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.names = ReadSimpleFilter("name", filter_)
461 self.use_locking = use_locking
463 self.query = query.Query(self.FIELDS, fields)
464 self.requested_data = self.query.RequestedData()
466 self.do_locking = None
469 def _GetNames(self, lu, all_names, lock_level):
470 """Helper function to determine names asked for in the query.
474 names = lu.acquired_locks[lock_level]
478 if self.wanted == locking.ALL_SET:
479 assert not self.names
480 # caller didn't specify names, so ordering is not important
481 return utils.NiceSort(names)
483 # caller specified names and we must keep the same order
485 assert not self.do_locking or lu.acquired_locks[lock_level]
487 missing = set(self.wanted).difference(names)
489 raise errors.OpExecError("Some items were removed before retrieving"
490 " their data: %s" % missing)
492 # Return expanded names
496 def FieldsQuery(cls, fields):
497 """Returns list of available fields.
499 @return: List of L{objects.QueryFieldDefinition}
502 return query.QueryFields(cls.FIELDS, fields)
504 def ExpandNames(self, lu):
505 """Expand names for this query.
507 See L{LogicalUnit.ExpandNames}.
510 raise NotImplementedError()
512 def DeclareLocks(self, lu, level):
513 """Declare locks for this query.
515 See L{LogicalUnit.DeclareLocks}.
518 raise NotImplementedError()
520 def _GetQueryData(self, lu):
521 """Collects all data for this query.
523 @return: Query data object
526 raise NotImplementedError()
528 def NewStyleQuery(self, lu):
529 """Collect data and execute query.
532 return query.GetQueryResponse(self.query, self._GetQueryData(lu))
534 def OldStyleQuery(self, lu):
535 """Collect data and execute query.
538 return self.query.OldStyleQuery(self._GetQueryData(lu))
541 def _GetWantedNodes(lu, nodes):
542 """Returns list of checked and expanded node names.
544 @type lu: L{LogicalUnit}
545 @param lu: the logical unit on whose behalf we execute
547 @param nodes: list of node names or None for all nodes
549 @return: the list of nodes, sorted
550 @raise errors.ProgrammerError: if the nodes parameter is wrong type
554 return [_ExpandNodeName(lu.cfg, name) for name in nodes]
556 return utils.NiceSort(lu.cfg.GetNodeList())
559 def _GetWantedInstances(lu, instances):
560 """Returns list of checked and expanded instance names.
562 @type lu: L{LogicalUnit}
563 @param lu: the logical unit on whose behalf we execute
564 @type instances: list
565 @param instances: list of instance names or None for all instances
567 @return: the list of instances, sorted
568 @raise errors.OpPrereqError: if the instances parameter is wrong type
569 @raise errors.OpPrereqError: if any of the passed instances is not found
573 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
575 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
579 def _GetUpdatedParams(old_params, update_dict,
580 use_default=True, use_none=False):
581 """Return the new version of a parameter dictionary.
583 @type old_params: dict
584 @param old_params: old parameters
585 @type update_dict: dict
586 @param update_dict: dict containing new parameter values, or
587 constants.VALUE_DEFAULT to reset the parameter to its default
589 @param use_default: boolean
590 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
591 values as 'to be deleted' values
592 @param use_none: boolean
593 @type use_none: whether to recognise C{None} values as 'to be
596 @return: the new parameter dictionary
599 params_copy = copy.deepcopy(old_params)
600 for key, val in update_dict.iteritems():
601 if ((use_default and val == constants.VALUE_DEFAULT) or
602 (use_none and val is None)):
608 params_copy[key] = val
612 def _CheckOutputFields(static, dynamic, selected):
613 """Checks whether all selected fields are valid.
615 @type static: L{utils.FieldSet}
616 @param static: static fields set
617 @type dynamic: L{utils.FieldSet}
618 @param dynamic: dynamic fields set
625 delta = f.NonMatching(selected)
627 raise errors.OpPrereqError("Unknown output fields selected: %s"
628 % ",".join(delta), errors.ECODE_INVAL)
631 def _CheckGlobalHvParams(params):
632 """Validates that given hypervisor params are not global ones.
634 This will ensure that instances don't get customised versions of
638 used_globals = constants.HVC_GLOBALS.intersection(params)
640 msg = ("The following hypervisor parameters are global and cannot"
641 " be customized at instance level, please modify them at"
642 " cluster level: %s" % utils.CommaJoin(used_globals))
643 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
646 def _CheckNodeOnline(lu, node, msg=None):
647 """Ensure that a given node is online.
649 @param lu: the LU on behalf of which we make the check
650 @param node: the node to check
651 @param msg: if passed, should be a message to replace the default one
652 @raise errors.OpPrereqError: if the node is offline
656 msg = "Can't use offline node"
657 if lu.cfg.GetNodeInfo(node).offline:
658 raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
661 def _CheckNodeNotDrained(lu, node):
662 """Ensure that a given node is not drained.
664 @param lu: the LU on behalf of which we make the check
665 @param node: the node to check
666 @raise errors.OpPrereqError: if the node is drained
669 if lu.cfg.GetNodeInfo(node).drained:
670 raise errors.OpPrereqError("Can't use drained node %s" % node,
674 def _CheckNodeVmCapable(lu, node):
675 """Ensure that a given node is vm capable.
677 @param lu: the LU on behalf of which we make the check
678 @param node: the node to check
679 @raise errors.OpPrereqError: if the node is not vm capable
682 if not lu.cfg.GetNodeInfo(node).vm_capable:
683 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
687 def _CheckNodeHasOS(lu, node, os_name, force_variant):
688 """Ensure that a node supports a given OS.
690 @param lu: the LU on behalf of which we make the check
691 @param node: the node to check
692 @param os_name: the OS to query about
693 @param force_variant: whether to ignore variant errors
694 @raise errors.OpPrereqError: if the node is not supporting the OS
697 result = lu.rpc.call_os_get(node, os_name)
698 result.Raise("OS '%s' not in supported OS list for node %s" %
700 prereq=True, ecode=errors.ECODE_INVAL)
701 if not force_variant:
702 _CheckOSVariant(result.payload, os_name)
705 def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
706 """Ensure that a node has the given secondary ip.
708 @type lu: L{LogicalUnit}
709 @param lu: the LU on behalf of which we make the check
711 @param node: the node to check
712 @type secondary_ip: string
713 @param secondary_ip: the ip to check
714 @type prereq: boolean
715 @param prereq: whether to throw a prerequisite or an execute error
716 @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
717 @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
720 result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
721 result.Raise("Failure checking secondary ip on node %s" % node,
722 prereq=prereq, ecode=errors.ECODE_ENVIRON)
723 if not result.payload:
724 msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
725 " please fix and re-run this command" % secondary_ip)
727 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
729 raise errors.OpExecError(msg)
732 def _GetClusterDomainSecret():
733 """Reads the cluster domain secret.
736 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
740 def _CheckInstanceDown(lu, instance, reason):
741 """Ensure that an instance is not running."""
742 if instance.admin_up:
743 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
744 (instance.name, reason), errors.ECODE_STATE)
746 pnode = instance.primary_node
747 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
748 ins_l.Raise("Can't contact node %s for instance information" % pnode,
749 prereq=True, ecode=errors.ECODE_ENVIRON)
751 if instance.name in ins_l.payload:
752 raise errors.OpPrereqError("Instance %s is running, %s" %
753 (instance.name, reason), errors.ECODE_STATE)
756 def _ExpandItemName(fn, name, kind):
757 """Expand an item name.
759 @param fn: the function to use for expansion
760 @param name: requested item name
761 @param kind: text description ('Node' or 'Instance')
762 @return: the resolved (full) name
763 @raise errors.OpPrereqError: if the item is not found
767 if full_name is None:
768 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
773 def _ExpandNodeName(cfg, name):
774 """Wrapper over L{_ExpandItemName} for nodes."""
775 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
778 def _ExpandInstanceName(cfg, name):
779 """Wrapper over L{_ExpandItemName} for instance."""
780 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
783 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
784 memory, vcpus, nics, disk_template, disks,
785 bep, hvp, hypervisor_name):
786 """Builds instance related env variables for hooks
788 This builds the hook environment from individual variables.
791 @param name: the name of the instance
792 @type primary_node: string
793 @param primary_node: the name of the instance's primary node
794 @type secondary_nodes: list
795 @param secondary_nodes: list of secondary nodes as strings
796 @type os_type: string
797 @param os_type: the name of the instance's OS
798 @type status: boolean
799 @param status: the should_run status of the instance
801 @param memory: the memory size of the instance
803 @param vcpus: the count of VCPUs the instance has
805 @param nics: list of tuples (ip, mac, mode, link) representing
806 the NICs the instance has
807 @type disk_template: string
808 @param disk_template: the disk template of the instance
810 @param disks: the list of (size, mode) pairs
812 @param bep: the backend parameters for the instance
814 @param hvp: the hypervisor parameters for the instance
815 @type hypervisor_name: string
816 @param hypervisor_name: the hypervisor for the instance
818 @return: the hook environment for this instance
827 "INSTANCE_NAME": name,
828 "INSTANCE_PRIMARY": primary_node,
829 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
830 "INSTANCE_OS_TYPE": os_type,
831 "INSTANCE_STATUS": str_status,
832 "INSTANCE_MEMORY": memory,
833 "INSTANCE_VCPUS": vcpus,
834 "INSTANCE_DISK_TEMPLATE": disk_template,
835 "INSTANCE_HYPERVISOR": hypervisor_name,
839 nic_count = len(nics)
840 for idx, (ip, mac, mode, link) in enumerate(nics):
843 env["INSTANCE_NIC%d_IP" % idx] = ip
844 env["INSTANCE_NIC%d_MAC" % idx] = mac
845 env["INSTANCE_NIC%d_MODE" % idx] = mode
846 env["INSTANCE_NIC%d_LINK" % idx] = link
847 if mode == constants.NIC_MODE_BRIDGED:
848 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
852 env["INSTANCE_NIC_COUNT"] = nic_count
855 disk_count = len(disks)
856 for idx, (size, mode) in enumerate(disks):
857 env["INSTANCE_DISK%d_SIZE" % idx] = size
858 env["INSTANCE_DISK%d_MODE" % idx] = mode
862 env["INSTANCE_DISK_COUNT"] = disk_count
864 for source, kind in [(bep, "BE"), (hvp, "HV")]:
865 for key, value in source.items():
866 env["INSTANCE_%s_%s" % (kind, key)] = value
871 def _NICListToTuple(lu, nics):
872 """Build a list of nic information tuples.
874 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
875 value in LUInstanceQueryData.
877 @type lu: L{LogicalUnit}
878 @param lu: the logical unit on whose behalf we execute
879 @type nics: list of L{objects.NIC}
880 @param nics: list of nics to convert to hooks tuples
884 cluster = lu.cfg.GetClusterInfo()
888 filled_params = cluster.SimpleFillNIC(nic.nicparams)
889 mode = filled_params[constants.NIC_MODE]
890 link = filled_params[constants.NIC_LINK]
891 hooks_nics.append((ip, mac, mode, link))
895 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
896 """Builds instance related env variables for hooks from an object.
898 @type lu: L{LogicalUnit}
899 @param lu: the logical unit on whose behalf we execute
900 @type instance: L{objects.Instance}
901 @param instance: the instance for which we should build the
904 @param override: dictionary with key/values that will override
907 @return: the hook environment dictionary
910 cluster = lu.cfg.GetClusterInfo()
911 bep = cluster.FillBE(instance)
912 hvp = cluster.FillHV(instance)
914 'name': instance.name,
915 'primary_node': instance.primary_node,
916 'secondary_nodes': instance.secondary_nodes,
917 'os_type': instance.os,
918 'status': instance.admin_up,
919 'memory': bep[constants.BE_MEMORY],
920 'vcpus': bep[constants.BE_VCPUS],
921 'nics': _NICListToTuple(lu, instance.nics),
922 'disk_template': instance.disk_template,
923 'disks': [(disk.size, disk.mode) for disk in instance.disks],
926 'hypervisor_name': instance.hypervisor,
929 args.update(override)
930 return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
933 def _AdjustCandidatePool(lu, exceptions):
934 """Adjust the candidate pool after node operations.
937 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
939 lu.LogInfo("Promoted nodes to master candidate role: %s",
940 utils.CommaJoin(node.name for node in mod_list))
941 for name in mod_list:
942 lu.context.ReaddNode(name)
943 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
945 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
949 def _DecideSelfPromotion(lu, exceptions=None):
950 """Decide whether I should promote myself as a master candidate.
953 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
954 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
955 # the new node will increase mc_max with one, so:
956 mc_should = min(mc_should + 1, cp_size)
957 return mc_now < mc_should
960 def _CheckNicsBridgesExist(lu, target_nics, target_node):
961 """Check that the brigdes needed by a list of nics exist.
964 cluster = lu.cfg.GetClusterInfo()
965 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
966 brlist = [params[constants.NIC_LINK] for params in paramslist
967 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
969 result = lu.rpc.call_bridges_exist(target_node, brlist)
970 result.Raise("Error checking bridges on destination node '%s'" %
971 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
974 def _CheckInstanceBridgesExist(lu, instance, node=None):
975 """Check that the brigdes needed by an instance exist.
979 node = instance.primary_node
980 _CheckNicsBridgesExist(lu, instance.nics, node)
983 def _CheckOSVariant(os_obj, name):
984 """Check whether an OS name conforms to the os variants specification.
986 @type os_obj: L{objects.OS}
987 @param os_obj: OS object to check
989 @param name: OS name passed by the user, to check for validity
992 if not os_obj.supported_variants:
994 variant = objects.OS.GetVariant(name)
996 raise errors.OpPrereqError("OS name must include a variant",
999 if variant not in os_obj.supported_variants:
1000 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1003 def _GetNodeInstancesInner(cfg, fn):
1004 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1007 def _GetNodeInstances(cfg, node_name):
1008 """Returns a list of all primary and secondary instances on a node.
1012 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1015 def _GetNodePrimaryInstances(cfg, node_name):
1016 """Returns primary instances on a node.
1019 return _GetNodeInstancesInner(cfg,
1020 lambda inst: node_name == inst.primary_node)
1023 def _GetNodeSecondaryInstances(cfg, node_name):
1024 """Returns secondary instances on a node.
1027 return _GetNodeInstancesInner(cfg,
1028 lambda inst: node_name in inst.secondary_nodes)
1031 def _GetStorageTypeArgs(cfg, storage_type):
1032 """Returns the arguments for a storage type.
1035 # Special case for file storage
1036 if storage_type == constants.ST_FILE:
1037 # storage.FileStorage wants a list of storage directories
1038 return [[cfg.GetFileStorageDir()]]
1043 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1046 for dev in instance.disks:
1047 cfg.SetDiskID(dev, node_name)
1049 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1050 result.Raise("Failed to get disk status from node %s" % node_name,
1051 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1053 for idx, bdev_status in enumerate(result.payload):
1054 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1060 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1061 """Check the sanity of iallocator and node arguments and use the
1062 cluster-wide iallocator if appropriate.
1064 Check that at most one of (iallocator, node) is specified. If none is
1065 specified, then the LU's opcode's iallocator slot is filled with the
1066 cluster-wide default iallocator.
1068 @type iallocator_slot: string
1069 @param iallocator_slot: the name of the opcode iallocator slot
1070 @type node_slot: string
1071 @param node_slot: the name of the opcode target node slot
1074 node = getattr(lu.op, node_slot, None)
1075 iallocator = getattr(lu.op, iallocator_slot, None)
1077 if node is not None and iallocator is not None:
1078 raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1080 elif node is None and iallocator is None:
1081 default_iallocator = lu.cfg.GetDefaultIAllocator()
1082 if default_iallocator:
1083 setattr(lu.op, iallocator_slot, default_iallocator)
1085 raise errors.OpPrereqError("No iallocator or node given and no"
1086 " cluster-wide default iallocator found."
1087 " Please specify either an iallocator or a"
1088 " node, or set a cluster-wide default"
1092 class LUClusterPostInit(LogicalUnit):
1093 """Logical unit for running hooks after cluster initialization.
1096 HPATH = "cluster-init"
1097 HTYPE = constants.HTYPE_CLUSTER
1099 def BuildHooksEnv(self):
1103 env = {"OP_TARGET": self.cfg.GetClusterName()}
1104 mn = self.cfg.GetMasterNode()
1105 return env, [], [mn]
1107 def Exec(self, feedback_fn):
1114 class LUClusterDestroy(LogicalUnit):
1115 """Logical unit for destroying the cluster.
1118 HPATH = "cluster-destroy"
1119 HTYPE = constants.HTYPE_CLUSTER
1121 def BuildHooksEnv(self):
1125 env = {"OP_TARGET": self.cfg.GetClusterName()}
1128 def CheckPrereq(self):
1129 """Check prerequisites.
1131 This checks whether the cluster is empty.
1133 Any errors are signaled by raising errors.OpPrereqError.
1136 master = self.cfg.GetMasterNode()
1138 nodelist = self.cfg.GetNodeList()
1139 if len(nodelist) != 1 or nodelist[0] != master:
1140 raise errors.OpPrereqError("There are still %d node(s) in"
1141 " this cluster." % (len(nodelist) - 1),
1143 instancelist = self.cfg.GetInstanceList()
1145 raise errors.OpPrereqError("There are still %d instance(s) in"
1146 " this cluster." % len(instancelist),
1149 def Exec(self, feedback_fn):
1150 """Destroys the cluster.
1153 master = self.cfg.GetMasterNode()
1155 # Run post hooks on master node before it's removed
1156 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1158 hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1160 # pylint: disable-msg=W0702
1161 self.LogWarning("Errors occurred running hooks on %s" % master)
1163 result = self.rpc.call_node_stop_master(master, False)
1164 result.Raise("Could not disable the master role")
1169 def _VerifyCertificate(filename):
1170 """Verifies a certificate for LUClusterVerify.
1172 @type filename: string
1173 @param filename: Path to PEM file
1177 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1178 utils.ReadFile(filename))
1179 except Exception, err: # pylint: disable-msg=W0703
1180 return (LUClusterVerify.ETYPE_ERROR,
1181 "Failed to load X509 certificate %s: %s" % (filename, err))
1184 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1185 constants.SSL_CERT_EXPIRATION_ERROR)
1188 fnamemsg = "While verifying %s: %s" % (filename, msg)
1193 return (None, fnamemsg)
1194 elif errcode == utils.CERT_WARNING:
1195 return (LUClusterVerify.ETYPE_WARNING, fnamemsg)
1196 elif errcode == utils.CERT_ERROR:
1197 return (LUClusterVerify.ETYPE_ERROR, fnamemsg)
1199 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1202 class LUClusterVerify(LogicalUnit):
1203 """Verifies the cluster status.
1206 HPATH = "cluster-verify"
1207 HTYPE = constants.HTYPE_CLUSTER
1210 TCLUSTER = "cluster"
1212 TINSTANCE = "instance"
1214 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1215 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1216 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1217 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1218 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1219 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1220 EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1221 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1222 EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1223 ENODEDRBD = (TNODE, "ENODEDRBD")
1224 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1225 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1226 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1227 ENODEHV = (TNODE, "ENODEHV")
1228 ENODELVM = (TNODE, "ENODELVM")
1229 ENODEN1 = (TNODE, "ENODEN1")
1230 ENODENET = (TNODE, "ENODENET")
1231 ENODEOS = (TNODE, "ENODEOS")
1232 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1233 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1234 ENODERPC = (TNODE, "ENODERPC")
1235 ENODESSH = (TNODE, "ENODESSH")
1236 ENODEVERSION = (TNODE, "ENODEVERSION")
1237 ENODESETUP = (TNODE, "ENODESETUP")
1238 ENODETIME = (TNODE, "ENODETIME")
1239 ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1241 ETYPE_FIELD = "code"
1242 ETYPE_ERROR = "ERROR"
1243 ETYPE_WARNING = "WARNING"
1245 _HOOKS_INDENT_RE = re.compile("^", re.M)
1247 class NodeImage(object):
1248 """A class representing the logical and physical status of a node.
1251 @ivar name: the node name to which this object refers
1252 @ivar volumes: a structure as returned from
1253 L{ganeti.backend.GetVolumeList} (runtime)
1254 @ivar instances: a list of running instances (runtime)
1255 @ivar pinst: list of configured primary instances (config)
1256 @ivar sinst: list of configured secondary instances (config)
1257 @ivar sbp: dictionary of {primary-node: list of instances} for all
1258 instances for which this node is secondary (config)
1259 @ivar mfree: free memory, as reported by hypervisor (runtime)
1260 @ivar dfree: free disk, as reported by the node (runtime)
1261 @ivar offline: the offline status (config)
1262 @type rpc_fail: boolean
1263 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1264 not whether the individual keys were correct) (runtime)
1265 @type lvm_fail: boolean
1266 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1267 @type hyp_fail: boolean
1268 @ivar hyp_fail: whether the RPC call didn't return the instance list
1269 @type ghost: boolean
1270 @ivar ghost: whether this is a known node or not (config)
1271 @type os_fail: boolean
1272 @ivar os_fail: whether the RPC call didn't return valid OS data
1274 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1275 @type vm_capable: boolean
1276 @ivar vm_capable: whether the node can host instances
1279 def __init__(self, offline=False, name=None, vm_capable=True):
1288 self.offline = offline
1289 self.vm_capable = vm_capable
1290 self.rpc_fail = False
1291 self.lvm_fail = False
1292 self.hyp_fail = False
1294 self.os_fail = False
1297 def ExpandNames(self):
1298 self.needed_locks = {
1299 locking.LEVEL_NODE: locking.ALL_SET,
1300 locking.LEVEL_INSTANCE: locking.ALL_SET,
1302 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1304 def _Error(self, ecode, item, msg, *args, **kwargs):
1305 """Format an error message.
1307 Based on the opcode's error_codes parameter, either format a
1308 parseable error code, or a simpler error string.
1310 This must be called only from Exec and functions called from Exec.
1313 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1315 # first complete the msg
1318 # then format the whole message
1319 if self.op.error_codes:
1320 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1326 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1327 # and finally report it via the feedback_fn
1328 self._feedback_fn(" - %s" % msg)
1330 def _ErrorIf(self, cond, *args, **kwargs):
1331 """Log an error message if the passed condition is True.
1334 cond = bool(cond) or self.op.debug_simulate_errors
1336 self._Error(*args, **kwargs)
1337 # do not mark the operation as failed for WARN cases only
1338 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1339 self.bad = self.bad or cond
1341 def _VerifyNode(self, ninfo, nresult):
1342 """Perform some basic validation on data returned from a node.
1344 - check the result data structure is well formed and has all the
1346 - check ganeti version
1348 @type ninfo: L{objects.Node}
1349 @param ninfo: the node to check
1350 @param nresult: the results from the node
1352 @return: whether overall this call was successful (and we can expect
1353 reasonable values in the respose)
1357 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1359 # main result, nresult should be a non-empty dict
1360 test = not nresult or not isinstance(nresult, dict)
1361 _ErrorIf(test, self.ENODERPC, node,
1362 "unable to verify node: no data returned")
1366 # compares ganeti version
1367 local_version = constants.PROTOCOL_VERSION
1368 remote_version = nresult.get("version", None)
1369 test = not (remote_version and
1370 isinstance(remote_version, (list, tuple)) and
1371 len(remote_version) == 2)
1372 _ErrorIf(test, self.ENODERPC, node,
1373 "connection to node returned invalid data")
1377 test = local_version != remote_version[0]
1378 _ErrorIf(test, self.ENODEVERSION, node,
1379 "incompatible protocol versions: master %s,"
1380 " node %s", local_version, remote_version[0])
1384 # node seems compatible, we can actually try to look into its results
1386 # full package version
1387 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1388 self.ENODEVERSION, node,
1389 "software version mismatch: master %s, node %s",
1390 constants.RELEASE_VERSION, remote_version[1],
1391 code=self.ETYPE_WARNING)
1393 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1394 if ninfo.vm_capable and isinstance(hyp_result, dict):
1395 for hv_name, hv_result in hyp_result.iteritems():
1396 test = hv_result is not None
1397 _ErrorIf(test, self.ENODEHV, node,
1398 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1400 hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1401 if ninfo.vm_capable and isinstance(hvp_result, list):
1402 for item, hv_name, hv_result in hvp_result:
1403 _ErrorIf(True, self.ENODEHV, node,
1404 "hypervisor %s parameter verify failure (source %s): %s",
1405 hv_name, item, hv_result)
1407 test = nresult.get(constants.NV_NODESETUP,
1408 ["Missing NODESETUP results"])
1409 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1414 def _VerifyNodeTime(self, ninfo, nresult,
1415 nvinfo_starttime, nvinfo_endtime):
1416 """Check the node time.
1418 @type ninfo: L{objects.Node}
1419 @param ninfo: the node to check
1420 @param nresult: the remote results for the node
1421 @param nvinfo_starttime: the start time of the RPC call
1422 @param nvinfo_endtime: the end time of the RPC call
1426 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1428 ntime = nresult.get(constants.NV_TIME, None)
1430 ntime_merged = utils.MergeTime(ntime)
1431 except (ValueError, TypeError):
1432 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1435 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1436 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1437 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1438 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1442 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1443 "Node time diverges by at least %s from master node time",
1446 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1447 """Check the node time.
1449 @type ninfo: L{objects.Node}
1450 @param ninfo: the node to check
1451 @param nresult: the remote results for the node
1452 @param vg_name: the configured VG name
1459 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1461 # checks vg existence and size > 20G
1462 vglist = nresult.get(constants.NV_VGLIST, None)
1464 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1466 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1467 constants.MIN_VG_SIZE)
1468 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1471 pvlist = nresult.get(constants.NV_PVLIST, None)
1472 test = pvlist is None
1473 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1475 # check that ':' is not present in PV names, since it's a
1476 # special character for lvcreate (denotes the range of PEs to
1478 for _, pvname, owner_vg in pvlist:
1479 test = ":" in pvname
1480 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1481 " '%s' of VG '%s'", pvname, owner_vg)
1483 def _VerifyNodeNetwork(self, ninfo, nresult):
1484 """Check the node time.
1486 @type ninfo: L{objects.Node}
1487 @param ninfo: the node to check
1488 @param nresult: the remote results for the node
1492 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1494 test = constants.NV_NODELIST not in nresult
1495 _ErrorIf(test, self.ENODESSH, node,
1496 "node hasn't returned node ssh connectivity data")
1498 if nresult[constants.NV_NODELIST]:
1499 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1500 _ErrorIf(True, self.ENODESSH, node,
1501 "ssh communication with node '%s': %s", a_node, a_msg)
1503 test = constants.NV_NODENETTEST not in nresult
1504 _ErrorIf(test, self.ENODENET, node,
1505 "node hasn't returned node tcp connectivity data")
1507 if nresult[constants.NV_NODENETTEST]:
1508 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1510 _ErrorIf(True, self.ENODENET, node,
1511 "tcp communication with node '%s': %s",
1512 anode, nresult[constants.NV_NODENETTEST][anode])
1514 test = constants.NV_MASTERIP not in nresult
1515 _ErrorIf(test, self.ENODENET, node,
1516 "node hasn't returned node master IP reachability data")
1518 if not nresult[constants.NV_MASTERIP]:
1519 if node == self.master_node:
1520 msg = "the master node cannot reach the master IP (not configured?)"
1522 msg = "cannot reach the master IP"
1523 _ErrorIf(True, self.ENODENET, node, msg)
1525 def _VerifyInstance(self, instance, instanceconfig, node_image,
1527 """Verify an instance.
1529 This function checks to see if the required block devices are
1530 available on the instance's node.
1533 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1534 node_current = instanceconfig.primary_node
1536 node_vol_should = {}
1537 instanceconfig.MapLVsByNode(node_vol_should)
1539 for node in node_vol_should:
1540 n_img = node_image[node]
1541 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1542 # ignore missing volumes on offline or broken nodes
1544 for volume in node_vol_should[node]:
1545 test = volume not in n_img.volumes
1546 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1547 "volume %s missing on node %s", volume, node)
1549 if instanceconfig.admin_up:
1550 pri_img = node_image[node_current]
1551 test = instance not in pri_img.instances and not pri_img.offline
1552 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1553 "instance not running on its primary node %s",
1556 for node, n_img in node_image.items():
1557 if node != node_current:
1558 test = instance in n_img.instances
1559 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1560 "instance should not run on node %s", node)
1562 diskdata = [(nname, success, status, idx)
1563 for (nname, disks) in diskstatus.items()
1564 for idx, (success, status) in enumerate(disks)]
1566 for nname, success, bdev_status, idx in diskdata:
1567 # the 'ghost node' construction in Exec() ensures that we have a
1569 snode = node_image[nname]
1570 bad_snode = snode.ghost or snode.offline
1571 _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
1572 self.EINSTANCEFAULTYDISK, instance,
1573 "couldn't retrieve status for disk/%s on %s: %s",
1574 idx, nname, bdev_status)
1575 _ErrorIf((instanceconfig.admin_up and success and
1576 bdev_status.ldisk_status == constants.LDS_FAULTY),
1577 self.EINSTANCEFAULTYDISK, instance,
1578 "disk/%s on %s is faulty", idx, nname)
1580 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1581 """Verify if there are any unknown volumes in the cluster.
