4 # Copyright (C) 2006, 2007, 2008, 2009, 2010 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
43 from ganeti import ssh
44 from ganeti import utils
45 from ganeti import errors
46 from ganeti import hypervisor
47 from ganeti import locking
48 from ganeti import constants
49 from ganeti import objects
50 from ganeti import serializer
51 from ganeti import ssconf
52 from ganeti import uidpool
53 from ganeti import compat
54 from ganeti import masterd
55 from ganeti import netutils
58 import ganeti.masterd.instance # pylint: disable-msg=W0611
60 # Common opcode attributes
62 #: output fields for a query operation
63 _POutputFields = ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString))
66 #: the shutdown timeout
67 _PShutdownTimeout = ("shutdown_timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT,
70 #: the force parameter
71 _PForce = ("force", False, ht.TBool)
73 #: a required instance name (for single-instance LUs)
74 _PInstanceName = ("instance_name", ht.NoDefault, ht.TNonEmptyString)
76 #: Whether to ignore offline nodes
77 _PIgnoreOfflineNodes = ("ignore_offline_nodes", False, ht.TBool)
79 #: a required node name (for single-node LUs)
80 _PNodeName = ("node_name", ht.NoDefault, ht.TNonEmptyString)
82 #: the migration type (live/non-live)
83 _PMigrationMode = ("mode", None,
84 ht.TOr(ht.TNone, ht.TElemOf(constants.HT_MIGRATION_MODES)))
86 #: the obsolete 'live' mode (boolean)
87 _PMigrationLive = ("live", None, ht.TMaybeBool)
91 class LogicalUnit(object):
92 """Logical Unit base class.
94 Subclasses must follow these rules:
95 - implement ExpandNames
96 - implement CheckPrereq (except when tasklets are used)
97 - implement Exec (except when tasklets are used)
98 - implement BuildHooksEnv
99 - redefine HPATH and HTYPE
100 - optionally redefine their run requirements:
101 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
103 Note that all commands require root permissions.
105 @ivar dry_run_result: the value (if any) that will be returned to the caller
106 in dry-run mode (signalled by opcode dry_run parameter)
107 @cvar _OP_PARAMS: a list of opcode attributes, their defaults values
108 they should get if not already defined, and types they must match
116 def __init__(self, processor, op, context, rpc):
117 """Constructor for LogicalUnit.
119 This needs to be overridden in derived classes in order to check op
123 self.proc = processor
125 self.cfg = context.cfg
126 self.context = context
128 # Dicts used to declare locking needs to mcpu
129 self.needed_locks = None
130 self.acquired_locks = {}
131 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
133 self.remove_locks = {}
134 # Used to force good behavior when calling helper functions
135 self.recalculate_locks = {}
138 self.Log = processor.Log # pylint: disable-msg=C0103
139 self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
140 self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
141 self.LogStep = processor.LogStep # pylint: disable-msg=C0103
142 # support for dry-run
143 self.dry_run_result = None
144 # support for generic debug attribute
145 if (not hasattr(self.op, "debug_level") or
146 not isinstance(self.op.debug_level, int)):
147 self.op.debug_level = 0
152 # The new kind-of-type-system
153 op_id = self.op.OP_ID
154 for attr_name, aval, test in self._OP_PARAMS:
155 if not hasattr(op, attr_name):
156 if aval == ht.NoDefault:
157 raise errors.OpPrereqError("Required parameter '%s.%s' missing" %
158 (op_id, attr_name), errors.ECODE_INVAL)
164 setattr(self.op, attr_name, dval)
165 attr_val = getattr(op, attr_name)
166 if test == ht.NoType:
169 if not callable(test):
170 raise errors.ProgrammerError("Validation for parameter '%s.%s' failed,"
171 " given type is not a proper type (%s)" %
172 (op_id, attr_name, test))
173 if not test(attr_val):
174 logging.error("OpCode %s, parameter %s, has invalid type %s/value %s",
175 self.op.OP_ID, attr_name, type(attr_val), attr_val)
176 raise errors.OpPrereqError("Parameter '%s.%s' fails validation" %
177 (op_id, attr_name), errors.ECODE_INVAL)
179 self.CheckArguments()
182 """Returns the SshRunner object
186 self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
189 ssh = property(fget=__GetSSH)
191 def CheckArguments(self):
192 """Check syntactic validity for the opcode arguments.
194 This method is for doing a simple syntactic check and ensure
195 validity of opcode parameters, without any cluster-related
196 checks. While the same can be accomplished in ExpandNames and/or
197 CheckPrereq, doing these separate is better because:
199 - ExpandNames is left as as purely a lock-related function
200 - CheckPrereq is run after we have acquired locks (and possible
203 The function is allowed to change the self.op attribute so that
204 later methods can no longer worry about missing parameters.
209 def ExpandNames(self):
210 """Expand names for this LU.
212 This method is called before starting to execute the opcode, and it should
213 update all the parameters of the opcode to their canonical form (e.g. a
214 short node name must be fully expanded after this method has successfully
215 completed). This way locking, hooks, logging, ecc. can work correctly.
217 LUs which implement this method must also populate the self.needed_locks
218 member, as a dict with lock levels as keys, and a list of needed lock names
221 - use an empty dict if you don't need any lock
222 - if you don't need any lock at a particular level omit that level
223 - don't put anything for the BGL level
224 - if you want all locks at a level use locking.ALL_SET as a value
226 If you need to share locks (rather than acquire them exclusively) at one
227 level you can modify self.share_locks, setting a true value (usually 1) for
228 that level. By default locks are not shared.
230 This function can also define a list of tasklets, which then will be
231 executed in order instead of the usual LU-level CheckPrereq and Exec
232 functions, if those are not defined by the LU.
236 # Acquire all nodes and one instance
237 self.needed_locks = {
238 locking.LEVEL_NODE: locking.ALL_SET,
239 locking.LEVEL_INSTANCE: ['instance1.example.com'],
241 # Acquire just two nodes
242 self.needed_locks = {
243 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
246 self.needed_locks = {} # No, you can't leave it to the default value None
249 # The implementation of this method is mandatory only if the new LU is
250 # concurrent, so that old LUs don't need to be changed all at the same
253 self.needed_locks = {} # Exclusive LUs don't need locks.
255 raise NotImplementedError
257 def DeclareLocks(self, level):
258 """Declare LU locking needs for a level
260 While most LUs can just declare their locking needs at ExpandNames time,
261 sometimes there's the need to calculate some locks after having acquired
262 the ones before. This function is called just before acquiring locks at a
263 particular level, but after acquiring the ones at lower levels, and permits
264 such calculations. It can be used to modify self.needed_locks, and by
265 default it does nothing.
267 This function is only called if you have something already set in
268 self.needed_locks for the level.
270 @param level: Locking level which is going to be locked
271 @type level: member of ganeti.locking.LEVELS
275 def CheckPrereq(self):
276 """Check prerequisites for this LU.
278 This method should check that the prerequisites for the execution
279 of this LU are fulfilled. It can do internode communication, but
280 it should be idempotent - no cluster or system changes are
283 The method should raise errors.OpPrereqError in case something is
284 not fulfilled. Its return value is ignored.
286 This method should also update all the parameters of the opcode to
287 their canonical form if it hasn't been done by ExpandNames before.
290 if self.tasklets is not None:
291 for (idx, tl) in enumerate(self.tasklets):
292 logging.debug("Checking prerequisites for tasklet %s/%s",
293 idx + 1, len(self.tasklets))
298 def Exec(self, feedback_fn):
301 This method should implement the actual work. It should raise
302 errors.OpExecError for failures that are somewhat dealt with in
306 if self.tasklets is not None:
307 for (idx, tl) in enumerate(self.tasklets):
308 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
311 raise NotImplementedError
313 def BuildHooksEnv(self):
314 """Build hooks environment for this LU.
316 This method should return a three-node tuple consisting of: a dict
317 containing the environment that will be used for running the
318 specific hook for this LU, a list of node names on which the hook
319 should run before the execution, and a list of node names on which
320 the hook should run after the execution.
322 The keys of the dict must not have 'GANETI_' prefixed as this will
323 be handled in the hooks runner. Also note additional keys will be
324 added by the hooks runner. If the LU doesn't define any
325 environment, an empty dict (and not None) should be returned.
327 No nodes should be returned as an empty list (and not None).
329 Note that if the HPATH for a LU class is None, this function will
333 raise NotImplementedError
335 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
336 """Notify the LU about the results of its hooks.
338 This method is called every time a hooks phase is executed, and notifies
339 the Logical Unit about the hooks' result. The LU can then use it to alter
340 its result based on the hooks. By default the method does nothing and the
341 previous result is passed back unchanged but any LU can define it if it
342 wants to use the local cluster hook-scripts somehow.
344 @param phase: one of L{constants.HOOKS_PHASE_POST} or
345 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
346 @param hook_results: the results of the multi-node hooks rpc call
347 @param feedback_fn: function used send feedback back to the caller
348 @param lu_result: the previous Exec result this LU had, or None
350 @return: the new Exec result, based on the previous result
354 # API must be kept, thus we ignore the unused argument and could
355 # be a function warnings
356 # pylint: disable-msg=W0613,R0201
359 def _ExpandAndLockInstance(self):
360 """Helper function to expand and lock an instance.
362 Many LUs that work on an instance take its name in self.op.instance_name
363 and need to expand it and then declare the expanded name for locking. This
364 function does it, and then updates self.op.instance_name to the expanded
365 name. It also initializes needed_locks as a dict, if this hasn't been done
369 if self.needed_locks is None:
370 self.needed_locks = {}
372 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
373 "_ExpandAndLockInstance called with instance-level locks set"
374 self.op.instance_name = _ExpandInstanceName(self.cfg,
375 self.op.instance_name)
376 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
378 def _LockInstancesNodes(self, primary_only=False):
379 """Helper function to declare instances' nodes for locking.
381 This function should be called after locking one or more instances to lock
382 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
383 with all primary or secondary nodes for instances already locked and
384 present in self.needed_locks[locking.LEVEL_INSTANCE].
386 It should be called from DeclareLocks, and for safety only works if
387 self.recalculate_locks[locking.LEVEL_NODE] is set.
389 In the future it may grow parameters to just lock some instance's nodes, or
390 to just lock primaries or secondary nodes, if needed.
392 If should be called in DeclareLocks in a way similar to::
394 if level == locking.LEVEL_NODE:
395 self._LockInstancesNodes()
397 @type primary_only: boolean
398 @param primary_only: only lock primary nodes of locked instances
401 assert locking.LEVEL_NODE in self.recalculate_locks, \
402 "_LockInstancesNodes helper function called with no nodes to recalculate"
404 # TODO: check if we're really been called with the instance locks held
406 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
407 # future we might want to have different behaviors depending on the value
408 # of self.recalculate_locks[locking.LEVEL_NODE]
410 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
411 instance = self.context.cfg.GetInstanceInfo(instance_name)
412 wanted_nodes.append(instance.primary_node)
414 wanted_nodes.extend(instance.secondary_nodes)
416 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
417 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
418 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
419 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
421 del self.recalculate_locks[locking.LEVEL_NODE]
424 class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
425 """Simple LU which runs no hooks.
427 This LU is intended as a parent for other LogicalUnits which will
428 run no hooks, in order to reduce duplicate code.
434 def BuildHooksEnv(self):
435 """Empty BuildHooksEnv for NoHooksLu.
437 This just raises an error.
440 assert False, "BuildHooksEnv called for NoHooksLUs"
444 """Tasklet base class.
446 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
447 they can mix legacy code with tasklets. Locking needs to be done in the LU,
448 tasklets know nothing about locks.
450 Subclasses must follow these rules:
451 - Implement CheckPrereq
455 def __init__(self, lu):
462 def CheckPrereq(self):
463 """Check prerequisites for this tasklets.
465 This method should check whether the prerequisites for the execution of
466 this tasklet are fulfilled. It can do internode communication, but it
467 should be idempotent - no cluster or system changes are allowed.
469 The method should raise errors.OpPrereqError in case something is not
470 fulfilled. Its return value is ignored.
472 This method should also update all parameters to their canonical form if it
473 hasn't been done before.
478 def Exec(self, feedback_fn):
479 """Execute the tasklet.
481 This method should implement the actual work. It should raise
482 errors.OpExecError for failures that are somewhat dealt with in code, or
486 raise NotImplementedError
489 def _GetWantedNodes(lu, nodes):
490 """Returns list of checked and expanded node names.
492 @type lu: L{LogicalUnit}
493 @param lu: the logical unit on whose behalf we execute
495 @param nodes: list of node names or None for all nodes
497 @return: the list of nodes, sorted
498 @raise errors.ProgrammerError: if the nodes parameter is wrong type
502 raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
503 " non-empty list of nodes whose name is to be expanded.")
505 wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
506 return utils.NiceSort(wanted)
509 def _GetWantedInstances(lu, instances):
510 """Returns list of checked and expanded instance names.
512 @type lu: L{LogicalUnit}
513 @param lu: the logical unit on whose behalf we execute
514 @type instances: list
515 @param instances: list of instance names or None for all instances
517 @return: the list of instances, sorted
518 @raise errors.OpPrereqError: if the instances parameter is wrong type
519 @raise errors.OpPrereqError: if any of the passed instances is not found
523 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
525 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
529 def _GetUpdatedParams(old_params, update_dict,
530 use_default=True, use_none=False):
531 """Return the new version of a parameter dictionary.
533 @type old_params: dict
534 @param old_params: old parameters
535 @type update_dict: dict
536 @param update_dict: dict containing new parameter values, or
537 constants.VALUE_DEFAULT to reset the parameter to its default
539 @param use_default: boolean
540 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
541 values as 'to be deleted' values
542 @param use_none: boolean
543 @type use_none: whether to recognise C{None} values as 'to be
546 @return: the new parameter dictionary
549 params_copy = copy.deepcopy(old_params)
550 for key, val in update_dict.iteritems():
551 if ((use_default and val == constants.VALUE_DEFAULT) or
552 (use_none and val is None)):
558 params_copy[key] = val
562 def _CheckOutputFields(static, dynamic, selected):
563 """Checks whether all selected fields are valid.
565 @type static: L{utils.FieldSet}
566 @param static: static fields set
567 @type dynamic: L{utils.FieldSet}
568 @param dynamic: dynamic fields set
575 delta = f.NonMatching(selected)
577 raise errors.OpPrereqError("Unknown output fields selected: %s"
578 % ",".join(delta), errors.ECODE_INVAL)
581 def _CheckGlobalHvParams(params):
582 """Validates that given hypervisor params are not global ones.
584 This will ensure that instances don't get customised versions of
588 used_globals = constants.HVC_GLOBALS.intersection(params)
590 msg = ("The following hypervisor parameters are global and cannot"
591 " be customized at instance level, please modify them at"
592 " cluster level: %s" % utils.CommaJoin(used_globals))
593 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
596 def _CheckNodeOnline(lu, node):
597 """Ensure that a given node is online.
599 @param lu: the LU on behalf of which we make the check
600 @param node: the node to check
601 @raise errors.OpPrereqError: if the node is offline
604 if lu.cfg.GetNodeInfo(node).offline:
605 raise errors.OpPrereqError("Can't use offline node %s" % node,
609 def _CheckNodeNotDrained(lu, node):
610 """Ensure that a given node is not drained.
612 @param lu: the LU on behalf of which we make the check
613 @param node: the node to check
614 @raise errors.OpPrereqError: if the node is drained
617 if lu.cfg.GetNodeInfo(node).drained:
618 raise errors.OpPrereqError("Can't use drained node %s" % node,
622 def _CheckNodeHasOS(lu, node, os_name, force_variant):
623 """Ensure that a node supports a given OS.
625 @param lu: the LU on behalf of which we make the check
626 @param node: the node to check
627 @param os_name: the OS to query about
628 @param force_variant: whether to ignore variant errors
629 @raise errors.OpPrereqError: if the node is not supporting the OS
632 result = lu.rpc.call_os_get(node, os_name)
633 result.Raise("OS '%s' not in supported OS list for node %s" %
635 prereq=True, ecode=errors.ECODE_INVAL)
636 if not force_variant:
637 _CheckOSVariant(result.payload, os_name)
640 def _RequireFileStorage():
641 """Checks that file storage is enabled.
643 @raise errors.OpPrereqError: when file storage is disabled
646 if not constants.ENABLE_FILE_STORAGE:
647 raise errors.OpPrereqError("File storage disabled at configure time",
651 def _CheckDiskTemplate(template):
652 """Ensure a given disk template is valid.
655 if template not in constants.DISK_TEMPLATES:
656 msg = ("Invalid disk template name '%s', valid templates are: %s" %
657 (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
658 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
659 if template == constants.DT_FILE:
660 _RequireFileStorage()
664 def _CheckStorageType(storage_type):
665 """Ensure a given storage type is valid.
668 if storage_type not in constants.VALID_STORAGE_TYPES:
669 raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
671 if storage_type == constants.ST_FILE:
672 _RequireFileStorage()
676 def _GetClusterDomainSecret():
677 """Reads the cluster domain secret.
680 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
684 def _CheckInstanceDown(lu, instance, reason):
685 """Ensure that an instance is not running."""
686 if instance.admin_up:
687 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
688 (instance.name, reason), errors.ECODE_STATE)
690 pnode = instance.primary_node
691 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
692 ins_l.Raise("Can't contact node %s for instance information" % pnode,
693 prereq=True, ecode=errors.ECODE_ENVIRON)
695 if instance.name in ins_l.payload:
696 raise errors.OpPrereqError("Instance %s is running, %s" %
697 (instance.name, reason), errors.ECODE_STATE)
700 def _ExpandItemName(fn, name, kind):
701 """Expand an item name.
703 @param fn: the function to use for expansion
704 @param name: requested item name
705 @param kind: text description ('Node' or 'Instance')
706 @return: the resolved (full) name
707 @raise errors.OpPrereqError: if the item is not found
711 if full_name is None:
712 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
717 def _ExpandNodeName(cfg, name):
718 """Wrapper over L{_ExpandItemName} for nodes."""
719 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
722 def _ExpandInstanceName(cfg, name):
723 """Wrapper over L{_ExpandItemName} for instance."""
724 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
727 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
728 memory, vcpus, nics, disk_template, disks,
729 bep, hvp, hypervisor_name):
730 """Builds instance related env variables for hooks
732 This builds the hook environment from individual variables.
735 @param name: the name of the instance
736 @type primary_node: string
737 @param primary_node: the name of the instance's primary node
738 @type secondary_nodes: list
739 @param secondary_nodes: list of secondary nodes as strings
740 @type os_type: string
741 @param os_type: the name of the instance's OS
742 @type status: boolean
743 @param status: the should_run status of the instance
745 @param memory: the memory size of the instance
747 @param vcpus: the count of VCPUs the instance has
749 @param nics: list of tuples (ip, mac, mode, link) representing
750 the NICs the instance has
751 @type disk_template: string
752 @param disk_template: the disk template of the instance
754 @param disks: the list of (size, mode) pairs
756 @param bep: the backend parameters for the instance
758 @param hvp: the hypervisor parameters for the instance
759 @type hypervisor_name: string
760 @param hypervisor_name: the hypervisor for the instance
762 @return: the hook environment for this instance
771 "INSTANCE_NAME": name,
772 "INSTANCE_PRIMARY": primary_node,
773 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
774 "INSTANCE_OS_TYPE": os_type,
775 "INSTANCE_STATUS": str_status,
776 "INSTANCE_MEMORY": memory,
777 "INSTANCE_VCPUS": vcpus,
778 "INSTANCE_DISK_TEMPLATE": disk_template,
779 "INSTANCE_HYPERVISOR": hypervisor_name,
783 nic_count = len(nics)
784 for idx, (ip, mac, mode, link) in enumerate(nics):
787 env["INSTANCE_NIC%d_IP" % idx] = ip
788 env["INSTANCE_NIC%d_MAC" % idx] = mac
789 env["INSTANCE_NIC%d_MODE" % idx] = mode
790 env["INSTANCE_NIC%d_LINK" % idx] = link
791 if mode == constants.NIC_MODE_BRIDGED:
792 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
796 env["INSTANCE_NIC_COUNT"] = nic_count
799 disk_count = len(disks)
800 for idx, (size, mode) in enumerate(disks):
801 env["INSTANCE_DISK%d_SIZE" % idx] = size
802 env["INSTANCE_DISK%d_MODE" % idx] = mode
806 env["INSTANCE_DISK_COUNT"] = disk_count
808 for source, kind in [(bep, "BE"), (hvp, "HV")]:
809 for key, value in source.items():
810 env["INSTANCE_%s_%s" % (kind, key)] = value
815 def _NICListToTuple(lu, nics):
816 """Build a list of nic information tuples.
818 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
819 value in LUQueryInstanceData.
821 @type lu: L{LogicalUnit}
822 @param lu: the logical unit on whose behalf we execute
823 @type nics: list of L{objects.NIC}
824 @param nics: list of nics to convert to hooks tuples
828 cluster = lu.cfg.GetClusterInfo()
832 filled_params = cluster.SimpleFillNIC(nic.nicparams)
833 mode = filled_params[constants.NIC_MODE]
834 link = filled_params[constants.NIC_LINK]
835 hooks_nics.append((ip, mac, mode, link))
839 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
840 """Builds instance related env variables for hooks from an object.
842 @type lu: L{LogicalUnit}
843 @param lu: the logical unit on whose behalf we execute
844 @type instance: L{objects.Instance}
845 @param instance: the instance for which we should build the
848 @param override: dictionary with key/values that will override
851 @return: the hook environment dictionary
854 cluster = lu.cfg.GetClusterInfo()
855 bep = cluster.FillBE(instance)
856 hvp = cluster.FillHV(instance)
858 'name': instance.name,
859 'primary_node': instance.primary_node,
860 'secondary_nodes': instance.secondary_nodes,
861 'os_type': instance.os,
862 'status': instance.admin_up,
863 'memory': bep[constants.BE_MEMORY],
864 'vcpus': bep[constants.BE_VCPUS],
865 'nics': _NICListToTuple(lu, instance.nics),
866 'disk_template': instance.disk_template,
867 'disks': [(disk.size, disk.mode) for disk in instance.disks],
870 'hypervisor_name': instance.hypervisor,
873 args.update(override)
874 return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
877 def _AdjustCandidatePool(lu, exceptions):
878 """Adjust the candidate pool after node operations.
881 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
883 lu.LogInfo("Promoted nodes to master candidate role: %s",
884 utils.CommaJoin(node.name for node in mod_list))
885 for name in mod_list:
886 lu.context.ReaddNode(name)
887 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
889 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
893 def _DecideSelfPromotion(lu, exceptions=None):
894 """Decide whether I should promote myself as a master candidate.
897 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
898 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
899 # the new node will increase mc_max with one, so:
900 mc_should = min(mc_should + 1, cp_size)
901 return mc_now < mc_should
904 def _CheckNicsBridgesExist(lu, target_nics, target_node):
905 """Check that the brigdes needed by a list of nics exist.
908 cluster = lu.cfg.GetClusterInfo()
909 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
910 brlist = [params[constants.NIC_LINK] for params in paramslist
911 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
913 result = lu.rpc.call_bridges_exist(target_node, brlist)
914 result.Raise("Error checking bridges on destination node '%s'" %
915 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
918 def _CheckInstanceBridgesExist(lu, instance, node=None):
919 """Check that the brigdes needed by an instance exist.
923 node = instance.primary_node
924 _CheckNicsBridgesExist(lu, instance.nics, node)
927 def _CheckOSVariant(os_obj, name):
928 """Check whether an OS name conforms to the os variants specification.
930 @type os_obj: L{objects.OS}
931 @param os_obj: OS object to check
933 @param name: OS name passed by the user, to check for validity
936 if not os_obj.supported_variants:
938 variant = objects.OS.GetVariant(name)
940 raise errors.OpPrereqError("OS name must include a variant",
943 if variant not in os_obj.supported_variants:
944 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
947 def _GetNodeInstancesInner(cfg, fn):
948 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
951 def _GetNodeInstances(cfg, node_name):
952 """Returns a list of all primary and secondary instances on a node.
956 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
959 def _GetNodePrimaryInstances(cfg, node_name):
960 """Returns primary instances on a node.
963 return _GetNodeInstancesInner(cfg,
964 lambda inst: node_name == inst.primary_node)
967 def _GetNodeSecondaryInstances(cfg, node_name):
968 """Returns secondary instances on a node.
971 return _GetNodeInstancesInner(cfg,
972 lambda inst: node_name in inst.secondary_nodes)
975 def _GetStorageTypeArgs(cfg, storage_type):
976 """Returns the arguments for a storage type.
979 # Special case for file storage
980 if storage_type == constants.ST_FILE:
981 # storage.FileStorage wants a list of storage directories
982 return [[cfg.GetFileStorageDir()]]
987 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
990 for dev in instance.disks:
991 cfg.SetDiskID(dev, node_name)
993 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
994 result.Raise("Failed to get disk status from node %s" % node_name,
995 prereq=prereq, ecode=errors.ECODE_ENVIRON)
997 for idx, bdev_status in enumerate(result.payload):
998 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1004 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1005 """Check the sanity of iallocator and node arguments and use the
1006 cluster-wide iallocator if appropriate.
1008 Check that at most one of (iallocator, node) is specified. If none is
1009 specified, then the LU's opcode's iallocator slot is filled with the
1010 cluster-wide default iallocator.
1012 @type iallocator_slot: string
1013 @param iallocator_slot: the name of the opcode iallocator slot
1014 @type node_slot: string
1015 @param node_slot: the name of the opcode target node slot
1018 node = getattr(lu.op, node_slot, None)
1019 iallocator = getattr(lu.op, iallocator_slot, None)
1021 if node is not None and iallocator is not None:
1022 raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1024 elif node is None and iallocator is None:
1025 default_iallocator = lu.cfg.GetDefaultIAllocator()
1026 if default_iallocator:
1027 setattr(lu.op, iallocator_slot, default_iallocator)
1029 raise errors.OpPrereqError("No iallocator or node given and no"
1030 " cluster-wide default iallocator found."
1031 " Please specify either an iallocator or a"
1032 " node, or set a cluster-wide default"
1036 class LUPostInitCluster(LogicalUnit):
1037 """Logical unit for running hooks after cluster initialization.
1040 HPATH = "cluster-init"
1041 HTYPE = constants.HTYPE_CLUSTER
1043 def BuildHooksEnv(self):
1047 env = {"OP_TARGET": self.cfg.GetClusterName()}
1048 mn = self.cfg.GetMasterNode()
1049 return env, [], [mn]
1051 def Exec(self, feedback_fn):
1058 class LUDestroyCluster(LogicalUnit):
1059 """Logical unit for destroying the cluster.
1062 HPATH = "cluster-destroy"
1063 HTYPE = constants.HTYPE_CLUSTER
1065 def BuildHooksEnv(self):
1069 env = {"OP_TARGET": self.cfg.GetClusterName()}
1072 def CheckPrereq(self):
1073 """Check prerequisites.
1075 This checks whether the cluster is empty.
1077 Any errors are signaled by raising errors.OpPrereqError.
1080 master = self.cfg.GetMasterNode()
1082 nodelist = self.cfg.GetNodeList()
1083 if len(nodelist) != 1 or nodelist[0] != master:
1084 raise errors.OpPrereqError("There are still %d node(s) in"
1085 " this cluster." % (len(nodelist) - 1),
1087 instancelist = self.cfg.GetInstanceList()
1089 raise errors.OpPrereqError("There are still %d instance(s) in"
1090 " this cluster." % len(instancelist),
1093 def Exec(self, feedback_fn):
1094 """Destroys the cluster.
1097 master = self.cfg.GetMasterNode()
1099 # Run post hooks on master node before it's removed
1100 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1102 hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1104 # pylint: disable-msg=W0702
1105 self.LogWarning("Errors occurred running hooks on %s" % master)
1107 result = self.rpc.call_node_stop_master(master, False)
1108 result.Raise("Could not disable the master role")
1113 def _VerifyCertificate(filename):
1114 """Verifies a certificate for LUVerifyCluster.
1116 @type filename: string
1117 @param filename: Path to PEM file
1121 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1122 utils.ReadFile(filename))
1123 except Exception, err: # pylint: disable-msg=W0703
1124 return (LUVerifyCluster.ETYPE_ERROR,
1125 "Failed to load X509 certificate %s: %s" % (filename, err))
1128 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1129 constants.SSL_CERT_EXPIRATION_ERROR)
1132 fnamemsg = "While verifying %s: %s" % (filename, msg)
1137 return (None, fnamemsg)
1138 elif errcode == utils.CERT_WARNING:
1139 return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
1140 elif errcode == utils.CERT_ERROR:
1141 return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
1143 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1146 class LUVerifyCluster(LogicalUnit):
1147 """Verifies the cluster status.
1150 HPATH = "cluster-verify"
1151 HTYPE = constants.HTYPE_CLUSTER
1153 ("skip_checks", ht.EmptyList,
1154 ht.TListOf(ht.TElemOf(constants.VERIFY_OPTIONAL_CHECKS))),
1155 ("verbose", False, ht.TBool),
1156 ("error_codes", False, ht.TBool),
1157 ("debug_simulate_errors", False, ht.TBool),
1161 TCLUSTER = "cluster"
1163 TINSTANCE = "instance"
1165 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1166 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1167 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1168 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1169 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1170 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1171 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1172 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1173 ENODEDRBD = (TNODE, "ENODEDRBD")
1174 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1175 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1176 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1177 ENODEHV = (TNODE, "ENODEHV")
1178 ENODELVM = (TNODE, "ENODELVM")
1179 ENODEN1 = (TNODE, "ENODEN1")
1180 ENODENET = (TNODE, "ENODENET")
1181 ENODEOS = (TNODE, "ENODEOS")
1182 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1183 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1184 ENODERPC = (TNODE, "ENODERPC")
1185 ENODESSH = (TNODE, "ENODESSH")
1186 ENODEVERSION = (TNODE, "ENODEVERSION")
1187 ENODESETUP = (TNODE, "ENODESETUP")
1188 ENODETIME = (TNODE, "ENODETIME")
1190 ETYPE_FIELD = "code"
1191 ETYPE_ERROR = "ERROR"
1192 ETYPE_WARNING = "WARNING"
1194 class NodeImage(object):
1195 """A class representing the logical and physical status of a node.
1198 @ivar name: the node name to which this object refers
1199 @ivar volumes: a structure as returned from
1200 L{ganeti.backend.GetVolumeList} (runtime)
1201 @ivar instances: a list of running instances (runtime)
1202 @ivar pinst: list of configured primary instances (config)
1203 @ivar sinst: list of configured secondary instances (config)
1204 @ivar sbp: diction of {secondary-node: list of instances} of all peers
1205 of this node (config)
1206 @ivar mfree: free memory, as reported by hypervisor (runtime)
1207 @ivar dfree: free disk, as reported by the node (runtime)
1208 @ivar offline: the offline status (config)
1209 @type rpc_fail: boolean
1210 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1211 not whether the individual keys were correct) (runtime)
1212 @type lvm_fail: boolean
1213 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1214 @type hyp_fail: boolean
1215 @ivar hyp_fail: whether the RPC call didn't return the instance list
1216 @type ghost: boolean
1217 @ivar ghost: whether this is a known node or not (config)
1218 @type os_fail: boolean
1219 @ivar os_fail: whether the RPC call didn't return valid OS data
1221 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1224 def __init__(self, offline=False, name=None):
1233 self.offline = offline
1234 self.rpc_fail = False
1235 self.lvm_fail = False
1236 self.hyp_fail = False
1238 self.os_fail = False
1241 def ExpandNames(self):
1242 self.needed_locks = {
1243 locking.LEVEL_NODE: locking.ALL_SET,
1244 locking.LEVEL_INSTANCE: locking.ALL_SET,
1246 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1248 def _Error(self, ecode, item, msg, *args, **kwargs):
1249 """Format an error message.
1251 Based on the opcode's error_codes parameter, either format a
1252 parseable error code, or a simpler error string.
1254 This must be called only from Exec and functions called from Exec.
1257 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1259 # first complete the msg
1262 # then format the whole message
1263 if self.op.error_codes:
1264 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1270 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1271 # and finally report it via the feedback_fn
1272 self._feedback_fn(" - %s" % msg)
1274 def _ErrorIf(self, cond, *args, **kwargs):
1275 """Log an error message if the passed condition is True.
1278 cond = bool(cond) or self.op.debug_simulate_errors
1280 self._Error(*args, **kwargs)
1281 # do not mark the operation as failed for WARN cases only
1282 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1283 self.bad = self.bad or cond
1285 def _VerifyNode(self, ninfo, nresult):
1286 """Perform some basic validation on data returned from a node.
1288 - check the result data structure is well formed and has all the
1290 - check ganeti version
1292 @type ninfo: L{objects.Node}
1293 @param ninfo: the node to check
1294 @param nresult: the results from the node
1296 @return: whether overall this call was successful (and we can expect
1297 reasonable values in the respose)
1301 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1303 # main result, nresult should be a non-empty dict
1304 test = not nresult or not isinstance(nresult, dict)
1305 _ErrorIf(test, self.ENODERPC, node,
1306 "unable to verify node: no data returned")
1310 # compares ganeti version
1311 local_version = constants.PROTOCOL_VERSION
1312 remote_version = nresult.get("version", None)
1313 test = not (remote_version and
1314 isinstance(remote_version, (list, tuple)) and
1315 len(remote_version) == 2)
1316 _ErrorIf(test, self.ENODERPC, node,
1317 "connection to node returned invalid data")
1321 test = local_version != remote_version[0]
1322 _ErrorIf(test, self.ENODEVERSION, node,
1323 "incompatible protocol versions: master %s,"
1324 " node %s", local_version, remote_version[0])
1328 # node seems compatible, we can actually try to look into its results
1330 # full package version
1331 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1332 self.ENODEVERSION, node,
1333 "software version mismatch: master %s, node %s",
1334 constants.RELEASE_VERSION, remote_version[1],
1335 code=self.ETYPE_WARNING)
1337 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1338 if isinstance(hyp_result, dict):
1339 for hv_name, hv_result in hyp_result.iteritems():
1340 test = hv_result is not None
1341 _ErrorIf(test, self.ENODEHV, node,
1342 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1345 test = nresult.get(constants.NV_NODESETUP,
1346 ["Missing NODESETUP results"])
1347 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1352 def _VerifyNodeTime(self, ninfo, nresult,
1353 nvinfo_starttime, nvinfo_endtime):
1354 """Check the node time.
1356 @type ninfo: L{objects.Node}
1357 @param ninfo: the node to check
1358 @param nresult: the remote results for the node
1359 @param nvinfo_starttime: the start time of the RPC call
1360 @param nvinfo_endtime: the end time of the RPC call
1364 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1366 ntime = nresult.get(constants.NV_TIME, None)
1368 ntime_merged = utils.MergeTime(ntime)
1369 except (ValueError, TypeError):
1370 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1373 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1374 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1375 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1376 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1380 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1381 "Node time diverges by at least %s from master node time",
1384 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1385 """Check the node time.
1387 @type ninfo: L{objects.Node}
1388 @param ninfo: the node to check
1389 @param nresult: the remote results for the node
1390 @param vg_name: the configured VG name
1397 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1399 # checks vg existence and size > 20G
1400 vglist = nresult.get(constants.NV_VGLIST, None)
1402 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1404 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1405 constants.MIN_VG_SIZE)
1406 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1409 pvlist = nresult.get(constants.NV_PVLIST, None)
1410 test = pvlist is None
1411 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1413 # check that ':' is not present in PV names, since it's a
1414 # special character for lvcreate (denotes the range of PEs to
1416 for _, pvname, owner_vg in pvlist:
1417 test = ":" in pvname
1418 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1419 " '%s' of VG '%s'", pvname, owner_vg)
1421 def _VerifyNodeNetwork(self, ninfo, nresult):
1422 """Check the node time.
