4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Module implementing the master-side code."""
24 # pylint: disable-msg=W0201,C0302
26 # W0201 since most LU attributes are defined in CheckPrereq or similar
29 # C0302: since we have waaaay to many lines in this module
44 from ganeti import ssh
45 from ganeti import utils
46 from ganeti import errors
47 from ganeti import hypervisor
48 from ganeti import locking
49 from ganeti import constants
50 from ganeti import objects
51 from ganeti import serializer
52 from ganeti import ssconf
53 from ganeti import uidpool
54 from ganeti import compat
55 from ganeti import masterd
56 from ganeti import netutils
57 from ganeti import query
58 from ganeti import qlang
59 from ganeti import opcodes
61 import ganeti.masterd.instance # pylint: disable-msg=W0611
64 def _SupportsOob(cfg, node):
65 """Tells if node supports OOB.
67 @type cfg: L{config.ConfigWriter}
68 @param cfg: The cluster configuration
69 @type node: L{objects.Node}
71 @return: The OOB script if supported or an empty string otherwise
74 return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
78 class LogicalUnit(object):
79 """Logical Unit base class.
81 Subclasses must follow these rules:
82 - implement ExpandNames
83 - implement CheckPrereq (except when tasklets are used)
84 - implement Exec (except when tasklets are used)
85 - implement BuildHooksEnv
86 - redefine HPATH and HTYPE
87 - optionally redefine their run requirements:
88 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
90 Note that all commands require root permissions.
92 @ivar dry_run_result: the value (if any) that will be returned to the caller
93 in dry-run mode (signalled by opcode dry_run parameter)
100 def __init__(self, processor, op, context, rpc):
101 """Constructor for LogicalUnit.
103 This needs to be overridden in derived classes in order to check op
107 self.proc = processor
109 self.cfg = context.cfg
110 self.context = context
112 # Dicts used to declare locking needs to mcpu
113 self.needed_locks = None
114 self.acquired_locks = {}
115 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
117 self.remove_locks = {}
118 # Used to force good behavior when calling helper functions
119 self.recalculate_locks = {}
122 self.Log = processor.Log # pylint: disable-msg=C0103
123 self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
124 self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
125 self.LogStep = processor.LogStep # pylint: disable-msg=C0103
126 # support for dry-run
127 self.dry_run_result = None
128 # support for generic debug attribute
129 if (not hasattr(self.op, "debug_level") or
130 not isinstance(self.op.debug_level, int)):
131 self.op.debug_level = 0
136 # Validate opcode parameters and set defaults
137 self.op.Validate(True)
139 self.CheckArguments()
142 """Returns the SshRunner object
146 self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
149 ssh = property(fget=__GetSSH)
151 def CheckArguments(self):
152 """Check syntactic validity for the opcode arguments.
154 This method is for doing a simple syntactic check and ensure
155 validity of opcode parameters, without any cluster-related
156 checks. While the same can be accomplished in ExpandNames and/or
157 CheckPrereq, doing these separate is better because:
159 - ExpandNames is left as as purely a lock-related function
160 - CheckPrereq is run after we have acquired locks (and possible
163 The function is allowed to change the self.op attribute so that
164 later methods can no longer worry about missing parameters.
169 def ExpandNames(self):
170 """Expand names for this LU.
172 This method is called before starting to execute the opcode, and it should
173 update all the parameters of the opcode to their canonical form (e.g. a
174 short node name must be fully expanded after this method has successfully
175 completed). This way locking, hooks, logging, etc. can work correctly.
177 LUs which implement this method must also populate the self.needed_locks
178 member, as a dict with lock levels as keys, and a list of needed lock names
181 - use an empty dict if you don't need any lock
182 - if you don't need any lock at a particular level omit that level
183 - don't put anything for the BGL level
184 - if you want all locks at a level use locking.ALL_SET as a value
186 If you need to share locks (rather than acquire them exclusively) at one
187 level you can modify self.share_locks, setting a true value (usually 1) for
188 that level. By default locks are not shared.
190 This function can also define a list of tasklets, which then will be
191 executed in order instead of the usual LU-level CheckPrereq and Exec
192 functions, if those are not defined by the LU.
196 # Acquire all nodes and one instance
197 self.needed_locks = {
198 locking.LEVEL_NODE: locking.ALL_SET,
199 locking.LEVEL_INSTANCE: ['instance1.example.com'],
201 # Acquire just two nodes
202 self.needed_locks = {
203 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
206 self.needed_locks = {} # No, you can't leave it to the default value None
209 # The implementation of this method is mandatory only if the new LU is
210 # concurrent, so that old LUs don't need to be changed all at the same
213 self.needed_locks = {} # Exclusive LUs don't need locks.
215 raise NotImplementedError
217 def DeclareLocks(self, level):
218 """Declare LU locking needs for a level
220 While most LUs can just declare their locking needs at ExpandNames time,
221 sometimes there's the need to calculate some locks after having acquired
222 the ones before. This function is called just before acquiring locks at a
223 particular level, but after acquiring the ones at lower levels, and permits
224 such calculations. It can be used to modify self.needed_locks, and by
225 default it does nothing.
227 This function is only called if you have something already set in
228 self.needed_locks for the level.
230 @param level: Locking level which is going to be locked
231 @type level: member of ganeti.locking.LEVELS
235 def CheckPrereq(self):
236 """Check prerequisites for this LU.
238 This method should check that the prerequisites for the execution
239 of this LU are fulfilled. It can do internode communication, but
240 it should be idempotent - no cluster or system changes are
243 The method should raise errors.OpPrereqError in case something is
244 not fulfilled. Its return value is ignored.
246 This method should also update all the parameters of the opcode to
247 their canonical form if it hasn't been done by ExpandNames before.
250 if self.tasklets is not None:
251 for (idx, tl) in enumerate(self.tasklets):
252 logging.debug("Checking prerequisites for tasklet %s/%s",
253 idx + 1, len(self.tasklets))
258 def Exec(self, feedback_fn):
261 This method should implement the actual work. It should raise
262 errors.OpExecError for failures that are somewhat dealt with in
266 if self.tasklets is not None:
267 for (idx, tl) in enumerate(self.tasklets):
268 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
271 raise NotImplementedError
273 def BuildHooksEnv(self):
274 """Build hooks environment for this LU.
276 This method should return a three-node tuple consisting of: a dict
277 containing the environment that will be used for running the
278 specific hook for this LU, a list of node names on which the hook
279 should run before the execution, and a list of node names on which
280 the hook should run after the execution.
282 The keys of the dict must not have 'GANETI_' prefixed as this will
283 be handled in the hooks runner. Also note additional keys will be
284 added by the hooks runner. If the LU doesn't define any
285 environment, an empty dict (and not None) should be returned.
287 No nodes should be returned as an empty list (and not None).
289 Note that if the HPATH for a LU class is None, this function will
293 raise NotImplementedError
295 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
296 """Notify the LU about the results of its hooks.
298 This method is called every time a hooks phase is executed, and notifies
299 the Logical Unit about the hooks' result. The LU can then use it to alter
300 its result based on the hooks. By default the method does nothing and the
301 previous result is passed back unchanged but any LU can define it if it
302 wants to use the local cluster hook-scripts somehow.
304 @param phase: one of L{constants.HOOKS_PHASE_POST} or
305 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
306 @param hook_results: the results of the multi-node hooks rpc call
307 @param feedback_fn: function used send feedback back to the caller
308 @param lu_result: the previous Exec result this LU had, or None
310 @return: the new Exec result, based on the previous result
314 # API must be kept, thus we ignore the unused argument and could
315 # be a function warnings
316 # pylint: disable-msg=W0613,R0201
319 def _ExpandAndLockInstance(self):
320 """Helper function to expand and lock an instance.
322 Many LUs that work on an instance take its name in self.op.instance_name
323 and need to expand it and then declare the expanded name for locking. This
324 function does it, and then updates self.op.instance_name to the expanded
325 name. It also initializes needed_locks as a dict, if this hasn't been done
329 if self.needed_locks is None:
330 self.needed_locks = {}
332 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
333 "_ExpandAndLockInstance called with instance-level locks set"
334 self.op.instance_name = _ExpandInstanceName(self.cfg,
335 self.op.instance_name)
336 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
338 def _LockInstancesNodes(self, primary_only=False):
339 """Helper function to declare instances' nodes for locking.
341 This function should be called after locking one or more instances to lock
342 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
343 with all primary or secondary nodes for instances already locked and
344 present in self.needed_locks[locking.LEVEL_INSTANCE].
346 It should be called from DeclareLocks, and for safety only works if
347 self.recalculate_locks[locking.LEVEL_NODE] is set.
349 In the future it may grow parameters to just lock some instance's nodes, or
350 to just lock primaries or secondary nodes, if needed.
352 If should be called in DeclareLocks in a way similar to::
354 if level == locking.LEVEL_NODE:
355 self._LockInstancesNodes()
357 @type primary_only: boolean
358 @param primary_only: only lock primary nodes of locked instances
361 assert locking.LEVEL_NODE in self.recalculate_locks, \
362 "_LockInstancesNodes helper function called with no nodes to recalculate"
364 # TODO: check if we're really been called with the instance locks held
366 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
367 # future we might want to have different behaviors depending on the value
368 # of self.recalculate_locks[locking.LEVEL_NODE]
370 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
371 instance = self.context.cfg.GetInstanceInfo(instance_name)
372 wanted_nodes.append(instance.primary_node)
374 wanted_nodes.extend(instance.secondary_nodes)
376 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
377 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
378 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
379 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
381 del self.recalculate_locks[locking.LEVEL_NODE]
384 class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
385 """Simple LU which runs no hooks.
387 This LU is intended as a parent for other LogicalUnits which will
388 run no hooks, in order to reduce duplicate code.
394 def BuildHooksEnv(self):
395 """Empty BuildHooksEnv for NoHooksLu.
397 This just raises an error.
400 assert False, "BuildHooksEnv called for NoHooksLUs"
404 """Tasklet base class.
406 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
407 they can mix legacy code with tasklets. Locking needs to be done in the LU,
408 tasklets know nothing about locks.
410 Subclasses must follow these rules:
411 - Implement CheckPrereq
415 def __init__(self, lu):
422 def CheckPrereq(self):
423 """Check prerequisites for this tasklets.
425 This method should check whether the prerequisites for the execution of
426 this tasklet are fulfilled. It can do internode communication, but it
427 should be idempotent - no cluster or system changes are allowed.
429 The method should raise errors.OpPrereqError in case something is not
430 fulfilled. Its return value is ignored.
432 This method should also update all parameters to their canonical form if it
433 hasn't been done before.
438 def Exec(self, feedback_fn):
439 """Execute the tasklet.
441 This method should implement the actual work. It should raise
442 errors.OpExecError for failures that are somewhat dealt with in code, or
446 raise NotImplementedError
450 """Base for query utility classes.
453 #: Attribute holding field definitions
456 def __init__(self, filter_, fields, use_locking):
457 """Initializes this class.
460 self.use_locking = use_locking
462 self.query = query.Query(self.FIELDS, fields, filter_=filter_,
464 self.requested_data = self.query.RequestedData()
465 self.names = self.query.RequestedNames()
467 # Sort only if no names were requested
468 self.sort_by_name = not self.names
470 self.do_locking = None
473 def _GetNames(self, lu, all_names, lock_level):
474 """Helper function to determine names asked for in the query.
478 names = lu.acquired_locks[lock_level]
482 if self.wanted == locking.ALL_SET:
483 assert not self.names
484 # caller didn't specify names, so ordering is not important
485 return utils.NiceSort(names)
487 # caller specified names and we must keep the same order
489 assert not self.do_locking or lu.acquired_locks[lock_level]
491 missing = set(self.wanted).difference(names)
493 raise errors.OpExecError("Some items were removed before retrieving"
494 " their data: %s" % missing)
496 # Return expanded names
500 def FieldsQuery(cls, fields):
501 """Returns list of available fields.
503 @return: List of L{objects.QueryFieldDefinition}
506 return query.QueryFields(cls.FIELDS, fields)
508 def ExpandNames(self, lu):
509 """Expand names for this query.
511 See L{LogicalUnit.ExpandNames}.
514 raise NotImplementedError()
516 def DeclareLocks(self, lu, level):
517 """Declare locks for this query.
519 See L{LogicalUnit.DeclareLocks}.
522 raise NotImplementedError()
524 def _GetQueryData(self, lu):
525 """Collects all data for this query.
527 @return: Query data object
530 raise NotImplementedError()
532 def NewStyleQuery(self, lu):
533 """Collect data and execute query.
536 return query.GetQueryResponse(self.query, self._GetQueryData(lu),
537 sort_by_name=self.sort_by_name)
539 def OldStyleQuery(self, lu):
540 """Collect data and execute query.
543 return self.query.OldStyleQuery(self._GetQueryData(lu),
544 sort_by_name=self.sort_by_name)
547 def _GetWantedNodes(lu, nodes):
548 """Returns list of checked and expanded node names.
550 @type lu: L{LogicalUnit}
551 @param lu: the logical unit on whose behalf we execute
553 @param nodes: list of node names or None for all nodes
555 @return: the list of nodes, sorted
556 @raise errors.ProgrammerError: if the nodes parameter is wrong type
560 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
562 return utils.NiceSort(lu.cfg.GetNodeList())
565 def _GetWantedInstances(lu, instances):
566 """Returns list of checked and expanded instance names.
568 @type lu: L{LogicalUnit}
569 @param lu: the logical unit on whose behalf we execute
570 @type instances: list
571 @param instances: list of instance names or None for all instances
573 @return: the list of instances, sorted
574 @raise errors.OpPrereqError: if the instances parameter is wrong type
575 @raise errors.OpPrereqError: if any of the passed instances is not found
579 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
581 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
585 def _GetUpdatedParams(old_params, update_dict,
586 use_default=True, use_none=False):
587 """Return the new version of a parameter dictionary.
589 @type old_params: dict
590 @param old_params: old parameters
591 @type update_dict: dict
592 @param update_dict: dict containing new parameter values, or
593 constants.VALUE_DEFAULT to reset the parameter to its default
595 @param use_default: boolean
596 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
597 values as 'to be deleted' values
598 @param use_none: boolean
599 @type use_none: whether to recognise C{None} values as 'to be
602 @return: the new parameter dictionary
605 params_copy = copy.deepcopy(old_params)
606 for key, val in update_dict.iteritems():
607 if ((use_default and val == constants.VALUE_DEFAULT) or
608 (use_none and val is None)):
614 params_copy[key] = val
618 def _CheckOutputFields(static, dynamic, selected):
619 """Checks whether all selected fields are valid.
621 @type static: L{utils.FieldSet}
622 @param static: static fields set
623 @type dynamic: L{utils.FieldSet}
624 @param dynamic: dynamic fields set
631 delta = f.NonMatching(selected)
633 raise errors.OpPrereqError("Unknown output fields selected: %s"
634 % ",".join(delta), errors.ECODE_INVAL)
637 def _CheckGlobalHvParams(params):
638 """Validates that given hypervisor params are not global ones.
640 This will ensure that instances don't get customised versions of
644 used_globals = constants.HVC_GLOBALS.intersection(params)
646 msg = ("The following hypervisor parameters are global and cannot"
647 " be customized at instance level, please modify them at"
648 " cluster level: %s" % utils.CommaJoin(used_globals))
649 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
652 def _CheckNodeOnline(lu, node, msg=None):
653 """Ensure that a given node is online.
655 @param lu: the LU on behalf of which we make the check
656 @param node: the node to check
657 @param msg: if passed, should be a message to replace the default one
658 @raise errors.OpPrereqError: if the node is offline
662 msg = "Can't use offline node"
663 if lu.cfg.GetNodeInfo(node).offline:
664 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
667 def _CheckNodeNotDrained(lu, node):
668 """Ensure that a given node is not drained.
670 @param lu: the LU on behalf of which we make the check
671 @param node: the node to check
672 @raise errors.OpPrereqError: if the node is drained
675 if lu.cfg.GetNodeInfo(node).drained:
676 raise errors.OpPrereqError("Can't use drained node %s" % node,
680 def _CheckNodeVmCapable(lu, node):
681 """Ensure that a given node is vm capable.
683 @param lu: the LU on behalf of which we make the check
684 @param node: the node to check
685 @raise errors.OpPrereqError: if the node is not vm capable
688 if not lu.cfg.GetNodeInfo(node).vm_capable:
689 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
693 def _CheckNodeHasOS(lu, node, os_name, force_variant):
694 """Ensure that a node supports a given OS.
696 @param lu: the LU on behalf of which we make the check
697 @param node: the node to check
698 @param os_name: the OS to query about
699 @param force_variant: whether to ignore variant errors
700 @raise errors.OpPrereqError: if the node is not supporting the OS
703 result = lu.rpc.call_os_get(node, os_name)
704 result.Raise("OS '%s' not in supported OS list for node %s" %
706 prereq=True, ecode=errors.ECODE_INVAL)
707 if not force_variant:
708 _CheckOSVariant(result.payload, os_name)
711 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
712 """Ensure that a node has the given secondary ip.
714 @type lu: L{LogicalUnit}
715 @param lu: the LU on behalf of which we make the check
717 @param node: the node to check
718 @type secondary_ip: string
719 @param secondary_ip: the ip to check
720 @type prereq: boolean
721 @param prereq: whether to throw a prerequisite or an execute error
722 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
723 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
726 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
727 result.Raise("Failure checking secondary ip on node %s" % node,
728 prereq=prereq, ecode=errors.ECODE_ENVIRON)
729 if not result.payload:
730 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
731 " please fix and re-run this command" % secondary_ip)
733 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
735 raise errors.OpExecError(msg)
738 def _GetClusterDomainSecret():
739 """Reads the cluster domain secret.
742 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
746 def _CheckInstanceDown(lu, instance, reason):
747 """Ensure that an instance is not running."""
748 if instance.admin_up:
749 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
750 (instance.name, reason), errors.ECODE_STATE)
752 pnode = instance.primary_node
753 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
754 ins_l.Raise("Can't contact node %s for instance information" % pnode,
755 prereq=True, ecode=errors.ECODE_ENVIRON)
757 if instance.name in ins_l.payload:
758 raise errors.OpPrereqError("Instance %s is running, %s" %
759 (instance.name, reason), errors.ECODE_STATE)
762 def _ExpandItemName(fn, name, kind):
763 """Expand an item name.
765 @param fn: the function to use for expansion
766 @param name: requested item name
767 @param kind: text description ('Node' or 'Instance')
768 @return: the resolved (full) name
769 @raise errors.OpPrereqError: if the item is not found
773 if full_name is None:
774 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
779 def _ExpandNodeName(cfg, name):
780 """Wrapper over L{_ExpandItemName} for nodes."""
781 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
784 def _ExpandInstanceName(cfg, name):
785 """Wrapper over L{_ExpandItemName} for instance."""
786 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
789 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
790 memory, vcpus, nics, disk_template, disks,
791 bep, hvp, hypervisor_name):
792 """Builds instance related env variables for hooks
794 This builds the hook environment from individual variables.
797 @param name: the name of the instance
798 @type primary_node: string
799 @param primary_node: the name of the instance's primary node
800 @type secondary_nodes: list
801 @param secondary_nodes: list of secondary nodes as strings
802 @type os_type: string
803 @param os_type: the name of the instance's OS
804 @type status: boolean
805 @param status: the should_run status of the instance
807 @param memory: the memory size of the instance
809 @param vcpus: the count of VCPUs the instance has
811 @param nics: list of tuples (ip, mac, mode, link) representing
812 the NICs the instance has
813 @type disk_template: string
814 @param disk_template: the disk template of the instance
816 @param disks: the list of (size, mode) pairs
818 @param bep: the backend parameters for the instance
820 @param hvp: the hypervisor parameters for the instance
821 @type hypervisor_name: string
822 @param hypervisor_name: the hypervisor for the instance
824 @return: the hook environment for this instance
833 "INSTANCE_NAME": name,
834 "INSTANCE_PRIMARY": primary_node,
835 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
836 "INSTANCE_OS_TYPE": os_type,
837 "INSTANCE_STATUS": str_status,
838 "INSTANCE_MEMORY": memory,
839 "INSTANCE_VCPUS": vcpus,
840 "INSTANCE_DISK_TEMPLATE": disk_template,
841 "INSTANCE_HYPERVISOR": hypervisor_name,
845 nic_count = len(nics)
846 for idx, (ip, mac, mode, link) in enumerate(nics):
849 env["INSTANCE_NIC%d_IP" % idx] = ip
850 env["INSTANCE_NIC%d_MAC" % idx] = mac
851 env["INSTANCE_NIC%d_MODE" % idx] = mode
852 env["INSTANCE_NIC%d_LINK" % idx] = link
853 if mode == constants.NIC_MODE_BRIDGED:
854 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
858 env["INSTANCE_NIC_COUNT"] = nic_count
861 disk_count = len(disks)
862 for idx, (size, mode) in enumerate(disks):
863 env["INSTANCE_DISK%d_SIZE" % idx] = size
864 env["INSTANCE_DISK%d_MODE" % idx] = mode
868 env["INSTANCE_DISK_COUNT"] = disk_count
870 for source, kind in [(bep, "BE"), (hvp, "HV")]:
871 for key, value in source.items():
872 env["INSTANCE_%s_%s" % (kind, key)] = value
877 def _NICListToTuple(lu, nics):
878 """Build a list of nic information tuples.
880 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
881 value in LUInstanceQueryData.
883 @type lu: L{LogicalUnit}
884 @param lu: the logical unit on whose behalf we execute
885 @type nics: list of L{objects.NIC}
886 @param nics: list of nics to convert to hooks tuples
890 cluster = lu.cfg.GetClusterInfo()
894 filled_params = cluster.SimpleFillNIC(nic.nicparams)
895 mode = filled_params[constants.NIC_MODE]
896 link = filled_params[constants.NIC_LINK]
897 hooks_nics.append((ip, mac, mode, link))
901 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
902 """Builds instance related env variables for hooks from an object.
904 @type lu: L{LogicalUnit}
905 @param lu: the logical unit on whose behalf we execute
906 @type instance: L{objects.Instance}
907 @param instance: the instance for which we should build the
910 @param override: dictionary with key/values that will override
913 @return: the hook environment dictionary
916 cluster = lu.cfg.GetClusterInfo()
917 bep = cluster.FillBE(instance)
918 hvp = cluster.FillHV(instance)
920 'name': instance.name,
921 'primary_node': instance.primary_node,
922 'secondary_nodes': instance.secondary_nodes,
923 'os_type': instance.os,
924 'status': instance.admin_up,
925 'memory': bep[constants.BE_MEMORY],
926 'vcpus': bep[constants.BE_VCPUS],
927 'nics': _NICListToTuple(lu, instance.nics),
928 'disk_template': instance.disk_template,
929 'disks': [(disk.size, disk.mode) for disk in instance.disks],
932 'hypervisor_name': instance.hypervisor,
935 args.update(override)
936 return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
939 def _AdjustCandidatePool(lu, exceptions):
940 """Adjust the candidate pool after node operations.
943 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
945 lu.LogInfo("Promoted nodes to master candidate role: %s",
946 utils.CommaJoin(node.name for node in mod_list))
947 for name in mod_list:
948 lu.context.ReaddNode(name)
949 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
951 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
955 def _DecideSelfPromotion(lu, exceptions=None):
956 """Decide whether I should promote myself as a master candidate.
959 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
960 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
961 # the new node will increase mc_max with one, so:
962 mc_should = min(mc_should + 1, cp_size)
963 return mc_now < mc_should
966 def _CheckNicsBridgesExist(lu, target_nics, target_node):
967 """Check that the brigdes needed by a list of nics exist.
970 cluster = lu.cfg.GetClusterInfo()
971 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
972 brlist = [params[constants.NIC_LINK] for params in paramslist
973 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
975 result = lu.rpc.call_bridges_exist(target_node, brlist)
976 result.Raise("Error checking bridges on destination node '%s'" %
977 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
980 def _CheckInstanceBridgesExist(lu, instance, node=None):
981 """Check that the brigdes needed by an instance exist.
985 node = instance.primary_node
986 _CheckNicsBridgesExist(lu, instance.nics, node)
989 def _CheckOSVariant(os_obj, name):
990 """Check whether an OS name conforms to the os variants specification.
992 @type os_obj: L{objects.OS}
993 @param os_obj: OS object to check
995 @param name: OS name passed by the user, to check for validity
998 if not os_obj.supported_variants:
1000 variant = objects.OS.GetVariant(name)
1002 raise errors.OpPrereqError("OS name must include a variant",
1005 if variant not in os_obj.supported_variants:
1006 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1009 def _GetNodeInstancesInner(cfg, fn):
1010 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1013 def _GetNodeInstances(cfg, node_name):
1014 """Returns a list of all primary and secondary instances on a node.
1018 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1021 def _GetNodePrimaryInstances(cfg, node_name):
1022 """Returns primary instances on a node.
1025 return _GetNodeInstancesInner(cfg,
1026 lambda inst: node_name == inst.primary_node)
1029 def _GetNodeSecondaryInstances(cfg, node_name):
1030 """Returns secondary instances on a node.
1033 return _GetNodeInstancesInner(cfg,
1034 lambda inst: node_name in inst.secondary_nodes)
1037 def _GetStorageTypeArgs(cfg, storage_type):
1038 """Returns the arguments for a storage type.
1041 # Special case for file storage
1042 if storage_type == constants.ST_FILE:
1043 # storage.FileStorage wants a list of storage directories
1044 return [[cfg.GetFileStorageDir(), cfg.GetSharedFileStorageDir()]]
1049 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1052 for dev in instance.disks:
1053 cfg.SetDiskID(dev, node_name)
1055 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1056 result.Raise("Failed to get disk status from node %s" % node_name,
1057 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1059 for idx, bdev_status in enumerate(result.payload):
1060 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1066 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1067 """Check the sanity of iallocator and node arguments and use the
1068 cluster-wide iallocator if appropriate.
1070 Check that at most one of (iallocator, node) is specified. If none is
1071 specified, then the LU's opcode's iallocator slot is filled with the
1072 cluster-wide default iallocator.
1074 @type iallocator_slot: string
1075 @param iallocator_slot: the name of the opcode iallocator slot
1076 @type node_slot: string
1077 @param node_slot: the name of the opcode target node slot
1080 node = getattr(lu.op, node_slot, None)
1081 iallocator = getattr(lu.op, iallocator_slot, None)
1083 if node is not None and iallocator is not None:
1084 raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1086 elif node is None and iallocator is None:
1087 default_iallocator = lu.cfg.GetDefaultIAllocator()
1088 if default_iallocator:
1089 setattr(lu.op, iallocator_slot, default_iallocator)
1091 raise errors.OpPrereqError("No iallocator or node given and no"
1092 " cluster-wide default iallocator found."
1093 " Please specify either an iallocator or a"
1094 " node, or set a cluster-wide default"
1098 class LUClusterPostInit(LogicalUnit):
1099 """Logical unit for running hooks after cluster initialization.
1102 HPATH = "cluster-init"
1103 HTYPE = constants.HTYPE_CLUSTER
1105 def BuildHooksEnv(self):
1109 env = {"OP_TARGET": self.cfg.GetClusterName()}
1110 mn = self.cfg.GetMasterNode()
1111 return env, [], [mn]
1113 def Exec(self, feedback_fn):
1120 class LUClusterDestroy(LogicalUnit):
1121 """Logical unit for destroying the cluster.
1124 HPATH = "cluster-destroy"
1125 HTYPE = constants.HTYPE_CLUSTER
1127 def BuildHooksEnv(self):
1131 env = {"OP_TARGET": self.cfg.GetClusterName()}
1134 def CheckPrereq(self):
1135 """Check prerequisites.
1137 This checks whether the cluster is empty.
1139 Any errors are signaled by raising errors.OpPrereqError.
1142 master = self.cfg.GetMasterNode()
1144 nodelist = self.cfg.GetNodeList()
1145 if len(nodelist) != 1 or nodelist[0] != master:
1146 raise errors.OpPrereqError("There are still %d node(s) in"
1147 " this cluster." % (len(nodelist) - 1),
1149 instancelist = self.cfg.GetInstanceList()
1151 raise errors.OpPrereqError("There are still %d instance(s) in"
1152 " this cluster." % len(instancelist),
1155 def Exec(self, feedback_fn):
1156 """Destroys the cluster.
1159 master = self.cfg.GetMasterNode()
1161 # Run post hooks on master node before it's removed
1162 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1164 hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1166 # pylint: disable-msg=W0702
1167 self.LogWarning("Errors occurred running hooks on %s" % master)
1169 result = self.rpc.call_node_stop_master(master, False)
1170 result.Raise("Could not disable the master role")
1175 def _VerifyCertificate(filename):
1176 """Verifies a certificate for LUClusterVerify.
1178 @type filename: string
1179 @param filename: Path to PEM file
1183 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1184 utils.ReadFile(filename))
1185 except Exception, err: # pylint: disable-msg=W0703
1186 return (LUClusterVerify.ETYPE_ERROR,
1187 "Failed to load X509 certificate %s: %s" % (filename, err))
1190 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1191 constants.SSL_CERT_EXPIRATION_ERROR)
1194 fnamemsg = "While verifying %s: %s" % (filename, msg)
1199 return (None, fnamemsg)
1200 elif errcode == utils.CERT_WARNING:
1201 return (LUClusterVerify.ETYPE_WARNING, fnamemsg)
1202 elif errcode == utils.CERT_ERROR:
1203 return (LUClusterVerify.ETYPE_ERROR, fnamemsg)
1205 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1208 class LUClusterVerify(LogicalUnit):
1209 """Verifies the cluster status.
1212 HPATH = "cluster-verify"
1213 HTYPE = constants.HTYPE_CLUSTER
1216 TCLUSTER = "cluster"
1218 TINSTANCE = "instance"
1220 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1221 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1222 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1223 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1224 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1225 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1226 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1227 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1228 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1229 ENODEDRBD = (TNODE, "ENODEDRBD")
1230 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1231 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1232 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1233 ENODEHV = (TNODE, "ENODEHV")
1234 ENODELVM = (TNODE, "ENODELVM")
1235 ENODEN1 = (TNODE, "ENODEN1")
1236 ENODENET = (TNODE, "ENODENET")
1237 ENODEOS = (TNODE, "ENODEOS")
1238 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1239 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1240 ENODERPC = (TNODE, "ENODERPC")
1241 ENODESSH = (TNODE, "ENODESSH")
1242 ENODEVERSION = (TNODE, "ENODEVERSION")
1243 ENODESETUP = (TNODE, "ENODESETUP")
1244 ENODETIME = (TNODE, "ENODETIME")
1245 ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1247 ETYPE_FIELD = "code"
1248 ETYPE_ERROR = "ERROR"
1249 ETYPE_WARNING = "WARNING"
1251 _HOOKS_INDENT_RE = re.compile("^", re.M)
1253 class NodeImage(object):
1254 """A class representing the logical and physical status of a node.
1257 @ivar name: the node name to which this object refers
1258 @ivar volumes: a structure as returned from
1259 L{ganeti.backend.GetVolumeList} (runtime)
1260 @ivar instances: a list of running instances (runtime)
1261 @ivar pinst: list of configured primary instances (config)
1262 @ivar sinst: list of configured secondary instances (config)
1263 @ivar sbp: dictionary of {primary-node: list of instances} for all
1264 instances for which this node is secondary (config)
1265 @ivar mfree: free memory, as reported by hypervisor (runtime)
1266 @ivar dfree: free disk, as reported by the node (runtime)
1267 @ivar offline: the offline status (config)
1268 @type rpc_fail: boolean
1269 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1270 not whether the individual keys were correct) (runtime)
1271 @type lvm_fail: boolean
1272 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1273 @type hyp_fail: boolean
1274 @ivar hyp_fail: whether the RPC call didn't return the instance list
1275 @type ghost: boolean
1276 @ivar ghost: whether this is a known node or not (config)
1277 @type os_fail: boolean
1278 @ivar os_fail: whether the RPC call didn't return valid OS data
1280 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1281 @type vm_capable: boolean
1282 @ivar vm_capable: whether the node can host instances
1285 def __init__(self, offline=False, name=None, vm_capable=True):
1294 self.offline = offline
1295 self.vm_capable = vm_capable
1296 self.rpc_fail = False
1297 self.lvm_fail = False
1298 self.hyp_fail = False
1300 self.os_fail = False
1303 def ExpandNames(self):
1304 self.needed_locks = {
1305 locking.LEVEL_NODE: locking.ALL_SET,
1306 locking.LEVEL_INSTANCE: locking.ALL_SET,
1308 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1310 def _Error(self, ecode, item, msg, *args, **kwargs):
1311 """Format an error message.
1313 Based on the opcode's error_codes parameter, either format a
1314 parseable error code, or a simpler error string.
1316 This must be called only from Exec and functions called from Exec.
1319 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1321 # first complete the msg
1324 # then format the whole message
1325 if self.op.error_codes:
1326 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1332 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1333 # and finally report it via the feedback_fn
1334 self._feedback_fn(" - %s" % msg)
1336 def _ErrorIf(self, cond, *args, **kwargs):
1337 """Log an error message if the passed condition is True.
1340 cond = bool(cond) or self.op.debug_simulate_errors
1342 self._Error(*args, **kwargs)
1343 # do not mark the operation as failed for WARN cases only
1344 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1345 self.bad = self.bad or cond
1347 def _VerifyNode(self, ninfo, nresult):
1348 """Perform some basic validation on data returned from a node.
1350 - check the result data structure is well formed and has all the
1352 - check ganeti version
1354 @type ninfo: L{objects.Node}
1355 @param ninfo: the node to check
1356 @param nresult: the results from the node
1358 @return: whether overall this call was successful (and we can expect
1359 reasonable values in the respose)
1363 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1365 # main result, nresult should be a non-empty dict
1366 test = not nresult or not isinstance(nresult, dict)
1367 _ErrorIf(test, self.ENODERPC, node,
1368 "unable to verify node: no data returned")
1372 # compares ganeti version
1373 local_version = constants.PROTOCOL_VERSION
1374 remote_version = nresult.get("version", None)
1375 test = not (remote_version and
1376 isinstance(remote_version, (list, tuple)) and
1377 len(remote_version) == 2)
1378 _ErrorIf(test, self.ENODERPC, node,
1379 "connection to node returned invalid data")
1383 test = local_version != remote_version[0]
1384 _ErrorIf(test, self.ENODEVERSION, node,
1385 "incompatible protocol versions: master %s,"
1386 " node %s", local_version, remote_version[0])
1390 # node seems compatible, we can actually try to look into its results
1392 # full package version
1393 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1394 self.ENODEVERSION, node,
1395 "software version mismatch: master %s, node %s",
1396 constants.RELEASE_VERSION, remote_version[1],
1397 code=self.ETYPE_WARNING)
1399 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1400 if ninfo.vm_capable and isinstance(hyp_result, dict):
1401 for hv_name, hv_result in hyp_result.iteritems():
1402 test = hv_result is not None
1403 _ErrorIf(test, self.ENODEHV, node,
1404 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1406 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1407 if ninfo.vm_capable and isinstance(hvp_result, list):
1408 for item, hv_name, hv_result in hvp_result:
1409 _ErrorIf(True, self.ENODEHV, node,
1410 "hypervisor %s parameter verify failure (source %s): %s",
1411 hv_name, item, hv_result)
1413 test = nresult.get(constants.NV_NODESETUP,
1414 ["Missing NODESETUP results"])
1415 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1420 def _VerifyNodeTime(self, ninfo, nresult,
1421 nvinfo_starttime, nvinfo_endtime):
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
1427 @param nvinfo_starttime: the start time of the RPC call
1428 @param nvinfo_endtime: the end time of the RPC call
1432 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1434 ntime = nresult.get(constants.NV_TIME, None)
1436 ntime_merged = utils.MergeTime(ntime)
1437 except (ValueError, TypeError):
1438 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1441 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1442 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1443 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1444 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1448 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1449 "Node time diverges by at least %s from master node time",
1452 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1453 """Check the node time.
1455 @type ninfo: L{objects.Node}
1456 @param ninfo: the node to check
1457 @param nresult: the remote results for the node
1458 @param vg_name: the configured VG name
1465 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1467 # checks vg existence and size > 20G
1468 vglist = nresult.get(constants.NV_VGLIST, None)
1470 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1472 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1473 constants.MIN_VG_SIZE)
1474 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1477 pvlist = nresult.get(constants.NV_PVLIST, None)
1478 test = pvlist is None
1479 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1481 # check that ':' is not present in PV names, since it's a
1482 # special character for lvcreate (denotes the range of PEs to
1484 for _, pvname, owner_vg in pvlist:
1485 test = ":" in pvname
1486 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1487 " '%s' of VG '%s'", pvname, owner_vg)
1489 def _VerifyNodeNetwork(self, ninfo, nresult):
1490 """Check the node time.
1492 @type ninfo: L{objects.Node}
1493 @param ninfo: the node to check
1494 @param nresult: the remote results for the node
1498 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1500 test = constants.NV_NODELIST not in nresult
1501 _ErrorIf(test, self.ENODESSH, node,
1502 "node hasn't returned node ssh connectivity data")
1504 if nresult[constants.NV_NODELIST]:
1505 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1506 _ErrorIf(True, self.ENODESSH, node,
1507 "ssh communication with node '%s': %s", a_node, a_msg)
1509 test = constants.NV_NODENETTEST not in nresult
1510 _ErrorIf(test, self.ENODENET, node,
1511 "node hasn't returned node tcp connectivity data")
1513 if nresult[constants.NV_NODENETTEST]:
1514 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1516 _ErrorIf(True, self.ENODENET, node,
1517 "tcp communication with node '%s': %s",
1518 anode, nresult[constants.NV_NODENETTEST][anode])
1520 test = constants.NV_MASTERIP not in nresult
1521 _ErrorIf(test, self.ENODENET, node,
1522 "node hasn't returned node master IP reachability data")
1524 if not nresult[constants.NV_MASTERIP]:
1525 if node == self.master_node:
1526 msg = "the master node cannot reach the master IP (not configured?)"
1528 msg = "cannot reach the master IP"
1529 _ErrorIf(True, self.ENODENET, node, msg)
1531 def _VerifyInstance(self, instance, instanceconfig, node_image,
1533 """Verify an instance.
1535 This function checks to see if the required block devices are
1536 available on the instance's node.
1539 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1540 node_current = instanceconfig.primary_node
1542 node_vol_should = {}
1543 instanceconfig.MapLVsByNode(node_vol_should)
1545 for node in node_vol_should:
1546 n_img = node_image[node]
1547 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1548 # ignore missing volumes on offline or broken nodes
1550 for volume in node_vol_should[node]:
1551 test = volume not in n_img.volumes
1552 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1553 "volume %s missing on node %s", volume, node)
1555 if instanceconfig.admin_up:
1556 pri_img = node_image[node_current]
1557 test = instance not in pri_img.instances and not pri_img.offline
1558 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1559 "instance not running on its primary node %s",
1562 for node, n_img in node_image.items():
1563 if node != node_current:
1564 test = instance in n_img.instances
1565 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1566 "instance should not run on node %s", node)
1568 diskdata = [(nname, success, status, idx)
1569 for (nname, disks) in diskstatus.items()
1570 for idx, (success, status) in enumerate(disks)]
1572 for nname, success, bdev_status, idx in diskdata:
1573 # the 'ghost node' construction in Exec() ensures that we have a
1575 snode = node_image[nname]
1576 bad_snode = snode.ghost or snode.offline
1577 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
1578 self.EINSTANCEFAULTYDISK, instance,
1579 "couldn't retrieve status for disk/%s on %s: %s",
1580 idx, nname, bdev_status)
1581 _ErrorIf((instanceconfig.admin_up and success and
1582 bdev_status.ldisk_status == constants.LDS_FAULTY),
1583 self.EINSTANCEFAULTYDISK, instance,
1584 "disk/%s on %s is faulty", idx, nname)
1586 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1587 """Verify if there are any unknown volumes in the cluster.
1589 The .os, .swap and backup volumes are ignored. All other volumes are
1590 reported as unknown.