1583 The .os, .swap and backup volumes are ignored. All other volumes are
1584 reported as unknown.
1586 @type reserved: L{ganeti.utils.FieldSet}
1587 @param reserved: a FieldSet of reserved volume names
1590 for node, n_img in node_image.items():
1591 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1592 # skip non-healthy nodes
1594 for volume in n_img.volumes:
1595 test = ((node not in node_vol_should or
1596 volume not in node_vol_should[node]) and
1597 not reserved.Matches(volume))
1598 self._ErrorIf(test, self.ENODEORPHANLV, node,
1599 "volume %s is unknown", volume)
1601 def _VerifyOrphanInstances(self, instancelist, node_image):
1602 """Verify the list of running instances.
1604 This checks what instances are running but unknown to the cluster.
1607 for node, n_img in node_image.items():
1608 for o_inst in n_img.instances:
1609 test = o_inst not in instancelist
1610 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1611 "instance %s on node %s should not exist", o_inst, node)
1613 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1614 """Verify N+1 Memory Resilience.
1616 Check that if one single node dies we can still start all the
1617 instances it was primary for.
1620 cluster_info = self.cfg.GetClusterInfo()
1621 for node, n_img in node_image.items():
1622 # This code checks that every node which is now listed as
1623 # secondary has enough memory to host all instances it is
1624 # supposed to should a single other node in the cluster fail.
1625 # FIXME: not ready for failover to an arbitrary node
1626 # FIXME: does not support file-backed instances
1627 # WARNING: we currently take into account down instances as well
1628 # as up ones, considering that even if they're down someone
1629 # might want to start them even in the event of a node failure.
1631 # we're skipping offline nodes from the N+1 warning, since
1632 # most likely we don't have good memory infromation from them;
1633 # we already list instances living on such nodes, and that's
1636 for prinode, instances in n_img.sbp.items():
1638 for instance in instances:
1639 bep = cluster_info.FillBE(instance_cfg[instance])
1640 if bep[constants.BE_AUTO_BALANCE]:
1641 needed_mem += bep[constants.BE_MEMORY]
1642 test = n_img.mfree < needed_mem
1643 self._ErrorIf(test, self.ENODEN1, node,
1644 "not enough memory to accomodate instance failovers"
1645 " should node %s fail", prinode)
1647 def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1649 """Verifies and computes the node required file checksums.
1651 @type ninfo: L{objects.Node}
1652 @param ninfo: the node to check
1653 @param nresult: the remote results for the node
1654 @param file_list: required list of files
1655 @param local_cksum: dictionary of local files and their checksums
1656 @param master_files: list of files that only masters should have
1660 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1662 remote_cksum = nresult.get(constants.NV_FILELIST, None)
1663 test = not isinstance(remote_cksum, dict)
1664 _ErrorIf(test, self.ENODEFILECHECK, node,
1665 "node hasn't returned file checksum data")
1669 for file_name in file_list:
1670 node_is_mc = ninfo.master_candidate
1671 must_have = (file_name not in master_files) or node_is_mc
1673 test1 = file_name not in remote_cksum
1675 test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1677 test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1678 _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1679 "file '%s' missing", file_name)
1680 _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1681 "file '%s' has wrong checksum", file_name)
1682 # not candidate and this is not a must-have file
1683 _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1684 "file '%s' should not exist on non master"
1685 " candidates (and the file is outdated)", file_name)
1686 # all good, except non-master/non-must have combination
1687 _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1688 "file '%s' should not exist"
1689 " on non master candidates", file_name)
1691 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1693 """Verifies and the node DRBD status.
1695 @type ninfo: L{objects.Node}
1696 @param ninfo: the node to check
1697 @param nresult: the remote results for the node
1698 @param instanceinfo: the dict of instances
1699 @param drbd_helper: the configured DRBD usermode helper
1700 @param drbd_map: the DRBD map as returned by
1701 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1705 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1708 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1709 test = (helper_result == None)
1710 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1711 "no drbd usermode helper returned")
1713 status, payload = helper_result
1715 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1716 "drbd usermode helper check unsuccessful: %s", payload)
1717 test = status and (payload != drbd_helper)
1718 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1719 "wrong drbd usermode helper: %s", payload)
1721 # compute the DRBD minors
1723 for minor, instance in drbd_map[node].items():
1724 test = instance not in instanceinfo
1725 _ErrorIf(test, self.ECLUSTERCFG, None,
1726 "ghost instance '%s' in temporary DRBD map", instance)
1727 # ghost instance should not be running, but otherwise we
1728 # don't give double warnings (both ghost instance and
1729 # unallocated minor in use)
1731 node_drbd[minor] = (instance, False)
1733 instance = instanceinfo[instance]
1734 node_drbd[minor] = (instance.name, instance.admin_up)
1736 # and now check them
1737 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1738 test = not isinstance(used_minors, (tuple, list))
1739 _ErrorIf(test, self.ENODEDRBD, node,
1740 "cannot parse drbd status file: %s", str(used_minors))
1742 # we cannot check drbd status
1745 for minor, (iname, must_exist) in node_drbd.items():
1746 test = minor not in used_minors and must_exist
1747 _ErrorIf(test, self.ENODEDRBD, node,
1748 "drbd minor %d of instance %s is not active", minor, iname)
1749 for minor in used_minors:
1750 test = minor not in node_drbd
1751 _ErrorIf(test, self.ENODEDRBD, node,
1752 "unallocated drbd minor %d is in use", minor)
1754 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1755 """Builds the node OS structures.
1757 @type ninfo: L{objects.Node}
1758 @param ninfo: the node to check
1759 @param nresult: the remote results for the node
1760 @param nimg: the node image object
1764 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1766 remote_os = nresult.get(constants.NV_OSLIST, None)
1767 test = (not isinstance(remote_os, list) or
1768 not compat.all(isinstance(v, list) and len(v) == 7
1769 for v in remote_os))
1771 _ErrorIf(test, self.ENODEOS, node,
1772 "node hasn't returned valid OS data")
1781 for (name, os_path, status, diagnose,
1782 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1784 if name not in os_dict:
1787 # parameters is a list of lists instead of list of tuples due to
1788 # JSON lacking a real tuple type, fix it:
1789 parameters = [tuple(v) for v in parameters]
1790 os_dict[name].append((os_path, status, diagnose,
1791 set(variants), set(parameters), set(api_ver)))
1793 nimg.oslist = os_dict
1795 def _VerifyNodeOS(self, ninfo, nimg, base):
1796 """Verifies the node OS list.
1798 @type ninfo: L{objects.Node}
1799 @param ninfo: the node to check
1800 @param nimg: the node image object
1801 @param base: the 'template' node we match against (e.g. from the master)
1805 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1807 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1809 for os_name, os_data in nimg.oslist.items():
1810 assert os_data, "Empty OS status for OS %s?!" % os_name
1811 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1812 _ErrorIf(not f_status, self.ENODEOS, node,
1813 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1814 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1815 "OS '%s' has multiple entries (first one shadows the rest): %s",
1816 os_name, utils.CommaJoin([v[0] for v in os_data]))
1817 # this will catched in backend too
1818 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1819 and not f_var, self.ENODEOS, node,
1820 "OS %s with API at least %d does not declare any variant",
1821 os_name, constants.OS_API_V15)
1822 # comparisons with the 'base' image
1823 test = os_name not in base.oslist
1824 _ErrorIf(test, self.ENODEOS, node,
1825 "Extra OS %s not present on reference node (%s)",
1829 assert base.oslist[os_name], "Base node has empty OS status?"
1830 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1832 # base OS is invalid, skipping
1834 for kind, a, b in [("API version", f_api, b_api),
1835 ("variants list", f_var, b_var),
1836 ("parameters", f_param, b_param)]:
1837 _ErrorIf(a != b, self.ENODEOS, node,
1838 "OS %s %s differs from reference node %s: %s vs. %s",
1839 kind, os_name, base.name,
1840 utils.CommaJoin(a), utils.CommaJoin(b))
1842 # check any missing OSes
1843 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1844 _ErrorIf(missing, self.ENODEOS, node,
1845 "OSes present on reference node %s but missing on this node: %s",
1846 base.name, utils.CommaJoin(missing))
1848 def _VerifyOob(self, ninfo, nresult):
1849 """Verifies out of band functionality of a node.
1851 @type ninfo: L{objects.Node}
1852 @param ninfo: the node to check
1853 @param nresult: the remote results for the node
1857 # We just have to verify the paths on master and/or master candidates
1858 # as the oob helper is invoked on the master
1859 if ((ninfo.master_candidate or ninfo.master_capable) and
1860 constants.NV_OOB_PATHS in nresult):
1861 for path_result in nresult[constants.NV_OOB_PATHS]:
1862 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1864 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1865 """Verifies and updates the node volume data.
1867 This function will update a L{NodeImage}'s internal structures
1868 with data from the remote call.
1870 @type ninfo: L{objects.Node}
1871 @param ninfo: the node to check
1872 @param nresult: the remote results for the node
1873 @param nimg: the node image object
1874 @param vg_name: the configured VG name
1878 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1880 nimg.lvm_fail = True
1881 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1884 elif isinstance(lvdata, basestring):
1885 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1886 utils.SafeEncode(lvdata))
1887 elif not isinstance(lvdata, dict):
1888 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1890 nimg.volumes = lvdata
1891 nimg.lvm_fail = False
1893 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1894 """Verifies and updates the node instance list.
1896 If the listing was successful, then updates this node's instance
1897 list. Otherwise, it marks the RPC call as failed for the instance
1900 @type ninfo: L{objects.Node}
1901 @param ninfo: the node to check
1902 @param nresult: the remote results for the node
1903 @param nimg: the node image object
1906 idata = nresult.get(constants.NV_INSTANCELIST, None)
1907 test = not isinstance(idata, list)
1908 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1909 " (instancelist): %s", utils.SafeEncode(str(idata)))
1911 nimg.hyp_fail = True
1913 nimg.instances = idata
1915 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1916 """Verifies and computes a node information map
1918 @type ninfo: L{objects.Node}
1919 @param ninfo: the node to check
1920 @param nresult: the remote results for the node
1921 @param nimg: the node image object
1922 @param vg_name: the configured VG name
1926 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1928 # try to read free memory (from the hypervisor)
1929 hv_info = nresult.get(constants.NV_HVINFO, None)
1930 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1931 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1934 nimg.mfree = int(hv_info["memory_free"])
1935 except (ValueError, TypeError):
1936 _ErrorIf(True, self.ENODERPC, node,
1937 "node returned invalid nodeinfo, check hypervisor")
1939 # FIXME: devise a free space model for file based instances as well
1940 if vg_name is not None:
1941 test = (constants.NV_VGLIST not in nresult or
1942 vg_name not in nresult[constants.NV_VGLIST])
1943 _ErrorIf(test, self.ENODELVM, node,
1944 "node didn't return data for the volume group '%s'"
1945 " - it is either missing or broken", vg_name)
1948 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1949 except (ValueError, TypeError):
1950 _ErrorIf(True, self.ENODERPC, node,
1951 "node returned invalid LVM info, check LVM status")
1953 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1954 """Gets per-disk status information for all instances.
1956 @type nodelist: list of strings
1957 @param nodelist: Node names
1958 @type node_image: dict of (name, L{objects.Node})
1959 @param node_image: Node objects
1960 @type instanceinfo: dict of (name, L{objects.Instance})
1961 @param instanceinfo: Instance objects
1962 @rtype: {instance: {node: [(succes, payload)]}}
1963 @return: a dictionary of per-instance dictionaries with nodes as
1964 keys and disk information as values; the disk information is a
1965 list of tuples (success, payload)
1968 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1971 node_disks_devonly = {}
1972 diskless_instances = set()
1973 diskless = constants.DT_DISKLESS
1975 for nname in nodelist:
1976 node_instances = list(itertools.chain(node_image[nname].pinst,
1977 node_image[nname].sinst))
1978 diskless_instances.update(inst for inst in node_instances
1979 if instanceinfo[inst].disk_template == diskless)
1980 disks = [(inst, disk)
1981 for inst in node_instances
1982 for disk in instanceinfo[inst].disks]
1985 # No need to collect data
1988 node_disks[nname] = disks
1990 # Creating copies as SetDiskID below will modify the objects and that can
1991 # lead to incorrect data returned from nodes
1992 devonly = [dev.Copy() for (_, dev) in disks]
1995 self.cfg.SetDiskID(dev, nname)
1997 node_disks_devonly[nname] = devonly
1999 assert len(node_disks) == len(node_disks_devonly)
2001 # Collect data from all nodes with disks
2002 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2005 assert len(result) == len(node_disks)
2009 for (nname, nres) in result.items():
2010 disks = node_disks[nname]
2013 # No data from this node
2014 data = len(disks) * [(False, "node offline")]
2017 _ErrorIf(msg, self.ENODERPC, nname,
2018 "while getting disk information: %s", msg)
2020 # No data from this node
2021 data = len(disks) * [(False, msg)]
2024 for idx, i in enumerate(nres.payload):
2025 if isinstance(i, (tuple, list)) and len(i) == 2:
2028 logging.warning("Invalid result from node %s, entry %d: %s",
2030 data.append((False, "Invalid result from the remote node"))
2032 for ((inst, _), status) in zip(disks, data):
2033 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2035 # Add empty entries for diskless instances.
2036 for inst in diskless_instances:
2037 assert inst not in instdisk
2040 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2041 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2042 compat.all(isinstance(s, (tuple, list)) and
2043 len(s) == 2 for s in statuses)
2044 for inst, nnames in instdisk.items()
2045 for nname, statuses in nnames.items())
2046 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2050 def _VerifyHVP(self, hvp_data):
2051 """Verifies locally the syntax of the hypervisor parameters.
2054 for item, hv_name, hv_params in hvp_data:
2055 msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2058 hv_class = hypervisor.GetHypervisor(hv_name)
2059 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2060 hv_class.CheckParameterSyntax(hv_params)
2061 except errors.GenericError, err:
2062 self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2065 def BuildHooksEnv(self):
2068 Cluster-Verify hooks just ran in the post phase and their failure makes
2069 the output be logged in the verify output and the verification to fail.
2072 all_nodes = self.cfg.GetNodeList()
2074 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2076 for node in self.cfg.GetAllNodesInfo().values():
2077 env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2079 return env, [], all_nodes
2081 def Exec(self, feedback_fn):
2082 """Verify integrity of cluster, performing various test on nodes.
2085 # This method has too many local variables. pylint: disable-msg=R0914
2087 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2088 verbose = self.op.verbose
2089 self._feedback_fn = feedback_fn
2090 feedback_fn("* Verifying global settings")
2091 for msg in self.cfg.VerifyConfig():
2092 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2094 # Check the cluster certificates
2095 for cert_filename in constants.ALL_CERT_FILES:
2096 (errcode, msg) = _VerifyCertificate(cert_filename)
2097 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2099 vg_name = self.cfg.GetVGName()
2100 drbd_helper = self.cfg.GetDRBDHelper()
2101 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2102 cluster = self.cfg.GetClusterInfo()
2103 nodelist = utils.NiceSort(self.cfg.GetNodeList())
2104 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2105 nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2106 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2107 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2108 for iname in instancelist)
2109 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2110 i_non_redundant = [] # Non redundant instances
2111 i_non_a_balanced = [] # Non auto-balanced instances
2112 n_offline = 0 # Count of offline nodes
2113 n_drained = 0 # Count of nodes being drained
2114 node_vol_should = {}
2116 # FIXME: verify OS list
2117 # do local checksums
2118 master_files = [constants.CLUSTER_CONF_FILE]
2119 master_node = self.master_node = self.cfg.GetMasterNode()
2120 master_ip = self.cfg.GetMasterIP()
2122 file_names = ssconf.SimpleStore().GetFileList()
2123 file_names.extend(constants.ALL_CERT_FILES)
2124 file_names.extend(master_files)
2125 if cluster.modify_etc_hosts:
2126 file_names.append(constants.ETC_HOSTS)
2128 local_checksums = utils.FingerprintFiles(file_names)
2130 # Compute the set of hypervisor parameters
2132 for hv_name in hypervisors:
2133 hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2134 for os_name, os_hvp in cluster.os_hvp.items():
2135 for hv_name, hv_params in os_hvp.items():
2138 full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2139 hvp_data.append(("os %s" % os_name, hv_name, full_params))
2140 # TODO: collapse identical parameter values in a single one
2141 for instance in instanceinfo.values():
2142 if not instance.hvparams:
2144 hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2145 cluster.FillHV(instance)))
2146 # and verify them locally
2147 self._VerifyHVP(hvp_data)
2149 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2150 node_verify_param = {
2151 constants.NV_FILELIST: file_names,
2152 constants.NV_NODELIST: [node.name for node in nodeinfo
2153 if not node.offline],
2154 constants.NV_HYPERVISOR: hypervisors,
2155 constants.NV_HVPARAMS: hvp_data,
2156 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2157 node.secondary_ip) for node in nodeinfo
2158 if not node.offline],
2159 constants.NV_INSTANCELIST: hypervisors,
2160 constants.NV_VERSION: None,
2161 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2162 constants.NV_NODESETUP: None,
2163 constants.NV_TIME: None,
2164 constants.NV_MASTERIP: (master_node, master_ip),
2165 constants.NV_OSLIST: None,
2166 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2169 if vg_name is not None:
2170 node_verify_param[constants.NV_VGLIST] = None
2171 node_verify_param[constants.NV_LVLIST] = vg_name
2172 node_verify_param[constants.NV_PVLIST] = [vg_name]
2173 node_verify_param[constants.NV_DRBDLIST] = None
2176 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2178 # Build our expected cluster state
2179 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2181 vm_capable=node.vm_capable))
2182 for node in nodeinfo)
2186 for node in nodeinfo:
2187 path = _SupportsOob(self.cfg, node)
2188 if path and path not in oob_paths:
2189 oob_paths.append(path)
2192 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2194 for instance in instancelist:
2195 inst_config = instanceinfo[instance]
2197 for nname in inst_config.all_nodes:
2198 if nname not in node_image:
2200 gnode = self.NodeImage(name=nname)
2202 node_image[nname] = gnode
2204 inst_config.MapLVsByNode(node_vol_should)
2206 pnode = inst_config.primary_node
2207 node_image[pnode].pinst.append(instance)
2209 for snode in inst_config.secondary_nodes:
2210 nimg = node_image[snode]
2211 nimg.sinst.append(instance)
2212 if pnode not in nimg.sbp:
2213 nimg.sbp[pnode] = []
2214 nimg.sbp[pnode].append(instance)
2216 # At this point, we have the in-memory data structures complete,
2217 # except for the runtime information, which we'll gather next
2219 # Due to the way our RPC system works, exact response times cannot be
2220 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2221 # time before and after executing the request, we can at least have a time
2223 nvinfo_starttime = time.time()
2224 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2225 self.cfg.GetClusterName())
2226 nvinfo_endtime = time.time()
2228 all_drbd_map = self.cfg.ComputeDRBDMap()
2230 feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2231 instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2233 feedback_fn("* Verifying node status")
2237 for node_i in nodeinfo:
2239 nimg = node_image[node]
2243 feedback_fn("* Skipping offline node %s" % (node,))
2247 if node == master_node:
2249 elif node_i.master_candidate:
2250 ntype = "master candidate"
2251 elif node_i.drained:
2257 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2259 msg = all_nvinfo[node].fail_msg
2260 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2262 nimg.rpc_fail = True
2265 nresult = all_nvinfo[node].payload
2267 nimg.call_ok = self._VerifyNode(node_i, nresult)
2268 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2269 self._VerifyNodeNetwork(node_i, nresult)
2270 self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2273 self._VerifyOob(node_i, nresult)
2276 self._VerifyNodeLVM(node_i, nresult, vg_name)
2277 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2280 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2281 self._UpdateNodeInstances(node_i, nresult, nimg)
2282 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2283 self._UpdateNodeOS(node_i, nresult, nimg)
2284 if not nimg.os_fail:
2285 if refos_img is None:
2287 self._VerifyNodeOS(node_i, nimg, refos_img)
2289 feedback_fn("* Verifying instance status")
2290 for instance in instancelist:
2292 feedback_fn("* Verifying instance %s" % instance)
2293 inst_config = instanceinfo[instance]
2294 self._VerifyInstance(instance, inst_config, node_image,
2296 inst_nodes_offline = []
2298 pnode = inst_config.primary_node
2299 pnode_img = node_image[pnode]
2300 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2301 self.ENODERPC, pnode, "instance %s, connection to"
2302 " primary node failed", instance)
2304 _ErrorIf(pnode_img.offline, self.EINSTANCEBADNODE, instance,
2305 "instance lives on offline node %s", inst_config.primary_node)
2307 # If the instance is non-redundant we cannot survive losing its primary
2308 # node, so we are not N+1 compliant. On the other hand we have no disk
2309 # templates with more than one secondary so that situation is not well
2311 # FIXME: does not support file-backed instances
2312 if not inst_config.secondary_nodes:
2313 i_non_redundant.append(instance)
2315 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2316 instance, "instance has multiple secondary nodes: %s",
2317 utils.CommaJoin(inst_config.secondary_nodes),
2318 code=self.ETYPE_WARNING)
2320 if inst_config.disk_template in constants.DTS_NET_MIRROR:
2321 pnode = inst_config.primary_node
2322 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2323 instance_groups = {}
2325 for node in instance_nodes:
2326 instance_groups.setdefault(nodeinfo_byname[node].group,
2330 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2331 # Sort so that we always list the primary node first.
2332 for group, nodes in sorted(instance_groups.items(),
2333 key=lambda (_, nodes): pnode in nodes,
2336 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2337 instance, "instance has primary and secondary nodes in"
2338 " different groups: %s", utils.CommaJoin(pretty_list),
2339 code=self.ETYPE_WARNING)
2341 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2342 i_non_a_balanced.append(instance)
2344 for snode in inst_config.secondary_nodes:
2345 s_img = node_image[snode]
2346 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2347 "instance %s, connection to secondary node failed", instance)
2350 inst_nodes_offline.append(snode)
2352 # warn that the instance lives on offline nodes
2353 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2354 "instance has offline secondary node(s) %s",
2355 utils.CommaJoin(inst_nodes_offline))
2356 # ... or ghost/non-vm_capable nodes
2357 for node in inst_config.all_nodes:
2358 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2359 "instance lives on ghost node %s", node)
2360 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2361 instance, "instance lives on non-vm_capable node %s", node)
2363 feedback_fn("* Verifying orphan volumes")
2364 reserved = utils.FieldSet(*cluster.reserved_lvs)
2365 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2367 feedback_fn("* Verifying orphan instances")
2368 self._VerifyOrphanInstances(instancelist, node_image)
2370 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2371 feedback_fn("* Verifying N+1 Memory redundancy")
2372 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2374 feedback_fn("* Other Notes")
2376 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2377 % len(i_non_redundant))
2379 if i_non_a_balanced:
2380 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2381 % len(i_non_a_balanced))
2384 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2387 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2391 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2392 """Analyze the post-hooks' result
2394 This method analyses the hook result, handles it, and sends some
2395 nicely-formatted feedback back to the user.
2397 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2398 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2399 @param hooks_results: the results of the multi-node hooks rpc call
2400 @param feedback_fn: function used send feedback back to the caller
2401 @param lu_result: previous Exec result
2402 @return: the new Exec result, based on the previous result
2406 # We only really run POST phase hooks, and are only interested in
2408 if phase == constants.HOOKS_PHASE_POST:
2409 # Used to change hooks' output to proper indentation
2410 feedback_fn("* Hooks Results")
2411 assert hooks_results, "invalid result from hooks"
2413 for node_name in hooks_results:
2414 res = hooks_results[node_name]
2416 test = msg and not res.offline
2417 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2418 "Communication failure in hooks execution: %s", msg)
2419 if res.offline or msg:
2420 # No need to investigate payload if node is offline or gave an error.
2421 # override manually lu_result here as _ErrorIf only
2422 # overrides self.bad
2425 for script, hkr, output in res.payload:
2426 test = hkr == constants.HKR_FAIL
2427 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2428 "Script %s failed, output:", script)
2430 output = self._HOOKS_INDENT_RE.sub(' ', output)
2431 feedback_fn("%s" % output)
2437 class LUClusterVerifyDisks(NoHooksLU):
2438 """Verifies the cluster disks status.
2443 def ExpandNames(self):
2444 self.needed_locks = {
2445 locking.LEVEL_NODE: locking.ALL_SET,
2446 locking.LEVEL_INSTANCE: locking.ALL_SET,
2448 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2450 def Exec(self, feedback_fn):
2451 """Verify integrity of cluster disks.
2453 @rtype: tuple of three items
2454 @return: a tuple of (dict of node-to-node_error, list of instances
2455 which need activate-disks, dict of instance: (node, volume) for
2459 result = res_nodes, res_instances, res_missing = {}, [], {}
2461 nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2462 instances = self.cfg.GetAllInstancesInfo().values()
2465 for inst in instances:
2467 if not inst.admin_up:
2469 inst.MapLVsByNode(inst_lvs)
2470 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2471 for node, vol_list in inst_lvs.iteritems():
2472 for vol in vol_list:
2473 nv_dict[(node, vol)] = inst
2478 node_lvs = self.rpc.call_lv_list(nodes, [])
2479 for node, node_res in node_lvs.items():
2480 if node_res.offline:
2482 msg = node_res.fail_msg
2484 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2485 res_nodes[node] = msg
2488 lvs = node_res.payload
2489 for lv_name, (_, _, lv_online) in lvs.items():
2490 inst = nv_dict.pop((node, lv_name), None)
2491 if (not lv_online and inst is not None
2492 and inst.name not in res_instances):
2493 res_instances.append(inst.name)
2495 # any leftover items in nv_dict are missing LVs, let's arrange the
2497 for key, inst in nv_dict.iteritems():
2498 if inst.name not in res_missing:
2499 res_missing[inst.name] = []
2500 res_missing[inst.name].append(key)
2505 class LUClusterRepairDiskSizes(NoHooksLU):
2506 """Verifies the cluster disks sizes.
2511 def ExpandNames(self):
2512 if self.op.instances:
2513 self.wanted_names = []
2514 for name in self.op.instances:
2515 full_name = _ExpandInstanceName(self.cfg, name)
2516 self.wanted_names.append(full_name)
2517 self.needed_locks = {
2518 locking.LEVEL_NODE: [],
2519 locking.LEVEL_INSTANCE: self.wanted_names,
2521 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2523 self.wanted_names = None
2524 self.needed_locks = {
2525 locking.LEVEL_NODE: locking.ALL_SET,
2526 locking.LEVEL_INSTANCE: locking.ALL_SET,
2528 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2530 def DeclareLocks(self, level):
2531 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2532 self._LockInstancesNodes(primary_only=True)
2534 def CheckPrereq(self):
2535 """Check prerequisites.
2537 This only checks the optional instance list against the existing names.
2540 if self.wanted_names is None:
2541 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2543 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2544 in self.wanted_names]
2546 def _EnsureChildSizes(self, disk):
2547 """Ensure children of the disk have the needed disk size.
2549 This is valid mainly for DRBD8 and fixes an issue where the
2550 children have smaller disk size.
2552 @param disk: an L{ganeti.objects.Disk} object
2555 if disk.dev_type == constants.LD_DRBD8:
2556 assert disk.children, "Empty children for DRBD8?"
2557 fchild = disk.children[0]
2558 mismatch = fchild.size < disk.size
2560 self.LogInfo("Child disk has size %d, parent %d, fixing",
2561 fchild.size, disk.size)
2562 fchild.size = disk.size
2564 # and we recurse on this child only, not on the metadev
2565 return self._EnsureChildSizes(fchild) or mismatch
2569 def Exec(self, feedback_fn):
2570 """Verify the size of cluster disks.