1424 @type ninfo: L{objects.Node}
1425 @param ninfo: the node to check
1426 @param nresult: the remote results for the node
1430 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1432 test = constants.NV_NODELIST not in nresult
1433 _ErrorIf(test, self.ENODESSH, node,
1434 "node hasn't returned node ssh connectivity data")
1436 if nresult[constants.NV_NODELIST]:
1437 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1438 _ErrorIf(True, self.ENODESSH, node,
1439 "ssh communication with node '%s': %s", a_node, a_msg)
1441 test = constants.NV_NODENETTEST not in nresult
1442 _ErrorIf(test, self.ENODENET, node,
1443 "node hasn't returned node tcp connectivity data")
1445 if nresult[constants.NV_NODENETTEST]:
1446 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1448 _ErrorIf(True, self.ENODENET, node,
1449 "tcp communication with node '%s': %s",
1450 anode, nresult[constants.NV_NODENETTEST][anode])
1452 test = constants.NV_MASTERIP not in nresult
1453 _ErrorIf(test, self.ENODENET, node,
1454 "node hasn't returned node master IP reachability data")
1456 if not nresult[constants.NV_MASTERIP]:
1457 if node == self.master_node:
1458 msg = "the master node cannot reach the master IP (not configured?)"
1460 msg = "cannot reach the master IP"
1461 _ErrorIf(True, self.ENODENET, node, msg)
1464 def _VerifyInstance(self, instance, instanceconfig, node_image):
1465 """Verify an instance.
1467 This function checks to see if the required block devices are
1468 available on the instance's node.
1471 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1472 node_current = instanceconfig.primary_node
1474 node_vol_should = {}
1475 instanceconfig.MapLVsByNode(node_vol_should)
1477 for node in node_vol_should:
1478 n_img = node_image[node]
1479 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1480 # ignore missing volumes on offline or broken nodes
1482 for volume in node_vol_should[node]:
1483 test = volume not in n_img.volumes
1484 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1485 "volume %s missing on node %s", volume, node)
1487 if instanceconfig.admin_up:
1488 pri_img = node_image[node_current]
1489 test = instance not in pri_img.instances and not pri_img.offline
1490 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1491 "instance not running on its primary node %s",
1494 for node, n_img in node_image.items():
1495 if (not node == node_current):
1496 test = instance in n_img.instances
1497 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1498 "instance should not run on node %s", node)
1500 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1501 """Verify if there are any unknown volumes in the cluster.
1503 The .os, .swap and backup volumes are ignored. All other volumes are
1504 reported as unknown.
1506 @type reserved: L{ganeti.utils.FieldSet}
1507 @param reserved: a FieldSet of reserved volume names
1510 for node, n_img in node_image.items():
1511 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1512 # skip non-healthy nodes
1514 for volume in n_img.volumes:
1515 test = ((node not in node_vol_should or
1516 volume not in node_vol_should[node]) and
1517 not reserved.Matches(volume))
1518 self._ErrorIf(test, self.ENODEORPHANLV, node,
1519 "volume %s is unknown", volume)
1521 def _VerifyOrphanInstances(self, instancelist, node_image):
1522 """Verify the list of running instances.
1524 This checks what instances are running but unknown to the cluster.
1527 for node, n_img in node_image.items():
1528 for o_inst in n_img.instances:
1529 test = o_inst not in instancelist
1530 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1531 "instance %s on node %s should not exist", o_inst, node)
1533 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1534 """Verify N+1 Memory Resilience.
1536 Check that if one single node dies we can still start all the
1537 instances it was primary for.
1540 for node, n_img in node_image.items():
1541 # This code checks that every node which is now listed as
1542 # secondary has enough memory to host all instances it is
1543 # supposed to should a single other node in the cluster fail.
1544 # FIXME: not ready for failover to an arbitrary node
1545 # FIXME: does not support file-backed instances
1546 # WARNING: we currently take into account down instances as well
1547 # as up ones, considering that even if they're down someone
1548 # might want to start them even in the event of a node failure.
1549 for prinode, instances in n_img.sbp.items():
1551 for instance in instances:
1552 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1553 if bep[constants.BE_AUTO_BALANCE]:
1554 needed_mem += bep[constants.BE_MEMORY]
1555 test = n_img.mfree < needed_mem
1556 self._ErrorIf(test, self.ENODEN1, node,
1557 "not enough memory on to accommodate"
1558 " failovers should peer node %s fail", prinode)
1560 def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1562 """Verifies and computes the node required file checksums.
1564 @type ninfo: L{objects.Node}
1565 @param ninfo: the node to check
1566 @param nresult: the remote results for the node
1567 @param file_list: required list of files
1568 @param local_cksum: dictionary of local files and their checksums
1569 @param master_files: list of files that only masters should have
1573 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1575 remote_cksum = nresult.get(constants.NV_FILELIST, None)
1576 test = not isinstance(remote_cksum, dict)
1577 _ErrorIf(test, self.ENODEFILECHECK, node,
1578 "node hasn't returned file checksum data")
1582 for file_name in file_list:
1583 node_is_mc = ninfo.master_candidate
1584 must_have = (file_name not in master_files) or node_is_mc
1586 test1 = file_name not in remote_cksum
1588 test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1590 test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1591 _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1592 "file '%s' missing", file_name)
1593 _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1594 "file '%s' has wrong checksum", file_name)
1595 # not candidate and this is not a must-have file
1596 _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1597 "file '%s' should not exist on non master"
1598 " candidates (and the file is outdated)", file_name)
1599 # all good, except non-master/non-must have combination
1600 _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1601 "file '%s' should not exist"
1602 " on non master candidates", file_name)
1604 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1606 """Verifies and the node DRBD status.
1608 @type ninfo: L{objects.Node}
1609 @param ninfo: the node to check
1610 @param nresult: the remote results for the node
1611 @param instanceinfo: the dict of instances
1612 @param drbd_helper: the configured DRBD usermode helper
1613 @param drbd_map: the DRBD map as returned by
1614 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1618 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1621 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1622 test = (helper_result == None)
1623 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1624 "no drbd usermode helper returned")
1626 status, payload = helper_result
1628 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1629 "drbd usermode helper check unsuccessful: %s", payload)
1630 test = status and (payload != drbd_helper)
1631 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1632 "wrong drbd usermode helper: %s", payload)
1634 # compute the DRBD minors
1636 for minor, instance in drbd_map[node].items():
1637 test = instance not in instanceinfo
1638 _ErrorIf(test, self.ECLUSTERCFG, None,
1639 "ghost instance '%s' in temporary DRBD map", instance)
1640 # ghost instance should not be running, but otherwise we
1641 # don't give double warnings (both ghost instance and
1642 # unallocated minor in use)
1644 node_drbd[minor] = (instance, False)
1646 instance = instanceinfo[instance]
1647 node_drbd[minor] = (instance.name, instance.admin_up)
1649 # and now check them
1650 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1651 test = not isinstance(used_minors, (tuple, list))
1652 _ErrorIf(test, self.ENODEDRBD, node,
1653 "cannot parse drbd status file: %s", str(used_minors))
1655 # we cannot check drbd status
1658 for minor, (iname, must_exist) in node_drbd.items():
1659 test = minor not in used_minors and must_exist
1660 _ErrorIf(test, self.ENODEDRBD, node,
1661 "drbd minor %d of instance %s is not active", minor, iname)
1662 for minor in used_minors:
1663 test = minor not in node_drbd
1664 _ErrorIf(test, self.ENODEDRBD, node,
1665 "unallocated drbd minor %d is in use", minor)
1667 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1668 """Builds the node OS structures.
1670 @type ninfo: L{objects.Node}
1671 @param ninfo: the node to check
1672 @param nresult: the remote results for the node
1673 @param nimg: the node image object
1677 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1679 remote_os = nresult.get(constants.NV_OSLIST, None)
1680 test = (not isinstance(remote_os, list) or
1681 not compat.all(isinstance(v, list) and len(v) == 7
1682 for v in remote_os))
1684 _ErrorIf(test, self.ENODEOS, node,
1685 "node hasn't returned valid OS data")
1694 for (name, os_path, status, diagnose,
1695 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1697 if name not in os_dict:
1700 # parameters is a list of lists instead of list of tuples due to
1701 # JSON lacking a real tuple type, fix it:
1702 parameters = [tuple(v) for v in parameters]
1703 os_dict[name].append((os_path, status, diagnose,
1704 set(variants), set(parameters), set(api_ver)))
1706 nimg.oslist = os_dict
1708 def _VerifyNodeOS(self, ninfo, nimg, base):
1709 """Verifies the node OS list.
1711 @type ninfo: L{objects.Node}
1712 @param ninfo: the node to check
1713 @param nimg: the node image object
1714 @param base: the 'template' node we match against (e.g. from the master)
1718 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1720 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1722 for os_name, os_data in nimg.oslist.items():
1723 assert os_data, "Empty OS status for OS %s?!" % os_name
1724 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1725 _ErrorIf(not f_status, self.ENODEOS, node,
1726 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1727 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1728 "OS '%s' has multiple entries (first one shadows the rest): %s",
1729 os_name, utils.CommaJoin([v[0] for v in os_data]))
1730 # this will catched in backend too
1731 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1732 and not f_var, self.ENODEOS, node,
1733 "OS %s with API at least %d does not declare any variant",
1734 os_name, constants.OS_API_V15)
1735 # comparisons with the 'base' image
1736 test = os_name not in base.oslist
1737 _ErrorIf(test, self.ENODEOS, node,
1738 "Extra OS %s not present on reference node (%s)",
1742 assert base.oslist[os_name], "Base node has empty OS status?"
1743 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1745 # base OS is invalid, skipping
1747 for kind, a, b in [("API version", f_api, b_api),
1748 ("variants list", f_var, b_var),
1749 ("parameters", f_param, b_param)]:
1750 _ErrorIf(a != b, self.ENODEOS, node,
1751 "OS %s %s differs from reference node %s: %s vs. %s",
1752 kind, os_name, base.name,
1753 utils.CommaJoin(a), utils.CommaJoin(b))
1755 # check any missing OSes
1756 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1757 _ErrorIf(missing, self.ENODEOS, node,
1758 "OSes present on reference node %s but missing on this node: %s",
1759 base.name, utils.CommaJoin(missing))
1761 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1762 """Verifies and updates the node volume data.
1764 This function will update a L{NodeImage}'s internal structures
1765 with data from the remote call.
1767 @type ninfo: L{objects.Node}
1768 @param ninfo: the node to check
1769 @param nresult: the remote results for the node
1770 @param nimg: the node image object
1771 @param vg_name: the configured VG name
1775 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1777 nimg.lvm_fail = True
1778 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1781 elif isinstance(lvdata, basestring):
1782 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1783 utils.SafeEncode(lvdata))
1784 elif not isinstance(lvdata, dict):
1785 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1787 nimg.volumes = lvdata
1788 nimg.lvm_fail = False
1790 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1791 """Verifies and updates the node instance list.
1793 If the listing was successful, then updates this node's instance
1794 list. Otherwise, it marks the RPC call as failed for the instance
1797 @type ninfo: L{objects.Node}
1798 @param ninfo: the node to check
1799 @param nresult: the remote results for the node
1800 @param nimg: the node image object
1803 idata = nresult.get(constants.NV_INSTANCELIST, None)
1804 test = not isinstance(idata, list)
1805 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1806 " (instancelist): %s", utils.SafeEncode(str(idata)))
1808 nimg.hyp_fail = True
1810 nimg.instances = idata
1812 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1813 """Verifies and computes a node information map
1815 @type ninfo: L{objects.Node}
1816 @param ninfo: the node to check
1817 @param nresult: the remote results for the node
1818 @param nimg: the node image object
1819 @param vg_name: the configured VG name
1823 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1825 # try to read free memory (from the hypervisor)
1826 hv_info = nresult.get(constants.NV_HVINFO, None)
1827 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1828 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1831 nimg.mfree = int(hv_info["memory_free"])
1832 except (ValueError, TypeError):
1833 _ErrorIf(True, self.ENODERPC, node,
1834 "node returned invalid nodeinfo, check hypervisor")
1836 # FIXME: devise a free space model for file based instances as well
1837 if vg_name is not None:
1838 test = (constants.NV_VGLIST not in nresult or
1839 vg_name not in nresult[constants.NV_VGLIST])
1840 _ErrorIf(test, self.ENODELVM, node,
1841 "node didn't return data for the volume group '%s'"
1842 " - it is either missing or broken", vg_name)
1845 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1846 except (ValueError, TypeError):
1847 _ErrorIf(True, self.ENODERPC, node,
1848 "node returned invalid LVM info, check LVM status")
1850 def BuildHooksEnv(self):
1853 Cluster-Verify hooks just ran in the post phase and their failure makes
1854 the output be logged in the verify output and the verification to fail.
1857 all_nodes = self.cfg.GetNodeList()
1859 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1861 for node in self.cfg.GetAllNodesInfo().values():
1862 env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1864 return env, [], all_nodes
1866 def Exec(self, feedback_fn):
1867 """Verify integrity of cluster, performing various test on nodes.
1871 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1872 verbose = self.op.verbose
1873 self._feedback_fn = feedback_fn
1874 feedback_fn("* Verifying global settings")
1875 for msg in self.cfg.VerifyConfig():
1876 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1878 # Check the cluster certificates
1879 for cert_filename in constants.ALL_CERT_FILES:
1880 (errcode, msg) = _VerifyCertificate(cert_filename)
1881 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
1883 vg_name = self.cfg.GetVGName()
1884 drbd_helper = self.cfg.GetDRBDHelper()
1885 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1886 cluster = self.cfg.GetClusterInfo()
1887 nodelist = utils.NiceSort(self.cfg.GetNodeList())
1888 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1889 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1890 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1891 for iname in instancelist)
1892 i_non_redundant = [] # Non redundant instances
1893 i_non_a_balanced = [] # Non auto-balanced instances
1894 n_offline = 0 # Count of offline nodes
1895 n_drained = 0 # Count of nodes being drained
1896 node_vol_should = {}
1898 # FIXME: verify OS list
1899 # do local checksums
1900 master_files = [constants.CLUSTER_CONF_FILE]
1901 master_node = self.master_node = self.cfg.GetMasterNode()
1902 master_ip = self.cfg.GetMasterIP()
1904 file_names = ssconf.SimpleStore().GetFileList()
1905 file_names.extend(constants.ALL_CERT_FILES)
1906 file_names.extend(master_files)
1907 if cluster.modify_etc_hosts:
1908 file_names.append(constants.ETC_HOSTS)
1910 local_checksums = utils.FingerprintFiles(file_names)
1912 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1913 node_verify_param = {
1914 constants.NV_FILELIST: file_names,
1915 constants.NV_NODELIST: [node.name for node in nodeinfo
1916 if not node.offline],
1917 constants.NV_HYPERVISOR: hypervisors,
1918 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1919 node.secondary_ip) for node in nodeinfo
1920 if not node.offline],
1921 constants.NV_INSTANCELIST: hypervisors,
1922 constants.NV_VERSION: None,
1923 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1924 constants.NV_NODESETUP: None,
1925 constants.NV_TIME: None,
1926 constants.NV_MASTERIP: (master_node, master_ip),
1927 constants.NV_OSLIST: None,
1930 if vg_name is not None:
1931 node_verify_param[constants.NV_VGLIST] = None
1932 node_verify_param[constants.NV_LVLIST] = vg_name
1933 node_verify_param[constants.NV_PVLIST] = [vg_name]
1934 node_verify_param[constants.NV_DRBDLIST] = None
1937 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
1939 # Build our expected cluster state
1940 node_image = dict((node.name, self.NodeImage(offline=node.offline,
1942 for node in nodeinfo)
1944 for instance in instancelist:
1945 inst_config = instanceinfo[instance]
1947 for nname in inst_config.all_nodes:
1948 if nname not in node_image:
1950 gnode = self.NodeImage(name=nname)
1952 node_image[nname] = gnode
1954 inst_config.MapLVsByNode(node_vol_should)
1956 pnode = inst_config.primary_node
1957 node_image[pnode].pinst.append(instance)
1959 for snode in inst_config.secondary_nodes:
1960 nimg = node_image[snode]
1961 nimg.sinst.append(instance)
1962 if pnode not in nimg.sbp:
1963 nimg.sbp[pnode] = []
1964 nimg.sbp[pnode].append(instance)
1966 # At this point, we have the in-memory data structures complete,
1967 # except for the runtime information, which we'll gather next
1969 # Due to the way our RPC system works, exact response times cannot be
1970 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
1971 # time before and after executing the request, we can at least have a time
1973 nvinfo_starttime = time.time()
1974 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1975 self.cfg.GetClusterName())
1976 nvinfo_endtime = time.time()
1978 all_drbd_map = self.cfg.ComputeDRBDMap()
1980 feedback_fn("* Verifying node status")
1984 for node_i in nodeinfo:
1986 nimg = node_image[node]
1990 feedback_fn("* Skipping offline node %s" % (node,))
1994 if node == master_node:
1996 elif node_i.master_candidate:
1997 ntype = "master candidate"
1998 elif node_i.drained:
2004 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2006 msg = all_nvinfo[node].fail_msg
2007 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2009 nimg.rpc_fail = True
2012 nresult = all_nvinfo[node].payload
2014 nimg.call_ok = self._VerifyNode(node_i, nresult)
2015 self._VerifyNodeNetwork(node_i, nresult)
2016 self._VerifyNodeLVM(node_i, nresult, vg_name)
2017 self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2019 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2021 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2023 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2024 self._UpdateNodeInstances(node_i, nresult, nimg)
2025 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2026 self._UpdateNodeOS(node_i, nresult, nimg)
2027 if not nimg.os_fail:
2028 if refos_img is None:
2030 self._VerifyNodeOS(node_i, nimg, refos_img)
2032 feedback_fn("* Verifying instance status")
2033 for instance in instancelist:
2035 feedback_fn("* Verifying instance %s" % instance)
2036 inst_config = instanceinfo[instance]
2037 self._VerifyInstance(instance, inst_config, node_image)
2038 inst_nodes_offline = []
2040 pnode = inst_config.primary_node
2041 pnode_img = node_image[pnode]
2042 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2043 self.ENODERPC, pnode, "instance %s, connection to"
2044 " primary node failed", instance)
2046 if pnode_img.offline:
2047 inst_nodes_offline.append(pnode)
2049 # If the instance is non-redundant we cannot survive losing its primary
2050 # node, so we are not N+1 compliant. On the other hand we have no disk
2051 # templates with more than one secondary so that situation is not well
2053 # FIXME: does not support file-backed instances
2054 if not inst_config.secondary_nodes:
2055 i_non_redundant.append(instance)
2056 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2057 instance, "instance has multiple secondary nodes: %s",
2058 utils.CommaJoin(inst_config.secondary_nodes),
2059 code=self.ETYPE_WARNING)
2061 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2062 i_non_a_balanced.append(instance)
2064 for snode in inst_config.secondary_nodes:
2065 s_img = node_image[snode]
2066 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2067 "instance %s, connection to secondary node failed", instance)
2070 inst_nodes_offline.append(snode)
2072 # warn that the instance lives on offline nodes
2073 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2074 "instance lives on offline node(s) %s",
2075 utils.CommaJoin(inst_nodes_offline))
2076 # ... or ghost nodes
2077 for node in inst_config.all_nodes:
2078 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2079 "instance lives on ghost node %s", node)
2081 feedback_fn("* Verifying orphan volumes")
2082 reserved = utils.FieldSet(*cluster.reserved_lvs)
2083 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2085 feedback_fn("* Verifying orphan instances")
2086 self._VerifyOrphanInstances(instancelist, node_image)
2088 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2089 feedback_fn("* Verifying N+1 Memory redundancy")
2090 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2092 feedback_fn("* Other Notes")
2094 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2095 % len(i_non_redundant))
2097 if i_non_a_balanced:
2098 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2099 % len(i_non_a_balanced))
2102 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2105 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2109 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2110 """Analyze the post-hooks' result
2112 This method analyses the hook result, handles it, and sends some
2113 nicely-formatted feedback back to the user.
2115 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2116 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2117 @param hooks_results: the results of the multi-node hooks rpc call
2118 @param feedback_fn: function used send feedback back to the caller
2119 @param lu_result: previous Exec result
2120 @return: the new Exec result, based on the previous result
2124 # We only really run POST phase hooks, and are only interested in
2126 if phase == constants.HOOKS_PHASE_POST:
2127 # Used to change hooks' output to proper indentation
2128 indent_re = re.compile('^', re.M)
2129 feedback_fn("* Hooks Results")
2130 assert hooks_results, "invalid result from hooks"
2132 for node_name in hooks_results:
2133 res = hooks_results[node_name]
2135 test = msg and not res.offline
2136 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2137 "Communication failure in hooks execution: %s", msg)
2138 if res.offline or msg:
2139 # No need to investigate payload if node is offline or gave an error.
2140 # override manually lu_result here as _ErrorIf only
2141 # overrides self.bad
2144 for script, hkr, output in res.payload:
2145 test = hkr == constants.HKR_FAIL
2146 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2147 "Script %s failed, output:", script)
2149 output = indent_re.sub(' ', output)
2150 feedback_fn("%s" % output)
2156 class LUVerifyDisks(NoHooksLU):
2157 """Verifies the cluster disks status.
2162 def ExpandNames(self):
2163 self.needed_locks = {
2164 locking.LEVEL_NODE: locking.ALL_SET,
2165 locking.LEVEL_INSTANCE: locking.ALL_SET,
2167 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2169 def Exec(self, feedback_fn):
2170 """Verify integrity of cluster disks.
2172 @rtype: tuple of three items
2173 @return: a tuple of (dict of node-to-node_error, list of instances
2174 which need activate-disks, dict of instance: (node, volume) for
2178 result = res_nodes, res_instances, res_missing = {}, [], {}
2180 vg_name = self.cfg.GetVGName()
2181 nodes = utils.NiceSort(self.cfg.GetNodeList())
2182 instances = [self.cfg.GetInstanceInfo(name)
2183 for name in self.cfg.GetInstanceList()]
2186 for inst in instances:
2188 if (not inst.admin_up or
2189 inst.disk_template not in constants.DTS_NET_MIRROR):
2191 inst.MapLVsByNode(inst_lvs)
2192 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2193 for node, vol_list in inst_lvs.iteritems():
2194 for vol in vol_list:
2195 nv_dict[(node, vol)] = inst
2200 node_lvs = self.rpc.call_lv_list(nodes, vg_name)
2204 node_res = node_lvs[node]
2205 if node_res.offline:
2207 msg = node_res.fail_msg
2209 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2210 res_nodes[node] = msg
2213 lvs = node_res.payload
2214 for lv_name, (_, _, lv_online) in lvs.items():
2215 inst = nv_dict.pop((node, lv_name), None)
2216 if (not lv_online and inst is not None
2217 and inst.name not in res_instances):
2218 res_instances.append(inst.name)
2220 # any leftover items in nv_dict are missing LVs, let's arrange the
2222 for key, inst in nv_dict.iteritems():
2223 if inst.name not in res_missing:
2224 res_missing[inst.name] = []
2225 res_missing[inst.name].append(key)
2230 class LURepairDiskSizes(NoHooksLU):
2231 """Verifies the cluster disks sizes.
2234 _OP_PARAMS = [("instances", ht.EmptyList, ht.TListOf(ht.TNonEmptyString))]
2237 def ExpandNames(self):
2238 if self.op.instances:
2239 self.wanted_names = []
2240 for name in self.op.instances:
2241 full_name = _ExpandInstanceName(self.cfg, name)
2242 self.wanted_names.append(full_name)
2243 self.needed_locks = {
2244 locking.LEVEL_NODE: [],
2245 locking.LEVEL_INSTANCE: self.wanted_names,
2247 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2249 self.wanted_names = None
2250 self.needed_locks = {
2251 locking.LEVEL_NODE: locking.ALL_SET,
2252 locking.LEVEL_INSTANCE: locking.ALL_SET,
2254 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2256 def DeclareLocks(self, level):
2257 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2258 self._LockInstancesNodes(primary_only=True)
2260 def CheckPrereq(self):
2261 """Check prerequisites.
2263 This only checks the optional instance list against the existing names.
2266 if self.wanted_names is None:
2267 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2269 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2270 in self.wanted_names]
2272 def _EnsureChildSizes(self, disk):
2273 """Ensure children of the disk have the needed disk size.
2275 This is valid mainly for DRBD8 and fixes an issue where the
2276 children have smaller disk size.
2278 @param disk: an L{ganeti.objects.Disk} object
2281 if disk.dev_type == constants.LD_DRBD8:
2282 assert disk.children, "Empty children for DRBD8?"
2283 fchild = disk.children[0]
2284 mismatch = fchild.size < disk.size
2286 self.LogInfo("Child disk has size %d, parent %d, fixing",
2287 fchild.size, disk.size)
2288 fchild.size = disk.size
2290 # and we recurse on this child only, not on the metadev
2291 return self._EnsureChildSizes(fchild) or mismatch
2295 def Exec(self, feedback_fn):
2296 """Verify the size of cluster disks.
2299 # TODO: check child disks too
2300 # TODO: check differences in size between primary/secondary nodes
2302 for instance in self.wanted_instances:
2303 pnode = instance.primary_node
2304 if pnode not in per_node_disks:
2305 per_node_disks[pnode] = []
2306 for idx, disk in enumerate(instance.disks):
2307 per_node_disks[pnode].append((instance, idx, disk))
2310 for node, dskl in per_node_disks.items():
2311 newl = [v[2].Copy() for v in dskl]
2313 self.cfg.SetDiskID(dsk, node)
2314 result = self.rpc.call_blockdev_getsizes(node, newl)
2316 self.LogWarning("Failure in blockdev_getsizes call to node"
2317 " %s, ignoring", node)
2319 if len(result.data) != len(dskl):
2320 self.LogWarning("Invalid result from node %s, ignoring node results",
2323 for ((instance, idx, disk), size) in zip(dskl, result.data):
2325 self.LogWarning("Disk %d of instance %s did not return size"
2326 " information, ignoring", idx, instance.name)
2328 if not isinstance(size, (int, long)):
2329 self.LogWarning("Disk %d of instance %s did not return valid"
2330 " size information, ignoring", idx, instance.name)
2333 if size != disk.size:
2334 self.LogInfo("Disk %d of instance %s has mismatched size,"
2335 " correcting: recorded %d, actual %d", idx,
2336 instance.name, disk.size, size)
2338 self.cfg.Update(instance, feedback_fn)
2339 changed.append((instance.name, idx, size))
2340 if self._EnsureChildSizes(disk):
2341 self.cfg.Update(instance, feedback_fn)
2342 changed.append((instance.name, idx, disk.size))
2346 class LURenameCluster(LogicalUnit):
2347 """Rename the cluster.
2350 HPATH = "cluster-rename"
2351 HTYPE = constants.HTYPE_CLUSTER
2352 _OP_PARAMS = [("name", ht.NoDefault, ht.TNonEmptyString)]
2354 def BuildHooksEnv(self):
2359 "OP_TARGET": self.cfg.GetClusterName(),
2360 "NEW_NAME": self.op.name,
2362 mn = self.cfg.GetMasterNode()
2363 all_nodes = self.cfg.GetNodeList()
2364 return env, [mn], all_nodes
2366 def CheckPrereq(self):
2367 """Verify that the passed name is a valid one.
2370 hostname = netutils.GetHostname(name=self.op.name,
2371 family=self.cfg.GetPrimaryIPFamily())
2373 new_name = hostname.name
2374 self.ip = new_ip = hostname.ip
2375 old_name = self.cfg.GetClusterName()
2376 old_ip = self.cfg.GetMasterIP()
2377 if new_name == old_name and new_ip == old_ip:
2378 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2379 " cluster has changed",
2381 if new_ip != old_ip:
2382 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2383 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2384 " reachable on the network" %
2385 new_ip, errors.ECODE_NOTUNIQUE)
2387 self.op.name = new_name
2389 def Exec(self, feedback_fn):
2390 """Rename the cluster.
2393 clustername = self.op.name
2396 # shutdown the master IP
2397 master = self.cfg.GetMasterNode()
2398 result = self.rpc.call_node_stop_master(master, False)
2399 result.Raise("Could not disable the master role")
2402 cluster = self.cfg.GetClusterInfo()
2403 cluster.cluster_name = clustername
2404 cluster.master_ip = ip
2405 self.cfg.Update(cluster, feedback_fn)
2407 # update the known hosts file
2408 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2409 node_list = self.cfg.GetNodeList()
2411 node_list.remove(master)
2414 result = self.rpc.call_upload_file(node_list,
2415 constants.SSH_KNOWN_HOSTS_FILE)
2416 for to_node, to_result in result.iteritems():
2417 msg = to_result.fail_msg
2419 msg = ("Copy of file %s to node %s failed: %s" %
2420 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
2421 self.proc.LogWarning(msg)
2424 result = self.rpc.call_node_start_master(master, False, False)
2425 msg = result.fail_msg
2427 self.LogWarning("Could not re-enable the master role on"
2428 " the master, please restart manually: %s", msg)
2433 class LUSetClusterParams(LogicalUnit):
2434 """Change the parameters of the cluster.
2437 HPATH = "cluster-modify"
2438 HTYPE = constants.HTYPE_CLUSTER
2440 ("vg_name", None, ht.TMaybeString),
2441 ("enabled_hypervisors", None,
2442 ht.TOr(ht.TAnd(ht.TListOf(ht.TElemOf(constants.HYPER_TYPES)), ht.TTrue),
2444 ("hvparams", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2446 ("beparams", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2448 ("os_hvp", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2450 ("osparams", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2452 ("candidate_pool_size", None, ht.TOr(ht.TStrictPositiveInt, ht.TNone)),
2453 ("uid_pool", None, ht.NoType),
2454 ("add_uids", None, ht.NoType),
2455 ("remove_uids", None, ht.NoType),
2456 ("maintain_node_health", None, ht.TMaybeBool),
2457 ("nicparams", None, ht.TOr(ht.TDict, ht.TNone)),
2458 ("drbd_helper", None, ht.TOr(ht.TString, ht.TNone)),
2459 ("default_iallocator", None, ht.TMaybeString),
2460 ("reserved_lvs", None, ht.TOr(ht.TListOf(ht.TNonEmptyString), ht.TNone)),
2461 ("hidden_os", None, ht.TOr(ht.TListOf(\
2464 ht.TMap(lambda v: v[0], ht.TElemOf(constants.DDMS_VALUES)))),
2466 ("blacklisted_os", None, ht.TOr(ht.TListOf(\
2469 ht.TMap(lambda v: v[0], ht.TElemOf(constants.DDMS_VALUES)))),
2474 def CheckArguments(self):
2478 if self.op.uid_pool:
2479 uidpool.CheckUidPool(self.op.uid_pool)
2481 if self.op.add_uids:
2482 uidpool.CheckUidPool(self.op.add_uids)
2484 if self.op.remove_uids:
2485 uidpool.CheckUidPool(self.op.remove_uids)
2487 def ExpandNames(self):
2488 # FIXME: in the future maybe other cluster params won't require checking on
2489 # all nodes to be modified.
2490 self.needed_locks = {
2491 locking.LEVEL_NODE: locking.ALL_SET,
2493 self.share_locks[locking.LEVEL_NODE] = 1
2495 def BuildHooksEnv(self):
2500 "OP_TARGET": self.cfg.GetClusterName(),
2501 "NEW_VG_NAME": self.op.vg_name,
2503 mn = self.cfg.GetMasterNode()
2504 return env, [mn], [mn]
2506 def CheckPrereq(self):
2507 """Check prerequisites.
2509 This checks whether the given params don't conflict and
2510 if the given volume group is valid.
2513 if self.op.vg_name is not None and not self.op.vg_name:
2514 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2515 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2516 " instances exist", errors.ECODE_INVAL)
2518 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2519 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2520 raise errors.OpPrereqError("Cannot disable drbd helper while"
2521 " drbd-based instances exist",
2524 node_list = self.acquired_locks[locking.LEVEL_NODE]
2526 # if vg_name not None, checks given volume group on all nodes
2528 vglist = self.rpc.call_vg_list(node_list)
2529 for node in node_list:
2530 msg = vglist[node].fail_msg
2532 # ignoring down node
2533 self.LogWarning("Error while gathering data on node %s"
2534 " (ignoring node): %s", node, msg)
2536 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2538 constants.MIN_VG_SIZE)
2540 raise errors.OpPrereqError("Error on node '%s': %s" %
2541 (node, vgstatus), errors.ECODE_ENVIRON)
2543 if self.op.drbd_helper:
2544 # checks given drbd helper on all nodes
2545 helpers = self.rpc.call_drbd_helper(node_list)
2546 for node in node_list:
2547 ninfo = self.cfg.GetNodeInfo(node)
2549 self.LogInfo("Not checking drbd helper on offline node %s", node)
2551 msg = helpers[node].fail_msg
2553 raise errors.OpPrereqError("Error checking drbd helper on node"
2554 " '%s': %s" % (node, msg),
2555 errors.ECODE_ENVIRON)
2556 node_helper = helpers[node].payload
2557 if node_helper != self.op.drbd_helper:
2558 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2559 (node, node_helper), errors.ECODE_ENVIRON)
2561 self.cluster = cluster = self.cfg.GetClusterInfo()
2562 # validate params changes
2563 if self.op.beparams:
2564 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2565 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2567 if self.op.nicparams:
2568 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2569 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2570 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2573 # check all instances for consistency
2574 for instance in self.cfg.GetAllInstancesInfo().values():
2575 for nic_idx, nic in enumerate(instance.nics):
2576 params_copy = copy.deepcopy(nic.nicparams)
2577 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2579 # check parameter syntax
2581 objects.NIC.CheckParameterSyntax(params_filled)
2582 except errors.ConfigurationError, err:
2583 nic_errors.append("Instance %s, nic/%d: %s" %
2584 (instance.name, nic_idx, err))
2586 # if we're moving instances to routed, check that they have an ip
2587 target_mode = params_filled[constants.NIC_MODE]
2588 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2589 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2590 (instance.name, nic_idx))
2592 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2593 "\n".join(nic_errors))
2595 # hypervisor list/parameters
2596 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2597 if self.op.hvparams:
2598 for hv_name, hv_dict in self.op.hvparams.items():
2599 if hv_name not in self.new_hvparams:
2600 self.new_hvparams[hv_name] = hv_dict
2602 self.new_hvparams[hv_name].update(hv_dict)
2604 # os hypervisor parameters
2605 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2607 for os_name, hvs in self.op.os_hvp.items():
2608 if os_name not in self.new_os_hvp:
2609 self.new_os_hvp[os_name] = hvs
2611 for hv_name, hv_dict in hvs.items():
2612 if hv_name not in self.new_os_hvp[os_name]:
2613 self.new_os_hvp[os_name][hv_name] = hv_dict
2615 self.new_os_hvp[os_name][hv_name].update(hv_dict)
2618 self.new_osp = objects.FillDict(cluster.osparams, {})
2619 if self.op.osparams:
2620 for os_name, osp in self.op.osparams.items():
2621 if os_name not in self.new_osp:
2622 self.new_osp[os_name] = {}
2624 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2627 if not self.new_osp[os_name]:
2628 # we removed all parameters
2629 del self.new_osp[os_name]
2631 # check the parameter validity (remote check)
2632 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2633 os_name, self.new_osp[os_name])
2635 # changes to the hypervisor list
2636 if self.op.enabled_hypervisors is not None:
2637 self.hv_list = self.op.enabled_hypervisors
2638 for hv in self.hv_list:
2639 # if the hypervisor doesn't already exist in the cluster
2640 # hvparams, we initialize it to empty, and then (in both
2641 # cases) we make sure to fill the defaults, as we might not
2642 # have a complete defaults list if the hypervisor wasn't
2644 if hv not in new_hvp:
2646 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2647 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2649 self.hv_list = cluster.enabled_hypervisors
2651 if self.op.hvparams or self.op.enabled_hypervisors is not None:
2652 # either the enabled list has changed, or the parameters have, validate
2653 for hv_name, hv_params in self.new_hvparams.items():
2654 if ((self.op.hvparams and hv_name in self.op.hvparams) or
2655 (self.op.enabled_hypervisors and
2656 hv_name in self.op.enabled_hypervisors)):
2657 # either this is a new hypervisor, or its parameters have changed
2658 hv_class = hypervisor.GetHypervisor(hv_name)
2659 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2660 hv_class.CheckParameterSyntax(hv_params)
2661 _CheckHVParams(self, node_list, hv_name, hv_params)
2664 # no need to check any newly-enabled hypervisors, since the
2665 # defaults have already been checked in the above code-block
2666 for os_name, os_hvp in self.new_os_hvp.items():
2667 for hv_name, hv_params in os_hvp.items():
2668 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2669 # we need to fill in the new os_hvp on top of the actual hv_p
2670 cluster_defaults = self.new_hvparams.get(hv_name, {})
2671 new_osp = objects.FillDict(cluster_defaults, hv_params)
2672 hv_class = hypervisor.GetHypervisor(hv_name)
2673 hv_class.CheckParameterSyntax(new_osp)
2674 _CheckHVParams(self, node_list, hv_name, new_osp)
2676 if self.op.default_iallocator:
2677 alloc_script = utils.FindFile(self.op.default_iallocator,
2678 constants.IALLOCATOR_SEARCH_PATH,
2680 if alloc_script is None:
2681 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2682 " specified" % self.op.default_iallocator,
2685 def Exec(self, feedback_fn):
2686 """Change the parameters of the cluster.
