocator is not None:
8529 # iallocator will select a new node in the same group
8530 self.needed_locks[locking.LEVEL_NODEGROUP] = []
8532 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
8533 self.op.iallocator, self.op.remote_node,
8534 self.op.disks, False, self.op.early_release)
8536 self.tasklets = [self.replacer]
8538 def DeclareLocks(self, level):
8539 if level == locking.LEVEL_NODEGROUP:
8540 assert self.op.remote_node is None
8541 assert self.op.iallocator is not None
8542 assert not self.needed_locks[locking.LEVEL_NODEGROUP]
8544 self.share_locks[locking.LEVEL_NODEGROUP] = 1
8545 self.needed_locks[locking.LEVEL_NODEGROUP] = \
8546 self.cfg.GetInstanceNodeGroups(self.op.instance_name)
8548 elif level == locking.LEVEL_NODE:
8549 if self.op.iallocator is not None:
8550 assert self.op.remote_node is None
8551 assert not self.needed_locks[locking.LEVEL_NODE]
8553 # Lock member nodes of all locked groups
8554 self.needed_locks[locking.LEVEL_NODE] = [node_name
8555 for group_uuid in self.glm.list_owned(locking.LEVEL_NODEGROUP)
8556 for node_name in self.cfg.GetNodeGroup(group_uuid).members]
8558 self._LockInstancesNodes()
8560 def BuildHooksEnv(self):
8563 This runs on the master, the primary and all the secondaries.
8566 instance = self.replacer.instance
8568 "MODE": self.op.mode,
8569 "NEW_SECONDARY": self.op.remote_node,
8570 "OLD_SECONDARY": instance.secondary_nodes[0],
8572 env.update(_BuildInstanceHookEnvByObject(self, instance))
8575 def BuildHooksNodes(self):
8576 """Build hooks nodes.
8579 instance = self.replacer.instance
8581 self.cfg.GetMasterNode(),
8582 instance.primary_node,
8584 if self.op.remote_node is not None:
8585 nl.append(self.op.remote_node)
8588 def CheckPrereq(self):
8589 """Check prerequisites.
8592 assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
8593 self.op.iallocator is None)
8595 owned_groups = self.glm.list_owned(locking.LEVEL_NODEGROUP)
8597 groups = self.cfg.GetInstanceNodeGroups(self.op.instance_name)
8598 if owned_groups != groups:
8599 raise errors.OpExecError("Node groups used by instance '%s' changed"
8600 " since lock was acquired, current list is %r,"
8601 " used to be '%s'" %
8602 (self.op.instance_name,
8603 utils.CommaJoin(groups),
8604 utils.CommaJoin(owned_groups)))
8606 return LogicalUnit.CheckPrereq(self)
8609 class TLReplaceDisks(Tasklet):
8610 """Replaces disks for an instance.
8612 Note: Locking is not within the scope of this class.
8615 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
8616 disks, delay_iallocator, early_release):
8617 """Initializes this class.
8620 Tasklet.__init__(self, lu)
8623 self.instance_name = instance_name
8625 self.iallocator_name = iallocator_name
8626 self.remote_node = remote_node
8628 self.delay_iallocator = delay_iallocator
8629 self.early_release = early_release
8632 self.instance = None
8633 self.new_node = None
8634 self.target_node = None
8635 self.other_node = None
8636 self.remote_node_info = None
8637 self.node_secondary_ip = None
8640 def CheckArguments(mode, remote_node, iallocator):
8641 """Helper function for users of this class.
8644 # check for valid parameter combination
8645 if mode == constants.REPLACE_DISK_CHG:
8646 if remote_node is None and iallocator is None:
8647 raise errors.OpPrereqError("When changing the secondary either an"
8648 " iallocator script must be used or the"
8649 " new node given", errors.ECODE_INVAL)
8651 if remote_node is not None and iallocator is not None:
8652 raise errors.OpPrereqError("Give either the iallocator or the new"
8653 " secondary, not both", errors.ECODE_INVAL)
8655 elif remote_node is not None or iallocator is not None:
8656 # Not replacing the secondary
8657 raise errors.OpPrereqError("The iallocator and new node options can"
8658 " only be used when changing the"
8659 " secondary node", errors.ECODE_INVAL)
8662 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
8663 """Compute a new secondary node using an IAllocator.
8666 ial = IAllocator(lu.cfg, lu.rpc,
8667 mode=constants.IALLOCATOR_MODE_RELOC,
8669 relocate_from=relocate_from)
8671 ial.Run(iallocator_name)
8674 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
8675 " %s" % (iallocator_name, ial.info),
8678 if len(ial.result) != ial.required_nodes:
8679 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8680 " of nodes (%s), required %s" %
8682 len(ial.result), ial.required_nodes),
8685 remote_node_name = ial.result[0]
8687 lu.LogInfo("Selected new secondary for instance '%s': %s",
8688 instance_name, remote_node_name)
8690 return remote_node_name
8692 def _FindFaultyDisks(self, node_name):
8693 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
8696 def _CheckDisksActivated(self, instance):
8697 """Checks if the instance disks are activated.
8699 @param instance: The instance to check disks
8700 @return: True if they are activated, False otherwise
8703 nodes = instance.all_nodes
8705 for idx, dev in enumerate(instance.disks):
8707 self.lu.LogInfo("Checking disk/%d on %s", idx, node)
8708 self.cfg.SetDiskID(dev, node)
8710 result = self.rpc.call_blockdev_find(node, dev)
8714 elif result.fail_msg or not result.payload:
8719 def CheckPrereq(self):
8720 """Check prerequisites.
8722 This checks that the instance is in the cluster.
8725 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
8726 assert instance is not None, \
8727 "Cannot retrieve locked instance %s" % self.instance_name
8729 if instance.disk_template != constants.DT_DRBD8:
8730 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
8731 " instances", errors.ECODE_INVAL)
8733 if len(instance.secondary_nodes) != 1:
8734 raise errors.OpPrereqError("The instance has a strange layout,"
8735 " expected one secondary but found %d" %
8736 len(instance.secondary_nodes),
8739 if not self.delay_iallocator:
8740 self._CheckPrereq2()
8742 def _CheckPrereq2(self):
8743 """Check prerequisites, second part.
8745 This function should always be part of CheckPrereq. It was separated and is
8746 now called from Exec because during node evacuation iallocator was only
8747 called with an unmodified cluster model, not taking planned changes into
8751 instance = self.instance
8752 secondary_node = instance.secondary_nodes[0]
8754 if self.iallocator_name is None:
8755 remote_node = self.remote_node
8757 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
8758 instance.name, instance.secondary_nodes)
8760 if remote_node is None:
8761 self.remote_node_info = None
8763 assert remote_node in self.lu.glm.list_owned(locking.LEVEL_NODE), \
8764 "Remote node '%s' is not locked" % remote_node
8766 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
8767 assert self.remote_node_info is not None, \
8768 "Cannot retrieve locked node %s" % remote_node
8770 if remote_node == self.instance.primary_node:
8771 raise errors.OpPrereqError("The specified node is the primary node of"
8772 " the instance", errors.ECODE_INVAL)
8774 if remote_node == secondary_node:
8775 raise errors.OpPrereqError("The specified node is already the"
8776 " secondary node of the instance",
8779 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
8780 constants.REPLACE_DISK_CHG):
8781 raise errors.OpPrereqError("Cannot specify disks to be replaced",
8784 if self.mode == constants.REPLACE_DISK_AUTO:
8785 if not self._CheckDisksActivated(instance):
8786 raise errors.OpPrereqError("Please run activate-disks on instance %s"
8787 " first" % self.instance_name,
8789 faulty_primary = self._FindFaultyDisks(instance.primary_node)
8790 faulty_secondary = self._FindFaultyDisks(secondary_node)
8792 if faulty_primary and faulty_secondary:
8793 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
8794 " one node and can not be repaired"
8795 " automatically" % self.instance_name,
8799 self.disks = faulty_primary
8800 self.target_node = instance.primary_node
8801 self.other_node = secondary_node
8802 check_nodes = [self.target_node, self.other_node]
8803 elif faulty_secondary:
8804 self.disks = faulty_secondary
8805 self.target_node = secondary_node
8806 self.other_node = instance.primary_node
8807 check_nodes = [self.target_node, self.other_node]
8813 # Non-automatic modes
8814 if self.mode == constants.REPLACE_DISK_PRI:
8815 self.target_node = instance.primary_node
8816 self.other_node = secondary_node
8817 check_nodes = [self.target_node, self.other_node]
8819 elif self.mode == constants.REPLACE_DISK_SEC:
8820 self.target_node = secondary_node
8821 self.other_node = instance.primary_node
8822 check_nodes = [self.target_node, self.other_node]
8824 elif self.mode == constants.REPLACE_DISK_CHG:
8825 self.new_node = remote_node
8826 self.other_node = instance.primary_node
8827 self.target_node = secondary_node
8828 check_nodes = [self.new_node, self.other_node]
8830 _CheckNodeNotDrained(self.lu, remote_node)
8831 _CheckNodeVmCapable(self.lu, remote_node)
8833 old_node_info = self.cfg.GetNodeInfo(secondary_node)
8834 assert old_node_info is not None
8835 if old_node_info.offline and not self.early_release:
8836 # doesn't make sense to delay the release
8837 self.early_release = True
8838 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
8839 " early-release mode", secondary_node)
8842 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
8845 # If not specified all disks should be replaced
8847 self.disks = range(len(self.instance.disks))
8849 for node in check_nodes:
8850 _CheckNodeOnline(self.lu, node)
8852 touched_nodes = frozenset(node_name for node_name in [self.new_node,
8855 if node_name is not None)
8857 # Release unneeded node locks
8858 _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
8860 # Release any owned node group
8861 if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
8862 _ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
8864 # Check whether disks are valid
8865 for disk_idx in self.disks:
8866 instance.FindDisk(disk_idx)
8868 # Get secondary node IP addresses
8869 self.node_secondary_ip = \
8870 dict((node_name, self.cfg.GetNodeInfo(node_name).secondary_ip)
8871 for node_name in touched_nodes)
8873 def Exec(self, feedback_fn):
8874 """Execute disk replacement.
8876 This dispatches the disk replacement to the appropriate handler.
8879 if self.delay_iallocator:
8880 self._CheckPrereq2()
8883 # Verify owned locks before starting operation
8884 owned_locks = self.lu.glm.list_owned(locking.LEVEL_NODE)
8885 assert set(owned_locks) == set(self.node_secondary_ip), \
8886 ("Incorrect node locks, owning %s, expected %s" %
8887 (owned_locks, self.node_secondary_ip.keys()))
8889 owned_locks = self.lu.glm.list_owned(locking.LEVEL_INSTANCE)
8890 assert list(owned_locks) == [self.instance_name], \
8891 "Instance '%s' not locked" % self.instance_name
8893 assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
8894 "Should not own any node group lock at this point"
8897 feedback_fn("No disks need replacement")
8900 feedback_fn("Replacing disk(s) %s for %s" %
8901 (utils.CommaJoin(self.disks), self.instance.name))
8903 activate_disks = (not self.instance.admin_up)
8905 # Activate the instance disks if we're replacing them on a down instance
8907 _StartInstanceDisks(self.lu, self.instance, True)
8910 # Should we replace the secondary node?