2573 # TODO: check child disks too
2574 # TODO: check differences in size between primary/secondary nodes
2576 for instance in self.wanted_instances:
2577 pnode = instance.primary_node
2578 if pnode not in per_node_disks:
2579 per_node_disks[pnode] = []
2580 for idx, disk in enumerate(instance.disks):
2581 per_node_disks[pnode].append((instance, idx, disk))
2584 for node, dskl in per_node_disks.items():
2585 newl = [v[2].Copy() for v in dskl]
2587 self.cfg.SetDiskID(dsk, node)
2588 result = self.rpc.call_blockdev_getsize(node, newl)
2590 self.LogWarning("Failure in blockdev_getsize call to node"
2591 " %s, ignoring", node)
2593 if len(result.payload) != len(dskl):
2594 logging.warning("Invalid result from node %s: len(dksl)=%d,"
2595 " result.payload=%s", node, len(dskl), result.payload)
2596 self.LogWarning("Invalid result from node %s, ignoring node results",
2599 for ((instance, idx, disk), size) in zip(dskl, result.payload):
2601 self.LogWarning("Disk %d of instance %s did not return size"
2602 " information, ignoring", idx, instance.name)
2604 if not isinstance(size, (int, long)):
2605 self.LogWarning("Disk %d of instance %s did not return valid"
2606 " size information, ignoring", idx, instance.name)
2609 if size != disk.size:
2610 self.LogInfo("Disk %d of instance %s has mismatched size,"
2611 " correcting: recorded %d, actual %d", idx,
2612 instance.name, disk.size, size)
2614 self.cfg.Update(instance, feedback_fn)
2615 changed.append((instance.name, idx, size))
2616 if self._EnsureChildSizes(disk):
2617 self.cfg.Update(instance, feedback_fn)
2618 changed.append((instance.name, idx, disk.size))
2622 class LUClusterRename(LogicalUnit):
2623 """Rename the cluster.
2626 HPATH = "cluster-rename"
2627 HTYPE = constants.HTYPE_CLUSTER
2629 def BuildHooksEnv(self):
2634 "OP_TARGET": self.cfg.GetClusterName(),
2635 "NEW_NAME": self.op.name,
2637 mn = self.cfg.GetMasterNode()
2638 all_nodes = self.cfg.GetNodeList()
2639 return env, [mn], all_nodes
2641 def CheckPrereq(self):
2642 """Verify that the passed name is a valid one.
2645 hostname = netutils.GetHostname(name=self.op.name,
2646 family=self.cfg.GetPrimaryIPFamily())
2648 new_name = hostname.name
2649 self.ip = new_ip = hostname.ip
2650 old_name = self.cfg.GetClusterName()
2651 old_ip = self.cfg.GetMasterIP()
2652 if new_name == old_name and new_ip == old_ip:
2653 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2654 " cluster has changed",
2656 if new_ip != old_ip:
2657 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2658 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2659 " reachable on the network" %
2660 new_ip, errors.ECODE_NOTUNIQUE)
2662 self.op.name = new_name
2664 def Exec(self, feedback_fn):
2665 """Rename the cluster.
2668 clustername = self.op.name
2671 # shutdown the master IP
2672 master = self.cfg.GetMasterNode()
2673 result = self.rpc.call_node_stop_master(master, False)
2674 result.Raise("Could not disable the master role")
2677 cluster = self.cfg.GetClusterInfo()
2678 cluster.cluster_name = clustername
2679 cluster.master_ip = ip
2680 self.cfg.Update(cluster, feedback_fn)
2682 # update the known hosts file
2683 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2684 node_list = self.cfg.GetOnlineNodeList()
2686 node_list.remove(master)
2689 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2691 result = self.rpc.call_node_start_master(master, False, False)
2692 msg = result.fail_msg
2694 self.LogWarning("Could not re-enable the master role on"
2695 " the master, please restart manually: %s", msg)
2700 class LUClusterSetParams(LogicalUnit):
2701 """Change the parameters of the cluster.
2704 HPATH = "cluster-modify"
2705 HTYPE = constants.HTYPE_CLUSTER
2708 def CheckArguments(self):
2712 if self.op.uid_pool:
2713 uidpool.CheckUidPool(self.op.uid_pool)
2715 if self.op.add_uids:
2716 uidpool.CheckUidPool(self.op.add_uids)
2718 if self.op.remove_uids:
2719 uidpool.CheckUidPool(self.op.remove_uids)
2721 def ExpandNames(self):
2722 # FIXME: in the future maybe other cluster params won't require checking on
2723 # all nodes to be modified.
2724 self.needed_locks = {
2725 locking.LEVEL_NODE: locking.ALL_SET,
2727 self.share_locks[locking.LEVEL_NODE] = 1
2729 def BuildHooksEnv(self):
2734 "OP_TARGET": self.cfg.GetClusterName(),
2735 "NEW_VG_NAME": self.op.vg_name,
2737 mn = self.cfg.GetMasterNode()
2738 return env, [mn], [mn]
2740 def CheckPrereq(self):
2741 """Check prerequisites.
2743 This checks whether the given params don't conflict and
2744 if the given volume group is valid.
2747 if self.op.vg_name is not None and not self.op.vg_name:
2748 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2749 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2750 " instances exist", errors.ECODE_INVAL)
2752 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2753 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2754 raise errors.OpPrereqError("Cannot disable drbd helper while"
2755 " drbd-based instances exist",
2758 node_list = self.acquired_locks[locking.LEVEL_NODE]
2760 # if vg_name not None, checks given volume group on all nodes
2762 vglist = self.rpc.call_vg_list(node_list)
2763 for node in node_list:
2764 msg = vglist[node].fail_msg
2766 # ignoring down node
2767 self.LogWarning("Error while gathering data on node %s"
2768 " (ignoring node): %s", node, msg)
2770 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2772 constants.MIN_VG_SIZE)
2774 raise errors.OpPrereqError("Error on node '%s': %s" %
2775 (node, vgstatus), errors.ECODE_ENVIRON)
2777 if self.op.drbd_helper:
2778 # checks given drbd helper on all nodes
2779 helpers = self.rpc.call_drbd_helper(node_list)
2780 for node in node_list:
2781 ninfo = self.cfg.GetNodeInfo(node)
2783 self.LogInfo("Not checking drbd helper on offline node %s", node)
2785 msg = helpers[node].fail_msg
2787 raise errors.OpPrereqError("Error checking drbd helper on node"
2788 " '%s': %s" % (node, msg),
2789 errors.ECODE_ENVIRON)
2790 node_helper = helpers[node].payload
2791 if node_helper != self.op.drbd_helper:
2792 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2793 (node, node_helper), errors.ECODE_ENVIRON)
2795 self.cluster = cluster = self.cfg.GetClusterInfo()
2796 # validate params changes
2797 if self.op.beparams:
2798 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2799 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2801 if self.op.ndparams:
2802 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2803 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2805 if self.op.nicparams:
2806 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2807 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2808 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2811 # check all instances for consistency
2812 for instance in self.cfg.GetAllInstancesInfo().values():
2813 for nic_idx, nic in enumerate(instance.nics):
2814 params_copy = copy.deepcopy(nic.nicparams)
2815 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2817 # check parameter syntax
2819 objects.NIC.CheckParameterSyntax(params_filled)
2820 except errors.ConfigurationError, err:
2821 nic_errors.append("Instance %s, nic/%d: %s" %
2822 (instance.name, nic_idx, err))
2824 # if we're moving instances to routed, check that they have an ip
2825 target_mode = params_filled[constants.NIC_MODE]
2826 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2827 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2828 (instance.name, nic_idx))
2830 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2831 "\n".join(nic_errors))
2833 # hypervisor list/parameters
2834 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2835 if self.op.hvparams:
2836 for hv_name, hv_dict in self.op.hvparams.items():
2837 if hv_name not in self.new_hvparams:
2838 self.new_hvparams[hv_name] = hv_dict
2840 self.new_hvparams[hv_name].update(hv_dict)
2842 # os hypervisor parameters
2843 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2845 for os_name, hvs in self.op.os_hvp.items():
2846 if os_name not in self.new_os_hvp:
2847 self.new_os_hvp[os_name] = hvs
2849 for hv_name, hv_dict in hvs.items():
2850 if hv_name not in self.new_os_hvp[os_name]:
2851 self.new_os_hvp[os_name][hv_name] = hv_dict
2853 self.new_os_hvp[os_name][hv_name].update(hv_dict)
2856 self.new_osp = objects.FillDict(cluster.osparams, {})
2857 if self.op.osparams:
2858 for os_name, osp in self.op.osparams.items():
2859 if os_name not in self.new_osp:
2860 self.new_osp[os_name] = {}
2862 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2865 if not self.new_osp[os_name]:
2866 # we removed all parameters
2867 del self.new_osp[os_name]
2869 # check the parameter validity (remote check)
2870 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2871 os_name, self.new_osp[os_name])
2873 # changes to the hypervisor list
2874 if self.op.enabled_hypervisors is not None:
2875 self.hv_list = self.op.enabled_hypervisors
2876 for hv in self.hv_list:
2877 # if the hypervisor doesn't already exist in the cluster
2878 # hvparams, we initialize it to empty, and then (in both
2879 # cases) we make sure to fill the defaults, as we might not
2880 # have a complete defaults list if the hypervisor wasn't
2882 if hv not in new_hvp:
2884 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2885 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2887 self.hv_list = cluster.enabled_hypervisors
2889 if self.op.hvparams or self.op.enabled_hypervisors is not None:
2890 # either the enabled list has changed, or the parameters have, validate
2891 for hv_name, hv_params in self.new_hvparams.items():
2892 if ((self.op.hvparams and hv_name in self.op.hvparams) or
2893 (self.op.enabled_hypervisors and
2894 hv_name in self.op.enabled_hypervisors)):
2895 # either this is a new hypervisor, or its parameters have changed
2896 hv_class = hypervisor.GetHypervisor(hv_name)
2897 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2898 hv_class.CheckParameterSyntax(hv_params)
2899 _CheckHVParams(self, node_list, hv_name, hv_params)
2902 # no need to check any newly-enabled hypervisors, since the
2903 # defaults have already been checked in the above code-block
2904 for os_name, os_hvp in self.new_os_hvp.items():
2905 for hv_name, hv_params in os_hvp.items():
2906 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2907 # we need to fill in the new os_hvp on top of the actual hv_p
2908 cluster_defaults = self.new_hvparams.get(hv_name, {})
2909 new_osp = objects.FillDict(cluster_defaults, hv_params)
2910 hv_class = hypervisor.GetHypervisor(hv_name)
2911 hv_class.CheckParameterSyntax(new_osp)
2912 _CheckHVParams(self, node_list, hv_name, new_osp)
2914 if self.op.default_iallocator:
2915 alloc_script = utils.FindFile(self.op.default_iallocator,
2916 constants.IALLOCATOR_SEARCH_PATH,
2918 if alloc_script is None:
2919 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2920 " specified" % self.op.default_iallocator,
2923 def Exec(self, feedback_fn):
2924 """Change the parameters of the cluster.
2927 if self.op.vg_name is not None:
2928 new_volume = self.op.vg_name
2931 if new_volume != self.cfg.GetVGName():
2932 self.cfg.SetVGName(new_volume)
2934 feedback_fn("Cluster LVM configuration already in desired"
2935 " state, not changing")
2936 if self.op.drbd_helper is not None:
2937 new_helper = self.op.drbd_helper
2940 if new_helper != self.cfg.GetDRBDHelper():
2941 self.cfg.SetDRBDHelper(new_helper)
2943 feedback_fn("Cluster DRBD helper already in desired state,"
2945 if self.op.hvparams:
2946 self.cluster.hvparams = self.new_hvparams
2948 self.cluster.os_hvp = self.new_os_hvp
2949 if self.op.enabled_hypervisors is not None:
2950 self.cluster.hvparams = self.new_hvparams
2951 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2952 if self.op.beparams:
2953 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2954 if self.op.nicparams:
2955 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2956 if self.op.osparams:
2957 self.cluster.osparams = self.new_osp
2958 if self.op.ndparams:
2959 self.cluster.ndparams = self.new_ndparams
2961 if self.op.candidate_pool_size is not None:
2962 self.cluster.candidate_pool_size = self.op.candidate_pool_size
2963 # we need to update the pool size here, otherwise the save will fail
2964 _AdjustCandidatePool(self, [])
2966 if self.op.maintain_node_health is not None:
2967 self.cluster.maintain_node_health = self.op.maintain_node_health
2969 if self.op.prealloc_wipe_disks is not None:
2970 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2972 if self.op.add_uids is not None:
2973 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2975 if self.op.remove_uids is not None:
2976 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2978 if self.op.uid_pool is not None:
2979 self.cluster.uid_pool = self.op.uid_pool
2981 if self.op.default_iallocator is not None:
2982 self.cluster.default_iallocator = self.op.default_iallocator
2984 if self.op.reserved_lvs is not None:
2985 self.cluster.reserved_lvs = self.op.reserved_lvs
2987 def helper_os(aname, mods, desc):
2989 lst = getattr(self.cluster, aname)
2990 for key, val in mods:
2991 if key == constants.DDM_ADD:
2993 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
2996 elif key == constants.DDM_REMOVE:
3000 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3002 raise errors.ProgrammerError("Invalid modification '%s'" % key)
3004 if self.op.hidden_os:
3005 helper_os("hidden_os", self.op.hidden_os, "hidden")
3007 if self.op.blacklisted_os:
3008 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3010 if self.op.master_netdev:
3011 master = self.cfg.GetMasterNode()
3012 feedback_fn("Shutting down master ip on the current netdev (%s)" %
3013 self.cluster.master_netdev)
3014 result = self.rpc.call_node_stop_master(master, False)
3015 result.Raise("Could not disable the master ip")
3016 feedback_fn("Changing master_netdev from %s to %s" %
3017 (self.cluster.master_netdev, self.op.master_netdev))
3018 self.cluster.master_netdev = self.op.master_netdev
3020 self.cfg.Update(self.cluster, feedback_fn)
3022 if self.op.master_netdev:
3023 feedback_fn("Starting the master ip on the new master netdev (%s)" %
3024 self.op.master_netdev)
3025 result = self.rpc.call_node_start_master(master, False, False)
3027 self.LogWarning("Could not re-enable the master ip on"
3028 " the master, please restart manually: %s",
3032 def _UploadHelper(lu, nodes, fname):
3033 """Helper for uploading a file and showing warnings.
3036 if os.path.exists(fname):
3037 result = lu.rpc.call_upload_file(nodes, fname)
3038 for to_node, to_result in result.items():
3039 msg = to_result.fail_msg
3041 msg = ("Copy of file %s to node %s failed: %s" %
3042 (fname, to_node, msg))
3043 lu.proc.LogWarning(msg)
3046 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3047 """Distribute additional files which are part of the cluster configuration.
3049 ConfigWriter takes care of distributing the config and ssconf files, but
3050 there are more files which should be distributed to all nodes. This function
3051 makes sure those are copied.
3053 @param lu: calling logical unit
3054 @param additional_nodes: list of nodes not in the config to distribute to
3055 @type additional_vm: boolean
3056 @param additional_vm: whether the additional nodes are vm-capable or not
3059 # 1. Gather target nodes
3060 myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3061 dist_nodes = lu.cfg.GetOnlineNodeList()
3062 nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3063 vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3064 if additional_nodes is not None:
3065 dist_nodes.extend(additional_nodes)
3067 vm_nodes.extend(additional_nodes)
3068 if myself.name in dist_nodes:
3069 dist_nodes.remove(myself.name)
3070 if myself.name in vm_nodes:
3071 vm_nodes.remove(myself.name)
3073 # 2. Gather files to distribute
3074 dist_files = set([constants.ETC_HOSTS,
3075 constants.SSH_KNOWN_HOSTS_FILE,
3076 constants.RAPI_CERT_FILE,
3077 constants.RAPI_USERS_FILE,
3078 constants.CONFD_HMAC_KEY,
3079 constants.CLUSTER_DOMAIN_SECRET_FILE,
3083 enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3084 for hv_name in enabled_hypervisors:
3085 hv_class = hypervisor.GetHypervisor(hv_name)
3086 vm_files.update(hv_class.GetAncillaryFiles())
3088 # 3. Perform the files upload
3089 for fname in dist_files:
3090 _UploadHelper(lu, dist_nodes, fname)
3091 for fname in vm_files:
3092 _UploadHelper(lu, vm_nodes, fname)
3095 class LUClusterRedistConf(NoHooksLU):
3096 """Force the redistribution of cluster configuration.
3098 This is a very simple LU.
3103 def ExpandNames(self):
3104 self.needed_locks = {
3105 locking.LEVEL_NODE: locking.ALL_SET,
3107 self.share_locks[locking.LEVEL_NODE] = 1
3109 def Exec(self, feedback_fn):
3110 """Redistribute the configuration.
3113 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3114 _RedistributeAncillaryFiles(self)
3117 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3118 """Sleep and poll for an instance's disk to sync.
3121 if not instance.disks or disks is not None and not disks:
3124 disks = _ExpandCheckDisks(instance, disks)
3127 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3129 node = instance.primary_node
3132 lu.cfg.SetDiskID(dev, node)
3134 # TODO: Convert to utils.Retry
3137 degr_retries = 10 # in seconds, as we sleep 1 second each time
3141 cumul_degraded = False
3142 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3143 msg = rstats.fail_msg
3145 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3148 raise errors.RemoteError("Can't contact node %s for mirror data,"
3149 " aborting." % node)
3152 rstats = rstats.payload
3154 for i, mstat in enumerate(rstats):
3156 lu.LogWarning("Can't compute data for node %s/%s",
3157 node, disks[i].iv_name)
3160 cumul_degraded = (cumul_degraded or
3161 (mstat.is_degraded and mstat.sync_percent is None))
3162 if mstat.sync_percent is not None:
3164 if mstat.estimated_time is not None:
3165 rem_time = ("%s remaining (estimated)" %
3166 utils.FormatSeconds(mstat.estimated_time))
3167 max_time = mstat.estimated_time
3169 rem_time = "no time estimate"
3170 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3171 (disks[i].iv_name, mstat.sync_percent, rem_time))
3173 # if we're done but degraded, let's do a few small retries, to
3174 # make sure we see a stable and not transient situation; therefore
3175 # we force restart of the loop
3176 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3177 logging.info("Degraded disks found, %d retries left", degr_retries)
3185 time.sleep(min(60, max_time))
3188 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3189 return not cumul_degraded
3192 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3193 """Check that mirrors are not degraded.
3195 The ldisk parameter, if True, will change the test from the
3196 is_degraded attribute (which represents overall non-ok status for
3197 the device(s)) to the ldisk (representing the local storage status).
3200 lu.cfg.SetDiskID(dev, node)
3204 if on_primary or dev.AssembleOnSecondary():
3205 rstats = lu.rpc.call_blockdev_find(node, dev)
3206 msg = rstats.fail_msg
3208 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3210 elif not rstats.payload:
3211 lu.LogWarning("Can't find disk on node %s", node)
3215 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3217 result = result and not rstats.payload.is_degraded
3220 for child in dev.children:
3221 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3226 class LUOobCommand(NoHooksLU):
3227 """Logical unit for OOB handling.
3231 _SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE)
3233 def CheckPrereq(self):
3234 """Check prerequisites.
3237 - the node exists in the configuration
3240 Any errors are signaled by raising errors.OpPrereqError.
3244 self.master_node = self.cfg.GetMasterNode()
3246 if self.op.node_names:
3247 if self.op.command in self._SKIP_MASTER:
3248 if self.master_node in self.op.node_names:
3249 master_node_obj = self.cfg.GetNodeInfo(self.master_node)
3250 master_oob_handler = _SupportsOob(self.cfg, master_node_obj)
3252 if master_oob_handler:
3253 additional_text = ("Run '%s %s %s' if you want to operate on the"
3254 " master regardless") % (master_oob_handler,
3258 additional_text = "The master node does not support out-of-band"
3260 raise errors.OpPrereqError(("Operating on the master node %s is not"
3261 " allowed for %s\n%s") %
3262 (self.master_node, self.op.command,
3263 additional_text), errors.ECODE_INVAL)
3265 self.op.node_names = self.cfg.GetNodeList()
3266 if self.op.command in self._SKIP_MASTER:
3267 self.op.node_names.remove(self.master_node)
3269 if self.op.command in self._SKIP_MASTER:
3270 assert self.master_node not in self.op.node_names
3272 for node_name in self.op.node_names:
3273 node = self.cfg.GetNodeInfo(node_name)
3276 raise errors.OpPrereqError("Node %s not found" % node_name,
3279 self.nodes.append(node)
3281 if (not self.op.ignore_status and
3282 (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
3283 raise errors.OpPrereqError(("Cannot power off node %s because it is"
3284 " not marked offline") % node_name,
3287 def ExpandNames(self):
3288 """Gather locks we need.
3291 if self.op.node_names:
3292 self.op.node_names = [_ExpandNodeName(self.cfg, name)
3293 for name in self.op.node_names]
3294 lock_names = self.op.node_names
3296 lock_names = locking.ALL_SET
3298 self.needed_locks = {
3299 locking.LEVEL_NODE: lock_names,
3302 def Exec(self, feedback_fn):
3303 """Execute OOB and return result if we expect any.
3306 master_node = self.master_node
3309 for node in self.nodes:
3310 node_entry = [(constants.RS_NORMAL, node.name)]
3311 ret.append(node_entry)
3313 oob_program = _SupportsOob(self.cfg, node)
3316 node_entry.append((constants.RS_UNAVAIL, None))
3319 logging.info("Executing out-of-band command '%s' using '%s' on %s",
3320 self.op.command, oob_program, node.name)
3321 result = self.rpc.call_run_oob(master_node, oob_program,
3322 self.op.command, node.name,
3326 self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3327 node.name, result.fail_msg)
3328 node_entry.append((constants.RS_NODATA, None))
3331 self._CheckPayload(result)
3332 except errors.OpExecError, err:
3333 self.LogWarning("The payload returned by '%s' is not valid: %s",
3335 node_entry.append((constants.RS_NODATA, None))
3337 if self.op.command == constants.OOB_HEALTH:
3338 # For health we should log important events
3339 for item, status in result.payload:
3340 if status in [constants.OOB_STATUS_WARNING,
3341 constants.OOB_STATUS_CRITICAL]:
3342 self.LogWarning("On node '%s' item '%s' has status '%s'",
3343 node.name, item, status)
3345 if self.op.command == constants.OOB_POWER_ON:
3347 elif self.op.command == constants.OOB_POWER_OFF:
3348 node.powered = False
3349 elif self.op.command == constants.OOB_POWER_STATUS:
3350 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3351 if powered != node.powered:
3352 logging.warning(("Recorded power state (%s) of node '%s' does not"
3353 " match actual power state (%s)"), node.powered,
3356 # For configuration changing commands we should update the node
3357 if self.op.command in (constants.OOB_POWER_ON,
3358 constants.OOB_POWER_OFF):
3359 self.cfg.Update(node, feedback_fn)
3361 node_entry.append((constants.RS_NORMAL, result.payload))
3365 def _CheckPayload(self, result):
3366 """Checks if the payload is valid.
3368 @param result: RPC result
3369 @raises errors.OpExecError: If payload is not valid
3373 if self.op.command == constants.OOB_HEALTH:
3374 if not isinstance(result.payload, list):
3375 errs.append("command 'health' is expected to return a list but got %s" %
3376 type(result.payload))
3378 for item, status in result.payload:
3379 if status not in constants.OOB_STATUSES:
3380 errs.append("health item '%s' has invalid status '%s'" %
3383 if self.op.command == constants.OOB_POWER_STATUS:
3384 if not isinstance(result.payload, dict):
3385 errs.append("power-status is expected to return a dict but got %s" %
3386 type(result.payload))
3388 if self.op.command in [
3389 constants.OOB_POWER_ON,
3390 constants.OOB_POWER_OFF,
3391 constants.OOB_POWER_CYCLE,
3393 if result.payload is not None:
3394 errs.append("%s is expected to not return payload but got '%s'" %
3395 (self.op.command, result.payload))
3398 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3399 utils.CommaJoin(errs))
3403 class LUOsDiagnose(NoHooksLU):
3404 """Logical unit for OS diagnose/query.
3409 _BLK = "blacklisted"
3411 _FIELDS_STATIC = utils.FieldSet()
3412 _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3413 "parameters", "api_versions", _HID, _BLK)
3415 def CheckArguments(self):
3417 raise errors.OpPrereqError("Selective OS query not supported",
3420 _CheckOutputFields(static=self._FIELDS_STATIC,
3421 dynamic=self._FIELDS_DYNAMIC,
3422 selected=self.op.output_fields)
3424 def ExpandNames(self):
3425 # Lock all nodes, in shared mode
3426 # Temporary removal of locks, should be reverted later
3427 # TODO: reintroduce locks when they are lighter-weight
3428 self.needed_locks = {}
3429 #self.share_locks[locking.LEVEL_NODE] = 1
3430 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3433 def _DiagnoseByOS(rlist):
3434 """Remaps a per-node return list into an a per-os per-node dictionary
3436 @param rlist: a map with node names as keys and OS objects as values
3439 @return: a dictionary with osnames as keys and as value another
3440 map, with nodes as keys and tuples of (path, status, diagnose,
3441 variants, parameters, api_versions) as values, eg::
3443 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3444 (/srv/..., False, "invalid api")],
3445 "node2": [(/srv/..., True, "", [], [])]}
3450 # we build here the list of nodes that didn't fail the RPC (at RPC
3451 # level), so that nodes with a non-responding node daemon don't
3452 # make all OSes invalid
3453 good_nodes = [node_name for node_name in rlist
3454 if not rlist[node_name].fail_msg]
3455 for node_name, nr in rlist.items():
3456 if nr.fail_msg or not nr.payload:
3458 for (name, path, status, diagnose, variants,
3459 params, api_versions) in nr.payload:
3460 if name not in all_os:
3461 # build a list of nodes for this os containing empty lists
3462 # for each node in node_list
3464 for nname in good_nodes:
3465 all_os[name][nname] = []
3466 # convert params from [name, help] to (name, help)
3467 params = [tuple(v) for v in params]
3468 all_os[name][node_name].append((path, status, diagnose,
3469 variants, params, api_versions))
3472 def Exec(self, feedback_fn):
3473 """Compute the list of OSes.
3476 valid_nodes = [node.name
3477 for node in self.cfg.GetAllNodesInfo().values()
3478 if not node.offline and node.vm_capable]
3479 node_data = self.rpc.call_os_diagnose(valid_nodes)
3480 pol = self._DiagnoseByOS(node_data)
3482 cluster = self.cfg.GetClusterInfo()
3484 for os_name in utils.NiceSort(pol.keys()):
3485 os_data = pol[os_name]
3488 (variants, params, api_versions) = null_state = (set(), set(), set())
3489 for idx, osl in enumerate(os_data.values()):
3490 valid = bool(valid and osl and osl[0][1])
3492 (variants, params, api_versions) = null_state
3494 node_variants, node_params, node_api = osl[0][3:6]
3495 if idx == 0: # first entry
3496 variants = set(node_variants)
3497 params = set(node_params)
3498 api_versions = set(node_api)
3499 else: # keep consistency
3500 variants.intersection_update(node_variants)
3501 params.intersection_update(node_params)
3502 api_versions.intersection_update(node_api)
3504 is_hid = os_name in cluster.hidden_os
3505 is_blk = os_name in cluster.blacklisted_os
3506 if ((self._HID not in self.op.output_fields and is_hid) or
3507 (self._BLK not in self.op.output_fields and is_blk) or
3508 (self._VLD not in self.op.output_fields and not valid)):
3511 for field in self.op.output_fields:
3514 elif field == self._VLD:
3516 elif field == "node_status":
3517 # this is just a copy of the dict
3519 for node_name, nos_list in os_data.items():
3520 val[node_name] = nos_list
3521 elif field == "variants":
3522 val = utils.NiceSort(list(variants))
3523 elif field == "parameters":
3525 elif field == "api_versions":
3526 val = list(api_versions)
3527 elif field == self._HID:
3529 elif field == self._BLK:
3532 raise errors.ParameterError(field)
3539 class LUNodeRemove(LogicalUnit):
3540 """Logical unit for removing a node.
3543 HPATH = "node-remove"
3544 HTYPE = constants.HTYPE_NODE
3546 def BuildHooksEnv(self):
3549 This doesn't run on the target node in the pre phase as a failed
3550 node would then be impossible to remove.
3554 "OP_TARGET": self.op.node_name,
3555 "NODE_NAME": self.op.node_name,
3557 all_nodes = self.cfg.GetNodeList()
3559 all_nodes.remove(self.op.node_name)
3561 logging.warning("Node %s which is about to be removed not found"
3562 " in the all nodes list", self.op.node_name)
3563 return env, all_nodes, all_nodes
3565 def CheckPrereq(self):
3566 """Check prerequisites.