1592 @type reserved: L{ganeti.utils.FieldSet}
1593 @param reserved: a FieldSet of reserved volume names
1596 for node, n_img in node_image.items():
1597 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1598 # skip non-healthy nodes
1600 for volume in n_img.volumes:
1601 test = ((node not in node_vol_should or
1602 volume not in node_vol_should[node]) and
1603 not reserved.Matches(volume))
1604 self._ErrorIf(test, self.ENODEORPHANLV, node,
1605 "volume %s is unknown", volume)
1607 def _VerifyOrphanInstances(self, instancelist, node_image):
1608 """Verify the list of running instances.
1610 This checks what instances are running but unknown to the cluster.
1613 for node, n_img in node_image.items():
1614 for o_inst in n_img.instances:
1615 test = o_inst not in instancelist
1616 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1617 "instance %s on node %s should not exist", o_inst, node)
1619 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1620 """Verify N+1 Memory Resilience.
1622 Check that if one single node dies we can still start all the
1623 instances it was primary for.
1626 cluster_info = self.cfg.GetClusterInfo()
1627 for node, n_img in node_image.items():
1628 # This code checks that every node which is now listed as
1629 # secondary has enough memory to host all instances it is
1630 # supposed to should a single other node in the cluster fail.
1631 # FIXME: not ready for failover to an arbitrary node
1632 # FIXME: does not support file-backed instances
1633 # WARNING: we currently take into account down instances as well
1634 # as up ones, considering that even if they're down someone
1635 # might want to start them even in the event of a node failure.
1637 # we're skipping offline nodes from the N+1 warning, since
1638 # most likely we don't have good memory infromation from them;
1639 # we already list instances living on such nodes, and that's
1642 for prinode, instances in n_img.sbp.items():
1644 for instance in instances:
1645 bep = cluster_info.FillBE(instance_cfg[instance])
1646 if bep[constants.BE_AUTO_BALANCE]:
1647 needed_mem += bep[constants.BE_MEMORY]
1648 test = n_img.mfree < needed_mem
1649 self._ErrorIf(test, self.ENODEN1, node,
1650 "not enough memory to accomodate instance failovers"
1651 " should node %s fail", prinode)
1653 def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1655 """Verifies and computes the node required file checksums.
1657 @type ninfo: L{objects.Node}
1658 @param ninfo: the node to check
1659 @param nresult: the remote results for the node
1660 @param file_list: required list of files
1661 @param local_cksum: dictionary of local files and their checksums
1662 @param master_files: list of files that only masters should have
1666 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1668 remote_cksum = nresult.get(constants.NV_FILELIST, None)
1669 test = not isinstance(remote_cksum, dict)
1670 _ErrorIf(test, self.ENODEFILECHECK, node,
1671 "node hasn't returned file checksum data")
1675 for file_name in file_list:
1676 node_is_mc = ninfo.master_candidate
1677 must_have = (file_name not in master_files) or node_is_mc
1679 test1 = file_name not in remote_cksum
1681 test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1683 test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1684 _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1685 "file '%s' missing", file_name)
1686 _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1687 "file '%s' has wrong checksum", file_name)
1688 # not candidate and this is not a must-have file
1689 _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1690 "file '%s' should not exist on non master"
1691 " candidates (and the file is outdated)", file_name)
1692 # all good, except non-master/non-must have combination
1693 _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1694 "file '%s' should not exist"
1695 " on non master candidates", file_name)
1697 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1699 """Verifies and the node DRBD status.
1701 @type ninfo: L{objects.Node}
1702 @param ninfo: the node to check
1703 @param nresult: the remote results for the node
1704 @param instanceinfo: the dict of instances
1705 @param drbd_helper: the configured DRBD usermode helper
1706 @param drbd_map: the DRBD map as returned by
1707 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1711 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1714 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1715 test = (helper_result == None)
1716 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1717 "no drbd usermode helper returned")
1719 status, payload = helper_result
1721 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1722 "drbd usermode helper check unsuccessful: %s", payload)
1723 test = status and (payload != drbd_helper)
1724 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1725 "wrong drbd usermode helper: %s", payload)
1727 # compute the DRBD minors
1729 for minor, instance in drbd_map[node].items():
1730 test = instance not in instanceinfo
1731 _ErrorIf(test, self.ECLUSTERCFG, None,
1732 "ghost instance '%s' in temporary DRBD map", instance)
1733 # ghost instance should not be running, but otherwise we
1734 # don't give double warnings (both ghost instance and
1735 # unallocated minor in use)
1737 node_drbd[minor] = (instance, False)
1739 instance = instanceinfo[instance]
1740 node_drbd[minor] = (instance.name, instance.admin_up)
1742 # and now check them
1743 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1744 test = not isinstance(used_minors, (tuple, list))
1745 _ErrorIf(test, self.ENODEDRBD, node,
1746 "cannot parse drbd status file: %s", str(used_minors))
1748 # we cannot check drbd status
1751 for minor, (iname, must_exist) in node_drbd.items():
1752 test = minor not in used_minors and must_exist
1753 _ErrorIf(test, self.ENODEDRBD, node,
1754 "drbd minor %d of instance %s is not active", minor, iname)
1755 for minor in used_minors:
1756 test = minor not in node_drbd
1757 _ErrorIf(test, self.ENODEDRBD, node,
1758 "unallocated drbd minor %d is in use", minor)
1760 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1761 """Builds the node OS structures.
1763 @type ninfo: L{objects.Node}
1764 @param ninfo: the node to check
1765 @param nresult: the remote results for the node
1766 @param nimg: the node image object
1770 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1772 remote_os = nresult.get(constants.NV_OSLIST, None)
1773 test = (not isinstance(remote_os, list) or
1774 not compat.all(isinstance(v, list) and len(v) == 7
1775 for v in remote_os))
1777 _ErrorIf(test, self.ENODEOS, node,
1778 "node hasn't returned valid OS data")
1787 for (name, os_path, status, diagnose,
1788 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1790 if name not in os_dict:
1793 # parameters is a list of lists instead of list of tuples due to
1794 # JSON lacking a real tuple type, fix it:
1795 parameters = [tuple(v) for v in parameters]
1796 os_dict[name].append((os_path, status, diagnose,
1797 set(variants), set(parameters), set(api_ver)))
1799 nimg.oslist = os_dict
1801 def _VerifyNodeOS(self, ninfo, nimg, base):
1802 """Verifies the node OS list.
1804 @type ninfo: L{objects.Node}
1805 @param ninfo: the node to check
1806 @param nimg: the node image object
1807 @param base: the 'template' node we match against (e.g. from the master)
1811 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1813 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1815 for os_name, os_data in nimg.oslist.items():
1816 assert os_data, "Empty OS status for OS %s?!" % os_name
1817 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1818 _ErrorIf(not f_status, self.ENODEOS, node,
1819 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1820 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1821 "OS '%s' has multiple entries (first one shadows the rest): %s",
1822 os_name, utils.CommaJoin([v[0] for v in os_data]))
1823 # this will catched in backend too
1824 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1825 and not f_var, self.ENODEOS, node,
1826 "OS %s with API at least %d does not declare any variant",
1827 os_name, constants.OS_API_V15)
1828 # comparisons with the 'base' image
1829 test = os_name not in base.oslist
1830 _ErrorIf(test, self.ENODEOS, node,
1831 "Extra OS %s not present on reference node (%s)",
1835 assert base.oslist[os_name], "Base node has empty OS status?"
1836 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1838 # base OS is invalid, skipping
1840 for kind, a, b in [("API version", f_api, b_api),
1841 ("variants list", f_var, b_var),
1842 ("parameters", f_param, b_param)]:
1843 _ErrorIf(a != b, self.ENODEOS, node,
1844 "OS %s %s differs from reference node %s: %s vs. %s",
1845 kind, os_name, base.name,
1846 utils.CommaJoin(a), utils.CommaJoin(b))
1848 # check any missing OSes
1849 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1850 _ErrorIf(missing, self.ENODEOS, node,
1851 "OSes present on reference node %s but missing on this node: %s",
1852 base.name, utils.CommaJoin(missing))
1854 def _VerifyOob(self, ninfo, nresult):
1855 """Verifies out of band functionality of a node.
1857 @type ninfo: L{objects.Node}
1858 @param ninfo: the node to check
1859 @param nresult: the remote results for the node
1863 # We just have to verify the paths on master and/or master candidates
1864 # as the oob helper is invoked on the master
1865 if ((ninfo.master_candidate or ninfo.master_capable) and
1866 constants.NV_OOB_PATHS in nresult):
1867 for path_result in nresult[constants.NV_OOB_PATHS]:
1868 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1870 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1871 """Verifies and updates the node volume data.
1873 This function will update a L{NodeImage}'s internal structures
1874 with data from the remote call.
1876 @type ninfo: L{objects.Node}
1877 @param ninfo: the node to check
1878 @param nresult: the remote results for the node
1879 @param nimg: the node image object
1880 @param vg_name: the configured VG name
1884 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1886 nimg.lvm_fail = True
1887 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1890 elif isinstance(lvdata, basestring):
1891 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1892 utils.SafeEncode(lvdata))
1893 elif not isinstance(lvdata, dict):
1894 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1896 nimg.volumes = lvdata
1897 nimg.lvm_fail = False
1899 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1900 """Verifies and updates the node instance list.
1902 If the listing was successful, then updates this node's instance
1903 list. Otherwise, it marks the RPC call as failed for the instance
1906 @type ninfo: L{objects.Node}
1907 @param ninfo: the node to check
1908 @param nresult: the remote results for the node
1909 @param nimg: the node image object
1912 idata = nresult.get(constants.NV_INSTANCELIST, None)
1913 test = not isinstance(idata, list)
1914 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1915 " (instancelist): %s", utils.SafeEncode(str(idata)))
1917 nimg.hyp_fail = True
1919 nimg.instances = idata
1921 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1922 """Verifies and computes a node information map
1924 @type ninfo: L{objects.Node}
1925 @param ninfo: the node to check
1926 @param nresult: the remote results for the node
1927 @param nimg: the node image object
1928 @param vg_name: the configured VG name
1932 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1934 # try to read free memory (from the hypervisor)
1935 hv_info = nresult.get(constants.NV_HVINFO, None)
1936 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1937 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1940 nimg.mfree = int(hv_info["memory_free"])
1941 except (ValueError, TypeError):
1942 _ErrorIf(True, self.ENODERPC, node,
1943 "node returned invalid nodeinfo, check hypervisor")
1945 # FIXME: devise a free space model for file based instances as well
1946 if vg_name is not None:
1947 test = (constants.NV_VGLIST not in nresult or
1948 vg_name not in nresult[constants.NV_VGLIST])
1949 _ErrorIf(test, self.ENODELVM, node,
1950 "node didn't return data for the volume group '%s'"
1951 " - it is either missing or broken", vg_name)
1954 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1955 except (ValueError, TypeError):
1956 _ErrorIf(True, self.ENODERPC, node,
1957 "node returned invalid LVM info, check LVM status")
1959 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1960 """Gets per-disk status information for all instances.
1962 @type nodelist: list of strings
1963 @param nodelist: Node names
1964 @type node_image: dict of (name, L{objects.Node})
1965 @param node_image: Node objects
1966 @type instanceinfo: dict of (name, L{objects.Instance})
1967 @param instanceinfo: Instance objects
1968 @rtype: {instance: {node: [(succes, payload)]}}
1969 @return: a dictionary of per-instance dictionaries with nodes as
1970 keys and disk information as values; the disk information is a
1971 list of tuples (success, payload)
1974 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1977 node_disks_devonly = {}
1978 diskless_instances = set()
1979 diskless = constants.DT_DISKLESS
1981 for nname in nodelist:
1982 node_instances = list(itertools.chain(node_image[nname].pinst,
1983 node_image[nname].sinst))
1984 diskless_instances.update(inst for inst in node_instances
1985 if instanceinfo[inst].disk_template == diskless)
1986 disks = [(inst, disk)
1987 for inst in node_instances
1988 for disk in instanceinfo[inst].disks]
1991 # No need to collect data
1994 node_disks[nname] = disks
1996 # Creating copies as SetDiskID below will modify the objects and that can
1997 # lead to incorrect data returned from nodes
1998 devonly = [dev.Copy() for (_, dev) in disks]
2001 self.cfg.SetDiskID(dev, nname)
2003 node_disks_devonly[nname] = devonly
2005 assert len(node_disks) == len(node_disks_devonly)
2007 # Collect data from all nodes with disks
2008 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2011 assert len(result) == len(node_disks)
2015 for (nname, nres) in result.items():
2016 disks = node_disks[nname]
2019 # No data from this node
2020 data = len(disks) * [(False, "node offline")]
2023 _ErrorIf(msg, self.ENODERPC, nname,
2024 "while getting disk information: %s", msg)
2026 # No data from this node
2027 data = len(disks) * [(False, msg)]
2030 for idx, i in enumerate(nres.payload):
2031 if isinstance(i, (tuple, list)) and len(i) == 2:
2034 logging.warning("Invalid result from node %s, entry %d: %s",
2036 data.append((False, "Invalid result from the remote node"))
2038 for ((inst, _), status) in zip(disks, data):
2039 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2041 # Add empty entries for diskless instances.
2042 for inst in diskless_instances:
2043 assert inst not in instdisk
2046 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2047 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2048 compat.all(isinstance(s, (tuple, list)) and
2049 len(s) == 2 for s in statuses)
2050 for inst, nnames in instdisk.items()
2051 for nname, statuses in nnames.items())
2052 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2056 def _VerifyHVP(self, hvp_data):
2057 """Verifies locally the syntax of the hypervisor parameters.
2060 for item, hv_name, hv_params in hvp_data:
2061 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2064 hv_class = hypervisor.GetHypervisor(hv_name)
2065 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2066 hv_class.CheckParameterSyntax(hv_params)
2067 except errors.GenericError, err:
2068 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2071 def BuildHooksEnv(self):
2074 Cluster-Verify hooks just ran in the post phase and their failure makes
2075 the output be logged in the verify output and the verification to fail.
2078 all_nodes = self.cfg.GetNodeList()
2080 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2082 for node in self.cfg.GetAllNodesInfo().values():
2083 env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2085 return env, [], all_nodes
2087 def Exec(self, feedback_fn):
2088 """Verify integrity of cluster, performing various test on nodes.
2091 # This method has too many local variables. pylint: disable-msg=R0914
2093 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2094 verbose = self.op.verbose
2095 self._feedback_fn = feedback_fn
2096 feedback_fn("* Verifying global settings")
2097 for msg in self.cfg.VerifyConfig():
2098 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2100 # Check the cluster certificates
2101 for cert_filename in constants.ALL_CERT_FILES:
2102 (errcode, msg) = _VerifyCertificate(cert_filename)
2103 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2105 vg_name = self.cfg.GetVGName()
2106 drbd_helper = self.cfg.GetDRBDHelper()
2107 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2108 cluster = self.cfg.GetClusterInfo()
2109 nodelist = utils.NiceSort(self.cfg.GetNodeList())
2110 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2111 nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2112 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2113 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2114 for iname in instancelist)
2115 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2116 i_non_redundant = [] # Non redundant instances
2117 i_non_a_balanced = [] # Non auto-balanced instances
2118 n_offline = 0 # Count of offline nodes
2119 n_drained = 0 # Count of nodes being drained
2120 node_vol_should = {}
2122 # FIXME: verify OS list
2123 # do local checksums
2124 master_files = [constants.CLUSTER_CONF_FILE]
2125 master_node = self.master_node = self.cfg.GetMasterNode()
2126 master_ip = self.cfg.GetMasterIP()
2128 file_names = ssconf.SimpleStore().GetFileList()
2129 file_names.extend(constants.ALL_CERT_FILES)
2130 file_names.extend(master_files)
2131 if cluster.modify_etc_hosts:
2132 file_names.append(constants.ETC_HOSTS)
2134 local_checksums = utils.FingerprintFiles(file_names)
2136 # Compute the set of hypervisor parameters
2138 for hv_name in hypervisors:
2139 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2140 for os_name, os_hvp in cluster.os_hvp.items():
2141 for hv_name, hv_params in os_hvp.items():
2144 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2145 hvp_data.append(("os %s" % os_name, hv_name, full_params))
2146 # TODO: collapse identical parameter values in a single one
2147 for instance in instanceinfo.values():
2148 if not instance.hvparams:
2150 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2151 cluster.FillHV(instance)))
2152 # and verify them locally
2153 self._VerifyHVP(hvp_data)
2155 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2156 node_verify_param = {
2157 constants.NV_FILELIST: file_names,
2158 constants.NV_NODELIST: [node.name for node in nodeinfo
2159 if not node.offline],
2160 constants.NV_HYPERVISOR: hypervisors,
2161 constants.NV_HVPARAMS: hvp_data,
2162 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2163 node.secondary_ip) for node in nodeinfo
2164 if not node.offline],
2165 constants.NV_INSTANCELIST: hypervisors,
2166 constants.NV_VERSION: None,
2167 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2168 constants.NV_NODESETUP: None,
2169 constants.NV_TIME: None,
2170 constants.NV_MASTERIP: (master_node, master_ip),
2171 constants.NV_OSLIST: None,
2172 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2175 if vg_name is not None:
2176 node_verify_param[constants.NV_VGLIST] = None
2177 node_verify_param[constants.NV_LVLIST] = vg_name
2178 node_verify_param[constants.NV_PVLIST] = [vg_name]
2179 node_verify_param[constants.NV_DRBDLIST] = None
2182 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2184 # Build our expected cluster state
2185 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2187 vm_capable=node.vm_capable))
2188 for node in nodeinfo)
2192 for node in nodeinfo:
2193 path = _SupportsOob(self.cfg, node)
2194 if path and path not in oob_paths:
2195 oob_paths.append(path)
2198 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2200 for instance in instancelist:
2201 inst_config = instanceinfo[instance]
2203 for nname in inst_config.all_nodes:
2204 if nname not in node_image:
2206 gnode = self.NodeImage(name=nname)
2208 node_image[nname] = gnode
2210 inst_config.MapLVsByNode(node_vol_should)
2212 pnode = inst_config.primary_node
2213 node_image[pnode].pinst.append(instance)
2215 for snode in inst_config.secondary_nodes:
2216 nimg = node_image[snode]
2217 nimg.sinst.append(instance)
2218 if pnode not in nimg.sbp:
2219 nimg.sbp[pnode] = []
2220 nimg.sbp[pnode].append(instance)
2222 # At this point, we have the in-memory data structures complete,
2223 # except for the runtime information, which we'll gather next
2225 # Due to the way our RPC system works, exact response times cannot be
2226 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2227 # time before and after executing the request, we can at least have a time
2229 nvinfo_starttime = time.time()
2230 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2231 self.cfg.GetClusterName())
2232 nvinfo_endtime = time.time()
2234 all_drbd_map = self.cfg.ComputeDRBDMap()
2236 feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2237 instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2239 feedback_fn("* Verifying node status")
2243 for node_i in nodeinfo:
2245 nimg = node_image[node]
2249 feedback_fn("* Skipping offline node %s" % (node,))
2253 if node == master_node:
2255 elif node_i.master_candidate:
2256 ntype = "master candidate"
2257 elif node_i.drained:
2263 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2265 msg = all_nvinfo[node].fail_msg
2266 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2268 nimg.rpc_fail = True
2271 nresult = all_nvinfo[node].payload
2273 nimg.call_ok = self._VerifyNode(node_i, nresult)
2274 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2275 self._VerifyNodeNetwork(node_i, nresult)
2276 self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2279 self._VerifyOob(node_i, nresult)
2282 self._VerifyNodeLVM(node_i, nresult, vg_name)
2283 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2286 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2287 self._UpdateNodeInstances(node_i, nresult, nimg)
2288 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2289 self._UpdateNodeOS(node_i, nresult, nimg)
2290 if not nimg.os_fail:
2291 if refos_img is None:
2293 self._VerifyNodeOS(node_i, nimg, refos_img)
2295 feedback_fn("* Verifying instance status")
2296 for instance in instancelist:
2298 feedback_fn("* Verifying instance %s" % instance)
2299 inst_config = instanceinfo[instance]
2300 self._VerifyInstance(instance, inst_config, node_image,
2302 inst_nodes_offline = []
2304 pnode = inst_config.primary_node
2305 pnode_img = node_image[pnode]
2306 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2307 self.ENODERPC, pnode, "instance %s, connection to"
2308 " primary node failed", instance)
2310 _ErrorIf(pnode_img.offline, self.EINSTANCEBADNODE, instance,
2311 "instance lives on offline node %s", inst_config.primary_node)
2313 # If the instance is non-redundant we cannot survive losing its primary
2314 # node, so we are not N+1 compliant. On the other hand we have no disk
2315 # templates with more than one secondary so that situation is not well
2317 # FIXME: does not support file-backed instances
2318 if not inst_config.secondary_nodes:
2319 i_non_redundant.append(instance)
2321 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2322 instance, "instance has multiple secondary nodes: %s",
2323 utils.CommaJoin(inst_config.secondary_nodes),
2324 code=self.ETYPE_WARNING)
2326 if inst_config.disk_template in constants.DTS_NET_MIRROR:
2327 pnode = inst_config.primary_node
2328 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2329 instance_groups = {}
2331 for node in instance_nodes:
2332 instance_groups.setdefault(nodeinfo_byname[node].group,
2336 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2337 # Sort so that we always list the primary node first.
2338 for group, nodes in sorted(instance_groups.items(),
2339 key=lambda (_, nodes): pnode in nodes,
2342 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2343 instance, "instance has primary and secondary nodes in"
2344 " different groups: %s", utils.CommaJoin(pretty_list),
2345 code=self.ETYPE_WARNING)
2347 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2348 i_non_a_balanced.append(instance)
2350 for snode in inst_config.secondary_nodes:
2351 s_img = node_image[snode]
2352 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2353 "instance %s, connection to secondary node failed", instance)
2356 inst_nodes_offline.append(snode)
2358 # warn that the instance lives on offline nodes
2359 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2360 "instance has offline secondary node(s) %s",
2361 utils.CommaJoin(inst_nodes_offline))
2362 # ... or ghost/non-vm_capable nodes
2363 for node in inst_config.all_nodes:
2364 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2365 "instance lives on ghost node %s", node)
2366 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2367 instance, "instance lives on non-vm_capable node %s", node)
2369 feedback_fn("* Verifying orphan volumes")
2370 reserved = utils.FieldSet(*cluster.reserved_lvs)
2371 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2373 feedback_fn("* Verifying orphan instances")
2374 self._VerifyOrphanInstances(instancelist, node_image)
2376 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2377 feedback_fn("* Verifying N+1 Memory redundancy")
2378 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2380 feedback_fn("* Other Notes")
2382 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2383 % len(i_non_redundant))
2385 if i_non_a_balanced:
2386 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2387 % len(i_non_a_balanced))
2390 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2393 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2397 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2398 """Analyze the post-hooks' result
2400 This method analyses the hook result, handles it, and sends some
2401 nicely-formatted feedback back to the user.
2403 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2404 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2405 @param hooks_results: the results of the multi-node hooks rpc call
2406 @param feedback_fn: function used send feedback back to the caller
2407 @param lu_result: previous Exec result
2408 @return: the new Exec result, based on the previous result
2412 # We only really run POST phase hooks, and are only interested in
2414 if phase == constants.HOOKS_PHASE_POST:
2415 # Used to change hooks' output to proper indentation
2416 feedback_fn("* Hooks Results")
2417 assert hooks_results, "invalid result from hooks"
2419 for node_name in hooks_results:
2420 res = hooks_results[node_name]
2422 test = msg and not res.offline
2423 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2424 "Communication failure in hooks execution: %s", msg)
2425 if res.offline or msg:
2426 # No need to investigate payload if node is offline or gave an error.
2427 # override manually lu_result here as _ErrorIf only
2428 # overrides self.bad
2431 for script, hkr, output in res.payload:
2432 test = hkr == constants.HKR_FAIL
2433 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2434 "Script %s failed, output:", script)
2436 output = self._HOOKS_INDENT_RE.sub(' ', output)
2437 feedback_fn("%s" % output)
2443 class LUClusterVerifyDisks(NoHooksLU):
2444 """Verifies the cluster disks status.
2449 def ExpandNames(self):
2450 self.needed_locks = {
2451 locking.LEVEL_NODE: locking.ALL_SET,
2452 locking.LEVEL_INSTANCE: locking.ALL_SET,
2454 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2456 def Exec(self, feedback_fn):
2457 """Verify integrity of cluster disks.
2459 @rtype: tuple of three items
2460 @return: a tuple of (dict of node-to-node_error, list of instances
2461 which need activate-disks, dict of instance: (node, volume) for
2465 result = res_nodes, res_instances, res_missing = {}, [], {}
2467 nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2468 instances = self.cfg.GetAllInstancesInfo().values()
2471 for inst in instances:
2473 if not inst.admin_up:
2475 inst.MapLVsByNode(inst_lvs)
2476 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2477 for node, vol_list in inst_lvs.iteritems():
2478 for vol in vol_list:
2479 nv_dict[(node, vol)] = inst
2484 node_lvs = self.rpc.call_lv_list(nodes, [])
2485 for node, node_res in node_lvs.items():
2486 if node_res.offline:
2488 msg = node_res.fail_msg
2490 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2491 res_nodes[node] = msg
2494 lvs = node_res.payload
2495 for lv_name, (_, _, lv_online) in lvs.items():
2496 inst = nv_dict.pop((node, lv_name), None)
2497 if (not lv_online and inst is not None
2498 and inst.name not in res_instances):
2499 res_instances.append(inst.name)
2501 # any leftover items in nv_dict are missing LVs, let's arrange the
2503 for key, inst in nv_dict.iteritems():
2504 if inst.name not in res_missing:
2505 res_missing[inst.name] = []
2506 res_missing[inst.name].append(key)
2511 class LUClusterRepairDiskSizes(NoHooksLU):
2512 """Verifies the cluster disks sizes.
2517 def ExpandNames(self):
2518 if self.op.instances:
2519 self.wanted_names = []
2520 for name in self.op.instances:
2521 full_name = _ExpandInstanceName(self.cfg, name)
2522 self.wanted_names.append(full_name)
2523 self.needed_locks = {
2524 locking.LEVEL_NODE: [],
2525 locking.LEVEL_INSTANCE: self.wanted_names,
2527 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2529 self.wanted_names = None
2530 self.needed_locks = {
2531 locking.LEVEL_NODE: locking.ALL_SET,
2532 locking.LEVEL_INSTANCE: locking.ALL_SET,
2534 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2536 def DeclareLocks(self, level):
2537 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2538 self._LockInstancesNodes(primary_only=True)
2540 def CheckPrereq(self):
2541 """Check prerequisites.
2543 This only checks the optional instance list against the existing names.
2546 if self.wanted_names is None:
2547 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2549 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2550 in self.wanted_names]
2552 def _EnsureChildSizes(self, disk):
2553 """Ensure children of the disk have the needed disk size.
2555 This is valid mainly for DRBD8 and fixes an issue where the
2556 children have smaller disk size.
2558 @param disk: an L{ganeti.objects.Disk} object
2561 if disk.dev_type == constants.LD_DRBD8:
2562 assert disk.children, "Empty children for DRBD8?"
2563 fchild = disk.children[0]
2564 mismatch = fchild.size < disk.size
2566 self.LogInfo("Child disk has size %d, parent %d, fixing",
2567 fchild.size, disk.size)
2568 fchild.size = disk.size
2570 # and we recurse on this child only, not on the metadev
2571 return self._EnsureChildSizes(fchild) or mismatch
2575 def Exec(self, feedback_fn):
2576 """Verify the size of cluster disks.
2579 # TODO: check child disks too
2580 # TODO: check differences in size between primary/secondary nodes
2582 for instance in self.wanted_instances:
2583 pnode = instance.primary_node
2584 if pnode not in per_node_disks:
2585 per_node_disks[pnode] = []
2586 for idx, disk in enumerate(instance.disks):
2587 per_node_disks[pnode].append((instance, idx, disk))
2590 for node, dskl in per_node_disks.items():
2591 newl = [v[2].Copy() for v in dskl]
2593 self.cfg.SetDiskID(dsk, node)
2594 result = self.rpc.call_blockdev_getsize(node, newl)
2596 self.LogWarning("Failure in blockdev_getsize call to node"
2597 " %s, ignoring", node)
2599 if len(result.payload) != len(dskl):
2600 logging.warning("Invalid result from node %s: len(dksl)=%d,"
2601 " result.payload=%s", node, len(dskl), result.payload)
2602 self.LogWarning("Invalid result from node %s, ignoring node results",
2605 for ((instance, idx, disk), size) in zip(dskl, result.payload):
2607 self.LogWarning("Disk %d of instance %s did not return size"
2608 " information, ignoring", idx, instance.name)
2610 if not isinstance(size, (int, long)):
2611 self.LogWarning("Disk %d of instance %s did not return valid"
2612 " size information, ignoring", idx, instance.name)
2615 if size != disk.size:
2616 self.LogInfo("Disk %d of instance %s has mismatched size,"
2617 " correcting: recorded %d, actual %d", idx,
2618 instance.name, disk.size, size)
2620 self.cfg.Update(instance, feedback_fn)
2621 changed.append((instance.name, idx, size))
2622 if self._EnsureChildSizes(disk):
2623 self.cfg.Update(instance, feedback_fn)
2624 changed.append((instance.name, idx, disk.size))
2628 class LUClusterRename(LogicalUnit):
2629 """Rename the cluster.
2632 HPATH = "cluster-rename"
2633 HTYPE = constants.HTYPE_CLUSTER
2635 def BuildHooksEnv(self):
2640 "OP_TARGET": self.cfg.GetClusterName(),
2641 "NEW_NAME": self.op.name,
2643 mn = self.cfg.GetMasterNode()
2644 all_nodes = self.cfg.GetNodeList()
2645 return env, [mn], all_nodes
2647 def CheckPrereq(self):
2648 """Verify that the passed name is a valid one.
2651 hostname = netutils.GetHostname(name=self.op.name,
2652 family=self.cfg.GetPrimaryIPFamily())
2654 new_name = hostname.name
2655 self.ip = new_ip = hostname.ip
2656 old_name = self.cfg.GetClusterName()
2657 old_ip = self.cfg.GetMasterIP()
2658 if new_name == old_name and new_ip == old_ip:
2659 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2660 " cluster has changed",
2662 if new_ip != old_ip:
2663 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2664 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2665 " reachable on the network" %
2666 new_ip, errors.ECODE_NOTUNIQUE)
2668 self.op.name = new_name
2670 def Exec(self, feedback_fn):
2671 """Rename the cluster.
2674 clustername = self.op.name
2677 # shutdown the master IP
2678 master = self.cfg.GetMasterNode()
2679 result = self.rpc.call_node_stop_master(master, False)
2680 result.Raise("Could not disable the master role")
2683 cluster = self.cfg.GetClusterInfo()
2684 cluster.cluster_name = clustername
2685 cluster.master_ip = ip
2686 self.cfg.Update(cluster, feedback_fn)
2688 # update the known hosts file
2689 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2690 node_list = self.cfg.GetOnlineNodeList()
2692 node_list.remove(master)
2695 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2697 result = self.rpc.call_node_start_master(master, False, False)
2698 msg = result.fail_msg
2700 self.LogWarning("Could not re-enable the master role on"
2701 " the master, please restart manually: %s", msg)
2706 class LUClusterSetParams(LogicalUnit):
2707 """Change the parameters of the cluster.
2710 HPATH = "cluster-modify"
2711 HTYPE = constants.HTYPE_CLUSTER
2714 def CheckArguments(self):
2718 if self.op.uid_pool:
2719 uidpool.CheckUidPool(self.op.uid_pool)
2721 if self.op.add_uids:
2722 uidpool.CheckUidPool(self.op.add_uids)
2724 if self.op.remove_uids:
2725 uidpool.CheckUidPool(self.op.remove_uids)
2727 def ExpandNames(self):
2728 # FIXME: in the future maybe other cluster params won't require checking on
2729 # all nodes to be modified.
2730 self.needed_locks = {
2731 locking.LEVEL_NODE: locking.ALL_SET,
2733 self.share_locks[locking.LEVEL_NODE] = 1
2735 def BuildHooksEnv(self):
2740 "OP_TARGET": self.cfg.GetClusterName(),
2741 "NEW_VG_NAME": self.op.vg_name,
2743 mn = self.cfg.GetMasterNode()
2744 return env, [mn], [mn]
2746 def CheckPrereq(self):
2747 """Check prerequisites.
2749 This checks whether the given params don't conflict and
2750 if the given volume group is valid.
2753 if self.op.vg_name is not None and not self.op.vg_name:
2754 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2755 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2756 " instances exist", errors.ECODE_INVAL)
2758 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2759 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2760 raise errors.OpPrereqError("Cannot disable drbd helper while"
2761 " drbd-based instances exist",
2764 node_list = self.acquired_locks[locking.LEVEL_NODE]
2766 # if vg_name not None, checks given volume group on all nodes
2768 vglist = self.rpc.call_vg_list(node_list)
2769 for node in node_list:
2770 msg = vglist[node].fail_msg
2772 # ignoring down node
2773 self.LogWarning("Error while gathering data on node %s"
2774 " (ignoring node): %s", node, msg)
2776 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2778 constants.MIN_VG_SIZE)
2780 raise errors.OpPrereqError("Error on node '%s': %s" %
2781 (node, vgstatus), errors.ECODE_ENVIRON)
2783 if self.op.drbd_helper:
2784 # checks given drbd helper on all nodes
2785 helpers = self.rpc.call_drbd_helper(node_list)
2786 for node in node_list:
2787 ninfo = self.cfg.GetNodeInfo(node)
2789 self.LogInfo("Not checking drbd helper on offline node %s", node)
2791 msg = helpers[node].fail_msg
2793 raise errors.OpPrereqError("Error checking drbd helper on node"
2794 " '%s': %s" % (node, msg),
2795 errors.ECODE_ENVIRON)
2796 node_helper = helpers[node].payload
2797 if node_helper != self.op.drbd_helper:
2798 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2799 (node, node_helper), errors.ECODE_ENVIRON)
2801 self.cluster = cluster = self.cfg.GetClusterInfo()
2802 # validate params changes
2803 if self.op.beparams:
2804 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2805 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2807 if self.op.ndparams:
2808 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2809 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2811 if self.op.nicparams:
2812 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2813 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2814 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2817 # check all instances for consistency
2818 for instance in self.cfg.GetAllInstancesInfo().values():
2819 for nic_idx, nic in enumerate(instance.nics):
2820 params_copy = copy.deepcopy(nic.nicparams)
2821 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2823 # check parameter syntax
2825 objects.NIC.CheckParameterSyntax(params_filled)
2826 except errors.ConfigurationError, err:
2827 nic_errors.append("Instance %s, nic/%d: %s" %
2828 (instance.name, nic_idx, err))
2830 # if we're moving instances to routed, check that they have an ip
2831 target_mode = params_filled[constants.NIC_MODE]
2832 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2833 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2834 (instance.name, nic_idx))
2836 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2837 "\n".join(nic_errors))
2839 # hypervisor list/parameters
2840 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2841 if self.op.hvparams:
2842 for hv_name, hv_dict in self.op.hvparams.items():
2843 if hv_name not in self.new_hvparams:
2844 self.new_hvparams[hv_name] = hv_dict
2846 self.new_hvparams[hv_name].update(hv_dict)
2848 # os hypervisor parameters
2849 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2851 for os_name, hvs in self.op.os_hvp.items():
2852 if os_name not in self.new_os_hvp:
2853 self.new_os_hvp[os_name] = hvs
2855 for hv_name, hv_dict in hvs.items():
2856 if hv_name not in self.new_os_hvp[os_name]:
2857 self.new_os_hvp[os_name][hv_name] = hv_dict
2859 self.new_os_hvp[os_name][hv_name].update(hv_dict)
2862 self.new_osp = objects.FillDict(cluster.osparams, {})
2863 if self.op.osparams:
2864 for os_name, osp in self.op.osparams.items():
2865 if os_name not in self.new_osp:
2866 self.new_osp[os_name] = {}
2868 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2871 if not self.new_osp[os_name]:
2872 # we removed all parameters
2873 del self.new_osp[os_name]
2875 # check the parameter validity (remote check)
2876 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2877 os_name, self.new_osp[os_name])
2879 # changes to the hypervisor list
2880 if self.op.enabled_hypervisors is not None:
2881 self.hv_list = self.op.enabled_hypervisors
2882 for hv in self.hv_list:
2883 # if the hypervisor doesn't already exist in the cluster
2884 # hvparams, we initialize it to empty, and then (in both
2885 # cases) we make sure to fill the defaults, as we might not
2886 # have a complete defaults list if the hypervisor wasn't
2888 if hv not in new_hvp:
2890 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2891 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2893 self.hv_list = cluster.enabled_hypervisors
2895 if self.op.hvparams or self.op.enabled_hypervisors is not None:
2896 # either the enabled list has changed, or the parameters have, validate
2897 for hv_name, hv_params in self.new_hvparams.items():
2898 if ((self.op.hvparams and hv_name in self.op.hvparams) or
2899 (self.op.enabled_hypervisors and
2900 hv_name in self.op.enabled_hypervisors)):
2901 # either this is a new hypervisor, or its parameters have changed
2902 hv_class = hypervisor.GetHypervisor(hv_name)
2903 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2904 hv_class.CheckParameterSyntax(hv_params)
2905 _CheckHVParams(self, node_list, hv_name, hv_params)
2908 # no need to check any newly-enabled hypervisors, since the
2909 # defaults have already been checked in the above code-block
2910 for os_name, os_hvp in self.new_os_hvp.items():
2911 for hv_name, hv_params in os_hvp.items():
2912 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2913 # we need to fill in the new os_hvp on top of the actual hv_p
2914 cluster_defaults = self.new_hvparams.get(hv_name, {})
2915 new_osp = objects.FillDict(cluster_defaults, hv_params)
2916 hv_class = hypervisor.GetHypervisor(hv_name)
2917 hv_class.CheckParameterSyntax(new_osp)
2918 _CheckHVParams(self, node_list, hv_name, new_osp)
2920 if self.op.default_iallocator:
2921 alloc_script = utils.FindFile(self.op.default_iallocator,
2922 constants.IALLOCATOR_SEARCH_PATH,
2924 if alloc_script is None:
2925 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2926 " specified" % self.op.default_iallocator,
2929 def Exec(self, feedback_fn):
2930 """Change the parameters of the cluster.