8911 if self.new_node is not None:
8912 fn = self._ExecDrbd8Secondary
8914 fn = self._ExecDrbd8DiskOnly
8916 result = fn(feedback_fn)
8918 # Deactivate the instance disks if we're replacing them on a
8921 _SafeShutdownInstanceDisks(self.lu, self.instance)
8924 # Verify owned locks
8925 owned_locks = self.lu.glm.list_owned(locking.LEVEL_NODE)
8926 nodes = frozenset(self.node_secondary_ip)
8927 assert ((self.early_release and not owned_locks) or
8928 (not self.early_release and not (set(owned_locks) - nodes))), \
8929 ("Not owning the correct locks, early_release=%s, owned=%r,"
8930 " nodes=%r" % (self.early_release, owned_locks, nodes))
8934 def _CheckVolumeGroup(self, nodes):
8935 self.lu.LogInfo("Checking volume groups")
8937 vgname = self.cfg.GetVGName()
8939 # Make sure volume group exists on all involved nodes
8940 results = self.rpc.call_vg_list(nodes)
8942 raise errors.OpExecError("Can't list volume groups on the nodes")
8946 res.Raise("Error checking node %s" % node)
8947 if vgname not in res.payload:
8948 raise errors.OpExecError("Volume group '%s' not found on node %s" %
8951 def _CheckDisksExistence(self, nodes):
8952 # Check disk existence
8953 for idx, dev in enumerate(self.instance.disks):
8954 if idx not in self.disks:
8958 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
8959 self.cfg.SetDiskID(dev, node)
8961 result = self.rpc.call_blockdev_find(node, dev)
8963 msg = result.fail_msg
8964 if msg or not result.payload:
8966 msg = "disk not found"
8967 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
8970 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
8971 for idx, dev in enumerate(self.instance.disks):
8972 if idx not in self.disks:
8975 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
8978 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
8980 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
8981 " replace disks for instance %s" %
8982 (node_name, self.instance.name))
8984 def _CreateNewStorage(self, node_name):
8987 for idx, dev in enumerate(self.instance.disks):
8988 if idx not in self.disks:
8991 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
8993 self.cfg.SetDiskID(dev, node_name)
8995 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
8996 names = _GenerateUniqueNames(self.lu, lv_names)
8998 vg_data = dev.children[0].logical_id[0]
8999 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
9000 logical_id=(vg_data, names[0]))
9001 vg_meta = dev.children[1].logical_id[0]
9002 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
9003 logical_id=(vg_meta, names[1]))
9005 new_lvs = [lv_data, lv_meta]
9006 old_lvs = dev.children
9007 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
9009 # we pass force_create=True to force the LVM creation
9010 for new_lv in new_lvs:
9011 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
9012 _GetInstanceInfoText(self.instance), False)
9016 def _CheckDevices(self, node_name, iv_names):
9017 for name, (dev, _, _) in iv_names.iteritems():
9018 self.cfg.SetDiskID(dev, node_name)
9020 result = self.rpc.call_blockdev_find(node_name, dev)
9022 msg = result.fail_msg
9023 if msg or not result.payload:
9025 msg = "disk not found"
9026 raise errors.OpExecError("Can't find DRBD device %s: %s" %
9029 if result.payload.is_degraded:
9030 raise errors.OpExecError("DRBD device %s is degraded!" % name)
9032 def _RemoveOldStorage(self, node_name, iv_names):
9033 for name, (_, old_lvs, _) in iv_names.iteritems():
9034 self.lu.LogInfo("Remove logical volumes for %s" % name)
9037 self.cfg.SetDiskID(lv, node_name)
9039 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
9041 self.lu.LogWarning("Can't remove old LV: %s" % msg,
9042 hint="remove unused LVs manually")
9044 def _ExecDrbd8DiskOnly(self, feedback_fn):
9045 """Replace a disk on the primary or secondary for DRBD 8.
9047 The algorithm for replace is quite complicated:
9049 1. for each disk to be replaced:
9051 1. create new LVs on the target node with unique names
9052 1. detach old LVs from the drbd device
9053 1. rename old LVs to name_replaced.<time_t>
9054 1. rename new LVs to old LVs
9055 1. attach the new LVs (with the old names now) to the drbd device
9057 1. wait for sync across all devices
9059 1. for each modified disk:
9061 1. remove old LVs (which have the name name_replaces.<time_t>)
9063 Failures are not very well handled.
9068 # Step: check device activation
9069 self.lu.LogStep(1, steps_total, "Check device existence")
9070 self._CheckDisksExistence([self.other_node, self.target_node])
9071 self._CheckVolumeGroup([self.target_node, self.other_node])
9073 # Step: check other node consistency
9074 self.lu.LogStep(2, steps_total, "Check peer consistency")
9075 self._CheckDisksConsistency(self.other_node,
9076 self.other_node == self.instance.primary_node,
9079 # Step: create new storage
9080 self.lu.LogStep(3, steps_total, "Allocate new storage")
9081 iv_names = self._CreateNewStorage(self.target_node)
9083 # Step: for each lv, detach+rename*2+attach
9084 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9085 for dev, old_lvs, new_lvs in iv_names.itervalues():
9086 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
9088 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
9090 result.Raise("Can't detach drbd from local storage on node"
9091 " %s for device %s" % (self.target_node, dev.iv_name))
9093 #cfg.Update(instance)
9095 # ok, we created the new LVs, so now we know we have the needed
9096 # storage; as such, we proceed on the target node to rename
9097 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
9098 # using the assumption that logical_id == physical_id (which in
9099 # turn is the unique_id on that node)
9101 # FIXME(iustin): use a better name for the replaced LVs
9102 temp_suffix = int(time.time())
9103 ren_fn = lambda d, suff: (d.physical_id[0],
9104 d.physical_id[1] + "_replaced-%s" % suff)
9106 # Build the rename list based on what LVs exist on the node
9107 rename_old_to_new = []
9108 for to_ren in old_lvs:
9109 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
9110 if not result.fail_msg and result.payload:
9112 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
9114 self.lu.LogInfo("Renaming the old LVs on the target node")
9115 result = self.rpc.call_blockdev_rename(self.target_node,
9117 result.Raise("Can't rename old LVs on node %s" % self.target_node)
9119 # Now we rename the new LVs to the old LVs
9120 self.lu.LogInfo("Renaming the new LVs on the target node")
9121 rename_new_to_old = [(new, old.physical_id)
9122 for old, new in zip(old_lvs, new_lvs)]
9123 result = self.rpc.call_blockdev_rename(self.target_node,
9125 result.Raise("Can't rename new LVs on node %s" % self.target_node)
9127 for old, new in zip(old_lvs, new_lvs):
9128 new.logical_id = old.logical_id
9129 self.cfg.SetDiskID(new, self.target_node)
9131 for disk in old_lvs:
9132 disk.logical_id = ren_fn(disk, temp_suffix)
9133 self.cfg.SetDiskID(disk, self.target_node)
9135 # Now that the new lvs have the old name, we can add them to the device
9136 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
9137 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
9139 msg = result.fail_msg
9141 for new_lv in new_lvs:
9142 msg2 = self.rpc.call_blockdev_remove(self.target_node,
9145 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
9146 hint=("cleanup manually the unused logical"
9148 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
9150 dev.children = new_lvs
9152 self.cfg.Update(self.instance, feedback_fn)
9155 if self.early_release:
9156 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9158 self._RemoveOldStorage(self.target_node, iv_names)
9159 # WARNING: we release both node locks here, do not do other RPCs
9160 # than WaitForSync to the primary node
9161 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9162 names=[self.target_node, self.other_node])
9165 # This can fail as the old devices are degraded and _WaitForSync
9166 # does a combined result over all disks, so we don't check its return value
9167 self.lu.LogStep(cstep, steps_total, "Sync devices")
9169 _WaitForSync(self.lu, self.instance)
9171 # Check all devices manually
9172 self._CheckDevices(self.instance.primary_node, iv_names)
9174 # Step: remove old storage
9175 if not self.early_release:
9176 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9178 self._RemoveOldStorage(self.target_node, iv_names)
9180 def _ExecDrbd8Secondary(self, feedback_fn):
9181 """Replace the secondary node for DRBD 8.
9183 The algorithm for replace is quite complicated:
9184 - for all disks of the instance:
9185 - create new LVs on the new node with same names
9186 - shutdown the drbd device on the old secondary
9187 - disconnect the drbd network on the primary
9188 - create the drbd device on the new secondary
9189 - network attach the drbd on the primary, using an artifice:
9190 the drbd code for Attach() will connect to the network if it
9191 finds a device which is connected to the good local disks but
9193 - wait for sync across all devices
9194 - remove all disks from the old secondary
9196 Failures are not very well handled.
9201 # Step: check device activation
9202 self.lu.LogStep(1, steps_total, "Check device existence")
9203 self._CheckDisksExistence([self.instance.primary_node])
9204 self._CheckVolumeGroup([self.instance.primary_node])
9206 # Step: check other node consistency
9207 self.lu.LogStep(2, steps_total, "Check peer consistency")
9208 self._CheckDisksConsistency(self.instance.primary_node, True, True)
9210 # Step: create new storage
9211 self.lu.LogStep(3, steps_total, "Allocate new storage")
9212 for idx, dev in enumerate(self.instance.disks):
9213 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
9214 (self.new_node, idx))
9215 # we pass force_create=True to force LVM creation
9216 for new_lv in dev.children:
9217 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
9218 _GetInstanceInfoText(self.instance), False)
9220 # Step 4: dbrd minors and drbd setups changes
9221 # after this, we must manually remove the drbd minors on both the
9222 # error and the success paths
9223 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
9224 minors = self.cfg.AllocateDRBDMinor([self.new_node
9225 for dev in self.instance.disks],
9227 logging.debug("Allocated minors %r", minors)
9230 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
9231 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
9232 (self.new_node, idx))
9233 # create new devices on new_node; note that we create two IDs:
9234 # one without port, so the drbd will be activated without
9235 # networking information on the new node at this stage, and one
9236 # with network, for the latter activation in step 4
9237 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
9238 if self.instance.primary_node == o_node1:
9241 assert self.instance.primary_node == o_node2, "Three-node instance?"
9244 new_alone_id = (self.instance.primary_node, self.new_node, None,
9245 p_minor, new_minor, o_secret)
9246 new_net_id = (self.instance.primary_node, self.new_node, o_port,
9247 p_minor, new_minor, o_secret)
9249 iv_names[idx] = (dev, dev.children, new_net_id)
9250 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
9252 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
9253 logical_id=new_alone_id,
9254 children=dev.children,
9257 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
9258 _GetInstanceInfoText(self.instance), False)
9259 except errors.GenericError:
9260 self.cfg.ReleaseDRBDMinors(self.instance.name)
9263 # We have new devices, shutdown the drbd on the old secondary
9264 for idx, dev in enumerate(self.instance.disks):
9265 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
9266 self.cfg.SetDiskID(dev, self.target_node)
9267 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
9269 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
9270 "node: %s" % (idx, msg),
9271 hint=("Please cleanup this device manually as"
9272 " soon as possible"))
9274 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
9275 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
9276 self.node_secondary_ip,
9277 self.instance.disks)\
9278 [self.instance.primary_node]
9280 msg = result.fail_msg
9282 # detaches didn't succeed (unlikely)
9283 self.cfg.ReleaseDRBDMinors(self.instance.name)
9284 raise errors.OpExecError("Can't detach the disks from the network on"
9285 " old node: %s" % (msg,))
9287 # if we managed to detach at least one, we update all the disks of
9288 # the instance to point to the new secondary
9289 self.lu.LogInfo("Updating instance configuration")
9290 for dev, _, new_logical_id in iv_names.itervalues():
9291 dev.logical_id = new_logical_id
9292 self.cfg.SetDiskID(dev, self.instance.primary_node)
9294 self.cfg.Update(self.instance, feedback_fn)
9296 # and now perform the drbd attach
9297 self.lu.LogInfo("Attaching primary drbds to new secondary"
9298 " (standalone => connected)")
9299 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
9301 self.node_secondary_ip,
9302 self.instance.disks,
9305 for to_node, to_result in result.items():
9306 msg = to_result.fail_msg
9308 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
9310 hint=("please do a gnt-instance info to see the"
9311 " status of disks"))
9313 if self.early_release:
9314 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9316 self._RemoveOldStorage(self.target_node, iv_names)
9317 # WARNING: we release all node locks here, do not do other RPCs
9318 # than WaitForSync to the primary node
9319 _ReleaseLocks(self.lu, locking.LEVEL_NODE,
9320 names=[self.instance.primary_node,
9325 # This can fail as the old devices are degraded and _WaitForSync
9326 # does a combined result over all disks, so we don't check its return value
9327 self.lu.LogStep(cstep, steps_total, "Sync devices")
9329 _WaitForSync(self.lu, self.instance)
9331 # Check all devices manually
9332 self._CheckDevices(self.instance.primary_node, iv_names)
9334 # Step: remove old storage
9335 if not self.early_release:
9336 self.lu.LogStep(cstep, steps_total, "Removing old storage")
9337 self._RemoveOldStorage(self.target_node, iv_names)
9340 class LURepairNodeStorage(NoHooksLU):
9341 """Repairs the volume group on a node.
9346 def CheckArguments(self):
9347 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
9349 storage_type = self.op.storage_type
9351 if (constants.SO_FIX_CONSISTENCY not in
9352 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
9353 raise errors.OpPrereqError("Storage units of type '%s' can not be"
9354 " repaired" % storage_type,
9357 def ExpandNames(self):
9358 self.needed_locks = {
9359 locking.LEVEL_NODE: [self.op.node_name],
9362 def _CheckFaultyDisks(self, instance, node_name):
9363 """Ensure faulty disks abort the opcode or at least warn."""
9365 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
9367 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
9368 " node '%s'" % (instance.name, node_name),
9370 except errors.OpPrereqError, err:
9371 if self.op.ignore_consistency:
9372 self.proc.LogWarning(str(err.args[0]))
9376 def CheckPrereq(self):
9377 """Check prerequisites.
9380 # Check whether any instance on this node has faulty disks
9381 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
9382 if not inst.admin_up:
9384 check_nodes = set(inst.all_nodes)
9385 check_nodes.discard(self.op.node_name)
9386 for inst_node_name in check_nodes:
9387 self._CheckFaultyDisks(inst, inst_node_name)
9389 def Exec(self, feedback_fn):
9390 feedback_fn("Repairing storage unit '%s' on %s ..." %
9391 (self.op.name, self.op.node_name))
9393 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
9394 result = self.rpc.call_storage_execute(self.op.node_name,
9395 self.op.storage_type, st_args,
9397 constants.SO_FIX_CONSISTENCY)
9398 result.Raise("Failed to repair storage unit '%s' on %s" %
9399 (self.op.name, self.op.node_name))
9402 class LUNodeEvacStrategy(NoHooksLU):
9403 """Computes the node evacuation strategy.