3569 - the node exists in the configuration
3570 - it does not have primary or secondary instances
3571 - it's not the master
3573 Any errors are signaled by raising errors.OpPrereqError.
3576 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3577 node = self.cfg.GetNodeInfo(self.op.node_name)
3578 assert node is not None
3580 instance_list = self.cfg.GetInstanceList()
3582 masternode = self.cfg.GetMasterNode()
3583 if node.name == masternode:
3584 raise errors.OpPrereqError("Node is the master node,"
3585 " you need to failover first.",
3588 for instance_name in instance_list:
3589 instance = self.cfg.GetInstanceInfo(instance_name)
3590 if node.name in instance.all_nodes:
3591 raise errors.OpPrereqError("Instance %s is still running on the node,"
3592 " please remove first." % instance_name,
3594 self.op.node_name = node.name
3597 def Exec(self, feedback_fn):
3598 """Removes the node from the cluster.
3602 logging.info("Stopping the node daemon and removing configs from node %s",
3605 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3607 # Promote nodes to master candidate as needed
3608 _AdjustCandidatePool(self, exceptions=[node.name])
3609 self.context.RemoveNode(node.name)
3611 # Run post hooks on the node before it's removed
3612 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3614 hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3616 # pylint: disable-msg=W0702
3617 self.LogWarning("Errors occurred running hooks on %s" % node.name)
3619 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3620 msg = result.fail_msg
3622 self.LogWarning("Errors encountered on the remote node while leaving"
3623 " the cluster: %s", msg)
3625 # Remove node from our /etc/hosts
3626 if self.cfg.GetClusterInfo().modify_etc_hosts:
3627 master_node = self.cfg.GetMasterNode()
3628 result = self.rpc.call_etc_hosts_modify(master_node,
3629 constants.ETC_HOSTS_REMOVE,
3631 result.Raise("Can't update hosts file with new host data")
3632 _RedistributeAncillaryFiles(self)
3635 class _NodeQuery(_QueryBase):
3636 FIELDS = query.NODE_FIELDS
3638 def ExpandNames(self, lu):
3639 lu.needed_locks = {}
3640 lu.share_locks[locking.LEVEL_NODE] = 1
3643 self.wanted = _GetWantedNodes(lu, self.names)
3645 self.wanted = locking.ALL_SET
3647 self.do_locking = (self.use_locking and
3648 query.NQ_LIVE in self.requested_data)
3651 # if we don't request only static fields, we need to lock the nodes
3652 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3654 def DeclareLocks(self, lu, level):
3657 def _GetQueryData(self, lu):
3658 """Computes the list of nodes and their attributes.
3661 all_info = lu.cfg.GetAllNodesInfo()
3663 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3665 # Gather data as requested
3666 if query.NQ_LIVE in self.requested_data:
3667 # filter out non-vm_capable nodes
3668 toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
3670 node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
3671 lu.cfg.GetHypervisorType())
3672 live_data = dict((name, nresult.payload)
3673 for (name, nresult) in node_data.items()
3674 if not nresult.fail_msg and nresult.payload)
3678 if query.NQ_INST in self.requested_data:
3679 node_to_primary = dict([(name, set()) for name in nodenames])
3680 node_to_secondary = dict([(name, set()) for name in nodenames])
3682 inst_data = lu.cfg.GetAllInstancesInfo()
3684 for inst in inst_data.values():
3685 if inst.primary_node in node_to_primary:
3686 node_to_primary[inst.primary_node].add(inst.name)
3687 for secnode in inst.secondary_nodes:
3688 if secnode in node_to_secondary:
3689 node_to_secondary[secnode].add(inst.name)
3691 node_to_primary = None
3692 node_to_secondary = None
3694 if query.NQ_OOB in self.requested_data:
3695 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3696 for name, node in all_info.iteritems())
3700 if query.NQ_GROUP in self.requested_data:
3701 groups = lu.cfg.GetAllNodeGroupsInfo()
3705 return query.NodeQueryData([all_info[name] for name in nodenames],
3706 live_data, lu.cfg.GetMasterNode(),
3707 node_to_primary, node_to_secondary, groups,
3708 oob_support, lu.cfg.GetClusterInfo())
3711 class LUNodeQuery(NoHooksLU):
3712 """Logical unit for querying nodes.
3715 # pylint: disable-msg=W0142
3718 def CheckArguments(self):
3719 self.nq = _NodeQuery(qlang.MakeSimpleFilter("name", self.op.names),
3720 self.op.output_fields, self.op.use_locking)
3722 def ExpandNames(self):
3723 self.nq.ExpandNames(self)
3725 def Exec(self, feedback_fn):
3726 return self.nq.OldStyleQuery(self)
3729 class LUNodeQueryvols(NoHooksLU):
3730 """Logical unit for getting volumes on node(s).
3734 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3735 _FIELDS_STATIC = utils.FieldSet("node")
3737 def CheckArguments(self):
3738 _CheckOutputFields(static=self._FIELDS_STATIC,
3739 dynamic=self._FIELDS_DYNAMIC,
3740 selected=self.op.output_fields)
3742 def ExpandNames(self):
3743 self.needed_locks = {}
3744 self.share_locks[locking.LEVEL_NODE] = 1
3745 if not self.op.nodes:
3746 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3748 self.needed_locks[locking.LEVEL_NODE] = \
3749 _GetWantedNodes(self, self.op.nodes)
3751 def Exec(self, feedback_fn):
3752 """Computes the list of nodes and their attributes.
3755 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3756 volumes = self.rpc.call_node_volumes(nodenames)
3758 ilist = [self.cfg.GetInstanceInfo(iname) for iname
3759 in self.cfg.GetInstanceList()]
3761 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3764 for node in nodenames:
3765 nresult = volumes[node]
3768 msg = nresult.fail_msg
3770 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3773 node_vols = nresult.payload[:]
3774 node_vols.sort(key=lambda vol: vol['dev'])
3776 for vol in node_vols:
3778 for field in self.op.output_fields:
3781 elif field == "phys":
3785 elif field == "name":
3787 elif field == "size":
3788 val = int(float(vol['size']))
3789 elif field == "instance":
3791 if node not in lv_by_node[inst]:
3793 if vol['name'] in lv_by_node[inst][node]:
3799 raise errors.ParameterError(field)
3800 node_output.append(str(val))
3802 output.append(node_output)
3807 class LUNodeQueryStorage(NoHooksLU):
3808 """Logical unit for getting information on storage units on node(s).
3811 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3814 def CheckArguments(self):
3815 _CheckOutputFields(static=self._FIELDS_STATIC,
3816 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3817 selected=self.op.output_fields)
3819 def ExpandNames(self):
3820 self.needed_locks = {}
3821 self.share_locks[locking.LEVEL_NODE] = 1
3824 self.needed_locks[locking.LEVEL_NODE] = \
3825 _GetWantedNodes(self, self.op.nodes)
3827 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3829 def Exec(self, feedback_fn):
3830 """Computes the list of nodes and their attributes.
3833 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3835 # Always get name to sort by
3836 if constants.SF_NAME in self.op.output_fields:
3837 fields = self.op.output_fields[:]
3839 fields = [constants.SF_NAME] + self.op.output_fields
3841 # Never ask for node or type as it's only known to the LU
3842 for extra in [constants.SF_NODE, constants.SF_TYPE]:
3843 while extra in fields:
3844 fields.remove(extra)
3846 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3847 name_idx = field_idx[constants.SF_NAME]
3849 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3850 data = self.rpc.call_storage_list(self.nodes,
3851 self.op.storage_type, st_args,
3852 self.op.name, fields)
3856 for node in utils.NiceSort(self.nodes):
3857 nresult = data[node]
3861 msg = nresult.fail_msg
3863 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3866 rows = dict([(row[name_idx], row) for row in nresult.payload])
3868 for name in utils.NiceSort(rows.keys()):
3873 for field in self.op.output_fields:
3874 if field == constants.SF_NODE:
3876 elif field == constants.SF_TYPE:
3877 val = self.op.storage_type
3878 elif field in field_idx:
3879 val = row[field_idx[field]]
3881 raise errors.ParameterError(field)
3890 class _InstanceQuery(_QueryBase):
3891 FIELDS = query.INSTANCE_FIELDS
3893 def ExpandNames(self, lu):
3894 lu.needed_locks = {}
3895 lu.share_locks[locking.LEVEL_INSTANCE] = 1
3896 lu.share_locks[locking.LEVEL_NODE] = 1
3899 self.wanted = _GetWantedInstances(lu, self.names)
3901 self.wanted = locking.ALL_SET
3903 self.do_locking = (self.use_locking and
3904 query.IQ_LIVE in self.requested_data)
3906 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3907 lu.needed_locks[locking.LEVEL_NODE] = []
3908 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3910 def DeclareLocks(self, lu, level):
3911 if level == locking.LEVEL_NODE and self.do_locking:
3912 lu._LockInstancesNodes() # pylint: disable-msg=W0212
3914 def _GetQueryData(self, lu):
3915 """Computes the list of instances and their attributes.
3918 cluster = lu.cfg.GetClusterInfo()
3919 all_info = lu.cfg.GetAllInstancesInfo()
3921 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3923 instance_list = [all_info[name] for name in instance_names]
3924 nodes = frozenset(itertools.chain(*(inst.all_nodes
3925 for inst in instance_list)))
3926 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3929 wrongnode_inst = set()
3931 # Gather data as requested
3932 if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
3934 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3936 result = node_data[name]
3938 # offline nodes will be in both lists
3939 assert result.fail_msg
3940 offline_nodes.append(name)
3942 bad_nodes.append(name)
3943 elif result.payload:
3944 for inst in result.payload:
3945 if all_info[inst].primary_node == name:
3946 live_data.update(result.payload)
3948 wrongnode_inst.add(inst)
3949 # else no instance is alive
3953 if query.IQ_DISKUSAGE in self.requested_data:
3954 disk_usage = dict((inst.name,
3955 _ComputeDiskSize(inst.disk_template,
3956 [{"size": disk.size}
3957 for disk in inst.disks]))
3958 for inst in instance_list)
3962 if query.IQ_CONSOLE in self.requested_data:
3964 for inst in instance_list:
3965 if inst.name in live_data:
3966 # Instance is running
3967 consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
3969 consinfo[inst.name] = None
3970 assert set(consinfo.keys()) == set(instance_names)
3974 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3975 disk_usage, offline_nodes, bad_nodes,
3976 live_data, wrongnode_inst, consinfo)
3979 class LUQuery(NoHooksLU):
3980 """Query for resources/items of a certain kind.
3983 # pylint: disable-msg=W0142
3986 def CheckArguments(self):
3987 qcls = _GetQueryImplementation(self.op.what)
3989 self.impl = qcls(self.op.filter, self.op.fields, False)
3991 def ExpandNames(self):
3992 self.impl.ExpandNames(self)
3994 def DeclareLocks(self, level):
3995 self.impl.DeclareLocks(self, level)
3997 def Exec(self, feedback_fn):
3998 return self.impl.NewStyleQuery(self)
4001 class LUQueryFields(NoHooksLU):
4002 """Query for resources/items of a certain kind.
4005 # pylint: disable-msg=W0142
4008 def CheckArguments(self):
4009 self.qcls = _GetQueryImplementation(self.op.what)
4011 def ExpandNames(self):
4012 self.needed_locks = {}
4014 def Exec(self, feedback_fn):
4015 return self.qcls.FieldsQuery(self.op.fields)
4018 class LUNodeModifyStorage(NoHooksLU):
4019 """Logical unit for modifying a storage volume on a node.
4024 def CheckArguments(self):
4025 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4027 storage_type = self.op.storage_type
4030 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4032 raise errors.OpPrereqError("Storage units of type '%s' can not be"
4033 " modified" % storage_type,
4036 diff = set(self.op.changes.keys()) - modifiable
4038 raise errors.OpPrereqError("The following fields can not be modified for"
4039 " storage units of type '%s': %r" %
4040 (storage_type, list(diff)),
4043 def ExpandNames(self):
4044 self.needed_locks = {
4045 locking.LEVEL_NODE: self.op.node_name,
4048 def Exec(self, feedback_fn):
4049 """Computes the list of nodes and their attributes.
4052 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4053 result = self.rpc.call_storage_modify(self.op.node_name,
4054 self.op.storage_type, st_args,
4055 self.op.name, self.op.changes)
4056 result.Raise("Failed to modify storage unit '%s' on %s" %
4057 (self.op.name, self.op.node_name))
4060 class LUNodeAdd(LogicalUnit):
4061 """Logical unit for adding node to the cluster.
4065 HTYPE = constants.HTYPE_NODE
4066 _NFLAGS = ["master_capable", "vm_capable"]
4068 def CheckArguments(self):
4069 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4070 # validate/normalize the node name
4071 self.hostname = netutils.GetHostname(name=self.op.node_name,
4072 family=self.primary_ip_family)
4073 self.op.node_name = self.hostname.name
4074 if self.op.readd and self.op.group:
4075 raise errors.OpPrereqError("Cannot pass a node group when a node is"
4076 " being readded", errors.ECODE_INVAL)
4078 def BuildHooksEnv(self):
4081 This will run on all nodes before, and on all nodes + the new node after.
4085 "OP_TARGET": self.op.node_name,
4086 "NODE_NAME": self.op.node_name,
4087 "NODE_PIP": self.op.primary_ip,
4088 "NODE_SIP": self.op.secondary_ip,
4089 "MASTER_CAPABLE": str(self.op.master_capable),
4090 "VM_CAPABLE": str(self.op.vm_capable),
4092 nodes_0 = self.cfg.GetNodeList()
4093 nodes_1 = nodes_0 + [self.op.node_name, ]
4094 return env, nodes_0, nodes_1
4096 def CheckPrereq(self):
4097 """Check prerequisites.
4100 - the new node is not already in the config
4102 - its parameters (single/dual homed) matches the cluster
4104 Any errors are signaled by raising errors.OpPrereqError.
4108 hostname = self.hostname
4109 node = hostname.name
4110 primary_ip = self.op.primary_ip = hostname.ip
4111 if self.op.secondary_ip is None:
4112 if self.primary_ip_family == netutils.IP6Address.family:
4113 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4114 " IPv4 address must be given as secondary",
4116 self.op.secondary_ip = primary_ip
4118 secondary_ip = self.op.secondary_ip
4119 if not netutils.IP4Address.IsValid(secondary_ip):
4120 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4121 " address" % secondary_ip, errors.ECODE_INVAL)
4123 node_list = cfg.GetNodeList()
4124 if not self.op.readd and node in node_list:
4125 raise errors.OpPrereqError("Node %s is already in the configuration" %
4126 node, errors.ECODE_EXISTS)
4127 elif self.op.readd and node not in node_list:
4128 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4131 self.changed_primary_ip = False
4133 for existing_node_name in node_list:
4134 existing_node = cfg.GetNodeInfo(existing_node_name)
4136 if self.op.readd and node == existing_node_name:
4137 if existing_node.secondary_ip != secondary_ip:
4138 raise errors.OpPrereqError("Readded node doesn't have the same IP"
4139 " address configuration as before",
4141 if existing_node.primary_ip != primary_ip:
4142 self.changed_primary_ip = True
4146 if (existing_node.primary_ip == primary_ip or
4147 existing_node.secondary_ip == primary_ip or
4148 existing_node.primary_ip == secondary_ip or
4149 existing_node.secondary_ip == secondary_ip):
4150 raise errors.OpPrereqError("New node ip address(es) conflict with"
4151 " existing node %s" % existing_node.name,
4152 errors.ECODE_NOTUNIQUE)
4154 # After this 'if' block, None is no longer a valid value for the
4155 # _capable op attributes
4157 old_node = self.cfg.GetNodeInfo(node)
4158 assert old_node is not None, "Can't retrieve locked node %s" % node
4159 for attr in self._NFLAGS:
4160 if getattr(self.op, attr) is None:
4161 setattr(self.op, attr, getattr(old_node, attr))
4163 for attr in self._NFLAGS:
4164 if getattr(self.op, attr) is None:
4165 setattr(self.op, attr, True)
4167 if self.op.readd and not self.op.vm_capable:
4168 pri, sec = cfg.GetNodeInstances(node)
4170 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4171 " flag set to false, but it already holds"
4172 " instances" % node,
4175 # check that the type of the node (single versus dual homed) is the
4176 # same as for the master
4177 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4178 master_singlehomed = myself.secondary_ip == myself.primary_ip
4179 newbie_singlehomed = secondary_ip == primary_ip
4180 if master_singlehomed != newbie_singlehomed:
4181 if master_singlehomed:
4182 raise errors.OpPrereqError("The master has no secondary ip but the"
4183 " new node has one",
4186 raise errors.OpPrereqError("The master has a secondary ip but the"
4187 " new node doesn't have one",
4190 # checks reachability
4191 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
4192 raise errors.OpPrereqError("Node not reachable by ping",
4193 errors.ECODE_ENVIRON)
4195 if not newbie_singlehomed:
4196 # check reachability from my secondary ip to newbie's secondary ip
4197 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
4198 source=myself.secondary_ip):
4199 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4200 " based ping to node daemon port",
4201 errors.ECODE_ENVIRON)
4208 if self.op.master_capable:
4209 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
4211 self.master_candidate = False
4214 self.new_node = old_node
4216 node_group = cfg.LookupNodeGroup(self.op.group)
4217 self.new_node = objects.Node(name=node,
4218 primary_ip=primary_ip,
4219 secondary_ip=secondary_ip,
4220 master_candidate=self.master_candidate,
4221 offline=False, drained=False,
4224 if self.op.ndparams:
4225 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
4227 def Exec(self, feedback_fn):
4228 """Adds the new node to the cluster.
4231 new_node = self.new_node
4232 node = new_node.name
4234 # We adding a new node so we assume it's powered
4235 new_node.powered = True
4237 # for re-adds, reset the offline/drained/master-candidate flags;
4238 # we need to reset here, otherwise offline would prevent RPC calls
4239 # later in the procedure; this also means that if the re-add
4240 # fails, we are left with a non-offlined, broken node
4242 new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
4243 self.LogInfo("Readding a node, the offline/drained flags were reset")
4244 # if we demote the node, we do cleanup later in the procedure
4245 new_node.master_candidate = self.master_candidate
4246 if self.changed_primary_ip:
4247 new_node.primary_ip = self.op.primary_ip
4249 # copy the master/vm_capable flags
4250 for attr in self._NFLAGS:
4251 setattr(new_node, attr, getattr(self.op, attr))
4253 # notify the user about any possible mc promotion
4254 if new_node.master_candidate:
4255 self.LogInfo("Node will be a master candidate")
4257 if self.op.ndparams:
4258 new_node.ndparams = self.op.ndparams
4260 new_node.ndparams = {}
4262 # check connectivity
4263 result = self.rpc.call_version([node])[node]
4264 result.Raise("Can't get version information from node %s" % node)
4265 if constants.PROTOCOL_VERSION == result.payload:
4266 logging.info("Communication to node %s fine, sw version %s match",
4267 node, result.payload)
4269 raise errors.OpExecError("Version mismatch master version %s,"
4270 " node version %s" %
4271 (constants.PROTOCOL_VERSION, result.payload))
4273 # Add node to our /etc/hosts, and add key to known_hosts
4274 if self.cfg.GetClusterInfo().modify_etc_hosts:
4275 master_node = self.cfg.GetMasterNode()
4276 result = self.rpc.call_etc_hosts_modify(master_node,
4277 constants.ETC_HOSTS_ADD,
4280 result.Raise("Can't update hosts file with new host data")
4282 if new_node.secondary_ip != new_node.primary_ip:
4283 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
4286 node_verify_list = [self.cfg.GetMasterNode()]
4287 node_verify_param = {
4288 constants.NV_NODELIST: [node],
4289 # TODO: do a node-net-test as well?
4292 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
4293 self.cfg.GetClusterName())
4294 for verifier in node_verify_list:
4295 result[verifier].Raise("Cannot communicate with node %s" % verifier)
4296 nl_payload = result[verifier].payload[constants.NV_NODELIST]
4298 for failed in nl_payload:
4299 feedback_fn("ssh/hostname verification failed"
4300 " (checking from %s): %s" %
4301 (verifier, nl_payload[failed]))
4302 raise errors.OpExecError("ssh/hostname verification failed.")
4305 _RedistributeAncillaryFiles(self)
4306 self.context.ReaddNode(new_node)
4307 # make sure we redistribute the config
4308 self.cfg.Update(new_node, feedback_fn)
4309 # and make sure the new node will not have old files around
4310 if not new_node.master_candidate:
4311 result = self.rpc.call_node_demote_from_mc(new_node.name)
4312 msg = result.fail_msg
4314 self.LogWarning("Node failed to demote itself from master"
4315 " candidate status: %s" % msg)
4317 _RedistributeAncillaryFiles(self, additional_nodes=[node],
4318 additional_vm=self.op.vm_capable)
4319 self.context.AddNode(new_node, self.proc.GetECId())
4322 class LUNodeSetParams(LogicalUnit):
4323 """Modifies the parameters of a node.
4325 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
4326 to the node role (as _ROLE_*)
4327 @cvar _R2F: a dictionary from node role to tuples of flags
4328 @cvar _FLAGS: a list of attribute names corresponding to the flags
4331 HPATH = "node-modify"
4332 HTYPE = constants.HTYPE_NODE
4334 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
4336 (True, False, False): _ROLE_CANDIDATE,
4337 (False, True, False): _ROLE_DRAINED,
4338 (False, False, True): _ROLE_OFFLINE,
4339 (False, False, False): _ROLE_REGULAR,
4341 _R2F = dict((v, k) for k, v in _F2R.items())
4342 _FLAGS = ["master_candidate", "drained", "offline"]
4344 def CheckArguments(self):
4345 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4346 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
4347 self.op.master_capable, self.op.vm_capable,
4348 self.op.secondary_ip, self.op.ndparams]
4349 if all_mods.count(None) == len(all_mods):
4350 raise errors.OpPrereqError("Please pass at least one modification",
4352 if all_mods.count(True) > 1:
4353 raise errors.OpPrereqError("Can't set the node into more than one"
4354 " state at the same time",
4357 # Boolean value that tells us whether we might be demoting from MC
4358 self.might_demote = (self.op.master_candidate == False or
4359 self.op.offline == True or
4360 self.op.drained == True or
4361 self.op.master_capable == False)
4363 if self.op.secondary_ip:
4364 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
4365 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4366 " address" % self.op.secondary_ip,
4369 self.lock_all = self.op.auto_promote and self.might_demote
4370 self.lock_instances = self.op.secondary_ip is not None
4372 def ExpandNames(self):
4374 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
4376 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
4378 if self.lock_instances:
4379 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4381 def DeclareLocks(self, level):
4382 # If we have locked all instances, before waiting to lock nodes, release
4383 # all the ones living on nodes unrelated to the current operation.
4384 if level == locking.LEVEL_NODE and self.lock_instances:
4385 instances_release = []
4387 self.affected_instances = []
4388 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4389 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
4390 instance = self.context.cfg.GetInstanceInfo(instance_name)
4391 i_mirrored = instance.disk_template in constants.DTS_NET_MIRROR
4392 if i_mirrored and self.op.node_name in instance.all_nodes:
4393 instances_keep.append(instance_name)
4394 self.affected_instances.append(instance)
4396 instances_release.append(instance_name)
4397 if instances_release:
4398 self.context.glm.release(locking.LEVEL_INSTANCE, instances_release)
4399 self.acquired_locks[locking.LEVEL_INSTANCE] = instances_keep
4401 def BuildHooksEnv(self):
4404 This runs on the master node.
4408 "OP_TARGET": self.op.node_name,
4409 "MASTER_CANDIDATE": str(self.op.master_candidate),
4410 "OFFLINE": str(self.op.offline),
4411 "DRAINED": str(self.op.drained),
4412 "MASTER_CAPABLE": str(self.op.master_capable),
4413 "VM_CAPABLE": str(self.op.vm_capable),
4415 nl = [self.cfg.GetMasterNode(),
4419 def CheckPrereq(self):
4420 """Check prerequisites.
4422 This only checks the instance list against the existing names.
4425 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
4427 if (self.op.master_candidate is not None or
4428 self.op.drained is not None or
4429 self.op.offline is not None):
4430 # we can't change the master's node flags
4431 if self.op.node_name == self.cfg.GetMasterNode():
4432 raise errors.OpPrereqError("The master role can be changed"
4433 " only via master-failover",
4436 if self.op.master_candidate and not node.master_capable:
4437 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
4438 " it a master candidate" % node.name,
4441 if self.op.vm_capable == False:
4442 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
4444 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
4445 " the vm_capable flag" % node.name,
4448 if node.master_candidate and self.might_demote and not self.lock_all:
4449 assert not self.op.auto_promote, "auto_promote set but lock_all not"
4450 # check if after removing the current node, we're missing master
4452 (mc_remaining, mc_should, _) = \
4453 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
4454 if mc_remaining < mc_should:
4455 raise errors.OpPrereqError("Not enough master candidates, please"
4456 " pass auto promote option to allow"
4457 " promotion", errors.ECODE_STATE)
4459 self.old_flags = old_flags = (node.master_candidate,
4460 node.drained, node.offline)
4461 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
4462 self.old_role = old_role = self._F2R[old_flags]
4464 # Check for ineffective changes
4465 for attr in self._FLAGS:
4466 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
4467 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
4468 setattr(self.op, attr, None)
4470 # Past this point, any flag change to False means a transition
4471 # away from the respective state, as only real changes are kept
4473 # TODO: We might query the real power state if it supports OOB
4474 if _SupportsOob(self.cfg, node):
4475 if self.op.offline is False and not (node.powered or
4476 self.op.powered == True):
4477 raise errors.OpPrereqError(("Please power on node %s first before you"
4478 " can reset offline state") %
4480 elif self.op.powered is not None:
4481 raise errors.OpPrereqError(("Unable to change powered state for node %s"
4482 " which does not support out-of-band"
4483 " handling") % self.op.node_name)
4485 # If we're being deofflined/drained, we'll MC ourself if needed
4486 if (self.op.drained == False or self.op.offline == False or
4487 (self.op.master_capable and not node.master_capable)):
4488 if _DecideSelfPromotion(self):
4489 self.op.master_candidate = True
4490 self.LogInfo("Auto-promoting node to master candidate")
4492 # If we're no longer master capable, we'll demote ourselves from MC
4493 if self.op.master_capable == False and node.master_candidate:
4494 self.LogInfo("Demoting from master candidate")
4495 self.op.master_candidate = False
4498 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
4499 if self.op.master_candidate:
4500 new_role = self._ROLE_CANDIDATE
4501 elif self.op.drained:
4502 new_role = self._ROLE_DRAINED
4503 elif self.op.offline:
4504 new_role = self._ROLE_OFFLINE
4505 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
4506 # False is still in new flags, which means we're un-setting (the
4508 new_role = self._ROLE_REGULAR
4509 else: # no new flags, nothing, keep old role
4512 self.new_role = new_role
4514 if old_role == self._ROLE_OFFLINE and new_role != old_role:
4515 # Trying to transition out of offline status
4516 result = self.rpc.call_version([node.name])[node.name]
4518 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
4519 " to report its version: %s" %
4520 (node.name, result.fail_msg),
4523 self.LogWarning("Transitioning node from offline to online state"
4524 " without using re-add. Please make sure the node"
4527 if self.op.secondary_ip:
4528 # Ok even without locking, because this can't be changed by any LU
4529 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
4530 master_singlehomed = master.secondary_ip == master.primary_ip
4531 if master_singlehomed and self.op.secondary_ip:
4532 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
4533 " homed cluster", errors.ECODE_INVAL)
4536 if self.affected_instances:
4537 raise errors.OpPrereqError("Cannot change secondary ip: offline"
4538 " node has instances (%s) configured"
4539 " to use it" % self.affected_instances)
4541 # On online nodes, check that no instances are running, and that
4542 # the node has the new ip and we can reach it.
4543 for instance in self.affected_instances:
4544 _CheckInstanceDown(self, instance, "cannot change secondary ip")
4546 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
4547 if master.name != node.name:
4548 # check reachability from master secondary ip to new secondary ip
4549 if not netutils.TcpPing(self.op.secondary_ip,
4550 constants.DEFAULT_NODED_PORT,
4551 source=master.secondary_ip):
4552 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4553 " based ping to node daemon port",
4554 errors.ECODE_ENVIRON)
4556 if self.op.ndparams:
4557 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
4558 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
4559 self.new_ndparams = new_ndparams
4561 def Exec(self, feedback_fn):
4566 old_role = self.old_role
4567 new_role = self.new_role
4571 if self.op.ndparams:
4572 node.ndparams = self.new_ndparams
4574 if self.op.powered is not None:
4575 node.powered = self.op.powered
4577 for attr in ["master_capable", "vm_capable"]:
4578 val = getattr(self.op, attr)
4580 setattr(node, attr, val)
4581 result.append((attr, str(val)))
4583 if new_role != old_role:
4584 # Tell the node to demote itself, if no longer MC and not offline
4585 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
4586 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
4588 self.LogWarning("Node failed to demote itself: %s", msg)
4590 new_flags = self._R2F[new_role]
4591 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
4593 result.append((desc, str(nf)))
4594 (node.master_candidate, node.drained, node.offline) = new_flags
4596 # we locked all nodes, we adjust the CP before updating this node
4598 _AdjustCandidatePool(self, [node.name])
4600 if self.op.secondary_ip:
4601 node.secondary_ip = self.op.secondary_ip
4602 result.append(("secondary_ip", self.op.secondary_ip))
4604 # this will trigger configuration file update, if needed
4605 self.cfg.Update(node, feedback_fn)
4607 # this will trigger job queue propagation or cleanup if the mc
4609 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
4610 self.context.ReaddNode(node)
4615 class LUNodePowercycle(NoHooksLU):
4616 """Powercycles a node.