2933 if self.op.vg_name is not None:
2934 new_volume = self.op.vg_name
2937 if new_volume != self.cfg.GetVGName():
2938 self.cfg.SetVGName(new_volume)
2940 feedback_fn("Cluster LVM configuration already in desired"
2941 " state, not changing")
2942 if self.op.drbd_helper is not None:
2943 new_helper = self.op.drbd_helper
2946 if new_helper != self.cfg.GetDRBDHelper():
2947 self.cfg.SetDRBDHelper(new_helper)
2949 feedback_fn("Cluster DRBD helper already in desired state,"
2951 if self.op.hvparams:
2952 self.cluster.hvparams = self.new_hvparams
2954 self.cluster.os_hvp = self.new_os_hvp
2955 if self.op.enabled_hypervisors is not None:
2956 self.cluster.hvparams = self.new_hvparams
2957 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2958 if self.op.beparams:
2959 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2960 if self.op.nicparams:
2961 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2962 if self.op.osparams:
2963 self.cluster.osparams = self.new_osp
2964 if self.op.ndparams:
2965 self.cluster.ndparams = self.new_ndparams
2967 if self.op.candidate_pool_size is not None:
2968 self.cluster.candidate_pool_size = self.op.candidate_pool_size
2969 # we need to update the pool size here, otherwise the save will fail
2970 _AdjustCandidatePool(self, [])
2972 if self.op.maintain_node_health is not None:
2973 self.cluster.maintain_node_health = self.op.maintain_node_health
2975 if self.op.prealloc_wipe_disks is not None:
2976 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2978 if self.op.add_uids is not None:
2979 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2981 if self.op.remove_uids is not None:
2982 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2984 if self.op.uid_pool is not None:
2985 self.cluster.uid_pool = self.op.uid_pool
2987 if self.op.default_iallocator is not None:
2988 self.cluster.default_iallocator = self.op.default_iallocator
2990 if self.op.reserved_lvs is not None:
2991 self.cluster.reserved_lvs = self.op.reserved_lvs
2993 def helper_os(aname, mods, desc):
2995 lst = getattr(self.cluster, aname)
2996 for key, val in mods:
2997 if key == constants.DDM_ADD:
2999 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3002 elif key == constants.DDM_REMOVE:
3006 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3008 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3010 if self.op.hidden_os:
3011 helper_os("hidden_os", self.op.hidden_os, "hidden")
3013 if self.op.blacklisted_os:
3014 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3016 if self.op.master_netdev:
3017 master = self.cfg.GetMasterNode()
3018 feedback_fn("Shutting down master ip on the current netdev (%s)" %
3019 self.cluster.master_netdev)
3020 result = self.rpc.call_node_stop_master(master, False)
3021 result.Raise("Could not disable the master ip")
3022 feedback_fn("Changing master_netdev from %s to %s" %
3023 (self.cluster.master_netdev, self.op.master_netdev))
3024 self.cluster.master_netdev = self.op.master_netdev
3026 self.cfg.Update(self.cluster, feedback_fn)
3028 if self.op.master_netdev:
3029 feedback_fn("Starting the master ip on the new master netdev (%s)" %
3030 self.op.master_netdev)
3031 result = self.rpc.call_node_start_master(master, False, False)
3033 self.LogWarning("Could not re-enable the master ip on"
3034 " the master, please restart manually: %s",
3038 def _UploadHelper(lu, nodes, fname):
3039 """Helper for uploading a file and showing warnings.
3042 if os.path.exists(fname):
3043 result = lu.rpc.call_upload_file(nodes, fname)
3044 for to_node, to_result in result.items():
3045 msg = to_result.fail_msg
3047 msg = ("Copy of file %s to node %s failed: %s" %
3048 (fname, to_node, msg))
3049 lu.proc.LogWarning(msg)
3052 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3053 """Distribute additional files which are part of the cluster configuration.
3055 ConfigWriter takes care of distributing the config and ssconf files, but
3056 there are more files which should be distributed to all nodes. This function
3057 makes sure those are copied.
3059 @param lu: calling logical unit
3060 @param additional_nodes: list of nodes not in the config to distribute to
3061 @type additional_vm: boolean
3062 @param additional_vm: whether the additional nodes are vm-capable or not
3065 # 1. Gather target nodes
3066 myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3067 dist_nodes = lu.cfg.GetOnlineNodeList()
3068 nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3069 vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3070 if additional_nodes is not None:
3071 dist_nodes.extend(additional_nodes)
3073 vm_nodes.extend(additional_nodes)
3074 if myself.name in dist_nodes:
3075 dist_nodes.remove(myself.name)
3076 if myself.name in vm_nodes:
3077 vm_nodes.remove(myself.name)
3079 # 2. Gather files to distribute
3080 dist_files = set([constants.ETC_HOSTS,
3081 constants.SSH_KNOWN_HOSTS_FILE,
3082 constants.RAPI_CERT_FILE,
3083 constants.RAPI_USERS_FILE,
3084 constants.CONFD_HMAC_KEY,
3085 constants.CLUSTER_DOMAIN_SECRET_FILE,
3089 enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3090 for hv_name in enabled_hypervisors:
3091 hv_class = hypervisor.GetHypervisor(hv_name)
3092 vm_files.update(hv_class.GetAncillaryFiles())
3094 # 3. Perform the files upload
3095 for fname in dist_files:
3096 _UploadHelper(lu, dist_nodes, fname)
3097 for fname in vm_files:
3098 _UploadHelper(lu, vm_nodes, fname)
3101 class LUClusterRedistConf(NoHooksLU):
3102 """Force the redistribution of cluster configuration.
3104 This is a very simple LU.
3109 def ExpandNames(self):
3110 self.needed_locks = {
3111 locking.LEVEL_NODE: locking.ALL_SET,
3113 self.share_locks[locking.LEVEL_NODE] = 1
3115 def Exec(self, feedback_fn):
3116 """Redistribute the configuration.
3119 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3120 _RedistributeAncillaryFiles(self)
3123 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3124 """Sleep and poll for an instance's disk to sync.
3127 if not instance.disks or disks is not None and not disks:
3130 disks = _ExpandCheckDisks(instance, disks)
3133 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3135 node = instance.primary_node
3138 lu.cfg.SetDiskID(dev, node)
3140 # TODO: Convert to utils.Retry
3143 degr_retries = 10 # in seconds, as we sleep 1 second each time
3147 cumul_degraded = False
3148 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3149 msg = rstats.fail_msg
3151 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3154 raise errors.RemoteError("Can't contact node %s for mirror data,"
3155 " aborting." % node)
3158 rstats = rstats.payload
3160 for i, mstat in enumerate(rstats):
3162 lu.LogWarning("Can't compute data for node %s/%s",
3163 node, disks[i].iv_name)
3166 cumul_degraded = (cumul_degraded or
3167 (mstat.is_degraded and mstat.sync_percent is None))
3168 if mstat.sync_percent is not None:
3170 if mstat.estimated_time is not None:
3171 rem_time = ("%s remaining (estimated)" %
3172 utils.FormatSeconds(mstat.estimated_time))
3173 max_time = mstat.estimated_time
3175 rem_time = "no time estimate"
3176 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3177 (disks[i].iv_name, mstat.sync_percent, rem_time))
3179 # if we're done but degraded, let's do a few small retries, to
3180 # make sure we see a stable and not transient situation; therefore
3181 # we force restart of the loop
3182 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3183 logging.info("Degraded disks found, %d retries left", degr_retries)
3191 time.sleep(min(60, max_time))
3194 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3195 return not cumul_degraded
3198 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3199 """Check that mirrors are not degraded.
3201 The ldisk parameter, if True, will change the test from the
3202 is_degraded attribute (which represents overall non-ok status for
3203 the device(s)) to the ldisk (representing the local storage status).
3206 lu.cfg.SetDiskID(dev, node)
3210 if on_primary or dev.AssembleOnSecondary():
3211 rstats = lu.rpc.call_blockdev_find(node, dev)
3212 msg = rstats.fail_msg
3214 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3216 elif not rstats.payload:
3217 lu.LogWarning("Can't find disk on node %s", node)
3221 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3223 result = result and not rstats.payload.is_degraded
3226 for child in dev.children:
3227 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3232 class LUOobCommand(NoHooksLU):
3233 """Logical unit for OOB handling.
3237 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
3239 def CheckPrereq(self):
3240 """Check prerequisites.
3243 - the node exists in the configuration
3246 Any errors are signaled by raising errors.OpPrereqError.
3250 self.master_node = self.cfg.GetMasterNode()
3252 if self.op.node_names:
3253 if self.op.command in self._SKIP_MASTER:
3254 if self.master_node in self.op.node_names:
3255 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
3256 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
3258 if master_oob_handler:
3259 additional_text = ("Run '%s %s %s' if you want to operate on the"
3260 " master regardless") % (master_oob_handler,
3264 additional_text = "The master node does not support out-of-band"
3266 raise errors.OpPrereqError(("Operating on the master node %s is not"
3267 " allowed for %s\n%s") %
3268 (self.master_node, self.op.command,
3269 additional_text), errors.ECODE_INVAL)
3271 self.op.node_names = self.cfg.GetNodeList()
3272 if self.op.command in self._SKIP_MASTER:
3273 self.op.node_names.remove(self.master_node)
3275 if self.op.command in self._SKIP_MASTER:
3276 assert self.master_node not in self.op.node_names
3278 for node_name in self.op.node_names:
3279 node = self.cfg.GetNodeInfo(node_name)
3282 raise errors.OpPrereqError("Node %s not found" % node_name,
3285 self.nodes.append(node)
3287 if (not self.op.ignore_status and
3288 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
3289 raise errors.OpPrereqError(("Cannot power off node %s because it is"
3290 " not marked offline") % node_name,
3293 def ExpandNames(self):
3294 """Gather locks we need.
3297 if self.op.node_names:
3298 self.op.node_names = [_ExpandNodeName(self.cfg, name)
3299 for name in self.op.node_names]
3300 lock_names = self.op.node_names
3302 lock_names = locking.ALL_SET
3304 self.needed_locks = {
3305 locking.LEVEL_NODE: lock_names,
3308 def Exec(self, feedback_fn):
3309 """Execute OOB and return result if we expect any.
3312 master_node = self.master_node
3315 for node in self.nodes:
3316 node_entry = [(constants.RS_NORMAL, node.name)]
3317 ret.append(node_entry)
3319 oob_program = _SupportsOob(self.cfg, node)
3322 node_entry.append((constants.RS_UNAVAIL, None))
3325 logging.info("Executing out-of-band command '%s' using '%s' on %s",
3326 self.op.command, oob_program, node.name)
3327 result = self.rpc.call_run_oob(master_node, oob_program,
3328 self.op.command, node.name,
3332 self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3333 node.name, result.fail_msg)
3334 node_entry.append((constants.RS_NODATA, None))
3337 self._CheckPayload(result)
3338 except errors.OpExecError, err:
3339 self.LogWarning("The payload returned by '%s' is not valid: %s",
3341 node_entry.append((constants.RS_NODATA, None))
3343 if self.op.command == constants.OOB_HEALTH:
3344 # For health we should log important events
3345 for item, status in result.payload:
3346 if status in [constants.OOB_STATUS_WARNING,
3347 constants.OOB_STATUS_CRITICAL]:
3348 self.LogWarning("On node '%s' item '%s' has status '%s'",
3349 node.name, item, status)
3351 if self.op.command == constants.OOB_POWER_ON:
3353 elif self.op.command == constants.OOB_POWER_OFF:
3354 node.powered = False
3355 elif self.op.command == constants.OOB_POWER_STATUS:
3356 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3357 if powered != node.powered:
3358 logging.warning(("Recorded power state (%s) of node '%s' does not"
3359 " match actual power state (%s)"), node.powered,
3362 # For configuration changing commands we should update the node
3363 if self.op.command in (constants.OOB_POWER_ON,
3364 constants.OOB_POWER_OFF):
3365 self.cfg.Update(node, feedback_fn)
3367 node_entry.append((constants.RS_NORMAL, result.payload))
3371 def _CheckPayload(self, result):
3372 """Checks if the payload is valid.
3374 @param result: RPC result
3375 @raises errors.OpExecError: If payload is not valid
3379 if self.op.command == constants.OOB_HEALTH:
3380 if not isinstance(result.payload, list):
3381 errs.append("command 'health' is expected to return a list but got %s" %
3382 type(result.payload))
3384 for item, status in result.payload:
3385 if status not in constants.OOB_STATUSES:
3386 errs.append("health item '%s' has invalid status '%s'" %
3389 if self.op.command == constants.OOB_POWER_STATUS:
3390 if not isinstance(result.payload, dict):
3391 errs.append("power-status is expected to return a dict but got %s" %
3392 type(result.payload))
3394 if self.op.command in [
3395 constants.OOB_POWER_ON,
3396 constants.OOB_POWER_OFF,
3397 constants.OOB_POWER_CYCLE,
3399 if result.payload is not None:
3400 errs.append("%s is expected to not return payload but got '%s'" %
3401 (self.op.command, result.payload))
3404 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3405 utils.CommaJoin(errs))
3409 class LUOsDiagnose(NoHooksLU):
3410 """Logical unit for OS diagnose/query.
3415 _BLK = "blacklisted"
3417 _FIELDS_STATIC = utils.FieldSet()
3418 _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3419 "parameters", "api_versions", _HID, _BLK)
3421 def CheckArguments(self):
3423 raise errors.OpPrereqError("Selective OS query not supported",
3426 _CheckOutputFields(static=self._FIELDS_STATIC,
3427 dynamic=self._FIELDS_DYNAMIC,
3428 selected=self.op.output_fields)
3430 def ExpandNames(self):
3431 # Lock all nodes, in shared mode
3432 # Temporary removal of locks, should be reverted later
3433 # TODO: reintroduce locks when they are lighter-weight
3434 self.needed_locks = {}
3435 #self.share_locks[locking.LEVEL_NODE] = 1
3436 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3439 def _DiagnoseByOS(rlist):
3440 """Remaps a per-node return list into an a per-os per-node dictionary
3442 @param rlist: a map with node names as keys and OS objects as values
3445 @return: a dictionary with osnames as keys and as value another
3446 map, with nodes as keys and tuples of (path, status, diagnose,
3447 variants, parameters, api_versions) as values, eg::
3449 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3450 (/srv/..., False, "invalid api")],
3451 "node2": [(/srv/..., True, "", [], [])]}
3456 # we build here the list of nodes that didn't fail the RPC (at RPC
3457 # level), so that nodes with a non-responding node daemon don't
3458 # make all OSes invalid
3459 good_nodes = [node_name for node_name in rlist
3460 if not rlist[node_name].fail_msg]
3461 for node_name, nr in rlist.items():
3462 if nr.fail_msg or not nr.payload:
3464 for (name, path, status, diagnose, variants,
3465 params, api_versions) in nr.payload:
3466 if name not in all_os:
3467 # build a list of nodes for this os containing empty lists
3468 # for each node in node_list
3470 for nname in good_nodes:
3471 all_os[name][nname] = []
3472 # convert params from [name, help] to (name, help)
3473 params = [tuple(v) for v in params]
3474 all_os[name][node_name].append((path, status, diagnose,
3475 variants, params, api_versions))
3478 def Exec(self, feedback_fn):
3479 """Compute the list of OSes.
3482 valid_nodes = [node.name
3483 for node in self.cfg.GetAllNodesInfo().values()
3484 if not node.offline and node.vm_capable]
3485 node_data = self.rpc.call_os_diagnose(valid_nodes)
3486 pol = self._DiagnoseByOS(node_data)
3488 cluster = self.cfg.GetClusterInfo()
3490 for os_name in utils.NiceSort(pol.keys()):
3491 os_data = pol[os_name]
3494 (variants, params, api_versions) = null_state = (set(), set(), set())
3495 for idx, osl in enumerate(os_data.values()):
3496 valid = bool(valid and osl and osl[0][1])
3498 (variants, params, api_versions) = null_state
3500 node_variants, node_params, node_api = osl[0][3:6]
3501 if idx == 0: # first entry
3502 variants = set(node_variants)
3503 params = set(node_params)
3504 api_versions = set(node_api)
3505 else: # keep consistency
3506 variants.intersection_update(node_variants)
3507 params.intersection_update(node_params)
3508 api_versions.intersection_update(node_api)
3510 is_hid = os_name in cluster.hidden_os
3511 is_blk = os_name in cluster.blacklisted_os
3512 if ((self._HID not in self.op.output_fields and is_hid) or
3513 (self._BLK not in self.op.output_fields and is_blk) or
3514 (self._VLD not in self.op.output_fields and not valid)):
3517 for field in self.op.output_fields:
3520 elif field == self._VLD:
3522 elif field == "node_status":
3523 # this is just a copy of the dict
3525 for node_name, nos_list in os_data.items():
3526 val[node_name] = nos_list
3527 elif field == "variants":
3528 val = utils.NiceSort(list(variants))
3529 elif field == "parameters":
3531 elif field == "api_versions":
3532 val = list(api_versions)
3533 elif field == self._HID:
3535 elif field == self._BLK:
3538 raise errors.ParameterError(field)
3545 class LUNodeRemove(LogicalUnit):
3546 """Logical unit for removing a node.
3549 HPATH = "node-remove"
3550 HTYPE = constants.HTYPE_NODE
3552 def BuildHooksEnv(self):
3555 This doesn't run on the target node in the pre phase as a failed
3556 node would then be impossible to remove.
3560 "OP_TARGET": self.op.node_name,
3561 "NODE_NAME": self.op.node_name,
3563 all_nodes = self.cfg.GetNodeList()
3565 all_nodes.remove(self.op.node_name)
3567 logging.warning("Node %s which is about to be removed not found"
3568 " in the all nodes list", self.op.node_name)
3569 return env, all_nodes, all_nodes
3571 def CheckPrereq(self):
3572 """Check prerequisites.
3575 - the node exists in the configuration
3576 - it does not have primary or secondary instances
3577 - it's not the master
3579 Any errors are signaled by raising errors.OpPrereqError.
3582 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3583 node = self.cfg.GetNodeInfo(self.op.node_name)
3584 assert node is not None
3586 instance_list = self.cfg.GetInstanceList()
3588 masternode = self.cfg.GetMasterNode()
3589 if node.name == masternode:
3590 raise errors.OpPrereqError("Node is the master node,"
3591 " you need to failover first.",
3594 for instance_name in instance_list:
3595 instance = self.cfg.GetInstanceInfo(instance_name)
3596 if node.name in instance.all_nodes:
3597 raise errors.OpPrereqError("Instance %s is still running on the node,"
3598 " please remove first." % instance_name,
3600 self.op.node_name = node.name
3603 def Exec(self, feedback_fn):
3604 """Removes the node from the cluster.
3608 logging.info("Stopping the node daemon and removing configs from node %s",
3611 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3613 # Promote nodes to master candidate as needed
3614 _AdjustCandidatePool(self, exceptions=[node.name])
3615 self.context.RemoveNode(node.name)
3617 # Run post hooks on the node before it's removed
3618 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3620 hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3622 # pylint: disable-msg=W0702
3623 self.LogWarning("Errors occurred running hooks on %s" % node.name)
3625 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3626 msg = result.fail_msg
3628 self.LogWarning("Errors encountered on the remote node while leaving"
3629 " the cluster: %s", msg)
3631 # Remove node from our /etc/hosts
3632 if self.cfg.GetClusterInfo().modify_etc_hosts:
3633 master_node = self.cfg.GetMasterNode()
3634 result = self.rpc.call_etc_hosts_modify(master_node,
3635 constants.ETC_HOSTS_REMOVE,
3637 result.Raise("Can't update hosts file with new host data")
3638 _RedistributeAncillaryFiles(self)
3641 class _NodeQuery(_QueryBase):
3642 FIELDS = query.NODE_FIELDS
3644 def ExpandNames(self, lu):
3645 lu.needed_locks = {}
3646 lu.share_locks[locking.LEVEL_NODE] = 1
3649 self.wanted = _GetWantedNodes(lu, self.names)
3651 self.wanted = locking.ALL_SET
3653 self.do_locking = (self.use_locking and
3654 query.NQ_LIVE in self.requested_data)
3657 # if we don't request only static fields, we need to lock the nodes
3658 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3660 def DeclareLocks(self, lu, level):
3663 def _GetQueryData(self, lu):
3664 """Computes the list of nodes and their attributes.
3667 all_info = lu.cfg.GetAllNodesInfo()
3669 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3671 # Gather data as requested
3672 if query.NQ_LIVE in self.requested_data:
3673 # filter out non-vm_capable nodes
3674 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
3676 node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
3677 lu.cfg.GetHypervisorType())
3678 live_data = dict((name, nresult.payload)
3679 for (name, nresult) in node_data.items()
3680 if not nresult.fail_msg and nresult.payload)
3684 if query.NQ_INST in self.requested_data:
3685 node_to_primary = dict([(name, set()) for name in nodenames])
3686 node_to_secondary = dict([(name, set()) for name in nodenames])
3688 inst_data = lu.cfg.GetAllInstancesInfo()
3690 for inst in inst_data.values():
3691 if inst.primary_node in node_to_primary:
3692 node_to_primary[inst.primary_node].add(inst.name)
3693 for secnode in inst.secondary_nodes:
3694 if secnode in node_to_secondary:
3695 node_to_secondary[secnode].add(inst.name)
3697 node_to_primary = None
3698 node_to_secondary = None
3700 if query.NQ_OOB in self.requested_data:
3701 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3702 for name, node in all_info.iteritems())
3706 if query.NQ_GROUP in self.requested_data:
3707 groups = lu.cfg.GetAllNodeGroupsInfo()
3711 return query.NodeQueryData([all_info[name] for name in nodenames],
3712 live_data, lu.cfg.GetMasterNode(),
3713 node_to_primary, node_to_secondary, groups,
3714 oob_support, lu.cfg.GetClusterInfo())
3717 class LUNodeQuery(NoHooksLU):
3718 """Logical unit for querying nodes.
3721 # pylint: disable-msg=W0142
3724 def CheckArguments(self):
3725 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
3726 self.op.output_fields, self.op.use_locking)
3728 def ExpandNames(self):
3729 self.nq.ExpandNames(self)
3731 def Exec(self, feedback_fn):
3732 return self.nq.OldStyleQuery(self)
3735 class LUNodeQueryvols(NoHooksLU):
3736 """Logical unit for getting volumes on node(s).
3740 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3741 _FIELDS_STATIC = utils.FieldSet("node")
3743 def CheckArguments(self):
3744 _CheckOutputFields(static=self._FIELDS_STATIC,
3745 dynamic=self._FIELDS_DYNAMIC,
3746 selected=self.op.output_fields)
3748 def ExpandNames(self):
3749 self.needed_locks = {}
3750 self.share_locks[locking.LEVEL_NODE] = 1
3751 if not self.op.nodes:
3752 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3754 self.needed_locks[locking.LEVEL_NODE] = \
3755 _GetWantedNodes(self, self.op.nodes)
3757 def Exec(self, feedback_fn):
3758 """Computes the list of nodes and their attributes.
3761 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3762 volumes = self.rpc.call_node_volumes(nodenames)
3764 ilist = [self.cfg.GetInstanceInfo(iname) for iname
3765 in self.cfg.GetInstanceList()]
3767 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3770 for node in nodenames:
3771 nresult = volumes[node]
3774 msg = nresult.fail_msg
3776 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3779 node_vols = nresult.payload[:]
3780 node_vols.sort(key=lambda vol: vol['dev'])
3782 for vol in node_vols:
3784 for field in self.op.output_fields:
3787 elif field == "phys":
3791 elif field == "name":
3793 elif field == "size":
3794 val = int(float(vol['size']))
3795 elif field == "instance":
3797 if node not in lv_by_node[inst]:
3799 if vol['name'] in lv_by_node[inst][node]:
3805 raise errors.ParameterError(field)
3806 node_output.append(str(val))
3808 output.append(node_output)
3813 class LUNodeQueryStorage(NoHooksLU):
3814 """Logical unit for getting information on storage units on node(s).
3817 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3820 def CheckArguments(self):
3821 _CheckOutputFields(static=self._FIELDS_STATIC,
3822 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3823 selected=self.op.output_fields)
3825 def ExpandNames(self):
3826 self.needed_locks = {}
3827 self.share_locks[locking.LEVEL_NODE] = 1
3830 self.needed_locks[locking.LEVEL_NODE] = \
3831 _GetWantedNodes(self, self.op.nodes)
3833 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3835 def Exec(self, feedback_fn):
3836 """Computes the list of nodes and their attributes.
3839 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3841 # Always get name to sort by
3842 if constants.SF_NAME in self.op.output_fields:
3843 fields = self.op.output_fields[:]
3845 fields = [constants.SF_NAME] + self.op.output_fields
3847 # Never ask for node or type as it's only known to the LU
3848 for extra in [constants.SF_NODE, constants.SF_TYPE]:
3849 while extra in fields:
3850 fields.remove(extra)
3852 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3853 name_idx = field_idx[constants.SF_NAME]
3855 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3856 data = self.rpc.call_storage_list(self.nodes,
3857 self.op.storage_type, st_args,
3858 self.op.name, fields)
3862 for node in utils.NiceSort(self.nodes):
3863 nresult = data[node]
3867 msg = nresult.fail_msg
3869 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3872 rows = dict([(row[name_idx], row) for row in nresult.payload])
3874 for name in utils.NiceSort(rows.keys()):
3879 for field in self.op.output_fields:
3880 if field == constants.SF_NODE:
3882 elif field == constants.SF_TYPE:
3883 val = self.op.storage_type
3884 elif field in field_idx:
3885 val = row[field_idx[field]]
3887 raise errors.ParameterError(field)
3896 class _InstanceQuery(_QueryBase):
3897 FIELDS = query.INSTANCE_FIELDS
3899 def ExpandNames(self, lu):
3900 lu.needed_locks = {}
3901 lu.share_locks[locking.LEVEL_INSTANCE] = 1
3902 lu.share_locks[locking.LEVEL_NODE] = 1
3905 self.wanted = _GetWantedInstances(lu, self.names)
3907 self.wanted = locking.ALL_SET
3909 self.do_locking = (self.use_locking and
3910 query.IQ_LIVE in self.requested_data)
3912 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3913 lu.needed_locks[locking.LEVEL_NODE] = []
3914 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3916 def DeclareLocks(self, lu, level):
3917 if level == locking.LEVEL_NODE and self.do_locking:
3918 lu._LockInstancesNodes() # pylint: disable-msg=W0212
3920 def _GetQueryData(self, lu):
3921 """Computes the list of instances and their attributes.
3924 cluster = lu.cfg.GetClusterInfo()
3925 all_info = lu.cfg.GetAllInstancesInfo()
3927 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3929 instance_list = [all_info[name] for name in instance_names]
3930 nodes = frozenset(itertools.chain(*(inst.all_nodes
3931 for inst in instance_list)))
3932 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3935 wrongnode_inst = set()
3937 # Gather data as requested
3938 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
3940 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3942 result = node_data[name]
3944 # offline nodes will be in both lists
3945 assert result.fail_msg
3946 offline_nodes.append(name)
3948 bad_nodes.append(name)
3949 elif result.payload:
3950 for inst in result.payload:
3951 if all_info[inst].primary_node == name:
3952 live_data.update(result.payload)
3954 wrongnode_inst.add(inst)
3955 # else no instance is alive
3959 if query.IQ_DISKUSAGE in self.requested_data:
3960 disk_usage = dict((inst.name,
3961 _ComputeDiskSize(inst.disk_template,
3962 [{"size": disk.size}
3963 for disk in inst.disks]))
3964 for inst in instance_list)
3968 if query.IQ_CONSOLE in self.requested_data:
3970 for inst in instance_list:
3971 if inst.name in live_data:
3972 # Instance is running
3973 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
3975 consinfo[inst.name] = None
3976 assert set(consinfo.keys()) == set(instance_names)
3980 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3981 disk_usage, offline_nodes, bad_nodes,
3982 live_data, wrongnode_inst, consinfo)
3985 class LUQuery(NoHooksLU):
3986 """Query for resources/items of a certain kind.
3989 # pylint: disable-msg=W0142
3992 def CheckArguments(self):
3993 qcls = _GetQueryImplementation(self.op.what)
3995 self.impl = qcls(self.op.filter, self.op.fields, False)
3997 def ExpandNames(self):
3998 self.impl.ExpandNames(self)
4000 def DeclareLocks(self, level):
4001 self.impl.DeclareLocks(self, level)
4003 def Exec(self, feedback_fn):
4004 return self.impl.NewStyleQuery(self)
4007 class LUQueryFields(NoHooksLU):
4008 """Query for resources/items of a certain kind.
4011 # pylint: disable-msg=W0142
4014 def CheckArguments(self):
4015 self.qcls = _GetQueryImplementation(self.op.what)
4017 def ExpandNames(self):
4018 self.needed_locks = {}
4020 def Exec(self, feedback_fn):
4021 return self.qcls.FieldsQuery(self.op.fields)
4024 class LUNodeModifyStorage(NoHooksLU):
4025 """Logical unit for modifying a storage volume on a node.
4030 def CheckArguments(self):
4031 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4033 storage_type = self.op.storage_type
4036 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4038 raise errors.OpPrereqError("Storage units of type '%s' can not be"
4039 " modified" % storage_type,
4042 diff = set(self.op.changes.keys()) - modifiable
4044 raise errors.OpPrereqError("The following fields can not be modified for"
4045 " storage units of type '%s': %r" %
4046 (storage_type, list(diff)),
4049 def ExpandNames(self):
4050 self.needed_locks = {
4051 locking.LEVEL_NODE: self.op.node_name,
4054 def Exec(self, feedback_fn):
4055 """Computes the list of nodes and their attributes.
4058 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4059 result = self.rpc.call_storage_modify(self.op.node_name,
4060 self.op.storage_type, st_args,
4061 self.op.name, self.op.changes)
4062 result.Raise("Failed to modify storage unit '%s' on %s" %
4063 (self.op.name, self.op.node_name))
4066 class LUNodeAdd(LogicalUnit):
4067 """Logical unit for adding node to the cluster.
4071 HTYPE = constants.HTYPE_NODE
4072 _NFLAGS = ["master_capable", "vm_capable"]
4074 def CheckArguments(self):
4075 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4076 # validate/normalize the node name
4077 self.hostname = netutils.GetHostname(name=self.op.node_name,
4078 family=self.primary_ip_family)
4079 self.op.node_name = self.hostname.name
4080 if self.op.readd and self.op.group:
4081 raise errors.OpPrereqError("Cannot pass a node group when a node is"
4082 " being readded", errors.ECODE_INVAL)
4084 def BuildHooksEnv(self):
4087 This will run on all nodes before, and on all nodes + the new node after.
4091 "OP_TARGET": self.op.node_name,
4092 "NODE_NAME": self.op.node_name,
4093 "NODE_PIP": self.op.primary_ip,
4094 "NODE_SIP": self.op.secondary_ip,
4095 "MASTER_CAPABLE": str(self.op.master_capable),
4096 "VM_CAPABLE": str(self.op.vm_capable),
4098 nodes_0 = self.cfg.GetNodeList()
4099 nodes_1 = nodes_0 + [self.op.node_name, ]
4100 return env, nodes_0, nodes_1
4102 def CheckPrereq(self):
4103 """Check prerequisites.
4106 - the new node is not already in the config
4108 - its parameters (single/dual homed) matches the cluster
4110 Any errors are signaled by raising errors.OpPrereqError.
4114 hostname = self.hostname
4115 node = hostname.name
4116 primary_ip = self.op.primary_ip = hostname.ip
4117 if self.op.secondary_ip is None:
4118 if self.primary_ip_family == netutils.IP6Address.family:
4119 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4120 " IPv4 address must be given as secondary",
4122 self.op.secondary_ip = primary_ip
4124 secondary_ip = self.op.secondary_ip
4125 if not netutils.IP4Address.IsValid(secondary_ip):
4126 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4127 " address" % secondary_ip, errors.ECODE_INVAL)
4129 node_list = cfg.GetNodeList()
4130 if not self.op.readd and node in node_list:
4131 raise errors.OpPrereqError("Node %s is already in the configuration" %
4132 node, errors.ECODE_EXISTS)
4133 elif self.op.readd and node not in node_list:
4134 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4137 self.changed_primary_ip = False
4139 for existing_node_name in node_list:
4140 existing_node = cfg.GetNodeInfo(existing_node_name)
4142 if self.op.readd and node == existing_node_name:
4143 if existing_node.secondary_ip != secondary_ip:
4144 raise errors.OpPrereqError("Readded node doesn't have the same IP"
4145 " address configuration as before",
4147 if existing_node.primary_ip != primary_ip:
4148 self.changed_primary_ip = True
4152 if (existing_node.primary_ip == primary_ip or
4153 existing_node.secondary_ip == primary_ip or
4154 existing_node.primary_ip == secondary_ip or
4155 existing_node.secondary_ip == secondary_ip):
4156 raise errors.OpPrereqError("New node ip address(es) conflict with"
4157 " existing node %s" % existing_node.name,
4158 errors.ECODE_NOTUNIQUE)
4160 # After this 'if' block, None is no longer a valid value for the
4161 # _capable op attributes
4163 old_node = self.cfg.GetNodeInfo(node)
4164 assert old_node is not None, "Can't retrieve locked node %s" % node
4165 for attr in self._NFLAGS:
4166 if getattr(self.op, attr) is None:
4167 setattr(self.op, attr, getattr(old_node, attr))
4169 for attr in self._NFLAGS:
4170 if getattr(self.op, attr) is None:
4171 setattr(self.op, attr, True)
4173 if self.op.readd and not self.op.vm_capable:
4174 pri, sec = cfg.GetNodeInstances(node)
4176 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4177 " flag set to false, but it already holds"
4178 " instances" % node,
4181 # check that the type of the node (single versus dual homed) is the
4182 # same as for the master
4183 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4184 master_singlehomed = myself.secondary_ip == myself.primary_ip
4185 newbie_singlehomed = secondary_ip == primary_ip
4186 if master_singlehomed != newbie_singlehomed:
4187 if master_singlehomed:
4188 raise errors.OpPrereqError("The master has no secondary ip but the"
4189 " new node has one",
4192 raise errors.OpPrereqError("The master has a secondary ip but the"
4193 " new node doesn't have one",
4196 # checks reachability
4197 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
4198 raise errors.OpPrereqError("Node not reachable by ping",
4199 errors.ECODE_ENVIRON)
4201 if not newbie_singlehomed:
4202 # check reachability from my secondary ip to newbie's secondary ip
4203 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
4204 source=myself.secondary_ip):
4205 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4206 " based ping to node daemon port",
4207 errors.ECODE_ENVIRON)
4214 if self.op.master_capable:
4215 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
4217 self.master_candidate = False
4220 self.new_node = old_node
4222 node_group = cfg.LookupNodeGroup(self.op.group)
4223 self.new_node = objects.Node(name=node,
4224 primary_ip=primary_ip,
4225 secondary_ip=secondary_ip,
4226 master_candidate=self.master_candidate,
4227 offline=False, drained=False,
4230 if self.op.ndparams:
4231 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
4233 def Exec(self, feedback_fn):
4234 """Adds the new node to the cluster.
4237 new_node = self.new_node
4238 node = new_node.name
4240 # We adding a new node so we assume it's powered
4241 new_node.powered = True
4243 # for re-adds, reset the offline/drained/master-candidate flags;
4244 # we need to reset here, otherwise offline would prevent RPC calls
4245 # later in the procedure; this also means that if the re-add
4246 # fails, we are left with a non-offlined, broken node
4248 new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
4249 self.LogInfo("Readding a node, the offline/drained flags were reset")
4250 # if we demote the node, we do cleanup later in the procedure
4251 new_node.master_candidate = self.master_candidate
4252 if self.changed_primary_ip:
4253 new_node.primary_ip = self.op.primary_ip
4255 # copy the master/vm_capable flags
4256 for attr in self._NFLAGS:
4257 setattr(new_node, attr, getattr(self.op, attr))
4259 # notify the user about any possible mc promotion
4260 if new_node.master_candidate:
4261 self.LogInfo("Node will be a master candidate")
4263 if self.op.ndparams:
4264 new_node.ndparams = self.op.ndparams
4266 new_node.ndparams = {}
4268 # check connectivity
4269 result = self.rpc.call_version([node])[node]
4270 result.Raise("Can't get version information from node %s" % node)
4271 if constants.PROTOCOL_VERSION == result.payload:
4272 logging.info("Communication to node %s fine, sw version %s match",
4273 node, result.payload)
4275 raise errors.OpExecError("Version mismatch master version %s,"
4276 " node version %s" %
4277 (constants.PROTOCOL_VERSION, result.payload))
4279 # Add node to our /etc/hosts, and add key to known_hosts
4280 if self.cfg.GetClusterInfo().modify_etc_hosts:
4281 master_node = self.cfg.GetMasterNode()
4282 result = self.rpc.call_etc_hosts_modify(master_node,
4283 constants.ETC_HOSTS_ADD,
4286 result.Raise("Can't update hosts file with new host data")
4288 if new_node.secondary_ip != new_node.primary_ip:
4289 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
4292 node_verify_list = [self.cfg.GetMasterNode()]
4293 node_verify_param = {
4294 constants.NV_NODELIST: [node],
4295 # TODO: do a node-net-test as well?
4298 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
4299 self.cfg.GetClusterName())
4300 for verifier in node_verify_list:
4301 result[verifier].Raise("Cannot communicate with node %s" % verifier)
4302 nl_payload = result[verifier].payload[constants.NV_NODELIST]
4304 for failed in nl_payload:
4305 feedback_fn("ssh/hostname verification failed"
4306 " (checking from %s): %s" %
4307 (verifier, nl_payload[failed]))
4308 raise errors.OpExecError("ssh/hostname verification failed.")
4311 _RedistributeAncillaryFiles(self)
4312 self.context.ReaddNode(new_node)
4313 # make sure we redistribute the config
4314 self.cfg.Update(new_node, feedback_fn)
4315 # and make sure the new node will not have old files around
4316 if not new_node.master_candidate:
4317 result = self.rpc.call_node_demote_from_mc(new_node.name)
4318 msg = result.fail_msg
4320 self.LogWarning("Node failed to demote itself from master"
4321 " candidate status: %s" % msg)
4323 _RedistributeAncillaryFiles(self, additional_nodes=[node],
4324 additional_vm=self.op.vm_capable)
4325 self.context.AddNode(new_node, self.proc.GetECId())
4328 class LUNodeSetParams(LogicalUnit):
4329 """Modifies the parameters of a node.
4331 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
4332 to the node role (as _ROLE_*)
4333 @cvar _R2F: a dictionary from node role to tuples of flags
4334 @cvar _FLAGS: a list of attribute names corresponding to the flags
4337 HPATH = "node-modify"
4338 HTYPE = constants.HTYPE_NODE
4340 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
4342 (True, False, False): _ROLE_CANDIDATE,
4343 (False, True, False): _ROLE_DRAINED,
4344 (False, False, True): _ROLE_OFFLINE,
4345 (False, False, False): _ROLE_REGULAR,
4347 _R2F = dict((v, k) for k, v in _F2R.items())
4348 _FLAGS = ["master_candidate", "drained", "offline"]
4350 def CheckArguments(self):
4351 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4352 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
4353 self.op.master_capable, self.op.vm_capable,
4354 self.op.secondary_ip, self.op.ndparams]
4355 if all_mods.count(None) == len(all_mods):
4356 raise errors.OpPrereqError("Please pass at least one modification",
4358 if all_mods.count(True) > 1:
4359 raise errors.OpPrereqError("Can't set the node into more than one"
4360 " state at the same time",
4363 # Boolean value that tells us whether we might be demoting from MC
4364 self.might_demote = (self.op.master_candidate == False or
4365 self.op.offline == True or
4366 self.op.drained == True or
4367 self.op.master_capable == False)
4369 if self.op.secondary_ip:
4370 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
4371 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4372 " address" % self.op.secondary_ip,
4375 self.lock_all = self.op.auto_promote and self.might_demote
4376 self.lock_instances = self.op.secondary_ip is not None
4378 def ExpandNames(self):
4380 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
4382 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
4384 if self.lock_instances:
4385 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4387 def DeclareLocks(self, level):
4388 # If we have locked all instances, before waiting to lock nodes, release
4389 # all the ones living on nodes unrelated to the current operation.
4390 if level == locking.LEVEL_NODE and self.lock_instances:
4391 instances_release = []
4393 self.affected_instances = []
4394 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4395 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
4396 instance = self.context.cfg.GetInstanceInfo(instance_name)
4397 i_mirrored = instance.disk_template in constants.DTS_NET_MIRROR
4398 if i_mirrored and self.op.node_name in instance.all_nodes:
4399 instances_keep.append(instance_name)
4400 self.affected_instances.append(instance)
4402 instances_release.append(instance_name)
4403 if instances_release:
4404 self.context.glm.release(locking.LEVEL_INSTANCE, instances_release)
4405 self.acquired_locks[locking.LEVEL_INSTANCE] = instances_keep
4407 def BuildHooksEnv(self):
4410 This runs on the master node.
4414 "OP_TARGET": self.op.node_name,
4415 "MASTER_CANDIDATE": str(self.op.master_candidate),
4416 "OFFLINE": str(self.op.offline),
4417 "DRAINED": str(self.op.drained),
4418 "MASTER_CAPABLE": str(self.op.master_capable),
4419 "VM_CAPABLE": str(self.op.vm_capable),
4421 nl = [self.cfg.GetMasterNode(),
4425 def CheckPrereq(self):
4426 """Check prerequisites.