2689 if self.op.vg_name is not None:
2690 new_volume = self.op.vg_name
2693 if new_volume != self.cfg.GetVGName():
2694 self.cfg.SetVGName(new_volume)
2696 feedback_fn("Cluster LVM configuration already in desired"
2697 " state, not changing")
2698 if self.op.drbd_helper is not None:
2699 new_helper = self.op.drbd_helper
2702 if new_helper != self.cfg.GetDRBDHelper():
2703 self.cfg.SetDRBDHelper(new_helper)
2705 feedback_fn("Cluster DRBD helper already in desired state,"
2707 if self.op.hvparams:
2708 self.cluster.hvparams = self.new_hvparams
2710 self.cluster.os_hvp = self.new_os_hvp
2711 if self.op.enabled_hypervisors is not None:
2712 self.cluster.hvparams = self.new_hvparams
2713 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2714 if self.op.beparams:
2715 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2716 if self.op.nicparams:
2717 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2718 if self.op.osparams:
2719 self.cluster.osparams = self.new_osp
2721 if self.op.candidate_pool_size is not None:
2722 self.cluster.candidate_pool_size = self.op.candidate_pool_size
2723 # we need to update the pool size here, otherwise the save will fail
2724 _AdjustCandidatePool(self, [])
2726 if self.op.maintain_node_health is not None:
2727 self.cluster.maintain_node_health = self.op.maintain_node_health
2729 if self.op.add_uids is not None:
2730 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2732 if self.op.remove_uids is not None:
2733 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2735 if self.op.uid_pool is not None:
2736 self.cluster.uid_pool = self.op.uid_pool
2738 if self.op.default_iallocator is not None:
2739 self.cluster.default_iallocator = self.op.default_iallocator
2741 if self.op.reserved_lvs is not None:
2742 self.cluster.reserved_lvs = self.op.reserved_lvs
2744 def helper_os(aname, mods, desc):
2746 lst = getattr(self.cluster, aname)
2747 for key, val in mods:
2748 if key == constants.DDM_ADD:
2750 feedback_fn("OS %s already in %s, ignoring", val, desc)
2753 elif key == constants.DDM_REMOVE:
2757 feedback_fn("OS %s not found in %s, ignoring", val, desc)
2759 raise errors.ProgrammerError("Invalid modification '%s'" % key)
2761 if self.op.hidden_os:
2762 helper_os("hidden_os", self.op.hidden_os, "hidden")
2764 if self.op.blacklisted_os:
2765 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
2767 self.cfg.Update(self.cluster, feedback_fn)
2770 def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2771 """Distribute additional files which are part of the cluster configuration.
2773 ConfigWriter takes care of distributing the config and ssconf files, but
2774 there are more files which should be distributed to all nodes. This function
2775 makes sure those are copied.
2777 @param lu: calling logical unit
2778 @param additional_nodes: list of nodes not in the config to distribute to
2781 # 1. Gather target nodes
2782 myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2783 dist_nodes = lu.cfg.GetOnlineNodeList()
2784 if additional_nodes is not None:
2785 dist_nodes.extend(additional_nodes)
2786 if myself.name in dist_nodes:
2787 dist_nodes.remove(myself.name)
2789 # 2. Gather files to distribute
2790 dist_files = set([constants.ETC_HOSTS,
2791 constants.SSH_KNOWN_HOSTS_FILE,
2792 constants.RAPI_CERT_FILE,
2793 constants.RAPI_USERS_FILE,
2794 constants.CONFD_HMAC_KEY,
2795 constants.CLUSTER_DOMAIN_SECRET_FILE,
2798 enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2799 for hv_name in enabled_hypervisors:
2800 hv_class = hypervisor.GetHypervisor(hv_name)
2801 dist_files.update(hv_class.GetAncillaryFiles())
2803 # 3. Perform the files upload
2804 for fname in dist_files:
2805 if os.path.exists(fname):
2806 result = lu.rpc.call_upload_file(dist_nodes, fname)
2807 for to_node, to_result in result.items():
2808 msg = to_result.fail_msg
2810 msg = ("Copy of file %s to node %s failed: %s" %
2811 (fname, to_node, msg))
2812 lu.proc.LogWarning(msg)
2815 class LURedistributeConfig(NoHooksLU):
2816 """Force the redistribution of cluster configuration.
2818 This is a very simple LU.
2823 def ExpandNames(self):
2824 self.needed_locks = {
2825 locking.LEVEL_NODE: locking.ALL_SET,
2827 self.share_locks[locking.LEVEL_NODE] = 1
2829 def Exec(self, feedback_fn):
2830 """Redistribute the configuration.
2833 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2834 _RedistributeAncillaryFiles(self)
2837 def _WaitForSync(lu, instance, disks=None, oneshot=False):
2838 """Sleep and poll for an instance's disk to sync.
2841 if not instance.disks or disks is not None and not disks:
2844 disks = _ExpandCheckDisks(instance, disks)
2847 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2849 node = instance.primary_node
2852 lu.cfg.SetDiskID(dev, node)
2854 # TODO: Convert to utils.Retry
2857 degr_retries = 10 # in seconds, as we sleep 1 second each time
2861 cumul_degraded = False
2862 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
2863 msg = rstats.fail_msg
2865 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2868 raise errors.RemoteError("Can't contact node %s for mirror data,"
2869 " aborting." % node)
2872 rstats = rstats.payload
2874 for i, mstat in enumerate(rstats):
2876 lu.LogWarning("Can't compute data for node %s/%s",
2877 node, disks[i].iv_name)
2880 cumul_degraded = (cumul_degraded or
2881 (mstat.is_degraded and mstat.sync_percent is None))
2882 if mstat.sync_percent is not None:
2884 if mstat.estimated_time is not None:
2885 rem_time = ("%s remaining (estimated)" %
2886 utils.FormatSeconds(mstat.estimated_time))
2887 max_time = mstat.estimated_time
2889 rem_time = "no time estimate"
2890 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2891 (disks[i].iv_name, mstat.sync_percent, rem_time))
2893 # if we're done but degraded, let's do a few small retries, to
2894 # make sure we see a stable and not transient situation; therefore
2895 # we force restart of the loop
2896 if (done or oneshot) and cumul_degraded and degr_retries > 0:
2897 logging.info("Degraded disks found, %d retries left", degr_retries)
2905 time.sleep(min(60, max_time))
2908 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2909 return not cumul_degraded
2912 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2913 """Check that mirrors are not degraded.
2915 The ldisk parameter, if True, will change the test from the
2916 is_degraded attribute (which represents overall non-ok status for
2917 the device(s)) to the ldisk (representing the local storage status).
2920 lu.cfg.SetDiskID(dev, node)
2924 if on_primary or dev.AssembleOnSecondary():
2925 rstats = lu.rpc.call_blockdev_find(node, dev)
2926 msg = rstats.fail_msg
2928 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2930 elif not rstats.payload:
2931 lu.LogWarning("Can't find disk on node %s", node)
2935 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2937 result = result and not rstats.payload.is_degraded
2940 for child in dev.children:
2941 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2946 class LUDiagnoseOS(NoHooksLU):
2947 """Logical unit for OS diagnose/query.
2952 ("names", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
2956 _BLK = "blacklisted"
2958 _FIELDS_STATIC = utils.FieldSet()
2959 _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
2960 "parameters", "api_versions", _HID, _BLK)
2962 def CheckArguments(self):
2964 raise errors.OpPrereqError("Selective OS query not supported",
2967 _CheckOutputFields(static=self._FIELDS_STATIC,
2968 dynamic=self._FIELDS_DYNAMIC,
2969 selected=self.op.output_fields)
2971 def ExpandNames(self):
2972 # Lock all nodes, in shared mode
2973 # Temporary removal of locks, should be reverted later
2974 # TODO: reintroduce locks when they are lighter-weight
2975 self.needed_locks = {}
2976 #self.share_locks[locking.LEVEL_NODE] = 1
2977 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2980 def _DiagnoseByOS(rlist):
2981 """Remaps a per-node return list into an a per-os per-node dictionary
2983 @param rlist: a map with node names as keys and OS objects as values
2986 @return: a dictionary with osnames as keys and as value another
2987 map, with nodes as keys and tuples of (path, status, diagnose,
2988 variants, parameters, api_versions) as values, eg::
2990 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
2991 (/srv/..., False, "invalid api")],
2992 "node2": [(/srv/..., True, "", [], [])]}
2997 # we build here the list of nodes that didn't fail the RPC (at RPC
2998 # level), so that nodes with a non-responding node daemon don't
2999 # make all OSes invalid
3000 good_nodes = [node_name for node_name in rlist
3001 if not rlist[node_name].fail_msg]
3002 for node_name, nr in rlist.items():
3003 if nr.fail_msg or not nr.payload:
3005 for (name, path, status, diagnose, variants,
3006 params, api_versions) in nr.payload:
3007 if name not in all_os:
3008 # build a list of nodes for this os containing empty lists
3009 # for each node in node_list
3011 for nname in good_nodes:
3012 all_os[name][nname] = []
3013 # convert params from [name, help] to (name, help)
3014 params = [tuple(v) for v in params]
3015 all_os[name][node_name].append((path, status, diagnose,
3016 variants, params, api_versions))
3019 def Exec(self, feedback_fn):
3020 """Compute the list of OSes.
3023 valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
3024 node_data = self.rpc.call_os_diagnose(valid_nodes)
3025 pol = self._DiagnoseByOS(node_data)
3027 cluster = self.cfg.GetClusterInfo()
3029 for os_name in utils.NiceSort(pol.keys()):
3030 os_data = pol[os_name]
3033 (variants, params, api_versions) = null_state = (set(), set(), set())
3034 for idx, osl in enumerate(os_data.values()):
3035 valid = bool(valid and osl and osl[0][1])
3037 (variants, params, api_versions) = null_state
3039 node_variants, node_params, node_api = osl[0][3:6]
3040 if idx == 0: # first entry
3041 variants = set(node_variants)
3042 params = set(node_params)
3043 api_versions = set(node_api)
3044 else: # keep consistency
3045 variants.intersection_update(node_variants)
3046 params.intersection_update(node_params)
3047 api_versions.intersection_update(node_api)
3049 is_hid = os_name in cluster.hidden_os
3050 is_blk = os_name in cluster.blacklisted_os
3051 if ((self._HID not in self.op.output_fields and is_hid) or
3052 (self._BLK not in self.op.output_fields and is_blk) or
3053 (self._VLD not in self.op.output_fields and not valid)):
3056 for field in self.op.output_fields:
3059 elif field == self._VLD:
3061 elif field == "node_status":
3062 # this is just a copy of the dict
3064 for node_name, nos_list in os_data.items():
3065 val[node_name] = nos_list
3066 elif field == "variants":
3067 val = utils.NiceSort(list(variants))
3068 elif field == "parameters":
3070 elif field == "api_versions":
3071 val = list(api_versions)
3072 elif field == self._HID:
3074 elif field == self._BLK:
3077 raise errors.ParameterError(field)
3084 class LURemoveNode(LogicalUnit):
3085 """Logical unit for removing a node.
3088 HPATH = "node-remove"
3089 HTYPE = constants.HTYPE_NODE
3094 def BuildHooksEnv(self):
3097 This doesn't run on the target node in the pre phase as a failed
3098 node would then be impossible to remove.
3102 "OP_TARGET": self.op.node_name,
3103 "NODE_NAME": self.op.node_name,
3105 all_nodes = self.cfg.GetNodeList()
3107 all_nodes.remove(self.op.node_name)
3109 logging.warning("Node %s which is about to be removed not found"
3110 " in the all nodes list", self.op.node_name)
3111 return env, all_nodes, all_nodes
3113 def CheckPrereq(self):
3114 """Check prerequisites.
3117 - the node exists in the configuration
3118 - it does not have primary or secondary instances
3119 - it's not the master
3121 Any errors are signaled by raising errors.OpPrereqError.
3124 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3125 node = self.cfg.GetNodeInfo(self.op.node_name)
3126 assert node is not None
3128 instance_list = self.cfg.GetInstanceList()
3130 masternode = self.cfg.GetMasterNode()
3131 if node.name == masternode:
3132 raise errors.OpPrereqError("Node is the master node,"
3133 " you need to failover first.",
3136 for instance_name in instance_list:
3137 instance = self.cfg.GetInstanceInfo(instance_name)
3138 if node.name in instance.all_nodes:
3139 raise errors.OpPrereqError("Instance %s is still running on the node,"
3140 " please remove first." % instance_name,
3142 self.op.node_name = node.name
3145 def Exec(self, feedback_fn):
3146 """Removes the node from the cluster.
3150 logging.info("Stopping the node daemon and removing configs from node %s",
3153 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3155 # Promote nodes to master candidate as needed
3156 _AdjustCandidatePool(self, exceptions=[node.name])
3157 self.context.RemoveNode(node.name)
3159 # Run post hooks on the node before it's removed
3160 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3162 hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3164 # pylint: disable-msg=W0702
3165 self.LogWarning("Errors occurred running hooks on %s" % node.name)
3167 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3168 msg = result.fail_msg
3170 self.LogWarning("Errors encountered on the remote node while leaving"
3171 " the cluster: %s", msg)
3173 # Remove node from our /etc/hosts
3174 if self.cfg.GetClusterInfo().modify_etc_hosts:
3175 master_node = self.cfg.GetMasterNode()
3176 result = self.rpc.call_etc_hosts_modify(master_node,
3177 constants.ETC_HOSTS_REMOVE,
3179 result.Raise("Can't update hosts file with new host data")
3180 _RedistributeAncillaryFiles(self)
3183 class LUQueryNodes(NoHooksLU):
3184 """Logical unit for querying nodes.
3187 # pylint: disable-msg=W0142
3190 ("names", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
3191 ("use_locking", False, ht.TBool),
3195 _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
3196 "master_candidate", "offline", "drained"]
3198 _FIELDS_DYNAMIC = utils.FieldSet(
3200 "mtotal", "mnode", "mfree",
3202 "ctotal", "cnodes", "csockets",
3205 _FIELDS_STATIC = utils.FieldSet(*[
3206 "pinst_cnt", "sinst_cnt",
3207 "pinst_list", "sinst_list",
3208 "pip", "sip", "tags",
3210 "role"] + _SIMPLE_FIELDS
3213 def CheckArguments(self):
3214 _CheckOutputFields(static=self._FIELDS_STATIC,
3215 dynamic=self._FIELDS_DYNAMIC,
3216 selected=self.op.output_fields)
3218 def ExpandNames(self):
3219 self.needed_locks = {}
3220 self.share_locks[locking.LEVEL_NODE] = 1
3223 self.wanted = _GetWantedNodes(self, self.op.names)
3225 self.wanted = locking.ALL_SET
3227 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3228 self.do_locking = self.do_node_query and self.op.use_locking
3230 # if we don't request only static fields, we need to lock the nodes
3231 self.needed_locks[locking.LEVEL_NODE] = self.wanted
3233 def Exec(self, feedback_fn):
3234 """Computes the list of nodes and their attributes.
3237 all_info = self.cfg.GetAllNodesInfo()
3239 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3240 elif self.wanted != locking.ALL_SET:
3241 nodenames = self.wanted
3242 missing = set(nodenames).difference(all_info.keys())
3244 raise errors.OpExecError(
3245 "Some nodes were removed before retrieving their data: %s" % missing)
3247 nodenames = all_info.keys()
3249 nodenames = utils.NiceSort(nodenames)
3250 nodelist = [all_info[name] for name in nodenames]
3252 # begin data gathering
3254 if self.do_node_query:
3256 node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3257 self.cfg.GetHypervisorType())
3258 for name in nodenames:
3259 nodeinfo = node_data[name]
3260 if not nodeinfo.fail_msg and nodeinfo.payload:
3261 nodeinfo = nodeinfo.payload
3262 fn = utils.TryConvert
3264 "mtotal": fn(int, nodeinfo.get('memory_total', None)),
3265 "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
3266 "mfree": fn(int, nodeinfo.get('memory_free', None)),
3267 "dtotal": fn(int, nodeinfo.get('vg_size', None)),
3268 "dfree": fn(int, nodeinfo.get('vg_free', None)),
3269 "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
3270 "bootid": nodeinfo.get('bootid', None),
3271 "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
3272 "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
3275 live_data[name] = {}
3277 live_data = dict.fromkeys(nodenames, {})
3279 node_to_primary = dict([(name, set()) for name in nodenames])
3280 node_to_secondary = dict([(name, set()) for name in nodenames])
3282 inst_fields = frozenset(("pinst_cnt", "pinst_list",
3283 "sinst_cnt", "sinst_list"))
3284 if inst_fields & frozenset(self.op.output_fields):
3285 inst_data = self.cfg.GetAllInstancesInfo()
3287 for inst in inst_data.values():
3288 if inst.primary_node in node_to_primary:
3289 node_to_primary[inst.primary_node].add(inst.name)
3290 for secnode in inst.secondary_nodes:
3291 if secnode in node_to_secondary:
3292 node_to_secondary[secnode].add(inst.name)
3294 master_node = self.cfg.GetMasterNode()
3296 # end data gathering
3299 for node in nodelist:
3301 for field in self.op.output_fields:
3302 if field in self._SIMPLE_FIELDS:
3303 val = getattr(node, field)
3304 elif field == "pinst_list":
3305 val = list(node_to_primary[node.name])
3306 elif field == "sinst_list":
3307 val = list(node_to_secondary[node.name])
3308 elif field == "pinst_cnt":
3309 val = len(node_to_primary[node.name])
3310 elif field == "sinst_cnt":
3311 val = len(node_to_secondary[node.name])
3312 elif field == "pip":
3313 val = node.primary_ip
3314 elif field == "sip":
3315 val = node.secondary_ip
3316 elif field == "tags":
3317 val = list(node.GetTags())
3318 elif field == "master":
3319 val = node.name == master_node
3320 elif self._FIELDS_DYNAMIC.Matches(field):
3321 val = live_data[node.name].get(field, None)
3322 elif field == "role":
3323 if node.name == master_node:
3325 elif node.master_candidate:
3334 raise errors.ParameterError(field)
3335 node_output.append(val)
3336 output.append(node_output)
3341 class LUQueryNodeVolumes(NoHooksLU):
3342 """Logical unit for getting volumes on node(s).
3346 ("nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
3347 ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
3350 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3351 _FIELDS_STATIC = utils.FieldSet("node")
3353 def CheckArguments(self):
3354 _CheckOutputFields(static=self._FIELDS_STATIC,
3355 dynamic=self._FIELDS_DYNAMIC,
3356 selected=self.op.output_fields)
3358 def ExpandNames(self):
3359 self.needed_locks = {}
3360 self.share_locks[locking.LEVEL_NODE] = 1
3361 if not self.op.nodes:
3362 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3364 self.needed_locks[locking.LEVEL_NODE] = \
3365 _GetWantedNodes(self, self.op.nodes)
3367 def Exec(self, feedback_fn):
3368 """Computes the list of nodes and their attributes.
3371 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3372 volumes = self.rpc.call_node_volumes(nodenames)
3374 ilist = [self.cfg.GetInstanceInfo(iname) for iname
3375 in self.cfg.GetInstanceList()]
3377 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3380 for node in nodenames:
3381 nresult = volumes[node]
3384 msg = nresult.fail_msg
3386 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3389 node_vols = nresult.payload[:]
3390 node_vols.sort(key=lambda vol: vol['dev'])
3392 for vol in node_vols:
3394 for field in self.op.output_fields:
3397 elif field == "phys":
3401 elif field == "name":
3403 elif field == "size":
3404 val = int(float(vol['size']))
3405 elif field == "instance":
3407 if node not in lv_by_node[inst]:
3409 if vol['name'] in lv_by_node[inst][node]:
3415 raise errors.ParameterError(field)
3416 node_output.append(str(val))
3418 output.append(node_output)
3423 class LUQueryNodeStorage(NoHooksLU):
3424 """Logical unit for getting information on storage units on node(s).
3427 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3429 ("nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
3430 ("storage_type", ht.NoDefault, _CheckStorageType),
3431 ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
3432 ("name", None, ht.TMaybeString),
3436 def CheckArguments(self):
3437 _CheckOutputFields(static=self._FIELDS_STATIC,
3438 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3439 selected=self.op.output_fields)
3441 def ExpandNames(self):
3442 self.needed_locks = {}
3443 self.share_locks[locking.LEVEL_NODE] = 1
3446 self.needed_locks[locking.LEVEL_NODE] = \
3447 _GetWantedNodes(self, self.op.nodes)
3449 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3451 def Exec(self, feedback_fn):
3452 """Computes the list of nodes and their attributes.
3455 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3457 # Always get name to sort by
3458 if constants.SF_NAME in self.op.output_fields:
3459 fields = self.op.output_fields[:]
3461 fields = [constants.SF_NAME] + self.op.output_fields
3463 # Never ask for node or type as it's only known to the LU
3464 for extra in [constants.SF_NODE, constants.SF_TYPE]:
3465 while extra in fields:
3466 fields.remove(extra)
3468 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3469 name_idx = field_idx[constants.SF_NAME]
3471 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3472 data = self.rpc.call_storage_list(self.nodes,
3473 self.op.storage_type, st_args,
3474 self.op.name, fields)
3478 for node in utils.NiceSort(self.nodes):
3479 nresult = data[node]
3483 msg = nresult.fail_msg
3485 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3488 rows = dict([(row[name_idx], row) for row in nresult.payload])
3490 for name in utils.NiceSort(rows.keys()):
3495 for field in self.op.output_fields:
3496 if field == constants.SF_NODE:
3498 elif field == constants.SF_TYPE:
3499 val = self.op.storage_type
3500 elif field in field_idx:
3501 val = row[field_idx[field]]
3503 raise errors.ParameterError(field)
3512 class LUModifyNodeStorage(NoHooksLU):
3513 """Logical unit for modifying a storage volume on a node.
3518 ("storage_type", ht.NoDefault, _CheckStorageType),
3519 ("name", ht.NoDefault, ht.TNonEmptyString),
3520 ("changes", ht.NoDefault, ht.TDict),
3524 def CheckArguments(self):
3525 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3527 storage_type = self.op.storage_type
3530 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3532 raise errors.OpPrereqError("Storage units of type '%s' can not be"
3533 " modified" % storage_type,
3536 diff = set(self.op.changes.keys()) - modifiable
3538 raise errors.OpPrereqError("The following fields can not be modified for"
3539 " storage units of type '%s': %r" %
3540 (storage_type, list(diff)),
3543 def ExpandNames(self):
3544 self.needed_locks = {
3545 locking.LEVEL_NODE: self.op.node_name,
3548 def Exec(self, feedback_fn):
3549 """Computes the list of nodes and their attributes.
3552 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3553 result = self.rpc.call_storage_modify(self.op.node_name,
3554 self.op.storage_type, st_args,
3555 self.op.name, self.op.changes)
3556 result.Raise("Failed to modify storage unit '%s' on %s" %
3557 (self.op.name, self.op.node_name))
3560 class LUAddNode(LogicalUnit):
3561 """Logical unit for adding node to the cluster.
3565 HTYPE = constants.HTYPE_NODE
3568 ("primary_ip", None, ht.NoType),
3569 ("secondary_ip", None, ht.TMaybeString),
3570 ("readd", False, ht.TBool),
3571 ("nodegroup", None, ht.TMaybeString)
3574 def CheckArguments(self):
3575 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
3576 # validate/normalize the node name
3577 self.hostname = netutils.GetHostname(name=self.op.node_name,
3578 family=self.primary_ip_family)
3579 self.op.node_name = self.hostname.name
3580 if self.op.readd and self.op.nodegroup:
3581 raise errors.OpPrereqError("Cannot pass a nodegroup when a node is"
3582 " being readded", errors.ECODE_INVAL)
3584 def BuildHooksEnv(self):
3587 This will run on all nodes before, and on all nodes + the new node after.
3591 "OP_TARGET": self.op.node_name,
3592 "NODE_NAME": self.op.node_name,
3593 "NODE_PIP": self.op.primary_ip,
3594 "NODE_SIP": self.op.secondary_ip,
3596 nodes_0 = self.cfg.GetNodeList()
3597 nodes_1 = nodes_0 + [self.op.node_name, ]
3598 return env, nodes_0, nodes_1
3600 def CheckPrereq(self):
3601 """Check prerequisites.
3604 - the new node is not already in the config
3606 - its parameters (single/dual homed) matches the cluster
3608 Any errors are signaled by raising errors.OpPrereqError.
3612 hostname = self.hostname
3613 node = hostname.name
3614 primary_ip = self.op.primary_ip = hostname.ip
3615 if self.op.secondary_ip is None:
3616 if self.primary_ip_family == netutils.IP6Address.family:
3617 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
3618 " IPv4 address must be given as secondary",
3620 self.op.secondary_ip = primary_ip
3622 secondary_ip = self.op.secondary_ip
3623 if not netutils.IP4Address.IsValid(secondary_ip):
3624 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
3625 " address" % secondary_ip, errors.ECODE_INVAL)
3627 node_list = cfg.GetNodeList()
3628 if not self.op.readd and node in node_list:
3629 raise errors.OpPrereqError("Node %s is already in the configuration" %
3630 node, errors.ECODE_EXISTS)
3631 elif self.op.readd and node not in node_list:
3632 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
3635 self.changed_primary_ip = False
3637 for existing_node_name in node_list:
3638 existing_node = cfg.GetNodeInfo(existing_node_name)
3640 if self.op.readd and node == existing_node_name:
3641 if existing_node.secondary_ip != secondary_ip:
3642 raise errors.OpPrereqError("Readded node doesn't have the same IP"
3643 " address configuration as before",
3645 if existing_node.primary_ip != primary_ip:
3646 self.changed_primary_ip = True
3650 if (existing_node.primary_ip == primary_ip or
3651 existing_node.secondary_ip == primary_ip or
3652 existing_node.primary_ip == secondary_ip or
3653 existing_node.secondary_ip == secondary_ip):
3654 raise errors.OpPrereqError("New node ip address(es) conflict with"
3655 " existing node %s" % existing_node.name,
3656 errors.ECODE_NOTUNIQUE)
3658 # check that the type of the node (single versus dual homed) is the
3659 # same as for the master
3660 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
3661 master_singlehomed = myself.secondary_ip == myself.primary_ip
3662 newbie_singlehomed = secondary_ip == primary_ip
3663 if master_singlehomed != newbie_singlehomed:
3664 if master_singlehomed:
3665 raise errors.OpPrereqError("The master has no private ip but the"
3666 " new node has one",
3669 raise errors.OpPrereqError("The master has a private ip but the"
3670 " new node doesn't have one",
3673 # checks reachability
3674 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
3675 raise errors.OpPrereqError("Node not reachable by ping",
3676 errors.ECODE_ENVIRON)
3678 if not newbie_singlehomed:
3679 # check reachability from my secondary ip to newbie's secondary ip
3680 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3681 source=myself.secondary_ip):
3682 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3683 " based ping to noded port",
3684 errors.ECODE_ENVIRON)
3691 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3694 self.new_node = self.cfg.GetNodeInfo(node)
3695 assert self.new_node is not None, "Can't retrieve locked node %s" % node
3697 nodegroup = cfg.LookupNodeGroup(self.op.nodegroup)
3698 self.new_node = objects.Node(name=node,
3699 primary_ip=primary_ip,
3700 secondary_ip=secondary_ip,
3701 master_candidate=self.master_candidate,
3702 offline=False, drained=False,
3703 nodegroup=nodegroup)
3705 def Exec(self, feedback_fn):
3706 """Adds the new node to the cluster.
3709 new_node = self.new_node
3710 node = new_node.name
3712 # for re-adds, reset the offline/drained/master-candidate flags;
3713 # we need to reset here, otherwise offline would prevent RPC calls
3714 # later in the procedure; this also means that if the re-add
3715 # fails, we are left with a non-offlined, broken node
3717 new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3718 self.LogInfo("Readding a node, the offline/drained flags were reset")
3719 # if we demote the node, we do cleanup later in the procedure
3720 new_node.master_candidate = self.master_candidate
3721 if self.changed_primary_ip:
3722 new_node.primary_ip = self.op.primary_ip
3724 # notify the user about any possible mc promotion
3725 if new_node.master_candidate:
3726 self.LogInfo("Node will be a master candidate")
3728 # check connectivity
3729 result = self.rpc.call_version([node])[node]
3730 result.Raise("Can't get version information from node %s" % node)
3731 if constants.PROTOCOL_VERSION == result.payload:
3732 logging.info("Communication to node %s fine, sw version %s match",
3733 node, result.payload)
3735 raise errors.OpExecError("Version mismatch master version %s,"
3736 " node version %s" %
3737 (constants.PROTOCOL_VERSION, result.payload))
3739 # Add node to our /etc/hosts, and add key to known_hosts
3740 if self.cfg.GetClusterInfo().modify_etc_hosts:
3741 master_node = self.cfg.GetMasterNode()
3742 result = self.rpc.call_etc_hosts_modify(master_node,
3743 constants.ETC_HOSTS_ADD,
3746 result.Raise("Can't update hosts file with new host data")
3748 if new_node.secondary_ip != new_node.primary_ip:
3749 result = self.rpc.call_node_has_ip_address(new_node.name,
3750 new_node.secondary_ip)
3751 result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3752 prereq=True, ecode=errors.ECODE_ENVIRON)
3753 if not result.payload:
3754 raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3755 " you gave (%s). Please fix and re-run this"
3756 " command." % new_node.secondary_ip)
3758 node_verify_list = [self.cfg.GetMasterNode()]
3759 node_verify_param = {
3760 constants.NV_NODELIST: [node],
3761 # TODO: do a node-net-test as well?
3764 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3765 self.cfg.GetClusterName())
3766 for verifier in node_verify_list:
3767 result[verifier].Raise("Cannot communicate with node %s" % verifier)
3768 nl_payload = result[verifier].payload[constants.NV_NODELIST]
3770 for failed in nl_payload:
3771 feedback_fn("ssh/hostname verification failed"
3772 " (checking from %s): %s" %
3773 (verifier, nl_payload[failed]))
3774 raise errors.OpExecError("ssh/hostname verification failed.")
3777 _RedistributeAncillaryFiles(self)
3778 self.context.ReaddNode(new_node)
3779 # make sure we redistribute the config
3780 self.cfg.Update(new_node, feedback_fn)
3781 # and make sure the new node will not have old files around
3782 if not new_node.master_candidate:
3783 result = self.rpc.call_node_demote_from_mc(new_node.name)
3784 msg = result.fail_msg
3786 self.LogWarning("Node failed to demote itself from master"
3787 " candidate status: %s" % msg)
3789 _RedistributeAncillaryFiles(self, additional_nodes=[node])
3790 self.context.AddNode(new_node, self.proc.GetECId())
3793 class LUSetNodeParams(LogicalUnit):
3794 """Modifies the parameters of a node.
3797 HPATH = "node-modify"
3798 HTYPE = constants.HTYPE_NODE
3801 ("master_candidate", None, ht.TMaybeBool),
3802 ("offline", None, ht.TMaybeBool),
3803 ("drained", None, ht.TMaybeBool),
3804 ("auto_promote", False, ht.TBool),
3809 def CheckArguments(self):
3810 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3811 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3812 if all_mods.count(None) == 3:
3813 raise errors.OpPrereqError("Please pass at least one modification",
3815 if all_mods.count(True) > 1:
3816 raise errors.OpPrereqError("Can't set the node into more than one"
3817 " state at the same time",
3820 # Boolean value that tells us whether we're offlining or draining the node
3821 self.offline_or_drain = (self.op.offline == True or
3822 self.op.drained == True)
3823 self.deoffline_or_drain = (self.op.offline == False or
3824 self.op.drained == False)
3825 self.might_demote = (self.op.master_candidate == False or
3826 self.offline_or_drain)
3828 self.lock_all = self.op.auto_promote and self.might_demote
3831 def ExpandNames(self):
3833 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
3835 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3837 def BuildHooksEnv(self):
3840 This runs on the master node.
3844 "OP_TARGET": self.op.node_name,
3845 "MASTER_CANDIDATE": str(self.op.master_candidate),
3846 "OFFLINE": str(self.op.offline),
3847 "DRAINED": str(self.op.drained),
3849 nl = [self.cfg.GetMasterNode(),
3853 def CheckPrereq(self):
3854 """Check prerequisites.
3856 This only checks the instance list against the existing names.
3859 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3861 if (self.op.master_candidate is not None or
3862 self.op.drained is not None or
3863 self.op.offline is not None):
3864 # we can't change the master's node flags
3865 if self.op.node_name == self.cfg.GetMasterNode():
3866 raise errors.OpPrereqError("The master role can be changed"
3867 " only via master-failover",
3871 if node.master_candidate and self.might_demote and not self.lock_all:
3872 assert not self.op.auto_promote, "auto-promote set but lock_all not"
3873 # check if after removing the current node, we're missing master
3875 (mc_remaining, mc_should, _) = \
3876 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
3877 if mc_remaining < mc_should:
3878 raise errors.OpPrereqError("Not enough master candidates, please"
3879 " pass auto_promote to allow promotion",
3882 if (self.op.master_candidate == True and
3883 ((node.offline and not self.op.offline == False) or
3884 (node.drained and not self.op.drained == False))):
3885 raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3886 " to master_candidate" % node.name,
3889 # If we're being deofflined/drained, we'll MC ourself if needed
3890 if (self.deoffline_or_drain and not self.offline_or_drain and not
3891 self.op.master_candidate == True and not node.master_candidate):
3892 self.op.master_candidate = _DecideSelfPromotion(self)
3893 if self.op.master_candidate:
3894 self.LogInfo("Autopromoting node to master candidate")
3898 def Exec(self, feedback_fn):
3907 if self.op.offline is not None:
3908 node.offline = self.op.offline
3909 result.append(("offline", str(self.op.offline)))
3910 if self.op.offline == True:
3911 if node.master_candidate:
3912 node.master_candidate = False
3914 result.append(("master_candidate", "auto-demotion due to offline"))
3916 node.drained = False
3917 result.append(("drained", "clear drained status due to offline"))
3919 if self.op.master_candidate is not None:
3920 node.master_candidate = self.op.master_candidate
3922 result.append(("master_candidate", str(self.op.master_candidate)))
3923 if self.op.master_candidate == False:
3924 rrc = self.rpc.call_node_demote_from_mc(node.name)
3927 self.LogWarning("Node failed to demote itself: %s" % msg)
3929 if self.op.drained is not None:
3930 node.drained = self.op.drained
3931 result.append(("drained", str(self.op.drained)))
3932 if self.op.drained == True:
3933 if node.master_candidate:
3934 node.master_candidate = False
3936 result.append(("master_candidate", "auto-demotion due to drain"))
3937 rrc = self.rpc.call_node_demote_from_mc(node.name)
3940 self.LogWarning("Node failed to demote itself: %s" % msg)
3942 node.offline = False
3943 result.append(("offline", "clear offline status due to drain"))
3945 # we locked all nodes, we adjust the CP before updating this node
3947 _AdjustCandidatePool(self, [node.name])
3949 # this will trigger configuration file update, if needed
3950 self.cfg.Update(node, feedback_fn)
3952 # this will trigger job queue propagation or cleanup
3954 self.context.ReaddNode(node)
3959 class LUPowercycleNode(NoHooksLU):
3960 """Powercycles a node.
3969 def CheckArguments(self):
3970 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3971 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
3972 raise errors.OpPrereqError("The node is the master and the force"
3973 " parameter was not set",
3976 def ExpandNames(self):
3977 """Locking for PowercycleNode.
3979 This is a last-resort option and shouldn't block on other
3980 jobs. Therefore, we grab no locks.