4621 def CheckArguments(self):
4622 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4623 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4624 raise errors.OpPrereqError("The node is the master and the force"
4625 " parameter was not set",
4628 def ExpandNames(self):
4629 """Locking for PowercycleNode.
4631 This is a last-resort option and shouldn't block on other
4632 jobs. Therefore, we grab no locks.
4635 self.needed_locks = {}
4637 def Exec(self, feedback_fn):
4641 result = self.rpc.call_node_powercycle(self.op.node_name,
4642 self.cfg.GetHypervisorType())
4643 result.Raise("Failed to schedule the reboot")
4644 return result.payload
4647 class LUClusterQuery(NoHooksLU):
4648 """Query cluster configuration.
4653 def ExpandNames(self):
4654 self.needed_locks = {}
4656 def Exec(self, feedback_fn):
4657 """Return cluster config.
4660 cluster = self.cfg.GetClusterInfo()
4663 # Filter just for enabled hypervisors
4664 for os_name, hv_dict in cluster.os_hvp.items():
4665 os_hvp[os_name] = {}
4666 for hv_name, hv_params in hv_dict.items():
4667 if hv_name in cluster.enabled_hypervisors:
4668 os_hvp[os_name][hv_name] = hv_params
4670 # Convert ip_family to ip_version
4671 primary_ip_version = constants.IP4_VERSION
4672 if cluster.primary_ip_family == netutils.IP6Address.family:
4673 primary_ip_version = constants.IP6_VERSION
4676 "software_version": constants.RELEASE_VERSION,
4677 "protocol_version": constants.PROTOCOL_VERSION,
4678 "config_version": constants.CONFIG_VERSION,
4679 "os_api_version": max(constants.OS_API_VERSIONS),
4680 "export_version": constants.EXPORT_VERSION,
4681 "architecture": (platform.architecture()[0], platform.machine()),
4682 "name": cluster.cluster_name,
4683 "master": cluster.master_node,
4684 "default_hypervisor": cluster.enabled_hypervisors[0],
4685 "enabled_hypervisors": cluster.enabled_hypervisors,
4686 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4687 for hypervisor_name in cluster.enabled_hypervisors]),
4689 "beparams": cluster.beparams,
4690 "osparams": cluster.osparams,
4691 "nicparams": cluster.nicparams,
4692 "ndparams": cluster.ndparams,
4693 "candidate_pool_size": cluster.candidate_pool_size,
4694 "master_netdev": cluster.master_netdev,
4695 "volume_group_name": cluster.volume_group_name,
4696 "drbd_usermode_helper": cluster.drbd_usermode_helper,
4697 "file_storage_dir": cluster.file_storage_dir,
4698 "maintain_node_health": cluster.maintain_node_health,
4699 "ctime": cluster.ctime,
4700 "mtime": cluster.mtime,
4701 "uuid": cluster.uuid,
4702 "tags": list(cluster.GetTags()),
4703 "uid_pool": cluster.uid_pool,
4704 "default_iallocator": cluster.default_iallocator,
4705 "reserved_lvs": cluster.reserved_lvs,
4706 "primary_ip_version": primary_ip_version,
4707 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4708 "hidden_os": cluster.hidden_os,
4709 "blacklisted_os": cluster.blacklisted_os,
4715 class LUClusterConfigQuery(NoHooksLU):
4716 """Return configuration values.
4720 _FIELDS_DYNAMIC = utils.FieldSet()
4721 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4722 "watcher_pause", "volume_group_name")
4724 def CheckArguments(self):
4725 _CheckOutputFields(static=self._FIELDS_STATIC,
4726 dynamic=self._FIELDS_DYNAMIC,
4727 selected=self.op.output_fields)
4729 def ExpandNames(self):
4730 self.needed_locks = {}
4732 def Exec(self, feedback_fn):
4733 """Dump a representation of the cluster config to the standard output.
4737 for field in self.op.output_fields:
4738 if field == "cluster_name":
4739 entry = self.cfg.GetClusterName()
4740 elif field == "master_node":
4741 entry = self.cfg.GetMasterNode()
4742 elif field == "drain_flag":
4743 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4744 elif field == "watcher_pause":
4745 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4746 elif field == "volume_group_name":
4747 entry = self.cfg.GetVGName()
4749 raise errors.ParameterError(field)
4750 values.append(entry)
4754 class LUInstanceActivateDisks(NoHooksLU):
4755 """Bring up an instance's disks.
4760 def ExpandNames(self):
4761 self._ExpandAndLockInstance()
4762 self.needed_locks[locking.LEVEL_NODE] = []
4763 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4765 def DeclareLocks(self, level):
4766 if level == locking.LEVEL_NODE:
4767 self._LockInstancesNodes()
4769 def CheckPrereq(self):
4770 """Check prerequisites.
4772 This checks that the instance is in the cluster.
4775 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4776 assert self.instance is not None, \
4777 "Cannot retrieve locked instance %s" % self.op.instance_name
4778 _CheckNodeOnline(self, self.instance.primary_node)
4780 def Exec(self, feedback_fn):
4781 """Activate the disks.
4784 disks_ok, disks_info = \
4785 _AssembleInstanceDisks(self, self.instance,
4786 ignore_size=self.op.ignore_size)
4788 raise errors.OpExecError("Cannot activate block devices")
4793 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4795 """Prepare the block devices for an instance.
4797 This sets up the block devices on all nodes.
4799 @type lu: L{LogicalUnit}
4800 @param lu: the logical unit on whose behalf we execute
4801 @type instance: L{objects.Instance}
4802 @param instance: the instance for whose disks we assemble
4803 @type disks: list of L{objects.Disk} or None
4804 @param disks: which disks to assemble (or all, if None)
4805 @type ignore_secondaries: boolean
4806 @param ignore_secondaries: if true, errors on secondary nodes
4807 won't result in an error return from the function
4808 @type ignore_size: boolean
4809 @param ignore_size: if true, the current known size of the disk
4810 will not be used during the disk activation, useful for cases
4811 when the size is wrong
4812 @return: False if the operation failed, otherwise a list of
4813 (host, instance_visible_name, node_visible_name)
4814 with the mapping from node devices to instance devices
4819 iname = instance.name
4820 disks = _ExpandCheckDisks(instance, disks)
4822 # With the two passes mechanism we try to reduce the window of
4823 # opportunity for the race condition of switching DRBD to primary
4824 # before handshaking occured, but we do not eliminate it
4826 # The proper fix would be to wait (with some limits) until the
4827 # connection has been made and drbd transitions from WFConnection
4828 # into any other network-connected state (Connected, SyncTarget,
4831 # 1st pass, assemble on all nodes in secondary mode
4832 for idx, inst_disk in enumerate(disks):
4833 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4835 node_disk = node_disk.Copy()
4836 node_disk.UnsetSize()
4837 lu.cfg.SetDiskID(node_disk, node)
4838 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
4839 msg = result.fail_msg
4841 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4842 " (is_primary=False, pass=1): %s",
4843 inst_disk.iv_name, node, msg)
4844 if not ignore_secondaries:
4847 # FIXME: race condition on drbd migration to primary
4849 # 2nd pass, do only the primary node
4850 for idx, inst_disk in enumerate(disks):
4853 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4854 if node != instance.primary_node:
4857 node_disk = node_disk.Copy()
4858 node_disk.UnsetSize()
4859 lu.cfg.SetDiskID(node_disk, node)
4860 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
4861 msg = result.fail_msg
4863 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4864 " (is_primary=True, pass=2): %s",
4865 inst_disk.iv_name, node, msg)
4868 dev_path = result.payload
4870 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4872 # leave the disks configured for the primary node
4873 # this is a workaround that would be fixed better by
4874 # improving the logical/physical id handling
4876 lu.cfg.SetDiskID(disk, instance.primary_node)
4878 return disks_ok, device_info
4881 def _StartInstanceDisks(lu, instance, force):
4882 """Start the disks of an instance.
4885 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4886 ignore_secondaries=force)
4888 _ShutdownInstanceDisks(lu, instance)
4889 if force is not None and not force:
4890 lu.proc.LogWarning("", hint="If the message above refers to a"
4892 " you can retry the operation using '--force'.")
4893 raise errors.OpExecError("Disk consistency error")
4896 class LUInstanceDeactivateDisks(NoHooksLU):
4897 """Shutdown an instance's disks.
4902 def ExpandNames(self):
4903 self._ExpandAndLockInstance()
4904 self.needed_locks[locking.LEVEL_NODE] = []
4905 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4907 def DeclareLocks(self, level):
4908 if level == locking.LEVEL_NODE:
4909 self._LockInstancesNodes()
4911 def CheckPrereq(self):
4912 """Check prerequisites.
4914 This checks that the instance is in the cluster.
4917 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4918 assert self.instance is not None, \
4919 "Cannot retrieve locked instance %s" % self.op.instance_name
4921 def Exec(self, feedback_fn):
4922 """Deactivate the disks
4925 instance = self.instance
4927 _ShutdownInstanceDisks(self, instance)
4929 _SafeShutdownInstanceDisks(self, instance)
4932 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4933 """Shutdown block devices of an instance.
4935 This function checks if an instance is running, before calling
4936 _ShutdownInstanceDisks.
4939 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4940 _ShutdownInstanceDisks(lu, instance, disks=disks)
4943 def _ExpandCheckDisks(instance, disks):
4944 """Return the instance disks selected by the disks list
4946 @type disks: list of L{objects.Disk} or None
4947 @param disks: selected disks
4948 @rtype: list of L{objects.Disk}
4949 @return: selected instance disks to act on
4953 return instance.disks
4955 if not set(disks).issubset(instance.disks):
4956 raise errors.ProgrammerError("Can only act on disks belonging to the"
4961 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4962 """Shutdown block devices of an instance.
4964 This does the shutdown on all nodes of the instance.
4966 If the ignore_primary is false, errors on the primary node are
4971 disks = _ExpandCheckDisks(instance, disks)
4974 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4975 lu.cfg.SetDiskID(top_disk, node)
4976 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4977 msg = result.fail_msg
4979 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4980 disk.iv_name, node, msg)
4981 if ((node == instance.primary_node and not ignore_primary) or
4982 (node != instance.primary_node and not result.offline)):
4987 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4988 """Checks if a node has enough free memory.
4990 This function check if a given node has the needed amount of free
4991 memory. In case the node has less memory or we cannot get the
4992 information from the node, this function raise an OpPrereqError
4995 @type lu: C{LogicalUnit}
4996 @param lu: a logical unit from which we get configuration data
4998 @param node: the node to check
4999 @type reason: C{str}
5000 @param reason: string to use in the error message
5001 @type requested: C{int}
5002 @param requested: the amount of memory in MiB to check for
5003 @type hypervisor_name: C{str}
5004 @param hypervisor_name: the hypervisor to ask for memory stats
5005 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
5006 we cannot check the node
5009 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
5010 nodeinfo[node].Raise("Can't get data from node %s" % node,
5011 prereq=True, ecode=errors.ECODE_ENVIRON)
5012 free_mem = nodeinfo[node].payload.get('memory_free', None)
5013 if not isinstance(free_mem, int):
5014 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
5015 " was '%s'" % (node, free_mem),
5016 errors.ECODE_ENVIRON)
5017 if requested > free_mem:
5018 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
5019 " needed %s MiB, available %s MiB" %
5020 (node, reason, requested, free_mem),
5024 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5025 """Checks if nodes have enough free disk space in the all VGs.
5027 This function check if all given nodes have the needed amount of
5028 free disk. In case any node has less disk or we cannot get the
5029 information from the node, this function raise an OpPrereqError
5032 @type lu: C{LogicalUnit}
5033 @param lu: a logical unit from which we get configuration data
5034 @type nodenames: C{list}
5035 @param nodenames: the list of node names to check
5036 @type req_sizes: C{dict}
5037 @param req_sizes: the hash of vg and corresponding amount of disk in
5039 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5040 or we cannot check the node
5043 for vg, req_size in req_sizes.items():
5044 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5047 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5048 """Checks if nodes have enough free disk space in the specified VG.
5050 This function check if all given nodes have the needed amount of
5051 free disk. In case any node has less disk or we cannot get the
5052 information from the node, this function raise an OpPrereqError
5055 @type lu: C{LogicalUnit}
5056 @param lu: a logical unit from which we get configuration data
5057 @type nodenames: C{list}
5058 @param nodenames: the list of node names to check
5060 @param vg: the volume group to check
5061 @type requested: C{int}
5062 @param requested: the amount of disk in MiB to check for
5063 @raise errors.OpPrereqError: if the node doesn't have enough disk,
5064 or we cannot check the node
5067 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
5068 for node in nodenames:
5069 info = nodeinfo[node]
5070 info.Raise("Cannot get current information from node %s" % node,
5071 prereq=True, ecode=errors.ECODE_ENVIRON)
5072 vg_free = info.payload.get("vg_free", None)
5073 if not isinstance(vg_free, int):
5074 raise errors.OpPrereqError("Can't compute free disk space on node"
5075 " %s for vg %s, result was '%s'" %
5076 (node, vg, vg_free), errors.ECODE_ENVIRON)
5077 if requested > vg_free:
5078 raise errors.OpPrereqError("Not enough disk space on target node %s"
5079 " vg %s: required %d MiB, available %d MiB" %
5080 (node, vg, requested, vg_free),
5084 class LUInstanceStartup(LogicalUnit):
5085 """Starts an instance.
5088 HPATH = "instance-start"
5089 HTYPE = constants.HTYPE_INSTANCE
5092 def CheckArguments(self):
5094 if self.op.beparams:
5095 # fill the beparams dict
5096 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5098 def ExpandNames(self):
5099 self._ExpandAndLockInstance()
5101 def BuildHooksEnv(self):
5104 This runs on master, primary and secondary nodes of the instance.
5108 "FORCE": self.op.force,
5110 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5111 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5114 def CheckPrereq(self):
5115 """Check prerequisites.
5117 This checks that the instance is in the cluster.
5120 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5121 assert self.instance is not None, \
5122 "Cannot retrieve locked instance %s" % self.op.instance_name
5125 if self.op.hvparams:
5126 # check hypervisor parameter syntax (locally)
5127 cluster = self.cfg.GetClusterInfo()
5128 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5129 filled_hvp = cluster.FillHV(instance)
5130 filled_hvp.update(self.op.hvparams)
5131 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5132 hv_type.CheckParameterSyntax(filled_hvp)
5133 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5135 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5137 if self.primary_offline and self.op.ignore_offline_nodes:
5138 self.proc.LogWarning("Ignoring offline primary node")
5140 if self.op.hvparams or self.op.beparams:
5141 self.proc.LogWarning("Overridden parameters are ignored")
5143 _CheckNodeOnline(self, instance.primary_node)
5145 bep = self.cfg.GetClusterInfo().FillBE(instance)
5147 # check bridges existence
5148 _CheckInstanceBridgesExist(self, instance)
5150 remote_info = self.rpc.call_instance_info(instance.primary_node,
5152 instance.hypervisor)
5153 remote_info.Raise("Error checking node %s" % instance.primary_node,
5154 prereq=True, ecode=errors.ECODE_ENVIRON)
5155 if not remote_info.payload: # not running already
5156 _CheckNodeFreeMemory(self, instance.primary_node,
5157 "starting instance %s" % instance.name,
5158 bep[constants.BE_MEMORY], instance.hypervisor)
5160 def Exec(self, feedback_fn):
5161 """Start the instance.
5164 instance = self.instance
5165 force = self.op.force
5167 self.cfg.MarkInstanceUp(instance.name)
5169 if self.primary_offline:
5170 assert self.op.ignore_offline_nodes
5171 self.proc.LogInfo("Primary node offline, marked instance as started")
5173 node_current = instance.primary_node
5175 _StartInstanceDisks(self, instance, force)
5177 result = self.rpc.call_instance_start(node_current, instance,
5178 self.op.hvparams, self.op.beparams)
5179 msg = result.fail_msg
5181 _ShutdownInstanceDisks(self, instance)
5182 raise errors.OpExecError("Could not start instance: %s" % msg)
5185 class LUInstanceReboot(LogicalUnit):
5186 """Reboot an instance.
5189 HPATH = "instance-reboot"
5190 HTYPE = constants.HTYPE_INSTANCE
5193 def ExpandNames(self):
5194 self._ExpandAndLockInstance()
5196 def BuildHooksEnv(self):
5199 This runs on master, primary and secondary nodes of the instance.
5203 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
5204 "REBOOT_TYPE": self.op.reboot_type,
5205 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5207 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5208 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5211 def CheckPrereq(self):
5212 """Check prerequisites.
5214 This checks that the instance is in the cluster.
5217 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5218 assert self.instance is not None, \
5219 "Cannot retrieve locked instance %s" % self.op.instance_name
5221 _CheckNodeOnline(self, instance.primary_node)
5223 # check bridges existence
5224 _CheckInstanceBridgesExist(self, instance)
5226 def Exec(self, feedback_fn):
5227 """Reboot the instance.
5230 instance = self.instance
5231 ignore_secondaries = self.op.ignore_secondaries
5232 reboot_type = self.op.reboot_type
5234 node_current = instance.primary_node
5236 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
5237 constants.INSTANCE_REBOOT_HARD]:
5238 for disk in instance.disks:
5239 self.cfg.SetDiskID(disk, node_current)
5240 result = self.rpc.call_instance_reboot(node_current, instance,
5242 self.op.shutdown_timeout)
5243 result.Raise("Could not reboot instance")
5245 result = self.rpc.call_instance_shutdown(node_current, instance,
5246 self.op.shutdown_timeout)
5247 result.Raise("Could not shutdown instance for full reboot")
5248 _ShutdownInstanceDisks(self, instance)
5249 _StartInstanceDisks(self, instance, ignore_secondaries)
5250 result = self.rpc.call_instance_start(node_current, instance, None, None)
5251 msg = result.fail_msg
5253 _ShutdownInstanceDisks(self, instance)
5254 raise errors.OpExecError("Could not start instance for"
5255 " full reboot: %s" % msg)
5257 self.cfg.MarkInstanceUp(instance.name)
5260 class LUInstanceShutdown(LogicalUnit):
5261 """Shutdown an instance.
5264 HPATH = "instance-stop"
5265 HTYPE = constants.HTYPE_INSTANCE
5268 def ExpandNames(self):
5269 self._ExpandAndLockInstance()
5271 def BuildHooksEnv(self):
5274 This runs on master, primary and secondary nodes of the instance.
5277 env = _BuildInstanceHookEnvByObject(self, self.instance)
5278 env["TIMEOUT"] = self.op.timeout
5279 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5282 def CheckPrereq(self):
5283 """Check prerequisites.
5285 This checks that the instance is in the cluster.
5288 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5289 assert self.instance is not None, \
5290 "Cannot retrieve locked instance %s" % self.op.instance_name
5292 self.primary_offline = \
5293 self.cfg.GetNodeInfo(self.instance.primary_node).offline
5295 if self.primary_offline and self.op.ignore_offline_nodes:
5296 self.proc.LogWarning("Ignoring offline primary node")
5298 _CheckNodeOnline(self, self.instance.primary_node)
5300 def Exec(self, feedback_fn):
5301 """Shutdown the instance.
5304 instance = self.instance
5305 node_current = instance.primary_node
5306 timeout = self.op.timeout
5308 self.cfg.MarkInstanceDown(instance.name)
5310 if self.primary_offline:
5311 assert self.op.ignore_offline_nodes
5312 self.proc.LogInfo("Primary node offline, marked instance as stopped")
5314 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
5315 msg = result.fail_msg
5317 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
5319 _ShutdownInstanceDisks(self, instance)
5322 class LUInstanceReinstall(LogicalUnit):
5323 """Reinstall an instance.
5326 HPATH = "instance-reinstall"
5327 HTYPE = constants.HTYPE_INSTANCE
5330 def ExpandNames(self):
5331 self._ExpandAndLockInstance()
5333 def BuildHooksEnv(self):
5336 This runs on master, primary and secondary nodes of the instance.
5339 env = _BuildInstanceHookEnvByObject(self, self.instance)
5340 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5343 def CheckPrereq(self):
5344 """Check prerequisites.
5346 This checks that the instance is in the cluster and is not running.
5349 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5350 assert instance is not None, \
5351 "Cannot retrieve locked instance %s" % self.op.instance_name
5352 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
5353 " offline, cannot reinstall")
5354 for node in instance.secondary_nodes:
5355 _CheckNodeOnline(self, node, "Instance secondary node offline,"
5356 " cannot reinstall")
5358 if instance.disk_template == constants.DT_DISKLESS:
5359 raise errors.OpPrereqError("Instance '%s' has no disks" %
5360 self.op.instance_name,
5362 _CheckInstanceDown(self, instance, "cannot reinstall")
5364 if self.op.os_type is not None:
5366 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
5367 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
5368 instance_os = self.op.os_type
5370 instance_os = instance.os
5372 nodelist = list(instance.all_nodes)
5374 if self.op.osparams:
5375 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
5376 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
5377 self.os_inst = i_osdict # the new dict (without defaults)
5381 self.instance = instance
5383 def Exec(self, feedback_fn):
5384 """Reinstall the instance.
5387 inst = self.instance
5389 if self.op.os_type is not None:
5390 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
5391 inst.os = self.op.os_type
5392 # Write to configuration
5393 self.cfg.Update(inst, feedback_fn)
5395 _StartInstanceDisks(self, inst, None)
5397 feedback_fn("Running the instance OS create scripts...")
5398 # FIXME: pass debug option from opcode to backend
5399 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
5400 self.op.debug_level,
5401 osparams=self.os_inst)
5402 result.Raise("Could not install OS for instance %s on node %s" %
5403 (inst.name, inst.primary_node))
5405 _ShutdownInstanceDisks(self, inst)
5408 class LUInstanceRecreateDisks(LogicalUnit):
5409 """Recreate an instance's missing disks.
5412 HPATH = "instance-recreate-disks"
5413 HTYPE = constants.HTYPE_INSTANCE
5416 def ExpandNames(self):
5417 self._ExpandAndLockInstance()
5419 def BuildHooksEnv(self):
5422 This runs on master, primary and secondary nodes of the instance.
5425 env = _BuildInstanceHookEnvByObject(self, self.instance)
5426 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5429 def CheckPrereq(self):
5430 """Check prerequisites.
5432 This checks that the instance is in the cluster and is not running.
5435 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5436 assert instance is not None, \
5437 "Cannot retrieve locked instance %s" % self.op.instance_name
5438 _CheckNodeOnline(self, instance.primary_node)
5440 if instance.disk_template == constants.DT_DISKLESS:
5441 raise errors.OpPrereqError("Instance '%s' has no disks" %
5442 self.op.instance_name, errors.ECODE_INVAL)
5443 _CheckInstanceDown(self, instance, "cannot recreate disks")
5445 if not self.op.disks:
5446 self.op.disks = range(len(instance.disks))
5448 for idx in self.op.disks:
5449 if idx >= len(instance.disks):
5450 raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
5453 self.instance = instance
5455 def Exec(self, feedback_fn):
5456 """Recreate the disks.
5460 for idx, _ in enumerate(self.instance.disks):
5461 if idx not in self.op.disks: # disk idx has not been passed in
5465 _CreateDisks(self, self.instance, to_skip=to_skip)
5468 class LUInstanceRename(LogicalUnit):
5469 """Rename an instance.
5472 HPATH = "instance-rename"
5473 HTYPE = constants.HTYPE_INSTANCE
5475 def CheckArguments(self):
5479 if self.op.ip_check and not self.op.name_check:
5480 # TODO: make the ip check more flexible and not depend on the name check
5481 raise errors.OpPrereqError("Cannot do ip check without a name check",
5484 def BuildHooksEnv(self):
5487 This runs on master, primary and secondary nodes of the instance.
5490 env = _BuildInstanceHookEnvByObject(self, self.instance)
5491 env["INSTANCE_NEW_NAME"] = self.op.new_name
5492 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5495 def CheckPrereq(self):
5496 """Check prerequisites.
5498 This checks that the instance is in the cluster and is not running.
5501 self.op.instance_name = _ExpandInstanceName(self.cfg,
5502 self.op.instance_name)
5503 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5504 assert instance is not None
5505 _CheckNodeOnline(self, instance.primary_node)
5506 _CheckInstanceDown(self, instance, "cannot rename")
5507 self.instance = instance
5509 new_name = self.op.new_name
5510 if self.op.name_check:
5511 hostname = netutils.GetHostname(name=new_name)
5512 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
5514 new_name = self.op.new_name = hostname.name
5515 if (self.op.ip_check and
5516 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
5517 raise errors.OpPrereqError("IP %s of instance %s already in use" %
5518 (hostname.ip, new_name),
5519 errors.ECODE_NOTUNIQUE)
5521 instance_list = self.cfg.GetInstanceList()
5522 if new_name in instance_list and new_name != instance.name:
5523 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5524 new_name, errors.ECODE_EXISTS)
5526 def Exec(self, feedback_fn):
5527 """Rename the instance.
5530 inst = self.instance
5531 old_name = inst.name
5533 rename_file_storage = False
5534 if (inst.disk_template == constants.DT_FILE and
5535 self.op.new_name != inst.name):
5536 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5537 rename_file_storage = True
5539 self.cfg.RenameInstance(inst.name, self.op.new_name)
5540 # Change the instance lock. This is definitely safe while we hold the BGL
5541 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5542 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5544 # re-read the instance from the configuration after rename
5545 inst = self.cfg.GetInstanceInfo(self.op.new_name)
5547 if rename_file_storage:
5548 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5549 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5550 old_file_storage_dir,
5551 new_file_storage_dir)
5552 result.Raise("Could not rename on node %s directory '%s' to '%s'"
5553 " (but the instance has been renamed in Ganeti)" %
5554 (inst.primary_node, old_file_storage_dir,
5555 new_file_storage_dir))
5557 _StartInstanceDisks(self, inst, None)
5559 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5560 old_name, self.op.debug_level)
5561 msg = result.fail_msg
5563 msg = ("Could not run OS rename script for instance %s on node %s"
5564 " (but the instance has been renamed in Ganeti): %s" %
5565 (inst.name, inst.primary_node, msg))
5566 self.proc.LogWarning(msg)
5568 _ShutdownInstanceDisks(self, inst)
5573 class LUInstanceRemove(LogicalUnit):
5574 """Remove an instance.
5577 HPATH = "instance-remove"
5578 HTYPE = constants.HTYPE_INSTANCE
5581 def ExpandNames(self):
5582 self._ExpandAndLockInstance()
5583 self.needed_locks[locking.LEVEL_NODE] = []
5584 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5586 def DeclareLocks(self, level):
5587 if level == locking.LEVEL_NODE:
5588 self._LockInstancesNodes()
5590 def BuildHooksEnv(self):
5593 This runs on master, primary and secondary nodes of the instance.
5596 env = _BuildInstanceHookEnvByObject(self, self.instance)
5597 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5598 nl = [self.cfg.GetMasterNode()]
5599 nl_post = list(self.instance.all_nodes) + nl
5600 return env, nl, nl_post
5602 def CheckPrereq(self):
5603 """Check prerequisites.
5605 This checks that the instance is in the cluster.
5608 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5609 assert self.instance is not None, \
5610 "Cannot retrieve locked instance %s" % self.op.instance_name
5612 def Exec(self, feedback_fn):
5613 """Remove the instance.
5616 instance = self.instance
5617 logging.info("Shutting down instance %s on node %s",
5618 instance.name, instance.primary_node)
5620 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5621 self.op.shutdown_timeout)
5622 msg = result.fail_msg
5624 if self.op.ignore_failures:
5625 feedback_fn("Warning: can't shutdown instance: %s" % msg)
5627 raise errors.OpExecError("Could not shutdown instance %s on"
5629 (instance.name, instance.primary_node, msg))
5631 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5634 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5635 """Utility function to remove an instance.