4428 This only checks the instance list against the existing names.
4431 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
4433 if (self.op.master_candidate is not None or
4434 self.op.drained is not None or
4435 self.op.offline is not None):
4436 # we can't change the master's node flags
4437 if self.op.node_name == self.cfg.GetMasterNode():
4438 raise errors.OpPrereqError("The master role can be changed"
4439 " only via master-failover",
4442 if self.op.master_candidate and not node.master_capable:
4443 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
4444 " it a master candidate" % node.name,
4447 if self.op.vm_capable == False:
4448 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
4450 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
4451 " the vm_capable flag" % node.name,
4454 if node.master_candidate and self.might_demote and not self.lock_all:
4455 assert not self.op.auto_promote, "auto_promote set but lock_all not"
4456 # check if after removing the current node, we're missing master
4458 (mc_remaining, mc_should, _) = \
4459 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
4460 if mc_remaining < mc_should:
4461 raise errors.OpPrereqError("Not enough master candidates, please"
4462 " pass auto promote option to allow"
4463 " promotion", errors.ECODE_STATE)
4465 self.old_flags = old_flags = (node.master_candidate,
4466 node.drained, node.offline)
4467 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
4468 self.old_role = old_role = self._F2R[old_flags]
4470 # Check for ineffective changes
4471 for attr in self._FLAGS:
4472 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
4473 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
4474 setattr(self.op, attr, None)
4476 # Past this point, any flag change to False means a transition
4477 # away from the respective state, as only real changes are kept
4479 # TODO: We might query the real power state if it supports OOB
4480 if _SupportsOob(self.cfg, node):
4481 if self.op.offline is False and not (node.powered or
4482 self.op.powered == True):
4483 raise errors.OpPrereqError(("Please power on node %s first before you"
4484 " can reset offline state") %
4486 elif self.op.powered is not None:
4487 raise errors.OpPrereqError(("Unable to change powered state for node %s"
4488 " which does not support out-of-band"
4489 " handling") % self.op.node_name)
4491 # If we're being deofflined/drained, we'll MC ourself if needed
4492 if (self.op.drained == False or self.op.offline == False or
4493 (self.op.master_capable and not node.master_capable)):
4494 if _DecideSelfPromotion(self):
4495 self.op.master_candidate = True
4496 self.LogInfo("Auto-promoting node to master candidate")
4498 # If we're no longer master capable, we'll demote ourselves from MC
4499 if self.op.master_capable == False and node.master_candidate:
4500 self.LogInfo("Demoting from master candidate")
4501 self.op.master_candidate = False
4504 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
4505 if self.op.master_candidate:
4506 new_role = self._ROLE_CANDIDATE
4507 elif self.op.drained:
4508 new_role = self._ROLE_DRAINED
4509 elif self.op.offline:
4510 new_role = self._ROLE_OFFLINE
4511 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
4512 # False is still in new flags, which means we're un-setting (the
4514 new_role = self._ROLE_REGULAR
4515 else: # no new flags, nothing, keep old role
4518 self.new_role = new_role
4520 if old_role == self._ROLE_OFFLINE and new_role != old_role:
4521 # Trying to transition out of offline status
4522 result = self.rpc.call_version([node.name])[node.name]
4524 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
4525 " to report its version: %s" %
4526 (node.name, result.fail_msg),
4529 self.LogWarning("Transitioning node from offline to online state"
4530 " without using re-add. Please make sure the node"
4533 if self.op.secondary_ip:
4534 # Ok even without locking, because this can't be changed by any LU
4535 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
4536 master_singlehomed = master.secondary_ip == master.primary_ip
4537 if master_singlehomed and self.op.secondary_ip:
4538 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
4539 " homed cluster", errors.ECODE_INVAL)
4542 if self.affected_instances:
4543 raise errors.OpPrereqError("Cannot change secondary ip: offline"
4544 " node has instances (%s) configured"
4545 " to use it" % self.affected_instances)
4547 # On online nodes, check that no instances are running, and that
4548 # the node has the new ip and we can reach it.
4549 for instance in self.affected_instances:
4550 _CheckInstanceDown(self, instance, "cannot change secondary ip")
4552 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
4553 if master.name != node.name:
4554 # check reachability from master secondary ip to new secondary ip
4555 if not netutils.TcpPing(self.op.secondary_ip,
4556 constants.DEFAULT_NODED_PORT,
4557 source=master.secondary_ip):
4558 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4559 " based ping to node daemon port",
4560 errors.ECODE_ENVIRON)
4562 if self.op.ndparams:
4563 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
4564 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
4565 self.new_ndparams = new_ndparams
4567 def Exec(self, feedback_fn):
4572 old_role = self.old_role
4573 new_role = self.new_role
4577 if self.op.ndparams:
4578 node.ndparams = self.new_ndparams
4580 if self.op.powered is not None:
4581 node.powered = self.op.powered
4583 for attr in ["master_capable", "vm_capable"]:
4584 val = getattr(self.op, attr)
4586 setattr(node, attr, val)
4587 result.append((attr, str(val)))
4589 if new_role != old_role:
4590 # Tell the node to demote itself, if no longer MC and not offline
4591 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
4592 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
4594 self.LogWarning("Node failed to demote itself: %s", msg)
4596 new_flags = self._R2F[new_role]
4597 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
4599 result.append((desc, str(nf)))
4600 (node.master_candidate, node.drained, node.offline) = new_flags
4602 # we locked all nodes, we adjust the CP before updating this node
4604 _AdjustCandidatePool(self, [node.name])
4606 if self.op.secondary_ip:
4607 node.secondary_ip = self.op.secondary_ip
4608 result.append(("secondary_ip", self.op.secondary_ip))
4610 # this will trigger configuration file update, if needed
4611 self.cfg.Update(node, feedback_fn)
4613 # this will trigger job queue propagation or cleanup if the mc
4615 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
4616 self.context.ReaddNode(node)
4621 class LUNodePowercycle(NoHooksLU):
4622 """Powercycles a node.
4627 def CheckArguments(self):
4628 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4629 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4630 raise errors.OpPrereqError("The node is the master and the force"
4631 " parameter was not set",
4634 def ExpandNames(self):
4635 """Locking for PowercycleNode.
4637 This is a last-resort option and shouldn't block on other
4638 jobs. Therefore, we grab no locks.
4641 self.needed_locks = {}
4643 def Exec(self, feedback_fn):
4647 result = self.rpc.call_node_powercycle(self.op.node_name,
4648 self.cfg.GetHypervisorType())
4649 result.Raise("Failed to schedule the reboot")
4650 return result.payload
4653 class LUClusterQuery(NoHooksLU):
4654 """Query cluster configuration.
4659 def ExpandNames(self):
4660 self.needed_locks = {}
4662 def Exec(self, feedback_fn):
4663 """Return cluster config.
4666 cluster = self.cfg.GetClusterInfo()
4669 # Filter just for enabled hypervisors
4670 for os_name, hv_dict in cluster.os_hvp.items():
4671 os_hvp[os_name] = {}
4672 for hv_name, hv_params in hv_dict.items():
4673 if hv_name in cluster.enabled_hypervisors:
4674 os_hvp[os_name][hv_name] = hv_params
4676 # Convert ip_family to ip_version
4677 primary_ip_version = constants.IP4_VERSION
4678 if cluster.primary_ip_family == netutils.IP6Address.family:
4679 primary_ip_version = constants.IP6_VERSION
4682 "software_version": constants.RELEASE_VERSION,
4683 "protocol_version": constants.PROTOCOL_VERSION,
4684 "config_version": constants.CONFIG_VERSION,
4685 "os_api_version": max(constants.OS_API_VERSIONS),
4686 "export_version": constants.EXPORT_VERSION,
4687 "architecture": (platform.architecture()[0], platform.machine()),
4688 "name": cluster.cluster_name,
4689 "master": cluster.master_node,
4690 "default_hypervisor": cluster.enabled_hypervisors[0],
4691 "enabled_hypervisors": cluster.enabled_hypervisors,
4692 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4693 for hypervisor_name in cluster.enabled_hypervisors]),
4695 "beparams": cluster.beparams,
4696 "osparams": cluster.osparams,
4697 "nicparams": cluster.nicparams,
4698 "ndparams": cluster.ndparams,
4699 "candidate_pool_size": cluster.candidate_pool_size,
4700 "master_netdev": cluster.master_netdev,
4701 "volume_group_name": cluster.volume_group_name,
4702 "drbd_usermode_helper": cluster.drbd_usermode_helper,
4703 "file_storage_dir": cluster.file_storage_dir,
4704 "shared_file_storage_dir": cluster.shared_file_storage_dir,
4705 "maintain_node_health": cluster.maintain_node_health,
4706 "ctime": cluster.ctime,
4707 "mtime": cluster.mtime,
4708 "uuid": cluster.uuid,
4709 "tags": list(cluster.GetTags()),
4710 "uid_pool": cluster.uid_pool,
4711 "default_iallocator": cluster.default_iallocator,
4712 "reserved_lvs": cluster.reserved_lvs,
4713 "primary_ip_version": primary_ip_version,
4714 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4715 "hidden_os": cluster.hidden_os,
4716 "blacklisted_os": cluster.blacklisted_os,
4722 class LUClusterConfigQuery(NoHooksLU):
4723 """Return configuration values.
4727 _FIELDS_DYNAMIC = utils.FieldSet()
4728 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4729 "watcher_pause", "volume_group_name")
4731 def CheckArguments(self):
4732 _CheckOutputFields(static=self._FIELDS_STATIC,
4733 dynamic=self._FIELDS_DYNAMIC,
4734 selected=self.op.output_fields)
4736 def ExpandNames(self):
4737 self.needed_locks = {}
4739 def Exec(self, feedback_fn):
4740 """Dump a representation of the cluster config to the standard output.
4744 for field in self.op.output_fields:
4745 if field == "cluster_name":
4746 entry = self.cfg.GetClusterName()
4747 elif field == "master_node":
4748 entry = self.cfg.GetMasterNode()
4749 elif field == "drain_flag":
4750 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4751 elif field == "watcher_pause":
4752 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4753 elif field == "volume_group_name":
4754 entry = self.cfg.GetVGName()
4756 raise errors.ParameterError(field)
4757 values.append(entry)
4761 class LUInstanceActivateDisks(NoHooksLU):
4762 """Bring up an instance's disks.
4767 def ExpandNames(self):
4768 self._ExpandAndLockInstance()
4769 self.needed_locks[locking.LEVEL_NODE] = []
4770 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4772 def DeclareLocks(self, level):
4773 if level == locking.LEVEL_NODE:
4774 self._LockInstancesNodes()
4776 def CheckPrereq(self):
4777 """Check prerequisites.
4779 This checks that the instance is in the cluster.
4782 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4783 assert self.instance is not None, \
4784 "Cannot retrieve locked instance %s" % self.op.instance_name
4785 _CheckNodeOnline(self, self.instance.primary_node)
4787 def Exec(self, feedback_fn):
4788 """Activate the disks.
4791 disks_ok, disks_info = \
4792 _AssembleInstanceDisks(self, self.instance,
4793 ignore_size=self.op.ignore_size)
4795 raise errors.OpExecError("Cannot activate block devices")
4800 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4802 """Prepare the block devices for an instance.
4804 This sets up the block devices on all nodes.
4806 @type lu: L{LogicalUnit}
4807 @param lu: the logical unit on whose behalf we execute
4808 @type instance: L{objects.Instance}
4809 @param instance: the instance for whose disks we assemble
4810 @type disks: list of L{objects.Disk} or None
4811 @param disks: which disks to assemble (or all, if None)
4812 @type ignore_secondaries: boolean
4813 @param ignore_secondaries: if true, errors on secondary nodes
4814 won't result in an error return from the function
4815 @type ignore_size: boolean
4816 @param ignore_size: if true, the current known size of the disk
4817 will not be used during the disk activation, useful for cases
4818 when the size is wrong
4819 @return: False if the operation failed, otherwise a list of
4820 (host, instance_visible_name, node_visible_name)
4821 with the mapping from node devices to instance devices
4826 iname = instance.name
4827 disks = _ExpandCheckDisks(instance, disks)
4829 # With the two passes mechanism we try to reduce the window of
4830 # opportunity for the race condition of switching DRBD to primary
4831 # before handshaking occured, but we do not eliminate it
4833 # The proper fix would be to wait (with some limits) until the
4834 # connection has been made and drbd transitions from WFConnection
4835 # into any other network-connected state (Connected, SyncTarget,
4838 # 1st pass, assemble on all nodes in secondary mode
4839 for idx, inst_disk in enumerate(disks):
4840 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4842 node_disk = node_disk.Copy()
4843 node_disk.UnsetSize()
4844 lu.cfg.SetDiskID(node_disk, node)
4845 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
4846 msg = result.fail_msg
4848 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4849 " (is_primary=False, pass=1): %s",
4850 inst_disk.iv_name, node, msg)
4851 if not ignore_secondaries:
4854 # FIXME: race condition on drbd migration to primary
4856 # 2nd pass, do only the primary node
4857 for idx, inst_disk in enumerate(disks):
4860 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4861 if node != instance.primary_node:
4864 node_disk = node_disk.Copy()
4865 node_disk.UnsetSize()
4866 lu.cfg.SetDiskID(node_disk, node)
4867 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
4868 msg = result.fail_msg
4870 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4871 " (is_primary=True, pass=2): %s",
4872 inst_disk.iv_name, node, msg)
4875 dev_path = result.payload
4877 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4879 # leave the disks configured for the primary node
4880 # this is a workaround that would be fixed better by
4881 # improving the logical/physical id handling
4883 lu.cfg.SetDiskID(disk, instance.primary_node)
4885 return disks_ok, device_info
4888 def _StartInstanceDisks(lu, instance, force):
4889 """Start the disks of an instance.
4892 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4893 ignore_secondaries=force)
4895 _ShutdownInstanceDisks(lu, instance)
4896 if force is not None and not force:
4897 lu.proc.LogWarning("", hint="If the message above refers to a"
4899 " you can retry the operation using '--force'.")
4900 raise errors.OpExecError("Disk consistency error")
4903 class LUInstanceDeactivateDisks(NoHooksLU):
4904 """Shutdown an instance's disks.
4909 def ExpandNames(self):
4910 self._ExpandAndLockInstance()
4911 self.needed_locks[locking.LEVEL_NODE] = []
4912 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4914 def DeclareLocks(self, level):
4915 if level == locking.LEVEL_NODE:
4916 self._LockInstancesNodes()
4918 def CheckPrereq(self):
4919 """Check prerequisites.
4921 This checks that the instance is in the cluster.
4924 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4925 assert self.instance is not None, \
4926 "Cannot retrieve locked instance %s" % self.op.instance_name
4928 def Exec(self, feedback_fn):
4929 """Deactivate the disks
4932 instance = self.instance
4934 _ShutdownInstanceDisks(self, instance)
4936 _SafeShutdownInstanceDisks(self, instance)
4939 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4940 """Shutdown block devices of an instance.
4942 This function checks if an instance is running, before calling
4943 _ShutdownInstanceDisks.
4946 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4947 _ShutdownInstanceDisks(lu, instance, disks=disks)
4950 def _ExpandCheckDisks(instance, disks):
4951 """Return the instance disks selected by the disks list
4953 @type disks: list of L{objects.Disk} or None
4954 @param disks: selected disks
4955 @rtype: list of L{objects.Disk}
4956 @return: selected instance disks to act on
4960 return instance.disks
4962 if not set(disks).issubset(instance.disks):
4963 raise errors.ProgrammerError("Can only act on disks belonging to the"
4968 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4969 """Shutdown block devices of an instance.
4971 This does the shutdown on all nodes of the instance.
4973 If the ignore_primary is false, errors on the primary node are
4978 disks = _ExpandCheckDisks(instance, disks)
4981 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4982 lu.cfg.SetDiskID(top_disk, node)
4983 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4984 msg = result.fail_msg
4986 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4987 disk.iv_name, node, msg)
4988 if ((node == instance.primary_node and not ignore_primary) or
4989 (node != instance.primary_node and not result.offline)):
4994 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4995 """Checks if a node has enough free memory.
4997 This function check if a given node has the needed amount of free
4998 memory. In case the node has less memory or we cannot get the
4999 information from the node, this function raise an OpPrereqError
5002 @type lu: C{LogicalUnit}
5003 @param lu: a logical unit from which we get configuration data
5005 @param node: the node to check
5006 @type reason: C{str}
5007 @param reason: string to use in the error message
5008 @type requested: C{int}
5009 @param requested: the amount of memory in MiB to check for
5010 @type hypervisor_name: C{str}
5011 @param hypervisor_name: the hypervisor to ask for memory stats
5012 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
5013 we cannot check the node
5016 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
5017 nodeinfo[node].Raise("Can't get data from node %s" % node,
5018 prereq=True, ecode=errors.ECODE_ENVIRON)
5019 free_mem = nodeinfo[node].payload.get('memory_free', None)
5020 if not isinstance(free_mem, int):
5021 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
5022 " was '%s'" % (node, free_mem),
5023 errors.ECODE_ENVIRON)
5024 if requested > free_mem:
5025 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
5026 " needed %s MiB, available %s MiB" %
5027 (node, reason, requested, free_mem),
5031 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5032 """Checks if nodes have enough free disk space in the all VGs.
5034 This function check if all given nodes have the needed amount of
5035 free disk. In case any node has less disk or we cannot get the
5036 information from the node, this function raise an OpPrereqError
5039 @type lu: C{LogicalUnit}
5040 @param lu: a logical unit from which we get configuration data
5041 @type nodenames: C{list}
5042 @param nodenames: the list of node names to check
5043 @type req_sizes: C{dict}
5044 @param req_sizes: the hash of vg and corresponding amount of disk in
5046 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5047 or we cannot check the node
5050 for vg, req_size in req_sizes.items():
5051 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5054 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5055 """Checks if nodes have enough free disk space in the specified VG.
5057 This function check if all given nodes have the needed amount of
5058 free disk. In case any node has less disk or we cannot get the
5059 information from the node, this function raise an OpPrereqError
5062 @type lu: C{LogicalUnit}
5063 @param lu: a logical unit from which we get configuration data
5064 @type nodenames: C{list}
5065 @param nodenames: the list of node names to check
5067 @param vg: the volume group to check
5068 @type requested: C{int}
5069 @param requested: the amount of disk in MiB to check for
5070 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5071 or we cannot check the node
5074 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
5075 for node in nodenames:
5076 info = nodeinfo[node]
5077 info.Raise("Cannot get current information from node %s" % node,
5078 prereq=True, ecode=errors.ECODE_ENVIRON)
5079 vg_free = info.payload.get("vg_free", None)
5080 if not isinstance(vg_free, int):
5081 raise errors.OpPrereqError("Can't compute free disk space on node"
5082 " %s for vg %s, result was '%s'" %
5083 (node, vg, vg_free), errors.ECODE_ENVIRON)
5084 if requested > vg_free:
5085 raise errors.OpPrereqError("Not enough disk space on target node %s"
5086 " vg %s: required %d MiB, available %d MiB" %
5087 (node, vg, requested, vg_free),
5091 class LUInstanceStartup(LogicalUnit):
5092 """Starts an instance.
5095 HPATH = "instance-start"
5096 HTYPE = constants.HTYPE_INSTANCE
5099 def CheckArguments(self):
5101 if self.op.beparams:
5102 # fill the beparams dict
5103 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5105 def ExpandNames(self):
5106 self._ExpandAndLockInstance()
5108 def BuildHooksEnv(self):
5111 This runs on master, primary and secondary nodes of the instance.
5115 "FORCE": self.op.force,
5117 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5118 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5121 def CheckPrereq(self):
5122 """Check prerequisites.
5124 This checks that the instance is in the cluster.
5127 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5128 assert self.instance is not None, \
5129 "Cannot retrieve locked instance %s" % self.op.instance_name
5132 if self.op.hvparams:
5133 # check hypervisor parameter syntax (locally)
5134 cluster = self.cfg.GetClusterInfo()
5135 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5136 filled_hvp = cluster.FillHV(instance)
5137 filled_hvp.update(self.op.hvparams)
5138 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5139 hv_type.CheckParameterSyntax(filled_hvp)
5140 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5142 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5144 if self.primary_offline and self.op.ignore_offline_nodes:
5145 self.proc.LogWarning("Ignoring offline primary node")
5147 if self.op.hvparams or self.op.beparams:
5148 self.proc.LogWarning("Overridden parameters are ignored")
5150 _CheckNodeOnline(self, instance.primary_node)
5152 bep = self.cfg.GetClusterInfo().FillBE(instance)
5154 # check bridges existence
5155 _CheckInstanceBridgesExist(self, instance)
5157 remote_info = self.rpc.call_instance_info(instance.primary_node,
5159 instance.hypervisor)
5160 remote_info.Raise("Error checking node %s" % instance.primary_node,
5161 prereq=True, ecode=errors.ECODE_ENVIRON)
5162 if not remote_info.payload: # not running already
5163 _CheckNodeFreeMemory(self, instance.primary_node,
5164 "starting instance %s" % instance.name,
5165 bep[constants.BE_MEMORY], instance.hypervisor)
5167 def Exec(self, feedback_fn):
5168 """Start the instance.
5171 instance = self.instance
5172 force = self.op.force
5174 self.cfg.MarkInstanceUp(instance.name)
5176 if self.primary_offline:
5177 assert self.op.ignore_offline_nodes
5178 self.proc.LogInfo("Primary node offline, marked instance as started")
5180 node_current = instance.primary_node
5182 _StartInstanceDisks(self, instance, force)
5184 result = self.rpc.call_instance_start(node_current, instance,
5185 self.op.hvparams, self.op.beparams)
5186 msg = result.fail_msg
5188 _ShutdownInstanceDisks(self, instance)
5189 raise errors.OpExecError("Could not start instance: %s" % msg)
5192 class LUInstanceReboot(LogicalUnit):
5193 """Reboot an instance.
5196 HPATH = "instance-reboot"
5197 HTYPE = constants.HTYPE_INSTANCE
5200 def ExpandNames(self):
5201 self._ExpandAndLockInstance()
5203 def BuildHooksEnv(self):
5206 This runs on master, primary and secondary nodes of the instance.
5210 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
5211 "REBOOT_TYPE": self.op.reboot_type,
5212 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5214 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5215 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5218 def CheckPrereq(self):
5219 """Check prerequisites.
5221 This checks that the instance is in the cluster.
5224 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5225 assert self.instance is not None, \
5226 "Cannot retrieve locked instance %s" % self.op.instance_name
5228 _CheckNodeOnline(self, instance.primary_node)
5230 # check bridges existence
5231 _CheckInstanceBridgesExist(self, instance)
5233 def Exec(self, feedback_fn):
5234 """Reboot the instance.
5237 instance = self.instance
5238 ignore_secondaries = self.op.ignore_secondaries
5239 reboot_type = self.op.reboot_type
5241 remote_info = self.rpc.call_instance_info(instance.primary_node,
5243 instance.hypervisor)
5244 remote_info.Raise("Error checking node %s" % instance.primary_node)
5245 instance_running = bool(remote_info.payload)
5247 node_current = instance.primary_node
5249 if instance_running and reboot_type in [constants.INSTANCE_REBOOT_SOFT,
5250 constants.INSTANCE_REBOOT_HARD]:
5251 for disk in instance.disks:
5252 self.cfg.SetDiskID(disk, node_current)
5253 result = self.rpc.call_instance_reboot(node_current, instance,
5255 self.op.shutdown_timeout)
5256 result.Raise("Could not reboot instance")
5258 if instance_running:
5259 result = self.rpc.call_instance_shutdown(node_current, instance,
5260 self.op.shutdown_timeout)
5261 result.Raise("Could not shutdown instance for full reboot")
5262 _ShutdownInstanceDisks(self, instance)
5264 self.LogInfo("Instance %s was already stopped, starting now",
5266 _StartInstanceDisks(self, instance, ignore_secondaries)
5267 result = self.rpc.call_instance_start(node_current, instance, None, None)
5268 msg = result.fail_msg
5270 _ShutdownInstanceDisks(self, instance)
5271 raise errors.OpExecError("Could not start instance for"
5272 " full reboot: %s" % msg)
5274 self.cfg.MarkInstanceUp(instance.name)
5277 class LUInstanceShutdown(LogicalUnit):
5278 """Shutdown an instance.
5281 HPATH = "instance-stop"
5282 HTYPE = constants.HTYPE_INSTANCE
5285 def ExpandNames(self):
5286 self._ExpandAndLockInstance()
5288 def BuildHooksEnv(self):
5291 This runs on master, primary and secondary nodes of the instance.
5294 env = _BuildInstanceHookEnvByObject(self, self.instance)
5295 env["TIMEOUT"] = self.op.timeout
5296 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5299 def CheckPrereq(self):
5300 """Check prerequisites.
5302 This checks that the instance is in the cluster.
5305 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5306 assert self.instance is not None, \
5307 "Cannot retrieve locked instance %s" % self.op.instance_name
5309 self.primary_offline = \
5310 self.cfg.GetNodeInfo(self.instance.primary_node).offline
5312 if self.primary_offline and self.op.ignore_offline_nodes:
5313 self.proc.LogWarning("Ignoring offline primary node")
5315 _CheckNodeOnline(self, self.instance.primary_node)
5317 def Exec(self, feedback_fn):
5318 """Shutdown the instance.
5321 instance = self.instance
5322 node_current = instance.primary_node
5323 timeout = self.op.timeout
5325 self.cfg.MarkInstanceDown(instance.name)
5327 if self.primary_offline:
5328 assert self.op.ignore_offline_nodes
5329 self.proc.LogInfo("Primary node offline, marked instance as stopped")
5331 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
5332 msg = result.fail_msg
5334 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
5336 _ShutdownInstanceDisks(self, instance)
5339 class LUInstanceReinstall(LogicalUnit):
5340 """Reinstall an instance.
5343 HPATH = "instance-reinstall"
5344 HTYPE = constants.HTYPE_INSTANCE
5347 def ExpandNames(self):
5348 self._ExpandAndLockInstance()
5350 def BuildHooksEnv(self):
5353 This runs on master, primary and secondary nodes of the instance.
5356 env = _BuildInstanceHookEnvByObject(self, self.instance)
5357 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5360 def CheckPrereq(self):
5361 """Check prerequisites.
5363 This checks that the instance is in the cluster and is not running.
5366 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5367 assert instance is not None, \
5368 "Cannot retrieve locked instance %s" % self.op.instance_name
5369 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
5370 " offline, cannot reinstall")
5371 for node in instance.secondary_nodes:
5372 _CheckNodeOnline(self, node, "Instance secondary node offline,"
5373 " cannot reinstall")
5375 if instance.disk_template == constants.DT_DISKLESS:
5376 raise errors.OpPrereqError("Instance '%s' has no disks" %
5377 self.op.instance_name,
5379 _CheckInstanceDown(self, instance, "cannot reinstall")
5381 if self.op.os_type is not None:
5383 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
5384 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
5385 instance_os = self.op.os_type
5387 instance_os = instance.os
5389 nodelist = list(instance.all_nodes)
5391 if self.op.osparams:
5392 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
5393 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
5394 self.os_inst = i_osdict # the new dict (without defaults)
5398 self.instance = instance
5400 def Exec(self, feedback_fn):
5401 """Reinstall the instance.
5404 inst = self.instance
5406 if self.op.os_type is not None:
5407 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
5408 inst.os = self.op.os_type
5409 # Write to configuration
5410 self.cfg.Update(inst, feedback_fn)
5412 _StartInstanceDisks(self, inst, None)
5414 feedback_fn("Running the instance OS create scripts...")
5415 # FIXME: pass debug option from opcode to backend
5416 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
5417 self.op.debug_level,
5418 osparams=self.os_inst)
5419 result.Raise("Could not install OS for instance %s on node %s" %
5420 (inst.name, inst.primary_node))
5422 _ShutdownInstanceDisks(self, inst)
5425 class LUInstanceRecreateDisks(LogicalUnit):
5426 """Recreate an instance's missing disks.
5429 HPATH = "instance-recreate-disks"
5430 HTYPE = constants.HTYPE_INSTANCE
5433 def ExpandNames(self):
5434 self._ExpandAndLockInstance()
5436 def BuildHooksEnv(self):
5439 This runs on master, primary and secondary nodes of the instance.
5442 env = _BuildInstanceHookEnvByObject(self, self.instance)
5443 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5446 def CheckPrereq(self):
5447 """Check prerequisites.
5449 This checks that the instance is in the cluster and is not running.
5452 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5453 assert instance is not None, \
5454 "Cannot retrieve locked instance %s" % self.op.instance_name
5455 _CheckNodeOnline(self, instance.primary_node)
5457 if instance.disk_template == constants.DT_DISKLESS:
5458 raise errors.OpPrereqError("Instance '%s' has no disks" %
5459 self.op.instance_name, errors.ECODE_INVAL)
5460 _CheckInstanceDown(self, instance, "cannot recreate disks")
5462 if not self.op.disks:
5463 self.op.disks = range(len(instance.disks))
5465 for idx in self.op.disks:
5466 if idx >= len(instance.disks):
5467 raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
5470 self.instance = instance
5472 def Exec(self, feedback_fn):
5473 """Recreate the disks.
5477 for idx, _ in enumerate(self.instance.disks):
5478 if idx not in self.op.disks: # disk idx has not been passed in
5482 _CreateDisks(self, self.instance, to_skip=to_skip)
5485 class LUInstanceRename(LogicalUnit):
5486 """Rename an instance.
5489 HPATH = "instance-rename"
5490 HTYPE = constants.HTYPE_INSTANCE
5492 def CheckArguments(self):
5496 if self.op.ip_check and not self.op.name_check:
5497 # TODO: make the ip check more flexible and not depend on the name check
5498 raise errors.OpPrereqError("Cannot do ip check without a name check",
5501 def BuildHooksEnv(self):
5504 This runs on master, primary and secondary nodes of the instance.
5507 env = _BuildInstanceHookEnvByObject(self, self.instance)
5508 env["INSTANCE_NEW_NAME"] = self.op.new_name
5509 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5512 def CheckPrereq(self):
5513 """Check prerequisites.
5515 This checks that the instance is in the cluster and is not running.
5518 self.op.instance_name = _ExpandInstanceName(self.cfg,
5519 self.op.instance_name)
5520 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5521 assert instance is not None
5522 _CheckNodeOnline(self, instance.primary_node)
5523 _CheckInstanceDown(self, instance, "cannot rename")
5524 self.instance = instance
5526 new_name = self.op.new_name
5527 if self.op.name_check:
5528 hostname = netutils.GetHostname(name=new_name)
5529 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
5531 if not utils.MatchNameComponent(self.op.new_name, [hostname.name]):
5532 raise errors.OpPrereqError(("Resolved hostname '%s' does not look the"
5533 " same as given hostname '%s'") %
5534 (hostname.name, self.op.new_name),
5536 new_name = self.op.new_name = hostname.name
5537 if (self.op.ip_check and
5538 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
5539 raise errors.OpPrereqError("IP %s of instance %s already in use" %
5540 (hostname.ip, new_name),
5541 errors.ECODE_NOTUNIQUE)
5543 instance_list = self.cfg.GetInstanceList()
5544 if new_name in instance_list and new_name != instance.name:
5545 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5546 new_name, errors.ECODE_EXISTS)
5548 def Exec(self, feedback_fn):
5549 """Rename the instance.
5552 inst = self.instance
5553 old_name = inst.name
5555 rename_file_storage = False
5556 if (inst.disk_template in (constants.DT_FILE, constants.DT_SHARED_FILE) and
5557 self.op.new_name != inst.name):
5558 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5559 rename_file_storage = True
5561 self.cfg.RenameInstance(inst.name, self.op.new_name)
5562 # Change the instance lock. This is definitely safe while we hold the BGL
5563 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5564 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5566 # re-read the instance from the configuration after rename
5567 inst = self.cfg.GetInstanceInfo(self.op.new_name)
5569 if rename_file_storage:
5570 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5571 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5572 old_file_storage_dir,
5573 new_file_storage_dir)
5574 result.Raise("Could not rename on node %s directory '%s' to '%s'"
5575 " (but the instance has been renamed in Ganeti)" %
5576 (inst.primary_node, old_file_storage_dir,
5577 new_file_storage_dir))
5579 _StartInstanceDisks(self, inst, None)
5581 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5582 old_name, self.op.debug_level)
5583 msg = result.fail_msg
5585 msg = ("Could not run OS rename script for instance %s on node %s"
5586 " (but the instance has been renamed in Ganeti): %s" %
5587 (inst.name, inst.primary_node, msg))
5588 self.proc.LogWarning(msg)
5590 _ShutdownInstanceDisks(self, inst)
5595 class LUInstanceRemove(LogicalUnit):
5596 """Remove an instance.
5599 HPATH = "instance-remove"
5600 HTYPE = constants.HTYPE_INSTANCE
5603 def ExpandNames(self):
5604 self._ExpandAndLockInstance()
5605 self.needed_locks[locking.LEVEL_NODE] = []
5606 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5608 def DeclareLocks(self, level):
5609 if level == locking.LEVEL_NODE:
5610 self._LockInstancesNodes()
5612 def BuildHooksEnv(self):
5615 This runs on master, primary and secondary nodes of the instance.
5618 env = _BuildInstanceHookEnvByObject(self, self.instance)
5619 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5620 nl = [self.cfg.GetMasterNode()]
5621 nl_post = list(self.instance.all_nodes) + nl
5622 return env, nl, nl_post
5624 def CheckPrereq(self):
5625 """Check prerequisites.
5627 This checks that the instance is in the cluster.
5630 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5631 assert self.instance is not None, \
5632 "Cannot retrieve locked instance %s" % self.op.instance_name
5634 def Exec(self, feedback_fn):
5635 """Remove the instance.
5638 instance = self.instance
5639 logging.info("Shutting down instance %s on node %s",
5640 instance.name, instance.primary_node)
5642 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5643 self.op.shutdown_timeout)
5644 msg = result.fail_msg
5646 if self.op.ignore_failures:
5647 feedback_fn("Warning: can't shutdown instance: %s" % msg)
5649 raise errors.OpExecError("Could not shutdown instance %s on"
5651 (instance.name, instance.primary_node, msg))
5653 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5656 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5657 """Utility function to remove an instance.
5660 logging.info("Removing block devices for instance %s", instance.name)
5662 if not _RemoveDisks(lu, instance):
5663 if not ignore_failures:
5664 raise errors.OpExecError("Can't remove instance's disks")
5665 feedback_fn("Warning: can't remove instance's disks")
5667 logging.info("Removing instance %s out of cluster config", instance.name)
5669 lu.cfg.RemoveInstance(instance.name)
5671 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5672 "Instance lock removal conflict"
5674 # Remove lock for the instance
5675 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5678 class LUInstanceQuery(NoHooksLU):
5679 """Logical unit for querying instances.
5682 # pylint: disable-msg=W0142
5685 def CheckArguments(self):
5686 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
5687 self.op.output_fields, self.op.use_locking)
5689 def ExpandNames(self):
5690 self.iq.ExpandNames(self)
5692 def DeclareLocks(self, level):
5693 self.iq.DeclareLocks(self, level)
5695 def Exec(self, feedback_fn):
5696 return self.iq.OldStyleQuery(self)
5699 class LUInstanceFailover(LogicalUnit):
5700 """Failover an instance.
5703 HPATH = "instance-failover"
5704 HTYPE = constants.HTYPE_INSTANCE
5707 def ExpandNames(self):
5708 self._ExpandAndLockInstance()
5709 self.needed_locks[locking.LEVEL_NODE] = []
5710 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5712 def DeclareLocks(self, level):
5713 if level == locking.LEVEL_NODE:
5714 self._LockInstancesNodes()
5716 def BuildHooksEnv(self):
5719 This runs on master, primary and secondary nodes of the instance.
5722 instance = self.instance
5723 source_node = instance.primary_node
5724 target_node = instance.secondary_nodes[0]
5726 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5727 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5728 "OLD_PRIMARY": source_node,
5729 "OLD_SECONDARY": target_node,
5730 "NEW_PRIMARY": target_node,
5731 "NEW_SECONDARY": source_node,
5733 env.update(_BuildInstanceHookEnvByObject(self, instance))
5734 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5736 nl_post.append(source_node)
5737 return env, nl, nl_post
5739 def CheckPrereq(self):
5740 """Check prerequisites.
5742 This checks that the instance is in the cluster.
5745 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5746 assert self.instance is not None, \
5747 "Cannot retrieve locked instance %s" % self.op.instance_name
5749 bep = self.cfg.GetClusterInfo().FillBE(instance)
5750 if instance.disk_template not in constants.DTS_NET_MIRROR:
5751 raise errors.OpPrereqError("Instance's disk layout is not"
5752 " network mirrored, cannot failover.",
5755 secondary_nodes = instance.secondary_nodes
5756 if not secondary_nodes:
5757 raise errors.ProgrammerError("no secondary node but using "
5758 "a mirrored disk template")
5760 target_node = secondary_nodes[0]
5761 _CheckNodeOnline(self, target_node)
5762 _CheckNodeNotDrained(self, target_node)
5763 if instance.admin_up:
5764 # check memory requirements on the secondary node
5765 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5766 instance.name, bep[constants.BE_MEMORY],
5767 instance.hypervisor)
5769 self.LogInfo("Not checking memory on the secondary node as"
5770 " instance will not be started")
5772 # check bridge existance
5773 _CheckInstanceBridgesExist(self, instance, node=target_node)
5775 def Exec(self, feedback_fn):
5776 """Failover an instance.
5778 The failover is done by shutting it down on its present node and
5779 starting it on the secondary.
5782 instance = self.instance
5783 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
5785 source_node = instance.primary_node
5786 target_node = instance.secondary_nodes[0]
5788 if instance.admin_up:
5789 feedback_fn("* checking disk consistency between source and target")
5790 for dev in instance.disks:
5791 # for drbd, these are drbd over lvm
5792 if not _CheckDiskConsistency(self, dev, target_node, False):
5793 if not self.op.ignore_consistency:
5794 raise errors.OpExecError("Disk %s is degraded on target node,"
5795 " aborting failover." % dev.iv_name)
5797 feedback_fn("* not checking disk consistency as instance is not running")
5799 feedback_fn("* shutting down instance on source node")
5800 logging.info("Shutting down instance %s on node %s",
5801 instance.name, source_node)
5803 result = self.rpc.call_instance_shutdown(source_node, instance,
5804 self.op.shutdown_timeout)
5805 msg = result.fail_msg
5807 if self.op.ignore_consistency or primary_node.offline:
5808 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5809 " Proceeding anyway. Please make sure node"
5810 " %s is down. Error details: %s",
5811 instance.name, source_node, source_node, msg)
5813 raise errors.OpExecError("Could not shutdown instance %s on"
5815 (instance.name, source_node, msg))
5817 feedback_fn("* deactivating the instance's disks on source node")
5818 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5819 raise errors.OpExecError("Can't shut down the instance's disks.")
5821 instance.primary_node = target_node
5822 # distribute new instance config to the other nodes
5823 self.cfg.Update(instance, feedback_fn)
5825 # Only start the instance if it's marked as up
5826 if instance.admin_up:
5827 feedback_fn("* activating the instance's disks on target node")
5828 logging.info("Starting instance %s on node %s",
5829 instance.name, target_node)
5831 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5832 ignore_secondaries=True)
5834 _ShutdownInstanceDisks(self, instance)
5835 raise errors.OpExecError("Can't activate the instance's disks")
5837 feedback_fn("* starting the instance on the target node")
5838 result = self.rpc.call_instance_start(target_node, instance, None, None)
5839 msg = result.fail_msg
5841 _ShutdownInstanceDisks(self, instance)
5842 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5843 (instance.name, target_node, msg))
5846 class LUInstanceMigrate(LogicalUnit):
5847 """Migrate an instance.
5849 This is migration without shutting down, compared to the failover,
5850 which is done with shutdown.