9408 def CheckArguments(self):
9409 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
9411 def ExpandNames(self):
9412 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
9413 self.needed_locks = locks = {}
9414 if self.op.remote_node is None:
9415 locks[locking.LEVEL_NODE] = locking.ALL_SET
9417 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9418 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
9420 def Exec(self, feedback_fn):
9421 if self.op.remote_node is not None:
9423 for node in self.op.nodes:
9424 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
9427 if i.primary_node == self.op.remote_node:
9428 raise errors.OpPrereqError("Node %s is the primary node of"
9429 " instance %s, cannot use it as"
9431 (self.op.remote_node, i.name),
9433 result.append([i.name, self.op.remote_node])
9435 ial = IAllocator(self.cfg, self.rpc,
9436 mode=constants.IALLOCATOR_MODE_MEVAC,
9437 evac_nodes=self.op.nodes)
9438 ial.Run(self.op.iallocator, validate=True)
9440 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
9446 class LUInstanceGrowDisk(LogicalUnit):
9447 """Grow a disk of an instance.
9451 HTYPE = constants.HTYPE_INSTANCE
9454 def ExpandNames(self):
9455 self._ExpandAndLockInstance()
9456 self.needed_locks[locking.LEVEL_NODE] = []
9457 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9459 def DeclareLocks(self, level):
9460 if level == locking.LEVEL_NODE:
9461 self._LockInstancesNodes()
9463 def BuildHooksEnv(self):
9466 This runs on the master, the primary and all the secondaries.
9470 "DISK": self.op.disk,
9471 "AMOUNT": self.op.amount,
9473 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9476 def BuildHooksNodes(self):
9477 """Build hooks nodes.
9480 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9483 def CheckPrereq(self):
9484 """Check prerequisites.
9486 This checks that the instance is in the cluster.
9489 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9490 assert instance is not None, \
9491 "Cannot retrieve locked instance %s" % self.op.instance_name
9492 nodenames = list(instance.all_nodes)
9493 for node in nodenames:
9494 _CheckNodeOnline(self, node)
9496 self.instance = instance
9498 if instance.disk_template not in constants.DTS_GROWABLE:
9499 raise errors.OpPrereqError("Instance's disk layout does not support"
9500 " growing", errors.ECODE_INVAL)
9502 self.disk = instance.FindDisk(self.op.disk)
9504 if instance.disk_template not in (constants.DT_FILE,
9505 constants.DT_SHARED_FILE):
9506 # TODO: check the free disk space for file, when that feature will be
9508 _CheckNodesFreeDiskPerVG(self, nodenames,
9509 self.disk.ComputeGrowth(self.op.amount))
9511 def Exec(self, feedback_fn):
9512 """Execute disk grow.
9515 instance = self.instance
9518 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
9520 raise errors.OpExecError("Cannot activate block device to grow")
9522 for node in instance.all_nodes:
9523 self.cfg.SetDiskID(disk, node)
9524 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
9525 result.Raise("Grow request failed to node %s" % node)
9527 # TODO: Rewrite code to work properly
9528 # DRBD goes into sync mode for a short amount of time after executing the
9529 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
9530 # calling "resize" in sync mode fails. Sleeping for a short amount of
9531 # time is a work-around.
9534 disk.RecordGrow(self.op.amount)
9535 self.cfg.Update(instance, feedback_fn)
9536 if self.op.wait_for_sync:
9537 disk_abort = not _WaitForSync(self, instance, disks=[disk])
9539 self.proc.LogWarning("Disk sync-ing has not returned a good"
9540 " status; please check the instance")
9541 if not instance.admin_up:
9542 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
9543 elif not instance.admin_up:
9544 self.proc.LogWarning("Not shutting down the disk even if the instance is"
9545 " not supposed to be running because no wait for"
9546 " sync mode was requested")
9549 class LUInstanceQueryData(NoHooksLU):
9550 """Query runtime instance data.
9555 def ExpandNames(self):
9556 self.needed_locks = {}
9558 # Use locking if requested or when non-static information is wanted
9559 if not (self.op.static or self.op.use_locking):
9560 self.LogWarning("Non-static data requested, locks need to be acquired")
9561 self.op.use_locking = True
9563 if self.op.instances or not self.op.use_locking:
9564 # Expand instance names right here
9565 self.wanted_names = _GetWantedInstances(self, self.op.instances)
9567 # Will use acquired locks
9568 self.wanted_names = None
9570 if self.op.use_locking:
9571 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9573 if self.wanted_names is None:
9574 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
9576 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
9578 self.needed_locks[locking.LEVEL_NODE] = []
9579 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9580 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9582 def DeclareLocks(self, level):
9583 if self.op.use_locking and level == locking.LEVEL_NODE:
9584 self._LockInstancesNodes()
9586 def CheckPrereq(self):
9587 """Check prerequisites.
9589 This only checks the optional instance list against the existing names.
9592 if self.wanted_names is None:
9593 assert self.op.use_locking, "Locking was not used"
9594 self.wanted_names = self.glm.list_owned(locking.LEVEL_INSTANCE)
9596 self.wanted_instances = [self.cfg.GetInstanceInfo(name)
9597 for name in self.wanted_names]
9599 def _ComputeBlockdevStatus(self, node, instance_name, dev):
9600 """Returns the status of a block device
9603 if self.op.static or not node:
9606 self.cfg.SetDiskID(dev, node)
9608 result = self.rpc.call_blockdev_find(node, dev)
9612 result.Raise("Can't compute disk status for %s" % instance_name)
9614 status = result.payload
9618 return (status.dev_path, status.major, status.minor,
9619 status.sync_percent, status.estimated_time,
9620 status.is_degraded, status.ldisk_status)
9622 def _ComputeDiskStatus(self, instance, snode, dev):
9623 """Compute block device status.
9626 if dev.dev_type in constants.LDS_DRBD:
9627 # we change the snode then (otherwise we use the one passed in)
9628 if dev.logical_id[0] == instance.primary_node:
9629 snode = dev.logical_id[1]
9631 snode = dev.logical_id[0]
9633 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
9635 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
9638 dev_children = [self._ComputeDiskStatus(instance, snode, child)
9639 for child in dev.children]
9644 "iv_name": dev.iv_name,
9645 "dev_type": dev.dev_type,
9646 "logical_id": dev.logical_id,
9647 "physical_id": dev.physical_id,
9648 "pstatus": dev_pstatus,
9649 "sstatus": dev_sstatus,
9650 "children": dev_children,
9655 def Exec(self, feedback_fn):
9656 """Gather and return data"""
9659 cluster = self.cfg.GetClusterInfo()
9661 for instance in self.wanted_instances:
9662 if not self.op.static:
9663 remote_info = self.rpc.call_instance_info(instance.primary_node,
9665 instance.hypervisor)
9666 remote_info.Raise("Error checking node %s" % instance.primary_node)
9667 remote_info = remote_info.payload
9668 if remote_info and "state" in remote_info:
9671 remote_state = "down"
9674 if instance.admin_up:
9677 config_state = "down"
9679 disks = [self._ComputeDiskStatus(instance, None, device)
9680 for device in instance.disks]
9682 result[instance.name] = {
9683 "name": instance.name,
9684 "config_state": config_state,
9685 "run_state": remote_state,
9686 "pnode": instance.primary_node,
9687 "snodes": instance.secondary_nodes,
9689 # this happens to be the same format used for hooks
9690 "nics": _NICListToTuple(self, instance.nics),
9691 "disk_template": instance.disk_template,
9693 "hypervisor": instance.hypervisor,
9694 "network_port": instance.network_port,
9695 "hv_instance": instance.hvparams,
9696 "hv_actual": cluster.FillHV(instance, skip_globals=True),
9697 "be_instance": instance.beparams,
9698 "be_actual": cluster.FillBE(instance),
9699 "os_instance": instance.osparams,
9700 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
9701 "serial_no": instance.serial_no,
9702 "mtime": instance.mtime,
9703 "ctime": instance.ctime,
9704 "uuid": instance.uuid,
9710 class LUInstanceSetParams(LogicalUnit):
9711 """Modifies an instances's parameters.
9714 HPATH = "instance-modify"
9715 HTYPE = constants.HTYPE_INSTANCE
9718 def CheckArguments(self):
9719 if not (self.op.nics or self.op.disks or self.op.disk_template or
9720 self.op.hvparams or self.op.beparams or self.op.os_name):
9721 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
9723 if self.op.hvparams:
9724 _CheckGlobalHvParams(self.op.hvparams)
9728 for disk_op, disk_dict in self.op.disks:
9729 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
9730 if disk_op == constants.DDM_REMOVE:
9733 elif disk_op == constants.DDM_ADD:
9736 if not isinstance(disk_op, int):
9737 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
9738 if not isinstance(disk_dict, dict):
9739 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
9740 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9742 if disk_op == constants.DDM_ADD:
9743 mode = disk_dict.setdefault(constants.IDISK_MODE, constants.DISK_RDWR)
9744 if mode not in constants.DISK_ACCESS_SET:
9745 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
9747 size = disk_dict.get(constants.IDISK_SIZE, None)
9749 raise errors.OpPrereqError("Required disk parameter size missing",
9753 except (TypeError, ValueError), err:
9754 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
9755 str(err), errors.ECODE_INVAL)
9756 disk_dict[constants.IDISK_SIZE] = size
9758 # modification of disk
9759 if constants.IDISK_SIZE in disk_dict:
9760 raise errors.OpPrereqError("Disk size change not possible, use"
9761 " grow-disk", errors.ECODE_INVAL)
9763 if disk_addremove > 1:
9764 raise errors.OpPrereqError("Only one disk add or remove operation"
9765 " supported at a time", errors.ECODE_INVAL)
9767 if self.op.disks and self.op.disk_template is not None:
9768 raise errors.OpPrereqError("Disk template conversion and other disk"
9769 " changes not supported at the same time",
9772 if (self.op.disk_template and
9773 self.op.disk_template in constants.DTS_INT_MIRROR and
9774 self.op.remote_node is None):
9775 raise errors.OpPrereqError("Changing the disk template to a mirrored"
9776 " one requires specifying a secondary node",
9781 for nic_op, nic_dict in self.op.nics:
9782 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
9783 if nic_op == constants.DDM_REMOVE:
9786 elif nic_op == constants.DDM_ADD:
9789 if not isinstance(nic_op, int):
9790 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
9791 if not isinstance(nic_dict, dict):
9792 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
9793 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9795 # nic_dict should be a dict
9796 nic_ip = nic_dict.get(constants.INIC_IP, None)
9797 if nic_ip is not None:
9798 if nic_ip.lower() == constants.VALUE_NONE:
9799 nic_dict[constants.INIC_IP] = None
9801 if not netutils.IPAddress.IsValid(nic_ip):
9802 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
9805 nic_bridge = nic_dict.get('bridge', None)
9806 nic_link = nic_dict.get(constants.INIC_LINK, None)
9807 if nic_bridge and nic_link:
9808 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
9809 " at the same time", errors.ECODE_INVAL)
9810 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
9811 nic_dict['bridge'] = None
9812 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
9813 nic_dict[constants.INIC_LINK] = None
9815 if nic_op == constants.DDM_ADD:
9816 nic_mac = nic_dict.get(constants.INIC_MAC, None)
9818 nic_dict[constants.INIC_MAC] = constants.VALUE_AUTO
9820 if constants.INIC_MAC in nic_dict:
9821 nic_mac = nic_dict[constants.INIC_MAC]
9822 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9823 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
9825 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
9826 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
9827 " modifying an existing nic",
9830 if nic_addremove > 1:
9831 raise errors.OpPrereqError("Only one NIC add or remove operation"
9832 " supported at a time", errors.ECODE_INVAL)
9834 def ExpandNames(self):
9835 self._ExpandAndLockInstance()
9836 self.needed_locks[locking.LEVEL_NODE] = []
9837 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9839 def DeclareLocks(self, level):
9840 if level == locking.LEVEL_NODE:
9841 self._LockInstancesNodes()
9842 if self.op.disk_template and self.op.remote_node:
9843 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9844 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
9846 def BuildHooksEnv(self):
9849 This runs on the master, primary and secondaries.
9853 if constants.BE_MEMORY in self.be_new:
9854 args['memory'] = self.be_new[constants.BE_MEMORY]
9855 if constants.BE_VCPUS in self.be_new:
9856 args['vcpus'] = self.be_new[constants.BE_VCPUS]
9857 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
9858 # information at all.