3983 self.needed_locks = {}
3985 def Exec(self, feedback_fn):
3989 result = self.rpc.call_node_powercycle(self.op.node_name,
3990 self.cfg.GetHypervisorType())
3991 result.Raise("Failed to schedule the reboot")
3992 return result.payload
3995 class LUQueryClusterInfo(NoHooksLU):
3996 """Query cluster configuration.
4001 def ExpandNames(self):
4002 self.needed_locks = {}
4004 def Exec(self, feedback_fn):
4005 """Return cluster config.
4008 cluster = self.cfg.GetClusterInfo()
4011 # Filter just for enabled hypervisors
4012 for os_name, hv_dict in cluster.os_hvp.items():
4013 os_hvp[os_name] = {}
4014 for hv_name, hv_params in hv_dict.items():
4015 if hv_name in cluster.enabled_hypervisors:
4016 os_hvp[os_name][hv_name] = hv_params
4018 # Convert ip_family to ip_version
4019 primary_ip_version = constants.IP4_VERSION
4020 if cluster.primary_ip_family == netutils.IP6Address.family:
4021 primary_ip_version = constants.IP6_VERSION
4024 "software_version": constants.RELEASE_VERSION,
4025 "protocol_version": constants.PROTOCOL_VERSION,
4026 "config_version": constants.CONFIG_VERSION,
4027 "os_api_version": max(constants.OS_API_VERSIONS),
4028 "export_version": constants.EXPORT_VERSION,
4029 "architecture": (platform.architecture()[0], platform.machine()),
4030 "name": cluster.cluster_name,
4031 "master": cluster.master_node,
4032 "default_hypervisor": cluster.enabled_hypervisors[0],
4033 "enabled_hypervisors": cluster.enabled_hypervisors,
4034 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4035 for hypervisor_name in cluster.enabled_hypervisors]),
4037 "beparams": cluster.beparams,
4038 "osparams": cluster.osparams,
4039 "nicparams": cluster.nicparams,
4040 "candidate_pool_size": cluster.candidate_pool_size,
4041 "master_netdev": cluster.master_netdev,
4042 "volume_group_name": cluster.volume_group_name,
4043 "drbd_usermode_helper": cluster.drbd_usermode_helper,
4044 "file_storage_dir": cluster.file_storage_dir,
4045 "maintain_node_health": cluster.maintain_node_health,
4046 "ctime": cluster.ctime,
4047 "mtime": cluster.mtime,
4048 "uuid": cluster.uuid,
4049 "tags": list(cluster.GetTags()),
4050 "uid_pool": cluster.uid_pool,
4051 "default_iallocator": cluster.default_iallocator,
4052 "reserved_lvs": cluster.reserved_lvs,
4053 "primary_ip_version": primary_ip_version,
4054 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4060 class LUQueryConfigValues(NoHooksLU):
4061 """Return configuration values.
4064 _OP_PARAMS = [_POutputFields]
4066 _FIELDS_DYNAMIC = utils.FieldSet()
4067 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4068 "watcher_pause", "volume_group_name")
4070 def CheckArguments(self):
4071 _CheckOutputFields(static=self._FIELDS_STATIC,
4072 dynamic=self._FIELDS_DYNAMIC,
4073 selected=self.op.output_fields)
4075 def ExpandNames(self):
4076 self.needed_locks = {}
4078 def Exec(self, feedback_fn):
4079 """Dump a representation of the cluster config to the standard output.
4083 for field in self.op.output_fields:
4084 if field == "cluster_name":
4085 entry = self.cfg.GetClusterName()
4086 elif field == "master_node":
4087 entry = self.cfg.GetMasterNode()
4088 elif field == "drain_flag":
4089 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4090 elif field == "watcher_pause":
4091 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4092 elif field == "volume_group_name":
4093 entry = self.cfg.GetVGName()
4095 raise errors.ParameterError(field)
4096 values.append(entry)
4100 class LUActivateInstanceDisks(NoHooksLU):
4101 """Bring up an instance's disks.
4106 ("ignore_size", False, ht.TBool),
4110 def ExpandNames(self):
4111 self._ExpandAndLockInstance()
4112 self.needed_locks[locking.LEVEL_NODE] = []
4113 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4115 def DeclareLocks(self, level):
4116 if level == locking.LEVEL_NODE:
4117 self._LockInstancesNodes()
4119 def CheckPrereq(self):
4120 """Check prerequisites.
4122 This checks that the instance is in the cluster.
4125 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4126 assert self.instance is not None, \
4127 "Cannot retrieve locked instance %s" % self.op.instance_name
4128 _CheckNodeOnline(self, self.instance.primary_node)
4130 def Exec(self, feedback_fn):
4131 """Activate the disks.
4134 disks_ok, disks_info = \
4135 _AssembleInstanceDisks(self, self.instance,
4136 ignore_size=self.op.ignore_size)
4138 raise errors.OpExecError("Cannot activate block devices")
4143 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4145 """Prepare the block devices for an instance.
4147 This sets up the block devices on all nodes.
4149 @type lu: L{LogicalUnit}
4150 @param lu: the logical unit on whose behalf we execute
4151 @type instance: L{objects.Instance}
4152 @param instance: the instance for whose disks we assemble
4153 @type disks: list of L{objects.Disk} or None
4154 @param disks: which disks to assemble (or all, if None)
4155 @type ignore_secondaries: boolean
4156 @param ignore_secondaries: if true, errors on secondary nodes
4157 won't result in an error return from the function
4158 @type ignore_size: boolean
4159 @param ignore_size: if true, the current known size of the disk
4160 will not be used during the disk activation, useful for cases
4161 when the size is wrong
4162 @return: False if the operation failed, otherwise a list of
4163 (host, instance_visible_name, node_visible_name)
4164 with the mapping from node devices to instance devices
4169 iname = instance.name
4170 disks = _ExpandCheckDisks(instance, disks)
4172 # With the two passes mechanism we try to reduce the window of
4173 # opportunity for the race condition of switching DRBD to primary
4174 # before handshaking occured, but we do not eliminate it
4176 # The proper fix would be to wait (with some limits) until the
4177 # connection has been made and drbd transitions from WFConnection
4178 # into any other network-connected state (Connected, SyncTarget,
4181 # 1st pass, assemble on all nodes in secondary mode
4182 for inst_disk in disks:
4183 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4185 node_disk = node_disk.Copy()
4186 node_disk.UnsetSize()
4187 lu.cfg.SetDiskID(node_disk, node)
4188 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4189 msg = result.fail_msg
4191 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4192 " (is_primary=False, pass=1): %s",
4193 inst_disk.iv_name, node, msg)
4194 if not ignore_secondaries:
4197 # FIXME: race condition on drbd migration to primary
4199 # 2nd pass, do only the primary node
4200 for inst_disk in disks:
4203 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4204 if node != instance.primary_node:
4207 node_disk = node_disk.Copy()
4208 node_disk.UnsetSize()
4209 lu.cfg.SetDiskID(node_disk, node)
4210 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4211 msg = result.fail_msg
4213 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4214 " (is_primary=True, pass=2): %s",
4215 inst_disk.iv_name, node, msg)
4218 dev_path = result.payload
4220 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4222 # leave the disks configured for the primary node
4223 # this is a workaround that would be fixed better by
4224 # improving the logical/physical id handling
4226 lu.cfg.SetDiskID(disk, instance.primary_node)
4228 return disks_ok, device_info
4231 def _StartInstanceDisks(lu, instance, force):
4232 """Start the disks of an instance.
4235 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4236 ignore_secondaries=force)
4238 _ShutdownInstanceDisks(lu, instance)
4239 if force is not None and not force:
4240 lu.proc.LogWarning("", hint="If the message above refers to a"
4242 " you can retry the operation using '--force'.")
4243 raise errors.OpExecError("Disk consistency error")
4246 class LUDeactivateInstanceDisks(NoHooksLU):
4247 """Shutdown an instance's disks.
4255 def ExpandNames(self):
4256 self._ExpandAndLockInstance()
4257 self.needed_locks[locking.LEVEL_NODE] = []
4258 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4260 def DeclareLocks(self, level):
4261 if level == locking.LEVEL_NODE:
4262 self._LockInstancesNodes()
4264 def CheckPrereq(self):
4265 """Check prerequisites.
4267 This checks that the instance is in the cluster.
4270 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4271 assert self.instance is not None, \
4272 "Cannot retrieve locked instance %s" % self.op.instance_name
4274 def Exec(self, feedback_fn):
4275 """Deactivate the disks
4278 instance = self.instance
4279 _SafeShutdownInstanceDisks(self, instance)
4282 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4283 """Shutdown block devices of an instance.
4285 This function checks if an instance is running, before calling
4286 _ShutdownInstanceDisks.
4289 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4290 _ShutdownInstanceDisks(lu, instance, disks=disks)
4293 def _ExpandCheckDisks(instance, disks):
4294 """Return the instance disks selected by the disks list
4296 @type disks: list of L{objects.Disk} or None
4297 @param disks: selected disks
4298 @rtype: list of L{objects.Disk}
4299 @return: selected instance disks to act on
4303 return instance.disks
4305 if not set(disks).issubset(instance.disks):
4306 raise errors.ProgrammerError("Can only act on disks belonging to the"
4311 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4312 """Shutdown block devices of an instance.
4314 This does the shutdown on all nodes of the instance.
4316 If the ignore_primary is false, errors on the primary node are
4321 disks = _ExpandCheckDisks(instance, disks)
4324 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4325 lu.cfg.SetDiskID(top_disk, node)
4326 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4327 msg = result.fail_msg
4329 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4330 disk.iv_name, node, msg)
4331 if not ignore_primary or node != instance.primary_node:
4336 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4337 """Checks if a node has enough free memory.
4339 This function check if a given node has the needed amount of free
4340 memory. In case the node has less memory or we cannot get the
4341 information from the node, this function raise an OpPrereqError
4344 @type lu: C{LogicalUnit}
4345 @param lu: a logical unit from which we get configuration data
4347 @param node: the node to check
4348 @type reason: C{str}
4349 @param reason: string to use in the error message
4350 @type requested: C{int}
4351 @param requested: the amount of memory in MiB to check for
4352 @type hypervisor_name: C{str}
4353 @param hypervisor_name: the hypervisor to ask for memory stats
4354 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4355 we cannot check the node
4358 nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4359 nodeinfo[node].Raise("Can't get data from node %s" % node,
4360 prereq=True, ecode=errors.ECODE_ENVIRON)
4361 free_mem = nodeinfo[node].payload.get('memory_free', None)
4362 if not isinstance(free_mem, int):
4363 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4364 " was '%s'" % (node, free_mem),
4365 errors.ECODE_ENVIRON)
4366 if requested > free_mem:
4367 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4368 " needed %s MiB, available %s MiB" %
4369 (node, reason, requested, free_mem),
4373 def _CheckNodesFreeDisk(lu, nodenames, requested):
4374 """Checks if nodes have enough free disk space in the default VG.
4376 This function check if all given nodes have the needed amount of
4377 free disk. In case any node has less disk or we cannot get the
4378 information from the node, this function raise an OpPrereqError
4381 @type lu: C{LogicalUnit}
4382 @param lu: a logical unit from which we get configuration data
4383 @type nodenames: C{list}
4384 @param nodenames: the list of node names to check
4385 @type requested: C{int}
4386 @param requested: the amount of disk in MiB to check for
4387 @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4388 we cannot check the node
4391 nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4392 lu.cfg.GetHypervisorType())
4393 for node in nodenames:
4394 info = nodeinfo[node]
4395 info.Raise("Cannot get current information from node %s" % node,
4396 prereq=True, ecode=errors.ECODE_ENVIRON)
4397 vg_free = info.payload.get("vg_free", None)
4398 if not isinstance(vg_free, int):
4399 raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4400 " result was '%s'" % (node, vg_free),
4401 errors.ECODE_ENVIRON)
4402 if requested > vg_free:
4403 raise errors.OpPrereqError("Not enough disk space on target node %s:"
4404 " required %d MiB, available %d MiB" %
4405 (node, requested, vg_free),
4409 class LUStartupInstance(LogicalUnit):
4410 """Starts an instance.
4413 HPATH = "instance-start"
4414 HTYPE = constants.HTYPE_INSTANCE
4418 _PIgnoreOfflineNodes,
4419 ("hvparams", ht.EmptyDict, ht.TDict),
4420 ("beparams", ht.EmptyDict, ht.TDict),
4424 def CheckArguments(self):
4426 if self.op.beparams:
4427 # fill the beparams dict
4428 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4430 def ExpandNames(self):
4431 self._ExpandAndLockInstance()
4433 def BuildHooksEnv(self):
4436 This runs on master, primary and secondary nodes of the instance.
4440 "FORCE": self.op.force,
4442 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4443 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4446 def CheckPrereq(self):
4447 """Check prerequisites.
4449 This checks that the instance is in the cluster.
4452 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4453 assert self.instance is not None, \
4454 "Cannot retrieve locked instance %s" % self.op.instance_name
4457 if self.op.hvparams:
4458 # check hypervisor parameter syntax (locally)
4459 cluster = self.cfg.GetClusterInfo()
4460 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4461 filled_hvp = cluster.FillHV(instance)
4462 filled_hvp.update(self.op.hvparams)
4463 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4464 hv_type.CheckParameterSyntax(filled_hvp)
4465 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4467 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
4469 if self.primary_offline and self.op.ignore_offline_nodes:
4470 self.proc.LogWarning("Ignoring offline primary node")
4472 if self.op.hvparams or self.op.beparams:
4473 self.proc.LogWarning("Overridden parameters are ignored")
4475 _CheckNodeOnline(self, instance.primary_node)
4477 bep = self.cfg.GetClusterInfo().FillBE(instance)
4479 # check bridges existence
4480 _CheckInstanceBridgesExist(self, instance)
4482 remote_info = self.rpc.call_instance_info(instance.primary_node,
4484 instance.hypervisor)
4485 remote_info.Raise("Error checking node %s" % instance.primary_node,
4486 prereq=True, ecode=errors.ECODE_ENVIRON)
4487 if not remote_info.payload: # not running already
4488 _CheckNodeFreeMemory(self, instance.primary_node,
4489 "starting instance %s" % instance.name,
4490 bep[constants.BE_MEMORY], instance.hypervisor)
4492 def Exec(self, feedback_fn):
4493 """Start the instance.
4496 instance = self.instance
4497 force = self.op.force
4499 self.cfg.MarkInstanceUp(instance.name)
4501 if self.primary_offline:
4502 assert self.op.ignore_offline_nodes
4503 self.proc.LogInfo("Primary node offline, marked instance as started")
4505 node_current = instance.primary_node
4507 _StartInstanceDisks(self, instance, force)
4509 result = self.rpc.call_instance_start(node_current, instance,
4510 self.op.hvparams, self.op.beparams)
4511 msg = result.fail_msg
4513 _ShutdownInstanceDisks(self, instance)
4514 raise errors.OpExecError("Could not start instance: %s" % msg)
4517 class LURebootInstance(LogicalUnit):
4518 """Reboot an instance.
4521 HPATH = "instance-reboot"
4522 HTYPE = constants.HTYPE_INSTANCE
4525 ("ignore_secondaries", False, ht.TBool),
4526 ("reboot_type", ht.NoDefault, ht.TElemOf(constants.REBOOT_TYPES)),
4531 def ExpandNames(self):
4532 self._ExpandAndLockInstance()
4534 def BuildHooksEnv(self):
4537 This runs on master, primary and secondary nodes of the instance.
4541 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4542 "REBOOT_TYPE": self.op.reboot_type,
4543 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
4545 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4546 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4549 def CheckPrereq(self):
4550 """Check prerequisites.
4552 This checks that the instance is in the cluster.
4555 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4556 assert self.instance is not None, \
4557 "Cannot retrieve locked instance %s" % self.op.instance_name
4559 _CheckNodeOnline(self, instance.primary_node)
4561 # check bridges existence
4562 _CheckInstanceBridgesExist(self, instance)
4564 def Exec(self, feedback_fn):
4565 """Reboot the instance.
4568 instance = self.instance
4569 ignore_secondaries = self.op.ignore_secondaries
4570 reboot_type = self.op.reboot_type
4572 node_current = instance.primary_node
4574 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4575 constants.INSTANCE_REBOOT_HARD]:
4576 for disk in instance.disks:
4577 self.cfg.SetDiskID(disk, node_current)
4578 result = self.rpc.call_instance_reboot(node_current, instance,
4580 self.op.shutdown_timeout)
4581 result.Raise("Could not reboot instance")
4583 result = self.rpc.call_instance_shutdown(node_current, instance,
4584 self.op.shutdown_timeout)
4585 result.Raise("Could not shutdown instance for full reboot")
4586 _ShutdownInstanceDisks(self, instance)
4587 _StartInstanceDisks(self, instance, ignore_secondaries)
4588 result = self.rpc.call_instance_start(node_current, instance, None, None)
4589 msg = result.fail_msg
4591 _ShutdownInstanceDisks(self, instance)
4592 raise errors.OpExecError("Could not start instance for"
4593 " full reboot: %s" % msg)
4595 self.cfg.MarkInstanceUp(instance.name)
4598 class LUShutdownInstance(LogicalUnit):
4599 """Shutdown an instance.
4602 HPATH = "instance-stop"
4603 HTYPE = constants.HTYPE_INSTANCE
4606 _PIgnoreOfflineNodes,
4607 ("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, ht.TPositiveInt),
4611 def ExpandNames(self):
4612 self._ExpandAndLockInstance()
4614 def BuildHooksEnv(self):
4617 This runs on master, primary and secondary nodes of the instance.
4620 env = _BuildInstanceHookEnvByObject(self, self.instance)
4621 env["TIMEOUT"] = self.op.timeout
4622 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4625 def CheckPrereq(self):
4626 """Check prerequisites.
4628 This checks that the instance is in the cluster.
4631 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4632 assert self.instance is not None, \
4633 "Cannot retrieve locked instance %s" % self.op.instance_name
4635 self.primary_offline = \
4636 self.cfg.GetNodeInfo(self.instance.primary_node).offline
4638 if self.primary_offline and self.op.ignore_offline_nodes:
4639 self.proc.LogWarning("Ignoring offline primary node")
4641 _CheckNodeOnline(self, self.instance.primary_node)
4643 def Exec(self, feedback_fn):
4644 """Shutdown the instance.
4647 instance = self.instance
4648 node_current = instance.primary_node
4649 timeout = self.op.timeout
4651 self.cfg.MarkInstanceDown(instance.name)
4653 if self.primary_offline:
4654 assert self.op.ignore_offline_nodes
4655 self.proc.LogInfo("Primary node offline, marked instance as stopped")
4657 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4658 msg = result.fail_msg
4660 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4662 _ShutdownInstanceDisks(self, instance)
4665 class LUReinstallInstance(LogicalUnit):
4666 """Reinstall an instance.
4669 HPATH = "instance-reinstall"
4670 HTYPE = constants.HTYPE_INSTANCE
4673 ("os_type", None, ht.TMaybeString),
4674 ("force_variant", False, ht.TBool),
4678 def ExpandNames(self):
4679 self._ExpandAndLockInstance()
4681 def BuildHooksEnv(self):
4684 This runs on master, primary and secondary nodes of the instance.
4687 env = _BuildInstanceHookEnvByObject(self, self.instance)
4688 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4691 def CheckPrereq(self):
4692 """Check prerequisites.
4694 This checks that the instance is in the cluster and is not running.
4697 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4698 assert instance is not None, \
4699 "Cannot retrieve locked instance %s" % self.op.instance_name
4700 _CheckNodeOnline(self, instance.primary_node)
4702 if instance.disk_template == constants.DT_DISKLESS:
4703 raise errors.OpPrereqError("Instance '%s' has no disks" %
4704 self.op.instance_name,
4706 _CheckInstanceDown(self, instance, "cannot reinstall")
4708 if self.op.os_type is not None:
4710 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4711 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4713 self.instance = instance
4715 def Exec(self, feedback_fn):
4716 """Reinstall the instance.
4719 inst = self.instance
4721 if self.op.os_type is not None:
4722 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4723 inst.os = self.op.os_type
4724 self.cfg.Update(inst, feedback_fn)
4726 _StartInstanceDisks(self, inst, None)
4728 feedback_fn("Running the instance OS create scripts...")
4729 # FIXME: pass debug option from opcode to backend
4730 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4731 self.op.debug_level)
4732 result.Raise("Could not install OS for instance %s on node %s" %
4733 (inst.name, inst.primary_node))
4735 _ShutdownInstanceDisks(self, inst)
4738 class LURecreateInstanceDisks(LogicalUnit):
4739 """Recreate an instance's missing disks.
4742 HPATH = "instance-recreate-disks"
4743 HTYPE = constants.HTYPE_INSTANCE
4746 ("disks", ht.EmptyList, ht.TListOf(ht.TPositiveInt)),
4750 def ExpandNames(self):
4751 self._ExpandAndLockInstance()
4753 def BuildHooksEnv(self):
4756 This runs on master, primary and secondary nodes of the instance.
4759 env = _BuildInstanceHookEnvByObject(self, self.instance)
4760 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4763 def CheckPrereq(self):
4764 """Check prerequisites.
4766 This checks that the instance is in the cluster and is not running.
4769 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4770 assert instance is not None, \
4771 "Cannot retrieve locked instance %s" % self.op.instance_name
4772 _CheckNodeOnline(self, instance.primary_node)
4774 if instance.disk_template == constants.DT_DISKLESS:
4775 raise errors.OpPrereqError("Instance '%s' has no disks" %
4776 self.op.instance_name, errors.ECODE_INVAL)
4777 _CheckInstanceDown(self, instance, "cannot recreate disks")
4779 if not self.op.disks:
4780 self.op.disks = range(len(instance.disks))
4782 for idx in self.op.disks:
4783 if idx >= len(instance.disks):
4784 raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4787 self.instance = instance
4789 def Exec(self, feedback_fn):
4790 """Recreate the disks.
4794 for idx, _ in enumerate(self.instance.disks):
4795 if idx not in self.op.disks: # disk idx has not been passed in
4799 _CreateDisks(self, self.instance, to_skip=to_skip)
4802 class LURenameInstance(LogicalUnit):
4803 """Rename an instance.
4806 HPATH = "instance-rename"
4807 HTYPE = constants.HTYPE_INSTANCE
4810 ("new_name", ht.NoDefault, ht.TNonEmptyString),
4811 ("ip_check", False, ht.TBool),
4812 ("name_check", True, ht.TBool),
4815 def CheckArguments(self):
4819 if self.op.ip_check and not self.op.name_check:
4820 # TODO: make the ip check more flexible and not depend on the name check
4821 raise errors.OpPrereqError("Cannot do ip check without a name check",
4824 def BuildHooksEnv(self):
4827 This runs on master, primary and secondary nodes of the instance.
4830 env = _BuildInstanceHookEnvByObject(self, self.instance)
4831 env["INSTANCE_NEW_NAME"] = self.op.new_name
4832 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4835 def CheckPrereq(self):
4836 """Check prerequisites.
4838 This checks that the instance is in the cluster and is not running.
4841 self.op.instance_name = _ExpandInstanceName(self.cfg,
4842 self.op.instance_name)
4843 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4844 assert instance is not None
4845 _CheckNodeOnline(self, instance.primary_node)
4846 _CheckInstanceDown(self, instance, "cannot rename")
4847 self.instance = instance
4849 new_name = self.op.new_name
4850 if self.op.name_check:
4851 hostname = netutils.GetHostname(name=new_name)
4852 new_name = self.op.new_name = hostname.name
4853 if (self.op.ip_check and
4854 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
4855 raise errors.OpPrereqError("IP %s of instance %s already in use" %
4856 (hostname.ip, new_name),
4857 errors.ECODE_NOTUNIQUE)
4859 instance_list = self.cfg.GetInstanceList()
4860 if new_name in instance_list:
4861 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4862 new_name, errors.ECODE_EXISTS)
4864 def Exec(self, feedback_fn):
4865 """Reinstall the instance.
4868 inst = self.instance
4869 old_name = inst.name
4871 if inst.disk_template == constants.DT_FILE:
4872 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4874 self.cfg.RenameInstance(inst.name, self.op.new_name)
4875 # Change the instance lock. This is definitely safe while we hold the BGL
4876 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4877 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4879 # re-read the instance from the configuration after rename
4880 inst = self.cfg.GetInstanceInfo(self.op.new_name)
4882 if inst.disk_template == constants.DT_FILE:
4883 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4884 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4885 old_file_storage_dir,
4886 new_file_storage_dir)
4887 result.Raise("Could not rename on node %s directory '%s' to '%s'"
4888 " (but the instance has been renamed in Ganeti)" %
4889 (inst.primary_node, old_file_storage_dir,
4890 new_file_storage_dir))
4892 _StartInstanceDisks(self, inst, None)
4894 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4895 old_name, self.op.debug_level)
4896 msg = result.fail_msg
4898 msg = ("Could not run OS rename script for instance %s on node %s"
4899 " (but the instance has been renamed in Ganeti): %s" %
4900 (inst.name, inst.primary_node, msg))
4901 self.proc.LogWarning(msg)
4903 _ShutdownInstanceDisks(self, inst)
4908 class LURemoveInstance(LogicalUnit):
4909 """Remove an instance.
4912 HPATH = "instance-remove"
4913 HTYPE = constants.HTYPE_INSTANCE
4916 ("ignore_failures", False, ht.TBool),
4921 def ExpandNames(self):
4922 self._ExpandAndLockInstance()
4923 self.needed_locks[locking.LEVEL_NODE] = []
4924 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4926 def DeclareLocks(self, level):
4927 if level == locking.LEVEL_NODE:
4928 self._LockInstancesNodes()
4930 def BuildHooksEnv(self):
4933 This runs on master, primary and secondary nodes of the instance.
4936 env = _BuildInstanceHookEnvByObject(self, self.instance)
4937 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
4938 nl = [self.cfg.GetMasterNode()]
4939 nl_post = list(self.instance.all_nodes) + nl
4940 return env, nl, nl_post
4942 def CheckPrereq(self):
4943 """Check prerequisites.
4945 This checks that the instance is in the cluster.
4948 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4949 assert self.instance is not None, \
4950 "Cannot retrieve locked instance %s" % self.op.instance_name
4952 def Exec(self, feedback_fn):
4953 """Remove the instance.
4956 instance = self.instance
4957 logging.info("Shutting down instance %s on node %s",
4958 instance.name, instance.primary_node)
4960 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
4961 self.op.shutdown_timeout)
4962 msg = result.fail_msg
4964 if self.op.ignore_failures:
4965 feedback_fn("Warning: can't shutdown instance: %s" % msg)
4967 raise errors.OpExecError("Could not shutdown instance %s on"
4969 (instance.name, instance.primary_node, msg))
4971 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
4974 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
4975 """Utility function to remove an instance.
4978 logging.info("Removing block devices for instance %s", instance.name)
4980 if not _RemoveDisks(lu, instance):
4981 if not ignore_failures:
4982 raise errors.OpExecError("Can't remove instance's disks")
4983 feedback_fn("Warning: can't remove instance's disks")
4985 logging.info("Removing instance %s out of cluster config", instance.name)
4987 lu.cfg.RemoveInstance(instance.name)
4989 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
4990 "Instance lock removal conflict"
4992 # Remove lock for the instance
4993 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
4996 class LUQueryInstances(NoHooksLU):
4997 """Logical unit for querying instances.
5000 # pylint: disable-msg=W0142
5002 ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
5003 ("names", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
5004 ("use_locking", False, ht.TBool),
5007 _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
5008 "serial_no", "ctime", "mtime", "uuid"]
5009 _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
5011 "disk_template", "ip", "mac", "bridge",
5012 "nic_mode", "nic_link",
5013 "sda_size", "sdb_size", "vcpus", "tags",
5014 "network_port", "beparams",
5015 r"(disk)\.(size)/([0-9]+)",
5016 r"(disk)\.(sizes)", "disk_usage",
5017 r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
5018 r"(nic)\.(bridge)/([0-9]+)",
5019 r"(nic)\.(macs|ips|modes|links|bridges)",
5020 r"(disk|nic)\.(count)",
5021 "hvparams", "custom_hvparams",
5022 "custom_beparams", "custom_nicparams",
5023 ] + _SIMPLE_FIELDS +
5025 for name in constants.HVS_PARAMETERS
5026 if name not in constants.HVC_GLOBALS] +
5028 for name in constants.BES_PARAMETERS])
5029 _FIELDS_DYNAMIC = utils.FieldSet("oper_state",
5035 def CheckArguments(self):
5036 _CheckOutputFields(static=self._FIELDS_STATIC,
5037 dynamic=self._FIELDS_DYNAMIC,
5038 selected=self.op.output_fields)
5040 def ExpandNames(self):
5041 self.needed_locks = {}
5042 self.share_locks[locking.LEVEL_INSTANCE] = 1
5043 self.share_locks[locking.LEVEL_NODE] = 1
5046 self.wanted = _GetWantedInstances(self, self.op.names)
5048 self.wanted = locking.ALL_SET
5050 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
5051 self.do_locking = self.do_node_query and self.op.use_locking
5053 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5054 self.needed_locks[locking.LEVEL_NODE] = []
5055 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5057 def DeclareLocks(self, level):
5058 if level == locking.LEVEL_NODE and self.do_locking:
5059 self._LockInstancesNodes()
5061 def Exec(self, feedback_fn):
5062 """Computes the list of nodes and their attributes.
5065 # pylint: disable-msg=R0912
5066 # way too many branches here
5067 all_info = self.cfg.GetAllInstancesInfo()
5068 if self.wanted == locking.ALL_SET:
5069 # caller didn't specify instance names, so ordering is not important
5071 instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5073 instance_names = all_info.keys()
5074 instance_names = utils.NiceSort(instance_names)
5076 # caller did specify names, so we must keep the ordering
5078 tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
5080 tgt_set = all_info.keys()
5081 missing = set(self.wanted).difference(tgt_set)
5083 raise errors.OpExecError("Some instances were removed before"
5084 " retrieving their data: %s" % missing)
5085 instance_names = self.wanted
5087 instance_list = [all_info[iname] for iname in instance_names]
5089 # begin data gathering
5091 nodes = frozenset([inst.primary_node for inst in instance_list])
5092 hv_list = list(set([inst.hypervisor for inst in instance_list]))
5096 if self.do_node_query:
5098 node_data = self.rpc.call_all_instances_info(nodes, hv_list)
5100 result = node_data[name]
5102 # offline nodes will be in both lists
5103 off_nodes.append(name)
5105 bad_nodes.append(name)
5108 live_data.update(result.payload)
5109 # else no instance is alive
5111 live_data = dict([(name, {}) for name in instance_names])
5113 # end data gathering
5118 cluster = self.cfg.GetClusterInfo()
5119 for instance in instance_list:
5121 i_hv = cluster.FillHV(instance, skip_globals=True)
5122 i_be = cluster.FillBE(instance)
5123 i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics]
5124 for field in self.op.output_fields:
5125 st_match = self._FIELDS_STATIC.Matches(field)
5126 if field in self._SIMPLE_FIELDS:
5127 val = getattr(instance, field)
5128 elif field == "pnode":
5129 val = instance.primary_node
5130 elif field == "snodes":
5131 val = list(instance.secondary_nodes)
5132 elif field == "admin_state":
5133 val = instance.admin_up
5134 elif field == "oper_state":
5135 if instance.primary_node in bad_nodes:
5138 val = bool(live_data.get(instance.name))
5139 elif field == "status":
5140 if instance.primary_node in off_nodes:
5141 val = "ERROR_nodeoffline"
5142 elif instance.primary_node in bad_nodes:
5143 val = "ERROR_nodedown"
5145 running = bool(live_data.get(instance.name))
5147 if instance.admin_up:
5152 if instance.admin_up:
5156 elif field == "oper_ram":
5157 if instance.primary_node in bad_nodes:
5159 elif instance.name in live_data:
5160 val = live_data[instance.name].get("memory", "?")
5163 elif field == "oper_vcpus":
5164 if instance.primary_node in bad_nodes:
5166 elif instance.name in live_data:
5167 val = live_data[instance.name].get("vcpus", "?")
5170 elif field == "vcpus":
5171 val = i_be[constants.BE_VCPUS]
5172 elif field == "disk_template":
5173 val = instance.disk_template
5176 val = instance.nics[0].ip
5179 elif field == "nic_mode":
5181 val = i_nicp[0][constants.NIC_MODE]
5184 elif field == "nic_link":
5186 val = i_nicp[0][constants.NIC_LINK]
5189 elif field == "bridge":
5190 if (instance.nics and
5191 i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
5192 val = i_nicp[0][constants.NIC_LINK]
5195 elif field == "mac":
5197 val = instance.nics[0].mac
5200 elif field == "custom_nicparams":
5201 val = [nic.nicparams for nic in instance.nics]
5202 elif field == "sda_size" or field == "sdb_size":
5203 idx = ord(field[2]) - ord('a')
5205 val = instance.FindDisk(idx).size
5206 except errors.OpPrereqError:
5208 elif field == "disk_usage": # total disk usage per node
5209 disk_sizes = [{'size': disk.size} for disk in instance.disks]
5210 val = _ComputeDiskSize(instance.disk_template, disk_sizes)
5211 elif field == "tags":
5212 val = list(instance.GetTags())
5213 elif field == "custom_hvparams":
5214 val = instance.hvparams # not filled!
5215 elif field == "hvparams":
5217 elif (field.startswith(HVPREFIX) and
5218 field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
5219 field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
5220 val = i_hv.get(field[len(HVPREFIX):], None)
5221 elif field == "custom_beparams":
5222 val = instance.beparams
5223 elif field == "beparams":
5225 elif (field.startswith(BEPREFIX) and
5226 field[len(BEPREFIX):] in constants.BES_PARAMETERS):
5227 val = i_be.get(field[len(BEPREFIX):], None)
5228 elif st_match and st_match.groups():
5229 # matches a variable list
5230 st_groups = st_match.groups()
5231 if st_groups and st_groups[0] == "disk":
5232 if st_groups[1] == "count":
5233 val = len(instance.disks)
5234 elif st_groups[1] == "sizes":
5235 val = [disk.size for disk in instance.disks]
5236 elif st_groups[1] == "size":
5238 val = instance.FindDisk(st_groups[2]).size
5239 except errors.OpPrereqError:
5242 assert False, "Unhandled disk parameter"
5243 elif st_groups[0] == "nic":
5244 if st_groups[1] == "count":
5245 val = len(instance.nics)
5246 elif st_groups[1] == "macs":
5247 val = [nic.mac for nic in instance.nics]
5248 elif st_groups[1] == "ips":
5249 val = [nic.ip for nic in instance.nics]
5250 elif st_groups[1] == "modes":
5251 val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
5252 elif st_groups[1] == "links":
5253 val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
5254 elif st_groups[1] == "bridges":
5257 if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
5258 val.append(nicp[constants.NIC_LINK])
5263 nic_idx = int(st_groups[2])
5264 if nic_idx >= len(instance.nics):
5267 if st_groups[1] == "mac":
5268 val = instance.nics[nic_idx].mac
5269 elif st_groups[1] == "ip":
5270 val = instance.nics[nic_idx].ip
5271 elif st_groups[1] == "mode":
5272 val = i_nicp[nic_idx][constants.NIC_MODE]
5273 elif st_groups[1] == "link":
5274 val = i_nicp[nic_idx][constants.NIC_LINK]
5275 elif st_groups[1] == "bridge":
5276 nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
5277 if nic_mode == constants.NIC_MODE_BRIDGED:
5278 val = i_nicp[nic_idx][constants.NIC_LINK]
5282 assert False, "Unhandled NIC parameter"
5284 assert False, ("Declared but unhandled variable parameter '%s'" %
5287 assert False, "Declared but unhandled parameter '%s'" % field
5294 class LUFailoverInstance(LogicalUnit):
5295 """Failover an instance.