5853 HPATH = "instance-migrate"
5854 HTYPE = constants.HTYPE_INSTANCE
5857 def CheckArguments(self):
5858 _CheckIAllocatorOrNode(self, "iallocator", "target_node")
5860 def ExpandNames(self):
5861 self._ExpandAndLockInstance()
5863 if self.op.target_node is not None:
5864 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5866 self.needed_locks[locking.LEVEL_NODE] = []
5867 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5869 self._migrater = TLMigrateInstance(self, self.op.instance_name,
5870 self.op.cleanup, self.op.iallocator,
5871 self.op.target_node)
5872 self.tasklets = [self._migrater]
5874 def DeclareLocks(self, level):
5875 if level == locking.LEVEL_NODE:
5876 instance = self.context.cfg.GetInstanceInfo(self.op.instance_name)
5877 if instance.disk_template in constants.DTS_EXT_MIRROR:
5878 if self.op.target_node is None:
5879 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5881 self.needed_locks[locking.LEVEL_NODE] = [instance.primary_node,
5882 self.op.target_node]
5883 del self.recalculate_locks[locking.LEVEL_NODE]
5885 self._LockInstancesNodes()
5887 def BuildHooksEnv(self):
5890 This runs on master, primary and secondary nodes of the instance.
5893 instance = self._migrater.instance
5894 source_node = instance.primary_node
5895 target_node = self._migrater.target_node
5896 env = _BuildInstanceHookEnvByObject(self, instance)
5897 env["MIGRATE_LIVE"] = self._migrater.live
5898 env["MIGRATE_CLEANUP"] = self.op.cleanup
5900 "OLD_PRIMARY": source_node,
5901 "NEW_PRIMARY": target_node,
5904 if instance.disk_template in constants.DTS_NET_MIRROR:
5905 env["OLD_SECONDARY"] = target_node
5906 env["NEW_SECONDARY"] = source_node
5908 env["OLD_SECONDARY"] = env["NEW_SECONDARY"] = None
5910 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5912 nl_post.append(source_node)
5913 return env, nl, nl_post
5916 class LUInstanceMove(LogicalUnit):
5917 """Move an instance by data-copying.
5920 HPATH = "instance-move"
5921 HTYPE = constants.HTYPE_INSTANCE
5924 def ExpandNames(self):
5925 self._ExpandAndLockInstance()
5926 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5927 self.op.target_node = target_node
5928 self.needed_locks[locking.LEVEL_NODE] = [target_node]
5929 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5931 def DeclareLocks(self, level):
5932 if level == locking.LEVEL_NODE:
5933 self._LockInstancesNodes(primary_only=True)
5935 def BuildHooksEnv(self):
5938 This runs on master, primary and secondary nodes of the instance.
5942 "TARGET_NODE": self.op.target_node,
5943 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5945 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5946 nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5947 self.op.target_node]
5950 def CheckPrereq(self):
5951 """Check prerequisites.
5953 This checks that the instance is in the cluster.
5956 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5957 assert self.instance is not None, \
5958 "Cannot retrieve locked instance %s" % self.op.instance_name
5960 node = self.cfg.GetNodeInfo(self.op.target_node)
5961 assert node is not None, \
5962 "Cannot retrieve locked node %s" % self.op.target_node
5964 self.target_node = target_node = node.name
5966 if target_node == instance.primary_node:
5967 raise errors.OpPrereqError("Instance %s is already on the node %s" %
5968 (instance.name, target_node),
5971 bep = self.cfg.GetClusterInfo().FillBE(instance)
5973 for idx, dsk in enumerate(instance.disks):
5974 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5975 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5976 " cannot copy" % idx, errors.ECODE_STATE)
5978 _CheckNodeOnline(self, target_node)
5979 _CheckNodeNotDrained(self, target_node)
5980 _CheckNodeVmCapable(self, target_node)
5982 if instance.admin_up:
5983 # check memory requirements on the secondary node
5984 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5985 instance.name, bep[constants.BE_MEMORY],
5986 instance.hypervisor)
5988 self.LogInfo("Not checking memory on the secondary node as"
5989 " instance will not be started")
5991 # check bridge existance
5992 _CheckInstanceBridgesExist(self, instance, node=target_node)
5994 def Exec(self, feedback_fn):
5995 """Move an instance.
5997 The move is done by shutting it down on its present node, copying
5998 the data over (slow) and starting it on the new node.
6001 instance = self.instance
6003 source_node = instance.primary_node
6004 target_node = self.target_node
6006 self.LogInfo("Shutting down instance %s on source node %s",
6007 instance.name, source_node)
6009 result = self.rpc.call_instance_shutdown(source_node, instance,
6010 self.op.shutdown_timeout)
6011 msg = result.fail_msg
6013 if self.op.ignore_consistency:
6014 self.proc.LogWarning("Could not shutdown instance %s on node %s."
6015 " Proceeding anyway. Please make sure node"
6016 " %s is down. Error details: %s",
6017 instance.name, source_node, source_node, msg)
6019 raise errors.OpExecError("Could not shutdown instance %s on"
6021 (instance.name, source_node, msg))
6023 # create the target disks
6025 _CreateDisks(self, instance, target_node=target_node)
6026 except errors.OpExecError:
6027 self.LogWarning("Device creation failed, reverting...")
6029 _RemoveDisks(self, instance, target_node=target_node)
6031 self.cfg.ReleaseDRBDMinors(instance.name)
6034 cluster_name = self.cfg.GetClusterInfo().cluster_name
6037 # activate, get path, copy the data over
6038 for idx, disk in enumerate(instance.disks):
6039 self.LogInfo("Copying data for disk %d", idx)
6040 result = self.rpc.call_blockdev_assemble(target_node, disk,
6041 instance.name, True, idx)
6043 self.LogWarning("Can't assemble newly created disk %d: %s",
6044 idx, result.fail_msg)
6045 errs.append(result.fail_msg)
6047 dev_path = result.payload
6048 result = self.rpc.call_blockdev_export(source_node, disk,
6049 target_node, dev_path,
6052 self.LogWarning("Can't copy data over for disk %d: %s",
6053 idx, result.fail_msg)
6054 errs.append(result.fail_msg)
6058 self.LogWarning("Some disks failed to copy, aborting")
6060 _RemoveDisks(self, instance, target_node=target_node)
6062 self.cfg.ReleaseDRBDMinors(instance.name)
6063 raise errors.OpExecError("Errors during disk copy: %s" %
6066 instance.primary_node = target_node
6067 self.cfg.Update(instance, feedback_fn)
6069 self.LogInfo("Removing the disks on the original node")
6070 _RemoveDisks(self, instance, target_node=source_node)
6072 # Only start the instance if it's marked as up
6073 if instance.admin_up:
6074 self.LogInfo("Starting instance %s on node %s",
6075 instance.name, target_node)
6077 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6078 ignore_secondaries=True)
6080 _ShutdownInstanceDisks(self, instance)
6081 raise errors.OpExecError("Can't activate the instance's disks")
6083 result = self.rpc.call_instance_start(target_node, instance, None, None)
6084 msg = result.fail_msg
6086 _ShutdownInstanceDisks(self, instance)
6087 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6088 (instance.name, target_node, msg))
6091 class LUNodeMigrate(LogicalUnit):
6092 """Migrate all instances from a node.
6095 HPATH = "node-migrate"
6096 HTYPE = constants.HTYPE_NODE
6099 def CheckArguments(self):
6100 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
6102 def ExpandNames(self):
6103 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6105 self.needed_locks = {}
6107 # Create tasklets for migrating instances for all instances on this node
6111 self.lock_all_nodes = False
6113 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
6114 logging.debug("Migrating instance %s", inst.name)
6115 names.append(inst.name)
6117 tasklets.append(TLMigrateInstance(self, inst.name, False,
6118 self.op.iallocator, None))
6120 if inst.disk_template in constants.DTS_EXT_MIRROR:
6121 # We need to lock all nodes, as the iallocator will choose the
6122 # destination nodes afterwards
6123 self.lock_all_nodes = True
6125 self.tasklets = tasklets
6127 # Declare node locks
6128 if self.lock_all_nodes:
6129 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6131 self.needed_locks[locking.LEVEL_NODE] = [self.op.node_name]
6132 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6134 # Declare instance locks
6135 self.needed_locks[locking.LEVEL_INSTANCE] = names
6137 def DeclareLocks(self, level):
6138 if level == locking.LEVEL_NODE and not self.lock_all_nodes:
6139 self._LockInstancesNodes()
6141 def BuildHooksEnv(self):
6144 This runs on the master, the primary and all the secondaries.
6148 "NODE_NAME": self.op.node_name,
6151 nl = [self.cfg.GetMasterNode()]
6153 return (env, nl, nl)
6156 class TLMigrateInstance(Tasklet):
6157 """Tasklet class for instance migration.
6160 @ivar live: whether the migration will be done live or non-live;
6161 this variable is initalized only after CheckPrereq has run
6164 def __init__(self, lu, instance_name, cleanup,
6165 iallocator=None, target_node=None):
6166 """Initializes this class.
6169 Tasklet.__init__(self, lu)
6172 self.instance_name = instance_name
6173 self.cleanup = cleanup
6174 self.live = False # will be overridden later
6175 self.iallocator = iallocator
6176 self.target_node = target_node
6178 def CheckPrereq(self):
6179 """Check prerequisites.
6181 This checks that the instance is in the cluster.
6184 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
6185 instance = self.cfg.GetInstanceInfo(instance_name)
6186 assert instance is not None
6187 self.instance = instance
6189 if instance.disk_template not in constants.DTS_MIRRORED:
6190 raise errors.OpPrereqError("Instance's disk layout '%s' does not allow"
6191 " migrations" % instance.disk_template,
6194 if instance.disk_template in constants.DTS_EXT_MIRROR:
6195 if [self.iallocator, self.target_node].count(None) != 1:
6196 raise errors.OpPrereqError("Do not specify both, iallocator and"
6197 " target node", errors.ECODE_INVAL)
6200 self._RunAllocator()
6202 # self.target_node is already populated, either directly or by the
6204 target_node = self.target_node
6206 if len(self.lu.tasklets) == 1:
6207 # It is safe to remove locks only when we're the only tasklet in the LU
6208 nodes_keep = [instance.primary_node, self.target_node]
6209 nodes_rel = [node for node in self.lu.acquired_locks[locking.LEVEL_NODE]
6210 if node not in nodes_keep]
6211 self.lu.context.glm.release(locking.LEVEL_NODE, nodes_rel)
6212 self.lu.acquired_locks[locking.LEVEL_NODE] = nodes_keep
6215 secondary_nodes = instance.secondary_nodes
6216 if not secondary_nodes:
6217 raise errors.ConfigurationError("No secondary node but using"
6218 " %s disk template" %
6219 instance.disk_template)
6220 target_node = secondary_nodes[0]
6222 i_be = self.cfg.GetClusterInfo().FillBE(instance)
6224 # check memory requirements on the secondary node
6225 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
6226 instance.name, i_be[constants.BE_MEMORY],
6227 instance.hypervisor)
6229 # check bridge existance
6230 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
6232 if not self.cleanup:
6233 _CheckNodeNotDrained(self.lu, target_node)
6234 result = self.rpc.call_instance_migratable(instance.primary_node,
6236 result.Raise("Can't migrate, please use failover",
6237 prereq=True, ecode=errors.ECODE_STATE)
6240 def _RunAllocator(self):
6241 """Run the allocator based on input opcode.
6244 ial = IAllocator(self.cfg, self.rpc,
6245 mode=constants.IALLOCATOR_MODE_RELOC,
6246 name=self.instance_name,
6247 # TODO See why hail breaks with a single node below
6248 relocate_from=[self.instance.primary_node,
6249 self.instance.primary_node],
6252 ial.Run(self.iallocator)
6255 raise errors.OpPrereqError("Can't compute nodes using"
6256 " iallocator '%s': %s" %
6257 (self.iallocator, ial.info),
6259 if len(ial.result) != ial.required_nodes:
6260 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6261 " of nodes (%s), required %s" %
6262 (self.iallocator, len(ial.result),
6263 ial.required_nodes), errors.ECODE_FAULT)
6264 self.target_node = ial.result[0]
6265 self.lu.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6266 self.instance_name, self.iallocator,
6267 utils.CommaJoin(ial.result))
6269 if self.lu.op.live is not None and self.lu.op.mode is not None:
6270 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
6271 " parameters are accepted",
6273 if self.lu.op.live is not None:
6275 self.lu.op.mode = constants.HT_MIGRATION_LIVE
6277 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
6278 # reset the 'live' parameter to None so that repeated
6279 # invocations of CheckPrereq do not raise an exception
6280 self.lu.op.live = None
6281 elif self.lu.op.mode is None:
6282 # read the default value from the hypervisor
6283 i_hv = self.cfg.GetClusterInfo().FillHV(self.instance, skip_globals=False)
6284 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
6286 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
6288 def _WaitUntilSync(self):
6289 """Poll with custom rpc for disk sync.
6291 This uses our own step-based rpc call.
6294 self.feedback_fn("* wait until resync is done")
6298 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
6300 self.instance.disks)
6302 for node, nres in result.items():
6303 nres.Raise("Cannot resync disks on node %s" % node)
6304 node_done, node_percent = nres.payload
6305 all_done = all_done and node_done
6306 if node_percent is not None:
6307 min_percent = min(min_percent, node_percent)
6309 if min_percent < 100:
6310 self.feedback_fn(" - progress: %.1f%%" % min_percent)
6313 def _EnsureSecondary(self, node):
6314 """Demote a node to secondary.
6317 self.feedback_fn("* switching node %s to secondary mode" % node)
6319 for dev in self.instance.disks:
6320 self.cfg.SetDiskID(dev, node)
6322 result = self.rpc.call_blockdev_close(node, self.instance.name,
6323 self.instance.disks)
6324 result.Raise("Cannot change disk to secondary on node %s" % node)
6326 def _GoStandalone(self):
6327 """Disconnect from the network.
6330 self.feedback_fn("* changing into standalone mode")
6331 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
6332 self.instance.disks)
6333 for node, nres in result.items():
6334 nres.Raise("Cannot disconnect disks node %s" % node)
6336 def _GoReconnect(self, multimaster):
6337 """Reconnect to the network.
6343 msg = "single-master"
6344 self.feedback_fn("* changing disks into %s mode" % msg)
6345 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
6346 self.instance.disks,
6347 self.instance.name, multimaster)
6348 for node, nres in result.items():
6349 nres.Raise("Cannot change disks config on node %s" % node)
6351 def _ExecCleanup(self):
6352 """Try to cleanup after a failed migration.
6354 The cleanup is done by:
6355 - check that the instance is running only on one node
6356 (and update the config if needed)
6357 - change disks on its secondary node to secondary
6358 - wait until disks are fully synchronized
6359 - disconnect from the network
6360 - change disks into single-master mode
6361 - wait again until disks are fully synchronized
6364 instance = self.instance
6365 target_node = self.target_node
6366 source_node = self.source_node
6368 # check running on only one node
6369 self.feedback_fn("* checking where the instance actually runs"
6370 " (if this hangs, the hypervisor might be in"
6372 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6373 for node, result in ins_l.items():
6374 result.Raise("Can't contact node %s" % node)
6376 runningon_source = instance.name in ins_l[source_node].payload
6377 runningon_target = instance.name in ins_l[target_node].payload
6379 if runningon_source and runningon_target:
6380 raise errors.OpExecError("Instance seems to be running on two nodes,"
6381 " or the hypervisor is confused. You will have"
6382 " to ensure manually that it runs only on one"
6383 " and restart this operation.")
6385 if not (runningon_source or runningon_target):
6386 raise errors.OpExecError("Instance does not seem to be running at all."
6387 " In this case, it's safer to repair by"
6388 " running 'gnt-instance stop' to ensure disk"
6389 " shutdown, and then restarting it.")
6391 if runningon_target:
6392 # the migration has actually succeeded, we need to update the config
6393 self.feedback_fn("* instance running on secondary node (%s),"
6394 " updating config" % target_node)
6395 instance.primary_node = target_node
6396 self.cfg.Update(instance, self.feedback_fn)
6397 demoted_node = source_node
6399 self.feedback_fn("* instance confirmed to be running on its"
6400 " primary node (%s)" % source_node)
6401 demoted_node = target_node
6403 if instance.disk_template in constants.DTS_NET_MIRROR:
6404 self._EnsureSecondary(demoted_node)
6406 self._WaitUntilSync()
6407 except errors.OpExecError:
6408 # we ignore here errors, since if the device is standalone, it
6409 # won't be able to sync
6411 self._GoStandalone()
6412 self._GoReconnect(False)
6413 self._WaitUntilSync()
6415 self.feedback_fn("* done")
6417 def _RevertDiskStatus(self):
6418 """Try to revert the disk status after a failed migration.
6421 target_node = self.target_node
6422 if self.instance.disk_template in constants.DTS_EXT_MIRROR:
6426 self._EnsureSecondary(target_node)
6427 self._GoStandalone()
6428 self._GoReconnect(False)
6429 self._WaitUntilSync()
6430 except errors.OpExecError, err:
6431 self.lu.LogWarning("Migration failed and I can't reconnect the"
6432 " drives: error '%s'\n"
6433 "Please look and recover the instance status" %
6436 def _AbortMigration(self):
6437 """Call the hypervisor code to abort a started migration.
6440 instance = self.instance
6441 target_node = self.target_node
6442 migration_info = self.migration_info
6444 abort_result = self.rpc.call_finalize_migration(target_node,
6448 abort_msg = abort_result.fail_msg
6450 logging.error("Aborting migration failed on target node %s: %s",
6451 target_node, abort_msg)
6452 # Don't raise an exception here, as we stil have to try to revert the
6453 # disk status, even if this step failed.
6455 def _ExecMigration(self):
6456 """Migrate an instance.
6458 The migrate is done by:
6459 - change the disks into dual-master mode
6460 - wait until disks are fully synchronized again
6461 - migrate the instance
6462 - change disks on the new secondary node (the old primary) to secondary
6463 - wait until disks are fully synchronized
6464 - change disks into single-master mode
6467 instance = self.instance
6468 target_node = self.target_node
6469 source_node = self.source_node
6471 self.feedback_fn("* checking disk consistency between source and target")
6472 for dev in instance.disks:
6473 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6474 raise errors.OpExecError("Disk %s is degraded or not fully"
6475 " synchronized on target node,"
6476 " aborting migrate." % dev.iv_name)
6478 # First get the migration information from the remote node
6479 result = self.rpc.call_migration_info(source_node, instance)
6480 msg = result.fail_msg
6482 log_err = ("Failed fetching source migration information from %s: %s" %
6484 logging.error(log_err)
6485 raise errors.OpExecError(log_err)
6487 self.migration_info = migration_info = result.payload
6489 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
6490 # Then switch the disks to master/master mode
6491 self._EnsureSecondary(target_node)
6492 self._GoStandalone()
6493 self._GoReconnect(True)
6494 self._WaitUntilSync()
6496 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6497 result = self.rpc.call_accept_instance(target_node,
6500 self.nodes_ip[target_node])
6502 msg = result.fail_msg
6504 logging.error("Instance pre-migration failed, trying to revert"
6505 " disk status: %s", msg)
6506 self.feedback_fn("Pre-migration failed, aborting")
6507 self._AbortMigration()
6508 self._RevertDiskStatus()
6509 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6510 (instance.name, msg))
6512 self.feedback_fn("* migrating instance to %s" % target_node)
6514 result = self.rpc.call_instance_migrate(source_node, instance,
6515 self.nodes_ip[target_node],
6517 msg = result.fail_msg
6519 logging.error("Instance migration failed, trying to revert"
6520 " disk status: %s", msg)
6521 self.feedback_fn("Migration failed, aborting")
6522 self._AbortMigration()
6523 self._RevertDiskStatus()
6524 raise errors.OpExecError("Could not migrate instance %s: %s" %
6525 (instance.name, msg))
6528 instance.primary_node = target_node
6529 # distribute new instance config to the other nodes
6530 self.cfg.Update(instance, self.feedback_fn)
6532 result = self.rpc.call_finalize_migration(target_node,
6536 msg = result.fail_msg
6538 logging.error("Instance migration succeeded, but finalization failed:"
6540 raise errors.OpExecError("Could not finalize instance migration: %s" %
6543 if self.instance.disk_template not in constants.DTS_EXT_MIRROR:
6544 self._EnsureSecondary(source_node)
6545 self._WaitUntilSync()
6546 self._GoStandalone()
6547 self._GoReconnect(False)
6548 self._WaitUntilSync()
6550 self.feedback_fn("* done")
6552 def Exec(self, feedback_fn):
6553 """Perform the migration.
6556 feedback_fn("Migrating instance %s" % self.instance.name)
6558 self.feedback_fn = feedback_fn
6560 self.source_node = self.instance.primary_node
6562 # FIXME: if we implement migrate-to-any in DRBD, this needs fixing
6563 if self.instance.disk_template in constants.DTS_NET_MIRROR:
6564 self.target_node = self.instance.secondary_nodes[0]
6565 # Otherwise self.target_node has been populated either
6566 # directly, or through an iallocator.
6568 self.all_nodes = [self.source_node, self.target_node]
6570 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6571 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6575 return self._ExecCleanup()
6577 return self._ExecMigration()
6580 def _CreateBlockDev(lu, node, instance, device, force_create,
6582 """Create a tree of block devices on a given node.
6584 If this device type has to be created on secondaries, create it and
6587 If not, just recurse to children keeping the same 'force' value.
6589 @param lu: the lu on whose behalf we execute
6590 @param node: the node on which to create the device
6591 @type instance: L{objects.Instance}
6592 @param instance: the instance which owns the device
6593 @type device: L{objects.Disk}
6594 @param device: the device to create
6595 @type force_create: boolean
6596 @param force_create: whether to force creation of this device; this
6597 will be change to True whenever we find a device which has
6598 CreateOnSecondary() attribute
6599 @param info: the extra 'metadata' we should attach to the device
6600 (this will be represented as a LVM tag)
6601 @type force_open: boolean
6602 @param force_open: this parameter will be passes to the
6603 L{backend.BlockdevCreate} function where it specifies
6604 whether we run on primary or not, and it affects both
6605 the child assembly and the device own Open() execution
6608 if device.CreateOnSecondary():
6612 for child in device.children:
6613 _CreateBlockDev(lu, node, instance, child, force_create,
6616 if not force_create:
6619 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6622 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6623 """Create a single block device on a given node.
6625 This will not recurse over children of the device, so they must be
6628 @param lu: the lu on whose behalf we execute
6629 @param node: the node on which to create the device
6630 @type instance: L{objects.Instance}
6631 @param instance: the instance which owns the device
6632 @type device: L{objects.Disk}
6633 @param device: the device to create
6634 @param info: the extra 'metadata' we should attach to the device
6635 (this will be represented as a LVM tag)
6636 @type force_open: boolean
6637 @param force_open: this parameter will be passes to the
6638 L{backend.BlockdevCreate} function where it specifies
6639 whether we run on primary or not, and it affects both
6640 the child assembly and the device own Open() execution
6643 lu.cfg.SetDiskID(device, node)
6644 result = lu.rpc.call_blockdev_create(node, device, device.size,
6645 instance.name, force_open, info)
6646 result.Raise("Can't create block device %s on"
6647 " node %s for instance %s" % (device, node, instance.name))
6648 if device.physical_id is None:
6649 device.physical_id = result.payload
6652 def _GenerateUniqueNames(lu, exts):
6653 """Generate a suitable LV name.
6655 This will generate a logical volume name for the given instance.
6660 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6661 results.append("%s%s" % (new_id, val))
6665 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgname, names, iv_name,
6667 """Generate a drbd8 device complete with its children.
6670 port = lu.cfg.AllocatePort()
6671 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6672 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6673 logical_id=(vgname, names[0]))
6674 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6675 logical_id=(vgname, names[1]))
6676 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6677 logical_id=(primary, secondary, port,
6680 children=[dev_data, dev_meta],
6685 def _GenerateDiskTemplate(lu, template_name,
6686 instance_name, primary_node,
6687 secondary_nodes, disk_info,
6688 file_storage_dir, file_driver,
6689 base_index, feedback_fn):
6690 """Generate the entire disk layout for a given template type.
6693 #TODO: compute space requirements
6695 vgname = lu.cfg.GetVGName()
6696 disk_count = len(disk_info)
6698 if template_name == constants.DT_DISKLESS:
6700 elif template_name == constants.DT_PLAIN:
6701 if len(secondary_nodes) != 0:
6702 raise errors.ProgrammerError("Wrong template configuration")
6704 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6705 for i in range(disk_count)])
6706 for idx, disk in enumerate(disk_info):
6707 disk_index = idx + base_index
6708 vg = disk.get("vg", vgname)
6709 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
6710 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6711 logical_id=(vg, names[idx]),
6712 iv_name="disk/%d" % disk_index,
6714 disks.append(disk_dev)
6715 elif template_name == constants.DT_DRBD8:
6716 if len(secondary_nodes) != 1:
6717 raise errors.ProgrammerError("Wrong template configuration")
6718 remote_node = secondary_nodes[0]
6719 minors = lu.cfg.AllocateDRBDMinor(
6720 [primary_node, remote_node] * len(disk_info), instance_name)
6723 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6724 for i in range(disk_count)]):
6725 names.append(lv_prefix + "_data")
6726 names.append(lv_prefix + "_meta")
6727 for idx, disk in enumerate(disk_info):
6728 disk_index = idx + base_index
6729 vg = disk.get("vg", vgname)
6730 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6731 disk["size"], vg, names[idx*2:idx*2+2],
6732 "disk/%d" % disk_index,
6733 minors[idx*2], minors[idx*2+1])
6734 disk_dev.mode = disk["mode"]
6735 disks.append(disk_dev)
6736 elif template_name == constants.DT_FILE:
6737 if len(secondary_nodes) != 0:
6738 raise errors.ProgrammerError("Wrong template configuration")
6740 opcodes.RequireFileStorage()
6742 for idx, disk in enumerate(disk_info):
6743 disk_index = idx + base_index
6744 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6745 iv_name="disk/%d" % disk_index,
6746 logical_id=(file_driver,
6747 "%s/disk%d" % (file_storage_dir,
6750 disks.append(disk_dev)
6751 elif template_name == constants.DT_SHARED_FILE:
6752 if len(secondary_nodes) != 0:
6753 raise errors.ProgrammerError("Wrong template configuration")
6755 opcodes.RequireSharedFileStorage()
6757 for idx, disk in enumerate(disk_info):
6758 disk_index = idx + base_index
6759 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6760 iv_name="disk/%d" % disk_index,
6761 logical_id=(file_driver,
6762 "%s/disk%d" % (file_storage_dir,
6765 disks.append(disk_dev)
6766 elif template_name == constants.DT_BLOCK:
6767 if len(secondary_nodes) != 0:
6768 raise errors.ProgrammerError("Wrong template configuration")
6770 for idx, disk in enumerate(disk_info):
6771 disk_index = idx + base_index
6772 disk_dev = objects.Disk(dev_type=constants.LD_BLOCKDEV, size=disk["size"],
6773 logical_id=(constants.BLOCKDEV_DRIVER_MANUAL,
6775 iv_name="disk/%d" % disk_index,
6777 disks.append(disk_dev)
6780 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6784 def _GetInstanceInfoText(instance):
6785 """Compute that text that should be added to the disk's metadata.
6788 return "originstname+%s" % instance.name
6791 def _CalcEta(time_taken, written, total_size):
6792 """Calculates the ETA based on size written and total size.
6794 @param time_taken: The time taken so far
6795 @param written: amount written so far
6796 @param total_size: The total size of data to be written
6797 @return: The remaining time in seconds
6800 avg_time = time_taken / float(written)
6801 return (total_size - written) * avg_time
6804 def _WipeDisks(lu, instance):
6805 """Wipes instance disks.
6807 @type lu: L{LogicalUnit}
6808 @param lu: the logical unit on whose behalf we execute
6809 @type instance: L{objects.Instance}
6810 @param instance: the instance whose disks we should create
6811 @return: the success of the wipe
6814 node = instance.primary_node
6816 for device in instance.disks:
6817 lu.cfg.SetDiskID(device, node)
6819 logging.info("Pause sync of instance %s disks", instance.name)
6820 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
6822 for idx, success in enumerate(result.payload):
6824 logging.warn("pause-sync of instance %s for disks %d failed",
6828 for idx, device in enumerate(instance.disks):
6829 lu.LogInfo("* Wiping disk %d", idx)
6830 logging.info("Wiping disk %d for instance %s, node %s",
6831 idx, instance.name, node)
6833 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
6834 # MAX_WIPE_CHUNK at max
6835 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
6836 constants.MIN_WIPE_CHUNK_PERCENT)
6841 start_time = time.time()
6843 while offset < size:
6844 wipe_size = min(wipe_chunk_size, size - offset)
6845 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
6846 result.Raise("Could not wipe disk %d at offset %d for size %d" %
6847 (idx, offset, wipe_size))
6850 if now - last_output >= 60:
6851 eta = _CalcEta(now - start_time, offset, size)
6852 lu.LogInfo(" - done: %.1f%% ETA: %s" %
6853 (offset / float(size) * 100, utils.FormatSeconds(eta)))
6856 logging.info("Resume sync of instance %s disks", instance.name)
6858 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
6860 for idx, success in enumerate(result.payload):
6862 lu.LogWarning("Warning: Resume sync of disk %d failed. Please have a"
6863 " look at the status and troubleshoot the issue.", idx)
6864 logging.warn("resume-sync of instance %s for disks %d failed",
6868 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6869 """Create all disks for an instance.
6871 This abstracts away some work from AddInstance.
6873 @type lu: L{LogicalUnit}
6874 @param lu: the logical unit on whose behalf we execute
6875 @type instance: L{objects.Instance}
6876 @param instance: the instance whose disks we should create
6878 @param to_skip: list of indices to skip
6879 @type target_node: string
6880 @param target_node: if passed, overrides the target node for creation
6882 @return: the success of the creation
6885 info = _GetInstanceInfoText(instance)
6886 if target_node is None:
6887 pnode = instance.primary_node
6888 all_nodes = instance.all_nodes
6893 if instance.disk_template in (constants.DT_FILE, constants.DT_SHARED_FILE):
6894 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6895 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6897 result.Raise("Failed to create directory '%s' on"
6898 " node %s" % (file_storage_dir, pnode))
6900 # Note: this needs to be kept in sync with adding of disks in
6901 # LUInstanceSetParams
6902 for idx, device in enumerate(instance.disks):
6903 if to_skip and idx in to_skip:
6905 logging.info("Creating volume %s for instance %s",
6906 device.iv_name, instance.name)
6908 for node in all_nodes:
6909 f_create = node == pnode
6910 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6913 def _RemoveDisks(lu, instance, target_node=None):
6914 """Remove all disks for an instance.
6916 This abstracts away some work from `AddInstance()` and
6917 `RemoveInstance()`. Note that in case some of the devices couldn't
6918 be removed, the removal will continue with the other ones (compare
6919 with `_CreateDisks()`).
6921 @type lu: L{LogicalUnit}
6922 @param lu: the logical unit on whose behalf we execute
6923 @type instance: L{objects.Instance}
6924 @param instance: the instance whose disks we should remove
6925 @type target_node: string
6926 @param target_node: used to override the node on which to remove the disks
6928 @return: the success of the removal
6931 logging.info("Removing block devices for instance %s", instance.name)
6934 for device in instance.disks:
6936 edata = [(target_node, device)]
6938 edata = device.ComputeNodeTree(instance.primary_node)
6939 for node, disk in edata:
6940 lu.cfg.SetDiskID(disk, node)
6941 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6943 lu.LogWarning("Could not remove block device %s on node %s,"
6944 " continuing anyway: %s", device.iv_name, node, msg)
6947 if instance.disk_template == constants.DT_FILE:
6948 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6952 tgt = instance.primary_node
6953 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6955 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6956 file_storage_dir, instance.primary_node, result.fail_msg)
6962 def _ComputeDiskSizePerVG(disk_template, disks):
6963 """Compute disk size requirements in the volume group
6966 def _compute(disks, payload):
6967 """Universal algorithm
6972 vgs[disk["vg"]] = vgs.get("vg", 0) + disk["size"] + payload
6976 # Required free disk space as a function of disk and swap space
6978 constants.DT_DISKLESS: {},
6979 constants.DT_PLAIN: _compute(disks, 0),
6980 # 128 MB are added for drbd metadata for each disk
6981 constants.DT_DRBD8: _compute(disks, 128),
6982 constants.DT_FILE: {},
6983 constants.DT_SHARED_FILE: {},
6986 if disk_template not in req_size_dict:
6987 raise errors.ProgrammerError("Disk template '%s' size requirement"
6988 " is unknown" % disk_template)
6990 return req_size_dict[disk_template]
6993 def _ComputeDiskSize(disk_template, disks):
6994 """Compute disk size requirements in the volume group
6997 # Required free disk space as a function of disk and swap space
6999 constants.DT_DISKLESS: None,
7000 constants.DT_PLAIN: sum(d["size"] for d in disks),
7001 # 128 MB are added for drbd metadata for each disk
7002 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
7003 constants.DT_FILE: None,
7004 constants.DT_SHARED_FILE: 0,
7005 constants.DT_BLOCK: 0,
7008 if disk_template not in req_size_dict:
7009 raise errors.ProgrammerError("Disk template '%s' size requirement"
7010 " is unknown" % disk_template)
7012 return req_size_dict[disk_template]
7015 def _FilterVmNodes(lu, nodenames):
7016 """Filters out non-vm_capable nodes from a list.
7018 @type lu: L{LogicalUnit}
7019 @param lu: the logical unit for which we check
7020 @type nodenames: list
7021 @param nodenames: the list of nodes on which we should check
7023 @return: the list of vm-capable nodes
7026 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
7027 return [name for name in nodenames if name not in vm_nodes]
7030 def _CheckHVParams(lu, nodenames, hvname, hvparams):
7031 """Hypervisor parameter validation.
7033 This function abstract the hypervisor parameter validation to be
7034 used in both instance create and instance modify.
7036 @type lu: L{LogicalUnit}
7037 @param lu: the logical unit for which we check
7038 @type nodenames: list
7039 @param nodenames: the list of nodes on which we should check
7040 @type hvname: string
7041 @param hvname: the name of the hypervisor we should use
7042 @type hvparams: dict
7043 @param hvparams: the parameters which we need to check
7044 @raise errors.OpPrereqError: if the parameters are not valid
7047 nodenames = _FilterVmNodes(lu, nodenames)
7048 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
7051 for node in nodenames:
7055 info.Raise("Hypervisor parameter validation failed on node %s" % node)
7058 def _CheckOSParams(lu, required, nodenames, osname, osparams):
7059 """OS parameters validation.
7061 @type lu: L{LogicalUnit}
7062 @param lu: the logical unit for which we check
7063 @type required: boolean
7064 @param required: whether the validation should fail if the OS is not
7066 @type nodenames: list
7067 @param nodenames: the list of nodes on which we should check
7068 @type osname: string
7069 @param osname: the name of the hypervisor we should use
7070 @type osparams: dict
7071 @param osparams: the parameters which we need to check
7072 @raise errors.OpPrereqError: if the parameters are not valid
7075 nodenames = _FilterVmNodes(lu, nodenames)
7076 result = lu.rpc.call_os_validate(required, nodenames, osname,
7077 [constants.OS_VALIDATE_PARAMETERS],
7079 for node, nres in result.items():
7080 # we don't check for offline cases since this should be run only
7081 # against the master node and/or an instance's nodes
7082 nres.Raise("OS Parameters validation failed on node %s" % node)
7083 if not nres.payload:
7084 lu.LogInfo("OS %s not found on node %s, validation skipped",
7088 class LUInstanceCreate(LogicalUnit):
7089 """Create an instance.
7092 HPATH = "instance-add"
7093 HTYPE = constants.HTYPE_INSTANCE
7096 def CheckArguments(self):
7100 # do not require name_check to ease forward/backward compatibility
7102 if self.op.no_install and self.op.start:
7103 self.LogInfo("No-installation mode selected, disabling startup")
7104 self.op.start = False
7105 # validate/normalize the instance name
7106 self.op.instance_name = \
7107 netutils.Hostname.GetNormalizedName(self.op.instance_name)
7109 if self.op.ip_check and not self.op.name_check:
7110 # TODO: make the ip check more flexible and not depend on the name check
7111 raise errors.OpPrereqError("Cannot do ip check without a name check",
7114 # check nics' parameter names
7115 for nic in self.op.nics:
7116 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
7118 # check disks. parameter names and consistent adopt/no-adopt strategy
7119 has_adopt = has_no_adopt = False
7120 for disk in self.op.disks:
7121 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
7126 if has_adopt and has_no_adopt:
7127 raise errors.OpPrereqError("Either all disks are adopted or none is",
7130 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
7131 raise errors.OpPrereqError("Disk adoption is not supported for the"
7132 " '%s' disk template" %
7133 self.op.disk_template,
7135 if self.op.iallocator is not None:
7136 raise errors.OpPrereqError("Disk adoption not allowed with an"
7137 " iallocator script", errors.ECODE_INVAL)
7138 if self.op.mode == constants.INSTANCE_IMPORT:
7139 raise errors.OpPrereqError("Disk adoption not allowed for"
7140 " instance import", errors.ECODE_INVAL)
7142 if self.op.disk_template in constants.DTS_MUST_ADOPT:
7143 raise errors.OpPrereqError("Disk template %s requires disk adoption,"
7144 " but no 'adopt' parameter given" %
7145 self.op.disk_template,
7148 self.adopt_disks = has_adopt
7150 # instance name verification
7151 if self.op.name_check:
7152 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
7153 self.op.instance_name = self.hostname1.name
7154 # used in CheckPrereq for ip ping check
7155 self.check_ip = self.hostname1.ip
7157 self.check_ip = None
7159 # file storage checks
7160 if (self.op.file_driver and
7161 not self.op.file_driver in constants.FILE_DRIVER):
7162 raise errors.OpPrereqError("Invalid file driver name '%s'" %
7163 self.op.file_driver, errors.ECODE_INVAL)
7165 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
7166 raise errors.OpPrereqError("File storage directory path not absolute",
7169 ### Node/iallocator related checks
7170 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
7172 if self.op.pnode is not None:
7173 if self.op.disk_template in constants.DTS_NET_MIRROR:
7174 if self.op.snode is None:
7175 raise errors.OpPrereqError("The networked disk templates need"
7176 " a mirror node", errors.ECODE_INVAL)
7178 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
7180 self.op.snode = None
7182 self._cds = _GetClusterDomainSecret()
7184 if self.op.mode == constants.INSTANCE_IMPORT:
7185 # On import force_variant must be True, because if we forced it at
7186 # initial install, our only chance when importing it back is that it
7188 self.op.force_variant = True
7190 if self.op.no_install:
7191 self.LogInfo("No-installation mode has no effect during import")
7193 elif self.op.mode == constants.INSTANCE_CREATE:
7194 if self.op.os_type is None:
7195 raise errors.OpPrereqError("No guest OS specified",
7197 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
7198 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
7199 " installation" % self.op.os_type,
7201 if self.op.disk_template is None:
7202 raise errors.OpPrereqError("No disk template specified",
7205 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7206 # Check handshake to ensure both clusters have the same domain secret
7207 src_handshake = self.op.source_handshake
7208 if not src_handshake:
7209 raise errors.OpPrereqError("Missing source handshake",
7212 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
7215 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
7218 # Load and check source CA
7219 self.source_x509_ca_pem = self.op.source_x509_ca
7220 if not self.source_x509_ca_pem:
7221 raise errors.OpPrereqError("Missing source X509 CA",
7225 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
7227 except OpenSSL.crypto.Error, err:
7228 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
7229 (err, ), errors.ECODE_INVAL)
7231 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
7232 if errcode is not None:
7233 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
7236 self.source_x509_ca = cert
7238 src_instance_name = self.op.source_instance_name
7239 if not src_instance_name:
7240 raise errors.OpPrereqError("Missing source instance name",
7243 self.source_instance_name = \
7244 netutils.GetHostname(name=src_instance_name).name
7247 raise errors.OpPrereqError("Invalid instance creation mode %r" %
7248 self.op.mode, errors.ECODE_INVAL)
7250 def ExpandNames(self):
7251 """ExpandNames for CreateInstance.