9861 nic_override = dict(self.op.nics)
9862 for idx, nic in enumerate(self.instance.nics):
9863 if idx in nic_override:
9864 this_nic_override = nic_override[idx]
9866 this_nic_override = {}
9867 if constants.INIC_IP in this_nic_override:
9868 ip = this_nic_override[constants.INIC_IP]
9871 if constants.INIC_MAC in this_nic_override:
9872 mac = this_nic_override[constants.INIC_MAC]
9875 if idx in self.nic_pnew:
9876 nicparams = self.nic_pnew[idx]
9878 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
9879 mode = nicparams[constants.NIC_MODE]
9880 link = nicparams[constants.NIC_LINK]
9881 args['nics'].append((ip, mac, mode, link))
9882 if constants.DDM_ADD in nic_override:
9883 ip = nic_override[constants.DDM_ADD].get(constants.INIC_IP, None)
9884 mac = nic_override[constants.DDM_ADD][constants.INIC_MAC]
9885 nicparams = self.nic_pnew[constants.DDM_ADD]
9886 mode = nicparams[constants.NIC_MODE]
9887 link = nicparams[constants.NIC_LINK]
9888 args['nics'].append((ip, mac, mode, link))
9889 elif constants.DDM_REMOVE in nic_override:
9890 del args['nics'][-1]
9892 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
9893 if self.op.disk_template:
9894 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
9898 def BuildHooksNodes(self):
9899 """Build hooks nodes.
9902 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9905 def CheckPrereq(self):
9906 """Check prerequisites.
9908 This only checks the instance list against the existing names.
9911 # checking the new params on the primary/secondary nodes
9913 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9914 cluster = self.cluster = self.cfg.GetClusterInfo()
9915 assert self.instance is not None, \
9916 "Cannot retrieve locked instance %s" % self.op.instance_name
9917 pnode = instance.primary_node
9918 nodelist = list(instance.all_nodes)
9921 if self.op.os_name and not self.op.force:
9922 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
9923 self.op.force_variant)
9924 instance_os = self.op.os_name
9926 instance_os = instance.os
9928 if self.op.disk_template:
9929 if instance.disk_template == self.op.disk_template:
9930 raise errors.OpPrereqError("Instance already has disk template %s" %
9931 instance.disk_template, errors.ECODE_INVAL)
9933 if (instance.disk_template,
9934 self.op.disk_template) not in self._DISK_CONVERSIONS:
9935 raise errors.OpPrereqError("Unsupported disk template conversion from"
9936 " %s to %s" % (instance.disk_template,
9937 self.op.disk_template),
9939 _CheckInstanceDown(self, instance, "cannot change disk template")
9940 if self.op.disk_template in constants.DTS_INT_MIRROR:
9941 if self.op.remote_node == pnode:
9942 raise errors.OpPrereqError("Given new secondary node %s is the same"
9943 " as the primary node of the instance" %
9944 self.op.remote_node, errors.ECODE_STATE)
9945 _CheckNodeOnline(self, self.op.remote_node)
9946 _CheckNodeNotDrained(self, self.op.remote_node)
9947 # FIXME: here we assume that the old instance type is DT_PLAIN
9948 assert instance.disk_template == constants.DT_PLAIN
9949 disks = [{constants.IDISK_SIZE: d.size,
9950 constants.IDISK_VG: d.logical_id[0]}
9951 for d in instance.disks]
9952 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
9953 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
9955 # hvparams processing
9956 if self.op.hvparams:
9957 hv_type = instance.hypervisor
9958 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
9959 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
9960 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
9963 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
9964 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
9965 self.hv_new = hv_new # the new actual values
9966 self.hv_inst = i_hvdict # the new dict (without defaults)
9968 self.hv_new = self.hv_inst = {}
9970 # beparams processing
9971 if self.op.beparams:
9972 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
9974 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
9975 be_new = cluster.SimpleFillBE(i_bedict)
9976 self.be_new = be_new # the new actual values
9977 self.be_inst = i_bedict # the new dict (without defaults)
9979 self.be_new = self.be_inst = {}
9981 # osparams processing
9982 if self.op.osparams:
9983 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
9984 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
9985 self.os_inst = i_osdict # the new dict (without defaults)
9991 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
9992 mem_check_list = [pnode]
9993 if be_new[constants.BE_AUTO_BALANCE]:
9994 # either we changed auto_balance to yes or it was from before
9995 mem_check_list.extend(instance.secondary_nodes)
9996 instance_info = self.rpc.call_instance_info(pnode, instance.name,
9997 instance.hypervisor)
9998 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
9999 instance.hypervisor)
10000 pninfo = nodeinfo[pnode]
10001 msg = pninfo.fail_msg
10003 # Assume the primary node is unreachable and go ahead
10004 self.warn.append("Can't get info from primary node %s: %s" %
10006 elif not isinstance(pninfo.payload.get('memory_free', None), int):
10007 self.warn.append("Node data from primary node %s doesn't contain"
10008 " free memory information" % pnode)
10009 elif instance_info.fail_msg:
10010 self.warn.append("Can't get instance runtime information: %s" %
10011 instance_info.fail_msg)
10013 if instance_info.payload:
10014 current_mem = int(instance_info.payload['memory'])
10016 # Assume instance not running
10017 # (there is a slight race condition here, but it's not very probable,
10018 # and we have no other way to check)
10020 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
10021 pninfo.payload['memory_free'])
10023 raise errors.OpPrereqError("This change will prevent the instance"
10024 " from starting, due to %d MB of memory"
10025 " missing on its primary node" % miss_mem,
10026 errors.ECODE_NORES)
10028 if be_new[constants.BE_AUTO_BALANCE]:
10029 for node, nres in nodeinfo.items():
10030 if node not in instance.secondary_nodes:
10032 msg = nres.fail_msg
10034 self.warn.append("Can't get info from secondary node %s: %s" %
10036 elif not isinstance(nres.payload.get('memory_free', None), int):
10037 self.warn.append("Secondary node %s didn't return free"
10038 " memory information" % node)
10039 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
10040 self.warn.append("Not enough memory to failover instance to"
10041 " secondary node %s" % node)
10045 self.nic_pinst = {}
10046 for nic_op, nic_dict in self.op.nics:
10047 if nic_op == constants.DDM_REMOVE:
10048 if not instance.nics:
10049 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
10050 errors.ECODE_INVAL)
10052 if nic_op != constants.DDM_ADD:
10054 if not instance.nics:
10055 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
10056 " no NICs" % nic_op,
10057 errors.ECODE_INVAL)
10058 if nic_op < 0 or nic_op >= len(instance.nics):
10059 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
10061 (nic_op, len(instance.nics) - 1),
10062 errors.ECODE_INVAL)
10063 old_nic_params = instance.nics[nic_op].nicparams
10064 old_nic_ip = instance.nics[nic_op].ip
10066 old_nic_params = {}
10069 update_params_dict = dict([(key, nic_dict[key])
10070 for key in constants.NICS_PARAMETERS
10071 if key in nic_dict])
10073 if 'bridge' in nic_dict:
10074 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
10076 new_nic_params = _GetUpdatedParams(old_nic_params,
10077 update_params_dict)
10078 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
10079 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
10080 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
10081 self.nic_pinst[nic_op] = new_nic_params
10082 self.nic_pnew[nic_op] = new_filled_nic_params
10083 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
10085 if new_nic_mode == constants.NIC_MODE_BRIDGED:
10086 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
10087 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
10089 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
10091 self.warn.append(msg)
10093 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
10094 if new_nic_mode == constants.NIC_MODE_ROUTED:
10095 if constants.INIC_IP in nic_dict:
10096 nic_ip = nic_dict[constants.INIC_IP]
10098 nic_ip = old_nic_ip
10100 raise errors.OpPrereqError('Cannot set the nic ip to None'
10101 ' on a routed nic', errors.ECODE_INVAL)
10102 if constants.INIC_MAC in nic_dict:
10103 nic_mac = nic_dict[constants.INIC_MAC]
10104 if nic_mac is None:
10105 raise errors.OpPrereqError('Cannot set the nic mac to None',
10106 errors.ECODE_INVAL)
10107 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
10108 # otherwise generate the mac
10109 nic_dict[constants.INIC_MAC] = \
10110 self.cfg.GenerateMAC(self.proc.GetECId())
10112 # or validate/reserve the current one
10114 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
10115 except errors.ReservationError:
10116 raise errors.OpPrereqError("MAC address %s already in use"
10117 " in cluster" % nic_mac,
10118 errors.ECODE_NOTUNIQUE)
10121 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
10122 raise errors.OpPrereqError("Disk operations not supported for"
10123 " diskless instances",
10124 errors.ECODE_INVAL)
10125 for disk_op, _ in self.op.disks:
10126 if disk_op == constants.DDM_REMOVE:
10127 if len(instance.disks) == 1:
10128 raise errors.OpPrereqError("Cannot remove the last disk of"
10129 " an instance", errors.ECODE_INVAL)
10130 _CheckInstanceDown(self, instance, "cannot remove disks")
10132 if (disk_op == constants.DDM_ADD and
10133 len(instance.disks) >= constants.MAX_DISKS):
10134 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
10135 " add more" % constants.MAX_DISKS,
10136 errors.ECODE_STATE)
10137 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
10139 if disk_op < 0 or disk_op >= len(instance.disks):
10140 raise errors.OpPrereqError("Invalid disk index %s, valid values"
10142 (disk_op, len(instance.disks)),
10143 errors.ECODE_INVAL)
10147 def _ConvertPlainToDrbd(self, feedback_fn):
10148 """Converts an instance from plain to drbd.
10151 feedback_fn("Converting template to drbd")
10152 instance = self.instance
10153 pnode = instance.primary_node
10154 snode = self.op.remote_node
10156 # create a fake disk info for _GenerateDiskTemplate
10157 disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
10158 constants.IDISK_VG: d.logical_id[0]}
10159 for d in instance.disks]
10160 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
10161 instance.name, pnode, [snode],
10162 disk_info, None, None, 0, feedback_fn)
10163 info = _GetInstanceInfoText(instance)
10164 feedback_fn("Creating aditional volumes...")
10165 # first, create the missing data and meta devices
10166 for disk in new_disks:
10167 # unfortunately this is... not too nice
10168 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
10170 for child in disk.children:
10171 _CreateSingleBlockDev(self, snode, instance, child, info, True)
10172 # at this stage, all new LVs have been created, we can rename the
10174 feedback_fn("Renaming original volumes...")
10175 rename_list = [(o, n.children[0].logical_id)
10176 for (o, n) in zip(instance.disks, new_disks)]
10177 result = self.rpc.call_blockdev_rename(pnode, rename_list)
10178 result.Raise("Failed to rename original LVs")
10180 feedback_fn("Initializing DRBD devices...")
10181 # all child devices are in place, we can now create the DRBD devices
10182 for disk in new_disks:
10183 for node in [pnode, snode]:
10184 f_create = node == pnode
10185 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
10187 # at this point, the instance has been modified
10188 instance.disk_template = constants.DT_DRBD8
10189 instance.disks = new_disks
10190 self.cfg.Update(instance, feedback_fn)
10192 # disks are created, waiting for sync
10193 disk_abort = not _WaitForSync(self, instance)
10195 raise errors.OpExecError("There are some degraded disks for"
10196 " this instance, please cleanup manually")
10198 def _ConvertDrbdToPlain(self, feedback_fn):
10199 """Converts an instance from drbd to plain.
10202 instance = self.instance
10203 assert len(instance.secondary_nodes) == 1
10204 pnode = instance.primary_node
10205 snode = instance.secondary_nodes[0]
10206 feedback_fn("Converting template to plain")
10208 old_disks = instance.disks
10209 new_disks = [d.children[0] for d in old_disks]
10211 # copy over size and mode
10212 for parent, child in zip(old_disks, new_disks):
10213 child.size = parent.size
10214 child.mode = parent.mode
10216 # update instance structure
10217 instance.disks = new_disks
10218 instance.disk_template = constants.DT_PLAIN
10219 self.cfg.Update(instance, feedback_fn)
10221 feedback_fn("Removing volumes on the secondary node...")
10222 for disk in old_disks:
10223 self.cfg.SetDiskID(disk, snode)
10224 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
10226 self.LogWarning("Could not remove block device %s on node %s,"
10227 " continuing anyway: %s", disk.iv_name, snode, msg)
10229 feedback_fn("Removing unneeded volumes on the primary node...")
10230 for idx, disk in enumerate(old_disks):
10231 meta = disk.children[1]
10232 self.cfg.SetDiskID(meta, pnode)
10233 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
10235 self.LogWarning("Could not remove metadata for disk %d on node %s,"
10236 " continuing anyway: %s", idx, pnode, msg)
10238 def Exec(self, feedback_fn):
10239 """Modifies an instance.
10241 All parameters take effect only at the next restart of the instance.
10244 # Process here the warnings from CheckPrereq, as we don't have a
10245 # feedback_fn there.