5298 HPATH = "instance-failover"
5299 HTYPE = constants.HTYPE_INSTANCE
5302 ("ignore_consistency", False, ht.TBool),
5307 def ExpandNames(self):
5308 self._ExpandAndLockInstance()
5309 self.needed_locks[locking.LEVEL_NODE] = []
5310 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5312 def DeclareLocks(self, level):
5313 if level == locking.LEVEL_NODE:
5314 self._LockInstancesNodes()
5316 def BuildHooksEnv(self):
5319 This runs on master, primary and secondary nodes of the instance.
5322 instance = self.instance
5323 source_node = instance.primary_node
5324 target_node = instance.secondary_nodes[0]
5326 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5327 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5328 "OLD_PRIMARY": source_node,
5329 "OLD_SECONDARY": target_node,
5330 "NEW_PRIMARY": target_node,
5331 "NEW_SECONDARY": source_node,
5333 env.update(_BuildInstanceHookEnvByObject(self, instance))
5334 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5336 nl_post.append(source_node)
5337 return env, nl, nl_post
5339 def CheckPrereq(self):
5340 """Check prerequisites.
5342 This checks that the instance is in the cluster.
5345 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5346 assert self.instance is not None, \
5347 "Cannot retrieve locked instance %s" % self.op.instance_name
5349 bep = self.cfg.GetClusterInfo().FillBE(instance)
5350 if instance.disk_template not in constants.DTS_NET_MIRROR:
5351 raise errors.OpPrereqError("Instance's disk layout is not"
5352 " network mirrored, cannot failover.",
5355 secondary_nodes = instance.secondary_nodes
5356 if not secondary_nodes:
5357 raise errors.ProgrammerError("no secondary node but using "
5358 "a mirrored disk template")
5360 target_node = secondary_nodes[0]
5361 _CheckNodeOnline(self, target_node)
5362 _CheckNodeNotDrained(self, target_node)
5363 if instance.admin_up:
5364 # check memory requirements on the secondary node
5365 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5366 instance.name, bep[constants.BE_MEMORY],
5367 instance.hypervisor)
5369 self.LogInfo("Not checking memory on the secondary node as"
5370 " instance will not be started")
5372 # check bridge existance
5373 _CheckInstanceBridgesExist(self, instance, node=target_node)
5375 def Exec(self, feedback_fn):
5376 """Failover an instance.
5378 The failover is done by shutting it down on its present node and
5379 starting it on the secondary.
5382 instance = self.instance
5383 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
5385 source_node = instance.primary_node
5386 target_node = instance.secondary_nodes[0]
5388 if instance.admin_up:
5389 feedback_fn("* checking disk consistency between source and target")
5390 for dev in instance.disks:
5391 # for drbd, these are drbd over lvm
5392 if not _CheckDiskConsistency(self, dev, target_node, False):
5393 if not self.op.ignore_consistency:
5394 raise errors.OpExecError("Disk %s is degraded on target node,"
5395 " aborting failover." % dev.iv_name)
5397 feedback_fn("* not checking disk consistency as instance is not running")
5399 feedback_fn("* shutting down instance on source node")
5400 logging.info("Shutting down instance %s on node %s",
5401 instance.name, source_node)
5403 result = self.rpc.call_instance_shutdown(source_node, instance,
5404 self.op.shutdown_timeout)
5405 msg = result.fail_msg
5407 if self.op.ignore_consistency or primary_node.offline:
5408 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5409 " Proceeding anyway. Please make sure node"
5410 " %s is down. Error details: %s",
5411 instance.name, source_node, source_node, msg)
5413 raise errors.OpExecError("Could not shutdown instance %s on"
5415 (instance.name, source_node, msg))
5417 feedback_fn("* deactivating the instance's disks on source node")
5418 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5419 raise errors.OpExecError("Can't shut down the instance's disks.")
5421 instance.primary_node = target_node
5422 # distribute new instance config to the other nodes
5423 self.cfg.Update(instance, feedback_fn)
5425 # Only start the instance if it's marked as up
5426 if instance.admin_up:
5427 feedback_fn("* activating the instance's disks on target node")
5428 logging.info("Starting instance %s on node %s",
5429 instance.name, target_node)
5431 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5432 ignore_secondaries=True)
5434 _ShutdownInstanceDisks(self, instance)
5435 raise errors.OpExecError("Can't activate the instance's disks")
5437 feedback_fn("* starting the instance on the target node")
5438 result = self.rpc.call_instance_start(target_node, instance, None, None)
5439 msg = result.fail_msg
5441 _ShutdownInstanceDisks(self, instance)
5442 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5443 (instance.name, target_node, msg))
5446 class LUMigrateInstance(LogicalUnit):
5447 """Migrate an instance.
5449 This is migration without shutting down, compared to the failover,
5450 which is done with shutdown.
5453 HPATH = "instance-migrate"
5454 HTYPE = constants.HTYPE_INSTANCE
5459 ("cleanup", False, ht.TBool),
5464 def ExpandNames(self):
5465 self._ExpandAndLockInstance()
5467 self.needed_locks[locking.LEVEL_NODE] = []
5468 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5470 self._migrater = TLMigrateInstance(self, self.op.instance_name,
5472 self.tasklets = [self._migrater]
5474 def DeclareLocks(self, level):
5475 if level == locking.LEVEL_NODE:
5476 self._LockInstancesNodes()
5478 def BuildHooksEnv(self):
5481 This runs on master, primary and secondary nodes of the instance.
5484 instance = self._migrater.instance
5485 source_node = instance.primary_node
5486 target_node = instance.secondary_nodes[0]
5487 env = _BuildInstanceHookEnvByObject(self, instance)
5488 env["MIGRATE_LIVE"] = self._migrater.live
5489 env["MIGRATE_CLEANUP"] = self.op.cleanup
5491 "OLD_PRIMARY": source_node,
5492 "OLD_SECONDARY": target_node,
5493 "NEW_PRIMARY": target_node,
5494 "NEW_SECONDARY": source_node,
5496 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5498 nl_post.append(source_node)
5499 return env, nl, nl_post
5502 class LUMoveInstance(LogicalUnit):
5503 """Move an instance by data-copying.
5506 HPATH = "instance-move"
5507 HTYPE = constants.HTYPE_INSTANCE
5510 ("target_node", ht.NoDefault, ht.TNonEmptyString),
5515 def ExpandNames(self):
5516 self._ExpandAndLockInstance()
5517 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5518 self.op.target_node = target_node
5519 self.needed_locks[locking.LEVEL_NODE] = [target_node]
5520 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5522 def DeclareLocks(self, level):
5523 if level == locking.LEVEL_NODE:
5524 self._LockInstancesNodes(primary_only=True)
5526 def BuildHooksEnv(self):
5529 This runs on master, primary and secondary nodes of the instance.
5533 "TARGET_NODE": self.op.target_node,
5534 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5536 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5537 nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5538 self.op.target_node]
5541 def CheckPrereq(self):
5542 """Check prerequisites.
5544 This checks that the instance is in the cluster.
5547 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5548 assert self.instance is not None, \
5549 "Cannot retrieve locked instance %s" % self.op.instance_name
5551 node = self.cfg.GetNodeInfo(self.op.target_node)
5552 assert node is not None, \
5553 "Cannot retrieve locked node %s" % self.op.target_node
5555 self.target_node = target_node = node.name
5557 if target_node == instance.primary_node:
5558 raise errors.OpPrereqError("Instance %s is already on the node %s" %
5559 (instance.name, target_node),
5562 bep = self.cfg.GetClusterInfo().FillBE(instance)
5564 for idx, dsk in enumerate(instance.disks):
5565 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5566 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5567 " cannot copy" % idx, errors.ECODE_STATE)
5569 _CheckNodeOnline(self, target_node)
5570 _CheckNodeNotDrained(self, target_node)
5572 if instance.admin_up:
5573 # check memory requirements on the secondary node
5574 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5575 instance.name, bep[constants.BE_MEMORY],
5576 instance.hypervisor)
5578 self.LogInfo("Not checking memory on the secondary node as"
5579 " instance will not be started")
5581 # check bridge existance
5582 _CheckInstanceBridgesExist(self, instance, node=target_node)
5584 def Exec(self, feedback_fn):
5585 """Move an instance.
5587 The move is done by shutting it down on its present node, copying
5588 the data over (slow) and starting it on the new node.
5591 instance = self.instance
5593 source_node = instance.primary_node
5594 target_node = self.target_node
5596 self.LogInfo("Shutting down instance %s on source node %s",
5597 instance.name, source_node)
5599 result = self.rpc.call_instance_shutdown(source_node, instance,
5600 self.op.shutdown_timeout)
5601 msg = result.fail_msg
5603 if self.op.ignore_consistency:
5604 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5605 " Proceeding anyway. Please make sure node"
5606 " %s is down. Error details: %s",
5607 instance.name, source_node, source_node, msg)
5609 raise errors.OpExecError("Could not shutdown instance %s on"
5611 (instance.name, source_node, msg))
5613 # create the target disks
5615 _CreateDisks(self, instance, target_node=target_node)
5616 except errors.OpExecError:
5617 self.LogWarning("Device creation failed, reverting...")
5619 _RemoveDisks(self, instance, target_node=target_node)
5621 self.cfg.ReleaseDRBDMinors(instance.name)
5624 cluster_name = self.cfg.GetClusterInfo().cluster_name
5627 # activate, get path, copy the data over
5628 for idx, disk in enumerate(instance.disks):
5629 self.LogInfo("Copying data for disk %d", idx)
5630 result = self.rpc.call_blockdev_assemble(target_node, disk,
5631 instance.name, True)
5633 self.LogWarning("Can't assemble newly created disk %d: %s",
5634 idx, result.fail_msg)
5635 errs.append(result.fail_msg)
5637 dev_path = result.payload
5638 result = self.rpc.call_blockdev_export(source_node, disk,
5639 target_node, dev_path,
5642 self.LogWarning("Can't copy data over for disk %d: %s",
5643 idx, result.fail_msg)
5644 errs.append(result.fail_msg)
5648 self.LogWarning("Some disks failed to copy, aborting")
5650 _RemoveDisks(self, instance, target_node=target_node)
5652 self.cfg.ReleaseDRBDMinors(instance.name)
5653 raise errors.OpExecError("Errors during disk copy: %s" %
5656 instance.primary_node = target_node
5657 self.cfg.Update(instance, feedback_fn)
5659 self.LogInfo("Removing the disks on the original node")
5660 _RemoveDisks(self, instance, target_node=source_node)
5662 # Only start the instance if it's marked as up
5663 if instance.admin_up:
5664 self.LogInfo("Starting instance %s on node %s",
5665 instance.name, target_node)
5667 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5668 ignore_secondaries=True)
5670 _ShutdownInstanceDisks(self, instance)
5671 raise errors.OpExecError("Can't activate the instance's disks")
5673 result = self.rpc.call_instance_start(target_node, instance, None, None)
5674 msg = result.fail_msg
5676 _ShutdownInstanceDisks(self, instance)
5677 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5678 (instance.name, target_node, msg))
5681 class LUMigrateNode(LogicalUnit):
5682 """Migrate all instances from a node.
5685 HPATH = "node-migrate"
5686 HTYPE = constants.HTYPE_NODE
5694 def ExpandNames(self):
5695 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5697 self.needed_locks = {
5698 locking.LEVEL_NODE: [self.op.node_name],
5701 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5703 # Create tasklets for migrating instances for all instances on this node
5707 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5708 logging.debug("Migrating instance %s", inst.name)
5709 names.append(inst.name)
5711 tasklets.append(TLMigrateInstance(self, inst.name, False))
5713 self.tasklets = tasklets
5715 # Declare instance locks
5716 self.needed_locks[locking.LEVEL_INSTANCE] = names
5718 def DeclareLocks(self, level):
5719 if level == locking.LEVEL_NODE:
5720 self._LockInstancesNodes()
5722 def BuildHooksEnv(self):
5725 This runs on the master, the primary and all the secondaries.
5729 "NODE_NAME": self.op.node_name,
5732 nl = [self.cfg.GetMasterNode()]
5734 return (env, nl, nl)
5737 class TLMigrateInstance(Tasklet):
5738 """Tasklet class for instance migration.
5741 @ivar live: whether the migration will be done live or non-live;
5742 this variable is initalized only after CheckPrereq has run
5745 def __init__(self, lu, instance_name, cleanup):
5746 """Initializes this class.
5749 Tasklet.__init__(self, lu)
5752 self.instance_name = instance_name
5753 self.cleanup = cleanup
5754 self.live = False # will be overridden later
5756 def CheckPrereq(self):
5757 """Check prerequisites.
5759 This checks that the instance is in the cluster.
5762 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5763 instance = self.cfg.GetInstanceInfo(instance_name)
5764 assert instance is not None
5766 if instance.disk_template != constants.DT_DRBD8:
5767 raise errors.OpPrereqError("Instance's disk layout is not"
5768 " drbd8, cannot migrate.", errors.ECODE_STATE)
5770 secondary_nodes = instance.secondary_nodes
5771 if not secondary_nodes:
5772 raise errors.ConfigurationError("No secondary node but using"
5773 " drbd8 disk template")
5775 i_be = self.cfg.GetClusterInfo().FillBE(instance)
5777 target_node = secondary_nodes[0]
5778 # check memory requirements on the secondary node
5779 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
5780 instance.name, i_be[constants.BE_MEMORY],
5781 instance.hypervisor)
5783 # check bridge existance
5784 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
5786 if not self.cleanup:
5787 _CheckNodeNotDrained(self.lu, target_node)
5788 result = self.rpc.call_instance_migratable(instance.primary_node,
5790 result.Raise("Can't migrate, please use failover",
5791 prereq=True, ecode=errors.ECODE_STATE)
5793 self.instance = instance
5795 if self.lu.op.live is not None and self.lu.op.mode is not None:
5796 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
5797 " parameters are accepted",
5799 if self.lu.op.live is not None:
5801 self.lu.op.mode = constants.HT_MIGRATION_LIVE
5803 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
5804 # reset the 'live' parameter to None so that repeated
5805 # invocations of CheckPrereq do not raise an exception
5806 self.lu.op.live = None
5807 elif self.lu.op.mode is None:
5808 # read the default value from the hypervisor
5809 i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
5810 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
5812 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
5814 def _WaitUntilSync(self):
5815 """Poll with custom rpc for disk sync.
5817 This uses our own step-based rpc call.
5820 self.feedback_fn("* wait until resync is done")
5824 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5826 self.instance.disks)
5828 for node, nres in result.items():
5829 nres.Raise("Cannot resync disks on node %s" % node)
5830 node_done, node_percent = nres.payload
5831 all_done = all_done and node_done
5832 if node_percent is not None:
5833 min_percent = min(min_percent, node_percent)
5835 if min_percent < 100:
5836 self.feedback_fn(" - progress: %.1f%%" % min_percent)
5839 def _EnsureSecondary(self, node):
5840 """Demote a node to secondary.
5843 self.feedback_fn("* switching node %s to secondary mode" % node)
5845 for dev in self.instance.disks:
5846 self.cfg.SetDiskID(dev, node)
5848 result = self.rpc.call_blockdev_close(node, self.instance.name,
5849 self.instance.disks)
5850 result.Raise("Cannot change disk to secondary on node %s" % node)
5852 def _GoStandalone(self):
5853 """Disconnect from the network.
5856 self.feedback_fn("* changing into standalone mode")
5857 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5858 self.instance.disks)
5859 for node, nres in result.items():
5860 nres.Raise("Cannot disconnect disks node %s" % node)
5862 def _GoReconnect(self, multimaster):
5863 """Reconnect to the network.
5869 msg = "single-master"
5870 self.feedback_fn("* changing disks into %s mode" % msg)
5871 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5872 self.instance.disks,
5873 self.instance.name, multimaster)
5874 for node, nres in result.items():
5875 nres.Raise("Cannot change disks config on node %s" % node)
5877 def _ExecCleanup(self):
5878 """Try to cleanup after a failed migration.
5880 The cleanup is done by:
5881 - check that the instance is running only on one node
5882 (and update the config if needed)
5883 - change disks on its secondary node to secondary
5884 - wait until disks are fully synchronized
5885 - disconnect from the network
5886 - change disks into single-master mode
5887 - wait again until disks are fully synchronized
5890 instance = self.instance
5891 target_node = self.target_node
5892 source_node = self.source_node
5894 # check running on only one node
5895 self.feedback_fn("* checking where the instance actually runs"
5896 " (if this hangs, the hypervisor might be in"
5898 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5899 for node, result in ins_l.items():
5900 result.Raise("Can't contact node %s" % node)
5902 runningon_source = instance.name in ins_l[source_node].payload
5903 runningon_target = instance.name in ins_l[target_node].payload
5905 if runningon_source and runningon_target:
5906 raise errors.OpExecError("Instance seems to be running on two nodes,"
5907 " or the hypervisor is confused. You will have"
5908 " to ensure manually that it runs only on one"
5909 " and restart this operation.")
5911 if not (runningon_source or runningon_target):
5912 raise errors.OpExecError("Instance does not seem to be running at all."
5913 " In this case, it's safer to repair by"
5914 " running 'gnt-instance stop' to ensure disk"
5915 " shutdown, and then restarting it.")
5917 if runningon_target:
5918 # the migration has actually succeeded, we need to update the config
5919 self.feedback_fn("* instance running on secondary node (%s),"
5920 " updating config" % target_node)
5921 instance.primary_node = target_node
5922 self.cfg.Update(instance, self.feedback_fn)
5923 demoted_node = source_node
5925 self.feedback_fn("* instance confirmed to be running on its"
5926 " primary node (%s)" % source_node)
5927 demoted_node = target_node
5929 self._EnsureSecondary(demoted_node)
5931 self._WaitUntilSync()
5932 except errors.OpExecError:
5933 # we ignore here errors, since if the device is standalone, it
5934 # won't be able to sync
5936 self._GoStandalone()
5937 self._GoReconnect(False)
5938 self._WaitUntilSync()
5940 self.feedback_fn("* done")
5942 def _RevertDiskStatus(self):
5943 """Try to revert the disk status after a failed migration.
5946 target_node = self.target_node
5948 self._EnsureSecondary(target_node)
5949 self._GoStandalone()
5950 self._GoReconnect(False)
5951 self._WaitUntilSync()
5952 except errors.OpExecError, err:
5953 self.lu.LogWarning("Migration failed and I can't reconnect the"
5954 " drives: error '%s'\n"
5955 "Please look and recover the instance status" %
5958 def _AbortMigration(self):
5959 """Call the hypervisor code to abort a started migration.
5962 instance = self.instance
5963 target_node = self.target_node
5964 migration_info = self.migration_info
5966 abort_result = self.rpc.call_finalize_migration(target_node,
5970 abort_msg = abort_result.fail_msg
5972 logging.error("Aborting migration failed on target node %s: %s",
5973 target_node, abort_msg)
5974 # Don't raise an exception here, as we stil have to try to revert the
5975 # disk status, even if this step failed.
5977 def _ExecMigration(self):
5978 """Migrate an instance.
5980 The migrate is done by:
5981 - change the disks into dual-master mode
5982 - wait until disks are fully synchronized again
5983 - migrate the instance
5984 - change disks on the new secondary node (the old primary) to secondary
5985 - wait until disks are fully synchronized
5986 - change disks into single-master mode
5989 instance = self.instance
5990 target_node = self.target_node
5991 source_node = self.source_node
5993 self.feedback_fn("* checking disk consistency between source and target")
5994 for dev in instance.disks:
5995 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
5996 raise errors.OpExecError("Disk %s is degraded or not fully"
5997 " synchronized on target node,"
5998 " aborting migrate." % dev.iv_name)
6000 # First get the migration information from the remote node
6001 result = self.rpc.call_migration_info(source_node, instance)
6002 msg = result.fail_msg
6004 log_err = ("Failed fetching source migration information from %s: %s" %
6006 logging.error(log_err)
6007 raise errors.OpExecError(log_err)
6009 self.migration_info = migration_info = result.payload
6011 # Then switch the disks to master/master mode
6012 self._EnsureSecondary(target_node)
6013 self._GoStandalone()
6014 self._GoReconnect(True)
6015 self._WaitUntilSync()
6017 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6018 result = self.rpc.call_accept_instance(target_node,
6021 self.nodes_ip[target_node])
6023 msg = result.fail_msg
6025 logging.error("Instance pre-migration failed, trying to revert"
6026 " disk status: %s", msg)
6027 self.feedback_fn("Pre-migration failed, aborting")
6028 self._AbortMigration()
6029 self._RevertDiskStatus()
6030 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6031 (instance.name, msg))
6033 self.feedback_fn("* migrating instance to %s" % target_node)
6035 result = self.rpc.call_instance_migrate(source_node, instance,
6036 self.nodes_ip[target_node],
6038 msg = result.fail_msg
6040 logging.error("Instance migration failed, trying to revert"
6041 " disk status: %s", msg)
6042 self.feedback_fn("Migration failed, aborting")
6043 self._AbortMigration()
6044 self._RevertDiskStatus()
6045 raise errors.OpExecError("Could not migrate instance %s: %s" %
6046 (instance.name, msg))
6049 instance.primary_node = target_node
6050 # distribute new instance config to the other nodes
6051 self.cfg.Update(instance, self.feedback_fn)
6053 result = self.rpc.call_finalize_migration(target_node,
6057 msg = result.fail_msg
6059 logging.error("Instance migration succeeded, but finalization failed:"
6061 raise errors.OpExecError("Could not finalize instance migration: %s" %
6064 self._EnsureSecondary(source_node)
6065 self._WaitUntilSync()
6066 self._GoStandalone()
6067 self._GoReconnect(False)
6068 self._WaitUntilSync()
6070 self.feedback_fn("* done")
6072 def Exec(self, feedback_fn):
6073 """Perform the migration.
6076 feedback_fn("Migrating instance %s" % self.instance.name)
6078 self.feedback_fn = feedback_fn
6080 self.source_node = self.instance.primary_node
6081 self.target_node = self.instance.secondary_nodes[0]
6082 self.all_nodes = [self.source_node, self.target_node]
6084 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6085 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6089 return self._ExecCleanup()
6091 return self._ExecMigration()
6094 def _CreateBlockDev(lu, node, instance, device, force_create,
6096 """Create a tree of block devices on a given node.
6098 If this device type has to be created on secondaries, create it and
6101 If not, just recurse to children keeping the same 'force' value.
6103 @param lu: the lu on whose behalf we execute
6104 @param node: the node on which to create the device
6105 @type instance: L{objects.Instance}
6106 @param instance: the instance which owns the device
6107 @type device: L{objects.Disk}
6108 @param device: the device to create
6109 @type force_create: boolean
6110 @param force_create: whether to force creation of this device; this
6111 will be change to True whenever we find a device which has
6112 CreateOnSecondary() attribute
6113 @param info: the extra 'metadata' we should attach to the device
6114 (this will be represented as a LVM tag)
6115 @type force_open: boolean
6116 @param force_open: this parameter will be passes to the
6117 L{backend.BlockdevCreate} function where it specifies
6118 whether we run on primary or not, and it affects both
6119 the child assembly and the device own Open() execution
6122 if device.CreateOnSecondary():
6126 for child in device.children:
6127 _CreateBlockDev(lu, node, instance, child, force_create,
6130 if not force_create:
6133 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6136 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6137 """Create a single block device on a given node.
6139 This will not recurse over children of the device, so they must be
6142 @param lu: the lu on whose behalf we execute
6143 @param node: the node on which to create the device
6144 @type instance: L{objects.Instance}
6145 @param instance: the instance which owns the device
6146 @type device: L{objects.Disk}
6147 @param device: the device to create
6148 @param info: the extra 'metadata' we should attach to the device
6149 (this will be represented as a LVM tag)
6150 @type force_open: boolean
6151 @param force_open: this parameter will be passes to the
6152 L{backend.BlockdevCreate} function where it specifies
6153 whether we run on primary or not, and it affects both
6154 the child assembly and the device own Open() execution
6157 lu.cfg.SetDiskID(device, node)
6158 result = lu.rpc.call_blockdev_create(node, device, device.size,
6159 instance.name, force_open, info)
6160 result.Raise("Can't create block device %s on"
6161 " node %s for instance %s" % (device, node, instance.name))
6162 if device.physical_id is None:
6163 device.physical_id = result.payload
6166 def _GenerateUniqueNames(lu, exts):
6167 """Generate a suitable LV name.
6169 This will generate a logical volume name for the given instance.
6174 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6175 results.append("%s%s" % (new_id, val))
6179 def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
6181 """Generate a drbd8 device complete with its children.
6184 port = lu.cfg.AllocatePort()
6185 vgname = lu.cfg.GetVGName()
6186 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6187 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6188 logical_id=(vgname, names[0]))
6189 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6190 logical_id=(vgname, names[1]))
6191 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6192 logical_id=(primary, secondary, port,
6195 children=[dev_data, dev_meta],
6200 def _GenerateDiskTemplate(lu, template_name,
6201 instance_name, primary_node,
6202 secondary_nodes, disk_info,
6203 file_storage_dir, file_driver,
6205 """Generate the entire disk layout for a given template type.
6208 #TODO: compute space requirements
6210 vgname = lu.cfg.GetVGName()
6211 disk_count = len(disk_info)
6213 if template_name == constants.DT_DISKLESS:
6215 elif template_name == constants.DT_PLAIN:
6216 if len(secondary_nodes) != 0:
6217 raise errors.ProgrammerError("Wrong template configuration")
6219 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6220 for i in range(disk_count)])
6221 for idx, disk in enumerate(disk_info):
6222 disk_index = idx + base_index
6223 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6224 logical_id=(vgname, names[idx]),
6225 iv_name="disk/%d" % disk_index,
6227 disks.append(disk_dev)
6228 elif template_name == constants.DT_DRBD8:
6229 if len(secondary_nodes) != 1:
6230 raise errors.ProgrammerError("Wrong template configuration")
6231 remote_node = secondary_nodes[0]
6232 minors = lu.cfg.AllocateDRBDMinor(
6233 [primary_node, remote_node] * len(disk_info), instance_name)
6236 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6237 for i in range(disk_count)]):
6238 names.append(lv_prefix + "_data")
6239 names.append(lv_prefix + "_meta")
6240 for idx, disk in enumerate(disk_info):
6241 disk_index = idx + base_index
6242 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6243 disk["size"], names[idx*2:idx*2+2],
6244 "disk/%d" % disk_index,
6245 minors[idx*2], minors[idx*2+1])
6246 disk_dev.mode = disk["mode"]
6247 disks.append(disk_dev)
6248 elif template_name == constants.DT_FILE:
6249 if len(secondary_nodes) != 0:
6250 raise errors.ProgrammerError("Wrong template configuration")
6252 _RequireFileStorage()
6254 for idx, disk in enumerate(disk_info):
6255 disk_index = idx + base_index
6256 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6257 iv_name="disk/%d" % disk_index,
6258 logical_id=(file_driver,
6259 "%s/disk%d" % (file_storage_dir,
6262 disks.append(disk_dev)
6264 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6268 def _GetInstanceInfoText(instance):
6269 """Compute that text that should be added to the disk's metadata.
6272 return "originstname+%s" % instance.name
6275 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6276 """Create all disks for an instance.
6278 This abstracts away some work from AddInstance.
6280 @type lu: L{LogicalUnit}
6281 @param lu: the logical unit on whose behalf we execute
6282 @type instance: L{objects.Instance}
6283 @param instance: the instance whose disks we should create
6285 @param to_skip: list of indices to skip
6286 @type target_node: string
6287 @param target_node: if passed, overrides the target node for creation
6289 @return: the success of the creation
6292 info = _GetInstanceInfoText(instance)
6293 if target_node is None:
6294 pnode = instance.primary_node
6295 all_nodes = instance.all_nodes
6300 if instance.disk_template == constants.DT_FILE:
6301 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6302 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6304 result.Raise("Failed to create directory '%s' on"
6305 " node %s" % (file_storage_dir, pnode))
6307 # Note: this needs to be kept in sync with adding of disks in
6308 # LUSetInstanceParams
6309 for idx, device in enumerate(instance.disks):
6310 if to_skip and idx in to_skip:
6312 logging.info("Creating volume %s for instance %s",
6313 device.iv_name, instance.name)
6315 for node in all_nodes:
6316 f_create = node == pnode
6317 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6320 def _RemoveDisks(lu, instance, target_node=None):
6321 """Remove all disks for an instance.
6323 This abstracts away some work from `AddInstance()` and
6324 `RemoveInstance()`. Note that in case some of the devices couldn't
6325 be removed, the removal will continue with the other ones (compare
6326 with `_CreateDisks()`).
6328 @type lu: L{LogicalUnit}
6329 @param lu: the logical unit on whose behalf we execute
6330 @type instance: L{objects.Instance}
6331 @param instance: the instance whose disks we should remove
6332 @type target_node: string
6333 @param target_node: used to override the node on which to remove the disks
6335 @return: the success of the removal
6338 logging.info("Removing block devices for instance %s", instance.name)
6341 for device in instance.disks:
6343 edata = [(target_node, device)]
6345 edata = device.ComputeNodeTree(instance.primary_node)
6346 for node, disk in edata:
6347 lu.cfg.SetDiskID(disk, node)
6348 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6350 lu.LogWarning("Could not remove block device %s on node %s,"
6351 " continuing anyway: %s", device.iv_name, node, msg)
6354 if instance.disk_template == constants.DT_FILE:
6355 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6359 tgt = instance.primary_node
6360 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6362 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6363 file_storage_dir, instance.primary_node, result.fail_msg)
6369 def _ComputeDiskSize(disk_template, disks):
6370 """Compute disk size requirements in the volume group
6373 # Required free disk space as a function of disk and swap space
6375 constants.DT_DISKLESS: None,
6376 constants.DT_PLAIN: sum(d["size"] for d in disks),
6377 # 128 MB are added for drbd metadata for each disk
6378 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6379 constants.DT_FILE: None,
6382 if disk_template not in req_size_dict:
6383 raise errors.ProgrammerError("Disk template '%s' size requirement"
6384 " is unknown" % disk_template)
6386 return req_size_dict[disk_template]
6389 def _CheckHVParams(lu, nodenames, hvname, hvparams):
6390 """Hypervisor parameter validation.
6392 This function abstract the hypervisor parameter validation to be
6393 used in both instance create and instance modify.
6395 @type lu: L{LogicalUnit}
6396 @param lu: the logical unit for which we check
6397 @type nodenames: list
6398 @param nodenames: the list of nodes on which we should check
6399 @type hvname: string
6400 @param hvname: the name of the hypervisor we should use
6401 @type hvparams: dict
6402 @param hvparams: the parameters which we need to check
6403 @raise errors.OpPrereqError: if the parameters are not valid
6406 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6409 for node in nodenames:
6413 info.Raise("Hypervisor parameter validation failed on node %s" % node)
6416 def _CheckOSParams(lu, required, nodenames, osname, osparams):
6417 """OS parameters validation.
6419 @type lu: L{LogicalUnit}
6420 @param lu: the logical unit for which we check
6421 @type required: boolean
6422 @param required: whether the validation should fail if the OS is not
6424 @type nodenames: list
6425 @param nodenames: the list of nodes on which we should check
6426 @type osname: string
6427 @param osname: the name of the hypervisor we should use
6428 @type osparams: dict
6429 @param osparams: the parameters which we need to check
6430 @raise errors.OpPrereqError: if the parameters are not valid
6433 result = lu.rpc.call_os_validate(required, nodenames, osname,
6434 [constants.OS_VALIDATE_PARAMETERS],
6436 for node, nres in result.items():
6437 # we don't check for offline cases since this should be run only
6438 # against the master node and/or an instance's nodes
6439 nres.Raise("OS Parameters validation failed on node %s" % node)
6440 if not nres.payload:
6441 lu.LogInfo("OS %s not found on node %s, validation skipped",
6445 class LUCreateInstance(LogicalUnit):
6446 """Create an instance.
6449 HPATH = "instance-add"
6450 HTYPE = constants.HTYPE_INSTANCE
6453 ("mode", ht.NoDefault, ht.TElemOf(constants.INSTANCE_CREATE_MODES)),
6454 ("start", True, ht.TBool),
6455 ("wait_for_sync", True, ht.TBool),
6456 ("ip_check", True, ht.TBool),
6457 ("name_check", True, ht.TBool),
6458 ("disks", ht.NoDefault, ht.TListOf(ht.TDict)),
6459 ("nics", ht.NoDefault, ht.TListOf(ht.TDict)),
6460 ("hvparams", ht.EmptyDict, ht.TDict),
6461 ("beparams", ht.EmptyDict, ht.TDict),
6462 ("osparams", ht.EmptyDict, ht.TDict),
6463 ("no_install", None, ht.TMaybeBool),
6464 ("os_type", None, ht.TMaybeString),
6465 ("force_variant", False, ht.TBool),
6466 ("source_handshake", None, ht.TOr(ht.TList, ht.TNone)),
6467 ("source_x509_ca", None, ht.TMaybeString),
6468 ("source_instance_name", None, ht.TMaybeString),
6469 ("src_node", None, ht.TMaybeString),
6470 ("src_path", None, ht.TMaybeString),
6471 ("pnode", None, ht.TMaybeString),
6472 ("snode", None, ht.TMaybeString),
6473 ("iallocator", None, ht.TMaybeString),
6474 ("hypervisor", None, ht.TMaybeString),
6475 ("disk_template", ht.NoDefault, _CheckDiskTemplate),
6476 ("identify_defaults", False, ht.TBool),
6477 ("file_driver", None, ht.TOr(ht.TNone, ht.TElemOf(constants.FILE_DRIVER))),
6478 ("file_storage_dir", None, ht.TMaybeString),
6482 def CheckArguments(self):
6486 # do not require name_check to ease forward/backward compatibility
6488 if self.op.no_install and self.op.start:
6489 self.LogInfo("No-installation mode selected, disabling startup")
6490 self.op.start = False
6491 # validate/normalize the instance name
6492 self.op.instance_name = \
6493 netutils.Hostname.GetNormalizedName(self.op.instance_name)
6495 if self.op.ip_check and not self.op.name_check:
6496 # TODO: make the ip check more flexible and not depend on the name check
6497 raise errors.OpPrereqError("Cannot do ip check without a name check",
6500 # check nics' parameter names
6501 for nic in self.op.nics:
6502 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6504 # check disks. parameter names and consistent adopt/no-adopt strategy
6505 has_adopt = has_no_adopt = False
6506 for disk in self.op.disks:
6507 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6512 if has_adopt and has_no_adopt:
6513 raise errors.OpPrereqError("Either all disks are adopted or none is",
6516 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6517 raise errors.OpPrereqError("Disk adoption is not supported for the"
6518 " '%s' disk template" %
6519 self.op.disk_template,
6521 if self.op.iallocator is not None:
6522 raise errors.OpPrereqError("Disk adoption not allowed with an"
6523 " iallocator script", errors.ECODE_INVAL)
6524 if self.op.mode == constants.INSTANCE_IMPORT:
6525 raise errors.OpPrereqError("Disk adoption not allowed for"
6526 " instance import", errors.ECODE_INVAL)
6528 self.adopt_disks = has_adopt
6530 # instance name verification
6531 if self.op.name_check:
6532 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
6533 self.op.instance_name = self.hostname1.name
6534 # used in CheckPrereq for ip ping check
6535 self.check_ip = self.hostname1.ip
6536 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6537 raise errors.OpPrereqError("Remote imports require names to be checked" %
6540 self.check_ip = None
6542 # file storage checks
6543 if (self.op.file_driver and
6544 not self.op.file_driver in constants.FILE_DRIVER):
6545 raise errors.OpPrereqError("Invalid file driver name '%s'" %
6546 self.op.file_driver, errors.ECODE_INVAL)
6548 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6549 raise errors.OpPrereqError("File storage directory path not absolute",
6552 ### Node/iallocator related checks
6553 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6555 if self.op.pnode is not None:
6556 if self.op.disk_template in constants.DTS_NET_MIRROR:
6557 if self.op.snode is None:
6558 raise errors.OpPrereqError("The networked disk templates need"
6559 " a mirror node", errors.ECODE_INVAL)
6561 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
6563 self.op.snode = None
6565 self._cds = _GetClusterDomainSecret()
6567 if self.op.mode == constants.INSTANCE_IMPORT:
6568 # On import force_variant must be True, because if we forced it at
6569 # initial install, our only chance when importing it back is that it
6571 self.op.force_variant = True
6573 if self.op.no_install:
6574 self.LogInfo("No-installation mode has no effect during import")
6576 elif self.op.mode == constants.INSTANCE_CREATE:
6577 if self.op.os_type is None:
6578 raise errors.OpPrereqError("No guest OS specified",
6580 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
6581 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
6582 " installation" % self.op.os_type,
6584 if self.op.disk_template is None:
6585 raise errors.OpPrereqError("No disk template specified",
6588 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6589 # Check handshake to ensure both clusters have the same domain secret
6590 src_handshake = self.op.source_handshake
6591 if not src_handshake:
6592 raise errors.OpPrereqError("Missing source handshake",
6595 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6598 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6601 # Load and check source CA
6602 self.source_x509_ca_pem = self.op.source_x509_ca
6603 if not self.source_x509_ca_pem:
6604 raise errors.OpPrereqError("Missing source X509 CA",
6608 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6610 except OpenSSL.crypto.Error, err:
6611 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6612 (err, ), errors.ECODE_INVAL)
6614 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6615 if errcode is not None:
6616 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6619 self.source_x509_ca = cert
6621 src_instance_name = self.op.source_instance_name
6622 if not src_instance_name:
6623 raise errors.OpPrereqError("Missing source instance name",
6626 self.source_instance_name = \
6627 netutils.GetHostname(name=src_instance_name).name
6630 raise errors.OpPrereqError("Invalid instance creation mode %r" %
6631 self.op.mode, errors.ECODE_INVAL)
6633 def ExpandNames(self):
6634 """ExpandNames for CreateInstance.