5638 logging.info("Removing block devices for instance %s", instance.name)
5640 if not _RemoveDisks(lu, instance):
5641 if not ignore_failures:
5642 raise errors.OpExecError("Can't remove instance's disks")
5643 feedback_fn("Warning: can't remove instance's disks")
5645 logging.info("Removing instance %s out of cluster config", instance.name)
5647 lu.cfg.RemoveInstance(instance.name)
5649 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5650 "Instance lock removal conflict"
5652 # Remove lock for the instance
5653 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5656 class LUInstanceQuery(NoHooksLU):
5657 """Logical unit for querying instances.
5660 # pylint: disable-msg=W0142
5663 def CheckArguments(self):
5664 self.iq = _InstanceQuery(qlang.MakeSimpleFilter("name", self.op.names),
5665 self.op.output_fields, self.op.use_locking)
5667 def ExpandNames(self):
5668 self.iq.ExpandNames(self)
5670 def DeclareLocks(self, level):
5671 self.iq.DeclareLocks(self, level)
5673 def Exec(self, feedback_fn):
5674 return self.iq.OldStyleQuery(self)
5677 class LUInstanceFailover(LogicalUnit):
5678 """Failover an instance.
5681 HPATH = "instance-failover"
5682 HTYPE = constants.HTYPE_INSTANCE
5685 def ExpandNames(self):
5686 self._ExpandAndLockInstance()
5687 self.needed_locks[locking.LEVEL_NODE] = []
5688 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5690 def DeclareLocks(self, level):
5691 if level == locking.LEVEL_NODE:
5692 self._LockInstancesNodes()
5694 def BuildHooksEnv(self):
5697 This runs on master, primary and secondary nodes of the instance.
5700 instance = self.instance
5701 source_node = instance.primary_node
5702 target_node = instance.secondary_nodes[0]
5704 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5705 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5706 "OLD_PRIMARY": source_node,
5707 "OLD_SECONDARY": target_node,
5708 "NEW_PRIMARY": target_node,
5709 "NEW_SECONDARY": source_node,
5711 env.update(_BuildInstanceHookEnvByObject(self, instance))
5712 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5714 nl_post.append(source_node)
5715 return env, nl, nl_post
5717 def CheckPrereq(self):
5718 """Check prerequisites.
5720 This checks that the instance is in the cluster.
5723 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5724 assert self.instance is not None, \
5725 "Cannot retrieve locked instance %s" % self.op.instance_name
5727 bep = self.cfg.GetClusterInfo().FillBE(instance)
5728 if instance.disk_template not in constants.DTS_NET_MIRROR:
5729 raise errors.OpPrereqError("Instance's disk layout is not"
5730 " network mirrored, cannot failover.",
5733 secondary_nodes = instance.secondary_nodes
5734 if not secondary_nodes:
5735 raise errors.ProgrammerError("no secondary node but using "
5736 "a mirrored disk template")
5738 target_node = secondary_nodes[0]
5739 _CheckNodeOnline(self, target_node)
5740 _CheckNodeNotDrained(self, target_node)
5741 if instance.admin_up:
5742 # check memory requirements on the secondary node
5743 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5744 instance.name, bep[constants.BE_MEMORY],
5745 instance.hypervisor)
5747 self.LogInfo("Not checking memory on the secondary node as"
5748 " instance will not be started")
5750 # check bridge existance
5751 _CheckInstanceBridgesExist(self, instance, node=target_node)
5753 def Exec(self, feedback_fn):
5754 """Failover an instance.
5756 The failover is done by shutting it down on its present node and
5757 starting it on the secondary.
5760 instance = self.instance
5761 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
5763 source_node = instance.primary_node
5764 target_node = instance.secondary_nodes[0]
5766 if instance.admin_up:
5767 feedback_fn("* checking disk consistency between source and target")
5768 for dev in instance.disks:
5769 # for drbd, these are drbd over lvm
5770 if not _CheckDiskConsistency(self, dev, target_node, False):
5771 if not self.op.ignore_consistency:
5772 raise errors.OpExecError("Disk %s is degraded on target node,"
5773 " aborting failover." % dev.iv_name)
5775 feedback_fn("* not checking disk consistency as instance is not running")
5777 feedback_fn("* shutting down instance on source node")
5778 logging.info("Shutting down instance %s on node %s",
5779 instance.name, source_node)
5781 result = self.rpc.call_instance_shutdown(source_node, instance,
5782 self.op.shutdown_timeout)
5783 msg = result.fail_msg
5785 if self.op.ignore_consistency or primary_node.offline:
5786 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5787 " Proceeding anyway. Please make sure node"
5788 " %s is down. Error details: %s",
5789 instance.name, source_node, source_node, msg)
5791 raise errors.OpExecError("Could not shutdown instance %s on"
5793 (instance.name, source_node, msg))
5795 feedback_fn("* deactivating the instance's disks on source node")
5796 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5797 raise errors.OpExecError("Can't shut down the instance's disks.")
5799 instance.primary_node = target_node
5800 # distribute new instance config to the other nodes
5801 self.cfg.Update(instance, feedback_fn)
5803 # Only start the instance if it's marked as up
5804 if instance.admin_up:
5805 feedback_fn("* activating the instance's disks on target node")
5806 logging.info("Starting instance %s on node %s",
5807 instance.name, target_node)
5809 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5810 ignore_secondaries=True)
5812 _ShutdownInstanceDisks(self, instance)
5813 raise errors.OpExecError("Can't activate the instance's disks")
5815 feedback_fn("* starting the instance on the target node")
5816 result = self.rpc.call_instance_start(target_node, instance, None, None)
5817 msg = result.fail_msg
5819 _ShutdownInstanceDisks(self, instance)
5820 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5821 (instance.name, target_node, msg))
5824 class LUInstanceMigrate(LogicalUnit):
5825 """Migrate an instance.
5827 This is migration without shutting down, compared to the failover,
5828 which is done with shutdown.
5831 HPATH = "instance-migrate"
5832 HTYPE = constants.HTYPE_INSTANCE
5835 def ExpandNames(self):
5836 self._ExpandAndLockInstance()
5838 self.needed_locks[locking.LEVEL_NODE] = []
5839 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5841 self._migrater = TLMigrateInstance(self, self.op.instance_name,
5843 self.tasklets = [self._migrater]
5845 def DeclareLocks(self, level):
5846 if level == locking.LEVEL_NODE:
5847 self._LockInstancesNodes()
5849 def BuildHooksEnv(self):
5852 This runs on master, primary and secondary nodes of the instance.
5855 instance = self._migrater.instance
5856 source_node = instance.primary_node
5857 target_node = instance.secondary_nodes[0]
5858 env = _BuildInstanceHookEnvByObject(self, instance)
5859 env["MIGRATE_LIVE"] = self._migrater.live
5860 env["MIGRATE_CLEANUP"] = self.op.cleanup
5862 "OLD_PRIMARY": source_node,
5863 "OLD_SECONDARY": target_node,
5864 "NEW_PRIMARY": target_node,
5865 "NEW_SECONDARY": source_node,
5867 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5869 nl_post.append(source_node)
5870 return env, nl, nl_post
5873 class LUInstanceMove(LogicalUnit):
5874 """Move an instance by data-copying.
5877 HPATH = "instance-move"
5878 HTYPE = constants.HTYPE_INSTANCE
5881 def ExpandNames(self):
5882 self._ExpandAndLockInstance()
5883 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5884 self.op.target_node = target_node
5885 self.needed_locks[locking.LEVEL_NODE] = [target_node]
5886 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5888 def DeclareLocks(self, level):
5889 if level == locking.LEVEL_NODE:
5890 self._LockInstancesNodes(primary_only=True)
5892 def BuildHooksEnv(self):
5895 This runs on master, primary and secondary nodes of the instance.
5899 "TARGET_NODE": self.op.target_node,
5900 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5902 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5903 nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5904 self.op.target_node]
5907 def CheckPrereq(self):
5908 """Check prerequisites.
5910 This checks that the instance is in the cluster.
5913 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5914 assert self.instance is not None, \
5915 "Cannot retrieve locked instance %s" % self.op.instance_name
5917 node = self.cfg.GetNodeInfo(self.op.target_node)
5918 assert node is not None, \
5919 "Cannot retrieve locked node %s" % self.op.target_node
5921 self.target_node = target_node = node.name
5923 if target_node == instance.primary_node:
5924 raise errors.OpPrereqError("Instance %s is already on the node %s" %
5925 (instance.name, target_node),
5928 bep = self.cfg.GetClusterInfo().FillBE(instance)
5930 for idx, dsk in enumerate(instance.disks):
5931 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5932 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5933 " cannot copy" % idx, errors.ECODE_STATE)
5935 _CheckNodeOnline(self, target_node)
5936 _CheckNodeNotDrained(self, target_node)
5937 _CheckNodeVmCapable(self, target_node)
5939 if instance.admin_up:
5940 # check memory requirements on the secondary node
5941 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5942 instance.name, bep[constants.BE_MEMORY],
5943 instance.hypervisor)
5945 self.LogInfo("Not checking memory on the secondary node as"
5946 " instance will not be started")
5948 # check bridge existance
5949 _CheckInstanceBridgesExist(self, instance, node=target_node)
5951 def Exec(self, feedback_fn):
5952 """Move an instance.
5954 The move is done by shutting it down on its present node, copying
5955 the data over (slow) and starting it on the new node.
5958 instance = self.instance
5960 source_node = instance.primary_node
5961 target_node = self.target_node
5963 self.LogInfo("Shutting down instance %s on source node %s",
5964 instance.name, source_node)
5966 result = self.rpc.call_instance_shutdown(source_node, instance,
5967 self.op.shutdown_timeout)
5968 msg = result.fail_msg
5970 if self.op.ignore_consistency:
5971 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5972 " Proceeding anyway. Please make sure node"
5973 " %s is down. Error details: %s",
5974 instance.name, source_node, source_node, msg)
5976 raise errors.OpExecError("Could not shutdown instance %s on"
5978 (instance.name, source_node, msg))
5980 # create the target disks
5982 _CreateDisks(self, instance, target_node=target_node)
5983 except errors.OpExecError:
5984 self.LogWarning("Device creation failed, reverting...")
5986 _RemoveDisks(self, instance, target_node=target_node)
5988 self.cfg.ReleaseDRBDMinors(instance.name)
5991 cluster_name = self.cfg.GetClusterInfo().cluster_name
5994 # activate, get path, copy the data over
5995 for idx, disk in enumerate(instance.disks):
5996 self.LogInfo("Copying data for disk %d", idx)
5997 result = self.rpc.call_blockdev_assemble(target_node, disk,
5998 instance.name, True, idx)
6000 self.LogWarning("Can't assemble newly created disk %d: %s",
6001 idx, result.fail_msg)
6002 errs.append(result.fail_msg)
6004 dev_path = result.payload
6005 result = self.rpc.call_blockdev_export(source_node, disk,
6006 target_node, dev_path,
6009 self.LogWarning("Can't copy data over for disk %d: %s",
6010 idx, result.fail_msg)
6011 errs.append(result.fail_msg)
6015 self.LogWarning("Some disks failed to copy, aborting")
6017 _RemoveDisks(self, instance, target_node=target_node)
6019 self.cfg.ReleaseDRBDMinors(instance.name)
6020 raise errors.OpExecError("Errors during disk copy: %s" %
6023 instance.primary_node = target_node
6024 self.cfg.Update(instance, feedback_fn)
6026 self.LogInfo("Removing the disks on the original node")
6027 _RemoveDisks(self, instance, target_node=source_node)
6029 # Only start the instance if it's marked as up
6030 if instance.admin_up:
6031 self.LogInfo("Starting instance %s on node %s",
6032 instance.name, target_node)
6034 disks_ok, _ = _AssembleInstanceDisks(self, instance,
6035 ignore_secondaries=True)
6037 _ShutdownInstanceDisks(self, instance)
6038 raise errors.OpExecError("Can't activate the instance's disks")
6040 result = self.rpc.call_instance_start(target_node, instance, None, None)
6041 msg = result.fail_msg
6043 _ShutdownInstanceDisks(self, instance)
6044 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6045 (instance.name, target_node, msg))
6048 class LUNodeMigrate(LogicalUnit):
6049 """Migrate all instances from a node.
6052 HPATH = "node-migrate"
6053 HTYPE = constants.HTYPE_NODE
6056 def ExpandNames(self):
6057 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6059 self.needed_locks = {
6060 locking.LEVEL_NODE: [self.op.node_name],
6063 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6065 # Create tasklets for migrating instances for all instances on this node
6069 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
6070 logging.debug("Migrating instance %s", inst.name)
6071 names.append(inst.name)
6073 tasklets.append(TLMigrateInstance(self, inst.name, False))
6075 self.tasklets = tasklets
6077 # Declare instance locks
6078 self.needed_locks[locking.LEVEL_INSTANCE] = names
6080 def DeclareLocks(self, level):
6081 if level == locking.LEVEL_NODE:
6082 self._LockInstancesNodes()
6084 def BuildHooksEnv(self):
6087 This runs on the master, the primary and all the secondaries.
6091 "NODE_NAME": self.op.node_name,
6094 nl = [self.cfg.GetMasterNode()]
6096 return (env, nl, nl)
6099 class TLMigrateInstance(Tasklet):
6100 """Tasklet class for instance migration.
6103 @ivar live: whether the migration will be done live or non-live;
6104 this variable is initalized only after CheckPrereq has run
6107 def __init__(self, lu, instance_name, cleanup):
6108 """Initializes this class.
6111 Tasklet.__init__(self, lu)
6114 self.instance_name = instance_name
6115 self.cleanup = cleanup
6116 self.live = False # will be overridden later
6118 def CheckPrereq(self):
6119 """Check prerequisites.
6121 This checks that the instance is in the cluster.
6124 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
6125 instance = self.cfg.GetInstanceInfo(instance_name)
6126 assert instance is not None
6128 if instance.disk_template != constants.DT_DRBD8:
6129 raise errors.OpPrereqError("Instance's disk layout is not"
6130 " drbd8, cannot migrate.", errors.ECODE_STATE)
6132 secondary_nodes = instance.secondary_nodes
6133 if not secondary_nodes:
6134 raise errors.ConfigurationError("No secondary node but using"
6135 " drbd8 disk template")
6137 i_be = self.cfg.GetClusterInfo().FillBE(instance)
6139 target_node = secondary_nodes[0]
6140 # check memory requirements on the secondary node
6141 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
6142 instance.name, i_be[constants.BE_MEMORY],
6143 instance.hypervisor)
6145 # check bridge existance
6146 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
6148 if not self.cleanup:
6149 _CheckNodeNotDrained(self.lu, target_node)
6150 result = self.rpc.call_instance_migratable(instance.primary_node,
6152 result.Raise("Can't migrate, please use failover",
6153 prereq=True, ecode=errors.ECODE_STATE)
6155 self.instance = instance
6157 if self.lu.op.live is not None and self.lu.op.mode is not None:
6158 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
6159 " parameters are accepted",
6161 if self.lu.op.live is not None:
6163 self.lu.op.mode = constants.HT_MIGRATION_LIVE
6165 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
6166 # reset the 'live' parameter to None so that repeated
6167 # invocations of CheckPrereq do not raise an exception
6168 self.lu.op.live = None
6169 elif self.lu.op.mode is None:
6170 # read the default value from the hypervisor
6171 i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
6172 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
6174 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
6176 def _WaitUntilSync(self):
6177 """Poll with custom rpc for disk sync.
6179 This uses our own step-based rpc call.
6182 self.feedback_fn("* wait until resync is done")
6186 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
6188 self.instance.disks)
6190 for node, nres in result.items():
6191 nres.Raise("Cannot resync disks on node %s" % node)
6192 node_done, node_percent = nres.payload
6193 all_done = all_done and node_done
6194 if node_percent is not None:
6195 min_percent = min(min_percent, node_percent)
6197 if min_percent < 100:
6198 self.feedback_fn(" - progress: %.1f%%" % min_percent)
6201 def _EnsureSecondary(self, node):
6202 """Demote a node to secondary.
6205 self.feedback_fn("* switching node %s to secondary mode" % node)
6207 for dev in self.instance.disks:
6208 self.cfg.SetDiskID(dev, node)
6210 result = self.rpc.call_blockdev_close(node, self.instance.name,
6211 self.instance.disks)
6212 result.Raise("Cannot change disk to secondary on node %s" % node)
6214 def _GoStandalone(self):
6215 """Disconnect from the network.
6218 self.feedback_fn("* changing into standalone mode")
6219 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
6220 self.instance.disks)
6221 for node, nres in result.items():
6222 nres.Raise("Cannot disconnect disks node %s" % node)
6224 def _GoReconnect(self, multimaster):
6225 """Reconnect to the network.
6231 msg = "single-master"
6232 self.feedback_fn("* changing disks into %s mode" % msg)
6233 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
6234 self.instance.disks,
6235 self.instance.name, multimaster)
6236 for node, nres in result.items():
6237 nres.Raise("Cannot change disks config on node %s" % node)
6239 def _ExecCleanup(self):
6240 """Try to cleanup after a failed migration.
6242 The cleanup is done by:
6243 - check that the instance is running only on one node
6244 (and update the config if needed)
6245 - change disks on its secondary node to secondary
6246 - wait until disks are fully synchronized
6247 - disconnect from the network
6248 - change disks into single-master mode
6249 - wait again until disks are fully synchronized
6252 instance = self.instance
6253 target_node = self.target_node
6254 source_node = self.source_node
6256 # check running on only one node
6257 self.feedback_fn("* checking where the instance actually runs"
6258 " (if this hangs, the hypervisor might be in"
6260 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6261 for node, result in ins_l.items():
6262 result.Raise("Can't contact node %s" % node)
6264 runningon_source = instance.name in ins_l[source_node].payload
6265 runningon_target = instance.name in ins_l[target_node].payload
6267 if runningon_source and runningon_target:
6268 raise errors.OpExecError("Instance seems to be running on two nodes,"
6269 " or the hypervisor is confused. You will have"
6270 " to ensure manually that it runs only on one"
6271 " and restart this operation.")
6273 if not (runningon_source or runningon_target):
6274 raise errors.OpExecError("Instance does not seem to be running at all."
6275 " In this case, it's safer to repair by"
6276 " running 'gnt-instance stop' to ensure disk"
6277 " shutdown, and then restarting it.")
6279 if runningon_target:
6280 # the migration has actually succeeded, we need to update the config
6281 self.feedback_fn("* instance running on secondary node (%s),"
6282 " updating config" % target_node)
6283 instance.primary_node = target_node
6284 self.cfg.Update(instance, self.feedback_fn)
6285 demoted_node = source_node
6287 self.feedback_fn("* instance confirmed to be running on its"
6288 " primary node (%s)" % source_node)
6289 demoted_node = target_node
6291 self._EnsureSecondary(demoted_node)
6293 self._WaitUntilSync()
6294 except errors.OpExecError:
6295 # we ignore here errors, since if the device is standalone, it
6296 # won't be able to sync
6298 self._GoStandalone()
6299 self._GoReconnect(False)
6300 self._WaitUntilSync()
6302 self.feedback_fn("* done")
6304 def _RevertDiskStatus(self):
6305 """Try to revert the disk status after a failed migration.
6308 target_node = self.target_node
6310 self._EnsureSecondary(target_node)
6311 self._GoStandalone()
6312 self._GoReconnect(False)
6313 self._WaitUntilSync()
6314 except errors.OpExecError, err:
6315 self.lu.LogWarning("Migration failed and I can't reconnect the"
6316 " drives: error '%s'\n"
6317 "Please look and recover the instance status" %
6320 def _AbortMigration(self):
6321 """Call the hypervisor code to abort a started migration.
6324 instance = self.instance
6325 target_node = self.target_node
6326 migration_info = self.migration_info
6328 abort_result = self.rpc.call_finalize_migration(target_node,
6332 abort_msg = abort_result.fail_msg
6334 logging.error("Aborting migration failed on target node %s: %s",
6335 target_node, abort_msg)
6336 # Don't raise an exception here, as we stil have to try to revert the
6337 # disk status, even if this step failed.
6339 def _ExecMigration(self):
6340 """Migrate an instance.
6342 The migrate is done by:
6343 - change the disks into dual-master mode
6344 - wait until disks are fully synchronized again
6345 - migrate the instance
6346 - change disks on the new secondary node (the old primary) to secondary
6347 - wait until disks are fully synchronized
6348 - change disks into single-master mode
6351 instance = self.instance
6352 target_node = self.target_node
6353 source_node = self.source_node
6355 self.feedback_fn("* checking disk consistency between source and target")
6356 for dev in instance.disks:
6357 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6358 raise errors.OpExecError("Disk %s is degraded or not fully"
6359 " synchronized on target node,"
6360 " aborting migrate." % dev.iv_name)
6362 # First get the migration information from the remote node
6363 result = self.rpc.call_migration_info(source_node, instance)
6364 msg = result.fail_msg
6366 log_err = ("Failed fetching source migration information from %s: %s" %
6368 logging.error(log_err)
6369 raise errors.OpExecError(log_err)
6371 self.migration_info = migration_info = result.payload
6373 # Then switch the disks to master/master mode
6374 self._EnsureSecondary(target_node)
6375 self._GoStandalone()
6376 self._GoReconnect(True)
6377 self._WaitUntilSync()
6379 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6380 result = self.rpc.call_accept_instance(target_node,
6383 self.nodes_ip[target_node])
6385 msg = result.fail_msg
6387 logging.error("Instance pre-migration failed, trying to revert"
6388 " disk status: %s", msg)
6389 self.feedback_fn("Pre-migration failed, aborting")
6390 self._AbortMigration()
6391 self._RevertDiskStatus()
6392 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6393 (instance.name, msg))
6395 self.feedback_fn("* migrating instance to %s" % target_node)
6397 result = self.rpc.call_instance_migrate(source_node, instance,
6398 self.nodes_ip[target_node],
6400 msg = result.fail_msg
6402 logging.error("Instance migration failed, trying to revert"
6403 " disk status: %s", msg)
6404 self.feedback_fn("Migration failed, aborting")
6405 self._AbortMigration()
6406 self._RevertDiskStatus()
6407 raise errors.OpExecError("Could not migrate instance %s: %s" %
6408 (instance.name, msg))
6411 instance.primary_node = target_node
6412 # distribute new instance config to the other nodes
6413 self.cfg.Update(instance, self.feedback_fn)
6415 result = self.rpc.call_finalize_migration(target_node,
6419 msg = result.fail_msg
6421 logging.error("Instance migration succeeded, but finalization failed:"
6423 raise errors.OpExecError("Could not finalize instance migration: %s" %
6426 self._EnsureSecondary(source_node)
6427 self._WaitUntilSync()
6428 self._GoStandalone()
6429 self._GoReconnect(False)
6430 self._WaitUntilSync()
6432 self.feedback_fn("* done")
6434 def Exec(self, feedback_fn):
6435 """Perform the migration.
6438 feedback_fn("Migrating instance %s" % self.instance.name)
6440 self.feedback_fn = feedback_fn
6442 self.source_node = self.instance.primary_node
6443 self.target_node = self.instance.secondary_nodes[0]
6444 self.all_nodes = [self.source_node, self.target_node]
6446 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6447 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6451 return self._ExecCleanup()
6453 return self._ExecMigration()
6456 def _CreateBlockDev(lu, node, instance, device, force_create,
6458 """Create a tree of block devices on a given node.
6460 If this device type has to be created on secondaries, create it and
6463 If not, just recurse to children keeping the same 'force' value.
6465 @param lu: the lu on whose behalf we execute
6466 @param node: the node on which to create the device
6467 @type instance: L{objects.Instance}
6468 @param instance: the instance which owns the device
6469 @type device: L{objects.Disk}
6470 @param device: the device to create
6471 @type force_create: boolean
6472 @param force_create: whether to force creation of this device; this
6473 will be change to True whenever we find a device which has
6474 CreateOnSecondary() attribute
6475 @param info: the extra 'metadata' we should attach to the device
6476 (this will be represented as a LVM tag)
6477 @type force_open: boolean
6478 @param force_open: this parameter will be passes to the
6479 L{backend.BlockdevCreate} function where it specifies
6480 whether we run on primary or not, and it affects both
6481 the child assembly and the device own Open() execution
6484 if device.CreateOnSecondary():
6488 for child in device.children:
6489 _CreateBlockDev(lu, node, instance, child, force_create,
6492 if not force_create:
6495 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6498 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6499 """Create a single block device on a given node.
6501 This will not recurse over children of the device, so they must be
6504 @param lu: the lu on whose behalf we execute
6505 @param node: the node on which to create the device
6506 @type instance: L{objects.Instance}
6507 @param instance: the instance which owns the device
6508 @type device: L{objects.Disk}
6509 @param device: the device to create
6510 @param info: the extra 'metadata' we should attach to the device
6511 (this will be represented as a LVM tag)
6512 @type force_open: boolean
6513 @param force_open: this parameter will be passes to the
6514 L{backend.BlockdevCreate} function where it specifies
6515 whether we run on primary or not, and it affects both
6516 the child assembly and the device own Open() execution
6519 lu.cfg.SetDiskID(device, node)
6520 result = lu.rpc.call_blockdev_create(node, device, device.size,
6521 instance.name, force_open, info)
6522 result.Raise("Can't create block device %s on"
6523 " node %s for instance %s" % (device, node, instance.name))
6524 if device.physical_id is None:
6525 device.physical_id = result.payload
6528 def _GenerateUniqueNames(lu, exts):
6529 """Generate a suitable LV name.
6531 This will generate a logical volume name for the given instance.
6536 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6537 results.append("%s%s" % (new_id, val))
6541 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgname, names, iv_name,
6543 """Generate a drbd8 device complete with its children.
6546 port = lu.cfg.AllocatePort()
6547 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6548 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6549 logical_id=(vgname, names[0]))
6550 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6551 logical_id=(vgname, names[1]))
6552 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6553 logical_id=(primary, secondary, port,
6556 children=[dev_data, dev_meta],
6561 def _GenerateDiskTemplate(lu, template_name,
6562 instance_name, primary_node,
6563 secondary_nodes, disk_info,
6564 file_storage_dir, file_driver,
6565 base_index, feedback_fn):
6566 """Generate the entire disk layout for a given template type.
6569 #TODO: compute space requirements
6571 vgname = lu.cfg.GetVGName()
6572 disk_count = len(disk_info)
6574 if template_name == constants.DT_DISKLESS:
6576 elif template_name == constants.DT_PLAIN:
6577 if len(secondary_nodes) != 0:
6578 raise errors.ProgrammerError("Wrong template configuration")
6580 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6581 for i in range(disk_count)])
6582 for idx, disk in enumerate(disk_info):
6583 disk_index = idx + base_index
6584 vg = disk.get("vg", vgname)
6585 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
6586 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6587 logical_id=(vg, names[idx]),
6588 iv_name="disk/%d" % disk_index,
6590 disks.append(disk_dev)
6591 elif template_name == constants.DT_DRBD8:
6592 if len(secondary_nodes) != 1:
6593 raise errors.ProgrammerError("Wrong template configuration")
6594 remote_node = secondary_nodes[0]
6595 minors = lu.cfg.AllocateDRBDMinor(
6596 [primary_node, remote_node] * len(disk_info), instance_name)
6599 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6600 for i in range(disk_count)]):
6601 names.append(lv_prefix + "_data")
6602 names.append(lv_prefix + "_meta")
6603 for idx, disk in enumerate(disk_info):
6604 disk_index = idx + base_index
6605 vg = disk.get("vg", vgname)
6606 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6607 disk["size"], vg, names[idx*2:idx*2+2],
6608 "disk/%d" % disk_index,
6609 minors[idx*2], minors[idx*2+1])
6610 disk_dev.mode = disk["mode"]
6611 disks.append(disk_dev)
6612 elif template_name == constants.DT_FILE:
6613 if len(secondary_nodes) != 0:
6614 raise errors.ProgrammerError("Wrong template configuration")
6616 opcodes.RequireFileStorage()
6618 for idx, disk in enumerate(disk_info):
6619 disk_index = idx + base_index
6620 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6621 iv_name="disk/%d" % disk_index,
6622 logical_id=(file_driver,
6623 "%s/disk%d" % (file_storage_dir,
6626 disks.append(disk_dev)
6628 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6632 def _GetInstanceInfoText(instance):
6633 """Compute that text that should be added to the disk's metadata.
6636 return "originstname+%s" % instance.name
6639 def _CalcEta(time_taken, written, total_size):
6640 """Calculates the ETA based on size written and total size.
6642 @param time_taken: The time taken so far
6643 @param written: amount written so far
6644 @param total_size: The total size of data to be written
6645 @return: The remaining time in seconds
6648 avg_time = time_taken / float(written)
6649 return (total_size - written) * avg_time
6652 def _WipeDisks(lu, instance):
6653 """Wipes instance disks.