10246 for warn in self.warn:
10247 feedback_fn("WARNING: %s" % warn)
10250 instance = self.instance
10252 for disk_op, disk_dict in self.op.disks:
10253 if disk_op == constants.DDM_REMOVE:
10254 # remove the last disk
10255 device = instance.disks.pop()
10256 device_idx = len(instance.disks)
10257 for node, disk in device.ComputeNodeTree(instance.primary_node):
10258 self.cfg.SetDiskID(disk, node)
10259 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
10261 self.LogWarning("Could not remove disk/%d on node %s: %s,"
10262 " continuing anyway", device_idx, node, msg)
10263 result.append(("disk/%d" % device_idx, "remove"))
10264 elif disk_op == constants.DDM_ADD:
10266 if instance.disk_template in (constants.DT_FILE,
10267 constants.DT_SHARED_FILE):
10268 file_driver, file_path = instance.disks[0].logical_id
10269 file_path = os.path.dirname(file_path)
10271 file_driver = file_path = None
10272 disk_idx_base = len(instance.disks)
10273 new_disk = _GenerateDiskTemplate(self,
10274 instance.disk_template,
10275 instance.name, instance.primary_node,
10276 instance.secondary_nodes,
10280 disk_idx_base, feedback_fn)[0]
10281 instance.disks.append(new_disk)
10282 info = _GetInstanceInfoText(instance)
10284 logging.info("Creating volume %s for instance %s",
10285 new_disk.iv_name, instance.name)
10286 # Note: this needs to be kept in sync with _CreateDisks
10288 for node in instance.all_nodes:
10289 f_create = node == instance.primary_node
10291 _CreateBlockDev(self, node, instance, new_disk,
10292 f_create, info, f_create)
10293 except errors.OpExecError, err:
10294 self.LogWarning("Failed to create volume %s (%s) on"
10296 new_disk.iv_name, new_disk, node, err)
10297 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
10298 (new_disk.size, new_disk.mode)))
10300 # change a given disk
10301 instance.disks[disk_op].mode = disk_dict[constants.IDISK_MODE]
10302 result.append(("disk.mode/%d" % disk_op,
10303 disk_dict[constants.IDISK_MODE]))
10305 if self.op.disk_template:
10306 r_shut = _ShutdownInstanceDisks(self, instance)
10308 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
10309 " proceed with disk template conversion")
10310 mode = (instance.disk_template, self.op.disk_template)
10312 self._DISK_CONVERSIONS[mode](self, feedback_fn)
10314 self.cfg.ReleaseDRBDMinors(instance.name)
10316 result.append(("disk_template", self.op.disk_template))
10319 for nic_op, nic_dict in self.op.nics:
10320 if nic_op == constants.DDM_REMOVE:
10321 # remove the last nic
10322 del instance.nics[-1]
10323 result.append(("nic.%d" % len(instance.nics), "remove"))
10324 elif nic_op == constants.DDM_ADD:
10325 # mac and bridge should be set, by now
10326 mac = nic_dict[constants.INIC_MAC]
10327 ip = nic_dict.get(constants.INIC_IP, None)
10328 nicparams = self.nic_pinst[constants.DDM_ADD]
10329 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
10330 instance.nics.append(new_nic)
10331 result.append(("nic.%d" % (len(instance.nics) - 1),
10332 "add:mac=%s,ip=%s,mode=%s,link=%s" %
10333 (new_nic.mac, new_nic.ip,
10334 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
10335 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
10338 for key in (constants.INIC_MAC, constants.INIC_IP):
10339 if key in nic_dict:
10340 setattr(instance.nics[nic_op], key, nic_dict[key])
10341 if nic_op in self.nic_pinst:
10342 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
10343 for key, val in nic_dict.iteritems():
10344 result.append(("nic.%s/%d" % (key, nic_op), val))
10347 if self.op.hvparams:
10348 instance.hvparams = self.hv_inst
10349 for key, val in self.op.hvparams.iteritems():
10350 result.append(("hv/%s" % key, val))
10353 if self.op.beparams:
10354 instance.beparams = self.be_inst
10355 for key, val in self.op.beparams.iteritems():
10356 result.append(("be/%s" % key, val))
10359 if self.op.os_name:
10360 instance.os = self.op.os_name
10363 if self.op.osparams:
10364 instance.osparams = self.os_inst
10365 for key, val in self.op.osparams.iteritems():
10366 result.append(("os/%s" % key, val))
10368 self.cfg.Update(instance, feedback_fn)
10372 _DISK_CONVERSIONS = {
10373 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
10374 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
10378 class LUBackupQuery(NoHooksLU):
10379 """Query the exports list
10384 def ExpandNames(self):
10385 self.needed_locks = {}
10386 self.share_locks[locking.LEVEL_NODE] = 1
10387 if not self.op.nodes:
10388 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10390 self.needed_locks[locking.LEVEL_NODE] = \
10391 _GetWantedNodes(self, self.op.nodes)
10393 def Exec(self, feedback_fn):
10394 """Compute the list of all the exported system images.
10397 @return: a dictionary with the structure node->(export-list)
10398 where export-list is a list of the instances exported on
10402 self.nodes = self.glm.list_owned(locking.LEVEL_NODE)
10403 rpcresult = self.rpc.call_export_list(self.nodes)
10405 for node in rpcresult:
10406 if rpcresult[node].fail_msg:
10407 result[node] = False
10409 result[node] = rpcresult[node].payload
10414 class LUBackupPrepare(NoHooksLU):
10415 """Prepares an instance for an export and returns useful information.
10420 def ExpandNames(self):
10421 self._ExpandAndLockInstance()
10423 def CheckPrereq(self):
10424 """Check prerequisites.
10427 instance_name = self.op.instance_name
10429 self.instance = self.cfg.GetInstanceInfo(instance_name)
10430 assert self.instance is not None, \
10431 "Cannot retrieve locked instance %s" % self.op.instance_name
10432 _CheckNodeOnline(self, self.instance.primary_node)
10434 self._cds = _GetClusterDomainSecret()
10436 def Exec(self, feedback_fn):
10437 """Prepares an instance for an export.
10440 instance = self.instance
10442 if self.op.mode == constants.EXPORT_MODE_REMOTE:
10443 salt = utils.GenerateSecret(8)
10445 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
10446 result = self.rpc.call_x509_cert_create(instance.primary_node,
10447 constants.RIE_CERT_VALIDITY)
10448 result.Raise("Can't create X509 key and certificate on %s" % result.node)
10450 (name, cert_pem) = result.payload
10452 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
10456 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
10457 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
10459 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
10465 class LUBackupExport(LogicalUnit):
10466 """Export an instance to an image in the cluster.
10469 HPATH = "instance-export"
10470 HTYPE = constants.HTYPE_INSTANCE
10473 def CheckArguments(self):
10474 """Check the arguments.
10477 self.x509_key_name = self.op.x509_key_name
10478 self.dest_x509_ca_pem = self.op.destination_x509_ca
10480 if self.op.mode == constants.EXPORT_MODE_REMOTE:
10481 if not self.x509_key_name:
10482 raise errors.OpPrereqError("Missing X509 key name for encryption",
10483 errors.ECODE_INVAL)
10485 if not self.dest_x509_ca_pem:
10486 raise errors.OpPrereqError("Missing destination X509 CA",
10487 errors.ECODE_INVAL)
10489 def ExpandNames(self):
10490 self._ExpandAndLockInstance()
10492 # Lock all nodes for local exports
10493 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10494 # FIXME: lock only instance primary and destination node
10496 # Sad but true, for now we have do lock all nodes, as we don't know where
10497 # the previous export might be, and in this LU we search for it and
10498 # remove it from its current node. In the future we could fix this by:
10499 # - making a tasklet to search (share-lock all), then create the
10500 # new one, then one to remove, after
10501 # - removing the removal operation altogether
10502 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10504 def DeclareLocks(self, level):
10505 """Last minute lock declaration."""
10506 # All nodes are locked anyway, so nothing to do here.
10508 def BuildHooksEnv(self):
10509 """Build hooks env.
10511 This will run on the master, primary node and target node.
10515 "EXPORT_MODE": self.op.mode,
10516 "EXPORT_NODE": self.op.target_node,
10517 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
10518 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
10519 # TODO: Generic function for boolean env variables
10520 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
10523 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
10527 def BuildHooksNodes(self):
10528 """Build hooks nodes.
10531 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
10533 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10534 nl.append(self.op.target_node)
10538 def CheckPrereq(self):
10539 """Check prerequisites.
10541 This checks that the instance and node names are valid.
10544 instance_name = self.op.instance_name
10546 self.instance = self.cfg.GetInstanceInfo(instance_name)
10547 assert self.instance is not None, \
10548 "Cannot retrieve locked instance %s" % self.op.instance_name
10549 _CheckNodeOnline(self, self.instance.primary_node)
10551 if (self.op.remove_instance and self.instance.admin_up and
10552 not self.op.shutdown):
10553 raise errors.OpPrereqError("Can not remove instance without shutting it"
10556 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10557 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
10558 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
10559 assert self.dst_node is not None
10561 _CheckNodeOnline(self, self.dst_node.name)
10562 _CheckNodeNotDrained(self, self.dst_node.name)
10565 self.dest_disk_info = None
10566 self.dest_x509_ca = None
10568 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
10569 self.dst_node = None
10571 if len(self.op.target_node) != len(self.instance.disks):
10572 raise errors.OpPrereqError(("Received destination information for %s"
10573 " disks, but instance %s has %s disks") %
10574 (len(self.op.target_node), instance_name,
10575 len(self.instance.disks)),
10576 errors.ECODE_INVAL)
10578 cds = _GetClusterDomainSecret()
10580 # Check X509 key name
10582 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
10583 except (TypeError, ValueError), err:
10584 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
10586 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
10587 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
10588 errors.ECODE_INVAL)
10590 # Load and verify CA
10592 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
10593 except OpenSSL.crypto.Error, err:
10594 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
10595 (err, ), errors.ECODE_INVAL)
10597 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
10598 if errcode is not None:
10599 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
10600 (msg, ), errors.ECODE_INVAL)
10602 self.dest_x509_ca = cert
10604 # Verify target information
10606 for idx, disk_data in enumerate(self.op.target_node):
10608 (host, port, magic) = \
10609 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
10610 except errors.GenericError, err:
10611 raise errors.OpPrereqError("Target info for disk %s: %s" %
10612 (idx, err), errors.ECODE_INVAL)
10614 disk_info.append((host, port, magic))
10616 assert len(disk_info) == len(self.op.target_node)
10617 self.dest_disk_info = disk_info
10620 raise errors.ProgrammerError("Unhandled export mode %r" %
10623 # instance disk type verification
10624 # TODO: Implement export support for file-based disks
10625 for disk in self.instance.disks:
10626 if disk.dev_type == constants.LD_FILE:
10627 raise errors.OpPrereqError("Export not supported for instances with"
10628 " file-based disks", errors.ECODE_INVAL)
10630 def _CleanupExports(self, feedback_fn):
10631 """Removes exports of current instance from all other nodes.
10633 If an instance in a cluster with nodes A..D was exported to node C, its
10634 exports will be removed from the nodes A, B and D.
10637 assert self.op.mode != constants.EXPORT_MODE_REMOTE
10639 nodelist = self.cfg.GetNodeList()
10640 nodelist.remove(self.dst_node.name)
10642 # on one-node clusters nodelist will be empty after the removal
10643 # if we proceed the backup would be removed because OpBackupQuery
10644 # substitutes an empty list with the full cluster node list.
10645 iname = self.instance.name
10647 feedback_fn("Removing old exports for instance %s" % iname)
10648 exportlist = self.rpc.call_export_list(nodelist)
10649 for node in exportlist:
10650 if exportlist[node].fail_msg:
10652 if iname in exportlist[node].payload:
10653 msg = self.rpc.call_export_remove(node, iname).fail_msg
10655 self.LogWarning("Could not remove older export for instance %s"
10656 " on node %s: %s", iname, node, msg)
10658 def Exec(self, feedback_fn):
10659 """Export an instance to an image in the cluster.