6636 Figure out the right locks for instance creation.
6639 self.needed_locks = {}
6641 instance_name = self.op.instance_name
6642 # this is just a preventive check, but someone might still add this
6643 # instance in the meantime, and creation will fail at lock-add time
6644 if instance_name in self.cfg.GetInstanceList():
6645 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6646 instance_name, errors.ECODE_EXISTS)
6648 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6650 if self.op.iallocator:
6651 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6653 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6654 nodelist = [self.op.pnode]
6655 if self.op.snode is not None:
6656 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6657 nodelist.append(self.op.snode)
6658 self.needed_locks[locking.LEVEL_NODE] = nodelist
6660 # in case of import lock the source node too
6661 if self.op.mode == constants.INSTANCE_IMPORT:
6662 src_node = self.op.src_node
6663 src_path = self.op.src_path
6665 if src_path is None:
6666 self.op.src_path = src_path = self.op.instance_name
6668 if src_node is None:
6669 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6670 self.op.src_node = None
6671 if os.path.isabs(src_path):
6672 raise errors.OpPrereqError("Importing an instance from an absolute"
6673 " path requires a source node option.",
6676 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6677 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6678 self.needed_locks[locking.LEVEL_NODE].append(src_node)
6679 if not os.path.isabs(src_path):
6680 self.op.src_path = src_path = \
6681 utils.PathJoin(constants.EXPORT_DIR, src_path)
6683 def _RunAllocator(self):
6684 """Run the allocator based on input opcode.
6687 nics = [n.ToDict() for n in self.nics]
6688 ial = IAllocator(self.cfg, self.rpc,
6689 mode=constants.IALLOCATOR_MODE_ALLOC,
6690 name=self.op.instance_name,
6691 disk_template=self.op.disk_template,
6694 vcpus=self.be_full[constants.BE_VCPUS],
6695 mem_size=self.be_full[constants.BE_MEMORY],
6698 hypervisor=self.op.hypervisor,
6701 ial.Run(self.op.iallocator)
6704 raise errors.OpPrereqError("Can't compute nodes using"
6705 " iallocator '%s': %s" %
6706 (self.op.iallocator, ial.info),
6708 if len(ial.result) != ial.required_nodes:
6709 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6710 " of nodes (%s), required %s" %
6711 (self.op.iallocator, len(ial.result),
6712 ial.required_nodes), errors.ECODE_FAULT)
6713 self.op.pnode = ial.result[0]
6714 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6715 self.op.instance_name, self.op.iallocator,
6716 utils.CommaJoin(ial.result))
6717 if ial.required_nodes == 2:
6718 self.op.snode = ial.result[1]
6720 def BuildHooksEnv(self):
6723 This runs on master, primary and secondary nodes of the instance.
6727 "ADD_MODE": self.op.mode,
6729 if self.op.mode == constants.INSTANCE_IMPORT:
6730 env["SRC_NODE"] = self.op.src_node
6731 env["SRC_PATH"] = self.op.src_path
6732 env["SRC_IMAGES"] = self.src_images
6734 env.update(_BuildInstanceHookEnv(
6735 name=self.op.instance_name,
6736 primary_node=self.op.pnode,
6737 secondary_nodes=self.secondaries,
6738 status=self.op.start,
6739 os_type=self.op.os_type,
6740 memory=self.be_full[constants.BE_MEMORY],
6741 vcpus=self.be_full[constants.BE_VCPUS],
6742 nics=_NICListToTuple(self, self.nics),
6743 disk_template=self.op.disk_template,
6744 disks=[(d["size"], d["mode"]) for d in self.disks],
6747 hypervisor_name=self.op.hypervisor,
6750 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6754 def _ReadExportInfo(self):
6755 """Reads the export information from disk.
6757 It will override the opcode source node and path with the actual
6758 information, if these two were not specified before.
6760 @return: the export information
6763 assert self.op.mode == constants.INSTANCE_IMPORT
6765 src_node = self.op.src_node
6766 src_path = self.op.src_path
6768 if src_node is None:
6769 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6770 exp_list = self.rpc.call_export_list(locked_nodes)
6772 for node in exp_list:
6773 if exp_list[node].fail_msg:
6775 if src_path in exp_list[node].payload:
6777 self.op.src_node = src_node = node
6778 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6782 raise errors.OpPrereqError("No export found for relative path %s" %
6783 src_path, errors.ECODE_INVAL)
6785 _CheckNodeOnline(self, src_node)
6786 result = self.rpc.call_export_info(src_node, src_path)
6787 result.Raise("No export or invalid export found in dir %s" % src_path)
6789 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6790 if not export_info.has_section(constants.INISECT_EXP):
6791 raise errors.ProgrammerError("Corrupted export config",
6792 errors.ECODE_ENVIRON)
6794 ei_version = export_info.get(constants.INISECT_EXP, "version")
6795 if (int(ei_version) != constants.EXPORT_VERSION):
6796 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6797 (ei_version, constants.EXPORT_VERSION),
6798 errors.ECODE_ENVIRON)
6801 def _ReadExportParams(self, einfo):
6802 """Use export parameters as defaults.
6804 In case the opcode doesn't specify (as in override) some instance
6805 parameters, then try to use them from the export information, if
6809 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6811 if self.op.disk_template is None:
6812 if einfo.has_option(constants.INISECT_INS, "disk_template"):
6813 self.op.disk_template = einfo.get(constants.INISECT_INS,
6816 raise errors.OpPrereqError("No disk template specified and the export"
6817 " is missing the disk_template information",
6820 if not self.op.disks:
6821 if einfo.has_option(constants.INISECT_INS, "disk_count"):
6823 # TODO: import the disk iv_name too
6824 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6825 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6826 disks.append({"size": disk_sz})
6827 self.op.disks = disks
6829 raise errors.OpPrereqError("No disk info specified and the export"
6830 " is missing the disk information",
6833 if (not self.op.nics and
6834 einfo.has_option(constants.INISECT_INS, "nic_count")):
6836 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6838 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6839 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6844 if (self.op.hypervisor is None and
6845 einfo.has_option(constants.INISECT_INS, "hypervisor")):
6846 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6847 if einfo.has_section(constants.INISECT_HYP):
6848 # use the export parameters but do not override the ones
6849 # specified by the user
6850 for name, value in einfo.items(constants.INISECT_HYP):
6851 if name not in self.op.hvparams:
6852 self.op.hvparams[name] = value
6854 if einfo.has_section(constants.INISECT_BEP):
6855 # use the parameters, without overriding
6856 for name, value in einfo.items(constants.INISECT_BEP):
6857 if name not in self.op.beparams:
6858 self.op.beparams[name] = value
6860 # try to read the parameters old style, from the main section
6861 for name in constants.BES_PARAMETERS:
6862 if (name not in self.op.beparams and
6863 einfo.has_option(constants.INISECT_INS, name)):
6864 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6866 if einfo.has_section(constants.INISECT_OSP):
6867 # use the parameters, without overriding
6868 for name, value in einfo.items(constants.INISECT_OSP):
6869 if name not in self.op.osparams:
6870 self.op.osparams[name] = value
6872 def _RevertToDefaults(self, cluster):
6873 """Revert the instance parameters to the default values.
6877 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
6878 for name in self.op.hvparams.keys():
6879 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
6880 del self.op.hvparams[name]
6882 be_defs = cluster.SimpleFillBE({})
6883 for name in self.op.beparams.keys():
6884 if name in be_defs and be_defs[name] == self.op.beparams[name]:
6885 del self.op.beparams[name]
6887 nic_defs = cluster.SimpleFillNIC({})
6888 for nic in self.op.nics:
6889 for name in constants.NICS_PARAMETERS:
6890 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
6893 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
6894 for name in self.op.osparams.keys():
6895 if name in os_defs and os_defs[name] == self.op.osparams[name]:
6896 del self.op.osparams[name]
6898 def CheckPrereq(self):
6899 """Check prerequisites.
6902 if self.op.mode == constants.INSTANCE_IMPORT:
6903 export_info = self._ReadExportInfo()
6904 self._ReadExportParams(export_info)
6906 _CheckDiskTemplate(self.op.disk_template)
6908 if (not self.cfg.GetVGName() and
6909 self.op.disk_template not in constants.DTS_NOT_LVM):
6910 raise errors.OpPrereqError("Cluster does not support lvm-based"
6911 " instances", errors.ECODE_STATE)
6913 if self.op.hypervisor is None:
6914 self.op.hypervisor = self.cfg.GetHypervisorType()
6916 cluster = self.cfg.GetClusterInfo()
6917 enabled_hvs = cluster.enabled_hypervisors
6918 if self.op.hypervisor not in enabled_hvs:
6919 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
6920 " cluster (%s)" % (self.op.hypervisor,
6921 ",".join(enabled_hvs)),
6924 # check hypervisor parameter syntax (locally)
6925 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6926 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
6928 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
6929 hv_type.CheckParameterSyntax(filled_hvp)
6930 self.hv_full = filled_hvp
6931 # check that we don't specify global parameters on an instance
6932 _CheckGlobalHvParams(self.op.hvparams)
6934 # fill and remember the beparams dict
6935 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6936 self.be_full = cluster.SimpleFillBE(self.op.beparams)
6938 # build os parameters
6939 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
6941 # now that hvp/bep are in final format, let's reset to defaults,
6943 if self.op.identify_defaults:
6944 self._RevertToDefaults(cluster)
6948 for idx, nic in enumerate(self.op.nics):
6949 nic_mode_req = nic.get("mode", None)
6950 nic_mode = nic_mode_req
6951 if nic_mode is None:
6952 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
6954 # in routed mode, for the first nic, the default ip is 'auto'
6955 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
6956 default_ip_mode = constants.VALUE_AUTO
6958 default_ip_mode = constants.VALUE_NONE
6960 # ip validity checks
6961 ip = nic.get("ip", default_ip_mode)
6962 if ip is None or ip.lower() == constants.VALUE_NONE:
6964 elif ip.lower() == constants.VALUE_AUTO:
6965 if not self.op.name_check:
6966 raise errors.OpPrereqError("IP address set to auto but name checks"
6967 " have been skipped",
6969 nic_ip = self.hostname1.ip
6971 if not netutils.IPAddress.IsValid(ip):
6972 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
6976 # TODO: check the ip address for uniqueness
6977 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
6978 raise errors.OpPrereqError("Routed nic mode requires an ip address",
6981 # MAC address verification
6982 mac = nic.get("mac", constants.VALUE_AUTO)
6983 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6984 mac = utils.NormalizeAndValidateMac(mac)
6987 self.cfg.ReserveMAC(mac, self.proc.GetECId())
6988 except errors.ReservationError:
6989 raise errors.OpPrereqError("MAC address %s already in use"
6990 " in cluster" % mac,
6991 errors.ECODE_NOTUNIQUE)
6993 # bridge verification
6994 bridge = nic.get("bridge", None)
6995 link = nic.get("link", None)
6997 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
6998 " at the same time", errors.ECODE_INVAL)
6999 elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7000 raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7007 nicparams[constants.NIC_MODE] = nic_mode_req
7009 nicparams[constants.NIC_LINK] = link
7011 check_params = cluster.SimpleFillNIC(nicparams)
7012 objects.NIC.CheckParameterSyntax(check_params)
7013 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7015 # disk checks/pre-build
7017 for disk in self.op.disks:
7018 mode = disk.get("mode", constants.DISK_RDWR)
7019 if mode not in constants.DISK_ACCESS_SET:
7020 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7021 mode, errors.ECODE_INVAL)
7022 size = disk.get("size", None)
7024 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7027 except (TypeError, ValueError):
7028 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7030 new_disk = {"size": size, "mode": mode}
7032 new_disk["adopt"] = disk["adopt"]
7033 self.disks.append(new_disk)
7035 if self.op.mode == constants.INSTANCE_IMPORT:
7037 # Check that the new instance doesn't have less disks than the export
7038 instance_disks = len(self.disks)
7039 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7040 if instance_disks < export_disks:
7041 raise errors.OpPrereqError("Not enough disks to import."
7042 " (instance: %d, export: %d)" %
7043 (instance_disks, export_disks),
7047 for idx in range(export_disks):
7048 option = 'disk%d_dump' % idx
7049 if export_info.has_option(constants.INISECT_INS, option):
7050 # FIXME: are the old os-es, disk sizes, etc. useful?
7051 export_name = export_info.get(constants.INISECT_INS, option)
7052 image = utils.PathJoin(self.op.src_path, export_name)
7053 disk_images.append(image)
7055 disk_images.append(False)
7057 self.src_images = disk_images
7059 old_name = export_info.get(constants.INISECT_INS, 'name')
7061 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7062 except (TypeError, ValueError), err:
7063 raise errors.OpPrereqError("Invalid export file, nic_count is not"
7064 " an integer: %s" % str(err),
7066 if self.op.instance_name == old_name:
7067 for idx, nic in enumerate(self.nics):
7068 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7069 nic_mac_ini = 'nic%d_mac' % idx
7070 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7072 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7074 # ip ping checks (we use the same ip that was resolved in ExpandNames)
7075 if self.op.ip_check:
7076 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7077 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7078 (self.check_ip, self.op.instance_name),
7079 errors.ECODE_NOTUNIQUE)
7081 #### mac address generation
7082 # By generating here the mac address both the allocator and the hooks get
7083 # the real final mac address rather than the 'auto' or 'generate' value.
7084 # There is a race condition between the generation and the instance object
7085 # creation, which means that we know the mac is valid now, but we're not
7086 # sure it will be when we actually add the instance. If things go bad
7087 # adding the instance will abort because of a duplicate mac, and the
7088 # creation job will fail.
7089 for nic in self.nics:
7090 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7091 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7095 if self.op.iallocator is not None:
7096 self._RunAllocator()
7098 #### node related checks
7100 # check primary node
7101 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7102 assert self.pnode is not None, \
7103 "Cannot retrieve locked node %s" % self.op.pnode
7105 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7106 pnode.name, errors.ECODE_STATE)
7108 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7109 pnode.name, errors.ECODE_STATE)
7111 self.secondaries = []
7113 # mirror node verification
7114 if self.op.disk_template in constants.DTS_NET_MIRROR:
7115 if self.op.snode == pnode.name:
7116 raise errors.OpPrereqError("The secondary node cannot be the"
7117 " primary node.", errors.ECODE_INVAL)
7118 _CheckNodeOnline(self, self.op.snode)
7119 _CheckNodeNotDrained(self, self.op.snode)
7120 self.secondaries.append(self.op.snode)
7122 nodenames = [pnode.name] + self.secondaries
7124 req_size = _ComputeDiskSize(self.op.disk_template,
7127 # Check lv size requirements, if not adopting
7128 if req_size is not None and not self.adopt_disks:
7129 _CheckNodesFreeDisk(self, nodenames, req_size)
7131 if self.adopt_disks: # instead, we must check the adoption data
7132 all_lvs = set([i["adopt"] for i in self.disks])
7133 if len(all_lvs) != len(self.disks):
7134 raise errors.OpPrereqError("Duplicate volume names given for adoption",
7136 for lv_name in all_lvs:
7138 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7139 except errors.ReservationError:
7140 raise errors.OpPrereqError("LV named %s used by another instance" %
7141 lv_name, errors.ECODE_NOTUNIQUE)
7143 node_lvs = self.rpc.call_lv_list([pnode.name],
7144 self.cfg.GetVGName())[pnode.name]
7145 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7146 node_lvs = node_lvs.payload
7147 delta = all_lvs.difference(node_lvs.keys())
7149 raise errors.OpPrereqError("Missing logical volume(s): %s" %
7150 utils.CommaJoin(delta),
7152 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7154 raise errors.OpPrereqError("Online logical volumes found, cannot"
7155 " adopt: %s" % utils.CommaJoin(online_lvs),
7157 # update the size of disk based on what is found
7158 for dsk in self.disks:
7159 dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
7161 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7163 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7164 # check OS parameters (remotely)
7165 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7167 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7169 # memory check on primary node
7171 _CheckNodeFreeMemory(self, self.pnode.name,
7172 "creating instance %s" % self.op.instance_name,
7173 self.be_full[constants.BE_MEMORY],
7176 self.dry_run_result = list(nodenames)
7178 def Exec(self, feedback_fn):
7179 """Create and add the instance to the cluster.
7182 instance = self.op.instance_name
7183 pnode_name = self.pnode.name
7185 ht_kind = self.op.hypervisor
7186 if ht_kind in constants.HTS_REQ_PORT:
7187 network_port = self.cfg.AllocatePort()
7191 if constants.ENABLE_FILE_STORAGE:
7192 # this is needed because os.path.join does not accept None arguments
7193 if self.op.file_storage_dir is None:
7194 string_file_storage_dir = ""
7196 string_file_storage_dir = self.op.file_storage_dir
7198 # build the full file storage dir path
7199 file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7200 string_file_storage_dir, instance)
7202 file_storage_dir = ""
7204 disks = _GenerateDiskTemplate(self,
7205 self.op.disk_template,
7206 instance, pnode_name,
7210 self.op.file_driver,
7213 iobj = objects.Instance(name=instance, os=self.op.os_type,
7214 primary_node=pnode_name,
7215 nics=self.nics, disks=disks,
7216 disk_template=self.op.disk_template,
7218 network_port=network_port,
7219 beparams=self.op.beparams,
7220 hvparams=self.op.hvparams,
7221 hypervisor=self.op.hypervisor,
7222 osparams=self.op.osparams,
7225 if self.adopt_disks:
7226 # rename LVs to the newly-generated names; we need to construct
7227 # 'fake' LV disks with the old data, plus the new unique_id
7228 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7230 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7231 rename_to.append(t_dsk.logical_id)
7232 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7233 self.cfg.SetDiskID(t_dsk, pnode_name)
7234 result = self.rpc.call_blockdev_rename(pnode_name,
7235 zip(tmp_disks, rename_to))
7236 result.Raise("Failed to rename adoped LVs")
7238 feedback_fn("* creating instance disks...")
7240 _CreateDisks(self, iobj)
7241 except errors.OpExecError:
7242 self.LogWarning("Device creation failed, reverting...")
7244 _RemoveDisks(self, iobj)
7246 self.cfg.ReleaseDRBDMinors(instance)
7249 feedback_fn("adding instance %s to cluster config" % instance)
7251 self.cfg.AddInstance(iobj, self.proc.GetECId())
7253 # Declare that we don't want to remove the instance lock anymore, as we've
7254 # added the instance to the config
7255 del self.remove_locks[locking.LEVEL_INSTANCE]
7256 # Unlock all the nodes
7257 if self.op.mode == constants.INSTANCE_IMPORT:
7258 nodes_keep = [self.op.src_node]
7259 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7260 if node != self.op.src_node]
7261 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7262 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7264 self.context.glm.release(locking.LEVEL_NODE)
7265 del self.acquired_locks[locking.LEVEL_NODE]
7267 if self.op.wait_for_sync:
7268 disk_abort = not _WaitForSync(self, iobj)
7269 elif iobj.disk_template in constants.DTS_NET_MIRROR:
7270 # make sure the disks are not degraded (still sync-ing is ok)
7272 feedback_fn("* checking mirrors status")
7273 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7278 _RemoveDisks(self, iobj)
7279 self.cfg.RemoveInstance(iobj.name)
7280 # Make sure the instance lock gets removed
7281 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7282 raise errors.OpExecError("There are some degraded disks for"
7285 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7286 if self.op.mode == constants.INSTANCE_CREATE:
7287 if not self.op.no_install:
7288 feedback_fn("* running the instance OS create scripts...")
7289 # FIXME: pass debug option from opcode to backend
7290 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7291 self.op.debug_level)
7292 result.Raise("Could not add os for instance %s"
7293 " on node %s" % (instance, pnode_name))
7295 elif self.op.mode == constants.INSTANCE_IMPORT:
7296 feedback_fn("* running the instance OS import scripts...")
7300 for idx, image in enumerate(self.src_images):
7304 # FIXME: pass debug option from opcode to backend
7305 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7306 constants.IEIO_FILE, (image, ),
7307 constants.IEIO_SCRIPT,
7308 (iobj.disks[idx], idx),
7310 transfers.append(dt)
7313 masterd.instance.TransferInstanceData(self, feedback_fn,
7314 self.op.src_node, pnode_name,
7315 self.pnode.secondary_ip,
7317 if not compat.all(import_result):
7318 self.LogWarning("Some disks for instance %s on node %s were not"
7319 " imported successfully" % (instance, pnode_name))
7321 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7322 feedback_fn("* preparing remote import...")
7323 connect_timeout = constants.RIE_CONNECT_TIMEOUT
7324 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7326 disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
7327 self.source_x509_ca,
7328 self._cds, timeouts)
7329 if not compat.all(disk_results):
7330 # TODO: Should the instance still be started, even if some disks
7331 # failed to import (valid for local imports, too)?
7332 self.LogWarning("Some disks for instance %s on node %s were not"
7333 " imported successfully" % (instance, pnode_name))
7335 # Run rename script on newly imported instance
7336 assert iobj.name == instance
7337 feedback_fn("Running rename script for %s" % instance)
7338 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7339 self.source_instance_name,
7340 self.op.debug_level)
7342 self.LogWarning("Failed to run rename script for %s on node"
7343 " %s: %s" % (instance, pnode_name, result.fail_msg))
7346 # also checked in the prereq part
7347 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7351 iobj.admin_up = True
7352 self.cfg.Update(iobj, feedback_fn)
7353 logging.info("Starting instance %s on node %s", instance, pnode_name)
7354 feedback_fn("* starting instance...")
7355 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7356 result.Raise("Could not start instance")
7358 return list(iobj.all_nodes)
7361 class LUConnectConsole(NoHooksLU):
7362 """Connect to an instance's console.
7364 This is somewhat special in that it returns the command line that
7365 you need to run on the master node in order to connect to the
7374 def ExpandNames(self):
7375 self._ExpandAndLockInstance()
7377 def CheckPrereq(self):
7378 """Check prerequisites.
7380 This checks that the instance is in the cluster.
7383 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7384 assert self.instance is not None, \
7385 "Cannot retrieve locked instance %s" % self.op.instance_name
7386 _CheckNodeOnline(self, self.instance.primary_node)
7388 def Exec(self, feedback_fn):
7389 """Connect to the console of an instance
7392 instance = self.instance
7393 node = instance.primary_node
7395 node_insts = self.rpc.call_instance_list([node],
7396 [instance.hypervisor])[node]
7397 node_insts.Raise("Can't get node information from %s" % node)
7399 if instance.name not in node_insts.payload:
7400 if instance.admin_up:
7401 state = "ERROR_down"
7403 state = "ADMIN_down"
7404 raise errors.OpExecError("Instance %s is not running (state %s)" %
7405 (instance.name, state))
7407 logging.debug("Connecting to console of %s on %s", instance.name, node)
7409 hyper = hypervisor.GetHypervisor(instance.hypervisor)
7410 cluster = self.cfg.GetClusterInfo()
7411 # beparams and hvparams are passed separately, to avoid editing the
7412 # instance and then saving the defaults in the instance itself.
7413 hvparams = cluster.FillHV(instance)
7414 beparams = cluster.FillBE(instance)
7415 console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
7418 return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
7421 class LUReplaceDisks(LogicalUnit):
7422 """Replace the disks of an instance.
7425 HPATH = "mirrors-replace"
7426 HTYPE = constants.HTYPE_INSTANCE
7429 ("mode", ht.NoDefault, ht.TElemOf(constants.REPLACE_MODES)),
7430 ("disks", ht.EmptyList, ht.TListOf(ht.TPositiveInt)),
7431 ("remote_node", None, ht.TMaybeString),
7432 ("iallocator", None, ht.TMaybeString),
7433 ("early_release", False, ht.TBool),
7437 def CheckArguments(self):
7438 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7441 def ExpandNames(self):
7442 self._ExpandAndLockInstance()
7444 if self.op.iallocator is not None:
7445 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7447 elif self.op.remote_node is not None:
7448 remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7449 self.op.remote_node = remote_node
7451 # Warning: do not remove the locking of the new secondary here
7452 # unless DRBD8.AddChildren is changed to work in parallel;
7453 # currently it doesn't since parallel invocations of
7454 # FindUnusedMinor will conflict
7455 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7456 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7459 self.needed_locks[locking.LEVEL_NODE] = []
7460 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7462 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7463 self.op.iallocator, self.op.remote_node,
7464 self.op.disks, False, self.op.early_release)
7466 self.tasklets = [self.replacer]
7468 def DeclareLocks(self, level):
7469 # If we're not already locking all nodes in the set we have to declare the
7470 # instance's primary/secondary nodes.
7471 if (level == locking.LEVEL_NODE and
7472 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7473 self._LockInstancesNodes()
7475 def BuildHooksEnv(self):
7478 This runs on the master, the primary and all the secondaries.
7481 instance = self.replacer.instance
7483 "MODE": self.op.mode,
7484 "NEW_SECONDARY": self.op.remote_node,
7485 "OLD_SECONDARY": instance.secondary_nodes[0],
7487 env.update(_BuildInstanceHookEnvByObject(self, instance))
7489 self.cfg.GetMasterNode(),
7490 instance.primary_node,
7492 if self.op.remote_node is not None:
7493 nl.append(self.op.remote_node)
7497 class TLReplaceDisks(Tasklet):
7498 """Replaces disks for an instance.
7500 Note: Locking is not within the scope of this class.
7503 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7504 disks, delay_iallocator, early_release):
7505 """Initializes this class.
7508 Tasklet.__init__(self, lu)
7511 self.instance_name = instance_name
7513 self.iallocator_name = iallocator_name
7514 self.remote_node = remote_node
7516 self.delay_iallocator = delay_iallocator
7517 self.early_release = early_release
7520 self.instance = None
7521 self.new_node = None
7522 self.target_node = None
7523 self.other_node = None
7524 self.remote_node_info = None
7525 self.node_secondary_ip = None
7528 def CheckArguments(mode, remote_node, iallocator):
7529 """Helper function for users of this class.
7532 # check for valid parameter combination
7533 if mode == constants.REPLACE_DISK_CHG:
7534 if remote_node is None and iallocator is None:
7535 raise errors.OpPrereqError("When changing the secondary either an"
7536 " iallocator script must be used or the"
7537 " new node given", errors.ECODE_INVAL)
7539 if remote_node is not None and iallocator is not None:
7540 raise errors.OpPrereqError("Give either the iallocator or the new"
7541 " secondary, not both", errors.ECODE_INVAL)
7543 elif remote_node is not None or iallocator is not None:
7544 # Not replacing the secondary
7545 raise errors.OpPrereqError("The iallocator and new node options can"
7546 " only be used when changing the"
7547 " secondary node", errors.ECODE_INVAL)
7550 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7551 """Compute a new secondary node using an IAllocator.
7554 ial = IAllocator(lu.cfg, lu.rpc,
7555 mode=constants.IALLOCATOR_MODE_RELOC,
7557 relocate_from=relocate_from)
7559 ial.Run(iallocator_name)
7562 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7563 " %s" % (iallocator_name, ial.info),
7566 if len(ial.result) != ial.required_nodes:
7567 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7568 " of nodes (%s), required %s" %
7570 len(ial.result), ial.required_nodes),
7573 remote_node_name = ial.result[0]
7575 lu.LogInfo("Selected new secondary for instance '%s': %s",
7576 instance_name, remote_node_name)
7578 return remote_node_name
7580 def _FindFaultyDisks(self, node_name):
7581 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7584 def CheckPrereq(self):
7585 """Check prerequisites.
7587 This checks that the instance is in the cluster.
7590 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7591 assert instance is not None, \
7592 "Cannot retrieve locked instance %s" % self.instance_name
7594 if instance.disk_template != constants.DT_DRBD8:
7595 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7596 " instances", errors.ECODE_INVAL)
7598 if len(instance.secondary_nodes) != 1:
7599 raise errors.OpPrereqError("The instance has a strange layout,"
7600 " expected one secondary but found %d" %
7601 len(instance.secondary_nodes),
7604 if not self.delay_iallocator:
7605 self._CheckPrereq2()
7607 def _CheckPrereq2(self):
7608 """Check prerequisites, second part.
7610 This function should always be part of CheckPrereq. It was separated and is
7611 now called from Exec because during node evacuation iallocator was only
7612 called with an unmodified cluster model, not taking planned changes into
7616 instance = self.instance
7617 secondary_node = instance.secondary_nodes[0]
7619 if self.iallocator_name is None:
7620 remote_node = self.remote_node
7622 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7623 instance.name, instance.secondary_nodes)
7625 if remote_node is not None:
7626 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7627 assert self.remote_node_info is not None, \
7628 "Cannot retrieve locked node %s" % remote_node
7630 self.remote_node_info = None
7632 if remote_node == self.instance.primary_node:
7633 raise errors.OpPrereqError("The specified node is the primary node of"
7634 " the instance.", errors.ECODE_INVAL)
7636 if remote_node == secondary_node:
7637 raise errors.OpPrereqError("The specified node is already the"
7638 " secondary node of the instance.",
7641 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7642 constants.REPLACE_DISK_CHG):
7643 raise errors.OpPrereqError("Cannot specify disks to be replaced",
7646 if self.mode == constants.REPLACE_DISK_AUTO:
7647 faulty_primary = self._FindFaultyDisks(instance.primary_node)
7648 faulty_secondary = self._FindFaultyDisks(secondary_node)
7650 if faulty_primary and faulty_secondary:
7651 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7652 " one node and can not be repaired"
7653 " automatically" % self.instance_name,
7657 self.disks = faulty_primary
7658 self.target_node = instance.primary_node
7659 self.other_node = secondary_node
7660 check_nodes = [self.target_node, self.other_node]
7661 elif faulty_secondary:
7662 self.disks = faulty_secondary
7663 self.target_node = secondary_node
7664 self.other_node = instance.primary_node
7665 check_nodes = [self.target_node, self.other_node]
7671 # Non-automatic modes
7672 if self.mode == constants.REPLACE_DISK_PRI:
7673 self.target_node = instance.primary_node
7674 self.other_node = secondary_node
7675 check_nodes = [self.target_node, self.other_node]
7677 elif self.mode == constants.REPLACE_DISK_SEC:
7678 self.target_node = secondary_node
7679 self.other_node = instance.primary_node
7680 check_nodes = [self.target_node, self.other_node]
7682 elif self.mode == constants.REPLACE_DISK_CHG:
7683 self.new_node = remote_node
7684 self.other_node = instance.primary_node
7685 self.target_node = secondary_node
7686 check_nodes = [self.new_node, self.other_node]
7688 _CheckNodeNotDrained(self.lu, remote_node)
7690 old_node_info = self.cfg.GetNodeInfo(secondary_node)
7691 assert old_node_info is not None
7692 if old_node_info.offline and not self.early_release:
7693 # doesn't make sense to delay the release
7694 self.early_release = True
7695 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7696 " early-release mode", secondary_node)
7699 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7702 # If not specified all disks should be replaced
7704 self.disks = range(len(self.instance.disks))
7706 for node in check_nodes:
7707 _CheckNodeOnline(self.lu, node)
7709 # Check whether disks are valid
7710 for disk_idx in self.disks:
7711 instance.FindDisk(disk_idx)
7713 # Get secondary node IP addresses
7716 for node_name in [self.target_node, self.other_node, self.new_node]:
7717 if node_name is not None:
7718 node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7720 self.node_secondary_ip = node_2nd_ip
7722 def Exec(self, feedback_fn):
7723 """Execute disk replacement.
7725 This dispatches the disk replacement to the appropriate handler.
7728 if self.delay_iallocator:
7729 self._CheckPrereq2()
7732 feedback_fn("No disks need replacement")
7735 feedback_fn("Replacing disk(s) %s for %s" %
7736 (utils.CommaJoin(self.disks), self.instance.name))
7738 activate_disks = (not self.instance.admin_up)
7740 # Activate the instance disks if we're replacing them on a down instance
7742 _StartInstanceDisks(self.lu, self.instance, True)
7745 # Should we replace the secondary node?
7746 if self.new_node is not None:
7747 fn = self._ExecDrbd8Secondary
7749 fn = self._ExecDrbd8DiskOnly
7751 return fn(feedback_fn)
7754 # Deactivate the instance disks if we're replacing them on a
7757 _SafeShutdownInstanceDisks(self.lu, self.instance)
7759 def _CheckVolumeGroup(self, nodes):
7760 self.lu.LogInfo("Checking volume groups")
7762 vgname = self.cfg.GetVGName()
7764 # Make sure volume group exists on all involved nodes
7765 results = self.rpc.call_vg_list(nodes)
7767 raise errors.OpExecError("Can't list volume groups on the nodes")
7771 res.Raise("Error checking node %s" % node)
7772 if vgname not in res.payload:
7773 raise errors.OpExecError("Volume group '%s' not found on node %s" %
7776 def _CheckDisksExistence(self, nodes):
7777 # Check disk existence
7778 for idx, dev in enumerate(self.instance.disks):
7779 if idx not in self.disks:
7783 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7784 self.cfg.SetDiskID(dev, node)
7786 result = self.rpc.call_blockdev_find(node, dev)
7788 msg = result.fail_msg
7789 if msg or not result.payload:
7791 msg = "disk not found"
7792 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7795 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7796 for idx, dev in enumerate(self.instance.disks):
7797 if idx not in self.disks:
7800 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7803 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7805 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7806 " replace disks for instance %s" %
7807 (node_name, self.instance.name))
7809 def _CreateNewStorage(self, node_name):
7810 vgname = self.cfg.GetVGName()
7813 for idx, dev in enumerate(self.instance.disks):
7814 if idx not in self.disks:
7817 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7819 self.cfg.SetDiskID(dev, node_name)
7821 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7822 names = _GenerateUniqueNames(self.lu, lv_names)
7824 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7825 logical_id=(vgname, names[0]))
7826 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7827 logical_id=(vgname, names[1]))
7829 new_lvs = [lv_data, lv_meta]
7830 old_lvs = dev.children
7831 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7833 # we pass force_create=True to force the LVM creation
7834 for new_lv in new_lvs:
7835 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7836 _GetInstanceInfoText(self.instance), False)
7840 def _CheckDevices(self, node_name, iv_names):
7841 for name, (dev, _, _) in iv_names.iteritems():
7842 self.cfg.SetDiskID(dev, node_name)
7844 result = self.rpc.call_blockdev_find(node_name, dev)
7846 msg = result.fail_msg
7847 if msg or not result.payload:
7849 msg = "disk not found"
7850 raise errors.OpExecError("Can't find DRBD device %s: %s" %
7853 if result.payload.is_degraded:
7854 raise errors.OpExecError("DRBD device %s is degraded!" % name)
7856 def _RemoveOldStorage(self, node_name, iv_names):
7857 for name, (_, old_lvs, _) in iv_names.iteritems():
7858 self.lu.LogInfo("Remove logical volumes for %s" % name)
7861 self.cfg.SetDiskID(lv, node_name)
7863 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7865 self.lu.LogWarning("Can't remove old LV: %s" % msg,
7866 hint="remove unused LVs manually")
7868 def _ReleaseNodeLock(self, node_name):
7869 """Releases the lock for a given node."""
7870 self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7872 def _ExecDrbd8DiskOnly(self, feedback_fn):
7873 """Replace a disk on the primary or secondary for DRBD 8.