7253 Figure out the right locks for instance creation.
7256 self.needed_locks = {}
7258 instance_name = self.op.instance_name
7259 # this is just a preventive check, but someone might still add this
7260 # instance in the meantime, and creation will fail at lock-add time
7261 if instance_name in self.cfg.GetInstanceList():
7262 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7263 instance_name, errors.ECODE_EXISTS)
7265 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
7267 if self.op.iallocator:
7268 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7270 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
7271 nodelist = [self.op.pnode]
7272 if self.op.snode is not None:
7273 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
7274 nodelist.append(self.op.snode)
7275 self.needed_locks[locking.LEVEL_NODE] = nodelist
7277 # in case of import lock the source node too
7278 if self.op.mode == constants.INSTANCE_IMPORT:
7279 src_node = self.op.src_node
7280 src_path = self.op.src_path
7282 if src_path is None:
7283 self.op.src_path = src_path = self.op.instance_name
7285 if src_node is None:
7286 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7287 self.op.src_node = None
7288 if os.path.isabs(src_path):
7289 raise errors.OpPrereqError("Importing an instance from an absolute"
7290 " path requires a source node option.",
7293 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
7294 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
7295 self.needed_locks[locking.LEVEL_NODE].append(src_node)
7296 if not os.path.isabs(src_path):
7297 self.op.src_path = src_path = \
7298 utils.PathJoin(constants.EXPORT_DIR, src_path)
7300 def _RunAllocator(self):
7301 """Run the allocator based on input opcode.
7304 nics = [n.ToDict() for n in self.nics]
7305 ial = IAllocator(self.cfg, self.rpc,
7306 mode=constants.IALLOCATOR_MODE_ALLOC,
7307 name=self.op.instance_name,
7308 disk_template=self.op.disk_template,
7311 vcpus=self.be_full[constants.BE_VCPUS],
7312 mem_size=self.be_full[constants.BE_MEMORY],
7315 hypervisor=self.op.hypervisor,
7318 ial.Run(self.op.iallocator)
7321 raise errors.OpPrereqError("Can't compute nodes using"
7322 " iallocator '%s': %s" %
7323 (self.op.iallocator, ial.info),
7325 if len(ial.result) != ial.required_nodes:
7326 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7327 " of nodes (%s), required %s" %
7328 (self.op.iallocator, len(ial.result),
7329 ial.required_nodes), errors.ECODE_FAULT)
7330 self.op.pnode = ial.result[0]
7331 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7332 self.op.instance_name, self.op.iallocator,
7333 utils.CommaJoin(ial.result))
7334 if ial.required_nodes == 2:
7335 self.op.snode = ial.result[1]
7337 def BuildHooksEnv(self):
7340 This runs on master, primary and secondary nodes of the instance.
7344 "ADD_MODE": self.op.mode,
7346 if self.op.mode == constants.INSTANCE_IMPORT:
7347 env["SRC_NODE"] = self.op.src_node
7348 env["SRC_PATH"] = self.op.src_path
7349 env["SRC_IMAGES"] = self.src_images
7351 env.update(_BuildInstanceHookEnv(
7352 name=self.op.instance_name,
7353 primary_node=self.op.pnode,
7354 secondary_nodes=self.secondaries,
7355 status=self.op.start,
7356 os_type=self.op.os_type,
7357 memory=self.be_full[constants.BE_MEMORY],
7358 vcpus=self.be_full[constants.BE_VCPUS],
7359 nics=_NICListToTuple(self, self.nics),
7360 disk_template=self.op.disk_template,
7361 disks=[(d["size"], d["mode"]) for d in self.disks],
7364 hypervisor_name=self.op.hypervisor,
7367 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
7371 def _ReadExportInfo(self):
7372 """Reads the export information from disk.
7374 It will override the opcode source node and path with the actual
7375 information, if these two were not specified before.
7377 @return: the export information
7380 assert self.op.mode == constants.INSTANCE_IMPORT
7382 src_node = self.op.src_node
7383 src_path = self.op.src_path
7385 if src_node is None:
7386 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
7387 exp_list = self.rpc.call_export_list(locked_nodes)
7389 for node in exp_list:
7390 if exp_list[node].fail_msg:
7392 if src_path in exp_list[node].payload:
7394 self.op.src_node = src_node = node
7395 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
7399 raise errors.OpPrereqError("No export found for relative path %s" %
7400 src_path, errors.ECODE_INVAL)
7402 _CheckNodeOnline(self, src_node)
7403 result = self.rpc.call_export_info(src_node, src_path)
7404 result.Raise("No export or invalid export found in dir %s" % src_path)
7406 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
7407 if not export_info.has_section(constants.INISECT_EXP):
7408 raise errors.ProgrammerError("Corrupted export config",
7409 errors.ECODE_ENVIRON)
7411 ei_version = export_info.get(constants.INISECT_EXP, "version")
7412 if (int(ei_version) != constants.EXPORT_VERSION):
7413 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
7414 (ei_version, constants.EXPORT_VERSION),
7415 errors.ECODE_ENVIRON)
7418 def _ReadExportParams(self, einfo):
7419 """Use export parameters as defaults.
7421 In case the opcode doesn't specify (as in override) some instance
7422 parameters, then try to use them from the export information, if
7426 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
7428 if self.op.disk_template is None:
7429 if einfo.has_option(constants.INISECT_INS, "disk_template"):
7430 self.op.disk_template = einfo.get(constants.INISECT_INS,
7433 raise errors.OpPrereqError("No disk template specified and the export"
7434 " is missing the disk_template information",
7437 if not self.op.disks:
7438 if einfo.has_option(constants.INISECT_INS, "disk_count"):
7440 # TODO: import the disk iv_name too
7441 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
7442 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
7443 disks.append({"size": disk_sz})
7444 self.op.disks = disks
7446 raise errors.OpPrereqError("No disk info specified and the export"
7447 " is missing the disk information",
7450 if (not self.op.nics and
7451 einfo.has_option(constants.INISECT_INS, "nic_count")):
7453 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
7455 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
7456 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
7461 if (self.op.hypervisor is None and
7462 einfo.has_option(constants.INISECT_INS, "hypervisor")):
7463 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
7464 if einfo.has_section(constants.INISECT_HYP):
7465 # use the export parameters but do not override the ones
7466 # specified by the user
7467 for name, value in einfo.items(constants.INISECT_HYP):
7468 if name not in self.op.hvparams:
7469 self.op.hvparams[name] = value
7471 if einfo.has_section(constants.INISECT_BEP):
7472 # use the parameters, without overriding
7473 for name, value in einfo.items(constants.INISECT_BEP):
7474 if name not in self.op.beparams:
7475 self.op.beparams[name] = value
7477 # try to read the parameters old style, from the main section
7478 for name in constants.BES_PARAMETERS:
7479 if (name not in self.op.beparams and
7480 einfo.has_option(constants.INISECT_INS, name)):
7481 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
7483 if einfo.has_section(constants.INISECT_OSP):
7484 # use the parameters, without overriding
7485 for name, value in einfo.items(constants.INISECT_OSP):
7486 if name not in self.op.osparams:
7487 self.op.osparams[name] = value
7489 def _RevertToDefaults(self, cluster):
7490 """Revert the instance parameters to the default values.
7494 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7495 for name in self.op.hvparams.keys():
7496 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7497 del self.op.hvparams[name]
7499 be_defs = cluster.SimpleFillBE({})
7500 for name in self.op.beparams.keys():
7501 if name in be_defs and be_defs[name] == self.op.beparams[name]:
7502 del self.op.beparams[name]
7504 nic_defs = cluster.SimpleFillNIC({})
7505 for nic in self.op.nics:
7506 for name in constants.NICS_PARAMETERS:
7507 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7510 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7511 for name in self.op.osparams.keys():
7512 if name in os_defs and os_defs[name] == self.op.osparams[name]:
7513 del self.op.osparams[name]
7515 def CheckPrereq(self):
7516 """Check prerequisites.
7519 if self.op.mode == constants.INSTANCE_IMPORT:
7520 export_info = self._ReadExportInfo()
7521 self._ReadExportParams(export_info)
7523 if (not self.cfg.GetVGName() and
7524 self.op.disk_template not in constants.DTS_NOT_LVM):
7525 raise errors.OpPrereqError("Cluster does not support lvm-based"
7526 " instances", errors.ECODE_STATE)
7528 if self.op.hypervisor is None:
7529 self.op.hypervisor = self.cfg.GetHypervisorType()
7531 cluster = self.cfg.GetClusterInfo()
7532 enabled_hvs = cluster.enabled_hypervisors
7533 if self.op.hypervisor not in enabled_hvs:
7534 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7535 " cluster (%s)" % (self.op.hypervisor,
7536 ",".join(enabled_hvs)),
7539 # check hypervisor parameter syntax (locally)
7540 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7541 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7543 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7544 hv_type.CheckParameterSyntax(filled_hvp)
7545 self.hv_full = filled_hvp
7546 # check that we don't specify global parameters on an instance
7547 _CheckGlobalHvParams(self.op.hvparams)
7549 # fill and remember the beparams dict
7550 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7551 self.be_full = cluster.SimpleFillBE(self.op.beparams)
7553 # build os parameters
7554 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7556 # now that hvp/bep are in final format, let's reset to defaults,
7558 if self.op.identify_defaults:
7559 self._RevertToDefaults(cluster)
7563 for idx, nic in enumerate(self.op.nics):
7564 nic_mode_req = nic.get("mode", None)
7565 nic_mode = nic_mode_req
7566 if nic_mode is None:
7567 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7569 # in routed mode, for the first nic, the default ip is 'auto'
7570 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7571 default_ip_mode = constants.VALUE_AUTO
7573 default_ip_mode = constants.VALUE_NONE
7575 # ip validity checks
7576 ip = nic.get("ip", default_ip_mode)
7577 if ip is None or ip.lower() == constants.VALUE_NONE:
7579 elif ip.lower() == constants.VALUE_AUTO:
7580 if not self.op.name_check:
7581 raise errors.OpPrereqError("IP address set to auto but name checks"
7582 " have been skipped",
7584 nic_ip = self.hostname1.ip
7586 if not netutils.IPAddress.IsValid(ip):
7587 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7591 # TODO: check the ip address for uniqueness
7592 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7593 raise errors.OpPrereqError("Routed nic mode requires an ip address",
7596 # MAC address verification
7597 mac = nic.get("mac", constants.VALUE_AUTO)
7598 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7599 mac = utils.NormalizeAndValidateMac(mac)
7602 self.cfg.ReserveMAC(mac, self.proc.GetECId())
7603 except errors.ReservationError:
7604 raise errors.OpPrereqError("MAC address %s already in use"
7605 " in cluster" % mac,
7606 errors.ECODE_NOTUNIQUE)
7608 # Build nic parameters
7609 link = nic.get(constants.INIC_LINK, None)
7612 nicparams[constants.NIC_MODE] = nic_mode_req
7614 nicparams[constants.NIC_LINK] = link
7616 check_params = cluster.SimpleFillNIC(nicparams)
7617 objects.NIC.CheckParameterSyntax(check_params)
7618 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7620 # disk checks/pre-build
7622 for disk in self.op.disks:
7623 mode = disk.get("mode", constants.DISK_RDWR)
7624 if mode not in constants.DISK_ACCESS_SET:
7625 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7626 mode, errors.ECODE_INVAL)
7627 size = disk.get("size", None)
7629 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7632 except (TypeError, ValueError):
7633 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7635 vg = disk.get("vg", self.cfg.GetVGName())
7636 new_disk = {"size": size, "mode": mode, "vg": vg}
7638 new_disk["adopt"] = disk["adopt"]
7639 self.disks.append(new_disk)
7641 if self.op.mode == constants.INSTANCE_IMPORT:
7643 # Check that the new instance doesn't have less disks than the export
7644 instance_disks = len(self.disks)
7645 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7646 if instance_disks < export_disks:
7647 raise errors.OpPrereqError("Not enough disks to import."
7648 " (instance: %d, export: %d)" %
7649 (instance_disks, export_disks),
7653 for idx in range(export_disks):
7654 option = 'disk%d_dump' % idx
7655 if export_info.has_option(constants.INISECT_INS, option):
7656 # FIXME: are the old os-es, disk sizes, etc. useful?
7657 export_name = export_info.get(constants.INISECT_INS, option)
7658 image = utils.PathJoin(self.op.src_path, export_name)
7659 disk_images.append(image)
7661 disk_images.append(False)
7663 self.src_images = disk_images
7665 old_name = export_info.get(constants.INISECT_INS, 'name')
7667 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7668 except (TypeError, ValueError), err:
7669 raise errors.OpPrereqError("Invalid export file, nic_count is not"
7670 " an integer: %s" % str(err),
7672 if self.op.instance_name == old_name:
7673 for idx, nic in enumerate(self.nics):
7674 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7675 nic_mac_ini = 'nic%d_mac' % idx
7676 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7678 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7680 # ip ping checks (we use the same ip that was resolved in ExpandNames)
7681 if self.op.ip_check:
7682 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7683 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7684 (self.check_ip, self.op.instance_name),
7685 errors.ECODE_NOTUNIQUE)
7687 #### mac address generation
7688 # By generating here the mac address both the allocator and the hooks get
7689 # the real final mac address rather than the 'auto' or 'generate' value.
7690 # There is a race condition between the generation and the instance object
7691 # creation, which means that we know the mac is valid now, but we're not
7692 # sure it will be when we actually add the instance. If things go bad
7693 # adding the instance will abort because of a duplicate mac, and the
7694 # creation job will fail.
7695 for nic in self.nics:
7696 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7697 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7701 if self.op.iallocator is not None:
7702 self._RunAllocator()
7704 #### node related checks
7706 # check primary node
7707 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7708 assert self.pnode is not None, \
7709 "Cannot retrieve locked node %s" % self.op.pnode
7711 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7712 pnode.name, errors.ECODE_STATE)
7714 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7715 pnode.name, errors.ECODE_STATE)
7716 if not pnode.vm_capable:
7717 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
7718 " '%s'" % pnode.name, errors.ECODE_STATE)
7720 self.secondaries = []
7722 # mirror node verification
7723 if self.op.disk_template in constants.DTS_NET_MIRROR:
7724 if self.op.snode == pnode.name:
7725 raise errors.OpPrereqError("The secondary node cannot be the"
7726 " primary node.", errors.ECODE_INVAL)
7727 _CheckNodeOnline(self, self.op.snode)
7728 _CheckNodeNotDrained(self, self.op.snode)
7729 _CheckNodeVmCapable(self, self.op.snode)
7730 self.secondaries.append(self.op.snode)
7732 nodenames = [pnode.name] + self.secondaries
7734 if not self.adopt_disks:
7735 # Check lv size requirements, if not adopting
7736 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
7737 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
7739 elif self.op.disk_template == constants.DT_PLAIN: # Check the adoption data
7740 all_lvs = set([i["vg"] + "/" + i["adopt"] for i in self.disks])
7741 if len(all_lvs) != len(self.disks):
7742 raise errors.OpPrereqError("Duplicate volume names given for adoption",
7744 for lv_name in all_lvs:
7746 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
7747 # to ReserveLV uses the same syntax
7748 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7749 except errors.ReservationError:
7750 raise errors.OpPrereqError("LV named %s used by another instance" %
7751 lv_name, errors.ECODE_NOTUNIQUE)
7753 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
7754 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
7756 node_lvs = self.rpc.call_lv_list([pnode.name],
7757 vg_names.payload.keys())[pnode.name]
7758 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7759 node_lvs = node_lvs.payload
7761 delta = all_lvs.difference(node_lvs.keys())
7763 raise errors.OpPrereqError("Missing logical volume(s): %s" %
7764 utils.CommaJoin(delta),
7766 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7768 raise errors.OpPrereqError("Online logical volumes found, cannot"
7769 " adopt: %s" % utils.CommaJoin(online_lvs),
7771 # update the size of disk based on what is found
7772 for dsk in self.disks:
7773 dsk["size"] = int(float(node_lvs[dsk["vg"] + "/" + dsk["adopt"]][0]))
7775 elif self.op.disk_template == constants.DT_BLOCK:
7776 # Normalize and de-duplicate device paths
7777 all_disks = set([os.path.abspath(i["adopt"]) for i in self.disks])
7778 if len(all_disks) != len(self.disks):
7779 raise errors.OpPrereqError("Duplicate disk names given for adoption",
7781 baddisks = [d for d in all_disks
7782 if not d.startswith(constants.ADOPTABLE_BLOCKDEV_ROOT)]
7784 raise errors.OpPrereqError("Device node(s) %s lie outside %s and"
7785 " cannot be adopted" %
7786 (", ".join(baddisks),
7787 constants.ADOPTABLE_BLOCKDEV_ROOT),
7790 node_disks = self.rpc.call_bdev_sizes([pnode.name],
7791 list(all_disks))[pnode.name]
7792 node_disks.Raise("Cannot get block device information from node %s" %
7794 node_disks = node_disks.payload
7795 delta = all_disks.difference(node_disks.keys())
7797 raise errors.OpPrereqError("Missing block device(s): %s" %
7798 utils.CommaJoin(delta),
7800 for dsk in self.disks:
7801 dsk["size"] = int(float(node_disks[dsk["adopt"]]))
7803 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7805 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7806 # check OS parameters (remotely)
7807 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7809 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7811 # memory check on primary node
7813 _CheckNodeFreeMemory(self, self.pnode.name,
7814 "creating instance %s" % self.op.instance_name,
7815 self.be_full[constants.BE_MEMORY],
7818 self.dry_run_result = list(nodenames)
7820 def Exec(self, feedback_fn):
7821 """Create and add the instance to the cluster.
7824 instance = self.op.instance_name
7825 pnode_name = self.pnode.name
7827 ht_kind = self.op.hypervisor
7828 if ht_kind in constants.HTS_REQ_PORT:
7829 network_port = self.cfg.AllocatePort()
7833 if constants.ENABLE_FILE_STORAGE or constants.ENABLE_SHARED_FILE_STORAGE:
7834 # this is needed because os.path.join does not accept None arguments
7835 if self.op.file_storage_dir is None:
7836 string_file_storage_dir = ""
7838 string_file_storage_dir = self.op.file_storage_dir
7840 # build the full file storage dir path
7841 if self.op.disk_template == constants.DT_SHARED_FILE:
7842 get_fsd_fn = self.cfg.GetSharedFileStorageDir
7844 get_fsd_fn = self.cfg.GetFileStorageDir
7846 file_storage_dir = utils.PathJoin(get_fsd_fn(),
7847 string_file_storage_dir, instance)
7849 file_storage_dir = ""
7851 disks = _GenerateDiskTemplate(self,
7852 self.op.disk_template,
7853 instance, pnode_name,
7857 self.op.file_driver,
7861 iobj = objects.Instance(name=instance, os=self.op.os_type,
7862 primary_node=pnode_name,
7863 nics=self.nics, disks=disks,
7864 disk_template=self.op.disk_template,
7866 network_port=network_port,
7867 beparams=self.op.beparams,
7868 hvparams=self.op.hvparams,
7869 hypervisor=self.op.hypervisor,
7870 osparams=self.op.osparams,
7873 if self.adopt_disks:
7874 if self.op.disk_template == constants.DT_PLAIN:
7875 # rename LVs to the newly-generated names; we need to construct
7876 # 'fake' LV disks with the old data, plus the new unique_id
7877 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7879 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7880 rename_to.append(t_dsk.logical_id)
7881 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7882 self.cfg.SetDiskID(t_dsk, pnode_name)
7883 result = self.rpc.call_blockdev_rename(pnode_name,
7884 zip(tmp_disks, rename_to))
7885 result.Raise("Failed to rename adoped LVs")
7887 feedback_fn("* creating instance disks...")
7889 _CreateDisks(self, iobj)
7890 except errors.OpExecError:
7891 self.LogWarning("Device creation failed, reverting...")
7893 _RemoveDisks(self, iobj)
7895 self.cfg.ReleaseDRBDMinors(instance)
7898 if self.cfg.GetClusterInfo().prealloc_wipe_disks:
7899 feedback_fn("* wiping instance disks...")
7901 _WipeDisks(self, iobj)
7902 except errors.OpExecError:
7903 self.LogWarning("Device wiping failed, reverting...")
7905 _RemoveDisks(self, iobj)
7907 self.cfg.ReleaseDRBDMinors(instance)
7910 feedback_fn("adding instance %s to cluster config" % instance)
7912 self.cfg.AddInstance(iobj, self.proc.GetECId())
7914 # Declare that we don't want to remove the instance lock anymore, as we've
7915 # added the instance to the config
7916 del self.remove_locks[locking.LEVEL_INSTANCE]
7917 # Unlock all the nodes
7918 if self.op.mode == constants.INSTANCE_IMPORT:
7919 nodes_keep = [self.op.src_node]
7920 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7921 if node != self.op.src_node]
7922 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7923 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7925 self.context.glm.release(locking.LEVEL_NODE)
7926 del self.acquired_locks[locking.LEVEL_NODE]
7928 if self.op.wait_for_sync:
7929 disk_abort = not _WaitForSync(self, iobj)
7930 elif iobj.disk_template in constants.DTS_NET_MIRROR:
7931 # make sure the disks are not degraded (still sync-ing is ok)
7933 feedback_fn("* checking mirrors status")
7934 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7939 _RemoveDisks(self, iobj)
7940 self.cfg.RemoveInstance(iobj.name)
7941 # Make sure the instance lock gets removed
7942 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7943 raise errors.OpExecError("There are some degraded disks for"
7946 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7947 if self.op.mode == constants.INSTANCE_CREATE:
7948 if not self.op.no_install:
7949 feedback_fn("* running the instance OS create scripts...")
7950 # FIXME: pass debug option from opcode to backend
7951 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7952 self.op.debug_level)
7953 result.Raise("Could not add os for instance %s"
7954 " on node %s" % (instance, pnode_name))
7956 elif self.op.mode == constants.INSTANCE_IMPORT:
7957 feedback_fn("* running the instance OS import scripts...")
7961 for idx, image in enumerate(self.src_images):
7965 # FIXME: pass debug option from opcode to backend
7966 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7967 constants.IEIO_FILE, (image, ),
7968 constants.IEIO_SCRIPT,
7969 (iobj.disks[idx], idx),
7971 transfers.append(dt)
7974 masterd.instance.TransferInstanceData(self, feedback_fn,
7975 self.op.src_node, pnode_name,
7976 self.pnode.secondary_ip,
7978 if not compat.all(import_result):
7979 self.LogWarning("Some disks for instance %s on node %s were not"
7980 " imported successfully" % (instance, pnode_name))
7982 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7983 feedback_fn("* preparing remote import...")
7984 # The source cluster will stop the instance before attempting to make a
7985 # connection. In some cases stopping an instance can take a long time,
7986 # hence the shutdown timeout is added to the connection timeout.
7987 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
7988 self.op.source_shutdown_timeout)
7989 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7991 assert iobj.primary_node == self.pnode.name
7993 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
7994 self.source_x509_ca,
7995 self._cds, timeouts)
7996 if not compat.all(disk_results):
7997 # TODO: Should the instance still be started, even if some disks
7998 # failed to import (valid for local imports, too)?
7999 self.LogWarning("Some disks for instance %s on node %s were not"
8000 " imported successfully" % (instance, pnode_name))
8002 # Run rename script on newly imported instance
8003 assert iobj.name == instance
8004 feedback_fn("Running rename script for %s" % instance)
8005 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
8006 self.source_instance_name,
8007 self.op.debug_level)
8009 self.LogWarning("Failed to run rename script for %s on node"
8010 " %s: %s" % (instance, pnode_name, result.fail_msg))
8013 # also checked in the prereq part
8014 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
8018 iobj.admin_up = True
8019 self.cfg.Update(iobj, feedback_fn)
8020 logging.info("Starting instance %s on node %s", instance, pnode_name)
8021 feedback_fn("* starting instance...")
8022 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
8023 result.Raise("Could not start instance")
8025 return list(iobj.all_nodes)
8028 class LUInstanceConsole(NoHooksLU):
8029 """Connect to an instance's console.
8031 This is somewhat special in that it returns the command line that
8032 you need to run on the master node in order to connect to the
8038 def ExpandNames(self):
8039 self._ExpandAndLockInstance()
8041 def CheckPrereq(self):
8042 """Check prerequisites.
8044 This checks that the instance is in the cluster.
8047 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8048 assert self.instance is not None, \
8049 "Cannot retrieve locked instance %s" % self.op.instance_name
8050 _CheckNodeOnline(self, self.instance.primary_node)
8052 def Exec(self, feedback_fn):
8053 """Connect to the console of an instance
8056 instance = self.instance
8057 node = instance.primary_node
8059 node_insts = self.rpc.call_instance_list([node],
8060 [instance.hypervisor])[node]
8061 node_insts.Raise("Can't get node information from %s" % node)
8063 if instance.name not in node_insts.payload:
8064 if instance.admin_up:
8065 state = constants.INSTST_ERRORDOWN
8067 state = constants.INSTST_ADMINDOWN
8068 raise errors.OpExecError("Instance %s is not running (state %s)" %
8069 (instance.name, state))
8071 logging.debug("Connecting to console of %s on %s", instance.name, node)
8073 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
8076 def _GetInstanceConsole(cluster, instance):
8077 """Returns console information for an instance.
8079 @type cluster: L{objects.Cluster}
8080 @type instance: L{objects.Instance}
8084 hyper = hypervisor.GetHypervisor(instance.hypervisor)
8085 # beparams and hvparams are passed separately, to avoid editing the
8086 # instance and then saving the defaults in the instance itself.
8087 hvparams = cluster.FillHV(instance)
8088 beparams = cluster.FillBE(instance)
8089 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
8091 assert console.instance == instance.name
8092 assert console.Validate()
8094 return console.ToDict()
8097 class LUInstanceReplaceDisks(LogicalUnit):
8098 """Replace the disks of an instance.
8101 HPATH = "mirrors-replace"
8102 HTYPE = constants.HTYPE_INSTANCE
8105 def CheckArguments(self):
8106 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
8109 def ExpandNames(self):
8110 self._ExpandAndLockInstance()
8112 if self.op.iallocator is not None:
8113 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8115 elif self.op.remote_node is not None:
8116 remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8117 self.op.remote_node = remote_node
8119 # Warning: do not remove the locking of the new secondary here
8120 # unless DRBD8.AddChildren is changed to work in parallel;
8121 # currently it doesn't since parallel invocations of
8122 # FindUnusedMinor will conflict
8123 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
8124 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
8127 self.needed_locks[locking.LEVEL_NODE] = []
8128 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8130 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
8131 self.op.iallocator, self.op.remote_node,
8132 self.op.disks, False, self.op.early_release)
8134 self.tasklets = [self.replacer]
8136 def DeclareLocks(self, level):
8137 # If we're not already locking all nodes in the set we have to declare the
8138 # instance's primary/secondary nodes.
8139 if (level == locking.LEVEL_NODE and
8140 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
8141 self._LockInstancesNodes()
8143 def BuildHooksEnv(self):
8146 This runs on the master, the primary and all the secondaries.
8149 instance = self.replacer.instance
8151 "MODE": self.op.mode,
8152 "NEW_SECONDARY": self.op.remote_node,
8153 "OLD_SECONDARY": instance.secondary_nodes[0],
8155 env.update(_BuildInstanceHookEnvByObject(self, instance))
8157 self.cfg.GetMasterNode(),
8158 instance.primary_node,
8160 if self.op.remote_node is not None:
8161 nl.append(self.op.remote_node)
8165 class TLReplaceDisks(Tasklet):
8166 """Replaces disks for an instance.
8168 Note: Locking is not within the scope of this class.
8171 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
8172 disks, delay_iallocator, early_release):
8173 """Initializes this class.
8176 Tasklet.__init__(self, lu)
8179 self.instance_name = instance_name
8181 self.iallocator_name = iallocator_name
8182 self.remote_node = remote_node
8184 self.delay_iallocator = delay_iallocator
8185 self.early_release = early_release
8188 self.instance = None
8189 self.new_node = None
8190 self.target_node = None
8191 self.other_node = None
8192 self.remote_node_info = None
8193 self.node_secondary_ip = None
8196 def CheckArguments(mode, remote_node, iallocator):
8197 """Helper function for users of this class.
8200 # check for valid parameter combination
8201 if mode == constants.REPLACE_DISK_CHG:
8202 if remote_node is None and iallocator is None:
8203 raise errors.OpPrereqError("When changing the secondary either an"
8204 " iallocator script must be used or the"
8205 " new node given", errors.ECODE_INVAL)
8207 if remote_node is not None and iallocator is not None:
8208 raise errors.OpPrereqError("Give either the iallocator or the new"
8209 " secondary, not both", errors.ECODE_INVAL)
8211 elif remote_node is not None or iallocator is not None:
8212 # Not replacing the secondary
8213 raise errors.OpPrereqError("The iallocator and new node options can"
8214 " only be used when changing the"
8215 " secondary node", errors.ECODE_INVAL)
8218 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
8219 """Compute a new secondary node using an IAllocator.
8222 ial = IAllocator(lu.cfg, lu.rpc,
8223 mode=constants.IALLOCATOR_MODE_RELOC,
8225 relocate_from=relocate_from)
8227 ial.Run(iallocator_name)
8230 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
8231 " %s" % (iallocator_name, ial.info),
8234 if len(ial.result) != ial.required_nodes:
8235 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8236 " of nodes (%s), required %s" %
8238 len(ial.result), ial.required_nodes),
8241 remote_node_name = ial.result[0]
8243 lu.LogInfo("Selected new secondary for instance '%s': %s",
8244 instance_name, remote_node_name)
8246 return remote_node_name
8248 def _FindFaultyDisks(self, node_name):
8249 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
8252 def CheckPrereq(self):
8253 """Check prerequisites.
8255 This checks that the instance is in the cluster.
8258 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
8259 assert instance is not None, \
8260 "Cannot retrieve locked instance %s" % self.instance_name
8262 if instance.disk_template != constants.DT_DRBD8:
8263 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
8264 " instances", errors.ECODE_INVAL)
8266 if len(instance.secondary_nodes) != 1:
8267 raise errors.OpPrereqError("The instance has a strange layout,"
8268 " expected one secondary but found %d" %
8269 len(instance.secondary_nodes),
8272 if not self.delay_iallocator:
8273 self._CheckPrereq2()
8275 def _CheckPrereq2(self):
8276 """Check prerequisites, second part.
8278 This function should always be part of CheckPrereq. It was separated and is
8279 now called from Exec because during node evacuation iallocator was only
8280 called with an unmodified cluster model, not taking planned changes into
8284 instance = self.instance
8285 secondary_node = instance.secondary_nodes[0]
8287 if self.iallocator_name is None:
8288 remote_node = self.remote_node
8290 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
8291 instance.name, instance.secondary_nodes)
8293 if remote_node is not None:
8294 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
8295 assert self.remote_node_info is not None, \
8296 "Cannot retrieve locked node %s" % remote_node
8298 self.remote_node_info = None
8300 if remote_node == self.instance.primary_node:
8301 raise errors.OpPrereqError("The specified node is the primary node of"
8302 " the instance.", errors.ECODE_INVAL)
8304 if remote_node == secondary_node:
8305 raise errors.OpPrereqError("The specified node is already the"
8306 " secondary node of the instance.",
8309 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
8310 constants.REPLACE_DISK_CHG):
8311 raise errors.OpPrereqError("Cannot specify disks to be replaced",
8314 if self.mode == constants.REPLACE_DISK_AUTO:
8315 faulty_primary = self._FindFaultyDisks(instance.primary_node)
8316 faulty_secondary = self._FindFaultyDisks(secondary_node)
8318 if faulty_primary and faulty_secondary:
8319 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
8320 " one node and can not be repaired"
8321 " automatically" % self.instance_name,
8325 self.disks = faulty_primary
8326 self.target_node = instance.primary_node
8327 self.other_node = secondary_node
8328 check_nodes = [self.target_node, self.other_node]
8329 elif faulty_secondary:
8330 self.disks = faulty_secondary
8331 self.target_node = secondary_node
8332 self.other_node = instance.primary_node
8333 check_nodes = [self.target_node, self.other_node]
8339 # Non-automatic modes
8340 if self.mode == constants.REPLACE_DISK_PRI:
8341 self.target_node = instance.primary_node
8342 self.other_node = secondary_node
8343 check_nodes = [self.target_node, self.other_node]
8345 elif self.mode == constants.REPLACE_DISK_SEC:
8346 self.target_node = secondary_node
8347 self.other_node = instance.primary_node
8348 check_nodes = [self.target_node, self.other_node]
8350 elif self.mode == constants.REPLACE_DISK_CHG:
8351 self.new_node = remote_node
8352 self.other_node = instance.primary_node
8353 self.target_node = secondary_node
8354 check_nodes = [self.new_node, self.other_node]
8356 _CheckNodeNotDrained(self.lu, remote_node)
8357 _CheckNodeVmCapable(self.lu, remote_node)
8359 old_node_info = self.cfg.GetNodeInfo(secondary_node)
8360 assert old_node_info is not None
8361 if old_node_info.offline and not self.early_release:
8362 # doesn't make sense to delay the release
8363 self.early_release = True
8364 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
8365 " early-release mode", secondary_node)
8368 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
8371 # If not specified all disks should be replaced
8373 self.disks = range(len(self.instance.disks))
8375 for node in check_nodes:
8376 _CheckNodeOnline(self.lu, node)
8378 # Check whether disks are valid
8379 for disk_idx in self.disks:
8380 instance.FindDisk(disk_idx)
8382 # Get secondary node IP addresses
8385 for node_name in [self.target_node, self.other_node, self.new_node]:
8386 if node_name is not None:
8387 node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
8389 self.node_secondary_ip = node_2nd_ip
8391 def Exec(self, feedback_fn):
8392 """Execute disk replacement.
8394 This dispatches the disk replacement to the appropriate handler.
8397 if self.delay_iallocator:
8398 self._CheckPrereq2()
8401 feedback_fn("No disks need replacement")
8404 feedback_fn("Replacing disk(s) %s for %s" %
8405 (utils.CommaJoin(self.disks), self.instance.name))
8407 activate_disks = (not self.instance.admin_up)
8409 # Activate the instance disks if we're replacing them on a down instance
8411 _StartInstanceDisks(self.lu, self.instance, True)
8414 # Should we replace the secondary node?
8415 if self.new_node is not None:
8416 fn = self._ExecDrbd8Secondary
8418 fn = self._ExecDrbd8DiskOnly
8420 return fn(feedback_fn)
8423 # Deactivate the instance disks if we're replacing them on a
8426 _SafeShutdownInstanceDisks(self.lu, self.instance)
8428 def _CheckVolumeGroup(self, nodes):
8429 self.lu.LogInfo("Checking volume groups")
8431 vgname = self.cfg.GetVGName()
8433 # Make sure volume group exists on all involved nodes
8434 results = self.rpc.call_vg_list(nodes)
8436 raise errors.OpExecError("Can't list volume groups on the nodes")
8440 res.Raise("Error checking node %s" % node)
8441 if vgname not in res.payload:
8442 raise errors.OpExecError("Volume group '%s' not found on node %s" %
8445 def _CheckDisksExistence(self, nodes):
8446 # Check disk existence
8447 for idx, dev in enumerate(self.instance.disks):
8448 if idx not in self.disks:
8452 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
8453 self.cfg.SetDiskID(dev, node)
8455 result = self.rpc.call_blockdev_find(node, dev)
8457 msg = result.fail_msg
8458 if msg or not result.payload:
8460 msg = "disk not found"
8461 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
8464 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
8465 for idx, dev in enumerate(self.instance.disks):
8466 if idx not in self.disks:
8469 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
8472 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
8474 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
8475 " replace disks for instance %s" %
8476 (node_name, self.instance.name))
8478 def _CreateNewStorage(self, node_name):
8479 vgname = self.cfg.GetVGName()
8482 for idx, dev in enumerate(self.instance.disks):
8483 if idx not in self.disks:
8486 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
8488 self.cfg.SetDiskID(dev, node_name)
8490 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
8491 names = _GenerateUniqueNames(self.lu, lv_names)
8493 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
8494 logical_id=(vgname, names[0]))
8495 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
8496 logical_id=(vgname, names[1]))
8498 new_lvs = [lv_data, lv_meta]
8499 old_lvs = dev.children
8500 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
8502 # we pass force_create=True to force the LVM creation
8503 for new_lv in new_lvs:
8504 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
8505 _GetInstanceInfoText(self.instance), False)
8509 def _CheckDevices(self, node_name, iv_names):
8510 for name, (dev, _, _) in iv_names.iteritems():
8511 self.cfg.SetDiskID(dev, node_name)
8513 result = self.rpc.call_blockdev_find(node_name, dev)
8515 msg = result.fail_msg
8516 if msg or not result.payload:
8518 msg = "disk not found"
8519 raise errors.OpExecError("Can't find DRBD device %s: %s" %
8522 if result.payload.is_degraded:
8523 raise errors.OpExecError("DRBD device %s is degraded!" % name)
8525 def _RemoveOldStorage(self, node_name, iv_names):
8526 for name, (_, old_lvs, _) in iv_names.iteritems():
8527 self.lu.LogInfo("Remove logical volumes for %s" % name)
8530 self.cfg.SetDiskID(lv, node_name)
8532 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
8534 self.lu.LogWarning("Can't remove old LV: %s" % msg,
8535 hint="remove unused LVs manually")
8537 def _ReleaseNodeLock(self, node_name):
8538 """Releases the lock for a given node."""
8539 self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
8541 def _ExecDrbd8DiskOnly(self, feedback_fn):
8542 """Replace a disk on the primary or secondary for DRBD 8.
8544 The algorithm for replace is quite complicated:
8546 1. for each disk to be replaced:
8548 1. create new LVs on the target node with unique names
8549 1. detach old LVs from the drbd device
8550 1. rename old LVs to name_replaced.<time_t>
8551 1. rename new LVs to old LVs
8552 1. attach the new LVs (with the old names now) to the drbd device
8554 1. wait for sync across all devices
8556 1. for each modified disk:
8558 1. remove old LVs (which have the name name_replaces.<time_t>)
8560 Failures are not very well handled.
8565 # Step: check device activation
8566 self.lu.LogStep(1, steps_total, "Check device existence")
8567 self._CheckDisksExistence([self.other_node, self.target_node])
8568 self._CheckVolumeGroup([self.target_node, self.other_node])
8570 # Step: check other node consistency
8571 self.lu.LogStep(2, steps_total, "Check peer consistency")
8572 self._CheckDisksConsistency(self.other_node,
8573 self.other_node == self.instance.primary_node,
8576 # Step: create new storage
8577 self.lu.LogStep(3, steps_total, "Allocate new storage")
8578 iv_names = self._CreateNewStorage(self.target_node)
8580 # Step: for each lv, detach+rename*2+attach
8581 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8582 for dev, old_lvs, new_lvs in iv_names.itervalues():
8583 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8585 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8587 result.Raise("Can't detach drbd from local storage on node"
8588 " %s for device %s" % (self.target_node, dev.iv_name))
8590 #cfg.Update(instance)
8592 # ok, we created the new LVs, so now we know we have the needed
8593 # storage; as such, we proceed on the target node to rename
8594 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8595 # using the assumption that logical_id == physical_id (which in
8596 # turn is the unique_id on that node)
8598 # FIXME(iustin): use a better name for the replaced LVs
8599 temp_suffix = int(time.time())
8600 ren_fn = lambda d, suff: (d.physical_id[0],
8601 d.physical_id[1] + "_replaced-%s" % suff)
8603 # Build the rename list based on what LVs exist on the node
8604 rename_old_to_new = []
8605 for to_ren in old_lvs:
8606 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8607 if not result.fail_msg and result.payload:
8609 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8611 self.lu.LogInfo("Renaming the old LVs on the target node")
8612 result = self.rpc.call_blockdev_rename(self.target_node,
8614 result.Raise("Can't rename old LVs on node %s" % self.target_node)
8616 # Now we rename the new LVs to the old LVs
8617 self.lu.LogInfo("Renaming the new LVs on the target node")
8618 rename_new_to_old = [(new, old.physical_id)
8619 for old, new in zip(old_lvs, new_lvs)]
8620 result = self.rpc.call_blockdev_rename(self.target_node,
8622 result.Raise("Can't rename new LVs on node %s" % self.target_node)
8624 for old, new in zip(old_lvs, new_lvs):
8625 new.logical_id = old.logical_id
8626 self.cfg.SetDiskID(new, self.target_node)
8628 for disk in old_lvs:
8629 disk.logical_id = ren_fn(disk, temp_suffix)
8630 self.cfg.SetDiskID(disk, self.target_node)
8632 # Now that the new lvs have the old name, we can add them to the device
8633 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8634 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8636 msg = result.fail_msg
8638 for new_lv in new_lvs:
8639 msg2 = self.rpc.call_blockdev_remove(self.target_node,
8642 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8643 hint=("cleanup manually the unused logical"
8645 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8647 dev.children = new_lvs
8649 self.cfg.Update(self.instance, feedback_fn)
8652 if self.early_release:
8653 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8655 self._RemoveOldStorage(self.target_node, iv_names)
8656 # WARNING: we release both node locks here, do not do other RPCs
8657 # than WaitForSync to the primary node
8658 self._ReleaseNodeLock([self.target_node, self.other_node])
8661 # This can fail as the old devices are degraded and _WaitForSync
8662 # does a combined result over all disks, so we don't check its return value
8663 self.lu.LogStep(cstep, steps_total, "Sync devices")
8665 _WaitForSync(self.lu, self.instance)
8667 # Check all devices manually
8668 self._CheckDevices(self.instance.primary_node, iv_names)
8670 # Step: remove old storage
8671 if not self.early_release:
8672 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8674 self._RemoveOldStorage(self.target_node, iv_names)
8676 def _ExecDrbd8Secondary(self, feedback_fn):
8677 """Replace the secondary node for DRBD 8.