10662 assert self.op.mode in constants.EXPORT_MODES
10664 instance = self.instance
10665 src_node = instance.primary_node
10667 if self.op.shutdown:
10668 # shutdown the instance, but not the disks
10669 feedback_fn("Shutting down instance %s" % instance.name)
10670 result = self.rpc.call_instance_shutdown(src_node, instance,
10671 self.op.shutdown_timeout)
10672 # TODO: Maybe ignore failures if ignore_remove_failures is set
10673 result.Raise("Could not shutdown instance %s on"
10674 " node %s" % (instance.name, src_node))
10676 # set the disks ID correctly since call_instance_start needs the
10677 # correct drbd minor to create the symlinks
10678 for disk in instance.disks:
10679 self.cfg.SetDiskID(disk, src_node)
10681 activate_disks = (not instance.admin_up)
10684 # Activate the instance disks if we'exporting a stopped instance
10685 feedback_fn("Activating disks for %s" % instance.name)
10686 _StartInstanceDisks(self, instance, None)
10689 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
10692 helper.CreateSnapshots()
10694 if (self.op.shutdown and instance.admin_up and
10695 not self.op.remove_instance):
10696 assert not activate_disks
10697 feedback_fn("Starting instance %s" % instance.name)
10698 result = self.rpc.call_instance_start(src_node, instance, None, None)
10699 msg = result.fail_msg
10701 feedback_fn("Failed to start instance: %s" % msg)
10702 _ShutdownInstanceDisks(self, instance)
10703 raise errors.OpExecError("Could not start instance: %s" % msg)
10705 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10706 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
10707 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
10708 connect_timeout = constants.RIE_CONNECT_TIMEOUT
10709 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
10711 (key_name, _, _) = self.x509_key_name
10714 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
10717 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
10718 key_name, dest_ca_pem,
10723 # Check for backwards compatibility
10724 assert len(dresults) == len(instance.disks)
10725 assert compat.all(isinstance(i, bool) for i in dresults), \
10726 "Not all results are boolean: %r" % dresults
10730 feedback_fn("Deactivating disks for %s" % instance.name)
10731 _ShutdownInstanceDisks(self, instance)
10733 if not (compat.all(dresults) and fin_resu):
10736 failures.append("export finalization")
10737 if not compat.all(dresults):
10738 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
10740 failures.append("disk export: disk(s) %s" % fdsk)
10742 raise errors.OpExecError("Export failed, errors in %s" %
10743 utils.CommaJoin(failures))
10745 # At this point, the export was successful, we can cleanup/finish
10747 # Remove instance if requested
10748 if self.op.remove_instance:
10749 feedback_fn("Removing instance %s" % instance.name)
10750 _RemoveInstance(self, feedback_fn, instance,
10751 self.op.ignore_remove_failures)
10753 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10754 self._CleanupExports(feedback_fn)
10756 return fin_resu, dresults
10759 class LUBackupRemove(NoHooksLU):
10760 """Remove exports related to the named instance.
10765 def ExpandNames(self):
10766 self.needed_locks = {}
10767 # We need all nodes to be locked in order for RemoveExport to work, but we
10768 # don't need to lock the instance itself, as nothing will happen to it (and
10769 # we can remove exports also for a removed instance)
10770 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10772 def Exec(self, feedback_fn):
10773 """Remove any export.
10776 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
10777 # If the instance was not found we'll try with the name that was passed in.
10778 # This will only work if it was an FQDN, though.
10780 if not instance_name:
10782 instance_name = self.op.instance_name
10784 locked_nodes = self.glm.list_owned(locking.LEVEL_NODE)
10785 exportlist = self.rpc.call_export_list(locked_nodes)
10787 for node in exportlist:
10788 msg = exportlist[node].fail_msg
10790 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
10792 if instance_name in exportlist[node].payload:
10794 result = self.rpc.call_export_remove(node, instance_name)
10795 msg = result.fail_msg
10797 logging.error("Could not remove export for instance %s"
10798 " on node %s: %s", instance_name, node, msg)
10800 if fqdn_warn and not found:
10801 feedback_fn("Export not found. If trying to remove an export belonging"
10802 " to a deleted instance please use its Fully Qualified"
10806 class LUGroupAdd(LogicalUnit):
10807 """Logical unit for creating node groups.
10810 HPATH = "group-add"
10811 HTYPE = constants.HTYPE_GROUP
10814 def ExpandNames(self):
10815 # We need the new group's UUID here so that we can create and acquire the
10816 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
10817 # that it should not check whether the UUID exists in the configuration.
10818 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
10819 self.needed_locks = {}
10820 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10822 def CheckPrereq(self):
10823 """Check prerequisites.
10825 This checks that the given group name is not an existing node group
10830 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10831 except errors.OpPrereqError:
10834 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
10835 " node group (UUID: %s)" %
10836 (self.op.group_name, existing_uuid),
10837 errors.ECODE_EXISTS)
10839 if self.op.ndparams:
10840 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10842 def BuildHooksEnv(self):
10843 """Build hooks env.
10847 "GROUP_NAME": self.op.group_name,
10850 def BuildHooksNodes(self):
10851 """Build hooks nodes.
10854 mn = self.cfg.GetMasterNode()
10855 return ([mn], [mn])
10857 def Exec(self, feedback_fn):
10858 """Add the node group to the cluster.
10861 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
10862 uuid=self.group_uuid,
10863 alloc_policy=self.op.alloc_policy,
10864 ndparams=self.op.ndparams)
10866 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
10867 del self.remove_locks[locking.LEVEL_NODEGROUP]
10870 class LUGroupAssignNodes(NoHooksLU):
10871 """Logical unit for assigning nodes to groups.
10876 def ExpandNames(self):
10877 # These raise errors.OpPrereqError on their own:
10878 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10879 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
10881 # We want to lock all the affected nodes and groups. We have readily
10882 # available the list of nodes, and the *destination* group. To gather the
10883 # list of "source" groups, we need to fetch node information.
10884 self.node_data = self.cfg.GetAllNodesInfo()
10885 affected_groups = set(self.node_data[node].group for node in self.op.nodes)
10886 affected_groups.add(self.group_uuid)
10888 self.needed_locks = {
10889 locking.LEVEL_NODEGROUP: list(affected_groups),
10890 locking.LEVEL_NODE: self.op.nodes,
10893 def CheckPrereq(self):
10894 """Check prerequisites.
10897 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10898 instance_data = self.cfg.GetAllInstancesInfo()
10900 if self.group is None:
10901 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10902 (self.op.group_name, self.group_uuid))
10904 (new_splits, previous_splits) = \
10905 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
10906 for node in self.op.nodes],
10907 self.node_data, instance_data)
10910 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
10912 if not self.op.force:
10913 raise errors.OpExecError("The following instances get split by this"
10914 " change and --force was not given: %s" %
10917 self.LogWarning("This operation will split the following instances: %s",
10920 if previous_splits:
10921 self.LogWarning("In addition, these already-split instances continue"
10922 " to be split across groups: %s",
10923 utils.CommaJoin(utils.NiceSort(previous_splits)))
10925 def Exec(self, feedback_fn):
10926 """Assign nodes to a new group.
10929 for node in self.op.nodes:
10930 self.node_data[node].group = self.group_uuid
10932 self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
10935 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
10936 """Check for split instances after a node assignment.
10938 This method considers a series of node assignments as an atomic operation,
10939 and returns information about split instances after applying the set of
10942 In particular, it returns information about newly split instances, and
10943 instances that were already split, and remain so after the change.
10945 Only instances whose disk template is listed in constants.DTS_INT_MIRROR are
10948 @type changes: list of (node_name, new_group_uuid) pairs.
10949 @param changes: list of node assignments to consider.
10950 @param node_data: a dict with data for all nodes
10951 @param instance_data: a dict with all instances to consider
10952 @rtype: a two-tuple
10953 @return: a list of instances that were previously okay and result split as a
10954 consequence of this change, and a list of instances that were previously
10955 split and this change does not fix.
10958 changed_nodes = dict((node, group) for node, group in changes
10959 if node_data[node].group != group)
10961 all_split_instances = set()
10962 previously_split_instances = set()
10964 def InstanceNodes(instance):
10965 return [instance.primary_node] + list(instance.secondary_nodes)
10967 for inst in instance_data.values():
10968 if inst.disk_template not in constants.DTS_INT_MIRROR:
10971 instance_nodes = InstanceNodes(inst)
10973 if len(set(node_data[node].group for node in instance_nodes)) > 1:
10974 previously_split_instances.add(inst.name)
10976 if len(set(changed_nodes.get(node, node_data[node].group)
10977 for node in instance_nodes)) > 1:
10978 all_split_instances.add(inst.name)
10980 return (list(all_split_instances - previously_split_instances),
10981 list(previously_split_instances & all_split_instances))
10984 class _GroupQuery(_QueryBase):
10985 FIELDS = query.GROUP_FIELDS
10987 def ExpandNames(self, lu):
10988 lu.needed_locks = {}
10990 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
10991 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
10994 self.wanted = [name_to_uuid[name]
10995 for name in utils.NiceSort(name_to_uuid.keys())]
10997 # Accept names to be either names or UUIDs.
11000 all_uuid = frozenset(self._all_groups.keys())
11002 for name in self.names:
11003 if name in all_uuid:
11004 self.wanted.append(name)
11005 elif name in name_to_uuid:
11006 self.wanted.append(name_to_uuid[name])
11008 missing.append(name)
11011 raise errors.OpPrereqError("Some groups do not exist: %s" %
11012 utils.CommaJoin(missing),
11013 errors.ECODE_NOENT)
11015 def DeclareLocks(self, lu, level):
11018 def _GetQueryData(self, lu):
11019 """Computes the list of node groups and their attributes.
11022 do_nodes = query.GQ_NODE in self.requested_data
11023 do_instances = query.GQ_INST in self.requested_data
11025 group_to_nodes = None
11026 group_to_instances = None
11028 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
11029 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
11030 # latter GetAllInstancesInfo() is not enough, for we have to go through
11031 # instance->node. Hence, we will need to process nodes even if we only need
11032 # instance information.
11033 if do_nodes or do_instances:
11034 all_nodes = lu.cfg.GetAllNodesInfo()
11035 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
11038 for node in all_nodes.values():
11039 if node.group in group_to_nodes:
11040 group_to_nodes[node.group].append(node.name)
11041 node_to_group[node.name] = node.group
11044 all_instances = lu.cfg.GetAllInstancesInfo()
11045 group_to_instances = dict((uuid, []) for uuid in self.wanted)
11047 for instance in all_instances.values():
11048 node = instance.primary_node
11049 if node in node_to_group:
11050 group_to_instances[node_to_group[node]].append(instance.name)
11053 # Do not pass on node information if it was not requested.
11054 group_to_nodes = None
11056 return query.GroupQueryData([self._all_groups[uuid]
11057 for uuid in self.wanted],
11058 group_to_nodes, group_to_instances)
11061 class LUGroupQuery(NoHooksLU):
11062 """Logical unit for querying node groups.
11067 def CheckArguments(self):
11068 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
11069 self.op.output_fields, False)
11071 def ExpandNames(self):
11072 self.gq.ExpandNames(self)
11074 def Exec(self, feedback_fn):
11075 return self.gq.OldStyleQuery(self)
11078 class LUGroupSetParams(LogicalUnit):
11079 """Modifies the parameters of a node group.
11082 HPATH = "group-modify"
11083 HTYPE = constants.HTYPE_GROUP
11086 def CheckArguments(self):
11089 self.op.alloc_policy,
11092 if all_changes.count(None) == len(all_changes):
11093 raise errors.OpPrereqError("Please pass at least one modification",
11094 errors.ECODE_INVAL)
11096 def ExpandNames(self):
11097 # This raises errors.OpPrereqError on its own:
11098 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11100 self.needed_locks = {
11101 locking.LEVEL_NODEGROUP: [self.group_uuid],
11104 def CheckPrereq(self):
11105 """Check prerequisites.
11108 self.group = self.cfg.GetNodeGroup(self.group_uuid)
11110 if self.group is None:
11111 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
11112 (self.op.group_name, self.group_uuid))
11114 if self.op.ndparams:
11115 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
11116 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
11117 self.new_ndparams = new_ndparams
11119 def BuildHooksEnv(self):
11120 """Build hooks env.
11124 "GROUP_NAME": self.op.group_name,
11125 "NEW_ALLOC_POLICY": self.op.alloc_policy,
11128 def BuildHooksNodes(self):
11129 """Build hooks nodes.
11132 mn = self.cfg.GetMasterNode()
11133 return ([mn], [mn])
11135 def Exec(self, feedback_fn):
11136 """Modifies the node group.
11141 if self.op.ndparams:
11142 self.group.ndparams = self.new_ndparams
11143 result.append(("ndparams", str(self.group.ndparams)))
11145 if self.op.alloc_policy:
11146 self.group.alloc_policy = self.op.alloc_policy
11148 self.cfg.Update(self.group, feedback_fn)
11153 class LUGroupRemove(LogicalUnit):
11154 HPATH = "group-remove"
11155 HTYPE = constants.HTYPE_GROUP
11158 def ExpandNames(self):
11159 # This will raises errors.OpPrereqError on its own:
11160 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11161 self.needed_locks = {
11162 locking.LEVEL_NODEGROUP: [self.group_uuid],
11165 def CheckPrereq(self):
11166 """Check prerequisites.
11168 This checks that the given group name exists as a node group, that is
11169 empty (i.e., contains no nodes), and that is not the last group of the
11173 # Verify that the group is empty.
11174 group_nodes = [node.name
11175 for node in self.cfg.GetAllNodesInfo().values()
11176 if node.group == self.group_uuid]
11179 raise errors.OpPrereqError("Group '%s' not empty, has the following"
11181 (self.op.group_name,
11182 utils.CommaJoin(utils.NiceSort(group_nodes))),
11183 errors.ECODE_STATE)
11185 # Verify the cluster would not be left group-less.
11186 if len(self.cfg.GetNodeGroupList()) == 1:
11187 raise errors.OpPrereqError("Group '%s' is the only group,"
11188 " cannot be removed" %
11189 self.op.group_name,
11190 errors.ECODE_STATE)
11192 def BuildHooksEnv(self):
11193 """Build hooks env.