7875 The algorithm for replace is quite complicated:
7877 1. for each disk to be replaced:
7879 1. create new LVs on the target node with unique names
7880 1. detach old LVs from the drbd device
7881 1. rename old LVs to name_replaced.<time_t>
7882 1. rename new LVs to old LVs
7883 1. attach the new LVs (with the old names now) to the drbd device
7885 1. wait for sync across all devices
7887 1. for each modified disk:
7889 1. remove old LVs (which have the name name_replaces.<time_t>)
7891 Failures are not very well handled.
7896 # Step: check device activation
7897 self.lu.LogStep(1, steps_total, "Check device existence")
7898 self._CheckDisksExistence([self.other_node, self.target_node])
7899 self._CheckVolumeGroup([self.target_node, self.other_node])
7901 # Step: check other node consistency
7902 self.lu.LogStep(2, steps_total, "Check peer consistency")
7903 self._CheckDisksConsistency(self.other_node,
7904 self.other_node == self.instance.primary_node,
7907 # Step: create new storage
7908 self.lu.LogStep(3, steps_total, "Allocate new storage")
7909 iv_names = self._CreateNewStorage(self.target_node)
7911 # Step: for each lv, detach+rename*2+attach
7912 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
7913 for dev, old_lvs, new_lvs in iv_names.itervalues():
7914 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
7916 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
7918 result.Raise("Can't detach drbd from local storage on node"
7919 " %s for device %s" % (self.target_node, dev.iv_name))
7921 #cfg.Update(instance)
7923 # ok, we created the new LVs, so now we know we have the needed
7924 # storage; as such, we proceed on the target node to rename
7925 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
7926 # using the assumption that logical_id == physical_id (which in
7927 # turn is the unique_id on that node)
7929 # FIXME(iustin): use a better name for the replaced LVs
7930 temp_suffix = int(time.time())
7931 ren_fn = lambda d, suff: (d.physical_id[0],
7932 d.physical_id[1] + "_replaced-%s" % suff)
7934 # Build the rename list based on what LVs exist on the node
7935 rename_old_to_new = []
7936 for to_ren in old_lvs:
7937 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
7938 if not result.fail_msg and result.payload:
7940 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
7942 self.lu.LogInfo("Renaming the old LVs on the target node")
7943 result = self.rpc.call_blockdev_rename(self.target_node,
7945 result.Raise("Can't rename old LVs on node %s" % self.target_node)
7947 # Now we rename the new LVs to the old LVs
7948 self.lu.LogInfo("Renaming the new LVs on the target node")
7949 rename_new_to_old = [(new, old.physical_id)
7950 for old, new in zip(old_lvs, new_lvs)]
7951 result = self.rpc.call_blockdev_rename(self.target_node,
7953 result.Raise("Can't rename new LVs on node %s" % self.target_node)
7955 for old, new in zip(old_lvs, new_lvs):
7956 new.logical_id = old.logical_id
7957 self.cfg.SetDiskID(new, self.target_node)
7959 for disk in old_lvs:
7960 disk.logical_id = ren_fn(disk, temp_suffix)
7961 self.cfg.SetDiskID(disk, self.target_node)
7963 # Now that the new lvs have the old name, we can add them to the device
7964 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
7965 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
7967 msg = result.fail_msg
7969 for new_lv in new_lvs:
7970 msg2 = self.rpc.call_blockdev_remove(self.target_node,
7973 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
7974 hint=("cleanup manually the unused logical"
7976 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
7978 dev.children = new_lvs
7980 self.cfg.Update(self.instance, feedback_fn)
7983 if self.early_release:
7984 self.lu.LogStep(cstep, steps_total, "Removing old storage")
7986 self._RemoveOldStorage(self.target_node, iv_names)
7987 # WARNING: we release both node locks here, do not do other RPCs
7988 # than WaitForSync to the primary node
7989 self._ReleaseNodeLock([self.target_node, self.other_node])
7992 # This can fail as the old devices are degraded and _WaitForSync
7993 # does a combined result over all disks, so we don't check its return value
7994 self.lu.LogStep(cstep, steps_total, "Sync devices")
7996 _WaitForSync(self.lu, self.instance)
7998 # Check all devices manually
7999 self._CheckDevices(self.instance.primary_node, iv_names)
8001 # Step: remove old storage
8002 if not self.early_release:
8003 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8005 self._RemoveOldStorage(self.target_node, iv_names)
8007 def _ExecDrbd8Secondary(self, feedback_fn):
8008 """Replace the secondary node for DRBD 8.
8010 The algorithm for replace is quite complicated:
8011 - for all disks of the instance:
8012 - create new LVs on the new node with same names
8013 - shutdown the drbd device on the old secondary
8014 - disconnect the drbd network on the primary
8015 - create the drbd device on the new secondary
8016 - network attach the drbd on the primary, using an artifice:
8017 the drbd code for Attach() will connect to the network if it
8018 finds a device which is connected to the good local disks but
8020 - wait for sync across all devices
8021 - remove all disks from the old secondary
8023 Failures are not very well handled.
8028 # Step: check device activation
8029 self.lu.LogStep(1, steps_total, "Check device existence")
8030 self._CheckDisksExistence([self.instance.primary_node])
8031 self._CheckVolumeGroup([self.instance.primary_node])
8033 # Step: check other node consistency
8034 self.lu.LogStep(2, steps_total, "Check peer consistency")
8035 self._CheckDisksConsistency(self.instance.primary_node, True, True)
8037 # Step: create new storage
8038 self.lu.LogStep(3, steps_total, "Allocate new storage")
8039 for idx, dev in enumerate(self.instance.disks):
8040 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8041 (self.new_node, idx))
8042 # we pass force_create=True to force LVM creation
8043 for new_lv in dev.children:
8044 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8045 _GetInstanceInfoText(self.instance), False)
8047 # Step 4: dbrd minors and drbd setups changes
8048 # after this, we must manually remove the drbd minors on both the
8049 # error and the success paths
8050 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8051 minors = self.cfg.AllocateDRBDMinor([self.new_node
8052 for dev in self.instance.disks],
8054 logging.debug("Allocated minors %r", minors)
8057 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8058 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8059 (self.new_node, idx))
8060 # create new devices on new_node; note that we create two IDs:
8061 # one without port, so the drbd will be activated without
8062 # networking information on the new node at this stage, and one
8063 # with network, for the latter activation in step 4
8064 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8065 if self.instance.primary_node == o_node1:
8068 assert self.instance.primary_node == o_node2, "Three-node instance?"
8071 new_alone_id = (self.instance.primary_node, self.new_node, None,
8072 p_minor, new_minor, o_secret)
8073 new_net_id = (self.instance.primary_node, self.new_node, o_port,
8074 p_minor, new_minor, o_secret)
8076 iv_names[idx] = (dev, dev.children, new_net_id)
8077 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8079 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8080 logical_id=new_alone_id,
8081 children=dev.children,
8084 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8085 _GetInstanceInfoText(self.instance), False)
8086 except errors.GenericError:
8087 self.cfg.ReleaseDRBDMinors(self.instance.name)
8090 # We have new devices, shutdown the drbd on the old secondary
8091 for idx, dev in enumerate(self.instance.disks):
8092 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8093 self.cfg.SetDiskID(dev, self.target_node)
8094 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8096 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8097 "node: %s" % (idx, msg),
8098 hint=("Please cleanup this device manually as"
8099 " soon as possible"))
8101 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8102 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8103 self.node_secondary_ip,
8104 self.instance.disks)\
8105 [self.instance.primary_node]
8107 msg = result.fail_msg
8109 # detaches didn't succeed (unlikely)
8110 self.cfg.ReleaseDRBDMinors(self.instance.name)
8111 raise errors.OpExecError("Can't detach the disks from the network on"
8112 " old node: %s" % (msg,))
8114 # if we managed to detach at least one, we update all the disks of
8115 # the instance to point to the new secondary
8116 self.lu.LogInfo("Updating instance configuration")
8117 for dev, _, new_logical_id in iv_names.itervalues():
8118 dev.logical_id = new_logical_id
8119 self.cfg.SetDiskID(dev, self.instance.primary_node)
8121 self.cfg.Update(self.instance, feedback_fn)
8123 # and now perform the drbd attach
8124 self.lu.LogInfo("Attaching primary drbds to new secondary"
8125 " (standalone => connected)")
8126 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8128 self.node_secondary_ip,
8129 self.instance.disks,
8132 for to_node, to_result in result.items():
8133 msg = to_result.fail_msg
8135 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8137 hint=("please do a gnt-instance info to see the"
8138 " status of disks"))
8140 if self.early_release:
8141 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8143 self._RemoveOldStorage(self.target_node, iv_names)
8144 # WARNING: we release all node locks here, do not do other RPCs
8145 # than WaitForSync to the primary node
8146 self._ReleaseNodeLock([self.instance.primary_node,
8151 # This can fail as the old devices are degraded and _WaitForSync
8152 # does a combined result over all disks, so we don't check its return value
8153 self.lu.LogStep(cstep, steps_total, "Sync devices")
8155 _WaitForSync(self.lu, self.instance)
8157 # Check all devices manually
8158 self._CheckDevices(self.instance.primary_node, iv_names)
8160 # Step: remove old storage
8161 if not self.early_release:
8162 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8163 self._RemoveOldStorage(self.target_node, iv_names)
8166 class LURepairNodeStorage(NoHooksLU):
8167 """Repairs the volume group on a node.
8172 ("storage_type", ht.NoDefault, _CheckStorageType),
8173 ("name", ht.NoDefault, ht.TNonEmptyString),
8174 ("ignore_consistency", False, ht.TBool),
8178 def CheckArguments(self):
8179 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8181 storage_type = self.op.storage_type
8183 if (constants.SO_FIX_CONSISTENCY not in
8184 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8185 raise errors.OpPrereqError("Storage units of type '%s' can not be"
8186 " repaired" % storage_type,
8189 def ExpandNames(self):
8190 self.needed_locks = {
8191 locking.LEVEL_NODE: [self.op.node_name],
8194 def _CheckFaultyDisks(self, instance, node_name):
8195 """Ensure faulty disks abort the opcode or at least warn."""
8197 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8199 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8200 " node '%s'" % (instance.name, node_name),
8202 except errors.OpPrereqError, err:
8203 if self.op.ignore_consistency:
8204 self.proc.LogWarning(str(err.args[0]))
8208 def CheckPrereq(self):
8209 """Check prerequisites.
8212 # Check whether any instance on this node has faulty disks
8213 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8214 if not inst.admin_up:
8216 check_nodes = set(inst.all_nodes)
8217 check_nodes.discard(self.op.node_name)
8218 for inst_node_name in check_nodes:
8219 self._CheckFaultyDisks(inst, inst_node_name)
8221 def Exec(self, feedback_fn):
8222 feedback_fn("Repairing storage unit '%s' on %s ..." %
8223 (self.op.name, self.op.node_name))
8225 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8226 result = self.rpc.call_storage_execute(self.op.node_name,
8227 self.op.storage_type, st_args,
8229 constants.SO_FIX_CONSISTENCY)
8230 result.Raise("Failed to repair storage unit '%s' on %s" %
8231 (self.op.name, self.op.node_name))
8234 class LUNodeEvacuationStrategy(NoHooksLU):
8235 """Computes the node evacuation strategy.
8239 ("nodes", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
8240 ("remote_node", None, ht.TMaybeString),
8241 ("iallocator", None, ht.TMaybeString),
8245 def CheckArguments(self):
8246 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8248 def ExpandNames(self):
8249 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8250 self.needed_locks = locks = {}
8251 if self.op.remote_node is None:
8252 locks[locking.LEVEL_NODE] = locking.ALL_SET
8254 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8255 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8257 def Exec(self, feedback_fn):
8258 if self.op.remote_node is not None:
8260 for node in self.op.nodes:
8261 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8264 if i.primary_node == self.op.remote_node:
8265 raise errors.OpPrereqError("Node %s is the primary node of"
8266 " instance %s, cannot use it as"
8268 (self.op.remote_node, i.name),
8270 result.append([i.name, self.op.remote_node])
8272 ial = IAllocator(self.cfg, self.rpc,
8273 mode=constants.IALLOCATOR_MODE_MEVAC,
8274 evac_nodes=self.op.nodes)
8275 ial.Run(self.op.iallocator, validate=True)
8277 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8283 class LUGrowDisk(LogicalUnit):
8284 """Grow a disk of an instance.
8288 HTYPE = constants.HTYPE_INSTANCE
8291 ("disk", ht.NoDefault, ht.TInt),
8292 ("amount", ht.NoDefault, ht.TInt),
8293 ("wait_for_sync", True, ht.TBool),
8297 def ExpandNames(self):
8298 self._ExpandAndLockInstance()
8299 self.needed_locks[locking.LEVEL_NODE] = []
8300 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8302 def DeclareLocks(self, level):
8303 if level == locking.LEVEL_NODE:
8304 self._LockInstancesNodes()
8306 def BuildHooksEnv(self):
8309 This runs on the master, the primary and all the secondaries.
8313 "DISK": self.op.disk,
8314 "AMOUNT": self.op.amount,
8316 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8317 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8320 def CheckPrereq(self):
8321 """Check prerequisites.
8323 This checks that the instance is in the cluster.
8326 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8327 assert instance is not None, \
8328 "Cannot retrieve locked instance %s" % self.op.instance_name
8329 nodenames = list(instance.all_nodes)
8330 for node in nodenames:
8331 _CheckNodeOnline(self, node)
8333 self.instance = instance
8335 if instance.disk_template not in constants.DTS_GROWABLE:
8336 raise errors.OpPrereqError("Instance's disk layout does not support"
8337 " growing.", errors.ECODE_INVAL)
8339 self.disk = instance.FindDisk(self.op.disk)
8341 if instance.disk_template != constants.DT_FILE:
8342 # TODO: check the free disk space for file, when that feature will be
8344 _CheckNodesFreeDisk(self, nodenames, self.op.amount)
8346 def Exec(self, feedback_fn):
8347 """Execute disk grow.
8350 instance = self.instance
8353 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8355 raise errors.OpExecError("Cannot activate block device to grow")
8357 for node in instance.all_nodes:
8358 self.cfg.SetDiskID(disk, node)
8359 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8360 result.Raise("Grow request failed to node %s" % node)
8362 # TODO: Rewrite code to work properly
8363 # DRBD goes into sync mode for a short amount of time after executing the
8364 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8365 # calling "resize" in sync mode fails. Sleeping for a short amount of
8366 # time is a work-around.
8369 disk.RecordGrow(self.op.amount)
8370 self.cfg.Update(instance, feedback_fn)
8371 if self.op.wait_for_sync:
8372 disk_abort = not _WaitForSync(self, instance, disks=[disk])
8374 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8375 " status.\nPlease check the instance.")
8376 if not instance.admin_up:
8377 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8378 elif not instance.admin_up:
8379 self.proc.LogWarning("Not shutting down the disk even if the instance is"
8380 " not supposed to be running because no wait for"
8381 " sync mode was requested.")
8384 class LUQueryInstanceData(NoHooksLU):
8385 """Query runtime instance data.
8389 ("instances", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
8390 ("static", False, ht.TBool),
8394 def ExpandNames(self):
8395 self.needed_locks = {}
8396 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8398 if self.op.instances:
8399 self.wanted_names = []
8400 for name in self.op.instances:
8401 full_name = _ExpandInstanceName(self.cfg, name)
8402 self.wanted_names.append(full_name)
8403 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8405 self.wanted_names = None
8406 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8408 self.needed_locks[locking.LEVEL_NODE] = []
8409 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8411 def DeclareLocks(self, level):
8412 if level == locking.LEVEL_NODE:
8413 self._LockInstancesNodes()
8415 def CheckPrereq(self):
8416 """Check prerequisites.
8418 This only checks the optional instance list against the existing names.
8421 if self.wanted_names is None:
8422 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8424 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8425 in self.wanted_names]
8427 def _ComputeBlockdevStatus(self, node, instance_name, dev):
8428 """Returns the status of a block device
8431 if self.op.static or not node:
8434 self.cfg.SetDiskID(dev, node)
8436 result = self.rpc.call_blockdev_find(node, dev)
8440 result.Raise("Can't compute disk status for %s" % instance_name)
8442 status = result.payload
8446 return (status.dev_path, status.major, status.minor,
8447 status.sync_percent, status.estimated_time,
8448 status.is_degraded, status.ldisk_status)
8450 def _ComputeDiskStatus(self, instance, snode, dev):
8451 """Compute block device status.
8454 if dev.dev_type in constants.LDS_DRBD:
8455 # we change the snode then (otherwise we use the one passed in)
8456 if dev.logical_id[0] == instance.primary_node:
8457 snode = dev.logical_id[1]
8459 snode = dev.logical_id[0]
8461 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8463 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8466 dev_children = [self._ComputeDiskStatus(instance, snode, child)
8467 for child in dev.children]
8472 "iv_name": dev.iv_name,
8473 "dev_type": dev.dev_type,
8474 "logical_id": dev.logical_id,
8475 "physical_id": dev.physical_id,
8476 "pstatus": dev_pstatus,
8477 "sstatus": dev_sstatus,
8478 "children": dev_children,
8485 def Exec(self, feedback_fn):
8486 """Gather and return data"""
8489 cluster = self.cfg.GetClusterInfo()
8491 for instance in self.wanted_instances:
8492 if not self.op.static:
8493 remote_info = self.rpc.call_instance_info(instance.primary_node,
8495 instance.hypervisor)
8496 remote_info.Raise("Error checking node %s" % instance.primary_node)
8497 remote_info = remote_info.payload
8498 if remote_info and "state" in remote_info:
8501 remote_state = "down"
8504 if instance.admin_up:
8507 config_state = "down"
8509 disks = [self._ComputeDiskStatus(instance, None, device)
8510 for device in instance.disks]
8513 "name": instance.name,
8514 "config_state": config_state,
8515 "run_state": remote_state,
8516 "pnode": instance.primary_node,
8517 "snodes": instance.secondary_nodes,
8519 # this happens to be the same format used for hooks
8520 "nics": _NICListToTuple(self, instance.nics),
8521 "disk_template": instance.disk_template,
8523 "hypervisor": instance.hypervisor,
8524 "network_port": instance.network_port,
8525 "hv_instance": instance.hvparams,
8526 "hv_actual": cluster.FillHV(instance, skip_globals=True),
8527 "be_instance": instance.beparams,
8528 "be_actual": cluster.FillBE(instance),
8529 "os_instance": instance.osparams,
8530 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8531 "serial_no": instance.serial_no,
8532 "mtime": instance.mtime,
8533 "ctime": instance.ctime,
8534 "uuid": instance.uuid,
8537 result[instance.name] = idict
8542 class LUSetInstanceParams(LogicalUnit):
8543 """Modifies an instances's parameters.
8546 HPATH = "instance-modify"
8547 HTYPE = constants.HTYPE_INSTANCE
8550 ("nics", ht.EmptyList, ht.TList),
8551 ("disks", ht.EmptyList, ht.TList),
8552 ("beparams", ht.EmptyDict, ht.TDict),
8553 ("hvparams", ht.EmptyDict, ht.TDict),
8554 ("disk_template", None, ht.TMaybeString),
8555 ("remote_node", None, ht.TMaybeString),
8556 ("os_name", None, ht.TMaybeString),
8557 ("force_variant", False, ht.TBool),
8558 ("osparams", None, ht.TOr(ht.TDict, ht.TNone)),
8563 def CheckArguments(self):
8564 if not (self.op.nics or self.op.disks or self.op.disk_template or
8565 self.op.hvparams or self.op.beparams or self.op.os_name):
8566 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8568 if self.op.hvparams:
8569 _CheckGlobalHvParams(self.op.hvparams)
8573 for disk_op, disk_dict in self.op.disks:
8574 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8575 if disk_op == constants.DDM_REMOVE:
8578 elif disk_op == constants.DDM_ADD:
8581 if not isinstance(disk_op, int):
8582 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8583 if not isinstance(disk_dict, dict):
8584 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8585 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8587 if disk_op == constants.DDM_ADD:
8588 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8589 if mode not in constants.DISK_ACCESS_SET:
8590 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8592 size = disk_dict.get('size', None)
8594 raise errors.OpPrereqError("Required disk parameter size missing",
8598 except (TypeError, ValueError), err:
8599 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8600 str(err), errors.ECODE_INVAL)
8601 disk_dict['size'] = size
8603 # modification of disk
8604 if 'size' in disk_dict:
8605 raise errors.OpPrereqError("Disk size change not possible, use"
8606 " grow-disk", errors.ECODE_INVAL)
8608 if disk_addremove > 1:
8609 raise errors.OpPrereqError("Only one disk add or remove operation"
8610 " supported at a time", errors.ECODE_INVAL)
8612 if self.op.disks and self.op.disk_template is not None:
8613 raise errors.OpPrereqError("Disk template conversion and other disk"
8614 " changes not supported at the same time",
8617 if self.op.disk_template:
8618 _CheckDiskTemplate(self.op.disk_template)
8619 if (self.op.disk_template in constants.DTS_NET_MIRROR and
8620 self.op.remote_node is None):
8621 raise errors.OpPrereqError("Changing the disk template to a mirrored"
8622 " one requires specifying a secondary node",
8627 for nic_op, nic_dict in self.op.nics:
8628 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8629 if nic_op == constants.DDM_REMOVE:
8632 elif nic_op == constants.DDM_ADD:
8635 if not isinstance(nic_op, int):
8636 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8637 if not isinstance(nic_dict, dict):
8638 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8639 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8641 # nic_dict should be a dict
8642 nic_ip = nic_dict.get('ip', None)
8643 if nic_ip is not None:
8644 if nic_ip.lower() == constants.VALUE_NONE:
8645 nic_dict['ip'] = None
8647 if not netutils.IPAddress.IsValid(nic_ip):
8648 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8651 nic_bridge = nic_dict.get('bridge', None)
8652 nic_link = nic_dict.get('link', None)
8653 if nic_bridge and nic_link:
8654 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8655 " at the same time", errors.ECODE_INVAL)
8656 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8657 nic_dict['bridge'] = None
8658 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8659 nic_dict['link'] = None
8661 if nic_op == constants.DDM_ADD:
8662 nic_mac = nic_dict.get('mac', None)
8664 nic_dict['mac'] = constants.VALUE_AUTO
8666 if 'mac' in nic_dict:
8667 nic_mac = nic_dict['mac']
8668 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8669 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8671 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8672 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8673 " modifying an existing nic",
8676 if nic_addremove > 1:
8677 raise errors.OpPrereqError("Only one NIC add or remove operation"
8678 " supported at a time", errors.ECODE_INVAL)
8680 def ExpandNames(self):
8681 self._ExpandAndLockInstance()
8682 self.needed_locks[locking.LEVEL_NODE] = []
8683 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8685 def DeclareLocks(self, level):
8686 if level == locking.LEVEL_NODE:
8687 self._LockInstancesNodes()
8688 if self.op.disk_template and self.op.remote_node:
8689 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8690 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8692 def BuildHooksEnv(self):
8695 This runs on the master, primary and secondaries.
8699 if constants.BE_MEMORY in self.be_new:
8700 args['memory'] = self.be_new[constants.BE_MEMORY]
8701 if constants.BE_VCPUS in self.be_new:
8702 args['vcpus'] = self.be_new[constants.BE_VCPUS]
8703 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8704 # information at all.
8707 nic_override = dict(self.op.nics)
8708 for idx, nic in enumerate(self.instance.nics):
8709 if idx in nic_override:
8710 this_nic_override = nic_override[idx]
8712 this_nic_override = {}
8713 if 'ip' in this_nic_override:
8714 ip = this_nic_override['ip']
8717 if 'mac' in this_nic_override:
8718 mac = this_nic_override['mac']
8721 if idx in self.nic_pnew:
8722 nicparams = self.nic_pnew[idx]
8724 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
8725 mode = nicparams[constants.NIC_MODE]
8726 link = nicparams[constants.NIC_LINK]
8727 args['nics'].append((ip, mac, mode, link))
8728 if constants.DDM_ADD in nic_override:
8729 ip = nic_override[constants.DDM_ADD].get('ip', None)
8730 mac = nic_override[constants.DDM_ADD]['mac']
8731 nicparams = self.nic_pnew[constants.DDM_ADD]
8732 mode = nicparams[constants.NIC_MODE]
8733 link = nicparams[constants.NIC_LINK]
8734 args['nics'].append((ip, mac, mode, link))
8735 elif constants.DDM_REMOVE in nic_override:
8736 del args['nics'][-1]
8738 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8739 if self.op.disk_template:
8740 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8741 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8744 def CheckPrereq(self):
8745 """Check prerequisites.
8747 This only checks the instance list against the existing names.
8750 # checking the new params on the primary/secondary nodes
8752 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8753 cluster = self.cluster = self.cfg.GetClusterInfo()
8754 assert self.instance is not None, \
8755 "Cannot retrieve locked instance %s" % self.op.instance_name
8756 pnode = instance.primary_node
8757 nodelist = list(instance.all_nodes)
8760 if self.op.os_name and not self.op.force:
8761 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8762 self.op.force_variant)
8763 instance_os = self.op.os_name
8765 instance_os = instance.os
8767 if self.op.disk_template:
8768 if instance.disk_template == self.op.disk_template:
8769 raise errors.OpPrereqError("Instance already has disk template %s" %
8770 instance.disk_template, errors.ECODE_INVAL)
8772 if (instance.disk_template,
8773 self.op.disk_template) not in self._DISK_CONVERSIONS:
8774 raise errors.OpPrereqError("Unsupported disk template conversion from"
8775 " %s to %s" % (instance.disk_template,
8776 self.op.disk_template),
8778 _CheckInstanceDown(self, instance, "cannot change disk template")
8779 if self.op.disk_template in constants.DTS_NET_MIRROR:
8780 if self.op.remote_node == pnode:
8781 raise errors.OpPrereqError("Given new secondary node %s is the same"
8782 " as the primary node of the instance" %
8783 self.op.remote_node, errors.ECODE_STATE)
8784 _CheckNodeOnline(self, self.op.remote_node)
8785 _CheckNodeNotDrained(self, self.op.remote_node)
8786 disks = [{"size": d.size} for d in instance.disks]
8787 required = _ComputeDiskSize(self.op.disk_template, disks)
8788 _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8790 # hvparams processing
8791 if self.op.hvparams:
8792 hv_type = instance.hypervisor
8793 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
8794 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
8795 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
8798 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
8799 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8800 self.hv_new = hv_new # the new actual values
8801 self.hv_inst = i_hvdict # the new dict (without defaults)
8803 self.hv_new = self.hv_inst = {}
8805 # beparams processing
8806 if self.op.beparams:
8807 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
8809 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
8810 be_new = cluster.SimpleFillBE(i_bedict)
8811 self.be_new = be_new # the new actual values
8812 self.be_inst = i_bedict # the new dict (without defaults)
8814 self.be_new = self.be_inst = {}
8816 # osparams processing
8817 if self.op.osparams:
8818 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
8819 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
8820 self.os_new = cluster.SimpleFillOS(instance_os, i_osdict)
8821 self.os_inst = i_osdict # the new dict (without defaults)
8823 self.os_new = self.os_inst = {}
8827 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
8828 mem_check_list = [pnode]
8829 if be_new[constants.BE_AUTO_BALANCE]:
8830 # either we changed auto_balance to yes or it was from before
8831 mem_check_list.extend(instance.secondary_nodes)
8832 instance_info = self.rpc.call_instance_info(pnode, instance.name,
8833 instance.hypervisor)
8834 nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8835 instance.hypervisor)
8836 pninfo = nodeinfo[pnode]
8837 msg = pninfo.fail_msg
8839 # Assume the primary node is unreachable and go ahead
8840 self.warn.append("Can't get info from primary node %s: %s" %
8842 elif not isinstance(pninfo.payload.get('memory_free', None), int):
8843 self.warn.append("Node data from primary node %s doesn't contain"
8844 " free memory information" % pnode)
8845 elif instance_info.fail_msg:
8846 self.warn.append("Can't get instance runtime information: %s" %
8847 instance_info.fail_msg)
8849 if instance_info.payload:
8850 current_mem = int(instance_info.payload['memory'])
8852 # Assume instance not running
8853 # (there is a slight race condition here, but it's not very probable,
8854 # and we have no other way to check)
8856 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8857 pninfo.payload['memory_free'])
8859 raise errors.OpPrereqError("This change will prevent the instance"
8860 " from starting, due to %d MB of memory"
8861 " missing on its primary node" % miss_mem,
8864 if be_new[constants.BE_AUTO_BALANCE]:
8865 for node, nres in nodeinfo.items():
8866 if node not in instance.secondary_nodes:
8870 self.warn.append("Can't get info from secondary node %s: %s" %
8872 elif not isinstance(nres.payload.get('memory_free', None), int):
8873 self.warn.append("Secondary node %s didn't return free"
8874 " memory information" % node)
8875 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8876 self.warn.append("Not enough memory to failover instance to"
8877 " secondary node %s" % node)
8882 for nic_op, nic_dict in self.op.nics:
8883 if nic_op == constants.DDM_REMOVE:
8884 if not instance.nics:
8885 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
8888 if nic_op != constants.DDM_ADD:
8890 if not instance.nics:
8891 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
8892 " no NICs" % nic_op,
8894 if nic_op < 0 or nic_op >= len(instance.nics):
8895 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
8897 (nic_op, len(instance.nics) - 1),
8899 old_nic_params = instance.nics[nic_op].nicparams
8900 old_nic_ip = instance.nics[nic_op].ip
8905 update_params_dict = dict([(key, nic_dict[key])
8906 for key in constants.NICS_PARAMETERS
8907 if key in nic_dict])
8909 if 'bridge' in nic_dict:
8910 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
8912 new_nic_params = _GetUpdatedParams(old_nic_params,
8914 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
8915 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
8916 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
8917 self.nic_pinst[nic_op] = new_nic_params
8918 self.nic_pnew[nic_op] = new_filled_nic_params
8919 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
8921 if new_nic_mode == constants.NIC_MODE_BRIDGED:
8922 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
8923 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
8925 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
8927 self.warn.append(msg)
8929 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
8930 if new_nic_mode == constants.NIC_MODE_ROUTED:
8931 if 'ip' in nic_dict:
8932 nic_ip = nic_dict['ip']
8936 raise errors.OpPrereqError('Cannot set the nic ip to None'
8937 ' on a routed nic', errors.ECODE_INVAL)
8938 if 'mac' in nic_dict:
8939 nic_mac = nic_dict['mac']
8941 raise errors.OpPrereqError('Cannot set the nic mac to None',
8943 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8944 # otherwise generate the mac
8945 nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
8947 # or validate/reserve the current one
8949 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
8950 except errors.ReservationError:
8951 raise errors.OpPrereqError("MAC address %s already in use"
8952 " in cluster" % nic_mac,
8953 errors.ECODE_NOTUNIQUE)
8956 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
8957 raise errors.OpPrereqError("Disk operations not supported for"
8958 " diskless instances",
8960 for disk_op, _ in self.op.disks:
8961 if disk_op == constants.DDM_REMOVE:
8962 if len(instance.disks) == 1:
8963 raise errors.OpPrereqError("Cannot remove the last disk of"
8964 " an instance", errors.ECODE_INVAL)
8965 _CheckInstanceDown(self, instance, "cannot remove disks")
8967 if (disk_op == constants.DDM_ADD and
8968 len(instance.nics) >= constants.MAX_DISKS):
8969 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
8970 " add more" % constants.MAX_DISKS,
8972 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
8974 if disk_op < 0 or disk_op >= len(instance.disks):
8975 raise errors.OpPrereqError("Invalid disk index %s, valid values"
8977 (disk_op, len(instance.disks)),
8982 def _ConvertPlainToDrbd(self, feedback_fn):
8983 """Converts an instance from plain to drbd.
8986 feedback_fn("Converting template to drbd")
8987 instance = self.instance
8988 pnode = instance.primary_node
8989 snode = self.op.remote_node
8991 # create a fake disk info for _GenerateDiskTemplate
8992 disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
8993 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
8994 instance.name, pnode, [snode],
8995 disk_info, None, None, 0)
8996 info = _GetInstanceInfoText(instance)
8997 feedback_fn("Creating aditional volumes...")
8998 # first, create the missing data and meta devices
8999 for disk in new_disks:
9000 # unfortunately this is... not too nice
9001 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9003 for child in disk.children:
9004 _CreateSingleBlockDev(self, snode, instance, child, info, True)
9005 # at this stage, all new LVs have been created, we can rename the
9007 feedback_fn("Renaming original volumes...")
9008 rename_list = [(o, n.children[0].logical_id)
9009 for (o, n) in zip(instance.disks, new_disks)]
9010 result = self.rpc.call_blockdev_rename(pnode, rename_list)
9011 result.Raise("Failed to rename original LVs")
9013 feedback_fn("Initializing DRBD devices...")
9014 # all child devices are in place, we can now create the DRBD devices
9015 for disk in new_disks:
9016 for node in [pnode, snode]:
9017 f_create = node == pnode
9018 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9020 # at this point, the instance has been modified
9021 instance.disk_template = constants.DT_DRBD8
9022 instance.disks = new_disks
9023 self.cfg.Update(instance, feedback_fn)
9025 # disks are created, waiting for sync
9026 disk_abort = not _WaitForSync(self, instance)
9028 raise errors.OpExecError("There are some degraded disks for"
9029 " this instance, please cleanup manually")
9031 def _ConvertDrbdToPlain(self, feedback_fn):
9032 """Converts an instance from drbd to plain.
9035 instance = self.instance
9036 assert len(instance.secondary_nodes) == 1
9037 pnode = instance.primary_node
9038 snode = instance.secondary_nodes[0]
9039 feedback_fn("Converting template to plain")
9041 old_disks = instance.disks
9042 new_disks = [d.children[0] for d in old_disks]
9044 # copy over size and mode
9045 for parent, child in zip(old_disks, new_disks):
9046 child.size = parent.size
9047 child.mode = parent.mode
9049 # update instance structure
9050 instance.disks = new_disks
9051 instance.disk_template = constants.DT_PLAIN
9052 self.cfg.Update(instance, feedback_fn)
9054 feedback_fn("Removing volumes on the secondary node...")
9055 for disk in old_disks:
9056 self.cfg.SetDiskID(disk, snode)
9057 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9059 self.LogWarning("Could not remove block device %s on node %s,"
9060 " continuing anyway: %s", disk.iv_name, snode, msg)
9062 feedback_fn("Removing unneeded volumes on the primary node...")
9063 for idx, disk in enumerate(old_disks):
9064 meta = disk.children[1]
9065 self.cfg.SetDiskID(meta, pnode)
9066 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9068 self.LogWarning("Could not remove metadata for disk %d on node %s,"
9069 " continuing anyway: %s", idx, pnode, msg)
9072 def Exec(self, feedback_fn):
9073 """Modifies an instance.
9075 All parameters take effect only at the next restart of the instance.
9078 # Process here the warnings from CheckPrereq, as we don't have a
9079 # feedback_fn there.