6655 @type lu: L{LogicalUnit}
6656 @param lu: the logical unit on whose behalf we execute
6657 @type instance: L{objects.Instance}
6658 @param instance: the instance whose disks we should create
6659 @return: the success of the wipe
6662 node = instance.primary_node
6663 logging.info("Pause sync of instance %s disks", instance.name)
6664 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
6666 for idx, success in enumerate(result.payload):
6668 logging.warn("pause-sync of instance %s for disks %d failed",
6672 for idx, device in enumerate(instance.disks):
6673 lu.LogInfo("* Wiping disk %d", idx)
6674 logging.info("Wiping disk %d for instance %s", idx, instance.name)
6676 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
6677 # MAX_WIPE_CHUNK at max
6678 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
6679 constants.MIN_WIPE_CHUNK_PERCENT)
6684 start_time = time.time()
6686 while offset < size:
6687 wipe_size = min(wipe_chunk_size, size - offset)
6688 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
6689 result.Raise("Could not wipe disk %d at offset %d for size %d" %
6690 (idx, offset, wipe_size))
6693 if now - last_output >= 60:
6694 eta = _CalcEta(now - start_time, offset, size)
6695 lu.LogInfo(" - done: %.1f%% ETA: %s" %
6696 (offset / float(size) * 100, utils.FormatSeconds(eta)))
6699 logging.info("Resume sync of instance %s disks", instance.name)
6701 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
6703 for idx, success in enumerate(result.payload):
6705 lu.LogWarning("Warning: Resume sync of disk %d failed. Please have a"
6706 " look at the status and troubleshoot the issue.", idx)
6707 logging.warn("resume-sync of instance %s for disks %d failed",
6711 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6712 """Create all disks for an instance.
6714 This abstracts away some work from AddInstance.
6716 @type lu: L{LogicalUnit}
6717 @param lu: the logical unit on whose behalf we execute
6718 @type instance: L{objects.Instance}
6719 @param instance: the instance whose disks we should create
6721 @param to_skip: list of indices to skip
6722 @type target_node: string
6723 @param target_node: if passed, overrides the target node for creation
6725 @return: the success of the creation
6728 info = _GetInstanceInfoText(instance)
6729 if target_node is None:
6730 pnode = instance.primary_node
6731 all_nodes = instance.all_nodes
6736 if instance.disk_template == constants.DT_FILE:
6737 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6738 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6740 result.Raise("Failed to create directory '%s' on"
6741 " node %s" % (file_storage_dir, pnode))
6743 # Note: this needs to be kept in sync with adding of disks in
6744 # LUInstanceSetParams
6745 for idx, device in enumerate(instance.disks):
6746 if to_skip and idx in to_skip:
6748 logging.info("Creating volume %s for instance %s",
6749 device.iv_name, instance.name)
6751 for node in all_nodes:
6752 f_create = node == pnode
6753 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6756 def _RemoveDisks(lu, instance, target_node=None):
6757 """Remove all disks for an instance.
6759 This abstracts away some work from `AddInstance()` and
6760 `RemoveInstance()`. Note that in case some of the devices couldn't
6761 be removed, the removal will continue with the other ones (compare
6762 with `_CreateDisks()`).
6764 @type lu: L{LogicalUnit}
6765 @param lu: the logical unit on whose behalf we execute
6766 @type instance: L{objects.Instance}
6767 @param instance: the instance whose disks we should remove
6768 @type target_node: string
6769 @param target_node: used to override the node on which to remove the disks
6771 @return: the success of the removal
6774 logging.info("Removing block devices for instance %s", instance.name)
6777 for device in instance.disks:
6779 edata = [(target_node, device)]
6781 edata = device.ComputeNodeTree(instance.primary_node)
6782 for node, disk in edata:
6783 lu.cfg.SetDiskID(disk, node)
6784 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6786 lu.LogWarning("Could not remove block device %s on node %s,"
6787 " continuing anyway: %s", device.iv_name, node, msg)
6790 if instance.disk_template == constants.DT_FILE:
6791 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6795 tgt = instance.primary_node
6796 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6798 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6799 file_storage_dir, instance.primary_node, result.fail_msg)
6805 def _ComputeDiskSizePerVG(disk_template, disks):
6806 """Compute disk size requirements in the volume group
6809 def _compute(disks, payload):
6810 """Universal algorithm
6815 vgs[disk["vg"]] = vgs.get("vg", 0) + disk["size"] + payload
6819 # Required free disk space as a function of disk and swap space
6821 constants.DT_DISKLESS: {},
6822 constants.DT_PLAIN: _compute(disks, 0),
6823 # 128 MB are added for drbd metadata for each disk
6824 constants.DT_DRBD8: _compute(disks, 128),
6825 constants.DT_FILE: {},
6828 if disk_template not in req_size_dict:
6829 raise errors.ProgrammerError("Disk template '%s' size requirement"
6830 " is unknown" % disk_template)
6832 return req_size_dict[disk_template]
6835 def _ComputeDiskSize(disk_template, disks):
6836 """Compute disk size requirements in the volume group
6839 # Required free disk space as a function of disk and swap space
6841 constants.DT_DISKLESS: None,
6842 constants.DT_PLAIN: sum(d["size"] for d in disks),
6843 # 128 MB are added for drbd metadata for each disk
6844 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6845 constants.DT_FILE: None,
6848 if disk_template not in req_size_dict:
6849 raise errors.ProgrammerError("Disk template '%s' size requirement"
6850 " is unknown" % disk_template)
6852 return req_size_dict[disk_template]
6855 def _FilterVmNodes(lu, nodenames):
6856 """Filters out non-vm_capable nodes from a list.
6858 @type lu: L{LogicalUnit}
6859 @param lu: the logical unit for which we check
6860 @type nodenames: list
6861 @param nodenames: the list of nodes on which we should check
6863 @return: the list of vm-capable nodes
6866 vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
6867 return [name for name in nodenames if name not in vm_nodes]
6870 def _CheckHVParams(lu, nodenames, hvname, hvparams):
6871 """Hypervisor parameter validation.
6873 This function abstract the hypervisor parameter validation to be
6874 used in both instance create and instance modify.
6876 @type lu: L{LogicalUnit}
6877 @param lu: the logical unit for which we check
6878 @type nodenames: list
6879 @param nodenames: the list of nodes on which we should check
6880 @type hvname: string
6881 @param hvname: the name of the hypervisor we should use
6882 @type hvparams: dict
6883 @param hvparams: the parameters which we need to check
6884 @raise errors.OpPrereqError: if the parameters are not valid
6887 nodenames = _FilterVmNodes(lu, nodenames)
6888 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6891 for node in nodenames:
6895 info.Raise("Hypervisor parameter validation failed on node %s" % node)
6898 def _CheckOSParams(lu, required, nodenames, osname, osparams):
6899 """OS parameters validation.
6901 @type lu: L{LogicalUnit}
6902 @param lu: the logical unit for which we check
6903 @type required: boolean
6904 @param required: whether the validation should fail if the OS is not
6906 @type nodenames: list
6907 @param nodenames: the list of nodes on which we should check
6908 @type osname: string
6909 @param osname: the name of the hypervisor we should use
6910 @type osparams: dict
6911 @param osparams: the parameters which we need to check
6912 @raise errors.OpPrereqError: if the parameters are not valid
6915 nodenames = _FilterVmNodes(lu, nodenames)
6916 result = lu.rpc.call_os_validate(required, nodenames, osname,
6917 [constants.OS_VALIDATE_PARAMETERS],
6919 for node, nres in result.items():
6920 # we don't check for offline cases since this should be run only
6921 # against the master node and/or an instance's nodes
6922 nres.Raise("OS Parameters validation failed on node %s" % node)
6923 if not nres.payload:
6924 lu.LogInfo("OS %s not found on node %s, validation skipped",
6928 class LUInstanceCreate(LogicalUnit):
6929 """Create an instance.
6932 HPATH = "instance-add"
6933 HTYPE = constants.HTYPE_INSTANCE
6936 def CheckArguments(self):
6940 # do not require name_check to ease forward/backward compatibility
6942 if self.op.no_install and self.op.start:
6943 self.LogInfo("No-installation mode selected, disabling startup")
6944 self.op.start = False
6945 # validate/normalize the instance name
6946 self.op.instance_name = \
6947 netutils.Hostname.GetNormalizedName(self.op.instance_name)
6949 if self.op.ip_check and not self.op.name_check:
6950 # TODO: make the ip check more flexible and not depend on the name check
6951 raise errors.OpPrereqError("Cannot do ip check without a name check",
6954 # check nics' parameter names
6955 for nic in self.op.nics:
6956 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6958 # check disks. parameter names and consistent adopt/no-adopt strategy
6959 has_adopt = has_no_adopt = False
6960 for disk in self.op.disks:
6961 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6966 if has_adopt and has_no_adopt:
6967 raise errors.OpPrereqError("Either all disks are adopted or none is",
6970 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6971 raise errors.OpPrereqError("Disk adoption is not supported for the"
6972 " '%s' disk template" %
6973 self.op.disk_template,
6975 if self.op.iallocator is not None:
6976 raise errors.OpPrereqError("Disk adoption not allowed with an"
6977 " iallocator script", errors.ECODE_INVAL)
6978 if self.op.mode == constants.INSTANCE_IMPORT:
6979 raise errors.OpPrereqError("Disk adoption not allowed for"
6980 " instance import", errors.ECODE_INVAL)
6982 self.adopt_disks = has_adopt
6984 # instance name verification
6985 if self.op.name_check:
6986 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
6987 self.op.instance_name = self.hostname1.name
6988 # used in CheckPrereq for ip ping check
6989 self.check_ip = self.hostname1.ip
6991 self.check_ip = None
6993 # file storage checks
6994 if (self.op.file_driver and
6995 not self.op.file_driver in constants.FILE_DRIVER):
6996 raise errors.OpPrereqError("Invalid file driver name '%s'" %
6997 self.op.file_driver, errors.ECODE_INVAL)
6999 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
7000 raise errors.OpPrereqError("File storage directory path not absolute",
7003 ### Node/iallocator related checks
7004 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
7006 if self.op.pnode is not None:
7007 if self.op.disk_template in constants.DTS_NET_MIRROR:
7008 if self.op.snode is None:
7009 raise errors.OpPrereqError("The networked disk templates need"
7010 " a mirror node", errors.ECODE_INVAL)
7012 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
7014 self.op.snode = None
7016 self._cds = _GetClusterDomainSecret()
7018 if self.op.mode == constants.INSTANCE_IMPORT:
7019 # On import force_variant must be True, because if we forced it at
7020 # initial install, our only chance when importing it back is that it
7022 self.op.force_variant = True
7024 if self.op.no_install:
7025 self.LogInfo("No-installation mode has no effect during import")
7027 elif self.op.mode == constants.INSTANCE_CREATE:
7028 if self.op.os_type is None:
7029 raise errors.OpPrereqError("No guest OS specified",
7031 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
7032 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
7033 " installation" % self.op.os_type,
7035 if self.op.disk_template is None:
7036 raise errors.OpPrereqError("No disk template specified",
7039 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7040 # Check handshake to ensure both clusters have the same domain secret
7041 src_handshake = self.op.source_handshake
7042 if not src_handshake:
7043 raise errors.OpPrereqError("Missing source handshake",
7046 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
7049 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
7052 # Load and check source CA
7053 self.source_x509_ca_pem = self.op.source_x509_ca
7054 if not self.source_x509_ca_pem:
7055 raise errors.OpPrereqError("Missing source X509 CA",
7059 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
7061 except OpenSSL.crypto.Error, err:
7062 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
7063 (err, ), errors.ECODE_INVAL)
7065 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
7066 if errcode is not None:
7067 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
7070 self.source_x509_ca = cert
7072 src_instance_name = self.op.source_instance_name
7073 if not src_instance_name:
7074 raise errors.OpPrereqError("Missing source instance name",
7077 self.source_instance_name = \
7078 netutils.GetHostname(name=src_instance_name).name
7081 raise errors.OpPrereqError("Invalid instance creation mode %r" %
7082 self.op.mode, errors.ECODE_INVAL)
7084 def ExpandNames(self):
7085 """ExpandNames for CreateInstance.
7087 Figure out the right locks for instance creation.
7090 self.needed_locks = {}
7092 instance_name = self.op.instance_name
7093 # this is just a preventive check, but someone might still add this
7094 # instance in the meantime, and creation will fail at lock-add time
7095 if instance_name in self.cfg.GetInstanceList():
7096 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7097 instance_name, errors.ECODE_EXISTS)
7099 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
7101 if self.op.iallocator:
7102 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7104 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
7105 nodelist = [self.op.pnode]
7106 if self.op.snode is not None:
7107 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
7108 nodelist.append(self.op.snode)
7109 self.needed_locks[locking.LEVEL_NODE] = nodelist
7111 # in case of import lock the source node too
7112 if self.op.mode == constants.INSTANCE_IMPORT:
7113 src_node = self.op.src_node
7114 src_path = self.op.src_path
7116 if src_path is None:
7117 self.op.src_path = src_path = self.op.instance_name
7119 if src_node is None:
7120 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7121 self.op.src_node = None
7122 if os.path.isabs(src_path):
7123 raise errors.OpPrereqError("Importing an instance from an absolute"
7124 " path requires a source node option.",
7127 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
7128 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
7129 self.needed_locks[locking.LEVEL_NODE].append(src_node)
7130 if not os.path.isabs(src_path):
7131 self.op.src_path = src_path = \
7132 utils.PathJoin(constants.EXPORT_DIR, src_path)
7134 def _RunAllocator(self):
7135 """Run the allocator based on input opcode.
7138 nics = [n.ToDict() for n in self.nics]
7139 ial = IAllocator(self.cfg, self.rpc,
7140 mode=constants.IALLOCATOR_MODE_ALLOC,
7141 name=self.op.instance_name,
7142 disk_template=self.op.disk_template,
7145 vcpus=self.be_full[constants.BE_VCPUS],
7146 mem_size=self.be_full[constants.BE_MEMORY],
7149 hypervisor=self.op.hypervisor,
7152 ial.Run(self.op.iallocator)
7155 raise errors.OpPrereqError("Can't compute nodes using"
7156 " iallocator '%s': %s" %
7157 (self.op.iallocator, ial.info),
7159 if len(ial.result) != ial.required_nodes:
7160 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7161 " of nodes (%s), required %s" %
7162 (self.op.iallocator, len(ial.result),
7163 ial.required_nodes), errors.ECODE_FAULT)
7164 self.op.pnode = ial.result[0]
7165 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7166 self.op.instance_name, self.op.iallocator,
7167 utils.CommaJoin(ial.result))
7168 if ial.required_nodes == 2:
7169 self.op.snode = ial.result[1]
7171 def BuildHooksEnv(self):
7174 This runs on master, primary and secondary nodes of the instance.
7178 "ADD_MODE": self.op.mode,
7180 if self.op.mode == constants.INSTANCE_IMPORT:
7181 env["SRC_NODE"] = self.op.src_node
7182 env["SRC_PATH"] = self.op.src_path
7183 env["SRC_IMAGES"] = self.src_images
7185 env.update(_BuildInstanceHookEnv(
7186 name=self.op.instance_name,
7187 primary_node=self.op.pnode,
7188 secondary_nodes=self.secondaries,
7189 status=self.op.start,
7190 os_type=self.op.os_type,
7191 memory=self.be_full[constants.BE_MEMORY],
7192 vcpus=self.be_full[constants.BE_VCPUS],
7193 nics=_NICListToTuple(self, self.nics),
7194 disk_template=self.op.disk_template,
7195 disks=[(d["size"], d["mode"]) for d in self.disks],
7198 hypervisor_name=self.op.hypervisor,
7201 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
7205 def _ReadExportInfo(self):
7206 """Reads the export information from disk.
7208 It will override the opcode source node and path with the actual
7209 information, if these two were not specified before.
7211 @return: the export information
7214 assert self.op.mode == constants.INSTANCE_IMPORT
7216 src_node = self.op.src_node
7217 src_path = self.op.src_path
7219 if src_node is None:
7220 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
7221 exp_list = self.rpc.call_export_list(locked_nodes)
7223 for node in exp_list:
7224 if exp_list[node].fail_msg:
7226 if src_path in exp_list[node].payload:
7228 self.op.src_node = src_node = node
7229 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
7233 raise errors.OpPrereqError("No export found for relative path %s" %
7234 src_path, errors.ECODE_INVAL)
7236 _CheckNodeOnline(self, src_node)
7237 result = self.rpc.call_export_info(src_node, src_path)
7238 result.Raise("No export or invalid export found in dir %s" % src_path)
7240 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
7241 if not export_info.has_section(constants.INISECT_EXP):
7242 raise errors.ProgrammerError("Corrupted export config",
7243 errors.ECODE_ENVIRON)
7245 ei_version = export_info.get(constants.INISECT_EXP, "version")
7246 if (int(ei_version) != constants.EXPORT_VERSION):
7247 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
7248 (ei_version, constants.EXPORT_VERSION),
7249 errors.ECODE_ENVIRON)
7252 def _ReadExportParams(self, einfo):
7253 """Use export parameters as defaults.
7255 In case the opcode doesn't specify (as in override) some instance
7256 parameters, then try to use them from the export information, if
7260 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
7262 if self.op.disk_template is None:
7263 if einfo.has_option(constants.INISECT_INS, "disk_template"):
7264 self.op.disk_template = einfo.get(constants.INISECT_INS,
7267 raise errors.OpPrereqError("No disk template specified and the export"
7268 " is missing the disk_template information",
7271 if not self.op.disks:
7272 if einfo.has_option(constants.INISECT_INS, "disk_count"):
7274 # TODO: import the disk iv_name too
7275 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
7276 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
7277 disks.append({"size": disk_sz})
7278 self.op.disks = disks
7280 raise errors.OpPrereqError("No disk info specified and the export"
7281 " is missing the disk information",
7284 if (not self.op.nics and
7285 einfo.has_option(constants.INISECT_INS, "nic_count")):
7287 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
7289 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
7290 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
7295 if (self.op.hypervisor is None and
7296 einfo.has_option(constants.INISECT_INS, "hypervisor")):
7297 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
7298 if einfo.has_section(constants.INISECT_HYP):
7299 # use the export parameters but do not override the ones
7300 # specified by the user
7301 for name, value in einfo.items(constants.INISECT_HYP):
7302 if name not in self.op.hvparams:
7303 self.op.hvparams[name] = value
7305 if einfo.has_section(constants.INISECT_BEP):
7306 # use the parameters, without overriding
7307 for name, value in einfo.items(constants.INISECT_BEP):
7308 if name not in self.op.beparams:
7309 self.op.beparams[name] = value
7311 # try to read the parameters old style, from the main section
7312 for name in constants.BES_PARAMETERS:
7313 if (name not in self.op.beparams and
7314 einfo.has_option(constants.INISECT_INS, name)):
7315 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
7317 if einfo.has_section(constants.INISECT_OSP):
7318 # use the parameters, without overriding
7319 for name, value in einfo.items(constants.INISECT_OSP):
7320 if name not in self.op.osparams:
7321 self.op.osparams[name] = value
7323 def _RevertToDefaults(self, cluster):
7324 """Revert the instance parameters to the default values.
7328 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7329 for name in self.op.hvparams.keys():
7330 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7331 del self.op.hvparams[name]
7333 be_defs = cluster.SimpleFillBE({})
7334 for name in self.op.beparams.keys():
7335 if name in be_defs and be_defs[name] == self.op.beparams[name]:
7336 del self.op.beparams[name]
7338 nic_defs = cluster.SimpleFillNIC({})
7339 for nic in self.op.nics:
7340 for name in constants.NICS_PARAMETERS:
7341 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7344 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7345 for name in self.op.osparams.keys():
7346 if name in os_defs and os_defs[name] == self.op.osparams[name]:
7347 del self.op.osparams[name]
7349 def CheckPrereq(self):
7350 """Check prerequisites.
7353 if self.op.mode == constants.INSTANCE_IMPORT:
7354 export_info = self._ReadExportInfo()
7355 self._ReadExportParams(export_info)
7357 if (not self.cfg.GetVGName() and
7358 self.op.disk_template not in constants.DTS_NOT_LVM):
7359 raise errors.OpPrereqError("Cluster does not support lvm-based"
7360 " instances", errors.ECODE_STATE)
7362 if self.op.hypervisor is None:
7363 self.op.hypervisor = self.cfg.GetHypervisorType()
7365 cluster = self.cfg.GetClusterInfo()
7366 enabled_hvs = cluster.enabled_hypervisors
7367 if self.op.hypervisor not in enabled_hvs:
7368 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7369 " cluster (%s)" % (self.op.hypervisor,
7370 ",".join(enabled_hvs)),
7373 # check hypervisor parameter syntax (locally)
7374 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7375 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7377 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7378 hv_type.CheckParameterSyntax(filled_hvp)
7379 self.hv_full = filled_hvp
7380 # check that we don't specify global parameters on an instance
7381 _CheckGlobalHvParams(self.op.hvparams)
7383 # fill and remember the beparams dict
7384 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7385 self.be_full = cluster.SimpleFillBE(self.op.beparams)
7387 # build os parameters
7388 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7390 # now that hvp/bep are in final format, let's reset to defaults,
7392 if self.op.identify_defaults:
7393 self._RevertToDefaults(cluster)
7397 for idx, nic in enumerate(self.op.nics):
7398 nic_mode_req = nic.get("mode", None)
7399 nic_mode = nic_mode_req
7400 if nic_mode is None:
7401 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7403 # in routed mode, for the first nic, the default ip is 'auto'
7404 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7405 default_ip_mode = constants.VALUE_AUTO
7407 default_ip_mode = constants.VALUE_NONE
7409 # ip validity checks
7410 ip = nic.get("ip", default_ip_mode)
7411 if ip is None or ip.lower() == constants.VALUE_NONE:
7413 elif ip.lower() == constants.VALUE_AUTO:
7414 if not self.op.name_check:
7415 raise errors.OpPrereqError("IP address set to auto but name checks"
7416 " have been skipped",
7418 nic_ip = self.hostname1.ip
7420 if not netutils.IPAddress.IsValid(ip):
7421 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7425 # TODO: check the ip address for uniqueness
7426 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7427 raise errors.OpPrereqError("Routed nic mode requires an ip address",
7430 # MAC address verification
7431 mac = nic.get("mac", constants.VALUE_AUTO)
7432 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7433 mac = utils.NormalizeAndValidateMac(mac)
7436 self.cfg.ReserveMAC(mac, self.proc.GetECId())
7437 except errors.ReservationError:
7438 raise errors.OpPrereqError("MAC address %s already in use"
7439 " in cluster" % mac,
7440 errors.ECODE_NOTUNIQUE)
7442 # bridge verification
7443 bridge = nic.get("bridge", None)
7444 link = nic.get("link", None)
7446 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7447 " at the same time", errors.ECODE_INVAL)
7448 elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7449 raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7456 nicparams[constants.NIC_MODE] = nic_mode_req
7458 nicparams[constants.NIC_LINK] = link
7460 check_params = cluster.SimpleFillNIC(nicparams)
7461 objects.NIC.CheckParameterSyntax(check_params)
7462 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7464 # disk checks/pre-build
7466 for disk in self.op.disks:
7467 mode = disk.get("mode", constants.DISK_RDWR)
7468 if mode not in constants.DISK_ACCESS_SET:
7469 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7470 mode, errors.ECODE_INVAL)
7471 size = disk.get("size", None)
7473 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7476 except (TypeError, ValueError):
7477 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7479 vg = disk.get("vg", self.cfg.GetVGName())
7480 new_disk = {"size": size, "mode": mode, "vg": vg}
7482 new_disk["adopt"] = disk["adopt"]
7483 self.disks.append(new_disk)
7485 if self.op.mode == constants.INSTANCE_IMPORT:
7487 # Check that the new instance doesn't have less disks than the export
7488 instance_disks = len(self.disks)
7489 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7490 if instance_disks < export_disks:
7491 raise errors.OpPrereqError("Not enough disks to import."
7492 " (instance: %d, export: %d)" %
7493 (instance_disks, export_disks),
7497 for idx in range(export_disks):
7498 option = 'disk%d_dump' % idx
7499 if export_info.has_option(constants.INISECT_INS, option):
7500 # FIXME: are the old os-es, disk sizes, etc. useful?
7501 export_name = export_info.get(constants.INISECT_INS, option)
7502 image = utils.PathJoin(self.op.src_path, export_name)
7503 disk_images.append(image)
7505 disk_images.append(False)
7507 self.src_images = disk_images
7509 old_name = export_info.get(constants.INISECT_INS, 'name')
7511 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7512 except (TypeError, ValueError), err:
7513 raise errors.OpPrereqError("Invalid export file, nic_count is not"
7514 " an integer: %s" % str(err),
7516 if self.op.instance_name == old_name:
7517 for idx, nic in enumerate(self.nics):
7518 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7519 nic_mac_ini = 'nic%d_mac' % idx
7520 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7522 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7524 # ip ping checks (we use the same ip that was resolved in ExpandNames)
7525 if self.op.ip_check:
7526 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7527 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7528 (self.check_ip, self.op.instance_name),
7529 errors.ECODE_NOTUNIQUE)
7531 #### mac address generation
7532 # By generating here the mac address both the allocator and the hooks get
7533 # the real final mac address rather than the 'auto' or 'generate' value.
7534 # There is a race condition between the generation and the instance object
7535 # creation, which means that we know the mac is valid now, but we're not
7536 # sure it will be when we actually add the instance. If things go bad
7537 # adding the instance will abort because of a duplicate mac, and the
7538 # creation job will fail.
7539 for nic in self.nics:
7540 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7541 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7545 if self.op.iallocator is not None:
7546 self._RunAllocator()
7548 #### node related checks
7550 # check primary node
7551 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7552 assert self.pnode is not None, \
7553 "Cannot retrieve locked node %s" % self.op.pnode
7555 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7556 pnode.name, errors.ECODE_STATE)
7558 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7559 pnode.name, errors.ECODE_STATE)
7560 if not pnode.vm_capable:
7561 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
7562 " '%s'" % pnode.name, errors.ECODE_STATE)
7564 self.secondaries = []
7566 # mirror node verification
7567 if self.op.disk_template in constants.DTS_NET_MIRROR:
7568 if self.op.snode == pnode.name:
7569 raise errors.OpPrereqError("The secondary node cannot be the"
7570 " primary node.", errors.ECODE_INVAL)
7571 _CheckNodeOnline(self, self.op.snode)
7572 _CheckNodeNotDrained(self, self.op.snode)
7573 _CheckNodeVmCapable(self, self.op.snode)
7574 self.secondaries.append(self.op.snode)
7576 nodenames = [pnode.name] + self.secondaries
7578 if not self.adopt_disks:
7579 # Check lv size requirements, if not adopting
7580 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
7581 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
7583 else: # instead, we must check the adoption data
7584 all_lvs = set([i["vg"] + "/" + i["adopt"] for i in self.disks])
7585 if len(all_lvs) != len(self.disks):
7586 raise errors.OpPrereqError("Duplicate volume names given for adoption",
7588 for lv_name in all_lvs:
7590 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
7591 # to ReserveLV uses the same syntax
7592 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7593 except errors.ReservationError:
7594 raise errors.OpPrereqError("LV named %s used by another instance" %
7595 lv_name, errors.ECODE_NOTUNIQUE)
7597 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
7598 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
7600 node_lvs = self.rpc.call_lv_list([pnode.name],
7601 vg_names.payload.keys())[pnode.name]
7602 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7603 node_lvs = node_lvs.payload
7605 delta = all_lvs.difference(node_lvs.keys())
7607 raise errors.OpPrereqError("Missing logical volume(s): %s" %
7608 utils.CommaJoin(delta),
7610 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7612 raise errors.OpPrereqError("Online logical volumes found, cannot"
7613 " adopt: %s" % utils.CommaJoin(online_lvs),
7615 # update the size of disk based on what is found
7616 for dsk in self.disks:
7617 dsk["size"] = int(float(node_lvs[dsk["vg"] + "/" + dsk["adopt"]][0]))
7619 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7621 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7622 # check OS parameters (remotely)
7623 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7625 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7627 # memory check on primary node
7629 _CheckNodeFreeMemory(self, self.pnode.name,
7630 "creating instance %s" % self.op.instance_name,
7631 self.be_full[constants.BE_MEMORY],
7634 self.dry_run_result = list(nodenames)
7636 def Exec(self, feedback_fn):
7637 """Create and add the instance to the cluster.