11197 "GROUP_NAME": self.op.group_name,
11200 def BuildHooksNodes(self):
11201 """Build hooks nodes.
11204 mn = self.cfg.GetMasterNode()
11205 return ([mn], [mn])
11207 def Exec(self, feedback_fn):
11208 """Remove the node group.
11212 self.cfg.RemoveNodeGroup(self.group_uuid)
11213 except errors.ConfigurationError:
11214 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
11215 (self.op.group_name, self.group_uuid))
11217 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
11220 class LUGroupRename(LogicalUnit):
11221 HPATH = "group-rename"
11222 HTYPE = constants.HTYPE_GROUP
11225 def ExpandNames(self):
11226 # This raises errors.OpPrereqError on its own:
11227 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
11229 self.needed_locks = {
11230 locking.LEVEL_NODEGROUP: [self.group_uuid],
11233 def CheckPrereq(self):
11234 """Check prerequisites.
11236 Ensures requested new name is not yet used.
11240 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
11241 except errors.OpPrereqError:
11244 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
11245 " node group (UUID: %s)" %
11246 (self.op.new_name, new_name_uuid),
11247 errors.ECODE_EXISTS)
11249 def BuildHooksEnv(self):
11250 """Build hooks env.
11254 "OLD_NAME": self.op.group_name,
11255 "NEW_NAME": self.op.new_name,
11258 def BuildHooksNodes(self):
11259 """Build hooks nodes.
11262 mn = self.cfg.GetMasterNode()
11264 all_nodes = self.cfg.GetAllNodesInfo()
11265 all_nodes.pop(mn, None)
11268 run_nodes.extend(node.name for node in all_nodes.values()
11269 if node.group == self.group_uuid)
11271 return (run_nodes, run_nodes)
11273 def Exec(self, feedback_fn):
11274 """Rename the node group.
11277 group = self.cfg.GetNodeGroup(self.group_uuid)
11280 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
11281 (self.op.group_name, self.group_uuid))
11283 group.name = self.op.new_name
11284 self.cfg.Update(group, feedback_fn)
11286 return self.op.new_name
11289 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
11290 """Generic tags LU.
11292 This is an abstract class which is the parent of all the other tags LUs.
11295 def ExpandNames(self):
11296 self.group_uuid = None
11297 self.needed_locks = {}
11298 if self.op.kind == constants.TAG_NODE:
11299 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
11300 self.needed_locks[locking.LEVEL_NODE] = self.op.name
11301 elif self.op.kind == constants.TAG_INSTANCE:
11302 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
11303 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
11304 elif self.op.kind == constants.TAG_NODEGROUP:
11305 self.group_uuid = self.cfg.LookupNodeGroup(self.op.name)
11307 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
11308 # not possible to acquire the BGL based on opcode parameters)
11310 def CheckPrereq(self):
11311 """Check prerequisites.
11314 if self.op.kind == constants.TAG_CLUSTER:
11315 self.target = self.cfg.GetClusterInfo()
11316 elif self.op.kind == constants.TAG_NODE:
11317 self.target = self.cfg.GetNodeInfo(self.op.name)
11318 elif self.op.kind == constants.TAG_INSTANCE:
11319 self.target = self.cfg.GetInstanceInfo(self.op.name)
11320 elif self.op.kind == constants.TAG_NODEGROUP:
11321 self.target = self.cfg.GetNodeGroup(self.group_uuid)
11323 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
11324 str(self.op.kind), errors.ECODE_INVAL)
11327 class LUTagsGet(TagsLU):
11328 """Returns the tags of a given object.
11333 def ExpandNames(self):
11334 TagsLU.ExpandNames(self)
11336 # Share locks as this is only a read operation
11337 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
11339 def Exec(self, feedback_fn):
11340 """Returns the tag list.
11343 return list(self.target.GetTags())
11346 class LUTagsSearch(NoHooksLU):
11347 """Searches the tags for a given pattern.
11352 def ExpandNames(self):
11353 self.needed_locks = {}
11355 def CheckPrereq(self):
11356 """Check prerequisites.
11358 This checks the pattern passed for validity by compiling it.
11362 self.re = re.compile(self.op.pattern)
11363 except re.error, err:
11364 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
11365 (self.op.pattern, err), errors.ECODE_INVAL)
11367 def Exec(self, feedback_fn):
11368 """Returns the tag list.
11372 tgts = [("/cluster", cfg.GetClusterInfo())]
11373 ilist = cfg.GetAllInstancesInfo().values()
11374 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
11375 nlist = cfg.GetAllNodesInfo().values()
11376 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
11377 tgts.extend(("/nodegroup/%s" % n.name, n)
11378 for n in cfg.GetAllNodeGroupsInfo().values())
11380 for path, target in tgts:
11381 for tag in target.GetTags():
11382 if self.re.search(tag):
11383 results.append((path, tag))
11387 class LUTagsSet(TagsLU):
11388 """Sets a tag on a given object.
11393 def CheckPrereq(self):
11394 """Check prerequisites.
11396 This checks the type and length of the tag name and value.
11399 TagsLU.CheckPrereq(self)
11400 for tag in self.op.tags:
11401 objects.TaggableObject.ValidateTag(tag)
11403 def Exec(self, feedback_fn):
11408 for tag in self.op.tags:
11409 self.target.AddTag(tag)
11410 except errors.TagError, err:
11411 raise errors.OpExecError("Error while setting tag: %s" % str(err))
11412 self.cfg.Update(self.target, feedback_fn)
11415 class LUTagsDel(TagsLU):
11416 """Delete a list of tags from a given object.
11421 def CheckPrereq(self):
11422 """Check prerequisites.
11424 This checks that we have the given tag.
11427 TagsLU.CheckPrereq(self)
11428 for tag in self.op.tags:
11429 objects.TaggableObject.ValidateTag(tag)
11430 del_tags = frozenset(self.op.tags)
11431 cur_tags = self.target.GetTags()
11433 diff_tags = del_tags - cur_tags
11435 diff_names = ("'%s'" % i for i in sorted(diff_tags))
11436 raise errors.OpPrereqError("Tag(s) %s not found" %
11437 (utils.CommaJoin(diff_names), ),
11438 errors.ECODE_NOENT)
11440 def Exec(self, feedback_fn):
11441 """Remove the tag from the object.
11444 for tag in self.op.tags:
11445 self.target.RemoveTag(tag)
11446 self.cfg.Update(self.target, feedback_fn)
11449 class LUTestDelay(NoHooksLU):
11450 """Sleep for a specified amount of time.
11452 This LU sleeps on the master and/or nodes for a specified amount of
11458 def ExpandNames(self):
11459 """Expand names and set required locks.
11461 This expands the node list, if any.
11464 self.needed_locks = {}
11465 if self.op.on_nodes:
11466 # _GetWantedNodes can be used here, but is not always appropriate to use
11467 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
11468 # more information.
11469 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
11470 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
11472 def _TestDelay(self):
11473 """Do the actual sleep.
11476 if self.op.on_master:
11477 if not utils.TestDelay(self.op.duration):
11478 raise errors.OpExecError("Error during master delay test")
11479 if self.op.on_nodes:
11480 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
11481 for node, node_result in result.items():
11482 node_result.Raise("Failure during rpc call to node %s" % node)
11484 def Exec(self, feedback_fn):
11485 """Execute the test delay opcode, with the wanted repetitions.
11488 if self.op.repeat == 0:
11491 top_value = self.op.repeat - 1
11492 for i in range(self.op.repeat):
11493 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
11497 class LUTestJqueue(NoHooksLU):
11498 """Utility LU to test some aspects of the job queue.
11503 # Must be lower than default timeout for WaitForJobChange to see whether it
11504 # notices changed jobs
11505 _CLIENT_CONNECT_TIMEOUT = 20.0
11506 _CLIENT_CONFIRM_TIMEOUT = 60.0
11509 def _NotifyUsingSocket(cls, cb, errcls):
11510 """Opens a Unix socket and waits for another program to connect.
11513 @param cb: Callback to send socket name to client
11514 @type errcls: class
11515 @param errcls: Exception class to use for errors
11518 # Using a temporary directory as there's no easy way to create temporary
11519 # sockets without writing a custom loop around tempfile.mktemp and
11521 tmpdir = tempfile.mkdtemp()
11523 tmpsock = utils.PathJoin(tmpdir, "sock")
11525 logging.debug("Creating temporary socket at %s", tmpsock)
11526 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
11531 # Send details to client
11534 # Wait for client to connect before continuing
11535 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
11537 (conn, _) = sock.accept()
11538 except socket.error, err:
11539 raise errcls("Client didn't connect in time (%s)" % err)
11543 # Remove as soon as client is connected
11544 shutil.rmtree(tmpdir)
11546 # Wait for client to close
11549 # pylint: disable-msg=E1101
11550 # Instance of '_socketobject' has no ... member
11551 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
11553 except socket.error, err:
11554 raise errcls("Client failed to confirm notification (%s)" % err)
11558 def _SendNotification(self, test, arg, sockname):
11559 """Sends a notification to the client.
11562 @param test: Test name
11563 @param arg: Test argument (depends on test)
11564 @type sockname: string
11565 @param sockname: Socket path
11568 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
11570 def _Notify(self, prereq, test, arg):
11571 """Notifies the client of a test.
11574 @param prereq: Whether this is a prereq-phase test
11576 @param test: Test name
11577 @param arg: Test argument (depends on test)
11581 errcls = errors.OpPrereqError
11583 errcls = errors.OpExecError
11585 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
11589 def CheckArguments(self):
11590 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
11591 self.expandnames_calls = 0
11593 def ExpandNames(self):
11594 checkargs_calls = getattr(self, "checkargs_calls", 0)
11595 if checkargs_calls < 1:
11596 raise errors.ProgrammerError("CheckArguments was not called")
11598 self.expandnames_calls += 1
11600 if self.op.notify_waitlock:
11601 self._Notify(True, constants.JQT_EXPANDNAMES, None)
11603 self.LogInfo("Expanding names")
11605 # Get lock on master node (just to get a lock, not for a particular reason)
11606 self.needed_locks = {
11607 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
11610 def Exec(self, feedback_fn):
11611 if self.expandnames_calls < 1:
11612 raise errors.ProgrammerError("ExpandNames was not called")
11614 if self.op.notify_exec:
11615 self._Notify(False, constants.JQT_EXEC, None)
11617 self.LogInfo("Executing")
11619 if self.op.log_messages:
11620 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
11621 for idx, msg in enumerate(self.op.log_messages):
11622 self.LogInfo("Sending log message %s", idx + 1)
11623 feedback_fn(constants.JQT_MSGPREFIX + msg)
11624 # Report how many test messages have been sent
11625 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
11628 raise errors.OpExecError("Opcode failure was requested")
11633 class IAllocator(object):
11634 """IAllocator framework.
11636 An IAllocator instance has three sets of attributes:
11637 - cfg that is needed to query the cluster
11638 - input data (all members of the _KEYS class attribute are required)
11639 - four buffer attributes (in|out_data|text), that represent the
11640 input (to the external script) in text and data structure format,
11641 and the output from it, again in two formats
11642 - the result variables from the script (success, info, nodes) for
11646 # pylint: disable-msg=R0902
11647 # lots of instance attributes
11649 "name", "mem_size", "disks", "disk_template",
11650 "os", "tags", "nics", "vcpus", "hypervisor",
11653 "name", "relocate_from",
11659 def __init__(self, cfg, rpc, mode, **kwargs):
11662 # init buffer variables
11663 self.in_text = self.out_text = self.in_data = self.out_data = None
11664 # init all input fields so that pylint is happy
11666 self.mem_size = self.disks = self.disk_template = None
11667 self.os = self.tags = self.nics = self.vcpus = None
11668 self.hypervisor = None
11669 self.relocate_from = None
11671 self.evac_nodes = None
11673 self.required_nodes = None
11674 # init result fields
11675 self.success = self.info = self.result = None
11676 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
11677 keyset = self._ALLO_KEYS
11678 fn = self._AddNewInstance
11679 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
11680 keyset = self._RELO_KEYS
11681 fn = self._AddRelocateInstance
11682 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
11683 keyset = self._EVAC_KEYS
11684 fn = self._AddEvacuateNodes
11686 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
11687 " IAllocator" % self.mode)
11689 if key not in keyset:
11690 raise errors.ProgrammerError("Invalid input parameter '%s' to"
11691 " IAllocator" % key)
11692 setattr(self, key, kwargs[key])
11695 if key not in kwargs:
11696 raise errors.ProgrammerError("Missing input parameter '%s' to"
11697 " IAllocator" % key)
11698 self._BuildInputData(fn)
11700 def _ComputeClusterData(self):
11701 """Compute the generic allocator input data.