8679 The algorithm for replace is quite complicated:
8680 - for all disks of the instance:
8681 - create new LVs on the new node with same names
8682 - shutdown the drbd device on the old secondary
8683 - disconnect the drbd network on the primary
8684 - create the drbd device on the new secondary
8685 - network attach the drbd on the primary, using an artifice:
8686 the drbd code for Attach() will connect to the network if it
8687 finds a device which is connected to the good local disks but
8689 - wait for sync across all devices
8690 - remove all disks from the old secondary
8692 Failures are not very well handled.
8697 # Step: check device activation
8698 self.lu.LogStep(1, steps_total, "Check device existence")
8699 self._CheckDisksExistence([self.instance.primary_node])
8700 self._CheckVolumeGroup([self.instance.primary_node])
8702 # Step: check other node consistency
8703 self.lu.LogStep(2, steps_total, "Check peer consistency")
8704 self._CheckDisksConsistency(self.instance.primary_node, True, True)
8706 # Step: create new storage
8707 self.lu.LogStep(3, steps_total, "Allocate new storage")
8708 for idx, dev in enumerate(self.instance.disks):
8709 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8710 (self.new_node, idx))
8711 # we pass force_create=True to force LVM creation
8712 for new_lv in dev.children:
8713 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8714 _GetInstanceInfoText(self.instance), False)
8716 # Step 4: dbrd minors and drbd setups changes
8717 # after this, we must manually remove the drbd minors on both the
8718 # error and the success paths
8719 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8720 minors = self.cfg.AllocateDRBDMinor([self.new_node
8721 for dev in self.instance.disks],
8723 logging.debug("Allocated minors %r", minors)
8726 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8727 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8728 (self.new_node, idx))
8729 # create new devices on new_node; note that we create two IDs:
8730 # one without port, so the drbd will be activated without
8731 # networking information on the new node at this stage, and one
8732 # with network, for the latter activation in step 4
8733 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8734 if self.instance.primary_node == o_node1:
8737 assert self.instance.primary_node == o_node2, "Three-node instance?"
8740 new_alone_id = (self.instance.primary_node, self.new_node, None,
8741 p_minor, new_minor, o_secret)
8742 new_net_id = (self.instance.primary_node, self.new_node, o_port,
8743 p_minor, new_minor, o_secret)
8745 iv_names[idx] = (dev, dev.children, new_net_id)
8746 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8748 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8749 logical_id=new_alone_id,
8750 children=dev.children,
8753 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8754 _GetInstanceInfoText(self.instance), False)
8755 except errors.GenericError:
8756 self.cfg.ReleaseDRBDMinors(self.instance.name)
8759 # We have new devices, shutdown the drbd on the old secondary
8760 for idx, dev in enumerate(self.instance.disks):
8761 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8762 self.cfg.SetDiskID(dev, self.target_node)
8763 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8765 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8766 "node: %s" % (idx, msg),
8767 hint=("Please cleanup this device manually as"
8768 " soon as possible"))
8770 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8771 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8772 self.node_secondary_ip,
8773 self.instance.disks)\
8774 [self.instance.primary_node]
8776 msg = result.fail_msg
8778 # detaches didn't succeed (unlikely)
8779 self.cfg.ReleaseDRBDMinors(self.instance.name)
8780 raise errors.OpExecError("Can't detach the disks from the network on"
8781 " old node: %s" % (msg,))
8783 # if we managed to detach at least one, we update all the disks of
8784 # the instance to point to the new secondary
8785 self.lu.LogInfo("Updating instance configuration")
8786 for dev, _, new_logical_id in iv_names.itervalues():
8787 dev.logical_id = new_logical_id
8788 self.cfg.SetDiskID(dev, self.instance.primary_node)
8790 self.cfg.Update(self.instance, feedback_fn)
8792 # and now perform the drbd attach
8793 self.lu.LogInfo("Attaching primary drbds to new secondary"
8794 " (standalone => connected)")
8795 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8797 self.node_secondary_ip,
8798 self.instance.disks,
8801 for to_node, to_result in result.items():
8802 msg = to_result.fail_msg
8804 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8806 hint=("please do a gnt-instance info to see the"
8807 " status of disks"))
8809 if self.early_release:
8810 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8812 self._RemoveOldStorage(self.target_node, iv_names)
8813 # WARNING: we release all node locks here, do not do other RPCs
8814 # than WaitForSync to the primary node
8815 self._ReleaseNodeLock([self.instance.primary_node,
8820 # This can fail as the old devices are degraded and _WaitForSync
8821 # does a combined result over all disks, so we don't check its return value
8822 self.lu.LogStep(cstep, steps_total, "Sync devices")
8824 _WaitForSync(self.lu, self.instance)
8826 # Check all devices manually
8827 self._CheckDevices(self.instance.primary_node, iv_names)
8829 # Step: remove old storage
8830 if not self.early_release:
8831 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8832 self._RemoveOldStorage(self.target_node, iv_names)
8835 class LURepairNodeStorage(NoHooksLU):
8836 """Repairs the volume group on a node.
8841 def CheckArguments(self):
8842 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8844 storage_type = self.op.storage_type
8846 if (constants.SO_FIX_CONSISTENCY not in
8847 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8848 raise errors.OpPrereqError("Storage units of type '%s' can not be"
8849 " repaired" % storage_type,
8852 def ExpandNames(self):
8853 self.needed_locks = {
8854 locking.LEVEL_NODE: [self.op.node_name],
8857 def _CheckFaultyDisks(self, instance, node_name):
8858 """Ensure faulty disks abort the opcode or at least warn."""
8860 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8862 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8863 " node '%s'" % (instance.name, node_name),
8865 except errors.OpPrereqError, err:
8866 if self.op.ignore_consistency:
8867 self.proc.LogWarning(str(err.args[0]))
8871 def CheckPrereq(self):
8872 """Check prerequisites.
8875 # Check whether any instance on this node has faulty disks
8876 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8877 if not inst.admin_up:
8879 check_nodes = set(inst.all_nodes)
8880 check_nodes.discard(self.op.node_name)
8881 for inst_node_name in check_nodes:
8882 self._CheckFaultyDisks(inst, inst_node_name)
8884 def Exec(self, feedback_fn):
8885 feedback_fn("Repairing storage unit '%s' on %s ..." %
8886 (self.op.name, self.op.node_name))
8888 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8889 result = self.rpc.call_storage_execute(self.op.node_name,
8890 self.op.storage_type, st_args,
8892 constants.SO_FIX_CONSISTENCY)
8893 result.Raise("Failed to repair storage unit '%s' on %s" %
8894 (self.op.name, self.op.node_name))
8897 class LUNodeEvacStrategy(NoHooksLU):
8898 """Computes the node evacuation strategy.
8903 def CheckArguments(self):
8904 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8906 def ExpandNames(self):
8907 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8908 self.needed_locks = locks = {}
8909 if self.op.remote_node is None:
8910 locks[locking.LEVEL_NODE] = locking.ALL_SET
8912 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8913 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8915 def Exec(self, feedback_fn):
8916 if self.op.remote_node is not None:
8918 for node in self.op.nodes:
8919 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8922 if i.primary_node == self.op.remote_node:
8923 raise errors.OpPrereqError("Node %s is the primary node of"
8924 " instance %s, cannot use it as"
8926 (self.op.remote_node, i.name),
8928 result.append([i.name, self.op.remote_node])
8930 ial = IAllocator(self.cfg, self.rpc,
8931 mode=constants.IALLOCATOR_MODE_MEVAC,
8932 evac_nodes=self.op.nodes)
8933 ial.Run(self.op.iallocator, validate=True)
8935 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8941 class LUInstanceGrowDisk(LogicalUnit):
8942 """Grow a disk of an instance.
8946 HTYPE = constants.HTYPE_INSTANCE
8949 def ExpandNames(self):
8950 self._ExpandAndLockInstance()
8951 self.needed_locks[locking.LEVEL_NODE] = []
8952 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8954 def DeclareLocks(self, level):
8955 if level == locking.LEVEL_NODE:
8956 self._LockInstancesNodes()
8958 def BuildHooksEnv(self):
8961 This runs on the master, the primary and all the secondaries.
8965 "DISK": self.op.disk,
8966 "AMOUNT": self.op.amount,
8968 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8969 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8972 def CheckPrereq(self):
8973 """Check prerequisites.
8975 This checks that the instance is in the cluster.
8978 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8979 assert instance is not None, \
8980 "Cannot retrieve locked instance %s" % self.op.instance_name
8981 nodenames = list(instance.all_nodes)
8982 for node in nodenames:
8983 _CheckNodeOnline(self, node)
8985 self.instance = instance
8987 if instance.disk_template not in constants.DTS_GROWABLE:
8988 raise errors.OpPrereqError("Instance's disk layout does not support"
8989 " growing.", errors.ECODE_INVAL)
8991 self.disk = instance.FindDisk(self.op.disk)
8993 if instance.disk_template not in (constants.DT_FILE,
8994 constants.DT_SHARED_FILE):
8995 # TODO: check the free disk space for file, when that feature will be
8997 _CheckNodesFreeDiskPerVG(self, nodenames,
8998 self.disk.ComputeGrowth(self.op.amount))
9000 def Exec(self, feedback_fn):
9001 """Execute disk grow.
9004 instance = self.instance
9007 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
9009 raise errors.OpExecError("Cannot activate block device to grow")
9011 for node in instance.all_nodes:
9012 self.cfg.SetDiskID(disk, node)
9013 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
9014 result.Raise("Grow request failed to node %s" % node)
9016 # TODO: Rewrite code to work properly
9017 # DRBD goes into sync mode for a short amount of time after executing the
9018 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
9019 # calling "resize" in sync mode fails. Sleeping for a short amount of
9020 # time is a work-around.
9023 disk.RecordGrow(self.op.amount)
9024 self.cfg.Update(instance, feedback_fn)
9025 if self.op.wait_for_sync:
9026 disk_abort = not _WaitForSync(self, instance, disks=[disk])
9028 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
9029 " status.\nPlease check the instance.")
9030 if not instance.admin_up:
9031 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
9032 elif not instance.admin_up:
9033 self.proc.LogWarning("Not shutting down the disk even if the instance is"
9034 " not supposed to be running because no wait for"
9035 " sync mode was requested.")
9038 class LUInstanceQueryData(NoHooksLU):
9039 """Query runtime instance data.
9044 def ExpandNames(self):
9045 self.needed_locks = {}
9046 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9048 if self.op.instances:
9049 self.wanted_names = []
9050 for name in self.op.instances:
9051 full_name = _ExpandInstanceName(self.cfg, name)
9052 self.wanted_names.append(full_name)
9053 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
9055 self.wanted_names = None
9056 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
9058 self.needed_locks[locking.LEVEL_NODE] = []
9059 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9061 def DeclareLocks(self, level):
9062 if level == locking.LEVEL_NODE:
9063 self._LockInstancesNodes()
9065 def CheckPrereq(self):
9066 """Check prerequisites.
9068 This only checks the optional instance list against the existing names.
9071 if self.wanted_names is None:
9072 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
9074 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
9075 in self.wanted_names]
9077 def _ComputeBlockdevStatus(self, node, instance_name, dev):
9078 """Returns the status of a block device
9081 if self.op.static or not node:
9084 self.cfg.SetDiskID(dev, node)
9086 result = self.rpc.call_blockdev_find(node, dev)
9090 result.Raise("Can't compute disk status for %s" % instance_name)
9092 status = result.payload
9096 return (status.dev_path, status.major, status.minor,
9097 status.sync_percent, status.estimated_time,
9098 status.is_degraded, status.ldisk_status)
9100 def _ComputeDiskStatus(self, instance, snode, dev):
9101 """Compute block device status.
9104 if dev.dev_type in constants.LDS_DRBD:
9105 # we change the snode then (otherwise we use the one passed in)
9106 if dev.logical_id[0] == instance.primary_node:
9107 snode = dev.logical_id[1]
9109 snode = dev.logical_id[0]
9111 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
9113 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
9116 dev_children = [self._ComputeDiskStatus(instance, snode, child)
9117 for child in dev.children]
9122 "iv_name": dev.iv_name,
9123 "dev_type": dev.dev_type,
9124 "logical_id": dev.logical_id,
9125 "physical_id": dev.physical_id,
9126 "pstatus": dev_pstatus,
9127 "sstatus": dev_sstatus,
9128 "children": dev_children,
9135 def Exec(self, feedback_fn):
9136 """Gather and return data"""
9139 cluster = self.cfg.GetClusterInfo()
9141 for instance in self.wanted_instances:
9142 if not self.op.static:
9143 remote_info = self.rpc.call_instance_info(instance.primary_node,
9145 instance.hypervisor)
9146 remote_info.Raise("Error checking node %s" % instance.primary_node)
9147 remote_info = remote_info.payload
9148 if remote_info and "state" in remote_info:
9151 remote_state = "down"
9154 if instance.admin_up:
9157 config_state = "down"
9159 disks = [self._ComputeDiskStatus(instance, None, device)
9160 for device in instance.disks]
9163 "name": instance.name,
9164 "config_state": config_state,
9165 "run_state": remote_state,
9166 "pnode": instance.primary_node,
9167 "snodes": instance.secondary_nodes,
9169 # this happens to be the same format used for hooks
9170 "nics": _NICListToTuple(self, instance.nics),
9171 "disk_template": instance.disk_template,
9173 "hypervisor": instance.hypervisor,
9174 "network_port": instance.network_port,
9175 "hv_instance": instance.hvparams,
9176 "hv_actual": cluster.FillHV(instance, skip_globals=True),
9177 "be_instance": instance.beparams,
9178 "be_actual": cluster.FillBE(instance),
9179 "os_instance": instance.osparams,
9180 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
9181 "serial_no": instance.serial_no,
9182 "mtime": instance.mtime,
9183 "ctime": instance.ctime,
9184 "uuid": instance.uuid,
9187 result[instance.name] = idict
9192 class LUInstanceSetParams(LogicalUnit):
9193 """Modifies an instances's parameters.
9196 HPATH = "instance-modify"
9197 HTYPE = constants.HTYPE_INSTANCE
9200 def CheckArguments(self):
9201 if not (self.op.nics or self.op.disks or self.op.disk_template or
9202 self.op.hvparams or self.op.beparams or self.op.os_name):
9203 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
9205 if self.op.hvparams:
9206 _CheckGlobalHvParams(self.op.hvparams)
9210 for disk_op, disk_dict in self.op.disks:
9211 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
9212 if disk_op == constants.DDM_REMOVE:
9215 elif disk_op == constants.DDM_ADD:
9218 if not isinstance(disk_op, int):
9219 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
9220 if not isinstance(disk_dict, dict):
9221 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
9222 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9224 if disk_op == constants.DDM_ADD:
9225 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
9226 if mode not in constants.DISK_ACCESS_SET:
9227 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
9229 size = disk_dict.get('size', None)
9231 raise errors.OpPrereqError("Required disk parameter size missing",
9235 except (TypeError, ValueError), err:
9236 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
9237 str(err), errors.ECODE_INVAL)
9238 disk_dict['size'] = size
9240 # modification of disk
9241 if 'size' in disk_dict:
9242 raise errors.OpPrereqError("Disk size change not possible, use"
9243 " grow-disk", errors.ECODE_INVAL)
9245 if disk_addremove > 1:
9246 raise errors.OpPrereqError("Only one disk add or remove operation"
9247 " supported at a time", errors.ECODE_INVAL)
9249 if self.op.disks and self.op.disk_template is not None:
9250 raise errors.OpPrereqError("Disk template conversion and other disk"
9251 " changes not supported at the same time",
9254 if (self.op.disk_template and
9255 self.op.disk_template in constants.DTS_NET_MIRROR and
9256 self.op.remote_node is None):
9257 raise errors.OpPrereqError("Changing the disk template to a mirrored"
9258 " one requires specifying a secondary node",
9263 for nic_op, nic_dict in self.op.nics:
9264 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
9265 if nic_op == constants.DDM_REMOVE:
9268 elif nic_op == constants.DDM_ADD:
9271 if not isinstance(nic_op, int):
9272 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
9273 if not isinstance(nic_dict, dict):
9274 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
9275 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9277 # nic_dict should be a dict
9278 nic_ip = nic_dict.get('ip', None)
9279 if nic_ip is not None:
9280 if nic_ip.lower() == constants.VALUE_NONE:
9281 nic_dict['ip'] = None
9283 if not netutils.IPAddress.IsValid(nic_ip):
9284 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
9287 nic_bridge = nic_dict.get('bridge', None)
9288 nic_link = nic_dict.get('link', None)
9289 if nic_bridge and nic_link:
9290 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
9291 " at the same time", errors.ECODE_INVAL)
9292 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
9293 nic_dict['bridge'] = None
9294 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
9295 nic_dict['link'] = None
9297 if nic_op == constants.DDM_ADD:
9298 nic_mac = nic_dict.get('mac', None)
9300 nic_dict['mac'] = constants.VALUE_AUTO
9302 if 'mac' in nic_dict:
9303 nic_mac = nic_dict['mac']
9304 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9305 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
9307 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
9308 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
9309 " modifying an existing nic",
9312 if nic_addremove > 1:
9313 raise errors.OpPrereqError("Only one NIC add or remove operation"
9314 " supported at a time", errors.ECODE_INVAL)
9316 def ExpandNames(self):
9317 self._ExpandAndLockInstance()
9318 self.needed_locks[locking.LEVEL_NODE] = []
9319 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9321 def DeclareLocks(self, level):
9322 if level == locking.LEVEL_NODE:
9323 self._LockInstancesNodes()
9324 if self.op.disk_template and self.op.remote_node:
9325 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9326 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
9328 def BuildHooksEnv(self):
9331 This runs on the master, primary and secondaries.
9335 if constants.BE_MEMORY in self.be_new:
9336 args['memory'] = self.be_new[constants.BE_MEMORY]
9337 if constants.BE_VCPUS in self.be_new:
9338 args['vcpus'] = self.be_new[constants.BE_VCPUS]
9339 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
9340 # information at all.
9343 nic_override = dict(self.op.nics)
9344 for idx, nic in enumerate(self.instance.nics):
9345 if idx in nic_override:
9346 this_nic_override = nic_override[idx]
9348 this_nic_override = {}
9349 if 'ip' in this_nic_override:
9350 ip = this_nic_override['ip']
9353 if 'mac' in this_nic_override:
9354 mac = this_nic_override['mac']
9357 if idx in self.nic_pnew:
9358 nicparams = self.nic_pnew[idx]
9360 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
9361 mode = nicparams[constants.NIC_MODE]
9362 link = nicparams[constants.NIC_LINK]
9363 args['nics'].append((ip, mac, mode, link))
9364 if constants.DDM_ADD in nic_override:
9365 ip = nic_override[constants.DDM_ADD].get('ip', None)
9366 mac = nic_override[constants.DDM_ADD]['mac']
9367 nicparams = self.nic_pnew[constants.DDM_ADD]
9368 mode = nicparams[constants.NIC_MODE]
9369 link = nicparams[constants.NIC_LINK]
9370 args['nics'].append((ip, mac, mode, link))
9371 elif constants.DDM_REMOVE in nic_override:
9372 del args['nics'][-1]
9374 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
9375 if self.op.disk_template:
9376 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
9377 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9380 def CheckPrereq(self):
9381 """Check prerequisites.
9383 This only checks the instance list against the existing names.
9386 # checking the new params on the primary/secondary nodes
9388 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9389 cluster = self.cluster = self.cfg.GetClusterInfo()
9390 assert self.instance is not None, \
9391 "Cannot retrieve locked instance %s" % self.op.instance_name
9392 pnode = instance.primary_node
9393 nodelist = list(instance.all_nodes)
9396 if self.op.os_name and not self.op.force:
9397 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
9398 self.op.force_variant)
9399 instance_os = self.op.os_name
9401 instance_os = instance.os
9403 if self.op.disk_template:
9404 if instance.disk_template == self.op.disk_template:
9405 raise errors.OpPrereqError("Instance already has disk template %s" %
9406 instance.disk_template, errors.ECODE_INVAL)
9408 if (instance.disk_template,
9409 self.op.disk_template) not in self._DISK_CONVERSIONS:
9410 raise errors.OpPrereqError("Unsupported disk template conversion from"
9411 " %s to %s" % (instance.disk_template,
9412 self.op.disk_template),
9414 _CheckInstanceDown(self, instance, "cannot change disk template")
9415 if self.op.disk_template in constants.DTS_NET_MIRROR:
9416 if self.op.remote_node == pnode:
9417 raise errors.OpPrereqError("Given new secondary node %s is the same"
9418 " as the primary node of the instance" %
9419 self.op.remote_node, errors.ECODE_STATE)
9420 _CheckNodeOnline(self, self.op.remote_node)
9421 _CheckNodeNotDrained(self, self.op.remote_node)
9422 # FIXME: here we assume that the old instance type is DT_PLAIN
9423 assert instance.disk_template == constants.DT_PLAIN
9424 disks = [{"size": d.size, "vg": d.logical_id[0]}
9425 for d in instance.disks]
9426 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
9427 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
9429 # hvparams processing
9430 if self.op.hvparams:
9431 hv_type = instance.hypervisor
9432 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
9433 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
9434 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
9437 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
9438 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
9439 self.hv_new = hv_new # the new actual values
9440 self.hv_inst = i_hvdict # the new dict (without defaults)
9442 self.hv_new = self.hv_inst = {}
9444 # beparams processing
9445 if self.op.beparams:
9446 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
9448 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
9449 be_new = cluster.SimpleFillBE(i_bedict)
9450 self.be_new = be_new # the new actual values
9451 self.be_inst = i_bedict # the new dict (without defaults)
9453 self.be_new = self.be_inst = {}
9455 # osparams processing
9456 if self.op.osparams:
9457 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
9458 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
9459 self.os_inst = i_osdict # the new dict (without defaults)
9465 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
9466 mem_check_list = [pnode]
9467 if be_new[constants.BE_AUTO_BALANCE]:
9468 # either we changed auto_balance to yes or it was from before
9469 mem_check_list.extend(instance.secondary_nodes)
9470 instance_info = self.rpc.call_instance_info(pnode, instance.name,
9471 instance.hypervisor)
9472 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
9473 instance.hypervisor)
9474 pninfo = nodeinfo[pnode]
9475 msg = pninfo.fail_msg
9477 # Assume the primary node is unreachable and go ahead
9478 self.warn.append("Can't get info from primary node %s: %s" %
9480 elif not isinstance(pninfo.payload.get('memory_free', None), int):
9481 self.warn.append("Node data from primary node %s doesn't contain"
9482 " free memory information" % pnode)
9483 elif instance_info.fail_msg:
9484 self.warn.append("Can't get instance runtime information: %s" %
9485 instance_info.fail_msg)
9487 if instance_info.payload:
9488 current_mem = int(instance_info.payload['memory'])
9490 # Assume instance not running
9491 # (there is a slight race condition here, but it's not very probable,
9492 # and we have no other way to check)
9494 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
9495 pninfo.payload['memory_free'])
9497 raise errors.OpPrereqError("This change will prevent the instance"
9498 " from starting, due to %d MB of memory"
9499 " missing on its primary node" % miss_mem,
9502 if be_new[constants.BE_AUTO_BALANCE]:
9503 for node, nres in nodeinfo.items():
9504 if node not in instance.secondary_nodes:
9508 self.warn.append("Can't get info from secondary node %s: %s" %
9510 elif not isinstance(nres.payload.get('memory_free', None), int):
9511 self.warn.append("Secondary node %s didn't return free"
9512 " memory information" % node)
9513 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
9514 self.warn.append("Not enough memory to failover instance to"
9515 " secondary node %s" % node)
9520 for nic_op, nic_dict in self.op.nics:
9521 if nic_op == constants.DDM_REMOVE:
9522 if not instance.nics:
9523 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
9526 if nic_op != constants.DDM_ADD:
9528 if not instance.nics:
9529 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
9530 " no NICs" % nic_op,
9532 if nic_op < 0 or nic_op >= len(instance.nics):
9533 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
9535 (nic_op, len(instance.nics) - 1),
9537 old_nic_params = instance.nics[nic_op].nicparams
9538 old_nic_ip = instance.nics[nic_op].ip
9543 update_params_dict = dict([(key, nic_dict[key])
9544 for key in constants.NICS_PARAMETERS
9545 if key in nic_dict])
9547 if 'bridge' in nic_dict:
9548 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
9550 new_nic_params = _GetUpdatedParams(old_nic_params,
9552 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9553 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9554 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9555 self.nic_pinst[nic_op] = new_nic_params
9556 self.nic_pnew[nic_op] = new_filled_nic_params
9557 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9559 if new_nic_mode == constants.NIC_MODE_BRIDGED:
9560 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9561 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9563 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9565 self.warn.append(msg)
9567 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9568 if new_nic_mode == constants.NIC_MODE_ROUTED:
9569 if 'ip' in nic_dict:
9570 nic_ip = nic_dict['ip']
9574 raise errors.OpPrereqError('Cannot set the nic ip to None'
9575 ' on a routed nic', errors.ECODE_INVAL)
9576 if 'mac' in nic_dict:
9577 nic_mac = nic_dict['mac']
9579 raise errors.OpPrereqError('Cannot set the nic mac to None',
9581 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9582 # otherwise generate the mac
9583 nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9585 # or validate/reserve the current one
9587 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9588 except errors.ReservationError:
9589 raise errors.OpPrereqError("MAC address %s already in use"
9590 " in cluster" % nic_mac,
9591 errors.ECODE_NOTUNIQUE)
9594 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9595 raise errors.OpPrereqError("Disk operations not supported for"
9596 " diskless instances",
9598 for disk_op, _ in self.op.disks:
9599 if disk_op == constants.DDM_REMOVE:
9600 if len(instance.disks) == 1:
9601 raise errors.OpPrereqError("Cannot remove the last disk of"
9602 " an instance", errors.ECODE_INVAL)
9603 _CheckInstanceDown(self, instance, "cannot remove disks")
9605 if (disk_op == constants.DDM_ADD and
9606 len(instance.disks) >= constants.MAX_DISKS):
9607 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9608 " add more" % constants.MAX_DISKS,
9610 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9612 if disk_op < 0 or disk_op >= len(instance.disks):
9613 raise errors.OpPrereqError("Invalid disk index %s, valid values"
9615 (disk_op, len(instance.disks)),
9620 def _ConvertPlainToDrbd(self, feedback_fn):
9621 """Converts an instance from plain to drbd.
9624 feedback_fn("Converting template to drbd")
9625 instance = self.instance
9626 pnode = instance.primary_node
9627 snode = self.op.remote_node
9629 # create a fake disk info for _GenerateDiskTemplate
9630 disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9631 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9632 instance.name, pnode, [snode],
9633 disk_info, None, None, 0, feedback_fn)
9634 info = _GetInstanceInfoText(instance)
9635 feedback_fn("Creating aditional volumes...")
9636 # first, create the missing data and meta devices
9637 for disk in new_disks:
9638 # unfortunately this is... not too nice
9639 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9641 for child in disk.children:
9642 _CreateSingleBlockDev(self, snode, instance, child, info, True)
9643 # at this stage, all new LVs have been created, we can rename the
9645 feedback_fn("Renaming original volumes...")
9646 rename_list = [(o, n.children[0].logical_id)
9647 for (o, n) in zip(instance.disks, new_disks)]
9648 result = self.rpc.call_blockdev_rename(pnode, rename_list)
9649 result.Raise("Failed to rename original LVs")
9651 feedback_fn("Initializing DRBD devices...")
9652 # all child devices are in place, we can now create the DRBD devices
9653 for disk in new_disks:
9654 for node in [pnode, snode]:
9655 f_create = node == pnode
9656 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9658 # at this point, the instance has been modified
9659 instance.disk_template = constants.DT_DRBD8
9660 instance.disks = new_disks
9661 self.cfg.Update(instance, feedback_fn)
9663 # disks are created, waiting for sync
9664 disk_abort = not _WaitForSync(self, instance)
9666 raise errors.OpExecError("There are some degraded disks for"
9667 " this instance, please cleanup manually")
9669 def _ConvertDrbdToPlain(self, feedback_fn):
9670 """Converts an instance from drbd to plain.
9673 instance = self.instance
9674 assert len(instance.secondary_nodes) == 1
9675 pnode = instance.primary_node
9676 snode = instance.secondary_nodes[0]
9677 feedback_fn("Converting template to plain")
9679 old_disks = instance.disks
9680 new_disks = [d.children[0] for d in old_disks]
9682 # copy over size and mode
9683 for parent, child in zip(old_disks, new_disks):
9684 child.size = parent.size
9685 child.mode = parent.mode
9687 # update instance structure
9688 instance.disks = new_disks
9689 instance.disk_template = constants.DT_PLAIN
9690 self.cfg.Update(instance, feedback_fn)
9692 feedback_fn("Removing volumes on the secondary node...")
9693 for disk in old_disks:
9694 self.cfg.SetDiskID(disk, snode)
9695 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9697 self.LogWarning("Could not remove block device %s on node %s,"
9698 " continuing anyway: %s", disk.iv_name, snode, msg)
9700 feedback_fn("Removing unneeded volumes on the primary node...")
9701 for idx, disk in enumerate(old_disks):
9702 meta = disk.children[1]
9703 self.cfg.SetDiskID(meta, pnode)
9704 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9706 self.LogWarning("Could not remove metadata for disk %d on node %s,"
9707 " continuing anyway: %s", idx, pnode, msg)
9709 def Exec(self, feedback_fn):
9710 """Modifies an instance.
9712 All parameters take effect only at the next restart of the instance.
9715 # Process here the warnings from CheckPrereq, as we don't have a
9716 # feedback_fn there.
9717 for warn in self.warn:
9718 feedback_fn("WARNING: %s" % warn)
9721 instance = self.instance
9723 for disk_op, disk_dict in self.op.disks:
9724 if disk_op == constants.DDM_REMOVE:
9725 # remove the last disk
9726 device = instance.disks.pop()
9727 device_idx = len(instance.disks)
9728 for node, disk in device.ComputeNodeTree(instance.primary_node):
9729 self.cfg.SetDiskID(disk, node)
9730 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9732 self.LogWarning("Could not remove disk/%d on node %s: %s,"
9733 " continuing anyway", device_idx, node, msg)
9734 result.append(("disk/%d" % device_idx, "remove"))
9735 elif disk_op == constants.DDM_ADD:
9737 if instance.disk_template in (constants.DT_FILE,
9738 constants.DT_SHARED_FILE):
9739 file_driver, file_path = instance.disks[0].logical_id
9740 file_path = os.path.dirname(file_path)
9742 file_driver = file_path = None
9743 disk_idx_base = len(instance.disks)
9744 new_disk = _GenerateDiskTemplate(self,
9745 instance.disk_template,
9746 instance.name, instance.primary_node,
9747 instance.secondary_nodes,
9751 disk_idx_base, feedback_fn)[0]
9752 instance.disks.append(new_disk)
9753 info = _GetInstanceInfoText(instance)
9755 logging.info("Creating volume %s for instance %s",
9756 new_disk.iv_name, instance.name)
9757 # Note: this needs to be kept in sync with _CreateDisks
9759 for node in instance.all_nodes:
9760 f_create = node == instance.primary_node
9762 _CreateBlockDev(self, node, instance, new_disk,
9763 f_create, info, f_create)
9764 except errors.OpExecError, err:
9765 self.LogWarning("Failed to create volume %s (%s) on"
9767 new_disk.iv_name, new_disk, node, err)
9768 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9769 (new_disk.size, new_disk.mode)))
9771 # change a given disk
9772 instance.disks[disk_op].mode = disk_dict['mode']
9773 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9775 if self.op.disk_template:
9776 r_shut = _ShutdownInstanceDisks(self, instance)
9778 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
9779 " proceed with disk template conversion")
9780 mode = (instance.disk_template, self.op.disk_template)
9782 self._DISK_CONVERSIONS[mode](self, feedback_fn)
9784 self.cfg.ReleaseDRBDMinors(instance.name)
9786 result.append(("disk_template", self.op.disk_template))
9789 for nic_op, nic_dict in self.op.nics:
9790 if nic_op == constants.DDM_REMOVE:
9791 # remove the last nic
9792 del instance.nics[-1]
9793 result.append(("nic.%d" % len(instance.nics), "remove"))
9794 elif nic_op == constants.DDM_ADD:
9795 # mac and bridge should be set, by now
9796 mac = nic_dict['mac']
9797 ip = nic_dict.get('ip', None)
9798 nicparams = self.nic_pinst[constants.DDM_ADD]
9799 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9800 instance.nics.append(new_nic)
9801 result.append(("nic.%d" % (len(instance.nics) - 1),
9802 "add:mac=%s,ip=%s,mode=%s,link=%s" %
9803 (new_nic.mac, new_nic.ip,
9804 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9805 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9808 for key in 'mac', 'ip':
9810 setattr(instance.nics[nic_op], key, nic_dict[key])
9811 if nic_op in self.nic_pinst:
9812 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9813 for key, val in nic_dict.iteritems():
9814 result.append(("nic.%s/%d" % (key, nic_op), val))
9817 if self.op.hvparams:
9818 instance.hvparams = self.hv_inst
9819 for key, val in self.op.hvparams.iteritems():
9820 result.append(("hv/%s" % key, val))
9823 if self.op.beparams:
9824 instance.beparams = self.be_inst
9825 for key, val in self.op.beparams.iteritems():
9826 result.append(("be/%s" % key, val))
9830 instance.os = self.op.os_name
9833 if self.op.osparams:
9834 instance.osparams = self.os_inst
9835 for key, val in self.op.osparams.iteritems():
9836 result.append(("os/%s" % key, val))
9838 self.cfg.Update(instance, feedback_fn)
9842 _DISK_CONVERSIONS = {
9843 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9844 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9848 class LUBackupQuery(NoHooksLU):
9849 """Query the exports list
9854 def ExpandNames(self):
9855 self.needed_locks = {}
9856 self.share_locks[locking.LEVEL_NODE] = 1
9857 if not self.op.nodes:
9858 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9860 self.needed_locks[locking.LEVEL_NODE] = \
9861 _GetWantedNodes(self, self.op.nodes)
9863 def Exec(self, feedback_fn):
9864 """Compute the list of all the exported system images.
9867 @return: a dictionary with the structure node->(export-list)
9868 where export-list is a list of the instances exported on
9872 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9873 rpcresult = self.rpc.call_export_list(self.nodes)
9875 for node in rpcresult:
9876 if rpcresult[node].fail_msg:
9877 result[node] = False
9879 result[node] = rpcresult[node].payload
9884 class LUBackupPrepare(NoHooksLU):
9885 """Prepares an instance for an export and returns useful information.
9890 def ExpandNames(self):
9891 self._ExpandAndLockInstance()
9893 def CheckPrereq(self):
9894 """Check prerequisites.
9897 instance_name = self.op.instance_name
9899 self.instance = self.cfg.GetInstanceInfo(instance_name)
9900 assert self.instance is not None, \
9901 "Cannot retrieve locked instance %s" % self.op.instance_name
9902 _CheckNodeOnline(self, self.instance.primary_node)
9904 self._cds = _GetClusterDomainSecret()
9906 def Exec(self, feedback_fn):
9907 """Prepares an instance for an export.
9910 instance = self.instance
9912 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9913 salt = utils.GenerateSecret(8)
9915 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9916 result = self.rpc.call_x509_cert_create(instance.primary_node,
9917 constants.RIE_CERT_VALIDITY)
9918 result.Raise("Can't create X509 key and certificate on %s" % result.node)
9920 (name, cert_pem) = result.payload
9922 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9926 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9927 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9929 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9935 class LUBackupExport(LogicalUnit):
9936 """Export an instance to an image in the cluster.
9939 HPATH = "instance-export"
9940 HTYPE = constants.HTYPE_INSTANCE
9943 def CheckArguments(self):
9944 """Check the arguments.
9947 self.x509_key_name = self.op.x509_key_name
9948 self.dest_x509_ca_pem = self.op.destination_x509_ca
9950 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9951 if not self.x509_key_name:
9952 raise errors.OpPrereqError("Missing X509 key name for encryption",
9955 if not self.dest_x509_ca_pem:
9956 raise errors.OpPrereqError("Missing destination X509 CA",
9959 def ExpandNames(self):
9960 self._ExpandAndLockInstance()
9962 # Lock all nodes for local exports
9963 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9964 # FIXME: lock only instance primary and destination node
9966 # Sad but true, for now we have do lock all nodes, as we don't know where
9967 # the previous export might be, and in this LU we search for it and
9968 # remove it from its current node. In the future we could fix this by:
9969 # - making a tasklet to search (share-lock all), then create the
9970 # new one, then one to remove, after
9971 # - removing the removal operation altogether
9972 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9974 def DeclareLocks(self, level):
9975 """Last minute lock declaration."""
9976 # All nodes are locked anyway, so nothing to do here.
9978 def BuildHooksEnv(self):
9981 This will run on the master, primary node and target node.
9985 "EXPORT_MODE": self.op.mode,
9986 "EXPORT_NODE": self.op.target_node,
9987 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9988 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9989 # TODO: Generic function for boolean env variables
9990 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9993 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9995 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9997 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9998 nl.append(self.op.target_node)
10002 def CheckPrereq(self):
10003 """Check prerequisites.