9080 for warn in self.warn:
9081 feedback_fn("WARNING: %s" % warn)
9084 instance = self.instance
9086 for disk_op, disk_dict in self.op.disks:
9087 if disk_op == constants.DDM_REMOVE:
9088 # remove the last disk
9089 device = instance.disks.pop()
9090 device_idx = len(instance.disks)
9091 for node, disk in device.ComputeNodeTree(instance.primary_node):
9092 self.cfg.SetDiskID(disk, node)
9093 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9095 self.LogWarning("Could not remove disk/%d on node %s: %s,"
9096 " continuing anyway", device_idx, node, msg)
9097 result.append(("disk/%d" % device_idx, "remove"))
9098 elif disk_op == constants.DDM_ADD:
9100 if instance.disk_template == constants.DT_FILE:
9101 file_driver, file_path = instance.disks[0].logical_id
9102 file_path = os.path.dirname(file_path)
9104 file_driver = file_path = None
9105 disk_idx_base = len(instance.disks)
9106 new_disk = _GenerateDiskTemplate(self,
9107 instance.disk_template,
9108 instance.name, instance.primary_node,
9109 instance.secondary_nodes,
9114 instance.disks.append(new_disk)
9115 info = _GetInstanceInfoText(instance)
9117 logging.info("Creating volume %s for instance %s",
9118 new_disk.iv_name, instance.name)
9119 # Note: this needs to be kept in sync with _CreateDisks
9121 for node in instance.all_nodes:
9122 f_create = node == instance.primary_node
9124 _CreateBlockDev(self, node, instance, new_disk,
9125 f_create, info, f_create)
9126 except errors.OpExecError, err:
9127 self.LogWarning("Failed to create volume %s (%s) on"
9129 new_disk.iv_name, new_disk, node, err)
9130 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9131 (new_disk.size, new_disk.mode)))
9133 # change a given disk
9134 instance.disks[disk_op].mode = disk_dict['mode']
9135 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9137 if self.op.disk_template:
9138 r_shut = _ShutdownInstanceDisks(self, instance)
9140 raise errors.OpExecError("Cannot shutdow instance disks, unable to"
9141 " proceed with disk template conversion")
9142 mode = (instance.disk_template, self.op.disk_template)
9144 self._DISK_CONVERSIONS[mode](self, feedback_fn)
9146 self.cfg.ReleaseDRBDMinors(instance.name)
9148 result.append(("disk_template", self.op.disk_template))
9151 for nic_op, nic_dict in self.op.nics:
9152 if nic_op == constants.DDM_REMOVE:
9153 # remove the last nic
9154 del instance.nics[-1]
9155 result.append(("nic.%d" % len(instance.nics), "remove"))
9156 elif nic_op == constants.DDM_ADD:
9157 # mac and bridge should be set, by now
9158 mac = nic_dict['mac']
9159 ip = nic_dict.get('ip', None)
9160 nicparams = self.nic_pinst[constants.DDM_ADD]
9161 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9162 instance.nics.append(new_nic)
9163 result.append(("nic.%d" % (len(instance.nics) - 1),
9164 "add:mac=%s,ip=%s,mode=%s,link=%s" %
9165 (new_nic.mac, new_nic.ip,
9166 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9167 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9170 for key in 'mac', 'ip':
9172 setattr(instance.nics[nic_op], key, nic_dict[key])
9173 if nic_op in self.nic_pinst:
9174 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9175 for key, val in nic_dict.iteritems():
9176 result.append(("nic.%s/%d" % (key, nic_op), val))
9179 if self.op.hvparams:
9180 instance.hvparams = self.hv_inst
9181 for key, val in self.op.hvparams.iteritems():
9182 result.append(("hv/%s" % key, val))
9185 if self.op.beparams:
9186 instance.beparams = self.be_inst
9187 for key, val in self.op.beparams.iteritems():
9188 result.append(("be/%s" % key, val))
9192 instance.os = self.op.os_name
9195 if self.op.osparams:
9196 instance.osparams = self.os_inst
9197 for key, val in self.op.osparams.iteritems():
9198 result.append(("os/%s" % key, val))
9200 self.cfg.Update(instance, feedback_fn)
9204 _DISK_CONVERSIONS = {
9205 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9206 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9210 class LUQueryExports(NoHooksLU):
9211 """Query the exports list
9215 ("nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
9216 ("use_locking", False, ht.TBool),
9220 def ExpandNames(self):
9221 self.needed_locks = {}
9222 self.share_locks[locking.LEVEL_NODE] = 1
9223 if not self.op.nodes:
9224 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9226 self.needed_locks[locking.LEVEL_NODE] = \
9227 _GetWantedNodes(self, self.op.nodes)
9229 def Exec(self, feedback_fn):
9230 """Compute the list of all the exported system images.
9233 @return: a dictionary with the structure node->(export-list)
9234 where export-list is a list of the instances exported on
9238 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9239 rpcresult = self.rpc.call_export_list(self.nodes)
9241 for node in rpcresult:
9242 if rpcresult[node].fail_msg:
9243 result[node] = False
9245 result[node] = rpcresult[node].payload
9250 class LUPrepareExport(NoHooksLU):
9251 """Prepares an instance for an export and returns useful information.
9256 ("mode", ht.NoDefault, ht.TElemOf(constants.EXPORT_MODES)),
9260 def ExpandNames(self):
9261 self._ExpandAndLockInstance()
9263 def CheckPrereq(self):
9264 """Check prerequisites.
9267 instance_name = self.op.instance_name
9269 self.instance = self.cfg.GetInstanceInfo(instance_name)
9270 assert self.instance is not None, \
9271 "Cannot retrieve locked instance %s" % self.op.instance_name
9272 _CheckNodeOnline(self, self.instance.primary_node)
9274 self._cds = _GetClusterDomainSecret()
9276 def Exec(self, feedback_fn):
9277 """Prepares an instance for an export.
9280 instance = self.instance
9282 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9283 salt = utils.GenerateSecret(8)
9285 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9286 result = self.rpc.call_x509_cert_create(instance.primary_node,
9287 constants.RIE_CERT_VALIDITY)
9288 result.Raise("Can't create X509 key and certificate on %s" % result.node)
9290 (name, cert_pem) = result.payload
9292 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9296 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9297 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9299 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9305 class LUExportInstance(LogicalUnit):
9306 """Export an instance to an image in the cluster.
9309 HPATH = "instance-export"
9310 HTYPE = constants.HTYPE_INSTANCE
9313 ("target_node", ht.NoDefault, ht.TOr(ht.TNonEmptyString, ht.TList)),
9314 ("shutdown", True, ht.TBool),
9316 ("remove_instance", False, ht.TBool),
9317 ("ignore_remove_failures", False, ht.TBool),
9318 ("mode", constants.EXPORT_MODE_LOCAL, ht.TElemOf(constants.EXPORT_MODES)),
9319 ("x509_key_name", None, ht.TOr(ht.TList, ht.TNone)),
9320 ("destination_x509_ca", None, ht.TMaybeString),
9324 def CheckArguments(self):
9325 """Check the arguments.
9328 self.x509_key_name = self.op.x509_key_name
9329 self.dest_x509_ca_pem = self.op.destination_x509_ca
9331 if self.op.remove_instance and not self.op.shutdown:
9332 raise errors.OpPrereqError("Can not remove instance without shutting it"
9335 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9336 if not self.x509_key_name:
9337 raise errors.OpPrereqError("Missing X509 key name for encryption",
9340 if not self.dest_x509_ca_pem:
9341 raise errors.OpPrereqError("Missing destination X509 CA",
9344 def ExpandNames(self):
9345 self._ExpandAndLockInstance()
9347 # Lock all nodes for local exports
9348 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9349 # FIXME: lock only instance primary and destination node
9351 # Sad but true, for now we have do lock all nodes, as we don't know where
9352 # the previous export might be, and in this LU we search for it and
9353 # remove it from its current node. In the future we could fix this by:
9354 # - making a tasklet to search (share-lock all), then create the
9355 # new one, then one to remove, after
9356 # - removing the removal operation altogether
9357 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9359 def DeclareLocks(self, level):
9360 """Last minute lock declaration."""
9361 # All nodes are locked anyway, so nothing to do here.
9363 def BuildHooksEnv(self):
9366 This will run on the master, primary node and target node.
9370 "EXPORT_MODE": self.op.mode,
9371 "EXPORT_NODE": self.op.target_node,
9372 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9373 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9374 # TODO: Generic function for boolean env variables
9375 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9378 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9380 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9382 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9383 nl.append(self.op.target_node)
9387 def CheckPrereq(self):
9388 """Check prerequisites.
9390 This checks that the instance and node names are valid.
9393 instance_name = self.op.instance_name
9395 self.instance = self.cfg.GetInstanceInfo(instance_name)
9396 assert self.instance is not None, \
9397 "Cannot retrieve locked instance %s" % self.op.instance_name
9398 _CheckNodeOnline(self, self.instance.primary_node)
9400 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9401 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9402 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9403 assert self.dst_node is not None
9405 _CheckNodeOnline(self, self.dst_node.name)
9406 _CheckNodeNotDrained(self, self.dst_node.name)
9409 self.dest_disk_info = None
9410 self.dest_x509_ca = None
9412 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9413 self.dst_node = None
9415 if len(self.op.target_node) != len(self.instance.disks):
9416 raise errors.OpPrereqError(("Received destination information for %s"
9417 " disks, but instance %s has %s disks") %
9418 (len(self.op.target_node), instance_name,
9419 len(self.instance.disks)),
9422 cds = _GetClusterDomainSecret()
9424 # Check X509 key name
9426 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9427 except (TypeError, ValueError), err:
9428 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9430 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9431 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9434 # Load and verify CA
9436 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9437 except OpenSSL.crypto.Error, err:
9438 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9439 (err, ), errors.ECODE_INVAL)
9441 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9442 if errcode is not None:
9443 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9444 (msg, ), errors.ECODE_INVAL)
9446 self.dest_x509_ca = cert
9448 # Verify target information
9450 for idx, disk_data in enumerate(self.op.target_node):
9452 (host, port, magic) = \
9453 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9454 except errors.GenericError, err:
9455 raise errors.OpPrereqError("Target info for disk %s: %s" %
9456 (idx, err), errors.ECODE_INVAL)
9458 disk_info.append((host, port, magic))
9460 assert len(disk_info) == len(self.op.target_node)
9461 self.dest_disk_info = disk_info
9464 raise errors.ProgrammerError("Unhandled export mode %r" %
9467 # instance disk type verification
9468 # TODO: Implement export support for file-based disks
9469 for disk in self.instance.disks:
9470 if disk.dev_type == constants.LD_FILE:
9471 raise errors.OpPrereqError("Export not supported for instances with"
9472 " file-based disks", errors.ECODE_INVAL)
9474 def _CleanupExports(self, feedback_fn):
9475 """Removes exports of current instance from all other nodes.
9477 If an instance in a cluster with nodes A..D was exported to node C, its
9478 exports will be removed from the nodes A, B and D.
9481 assert self.op.mode != constants.EXPORT_MODE_REMOTE
9483 nodelist = self.cfg.GetNodeList()
9484 nodelist.remove(self.dst_node.name)
9486 # on one-node clusters nodelist will be empty after the removal
9487 # if we proceed the backup would be removed because OpQueryExports
9488 # substitutes an empty list with the full cluster node list.
9489 iname = self.instance.name
9491 feedback_fn("Removing old exports for instance %s" % iname)
9492 exportlist = self.rpc.call_export_list(nodelist)
9493 for node in exportlist:
9494 if exportlist[node].fail_msg:
9496 if iname in exportlist[node].payload:
9497 msg = self.rpc.call_export_remove(node, iname).fail_msg
9499 self.LogWarning("Could not remove older export for instance %s"
9500 " on node %s: %s", iname, node, msg)
9502 def Exec(self, feedback_fn):
9503 """Export an instance to an image in the cluster.
9506 assert self.op.mode in constants.EXPORT_MODES
9508 instance = self.instance
9509 src_node = instance.primary_node
9511 if self.op.shutdown:
9512 # shutdown the instance, but not the disks
9513 feedback_fn("Shutting down instance %s" % instance.name)
9514 result = self.rpc.call_instance_shutdown(src_node, instance,
9515 self.op.shutdown_timeout)
9516 # TODO: Maybe ignore failures if ignore_remove_failures is set
9517 result.Raise("Could not shutdown instance %s on"
9518 " node %s" % (instance.name, src_node))
9520 # set the disks ID correctly since call_instance_start needs the
9521 # correct drbd minor to create the symlinks
9522 for disk in instance.disks:
9523 self.cfg.SetDiskID(disk, src_node)
9525 activate_disks = (not instance.admin_up)
9528 # Activate the instance disks if we'exporting a stopped instance
9529 feedback_fn("Activating disks for %s" % instance.name)
9530 _StartInstanceDisks(self, instance, None)
9533 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9536 helper.CreateSnapshots()
9538 if (self.op.shutdown and instance.admin_up and
9539 not self.op.remove_instance):
9540 assert not activate_disks
9541 feedback_fn("Starting instance %s" % instance.name)
9542 result = self.rpc.call_instance_start(src_node, instance, None, None)
9543 msg = result.fail_msg
9545 feedback_fn("Failed to start instance: %s" % msg)
9546 _ShutdownInstanceDisks(self, instance)
9547 raise errors.OpExecError("Could not start instance: %s" % msg)
9549 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9550 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9551 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9552 connect_timeout = constants.RIE_CONNECT_TIMEOUT
9553 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9555 (key_name, _, _) = self.x509_key_name
9558 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9561 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9562 key_name, dest_ca_pem,
9567 # Check for backwards compatibility
9568 assert len(dresults) == len(instance.disks)
9569 assert compat.all(isinstance(i, bool) for i in dresults), \
9570 "Not all results are boolean: %r" % dresults
9574 feedback_fn("Deactivating disks for %s" % instance.name)
9575 _ShutdownInstanceDisks(self, instance)
9577 if not (compat.all(dresults) and fin_resu):
9580 failures.append("export finalization")
9581 if not compat.all(dresults):
9582 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9584 failures.append("disk export: disk(s) %s" % fdsk)
9586 raise errors.OpExecError("Export failed, errors in %s" %
9587 utils.CommaJoin(failures))
9589 # At this point, the export was successful, we can cleanup/finish
9591 # Remove instance if requested
9592 if self.op.remove_instance:
9593 feedback_fn("Removing instance %s" % instance.name)
9594 _RemoveInstance(self, feedback_fn, instance,
9595 self.op.ignore_remove_failures)
9597 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9598 self._CleanupExports(feedback_fn)
9600 return fin_resu, dresults
9603 class LURemoveExport(NoHooksLU):
9604 """Remove exports related to the named instance.
9612 def ExpandNames(self):
9613 self.needed_locks = {}
9614 # We need all nodes to be locked in order for RemoveExport to work, but we
9615 # don't need to lock the instance itself, as nothing will happen to it (and
9616 # we can remove exports also for a removed instance)
9617 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9619 def Exec(self, feedback_fn):
9620 """Remove any export.
9623 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9624 # If the instance was not found we'll try with the name that was passed in.
9625 # This will only work if it was an FQDN, though.
9627 if not instance_name:
9629 instance_name = self.op.instance_name
9631 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9632 exportlist = self.rpc.call_export_list(locked_nodes)
9634 for node in exportlist:
9635 msg = exportlist[node].fail_msg
9637 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9639 if instance_name in exportlist[node].payload:
9641 result = self.rpc.call_export_remove(node, instance_name)
9642 msg = result.fail_msg
9644 logging.error("Could not remove export for instance %s"
9645 " on node %s: %s", instance_name, node, msg)
9647 if fqdn_warn and not found:
9648 feedback_fn("Export not found. If trying to remove an export belonging"
9649 " to a deleted instance please use its Fully Qualified"
9653 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9656 This is an abstract class which is the parent of all the other tags LUs.
9660 def ExpandNames(self):
9661 self.needed_locks = {}
9662 if self.op.kind == constants.TAG_NODE:
9663 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9664 self.needed_locks[locking.LEVEL_NODE] = self.op.name
9665 elif self.op.kind == constants.TAG_INSTANCE:
9666 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9667 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9669 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
9670 # not possible to acquire the BGL based on opcode parameters)
9672 def CheckPrereq(self):
9673 """Check prerequisites.
9676 if self.op.kind == constants.TAG_CLUSTER:
9677 self.target = self.cfg.GetClusterInfo()
9678 elif self.op.kind == constants.TAG_NODE:
9679 self.target = self.cfg.GetNodeInfo(self.op.name)
9680 elif self.op.kind == constants.TAG_INSTANCE:
9681 self.target = self.cfg.GetInstanceInfo(self.op.name)
9683 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9684 str(self.op.kind), errors.ECODE_INVAL)
9687 class LUGetTags(TagsLU):
9688 """Returns the tags of a given object.
9692 ("kind", ht.NoDefault, ht.TElemOf(constants.VALID_TAG_TYPES)),
9693 # Name is only meaningful for nodes and instances
9694 ("name", ht.NoDefault, ht.TMaybeString),
9698 def ExpandNames(self):
9699 TagsLU.ExpandNames(self)
9701 # Share locks as this is only a read operation
9702 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9704 def Exec(self, feedback_fn):
9705 """Returns the tag list.
9708 return list(self.target.GetTags())
9711 class LUSearchTags(NoHooksLU):
9712 """Searches the tags for a given pattern.
9716 ("pattern", ht.NoDefault, ht.TNonEmptyString),
9720 def ExpandNames(self):
9721 self.needed_locks = {}
9723 def CheckPrereq(self):
9724 """Check prerequisites.
9726 This checks the pattern passed for validity by compiling it.
9730 self.re = re.compile(self.op.pattern)
9731 except re.error, err:
9732 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9733 (self.op.pattern, err), errors.ECODE_INVAL)
9735 def Exec(self, feedback_fn):
9736 """Returns the tag list.
9740 tgts = [("/cluster", cfg.GetClusterInfo())]
9741 ilist = cfg.GetAllInstancesInfo().values()
9742 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9743 nlist = cfg.GetAllNodesInfo().values()
9744 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9746 for path, target in tgts:
9747 for tag in target.GetTags():
9748 if self.re.search(tag):
9749 results.append((path, tag))
9753 class LUAddTags(TagsLU):
9754 """Sets a tag on a given object.
9758 ("kind", ht.NoDefault, ht.TElemOf(constants.VALID_TAG_TYPES)),
9759 # Name is only meaningful for nodes and instances
9760 ("name", ht.NoDefault, ht.TMaybeString),
9761 ("tags", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
9765 def CheckPrereq(self):
9766 """Check prerequisites.
9768 This checks the type and length of the tag name and value.
9771 TagsLU.CheckPrereq(self)
9772 for tag in self.op.tags:
9773 objects.TaggableObject.ValidateTag(tag)
9775 def Exec(self, feedback_fn):
9780 for tag in self.op.tags:
9781 self.target.AddTag(tag)
9782 except errors.TagError, err:
9783 raise errors.OpExecError("Error while setting tag: %s" % str(err))
9784 self.cfg.Update(self.target, feedback_fn)
9787 class LUDelTags(TagsLU):
9788 """Delete a list of tags from a given object.
9792 ("kind", ht.NoDefault, ht.TElemOf(constants.VALID_TAG_TYPES)),
9793 # Name is only meaningful for nodes and instances
9794 ("name", ht.NoDefault, ht.TMaybeString),
9795 ("tags", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
9799 def CheckPrereq(self):
9800 """Check prerequisites.
9802 This checks that we have the given tag.
9805 TagsLU.CheckPrereq(self)
9806 for tag in self.op.tags:
9807 objects.TaggableObject.ValidateTag(tag)
9808 del_tags = frozenset(self.op.tags)
9809 cur_tags = self.target.GetTags()
9811 diff_tags = del_tags - cur_tags
9813 diff_names = ("'%s'" % i for i in sorted(diff_tags))
9814 raise errors.OpPrereqError("Tag(s) %s not found" %
9815 (utils.CommaJoin(diff_names), ),
9818 def Exec(self, feedback_fn):
9819 """Remove the tag from the object.
9822 for tag in self.op.tags:
9823 self.target.RemoveTag(tag)
9824 self.cfg.Update(self.target, feedback_fn)
9827 class LUTestDelay(NoHooksLU):
9828 """Sleep for a specified amount of time.
9830 This LU sleeps on the master and/or nodes for a specified amount of
9835 ("duration", ht.NoDefault, ht.TFloat),
9836 ("on_master", True, ht.TBool),
9837 ("on_nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
9838 ("repeat", 0, ht.TPositiveInt)
9842 def ExpandNames(self):
9843 """Expand names and set required locks.
9845 This expands the node list, if any.
9848 self.needed_locks = {}
9849 if self.op.on_nodes:
9850 # _GetWantedNodes can be used here, but is not always appropriate to use
9851 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9853 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9854 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9856 def _TestDelay(self):
9857 """Do the actual sleep.
9860 if self.op.on_master:
9861 if not utils.TestDelay(self.op.duration):
9862 raise errors.OpExecError("Error during master delay test")
9863 if self.op.on_nodes:
9864 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9865 for node, node_result in result.items():
9866 node_result.Raise("Failure during rpc call to node %s" % node)
9868 def Exec(self, feedback_fn):
9869 """Execute the test delay opcode, with the wanted repetitions.
9872 if self.op.repeat == 0:
9875 top_value = self.op.repeat - 1
9876 for i in range(self.op.repeat):
9877 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
9881 class LUTestJobqueue(NoHooksLU):
9882 """Utility LU to test some aspects of the job queue.
9886 ("notify_waitlock", False, ht.TBool),
9887 ("notify_exec", False, ht.TBool),
9888 ("log_messages", ht.EmptyList, ht.TListOf(ht.TString)),
9889 ("fail", False, ht.TBool),
9893 # Must be lower than default timeout for WaitForJobChange to see whether it
9894 # notices changed jobs
9895 _CLIENT_CONNECT_TIMEOUT = 20.0
9896 _CLIENT_CONFIRM_TIMEOUT = 60.0
9899 def _NotifyUsingSocket(cls, cb, errcls):
9900 """Opens a Unix socket and waits for another program to connect.
9903 @param cb: Callback to send socket name to client
9905 @param errcls: Exception class to use for errors
9908 # Using a temporary directory as there's no easy way to create temporary
9909 # sockets without writing a custom loop around tempfile.mktemp and
9911 tmpdir = tempfile.mkdtemp()
9913 tmpsock = utils.PathJoin(tmpdir, "sock")
9915 logging.debug("Creating temporary socket at %s", tmpsock)
9916 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
9921 # Send details to client
9924 # Wait for client to connect before continuing
9925 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
9927 (conn, _) = sock.accept()
9928 except socket.error, err:
9929 raise errcls("Client didn't connect in time (%s)" % err)
9933 # Remove as soon as client is connected
9934 shutil.rmtree(tmpdir)
9936 # Wait for client to close
9939 # pylint: disable-msg=E1101
9940 # Instance of '_socketobject' has no ... member
9941 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
9943 except socket.error, err:
9944 raise errcls("Client failed to confirm notification (%s)" % err)
9948 def _SendNotification(self, test, arg, sockname):
9949 """Sends a notification to the client.
9952 @param test: Test name
9953 @param arg: Test argument (depends on test)
9954 @type sockname: string
9955 @param sockname: Socket path
9958 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
9960 def _Notify(self, prereq, test, arg):
9961 """Notifies the client of a test.
9964 @param prereq: Whether this is a prereq-phase test
9966 @param test: Test name
9967 @param arg: Test argument (depends on test)
9971 errcls = errors.OpPrereqError
9973 errcls = errors.OpExecError
9975 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
9979 def CheckArguments(self):
9980 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
9981 self.expandnames_calls = 0
9983 def ExpandNames(self):
9984 checkargs_calls = getattr(self, "checkargs_calls", 0)
9985 if checkargs_calls < 1:
9986 raise errors.ProgrammerError("CheckArguments was not called")
9988 self.expandnames_calls += 1
9990 if self.op.notify_waitlock:
9991 self._Notify(True, constants.JQT_EXPANDNAMES, None)
9993 self.LogInfo("Expanding names")
9995 # Get lock on master node (just to get a lock, not for a particular reason)
9996 self.needed_locks = {
9997 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10000 def Exec(self, feedback_fn):
10001 if self.expandnames_calls < 1:
10002 raise errors.ProgrammerError("ExpandNames was not called")
10004 if self.op.notify_exec:
10005 self._Notify(False, constants.JQT_EXEC, None)
10007 self.LogInfo("Executing")
10009 if self.op.log_messages:
10010 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10011 for idx, msg in enumerate(self.op.log_messages):
10012 self.LogInfo("Sending log message %s", idx + 1)
10013 feedback_fn(constants.JQT_MSGPREFIX + msg)
10014 # Report how many test messages have been sent
10015 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10018 raise errors.OpExecError("Opcode failure was requested")
10023 class IAllocator(object):
10024 """IAllocator framework.
10026 An IAllocator instance has three sets of attributes:
10027 - cfg that is needed to query the cluster
10028 - input data (all members of the _KEYS class attribute are required)
10029 - four buffer attributes (in|out_data|text), that represent the
10030 input (to the external script) in text and data structure format,
10031 and the output from it, again in two formats
10032 - the result variables from the script (success, info, nodes) for
10036 # pylint: disable-msg=R0902
10037 # lots of instance attributes
10039 "name", "mem_size", "disks", "disk_template",
10040 "os", "tags", "nics", "vcpus", "hypervisor",
10043 "name", "relocate_from",
10049 def __init__(self, cfg, rpc, mode, **kwargs):
10052 # init buffer variables
10053 self.in_text = self.out_text = self.in_data = self.out_data = None
10054 # init all input fields so that pylint is happy
10056 self.mem_size = self.disks = self.disk_template = None
10057 self.os = self.tags = self.nics = self.vcpus = None
10058 self.hypervisor = None
10059 self.relocate_from = None
10061 self.evac_nodes = None
10063 self.required_nodes = None
10064 # init result fields
10065 self.success = self.info = self.result = None
10066 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10067 keyset = self._ALLO_KEYS
10068 fn = self._AddNewInstance
10069 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10070 keyset = self._RELO_KEYS
10071 fn = self._AddRelocateInstance
10072 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10073 keyset = self._EVAC_KEYS
10074 fn = self._AddEvacuateNodes
10076 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10077 " IAllocator" % self.mode)
10079 if key not in keyset:
10080 raise errors.ProgrammerError("Invalid input parameter '%s' to"
10081 " IAllocator" % key)
10082 setattr(self, key, kwargs[key])
10085 if key not in kwargs:
10086 raise errors.ProgrammerError("Missing input parameter '%s' to"
10087 " IAllocator" % key)
10088 self._BuildInputData(fn)
10090 def _ComputeClusterData(self):
10091 """Compute the generic allocator input data.
10093 This is the data that is independent of the actual operation.
10097 cluster_info = cfg.GetClusterInfo()
10100 "version": constants.IALLOCATOR_VERSION,
10101 "cluster_name": cfg.GetClusterName(),
10102 "cluster_tags": list(cluster_info.GetTags()),
10103 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10104 # we don't have job IDs
10106 iinfo = cfg.GetAllInstancesInfo().values()
10107 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10111 node_list = cfg.GetNodeList()
10113 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10114 hypervisor_name = self.hypervisor
10115 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10116 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10117 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10118 hypervisor_name = cluster_info.enabled_hypervisors[0]
10120 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10123 self.rpc.call_all_instances_info(node_list,
10124 cluster_info.enabled_hypervisors)
10125 for nname, nresult in node_data.items():
10126 # first fill in static (config-based) values
10127 ninfo = cfg.GetNodeInfo(nname)
10129 "tags": list(ninfo.GetTags()),
10130 "primary_ip": ninfo.primary_ip,
10131 "secondary_ip": ninfo.secondary_ip,
10132 "offline": ninfo.offline,
10133 "drained": ninfo.drained,
10134 "master_candidate": ninfo.master_candidate,
10137 if not (ninfo.offline or ninfo.drained):
10138 nresult.Raise("Can't get data for node %s" % nname)
10139 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10141 remote_info = nresult.payload
10143 for attr in ['memory_total', 'memory_free', 'memory_dom0',
10144 'vg_size', 'vg_free', 'cpu_total']:
10145 if attr not in remote_info:
10146 raise errors.OpExecError("Node '%s' didn't return attribute"
10147 " '%s'" % (nname, attr))
10148 if not isinstance(remote_info[attr], int):
10149 raise errors.OpExecError("Node '%s' returned invalid value"
10151 (nname, attr, remote_info[attr]))
10152 # compute memory used by primary instances
10153 i_p_mem = i_p_up_mem = 0
10154 for iinfo, beinfo in i_list:
10155 if iinfo.primary_node == nname:
10156 i_p_mem += beinfo[constants.BE_MEMORY]
10157 if iinfo.name not in node_iinfo[nname].payload:
10160 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10161 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10162 remote_info['memory_free'] -= max(0, i_mem_diff)
10165 i_p_up_mem += beinfo[constants.BE_MEMORY]
10167 # compute memory used by instances
10169 "total_memory": remote_info['memory_total'],
10170 "reserved_memory": remote_info['memory_dom0'],
10171 "free_memory": remote_info['memory_free'],
10172 "total_disk": remote_info['vg_size'],
10173 "free_disk": remote_info['vg_free'],
10174 "total_cpus": remote_info['cpu_total'],
10175 "i_pri_memory": i_p_mem,
10176 "i_pri_up_memory": i_p_up_mem,
10178 pnr.update(pnr_dyn)
10180 node_results[nname] = pnr
10181 data["nodes"] = node_results
10185 for iinfo, beinfo in i_list:
10187 for nic in iinfo.nics:
10188 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10189 nic_dict = {"mac": nic.mac,
10191 "mode": filled_params[constants.NIC_MODE],
10192 "link": filled_params[constants.NIC_LINK],
10194 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10195 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10196 nic_data.append(nic_dict)
10198 "tags": list(iinfo.GetTags()),
10199 "admin_up": iinfo.admin_up,
10200 "vcpus": beinfo[constants.BE_VCPUS],
10201 "memory": beinfo[constants.BE_MEMORY],
10203 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10205 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10206 "disk_template": iinfo.disk_template,
10207 "hypervisor": iinfo.hypervisor,
10209 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10211 instance_data[iinfo.name] = pir
10213 data["instances"] = instance_data
10215 self.in_data = data
10217 def _AddNewInstance(self):
10218 """Add new instance data to allocator structure.
10220 This in combination with _AllocatorGetClusterData will create the
10221 correct structure needed as input for the allocator.
10223 The checks for the completeness of the opcode must have already been
10227 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10229 if self.disk_template in constants.DTS_NET_MIRROR:
10230 self.required_nodes = 2
10232 self.required_nodes = 1
10235 "disk_template": self.disk_template,
10238 "vcpus": self.vcpus,
10239 "memory": self.mem_size,
10240 "disks": self.disks,
10241 "disk_space_total": disk_space,
10243 "required_nodes": self.required_nodes,
10247 def _AddRelocateInstance(self):
10248 """Add relocate instance data to allocator structure.
10250 This in combination with _IAllocatorGetClusterData will create the
10251 correct structure needed as input for the allocator.
10253 The checks for the completeness of the opcode must have already been
10257 instance = self.cfg.GetInstanceInfo(self.name)
10258 if instance is None:
10259 raise errors.ProgrammerError("Unknown instance '%s' passed to"
10260 " IAllocator" % self.name)
10262 if instance.disk_template not in constants.DTS_NET_MIRROR:
10263 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
10264 errors.ECODE_INVAL)
10266 if len(instance.secondary_nodes) != 1:
10267 raise errors.OpPrereqError("Instance has not exactly one secondary node",
10268 errors.ECODE_STATE)
10270 self.required_nodes = 1
10271 disk_sizes = [{'size': disk.size} for disk in instance.disks]
10272 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
10276 "disk_space_total": disk_space,
10277 "required_nodes": self.required_nodes,
10278 "relocate_from": self.relocate_from,
10282 def _AddEvacuateNodes(self):
10283 """Add evacuate nodes data to allocator structure.
10287 "evac_nodes": self.evac_nodes
10291 def _BuildInputData(self, fn):
10292 """Build input data structures.
10295 self._ComputeClusterData()
10298 request["type"] = self.mode
10299 self.in_data["request"] = request
10301 self.in_text = serializer.Dump(self.in_data)
10303 def Run(self, name, validate=True, call_fn=None):
10304 """Run an instance allocator and return the results.
10307 if call_fn is None:
10308 call_fn = self.rpc.call_iallocator_runner
10310 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
10311 result.Raise("Failure while running the iallocator script")
10313 self.out_text = result.payload
10315 self._ValidateResult()
10317 def _ValidateResult(self):
10318 """Process the allocator results.
10320 This will process and if successful save the result in
10321 self.out_data and the other parameters.
10325 rdict = serializer.Load(self.out_text)
10326 except Exception, err:
10327 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
10329 if not isinstance(rdict, dict):
10330 raise errors.OpExecError("Can't parse iallocator results: not a dict")
10332 # TODO: remove backwards compatiblity in later versions
10333 if "nodes" in rdict and "result" not in rdict:
10334 rdict["result"] = rdict["nodes"]
10337 for key in "success", "info", "result":
10338 if key not in rdict:
10339 raise errors.OpExecError("Can't parse iallocator results:"
10340 " missing key '%s'" % key)
10341 setattr(self, key, rdict[key])
10343 if not isinstance(rdict["result"], list):
10344 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
10346 self.out_data = rdict
10349 class LUTestAllocator(NoHooksLU):
10350 """Run allocator tests.
10352 This LU runs the allocator tests
10356 ("direction", ht.NoDefault,
10357 ht.TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
10358 ("mode", ht.NoDefault, ht.TElemOf(constants.VALID_IALLOCATOR_MODES)),
10359 ("name", ht.NoDefault, ht.TNonEmptyString),
10360 ("nics", ht.NoDefault, ht.TOr(ht.TNone, ht.TListOf(
10361 ht.TDictOf(ht.TElemOf(["mac", "ip", "bridge"]),
10362 ht.TOr(ht.TNone, ht.TNonEmptyString))))),
10363 ("disks", ht.NoDefault, ht.TOr(ht.TNone, ht.TList)),
10364 ("hypervisor", None, ht.TMaybeString),
10365 ("allocator", None, ht.TMaybeString),
10366 ("tags", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
10367 ("mem_size", None, ht.TOr(ht.TNone, ht.TPositiveInt)),
10368 ("vcpus", None, ht.TOr(ht.TNone, ht.TPositiveInt)),
10369 ("os", None, ht.TMaybeString),
10370 ("disk_template", None, ht.TMaybeString),
10371 ("evac_nodes", None, ht.TOr(ht.TNone, ht.TListOf(ht.TNonEmptyString))),
10374 def CheckPrereq(self):
10375 """Check prerequisites.
10377 This checks the opcode parameters depending on the director and mode test.
10380 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10381 for attr in ["mem_size", "disks", "disk_template",
10382 "os", "tags", "nics", "vcpus"]:
10383 if not hasattr(self.op, attr):
10384 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
10385 attr, errors.ECODE_INVAL)
10386 iname = self.cfg.ExpandInstanceName(self.op.name)
10387 if iname is not None:
10388 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
10389 iname, errors.ECODE_EXISTS)
10390 if not isinstance(self.op.nics, list):
10391 raise errors.OpPrereqError("Invalid parameter 'nics'",
10392 errors.ECODE_INVAL)
10393 if not isinstance(self.op.disks, list):
10394 raise errors.OpPrereqError("Invalid parameter 'disks'",
10395 errors.ECODE_INVAL)
10396 for row in self.op.disks:
10397 if (not isinstance(row, dict) or
10398 "size" not in row or
10399 not isinstance(row["size"], int) or
10400 "mode" not in row or
10401 row["mode"] not in ['r', 'w']):
10402 raise errors.OpPrereqError("Invalid contents of the 'disks'"
10403 " parameter", errors.ECODE_INVAL)
10404 if self.op.hypervisor is None:
10405 self.op.hypervisor = self.cfg.GetHypervisorType()
10406 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10407 fname = _ExpandInstanceName(self.cfg, self.op.name)
10408 self.op.name = fname
10409 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
10410 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10411 if not hasattr(self.op, "evac_nodes"):
10412 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
10413 " opcode input", errors.ECODE_INVAL)
10415 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
10416 self.op.mode, errors.ECODE_INVAL)
10418 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
10419 if self.op.allocator is None:
10420 raise errors.OpPrereqError("Missing allocator name",
10421 errors.ECODE_INVAL)
10422 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
10423 raise errors.OpPrereqError("Wrong allocator test '%s'" %
10424 self.op.direction, errors.ECODE_INVAL)
10426 def Exec(self, feedback_fn):
10427 """Run the allocator test.
10430 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10431 ial = IAllocator(self.cfg, self.rpc,
10434 mem_size=self.op.mem_size,
10435 disks=self.op.disks,
10436 disk_template=self.op.disk_template,
10440 vcpus=self.op.vcpus,
10441 hypervisor=self.op.hypervisor,
10443 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10444 ial = IAllocator(self.cfg, self.rpc,
10447 relocate_from=list(self.relocate_from),
10449 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10450 ial = IAllocator(self.cfg, self.rpc,
10452 evac_nodes=self.op.evac_nodes)
10454 raise errors.ProgrammerError("Uncatched mode %s in"
10455 " LUTestAllocator.Exec", self.op.mode)
10457 if self.op.direction == constants.IALLOCATOR_DIR_IN:
10458 result = ial.in_text
10460 ial.Run(self.op.allocator, validate=False)
10461 result = ial.out_text