11703 This is the data that is independent of the actual operation.
11707 cluster_info = cfg.GetClusterInfo()
11710 "version": constants.IALLOCATOR_VERSION,
11711 "cluster_name": cfg.GetClusterName(),
11712 "cluster_tags": list(cluster_info.GetTags()),
11713 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
11714 # we don't have job IDs
11716 ninfo = cfg.GetAllNodesInfo()
11717 iinfo = cfg.GetAllInstancesInfo().values()
11718 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
11721 node_list = [n.name for n in ninfo.values() if n.vm_capable]
11723 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
11724 hypervisor_name = self.hypervisor
11725 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
11726 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
11727 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
11728 hypervisor_name = cluster_info.enabled_hypervisors[0]
11730 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
11733 self.rpc.call_all_instances_info(node_list,
11734 cluster_info.enabled_hypervisors)
11736 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
11738 config_ndata = self._ComputeBasicNodeData(ninfo)
11739 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
11740 i_list, config_ndata)
11741 assert len(data["nodes"]) == len(ninfo), \
11742 "Incomplete node data computed"
11744 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
11746 self.in_data = data
11749 def _ComputeNodeGroupData(cfg):
11750 """Compute node groups data.
11754 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
11756 "name": gdata.name,
11757 "alloc_policy": gdata.alloc_policy,
11762 def _ComputeBasicNodeData(node_cfg):
11763 """Compute global node data.
11766 @returns: a dict of name: (node dict, node config)
11770 for ninfo in node_cfg.values():
11771 # fill in static (config-based) values
11773 "tags": list(ninfo.GetTags()),
11774 "primary_ip": ninfo.primary_ip,
11775 "secondary_ip": ninfo.secondary_ip,
11776 "offline": ninfo.offline,
11777 "drained": ninfo.drained,
11778 "master_candidate": ninfo.master_candidate,
11779 "group": ninfo.group,
11780 "master_capable": ninfo.master_capable,
11781 "vm_capable": ninfo.vm_capable,
11784 node_results[ninfo.name] = pnr
11786 return node_results
11789 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
11791 """Compute global node data.
11793 @param node_results: the basic node structures as filled from the config
11796 # make a copy of the current dict
11797 node_results = dict(node_results)
11798 for nname, nresult in node_data.items():
11799 assert nname in node_results, "Missing basic data for node %s" % nname
11800 ninfo = node_cfg[nname]
11802 if not (ninfo.offline or ninfo.drained):
11803 nresult.Raise("Can't get data for node %s" % nname)
11804 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
11806 remote_info = nresult.payload
11808 for attr in ['memory_total', 'memory_free', 'memory_dom0',
11809 'vg_size', 'vg_free', 'cpu_total']:
11810 if attr not in remote_info:
11811 raise errors.OpExecError("Node '%s' didn't return attribute"
11812 " '%s'" % (nname, attr))
11813 if not isinstance(remote_info[attr], int):
11814 raise errors.OpExecError("Node '%s' returned invalid value"
11816 (nname, attr, remote_info[attr]))
11817 # compute memory used by primary instances
11818 i_p_mem = i_p_up_mem = 0
11819 for iinfo, beinfo in i_list:
11820 if iinfo.primary_node == nname:
11821 i_p_mem += beinfo[constants.BE_MEMORY]
11822 if iinfo.name not in node_iinfo[nname].payload:
11825 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
11826 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
11827 remote_info['memory_free'] -= max(0, i_mem_diff)
11830 i_p_up_mem += beinfo[constants.BE_MEMORY]
11832 # compute memory used by instances
11834 "total_memory": remote_info['memory_total'],
11835 "reserved_memory": remote_info['memory_dom0'],
11836 "free_memory": remote_info['memory_free'],
11837 "total_disk": remote_info['vg_size'],
11838 "free_disk": remote_info['vg_free'],
11839 "total_cpus": remote_info['cpu_total'],
11840 "i_pri_memory": i_p_mem,
11841 "i_pri_up_memory": i_p_up_mem,
11843 pnr_dyn.update(node_results[nname])
11844 node_results[nname] = pnr_dyn
11846 return node_results
11849 def _ComputeInstanceData(cluster_info, i_list):
11850 """Compute global instance data.
11854 for iinfo, beinfo in i_list:
11856 for nic in iinfo.nics:
11857 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
11858 nic_dict = {"mac": nic.mac,
11860 "mode": filled_params[constants.NIC_MODE],
11861 "link": filled_params[constants.NIC_LINK],
11863 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
11864 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
11865 nic_data.append(nic_dict)
11867 "tags": list(iinfo.GetTags()),
11868 "admin_up": iinfo.admin_up,
11869 "vcpus": beinfo[constants.BE_VCPUS],
11870 "memory": beinfo[constants.BE_MEMORY],
11872 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
11874 "disks": [{constants.IDISK_SIZE: dsk.size,
11875 constants.IDISK_MODE: dsk.mode}
11876 for dsk in iinfo.disks],
11877 "disk_template": iinfo.disk_template,
11878 "hypervisor": iinfo.hypervisor,
11880 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
11882 instance_data[iinfo.name] = pir
11884 return instance_data
11886 def _AddNewInstance(self):
11887 """Add new instance data to allocator structure.
11889 This in combination with _AllocatorGetClusterData will create the
11890 correct structure needed as input for the allocator.
11892 The checks for the completeness of the opcode must have already been
11896 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
11898 if self.disk_template in constants.DTS_INT_MIRROR:
11899 self.required_nodes = 2
11901 self.required_nodes = 1
11904 "disk_template": self.disk_template,
11907 "vcpus": self.vcpus,
11908 "memory": self.mem_size,
11909 "disks": self.disks,
11910 "disk_space_total": disk_space,
11912 "required_nodes": self.required_nodes,
11916 def _AddRelocateInstance(self):
11917 """Add relocate instance data to allocator structure.
11919 This in combination with _IAllocatorGetClusterData will create the
11920 correct structure needed as input for the allocator.
11922 The checks for the completeness of the opcode must have already been
11926 instance = self.cfg.GetInstanceInfo(self.name)
11927 if instance is None:
11928 raise errors.ProgrammerError("Unknown instance '%s' passed to"
11929 " IAllocator" % self.name)
11931 if instance.disk_template not in constants.DTS_MIRRORED:
11932 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
11933 errors.ECODE_INVAL)
11935 if instance.disk_template in constants.DTS_INT_MIRROR and \
11936 len(instance.secondary_nodes) != 1:
11937 raise errors.OpPrereqError("Instance has not exactly one secondary node",
11938 errors.ECODE_STATE)
11940 self.required_nodes = 1
11941 disk_sizes = [{constants.IDISK_SIZE: disk.size} for disk in instance.disks]
11942 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
11946 "disk_space_total": disk_space,
11947 "required_nodes": self.required_nodes,
11948 "relocate_from": self.relocate_from,
11952 def _AddEvacuateNodes(self):
11953 """Add evacuate nodes data to allocator structure.
11957 "evac_nodes": self.evac_nodes
11961 def _BuildInputData(self, fn):
11962 """Build input data structures.
11965 self._ComputeClusterData()
11968 request["type"] = self.mode
11969 self.in_data["request"] = request
11971 self.in_text = serializer.Dump(self.in_data)
11973 def Run(self, name, validate=True, call_fn=None):
11974 """Run an instance allocator and return the results.
11977 if call_fn is None:
11978 call_fn = self.rpc.call_iallocator_runner
11980 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
11981 result.Raise("Failure while running the iallocator script")
11983 self.out_text = result.payload
11985 self._ValidateResult()
11987 def _ValidateResult(self):
11988 """Process the allocator results.
11990 This will process and if successful save the result in
11991 self.out_data and the other parameters.
11995 rdict = serializer.Load(self.out_text)
11996 except Exception, err:
11997 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
11999 if not isinstance(rdict, dict):
12000 raise errors.OpExecError("Can't parse iallocator results: not a dict")
12002 # TODO: remove backwards compatiblity in later versions
12003 if "nodes" in rdict and "result" not in rdict:
12004 rdict["result"] = rdict["nodes"]
12007 for key in "success", "info", "result":
12008 if key not in rdict:
12009 raise errors.OpExecError("Can't parse iallocator results:"
12010 " missing key '%s'" % key)
12011 setattr(self, key, rdict[key])
12013 if not isinstance(rdict["result"], list):
12014 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
12017 if self.mode == constants.IALLOCATOR_MODE_RELOC:
12018 assert self.relocate_from is not None
12019 assert self.required_nodes == 1
12021 node2group = dict((name, ndata["group"])
12022 for (name, ndata) in self.in_data["nodes"].items())
12024 fn = compat.partial(self._NodesToGroups, node2group,
12025 self.in_data["nodegroups"])
12027 request_groups = fn(self.relocate_from)
12028 result_groups = fn(rdict["result"])
12030 if result_groups != request_groups:
12031 raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
12032 " differ from original groups (%s)" %
12033 (utils.CommaJoin(result_groups),
12034 utils.CommaJoin(request_groups)))
12036 self.out_data = rdict
12039 def _NodesToGroups(node2group, groups, nodes):
12040 """Returns a list of unique group names for a list of nodes.
12042 @type node2group: dict
12043 @param node2group: Map from node name to group UUID
12045 @param groups: Group information
12047 @param nodes: Node names
12054 group_uuid = node2group[node]
12056 # Ignore unknown node
12060 group = groups[group_uuid]
12062 # Can't find group, let's use UUID
12063 group_name = group_uuid
12065 group_name = group["name"]
12067 result.add(group_name)
12069 return sorted(result)
12072 class LUTestAllocator(NoHooksLU):
12073 """Run allocator tests.
12075 This LU runs the allocator tests
12078 def CheckPrereq(self):
12079 """Check prerequisites.
12081 This checks the opcode parameters depending on the director and mode test.
12084 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
12085 for attr in ["mem_size", "disks", "disk_template",
12086 "os", "tags", "nics", "vcpus"]:
12087 if not hasattr(self.op, attr):
12088 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
12089 attr, errors.ECODE_INVAL)
12090 iname = self.cfg.ExpandInstanceName(self.op.name)
12091 if iname is not None:
12092 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
12093 iname, errors.ECODE_EXISTS)
12094 if not isinstance(self.op.nics, list):
12095 raise errors.OpPrereqError("Invalid parameter 'nics'",
12096 errors.ECODE_INVAL)
12097 if not isinstance(self.op.disks, list):
12098 raise errors.OpPrereqError("Invalid parameter 'disks'",
12099 errors.ECODE_INVAL)
12100 for row in self.op.disks:
12101 if (not isinstance(row, dict) or
12102 "size" not in row or
12103 not isinstance(row["size"], int) or
12104 "mode" not in row or
12105 row["mode"] not in ['r', 'w']):
12106 raise errors.OpPrereqError("Invalid contents of the 'disks'"
12107 " parameter", errors.ECODE_INVAL)
12108 if self.op.hypervisor is None:
12109 self.op.hypervisor = self.cfg.GetHypervisorType()
12110 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
12111 fname = _ExpandInstanceName(self.cfg, self.op.name)
12112 self.op.name = fname
12113 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
12114 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
12115 if not hasattr(self.op, "evac_nodes"):
12116 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
12117 " opcode input", errors.ECODE_INVAL)
12119 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
12120 self.op.mode, errors.ECODE_INVAL)
12122 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
12123 if self.op.allocator is None:
12124 raise errors.OpPrereqError("Missing allocator name",
12125 errors.ECODE_INVAL)
12126 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
12127 raise errors.OpPrereqError("Wrong allocator test '%s'" %
12128 self.op.direction, errors.ECODE_INVAL)
12130 def Exec(self, feedback_fn):
12131 """Run the allocator test.
12134 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
12135 ial = IAllocator(self.cfg, self.rpc,
12138 mem_size=self.op.mem_size,
12139 disks=self.op.disks,
12140 disk_template=self.op.disk_template,
12144 vcpus=self.op.vcpus,
12145 hypervisor=self.op.hypervisor,
12147 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
12148 ial = IAllocator(self.cfg, self.rpc,
12151 relocate_from=list(self.relocate_from),
12153 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
12154 ial = IAllocator(self.cfg, self.rpc,
12156 evac_nodes=self.op.evac_nodes)
12158 raise errors.ProgrammerError("Uncatched mode %s in"
12159 " LUTestAllocator.Exec", self.op.mode)
12161 if self.op.direction == constants.IALLOCATOR_DIR_IN:
12162 result = ial.in_text
12164 ial.Run(self.op.allocator, validate=False)
12165 result = ial.out_text
12169 #: Query type implementations
12171 constants.QR_INSTANCE: _InstanceQuery,
12172 constants.QR_NODE: _NodeQuery,
12173 constants.QR_GROUP: _GroupQuery,
12174 constants.QR_OS: _OsQuery,
12177 assert set(_QUERY_IMPL.keys()) == constants.QR_VIA_OP
12180 def _GetQueryImplementation(name):
12181 """Returns the implemtnation for a query type.
12183 @param name: Query type, must be one of L{constants.QR_VIA_OP}
12187 return _QUERY_IMPL[name]
12189 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
12190 errors.ECODE_INVAL)