10005 This checks that the instance and node names are valid.
10008 instance_name = self.op.instance_name
10010 self.instance = self.cfg.GetInstanceInfo(instance_name)
10011 assert self.instance is not None, \
10012 "Cannot retrieve locked instance %s" % self.op.instance_name
10013 _CheckNodeOnline(self, self.instance.primary_node)
10015 if (self.op.remove_instance and self.instance.admin_up and
10016 not self.op.shutdown):
10017 raise errors.OpPrereqError("Can not remove instance without shutting it"
10020 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10021 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
10022 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
10023 assert self.dst_node is not None
10025 _CheckNodeOnline(self, self.dst_node.name)
10026 _CheckNodeNotDrained(self, self.dst_node.name)
10029 self.dest_disk_info = None
10030 self.dest_x509_ca = None
10032 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
10033 self.dst_node = None
10035 if len(self.op.target_node) != len(self.instance.disks):
10036 raise errors.OpPrereqError(("Received destination information for %s"
10037 " disks, but instance %s has %s disks") %
10038 (len(self.op.target_node), instance_name,
10039 len(self.instance.disks)),
10040 errors.ECODE_INVAL)
10042 cds = _GetClusterDomainSecret()
10044 # Check X509 key name
10046 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
10047 except (TypeError, ValueError), err:
10048 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
10050 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
10051 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
10052 errors.ECODE_INVAL)
10054 # Load and verify CA
10056 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
10057 except OpenSSL.crypto.Error, err:
10058 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
10059 (err, ), errors.ECODE_INVAL)
10061 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
10062 if errcode is not None:
10063 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
10064 (msg, ), errors.ECODE_INVAL)
10066 self.dest_x509_ca = cert
10068 # Verify target information
10070 for idx, disk_data in enumerate(self.op.target_node):
10072 (host, port, magic) = \
10073 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
10074 except errors.GenericError, err:
10075 raise errors.OpPrereqError("Target info for disk %s: %s" %
10076 (idx, err), errors.ECODE_INVAL)
10078 disk_info.append((host, port, magic))
10080 assert len(disk_info) == len(self.op.target_node)
10081 self.dest_disk_info = disk_info
10084 raise errors.ProgrammerError("Unhandled export mode %r" %
10087 # instance disk type verification
10088 # TODO: Implement export support for file-based disks
10089 for disk in self.instance.disks:
10090 if disk.dev_type == constants.LD_FILE:
10091 raise errors.OpPrereqError("Export not supported for instances with"
10092 " file-based disks", errors.ECODE_INVAL)
10094 def _CleanupExports(self, feedback_fn):
10095 """Removes exports of current instance from all other nodes.
10097 If an instance in a cluster with nodes A..D was exported to node C, its
10098 exports will be removed from the nodes A, B and D.
10101 assert self.op.mode != constants.EXPORT_MODE_REMOTE
10103 nodelist = self.cfg.GetNodeList()
10104 nodelist.remove(self.dst_node.name)
10106 # on one-node clusters nodelist will be empty after the removal
10107 # if we proceed the backup would be removed because OpBackupQuery
10108 # substitutes an empty list with the full cluster node list.
10109 iname = self.instance.name
10111 feedback_fn("Removing old exports for instance %s" % iname)
10112 exportlist = self.rpc.call_export_list(nodelist)
10113 for node in exportlist:
10114 if exportlist[node].fail_msg:
10116 if iname in exportlist[node].payload:
10117 msg = self.rpc.call_export_remove(node, iname).fail_msg
10119 self.LogWarning("Could not remove older export for instance %s"
10120 " on node %s: %s", iname, node, msg)
10122 def Exec(self, feedback_fn):
10123 """Export an instance to an image in the cluster.
10126 assert self.op.mode in constants.EXPORT_MODES
10128 instance = self.instance
10129 src_node = instance.primary_node
10131 if self.op.shutdown:
10132 # shutdown the instance, but not the disks
10133 feedback_fn("Shutting down instance %s" % instance.name)
10134 result = self.rpc.call_instance_shutdown(src_node, instance,
10135 self.op.shutdown_timeout)
10136 # TODO: Maybe ignore failures if ignore_remove_failures is set
10137 result.Raise("Could not shutdown instance %s on"
10138 " node %s" % (instance.name, src_node))
10140 # set the disks ID correctly since call_instance_start needs the
10141 # correct drbd minor to create the symlinks
10142 for disk in instance.disks:
10143 self.cfg.SetDiskID(disk, src_node)
10145 activate_disks = (not instance.admin_up)
10148 # Activate the instance disks if we'exporting a stopped instance
10149 feedback_fn("Activating disks for %s" % instance.name)
10150 _StartInstanceDisks(self, instance, None)
10153 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
10156 helper.CreateSnapshots()
10158 if (self.op.shutdown and instance.admin_up and
10159 not self.op.remove_instance):
10160 assert not activate_disks
10161 feedback_fn("Starting instance %s" % instance.name)
10162 result = self.rpc.call_instance_start(src_node, instance, None, None)
10163 msg = result.fail_msg
10165 feedback_fn("Failed to start instance: %s" % msg)
10166 _ShutdownInstanceDisks(self, instance)
10167 raise errors.OpExecError("Could not start instance: %s" % msg)
10169 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10170 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
10171 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
10172 connect_timeout = constants.RIE_CONNECT_TIMEOUT
10173 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
10175 (key_name, _, _) = self.x509_key_name
10178 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
10181 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
10182 key_name, dest_ca_pem,
10187 # Check for backwards compatibility
10188 assert len(dresults) == len(instance.disks)
10189 assert compat.all(isinstance(i, bool) for i in dresults), \
10190 "Not all results are boolean: %r" % dresults
10194 feedback_fn("Deactivating disks for %s" % instance.name)
10195 _ShutdownInstanceDisks(self, instance)
10197 if not (compat.all(dresults) and fin_resu):
10200 failures.append("export finalization")
10201 if not compat.all(dresults):
10202 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
10204 failures.append("disk export: disk(s) %s" % fdsk)
10206 raise errors.OpExecError("Export failed, errors in %s" %
10207 utils.CommaJoin(failures))
10209 # At this point, the export was successful, we can cleanup/finish
10211 # Remove instance if requested
10212 if self.op.remove_instance:
10213 feedback_fn("Removing instance %s" % instance.name)
10214 _RemoveInstance(self, feedback_fn, instance,
10215 self.op.ignore_remove_failures)
10217 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10218 self._CleanupExports(feedback_fn)
10220 return fin_resu, dresults
10223 class LUBackupRemove(NoHooksLU):
10224 """Remove exports related to the named instance.
10229 def ExpandNames(self):
10230 self.needed_locks = {}
10231 # We need all nodes to be locked in order for RemoveExport to work, but we
10232 # don't need to lock the instance itself, as nothing will happen to it (and
10233 # we can remove exports also for a removed instance)
10234 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10236 def Exec(self, feedback_fn):
10237 """Remove any export.
10240 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
10241 # If the instance was not found we'll try with the name that was passed in.
10242 # This will only work if it was an FQDN, though.
10244 if not instance_name:
10246 instance_name = self.op.instance_name
10248 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
10249 exportlist = self.rpc.call_export_list(locked_nodes)
10251 for node in exportlist:
10252 msg = exportlist[node].fail_msg
10254 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
10256 if instance_name in exportlist[node].payload:
10258 result = self.rpc.call_export_remove(node, instance_name)
10259 msg = result.fail_msg
10261 logging.error("Could not remove export for instance %s"
10262 " on node %s: %s", instance_name, node, msg)
10264 if fqdn_warn and not found:
10265 feedback_fn("Export not found. If trying to remove an export belonging"
10266 " to a deleted instance please use its Fully Qualified"
10270 class LUGroupAdd(LogicalUnit):
10271 """Logical unit for creating node groups.
10274 HPATH = "group-add"
10275 HTYPE = constants.HTYPE_GROUP
10278 def ExpandNames(self):
10279 # We need the new group's UUID here so that we can create and acquire the
10280 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
10281 # that it should not check whether the UUID exists in the configuration.
10282 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
10283 self.needed_locks = {}
10284 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10286 def CheckPrereq(self):
10287 """Check prerequisites.
10289 This checks that the given group name is not an existing node group
10294 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10295 except errors.OpPrereqError:
10298 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
10299 " node group (UUID: %s)" %
10300 (self.op.group_name, existing_uuid),
10301 errors.ECODE_EXISTS)
10303 if self.op.ndparams:
10304 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10306 def BuildHooksEnv(self):
10307 """Build hooks env.
10311 "GROUP_NAME": self.op.group_name,
10313 mn = self.cfg.GetMasterNode()
10314 return env, [mn], [mn]
10316 def Exec(self, feedback_fn):
10317 """Add the node group to the cluster.
10320 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
10321 uuid=self.group_uuid,
10322 alloc_policy=self.op.alloc_policy,
10323 ndparams=self.op.ndparams)
10325 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
10326 del self.remove_locks[locking.LEVEL_NODEGROUP]
10329 class LUGroupAssignNodes(NoHooksLU):
10330 """Logical unit for assigning nodes to groups.
10335 def ExpandNames(self):
10336 # These raise errors.OpPrereqError on their own:
10337 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10338 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
10340 # We want to lock all the affected nodes and groups. We have readily
10341 # available the list of nodes, and the *destination* group. To gather the
10342 # list of "source" groups, we need to fetch node information.
10343 self.node_data = self.cfg.GetAllNodesInfo()
10344 affected_groups = set(self.node_data[node].group for node in self.op.nodes)
10345 affected_groups.add(self.group_uuid)
10347 self.needed_locks = {
10348 locking.LEVEL_NODEGROUP: list(affected_groups),
10349 locking.LEVEL_NODE: self.op.nodes,
10352 def CheckPrereq(self):
10353 """Check prerequisites.
10356 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10357 instance_data = self.cfg.GetAllInstancesInfo()
10359 if self.group is None:
10360 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10361 (self.op.group_name, self.group_uuid))
10363 (new_splits, previous_splits) = \
10364 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
10365 for node in self.op.nodes],
10366 self.node_data, instance_data)
10369 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
10371 if not self.op.force:
10372 raise errors.OpExecError("The following instances get split by this"
10373 " change and --force was not given: %s" %
10376 self.LogWarning("This operation will split the following instances: %s",
10379 if previous_splits:
10380 self.LogWarning("In addition, these already-split instances continue"
10381 " to be spit across groups: %s",
10382 utils.CommaJoin(utils.NiceSort(previous_splits)))
10384 def Exec(self, feedback_fn):
10385 """Assign nodes to a new group.
10388 for node in self.op.nodes:
10389 self.node_data[node].group = self.group_uuid
10391 self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
10394 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
10395 """Check for split instances after a node assignment.
10397 This method considers a series of node assignments as an atomic operation,
10398 and returns information about split instances after applying the set of
10401 In particular, it returns information about newly split instances, and
10402 instances that were already split, and remain so after the change.
10404 Only instances whose disk template is listed in constants.DTS_NET_MIRROR are
10407 @type changes: list of (node_name, new_group_uuid) pairs.
10408 @param changes: list of node assignments to consider.
10409 @param node_data: a dict with data for all nodes
10410 @param instance_data: a dict with all instances to consider
10411 @rtype: a two-tuple
10412 @return: a list of instances that were previously okay and result split as a
10413 consequence of this change, and a list of instances that were previously
10414 split and this change does not fix.
10417 changed_nodes = dict((node, group) for node, group in changes
10418 if node_data[node].group != group)
10420 all_split_instances = set()
10421 previously_split_instances = set()
10423 def InstanceNodes(instance):
10424 return [instance.primary_node] + list(instance.secondary_nodes)
10426 for inst in instance_data.values():
10427 if inst.disk_template not in constants.DTS_NET_MIRROR:
10430 instance_nodes = InstanceNodes(inst)
10432 if len(set(node_data[node].group for node in instance_nodes)) > 1:
10433 previously_split_instances.add(inst.name)
10435 if len(set(changed_nodes.get(node, node_data[node].group)
10436 for node in instance_nodes)) > 1:
10437 all_split_instances.add(inst.name)
10439 return (list(all_split_instances - previously_split_instances),
10440 list(previously_split_instances & all_split_instances))
10443 class _GroupQuery(_QueryBase):
10444 FIELDS = query.GROUP_FIELDS
10446 def ExpandNames(self, lu):
10447 lu.needed_locks = {}
10449 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
10450 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
10453 self.wanted = [name_to_uuid[name]
10454 for name in utils.NiceSort(name_to_uuid.keys())]
10456 # Accept names to be either names or UUIDs.
10459 all_uuid = frozenset(self._all_groups.keys())
10461 for name in self.names:
10462 if name in all_uuid:
10463 self.wanted.append(name)
10464 elif name in name_to_uuid:
10465 self.wanted.append(name_to_uuid[name])
10467 missing.append(name)
10470 raise errors.OpPrereqError("Some groups do not exist: %s" % missing,
10471 errors.ECODE_NOENT)
10473 def DeclareLocks(self, lu, level):
10476 def _GetQueryData(self, lu):
10477 """Computes the list of node groups and their attributes.
10480 do_nodes = query.GQ_NODE in self.requested_data
10481 do_instances = query.GQ_INST in self.requested_data
10483 group_to_nodes = None
10484 group_to_instances = None
10486 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
10487 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
10488 # latter GetAllInstancesInfo() is not enough, for we have to go through
10489 # instance->node. Hence, we will need to process nodes even if we only need
10490 # instance information.
10491 if do_nodes or do_instances:
10492 all_nodes = lu.cfg.GetAllNodesInfo()
10493 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
10496 for node in all_nodes.values():
10497 if node.group in group_to_nodes:
10498 group_to_nodes[node.group].append(node.name)
10499 node_to_group[node.name] = node.group
10502 all_instances = lu.cfg.GetAllInstancesInfo()
10503 group_to_instances = dict((uuid, []) for uuid in self.wanted)
10505 for instance in all_instances.values():
10506 node = instance.primary_node
10507 if node in node_to_group:
10508 group_to_instances[node_to_group[node]].append(instance.name)
10511 # Do not pass on node information if it was not requested.
10512 group_to_nodes = None
10514 return query.GroupQueryData([self._all_groups[uuid]
10515 for uuid in self.wanted],
10516 group_to_nodes, group_to_instances)
10519 class LUGroupQuery(NoHooksLU):
10520 """Logical unit for querying node groups.
10525 def CheckArguments(self):
10526 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
10527 self.op.output_fields, False)
10529 def ExpandNames(self):
10530 self.gq.ExpandNames(self)
10532 def Exec(self, feedback_fn):
10533 return self.gq.OldStyleQuery(self)
10536 class LUGroupSetParams(LogicalUnit):
10537 """Modifies the parameters of a node group.
10540 HPATH = "group-modify"
10541 HTYPE = constants.HTYPE_GROUP
10544 def CheckArguments(self):
10547 self.op.alloc_policy,
10550 if all_changes.count(None) == len(all_changes):
10551 raise errors.OpPrereqError("Please pass at least one modification",
10552 errors.ECODE_INVAL)
10554 def ExpandNames(self):
10555 # This raises errors.OpPrereqError on its own:
10556 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10558 self.needed_locks = {
10559 locking.LEVEL_NODEGROUP: [self.group_uuid],
10562 def CheckPrereq(self):
10563 """Check prerequisites.
10566 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10568 if self.group is None:
10569 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10570 (self.op.group_name, self.group_uuid))
10572 if self.op.ndparams:
10573 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
10574 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10575 self.new_ndparams = new_ndparams
10577 def BuildHooksEnv(self):
10578 """Build hooks env.
10582 "GROUP_NAME": self.op.group_name,
10583 "NEW_ALLOC_POLICY": self.op.alloc_policy,
10585 mn = self.cfg.GetMasterNode()
10586 return env, [mn], [mn]
10588 def Exec(self, feedback_fn):
10589 """Modifies the node group.
10594 if self.op.ndparams:
10595 self.group.ndparams = self.new_ndparams
10596 result.append(("ndparams", str(self.group.ndparams)))
10598 if self.op.alloc_policy:
10599 self.group.alloc_policy = self.op.alloc_policy
10601 self.cfg.Update(self.group, feedback_fn)
10606 class LUGroupRemove(LogicalUnit):
10607 HPATH = "group-remove"
10608 HTYPE = constants.HTYPE_GROUP
10611 def ExpandNames(self):
10612 # This will raises errors.OpPrereqError on its own:
10613 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10614 self.needed_locks = {
10615 locking.LEVEL_NODEGROUP: [self.group_uuid],
10618 def CheckPrereq(self):
10619 """Check prerequisites.
10621 This checks that the given group name exists as a node group, that is
10622 empty (i.e., contains no nodes), and that is not the last group of the
10626 # Verify that the group is empty.
10627 group_nodes = [node.name
10628 for node in self.cfg.GetAllNodesInfo().values()
10629 if node.group == self.group_uuid]
10632 raise errors.OpPrereqError("Group '%s' not empty, has the following"
10634 (self.op.group_name,
10635 utils.CommaJoin(utils.NiceSort(group_nodes))),
10636 errors.ECODE_STATE)
10638 # Verify the cluster would not be left group-less.
10639 if len(self.cfg.GetNodeGroupList()) == 1:
10640 raise errors.OpPrereqError("Group '%s' is the only group,"
10641 " cannot be removed" %
10642 self.op.group_name,
10643 errors.ECODE_STATE)
10645 def BuildHooksEnv(self):
10646 """Build hooks env.
10650 "GROUP_NAME": self.op.group_name,
10652 mn = self.cfg.GetMasterNode()
10653 return env, [mn], [mn]
10655 def Exec(self, feedback_fn):
10656 """Remove the node group.
10660 self.cfg.RemoveNodeGroup(self.group_uuid)
10661 except errors.ConfigurationError:
10662 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
10663 (self.op.group_name, self.group_uuid))
10665 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10668 class LUGroupRename(LogicalUnit):
10669 HPATH = "group-rename"
10670 HTYPE = constants.HTYPE_GROUP
10673 def ExpandNames(self):
10674 # This raises errors.OpPrereqError on its own:
10675 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10677 self.needed_locks = {
10678 locking.LEVEL_NODEGROUP: [self.group_uuid],
10681 def CheckPrereq(self):
10682 """Check prerequisites.
10684 Ensures requested new name is not yet used.
10688 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
10689 except errors.OpPrereqError:
10692 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
10693 " node group (UUID: %s)" %
10694 (self.op.new_name, new_name_uuid),
10695 errors.ECODE_EXISTS)
10697 def BuildHooksEnv(self):
10698 """Build hooks env.
10702 "OLD_NAME": self.op.group_name,
10703 "NEW_NAME": self.op.new_name,
10706 mn = self.cfg.GetMasterNode()
10707 all_nodes = self.cfg.GetAllNodesInfo()
10709 all_nodes.pop(mn, None)
10711 for node in all_nodes.values():
10712 if node.group == self.group_uuid:
10713 run_nodes.append(node.name)
10715 return env, run_nodes, run_nodes
10717 def Exec(self, feedback_fn):
10718 """Rename the node group.
10721 group = self.cfg.GetNodeGroup(self.group_uuid)
10724 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10725 (self.op.group_name, self.group_uuid))
10727 group.name = self.op.new_name
10728 self.cfg.Update(group, feedback_fn)
10730 return self.op.new_name
10733 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
10734 """Generic tags LU.
10736 This is an abstract class which is the parent of all the other tags LUs.
10740 def ExpandNames(self):
10741 self.needed_locks = {}
10742 if self.op.kind == constants.TAG_NODE:
10743 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
10744 self.needed_locks[locking.LEVEL_NODE] = self.op.name
10745 elif self.op.kind == constants.TAG_INSTANCE:
10746 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
10747 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
10749 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
10750 # not possible to acquire the BGL based on opcode parameters)
10752 def CheckPrereq(self):
10753 """Check prerequisites.
10756 if self.op.kind == constants.TAG_CLUSTER:
10757 self.target = self.cfg.GetClusterInfo()
10758 elif self.op.kind == constants.TAG_NODE:
10759 self.target = self.cfg.GetNodeInfo(self.op.name)
10760 elif self.op.kind == constants.TAG_INSTANCE:
10761 self.target = self.cfg.GetInstanceInfo(self.op.name)
10763 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
10764 str(self.op.kind), errors.ECODE_INVAL)
10767 class LUTagsGet(TagsLU):
10768 """Returns the tags of a given object.
10773 def ExpandNames(self):
10774 TagsLU.ExpandNames(self)
10776 # Share locks as this is only a read operation
10777 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
10779 def Exec(self, feedback_fn):
10780 """Returns the tag list.
10783 return list(self.target.GetTags())
10786 class LUTagsSearch(NoHooksLU):
10787 """Searches the tags for a given pattern.
10792 def ExpandNames(self):
10793 self.needed_locks = {}
10795 def CheckPrereq(self):
10796 """Check prerequisites.
10798 This checks the pattern passed for validity by compiling it.
10802 self.re = re.compile(self.op.pattern)
10803 except re.error, err:
10804 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
10805 (self.op.pattern, err), errors.ECODE_INVAL)
10807 def Exec(self, feedback_fn):
10808 """Returns the tag list.
10812 tgts = [("/cluster", cfg.GetClusterInfo())]
10813 ilist = cfg.GetAllInstancesInfo().values()
10814 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
10815 nlist = cfg.GetAllNodesInfo().values()
10816 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
10818 for path, target in tgts:
10819 for tag in target.GetTags():
10820 if self.re.search(tag):
10821 results.append((path, tag))
10825 class LUTagsSet(TagsLU):
10826 """Sets a tag on a given object.
10831 def CheckPrereq(self):
10832 """Check prerequisites.
10834 This checks the type and length of the tag name and value.
10837 TagsLU.CheckPrereq(self)
10838 for tag in self.op.tags:
10839 objects.TaggableObject.ValidateTag(tag)
10841 def Exec(self, feedback_fn):
10846 for tag in self.op.tags:
10847 self.target.AddTag(tag)
10848 except errors.TagError, err:
10849 raise errors.OpExecError("Error while setting tag: %s" % str(err))
10850 self.cfg.Update(self.target, feedback_fn)
10853 class LUTagsDel(TagsLU):
10854 """Delete a list of tags from a given object.
10859 def CheckPrereq(self):
10860 """Check prerequisites.
10862 This checks that we have the given tag.
10865 TagsLU.CheckPrereq(self)
10866 for tag in self.op.tags:
10867 objects.TaggableObject.ValidateTag(tag)
10868 del_tags = frozenset(self.op.tags)
10869 cur_tags = self.target.GetTags()
10871 diff_tags = del_tags - cur_tags
10873 diff_names = ("'%s'" % i for i in sorted(diff_tags))
10874 raise errors.OpPrereqError("Tag(s) %s not found" %
10875 (utils.CommaJoin(diff_names), ),
10876 errors.ECODE_NOENT)
10878 def Exec(self, feedback_fn):
10879 """Remove the tag from the object.
10882 for tag in self.op.tags:
10883 self.target.RemoveTag(tag)
10884 self.cfg.Update(self.target, feedback_fn)
10887 class LUTestDelay(NoHooksLU):
10888 """Sleep for a specified amount of time.
10890 This LU sleeps on the master and/or nodes for a specified amount of
10896 def ExpandNames(self):
10897 """Expand names and set required locks.
10899 This expands the node list, if any.
10902 self.needed_locks = {}
10903 if self.op.on_nodes:
10904 # _GetWantedNodes can be used here, but is not always appropriate to use
10905 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
10906 # more information.
10907 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
10908 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
10910 def _TestDelay(self):
10911 """Do the actual sleep.
10914 if self.op.on_master:
10915 if not utils.TestDelay(self.op.duration):
10916 raise errors.OpExecError("Error during master delay test")
10917 if self.op.on_nodes:
10918 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
10919 for node, node_result in result.items():
10920 node_result.Raise("Failure during rpc call to node %s" % node)
10922 def Exec(self, feedback_fn):
10923 """Execute the test delay opcode, with the wanted repetitions.
10926 if self.op.repeat == 0:
10929 top_value = self.op.repeat - 1
10930 for i in range(self.op.repeat):
10931 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
10935 class LUTestJqueue(NoHooksLU):
10936 """Utility LU to test some aspects of the job queue.
10941 # Must be lower than default timeout for WaitForJobChange to see whether it
10942 # notices changed jobs
10943 _CLIENT_CONNECT_TIMEOUT = 20.0
10944 _CLIENT_CONFIRM_TIMEOUT = 60.0
10947 def _NotifyUsingSocket(cls, cb, errcls):
10948 """Opens a Unix socket and waits for another program to connect.
10951 @param cb: Callback to send socket name to client
10952 @type errcls: class
10953 @param errcls: Exception class to use for errors
10956 # Using a temporary directory as there's no easy way to create temporary
10957 # sockets without writing a custom loop around tempfile.mktemp and
10959 tmpdir = tempfile.mkdtemp()
10961 tmpsock = utils.PathJoin(tmpdir, "sock")
10963 logging.debug("Creating temporary socket at %s", tmpsock)
10964 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10969 # Send details to client
10972 # Wait for client to connect before continuing
10973 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10975 (conn, _) = sock.accept()
10976 except socket.error, err:
10977 raise errcls("Client didn't connect in time (%s)" % err)
10981 # Remove as soon as client is connected
10982 shutil.rmtree(tmpdir)
10984 # Wait for client to close
10987 # pylint: disable-msg=E1101
10988 # Instance of '_socketobject' has no ... member
10989 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10991 except socket.error, err:
10992 raise errcls("Client failed to confirm notification (%s)" % err)
10996 def _SendNotification(self, test, arg, sockname):
10997 """Sends a notification to the client.
11000 @param test: Test name
11001 @param arg: Test argument (depends on test)
11002 @type sockname: string
11003 @param sockname: Socket path
11006 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
11008 def _Notify(self, prereq, test, arg):
11009 """Notifies the client of a test.
11012 @param prereq: Whether this is a prereq-phase test
11014 @param test: Test name
11015 @param arg: Test argument (depends on test)
11019 errcls = errors.OpPrereqError
11021 errcls = errors.OpExecError
11023 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
11027 def CheckArguments(self):
11028 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
11029 self.expandnames_calls = 0
11031 def ExpandNames(self):
11032 checkargs_calls = getattr(self, "checkargs_calls", 0)
11033 if checkargs_calls < 1:
11034 raise errors.ProgrammerError("CheckArguments was not called")
11036 self.expandnames_calls += 1
11038 if self.op.notify_waitlock:
11039 self._Notify(True, constants.JQT_EXPANDNAMES, None)
11041 self.LogInfo("Expanding names")
11043 # Get lock on master node (just to get a lock, not for a particular reason)
11044 self.needed_locks = {
11045 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
11048 def Exec(self, feedback_fn):
11049 if self.expandnames_calls < 1:
11050 raise errors.ProgrammerError("ExpandNames was not called")
11052 if self.op.notify_exec:
11053 self._Notify(False, constants.JQT_EXEC, None)
11055 self.LogInfo("Executing")
11057 if self.op.log_messages:
11058 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
11059 for idx, msg in enumerate(self.op.log_messages):
11060 self.LogInfo("Sending log message %s", idx + 1)
11061 feedback_fn(constants.JQT_MSGPREFIX + msg)
11062 # Report how many test messages have been sent
11063 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
11066 raise errors.OpExecError("Opcode failure was requested")
11071 class IAllocator(object):
11072 """IAllocator framework.
11074 An IAllocator instance has three sets of attributes:
11075 - cfg that is needed to query the cluster
11076 - input data (all members of the _KEYS class attribute are required)
11077 - four buffer attributes (in|out_data|text), that represent the
11078 input (to the external script) in text and data structure format,
11079 and the output from it, again in two formats
11080 - the result variables from the script (success, info, nodes) for
11084 # pylint: disable-msg=R0902
11085 # lots of instance attributes
11087 "name", "mem_size", "disks", "disk_template",
11088 "os", "tags", "nics", "vcpus", "hypervisor",
11091 "name", "relocate_from",
11097 def __init__(self, cfg, rpc, mode, **kwargs):
11100 # init buffer variables
11101 self.in_text = self.out_text = self.in_data = self.out_data = None
11102 # init all input fields so that pylint is happy
11104 self.mem_size = self.disks = self.disk_template = None
11105 self.os = self.tags = self.nics = self.vcpus = None
11106 self.hypervisor = None
11107 self.relocate_from = None
11109 self.evac_nodes = None
11111 self.required_nodes = None
11112 # init result fields
11113 self.success = self.info = self.result = None
11114 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
11115 keyset = self._ALLO_KEYS
11116 fn = self._AddNewInstance
11117 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
11118 keyset = self._RELO_KEYS
11119 fn = self._AddRelocateInstance
11120 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
11121 keyset = self._EVAC_KEYS
11122 fn = self._AddEvacuateNodes
11124 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
11125 " IAllocator" % self.mode)
11127 if key not in keyset:
11128 raise errors.ProgrammerError("Invalid input parameter '%s' to"
11129 " IAllocator" % key)
11130 setattr(self, key, kwargs[key])
11133 if key not in kwargs:
11134 raise errors.ProgrammerError("Missing input parameter '%s' to"
11135 " IAllocator" % key)
11136 self._BuildInputData(fn)
11138 def _ComputeClusterData(self):
11139 """Compute the generic allocator input data.
11141 This is the data that is independent of the actual operation.
11145 cluster_info = cfg.GetClusterInfo()
11148 "version": constants.IALLOCATOR_VERSION,
11149 "cluster_name": cfg.GetClusterName(),
11150 "cluster_tags": list(cluster_info.GetTags()),
11151 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
11152 # we don't have job IDs
11154 ninfo = cfg.GetAllNodesInfo()
11155 iinfo = cfg.GetAllInstancesInfo().values()
11156 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
11159 node_list = [n.name for n in ninfo.values() if n.vm_capable]
11161 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
11162 hypervisor_name = self.hypervisor
11163 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
11164 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
11165 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
11166 hypervisor_name = cluster_info.enabled_hypervisors[0]
11168 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
11171 self.rpc.call_all_instances_info(node_list,
11172 cluster_info.enabled_hypervisors)
11174 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
11176 config_ndata = self._ComputeBasicNodeData(ninfo)
11177 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
11178 i_list, config_ndata)
11179 assert len(data["nodes"]) == len(ninfo), \
11180 "Incomplete node data computed"
11182 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
11184 self.in_data = data
11187 def _ComputeNodeGroupData(cfg):
11188 """Compute node groups data.
11192 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
11194 "name": gdata.name,
11195 "alloc_policy": gdata.alloc_policy,
11200 def _ComputeBasicNodeData(node_cfg):
11201 """Compute global node data.
11204 @returns: a dict of name: (node dict, node config)
11208 for ninfo in node_cfg.values():
11209 # fill in static (config-based) values
11211 "tags": list(ninfo.GetTags()),
11212 "primary_ip": ninfo.primary_ip,
11213 "secondary_ip": ninfo.secondary_ip,
11214 "offline": ninfo.offline,
11215 "drained": ninfo.drained,
11216 "master_candidate": ninfo.master_candidate,
11217 "group": ninfo.group,
11218 "master_capable": ninfo.master_capable,
11219 "vm_capable": ninfo.vm_capable,
11222 node_results[ninfo.name] = pnr
11224 return node_results
11227 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
11229 """Compute global node data.
11231 @param node_results: the basic node structures as filled from the config
11234 # make a copy of the current dict
11235 node_results = dict(node_results)
11236 for nname, nresult in node_data.items():
11237 assert nname in node_results, "Missing basic data for node %s" % nname
11238 ninfo = node_cfg[nname]
11240 if not (ninfo.offline or ninfo.drained):
11241 nresult.Raise("Can't get data for node %s" % nname)
11242 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
11244 remote_info = nresult.payload
11246 for attr in ['memory_total', 'memory_free', 'memory_dom0',
11247 'vg_size', 'vg_free', 'cpu_total']:
11248 if attr not in remote_info:
11249 raise errors.OpExecError("Node '%s' didn't return attribute"
11250 " '%s'" % (nname, attr))
11251 if not isinstance(remote_info[attr], int):
11252 raise errors.OpExecError("Node '%s' returned invalid value"
11254 (nname, attr, remote_info[attr]))
11255 # compute memory used by primary instances
11256 i_p_mem = i_p_up_mem = 0
11257 for iinfo, beinfo in i_list:
11258 if iinfo.primary_node == nname:
11259 i_p_mem += beinfo[constants.BE_MEMORY]
11260 if iinfo.name not in node_iinfo[nname].payload:
11263 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
11264 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
11265 remote_info['memory_free'] -= max(0, i_mem_diff)
11268 i_p_up_mem += beinfo[constants.BE_MEMORY]
11270 # compute memory used by instances
11272 "total_memory": remote_info['memory_total'],
11273 "reserved_memory": remote_info['memory_dom0'],
11274 "free_memory": remote_info['memory_free'],
11275 "total_disk": remote_info['vg_size'],
11276 "free_disk": remote_info['vg_free'],
11277 "total_cpus": remote_info['cpu_total'],
11278 "i_pri_memory": i_p_mem,
11279 "i_pri_up_memory": i_p_up_mem,
11281 pnr_dyn.update(node_results[nname])
11282 node_results[nname] = pnr_dyn
11284 return node_results
11287 def _ComputeInstanceData(cluster_info, i_list):
11288 """Compute global instance data.
11292 for iinfo, beinfo in i_list:
11294 for nic in iinfo.nics:
11295 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
11296 nic_dict = {"mac": nic.mac,
11298 "mode": filled_params[constants.NIC_MODE],
11299 "link": filled_params[constants.NIC_LINK],
11301 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
11302 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
11303 nic_data.append(nic_dict)
11305 "tags": list(iinfo.GetTags()),
11306 "admin_up": iinfo.admin_up,
11307 "vcpus": beinfo[constants.BE_VCPUS],
11308 "memory": beinfo[constants.BE_MEMORY],
11310 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
11312 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
11313 "disk_template": iinfo.disk_template,
11314 "hypervisor": iinfo.hypervisor,
11316 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
11318 instance_data[iinfo.name] = pir
11320 return instance_data
11322 def _AddNewInstance(self):
11323 """Add new instance data to allocator structure.
11325 This in combination with _AllocatorGetClusterData will create the
11326 correct structure needed as input for the allocator.
11328 The checks for the completeness of the opcode must have already been
11332 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
11334 if self.disk_template in constants.DTS_NET_MIRROR:
11335 self.required_nodes = 2
11337 self.required_nodes = 1
11340 "disk_template": self.disk_template,
11343 "vcpus": self.vcpus,
11344 "memory": self.mem_size,
11345 "disks": self.disks,
11346 "disk_space_total": disk_space,
11348 "required_nodes": self.required_nodes,
11352 def _AddRelocateInstance(self):
11353 """Add relocate instance data to allocator structure.
11355 This in combination with _IAllocatorGetClusterData will create the
11356 correct structure needed as input for the allocator.
11358 The checks for the completeness of the opcode must have already been
11362 instance = self.cfg.GetInstanceInfo(self.name)
11363 if instance is None:
11364 raise errors.ProgrammerError("Unknown instance '%s' passed to"
11365 " IAllocator" % self.name)
11367 if instance.disk_template not in constants.DTS_MIRRORED:
11368 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
11369 errors.ECODE_INVAL)
11371 if instance.disk_template in constants.DTS_NET_MIRROR and \
11372 len(instance.secondary_nodes) != 1:
11373 raise errors.OpPrereqError("Instance has not exactly one secondary node",
11374 errors.ECODE_STATE)
11376 self.required_nodes = 1
11377 disk_sizes = [{'size': disk.size} for disk in instance.disks]
11378 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
11382 "disk_space_total": disk_space,
11383 "required_nodes": self.required_nodes,
11384 "relocate_from": self.relocate_from,
11388 def _AddEvacuateNodes(self):
11389 """Add evacuate nodes data to allocator structure.
11393 "evac_nodes": self.evac_nodes
11397 def _BuildInputData(self, fn):
11398 """Build input data structures.
11401 self._ComputeClusterData()
11404 request["type"] = self.mode
11405 self.in_data["request"] = request
11407 self.in_text = serializer.Dump(self.in_data)
11409 def Run(self, name, validate=True, call_fn=None):
11410 """Run an instance allocator and return the results.
11413 if call_fn is None:
11414 call_fn = self.rpc.call_iallocator_runner
11416 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
11417 result.Raise("Failure while running the iallocator script")
11419 self.out_text = result.payload
11421 self._ValidateResult()
11423 def _ValidateResult(self):
11424 """Process the allocator results.
11426 This will process and if successful save the result in
11427 self.out_data and the other parameters.
11431 rdict = serializer.Load(self.out_text)
11432 except Exception, err:
11433 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
11435 if not isinstance(rdict, dict):
11436 raise errors.OpExecError("Can't parse iallocator results: not a dict")
11438 # TODO: remove backwards compatiblity in later versions
11439 if "nodes" in rdict and "result" not in rdict:
11440 rdict["result"] = rdict["nodes"]
11443 for key in "success", "info", "result":
11444 if key not in rdict:
11445 raise errors.OpExecError("Can't parse iallocator results:"
11446 " missing key '%s'" % key)
11447 setattr(self, key, rdict[key])
11449 if not isinstance(rdict["result"], list):
11450 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
11452 self.out_data = rdict
11455 class LUTestAllocator(NoHooksLU):
11456 """Run allocator tests.
11458 This LU runs the allocator tests
11461 def CheckPrereq(self):
11462 """Check prerequisites.
11464 This checks the opcode parameters depending on the director and mode test.
11467 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11468 for attr in ["mem_size", "disks", "disk_template",
11469 "os", "tags", "nics", "vcpus"]:
11470 if not hasattr(self.op, attr):
11471 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
11472 attr, errors.ECODE_INVAL)
11473 iname = self.cfg.ExpandInstanceName(self.op.name)
11474 if iname is not None:
11475 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
11476 iname, errors.ECODE_EXISTS)
11477 if not isinstance(self.op.nics, list):
11478 raise errors.OpPrereqError("Invalid parameter 'nics'",
11479 errors.ECODE_INVAL)
11480 if not isinstance(self.op.disks, list):
11481 raise errors.OpPrereqError("Invalid parameter 'disks'",
11482 errors.ECODE_INVAL)
11483 for row in self.op.disks:
11484 if (not isinstance(row, dict) or
11485 "size" not in row or
11486 not isinstance(row["size"], int) or
11487 "mode" not in row or
11488 row["mode"] not in ['r', 'w']):
11489 raise errors.OpPrereqError("Invalid contents of the 'disks'"
11490 " parameter", errors.ECODE_INVAL)
11491 if self.op.hypervisor is None:
11492 self.op.hypervisor = self.cfg.GetHypervisorType()
11493 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11494 fname = _ExpandInstanceName(self.cfg, self.op.name)
11495 self.op.name = fname
11496 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
11497 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11498 if not hasattr(self.op, "evac_nodes"):
11499 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
11500 " opcode input", errors.ECODE_INVAL)
11502 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
11503 self.op.mode, errors.ECODE_INVAL)
11505 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
11506 if self.op.allocator is None:
11507 raise errors.OpPrereqError("Missing allocator name",
11508 errors.ECODE_INVAL)
11509 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
11510 raise errors.OpPrereqError("Wrong allocator test '%s'" %
11511 self.op.direction, errors.ECODE_INVAL)
11513 def Exec(self, feedback_fn):
11514 """Run the allocator test.
11517 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11518 ial = IAllocator(self.cfg, self.rpc,
11521 mem_size=self.op.mem_size,
11522 disks=self.op.disks,
11523 disk_template=self.op.disk_template,
11527 vcpus=self.op.vcpus,
11528 hypervisor=self.op.hypervisor,
11530 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11531 ial = IAllocator(self.cfg, self.rpc,
11534 relocate_from=list(self.relocate_from),
11536 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11537 ial = IAllocator(self.cfg, self.rpc,
11539 evac_nodes=self.op.evac_nodes)
11541 raise errors.ProgrammerError("Uncatched mode %s in"
11542 " LUTestAllocator.Exec", self.op.mode)
11544 if self.op.direction == constants.IALLOCATOR_DIR_IN:
11545 result = ial.in_text
11547 ial.Run(self.op.allocator, validate=False)
11548 result = ial.out_text
11552 #: Query type implementations
11554 constants.QR_INSTANCE: _InstanceQuery,
11555 constants.QR_NODE: _NodeQuery,
11556 constants.QR_GROUP: _GroupQuery,
11560 def _GetQueryImplementation(name):
11561 """Returns the implemtnation for a query type.
11563 @param name: Query type, must be one of L{constants.QR_OP_QUERY}
11567 return _QUERY_IMPL[name]
11569 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
11570 errors.ECODE_INVAL)