7640 instance = self.op.instance_name
7641 pnode_name = self.pnode.name
7643 ht_kind = self.op.hypervisor
7644 if ht_kind in constants.HTS_REQ_PORT:
7645 network_port = self.cfg.AllocatePort()
7649 if constants.ENABLE_FILE_STORAGE:
7650 # this is needed because os.path.join does not accept None arguments
7651 if self.op.file_storage_dir is None:
7652 string_file_storage_dir = ""
7654 string_file_storage_dir = self.op.file_storage_dir
7656 # build the full file storage dir path
7657 file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7658 string_file_storage_dir, instance)
7660 file_storage_dir = ""
7662 disks = _GenerateDiskTemplate(self,
7663 self.op.disk_template,
7664 instance, pnode_name,
7668 self.op.file_driver,
7672 iobj = objects.Instance(name=instance, os=self.op.os_type,
7673 primary_node=pnode_name,
7674 nics=self.nics, disks=disks,
7675 disk_template=self.op.disk_template,
7677 network_port=network_port,
7678 beparams=self.op.beparams,
7679 hvparams=self.op.hvparams,
7680 hypervisor=self.op.hypervisor,
7681 osparams=self.op.osparams,
7684 if self.adopt_disks:
7685 # rename LVs to the newly-generated names; we need to construct
7686 # 'fake' LV disks with the old data, plus the new unique_id
7687 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7689 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7690 rename_to.append(t_dsk.logical_id)
7691 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7692 self.cfg.SetDiskID(t_dsk, pnode_name)
7693 result = self.rpc.call_blockdev_rename(pnode_name,
7694 zip(tmp_disks, rename_to))
7695 result.Raise("Failed to rename adoped LVs")
7697 feedback_fn("* creating instance disks...")
7699 _CreateDisks(self, iobj)
7700 except errors.OpExecError:
7701 self.LogWarning("Device creation failed, reverting...")
7703 _RemoveDisks(self, iobj)
7705 self.cfg.ReleaseDRBDMinors(instance)
7708 if self.cfg.GetClusterInfo().prealloc_wipe_disks:
7709 feedback_fn("* wiping instance disks...")
7711 _WipeDisks(self, iobj)
7712 except errors.OpExecError:
7713 self.LogWarning("Device wiping failed, reverting...")
7715 _RemoveDisks(self, iobj)
7717 self.cfg.ReleaseDRBDMinors(instance)
7720 feedback_fn("adding instance %s to cluster config" % instance)
7722 self.cfg.AddInstance(iobj, self.proc.GetECId())
7724 # Declare that we don't want to remove the instance lock anymore, as we've
7725 # added the instance to the config
7726 del self.remove_locks[locking.LEVEL_INSTANCE]
7727 # Unlock all the nodes
7728 if self.op.mode == constants.INSTANCE_IMPORT:
7729 nodes_keep = [self.op.src_node]
7730 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7731 if node != self.op.src_node]
7732 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7733 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7735 self.context.glm.release(locking.LEVEL_NODE)
7736 del self.acquired_locks[locking.LEVEL_NODE]
7738 if self.op.wait_for_sync:
7739 disk_abort = not _WaitForSync(self, iobj)
7740 elif iobj.disk_template in constants.DTS_NET_MIRROR:
7741 # make sure the disks are not degraded (still sync-ing is ok)
7743 feedback_fn("* checking mirrors status")
7744 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7749 _RemoveDisks(self, iobj)
7750 self.cfg.RemoveInstance(iobj.name)
7751 # Make sure the instance lock gets removed
7752 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7753 raise errors.OpExecError("There are some degraded disks for"
7756 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7757 if self.op.mode == constants.INSTANCE_CREATE:
7758 if not self.op.no_install:
7759 feedback_fn("* running the instance OS create scripts...")
7760 # FIXME: pass debug option from opcode to backend
7761 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7762 self.op.debug_level)
7763 result.Raise("Could not add os for instance %s"
7764 " on node %s" % (instance, pnode_name))
7766 elif self.op.mode == constants.INSTANCE_IMPORT:
7767 feedback_fn("* running the instance OS import scripts...")
7771 for idx, image in enumerate(self.src_images):
7775 # FIXME: pass debug option from opcode to backend
7776 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7777 constants.IEIO_FILE, (image, ),
7778 constants.IEIO_SCRIPT,
7779 (iobj.disks[idx], idx),
7781 transfers.append(dt)
7784 masterd.instance.TransferInstanceData(self, feedback_fn,
7785 self.op.src_node, pnode_name,
7786 self.pnode.secondary_ip,
7788 if not compat.all(import_result):
7789 self.LogWarning("Some disks for instance %s on node %s were not"
7790 " imported successfully" % (instance, pnode_name))
7792 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7793 feedback_fn("* preparing remote import...")
7794 # The source cluster will stop the instance before attempting to make a
7795 # connection. In some cases stopping an instance can take a long time,
7796 # hence the shutdown timeout is added to the connection timeout.
7797 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
7798 self.op.source_shutdown_timeout)
7799 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7801 assert iobj.primary_node == self.pnode.name
7803 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
7804 self.source_x509_ca,
7805 self._cds, timeouts)
7806 if not compat.all(disk_results):
7807 # TODO: Should the instance still be started, even if some disks
7808 # failed to import (valid for local imports, too)?
7809 self.LogWarning("Some disks for instance %s on node %s were not"
7810 " imported successfully" % (instance, pnode_name))
7812 # Run rename script on newly imported instance
7813 assert iobj.name == instance
7814 feedback_fn("Running rename script for %s" % instance)
7815 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7816 self.source_instance_name,
7817 self.op.debug_level)
7819 self.LogWarning("Failed to run rename script for %s on node"
7820 " %s: %s" % (instance, pnode_name, result.fail_msg))
7823 # also checked in the prereq part
7824 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7828 iobj.admin_up = True
7829 self.cfg.Update(iobj, feedback_fn)
7830 logging.info("Starting instance %s on node %s", instance, pnode_name)
7831 feedback_fn("* starting instance...")
7832 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7833 result.Raise("Could not start instance")
7835 return list(iobj.all_nodes)
7838 class LUInstanceConsole(NoHooksLU):
7839 """Connect to an instance's console.
7841 This is somewhat special in that it returns the command line that
7842 you need to run on the master node in order to connect to the
7848 def ExpandNames(self):
7849 self._ExpandAndLockInstance()
7851 def CheckPrereq(self):
7852 """Check prerequisites.
7854 This checks that the instance is in the cluster.
7857 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7858 assert self.instance is not None, \
7859 "Cannot retrieve locked instance %s" % self.op.instance_name
7860 _CheckNodeOnline(self, self.instance.primary_node)
7862 def Exec(self, feedback_fn):
7863 """Connect to the console of an instance
7866 instance = self.instance
7867 node = instance.primary_node
7869 node_insts = self.rpc.call_instance_list([node],
7870 [instance.hypervisor])[node]
7871 node_insts.Raise("Can't get node information from %s" % node)
7873 if instance.name not in node_insts.payload:
7874 if instance.admin_up:
7875 state = "ERROR_down"
7877 state = "ADMIN_down"
7878 raise errors.OpExecError("Instance %s is not running (state %s)" %
7879 (instance.name, state))
7881 logging.debug("Connecting to console of %s on %s", instance.name, node)
7883 return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
7886 def _GetInstanceConsole(cluster, instance):
7887 """Returns console information for an instance.
7889 @type cluster: L{objects.Cluster}
7890 @type instance: L{objects.Instance}
7894 hyper = hypervisor.GetHypervisor(instance.hypervisor)
7895 # beparams and hvparams are passed separately, to avoid editing the
7896 # instance and then saving the defaults in the instance itself.
7897 hvparams = cluster.FillHV(instance)
7898 beparams = cluster.FillBE(instance)
7899 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
7901 assert console.instance == instance.name
7902 assert console.Validate()
7904 return console.ToDict()
7907 class LUInstanceReplaceDisks(LogicalUnit):
7908 """Replace the disks of an instance.
7911 HPATH = "mirrors-replace"
7912 HTYPE = constants.HTYPE_INSTANCE
7915 def CheckArguments(self):
7916 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7919 def ExpandNames(self):
7920 self._ExpandAndLockInstance()
7922 if self.op.iallocator is not None:
7923 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7925 elif self.op.remote_node is not None:
7926 remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7927 self.op.remote_node = remote_node
7929 # Warning: do not remove the locking of the new secondary here
7930 # unless DRBD8.AddChildren is changed to work in parallel;
7931 # currently it doesn't since parallel invocations of
7932 # FindUnusedMinor will conflict
7933 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7934 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7937 self.needed_locks[locking.LEVEL_NODE] = []
7938 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7940 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7941 self.op.iallocator, self.op.remote_node,
7942 self.op.disks, False, self.op.early_release)
7944 self.tasklets = [self.replacer]
7946 def DeclareLocks(self, level):
7947 # If we're not already locking all nodes in the set we have to declare the
7948 # instance's primary/secondary nodes.
7949 if (level == locking.LEVEL_NODE and
7950 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7951 self._LockInstancesNodes()
7953 def BuildHooksEnv(self):
7956 This runs on the master, the primary and all the secondaries.
7959 instance = self.replacer.instance
7961 "MODE": self.op.mode,
7962 "NEW_SECONDARY": self.op.remote_node,
7963 "OLD_SECONDARY": instance.secondary_nodes[0],
7965 env.update(_BuildInstanceHookEnvByObject(self, instance))
7967 self.cfg.GetMasterNode(),
7968 instance.primary_node,
7970 if self.op.remote_node is not None:
7971 nl.append(self.op.remote_node)
7975 class TLReplaceDisks(Tasklet):
7976 """Replaces disks for an instance.
7978 Note: Locking is not within the scope of this class.
7981 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7982 disks, delay_iallocator, early_release):
7983 """Initializes this class.
7986 Tasklet.__init__(self, lu)
7989 self.instance_name = instance_name
7991 self.iallocator_name = iallocator_name
7992 self.remote_node = remote_node
7994 self.delay_iallocator = delay_iallocator
7995 self.early_release = early_release
7998 self.instance = None
7999 self.new_node = None
8000 self.target_node = None
8001 self.other_node = None
8002 self.remote_node_info = None
8003 self.node_secondary_ip = None
8006 def CheckArguments(mode, remote_node, iallocator):
8007 """Helper function for users of this class.
8010 # check for valid parameter combination
8011 if mode == constants.REPLACE_DISK_CHG:
8012 if remote_node is None and iallocator is None:
8013 raise errors.OpPrereqError("When changing the secondary either an"
8014 " iallocator script must be used or the"
8015 " new node given", errors.ECODE_INVAL)
8017 if remote_node is not None and iallocator is not None:
8018 raise errors.OpPrereqError("Give either the iallocator or the new"
8019 " secondary, not both", errors.ECODE_INVAL)
8021 elif remote_node is not None or iallocator is not None:
8022 # Not replacing the secondary
8023 raise errors.OpPrereqError("The iallocator and new node options can"
8024 " only be used when changing the"
8025 " secondary node", errors.ECODE_INVAL)
8028 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
8029 """Compute a new secondary node using an IAllocator.
8032 ial = IAllocator(lu.cfg, lu.rpc,
8033 mode=constants.IALLOCATOR_MODE_RELOC,
8035 relocate_from=relocate_from)
8037 ial.Run(iallocator_name)
8040 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
8041 " %s" % (iallocator_name, ial.info),
8044 if len(ial.result) != ial.required_nodes:
8045 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8046 " of nodes (%s), required %s" %
8048 len(ial.result), ial.required_nodes),
8051 remote_node_name = ial.result[0]
8053 lu.LogInfo("Selected new secondary for instance '%s': %s",
8054 instance_name, remote_node_name)
8056 return remote_node_name
8058 def _FindFaultyDisks(self, node_name):
8059 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
8062 def CheckPrereq(self):
8063 """Check prerequisites.
8065 This checks that the instance is in the cluster.
8068 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
8069 assert instance is not None, \
8070 "Cannot retrieve locked instance %s" % self.instance_name
8072 if instance.disk_template != constants.DT_DRBD8:
8073 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
8074 " instances", errors.ECODE_INVAL)
8076 if len(instance.secondary_nodes) != 1:
8077 raise errors.OpPrereqError("The instance has a strange layout,"
8078 " expected one secondary but found %d" %
8079 len(instance.secondary_nodes),
8082 if not self.delay_iallocator:
8083 self._CheckPrereq2()
8085 def _CheckPrereq2(self):
8086 """Check prerequisites, second part.
8088 This function should always be part of CheckPrereq. It was separated and is
8089 now called from Exec because during node evacuation iallocator was only
8090 called with an unmodified cluster model, not taking planned changes into
8094 instance = self.instance
8095 secondary_node = instance.secondary_nodes[0]
8097 if self.iallocator_name is None:
8098 remote_node = self.remote_node
8100 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
8101 instance.name, instance.secondary_nodes)
8103 if remote_node is not None:
8104 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
8105 assert self.remote_node_info is not None, \
8106 "Cannot retrieve locked node %s" % remote_node
8108 self.remote_node_info = None
8110 if remote_node == self.instance.primary_node:
8111 raise errors.OpPrereqError("The specified node is the primary node of"
8112 " the instance.", errors.ECODE_INVAL)
8114 if remote_node == secondary_node:
8115 raise errors.OpPrereqError("The specified node is already the"
8116 " secondary node of the instance.",
8119 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
8120 constants.REPLACE_DISK_CHG):
8121 raise errors.OpPrereqError("Cannot specify disks to be replaced",
8124 if self.mode == constants.REPLACE_DISK_AUTO:
8125 faulty_primary = self._FindFaultyDisks(instance.primary_node)
8126 faulty_secondary = self._FindFaultyDisks(secondary_node)
8128 if faulty_primary and faulty_secondary:
8129 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
8130 " one node and can not be repaired"
8131 " automatically" % self.instance_name,
8135 self.disks = faulty_primary
8136 self.target_node = instance.primary_node
8137 self.other_node = secondary_node
8138 check_nodes = [self.target_node, self.other_node]
8139 elif faulty_secondary:
8140 self.disks = faulty_secondary
8141 self.target_node = secondary_node
8142 self.other_node = instance.primary_node
8143 check_nodes = [self.target_node, self.other_node]
8149 # Non-automatic modes
8150 if self.mode == constants.REPLACE_DISK_PRI:
8151 self.target_node = instance.primary_node
8152 self.other_node = secondary_node
8153 check_nodes = [self.target_node, self.other_node]
8155 elif self.mode == constants.REPLACE_DISK_SEC:
8156 self.target_node = secondary_node
8157 self.other_node = instance.primary_node
8158 check_nodes = [self.target_node, self.other_node]
8160 elif self.mode == constants.REPLACE_DISK_CHG:
8161 self.new_node = remote_node
8162 self.other_node = instance.primary_node
8163 self.target_node = secondary_node
8164 check_nodes = [self.new_node, self.other_node]
8166 _CheckNodeNotDrained(self.lu, remote_node)
8167 _CheckNodeVmCapable(self.lu, remote_node)
8169 old_node_info = self.cfg.GetNodeInfo(secondary_node)
8170 assert old_node_info is not None
8171 if old_node_info.offline and not self.early_release:
8172 # doesn't make sense to delay the release
8173 self.early_release = True
8174 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
8175 " early-release mode", secondary_node)
8178 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
8181 # If not specified all disks should be replaced
8183 self.disks = range(len(self.instance.disks))
8185 for node in check_nodes:
8186 _CheckNodeOnline(self.lu, node)
8188 # Check whether disks are valid
8189 for disk_idx in self.disks:
8190 instance.FindDisk(disk_idx)
8192 # Get secondary node IP addresses
8195 for node_name in [self.target_node, self.other_node, self.new_node]:
8196 if node_name is not None:
8197 node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
8199 self.node_secondary_ip = node_2nd_ip
8201 def Exec(self, feedback_fn):
8202 """Execute disk replacement.
8204 This dispatches the disk replacement to the appropriate handler.
8207 if self.delay_iallocator:
8208 self._CheckPrereq2()
8211 feedback_fn("No disks need replacement")
8214 feedback_fn("Replacing disk(s) %s for %s" %
8215 (utils.CommaJoin(self.disks), self.instance.name))
8217 activate_disks = (not self.instance.admin_up)
8219 # Activate the instance disks if we're replacing them on a down instance
8221 _StartInstanceDisks(self.lu, self.instance, True)
8224 # Should we replace the secondary node?
8225 if self.new_node is not None:
8226 fn = self._ExecDrbd8Secondary
8228 fn = self._ExecDrbd8DiskOnly
8230 return fn(feedback_fn)
8233 # Deactivate the instance disks if we're replacing them on a
8236 _SafeShutdownInstanceDisks(self.lu, self.instance)
8238 def _CheckVolumeGroup(self, nodes):
8239 self.lu.LogInfo("Checking volume groups")
8241 vgname = self.cfg.GetVGName()
8243 # Make sure volume group exists on all involved nodes
8244 results = self.rpc.call_vg_list(nodes)
8246 raise errors.OpExecError("Can't list volume groups on the nodes")
8250 res.Raise("Error checking node %s" % node)
8251 if vgname not in res.payload:
8252 raise errors.OpExecError("Volume group '%s' not found on node %s" %
8255 def _CheckDisksExistence(self, nodes):
8256 # Check disk existence
8257 for idx, dev in enumerate(self.instance.disks):
8258 if idx not in self.disks:
8262 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
8263 self.cfg.SetDiskID(dev, node)
8265 result = self.rpc.call_blockdev_find(node, dev)
8267 msg = result.fail_msg
8268 if msg or not result.payload:
8270 msg = "disk not found"
8271 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
8274 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
8275 for idx, dev in enumerate(self.instance.disks):
8276 if idx not in self.disks:
8279 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
8282 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
8284 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
8285 " replace disks for instance %s" %
8286 (node_name, self.instance.name))
8288 def _CreateNewStorage(self, node_name):
8289 vgname = self.cfg.GetVGName()
8292 for idx, dev in enumerate(self.instance.disks):
8293 if idx not in self.disks:
8296 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
8298 self.cfg.SetDiskID(dev, node_name)
8300 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
8301 names = _GenerateUniqueNames(self.lu, lv_names)
8303 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
8304 logical_id=(vgname, names[0]))
8305 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
8306 logical_id=(vgname, names[1]))
8308 new_lvs = [lv_data, lv_meta]
8309 old_lvs = dev.children
8310 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
8312 # we pass force_create=True to force the LVM creation
8313 for new_lv in new_lvs:
8314 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
8315 _GetInstanceInfoText(self.instance), False)
8319 def _CheckDevices(self, node_name, iv_names):
8320 for name, (dev, _, _) in iv_names.iteritems():
8321 self.cfg.SetDiskID(dev, node_name)
8323 result = self.rpc.call_blockdev_find(node_name, dev)
8325 msg = result.fail_msg
8326 if msg or not result.payload:
8328 msg = "disk not found"
8329 raise errors.OpExecError("Can't find DRBD device %s: %s" %
8332 if result.payload.is_degraded:
8333 raise errors.OpExecError("DRBD device %s is degraded!" % name)
8335 def _RemoveOldStorage(self, node_name, iv_names):
8336 for name, (_, old_lvs, _) in iv_names.iteritems():
8337 self.lu.LogInfo("Remove logical volumes for %s" % name)
8340 self.cfg.SetDiskID(lv, node_name)
8342 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
8344 self.lu.LogWarning("Can't remove old LV: %s" % msg,
8345 hint="remove unused LVs manually")
8347 def _ReleaseNodeLock(self, node_name):
8348 """Releases the lock for a given node."""
8349 self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
8351 def _ExecDrbd8DiskOnly(self, feedback_fn):
8352 """Replace a disk on the primary or secondary for DRBD 8.
8354 The algorithm for replace is quite complicated:
8356 1. for each disk to be replaced:
8358 1. create new LVs on the target node with unique names
8359 1. detach old LVs from the drbd device
8360 1. rename old LVs to name_replaced.<time_t>
8361 1. rename new LVs to old LVs
8362 1. attach the new LVs (with the old names now) to the drbd device
8364 1. wait for sync across all devices
8366 1. for each modified disk:
8368 1. remove old LVs (which have the name name_replaces.<time_t>)
8370 Failures are not very well handled.
8375 # Step: check device activation
8376 self.lu.LogStep(1, steps_total, "Check device existence")
8377 self._CheckDisksExistence([self.other_node, self.target_node])
8378 self._CheckVolumeGroup([self.target_node, self.other_node])
8380 # Step: check other node consistency
8381 self.lu.LogStep(2, steps_total, "Check peer consistency")
8382 self._CheckDisksConsistency(self.other_node,
8383 self.other_node == self.instance.primary_node,
8386 # Step: create new storage
8387 self.lu.LogStep(3, steps_total, "Allocate new storage")
8388 iv_names = self._CreateNewStorage(self.target_node)
8390 # Step: for each lv, detach+rename*2+attach
8391 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8392 for dev, old_lvs, new_lvs in iv_names.itervalues():
8393 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8395 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8397 result.Raise("Can't detach drbd from local storage on node"
8398 " %s for device %s" % (self.target_node, dev.iv_name))
8400 #cfg.Update(instance)
8402 # ok, we created the new LVs, so now we know we have the needed
8403 # storage; as such, we proceed on the target node to rename
8404 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8405 # using the assumption that logical_id == physical_id (which in
8406 # turn is the unique_id on that node)
8408 # FIXME(iustin): use a better name for the replaced LVs
8409 temp_suffix = int(time.time())
8410 ren_fn = lambda d, suff: (d.physical_id[0],
8411 d.physical_id[1] + "_replaced-%s" % suff)
8413 # Build the rename list based on what LVs exist on the node
8414 rename_old_to_new = []
8415 for to_ren in old_lvs:
8416 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8417 if not result.fail_msg and result.payload:
8419 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8421 self.lu.LogInfo("Renaming the old LVs on the target node")
8422 result = self.rpc.call_blockdev_rename(self.target_node,
8424 result.Raise("Can't rename old LVs on node %s" % self.target_node)
8426 # Now we rename the new LVs to the old LVs
8427 self.lu.LogInfo("Renaming the new LVs on the target node")
8428 rename_new_to_old = [(new, old.physical_id)
8429 for old, new in zip(old_lvs, new_lvs)]
8430 result = self.rpc.call_blockdev_rename(self.target_node,
8432 result.Raise("Can't rename new LVs on node %s" % self.target_node)
8434 for old, new in zip(old_lvs, new_lvs):
8435 new.logical_id = old.logical_id
8436 self.cfg.SetDiskID(new, self.target_node)
8438 for disk in old_lvs:
8439 disk.logical_id = ren_fn(disk, temp_suffix)
8440 self.cfg.SetDiskID(disk, self.target_node)
8442 # Now that the new lvs have the old name, we can add them to the device
8443 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8444 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8446 msg = result.fail_msg
8448 for new_lv in new_lvs:
8449 msg2 = self.rpc.call_blockdev_remove(self.target_node,
8452 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8453 hint=("cleanup manually the unused logical"
8455 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8457 dev.children = new_lvs
8459 self.cfg.Update(self.instance, feedback_fn)
8462 if self.early_release:
8463 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8465 self._RemoveOldStorage(self.target_node, iv_names)
8466 # WARNING: we release both node locks here, do not do other RPCs
8467 # than WaitForSync to the primary node
8468 self._ReleaseNodeLock([self.target_node, self.other_node])
8471 # This can fail as the old devices are degraded and _WaitForSync
8472 # does a combined result over all disks, so we don't check its return value
8473 self.lu.LogStep(cstep, steps_total, "Sync devices")
8475 _WaitForSync(self.lu, self.instance)
8477 # Check all devices manually
8478 self._CheckDevices(self.instance.primary_node, iv_names)
8480 # Step: remove old storage
8481 if not self.early_release:
8482 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8484 self._RemoveOldStorage(self.target_node, iv_names)
8486 def _ExecDrbd8Secondary(self, feedback_fn):
8487 """Replace the secondary node for DRBD 8.
8489 The algorithm for replace is quite complicated:
8490 - for all disks of the instance:
8491 - create new LVs on the new node with same names
8492 - shutdown the drbd device on the old secondary
8493 - disconnect the drbd network on the primary
8494 - create the drbd device on the new secondary
8495 - network attach the drbd on the primary, using an artifice:
8496 the drbd code for Attach() will connect to the network if it
8497 finds a device which is connected to the good local disks but
8499 - wait for sync across all devices
8500 - remove all disks from the old secondary
8502 Failures are not very well handled.
8507 # Step: check device activation
8508 self.lu.LogStep(1, steps_total, "Check device existence")
8509 self._CheckDisksExistence([self.instance.primary_node])
8510 self._CheckVolumeGroup([self.instance.primary_node])
8512 # Step: check other node consistency
8513 self.lu.LogStep(2, steps_total, "Check peer consistency")
8514 self._CheckDisksConsistency(self.instance.primary_node, True, True)
8516 # Step: create new storage
8517 self.lu.LogStep(3, steps_total, "Allocate new storage")
8518 for idx, dev in enumerate(self.instance.disks):
8519 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8520 (self.new_node, idx))
8521 # we pass force_create=True to force LVM creation
8522 for new_lv in dev.children:
8523 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8524 _GetInstanceInfoText(self.instance), False)
8526 # Step 4: dbrd minors and drbd setups changes
8527 # after this, we must manually remove the drbd minors on both the
8528 # error and the success paths
8529 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8530 minors = self.cfg.AllocateDRBDMinor([self.new_node
8531 for dev in self.instance.disks],
8533 logging.debug("Allocated minors %r", minors)
8536 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8537 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8538 (self.new_node, idx))
8539 # create new devices on new_node; note that we create two IDs:
8540 # one without port, so the drbd will be activated without
8541 # networking information on the new node at this stage, and one
8542 # with network, for the latter activation in step 4
8543 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8544 if self.instance.primary_node == o_node1:
8547 assert self.instance.primary_node == o_node2, "Three-node instance?"
8550 new_alone_id = (self.instance.primary_node, self.new_node, None,
8551 p_minor, new_minor, o_secret)
8552 new_net_id = (self.instance.primary_node, self.new_node, o_port,
8553 p_minor, new_minor, o_secret)
8555 iv_names[idx] = (dev, dev.children, new_net_id)
8556 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8558 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8559 logical_id=new_alone_id,
8560 children=dev.children,
8563 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8564 _GetInstanceInfoText(self.instance), False)
8565 except errors.GenericError:
8566 self.cfg.ReleaseDRBDMinors(self.instance.name)
8569 # We have new devices, shutdown the drbd on the old secondary
8570 for idx, dev in enumerate(self.instance.disks):
8571 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8572 self.cfg.SetDiskID(dev, self.target_node)
8573 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8575 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8576 "node: %s" % (idx, msg),
8577 hint=("Please cleanup this device manually as"
8578 " soon as possible"))
8580 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8581 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8582 self.node_secondary_ip,
8583 self.instance.disks)\
8584 [self.instance.primary_node]
8586 msg = result.fail_msg
8588 # detaches didn't succeed (unlikely)
8589 self.cfg.ReleaseDRBDMinors(self.instance.name)
8590 raise errors.OpExecError("Can't detach the disks from the network on"
8591 " old node: %s" % (msg,))
8593 # if we managed to detach at least one, we update all the disks of
8594 # the instance to point to the new secondary
8595 self.lu.LogInfo("Updating instance configuration")
8596 for dev, _, new_logical_id in iv_names.itervalues():
8597 dev.logical_id = new_logical_id
8598 self.cfg.SetDiskID(dev, self.instance.primary_node)
8600 self.cfg.Update(self.instance, feedback_fn)
8602 # and now perform the drbd attach
8603 self.lu.LogInfo("Attaching primary drbds to new secondary"
8604 " (standalone => connected)")
8605 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8607 self.node_secondary_ip,
8608 self.instance.disks,
8611 for to_node, to_result in result.items():
8612 msg = to_result.fail_msg
8614 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8616 hint=("please do a gnt-instance info to see the"
8617 " status of disks"))
8619 if self.early_release:
8620 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8622 self._RemoveOldStorage(self.target_node, iv_names)
8623 # WARNING: we release all node locks here, do not do other RPCs
8624 # than WaitForSync to the primary node
8625 self._ReleaseNodeLock([self.instance.primary_node,
8630 # This can fail as the old devices are degraded and _WaitForSync
8631 # does a combined result over all disks, so we don't check its return value
8632 self.lu.LogStep(cstep, steps_total, "Sync devices")
8634 _WaitForSync(self.lu, self.instance)
8636 # Check all devices manually
8637 self._CheckDevices(self.instance.primary_node, iv_names)
8639 # Step: remove old storage
8640 if not self.early_release:
8641 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8642 self._RemoveOldStorage(self.target_node, iv_names)
8645 class LURepairNodeStorage(NoHooksLU):
8646 """Repairs the volume group on a node.
8651 def CheckArguments(self):
8652 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8654 storage_type = self.op.storage_type
8656 if (constants.SO_FIX_CONSISTENCY not in
8657 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8658 raise errors.OpPrereqError("Storage units of type '%s' can not be"
8659 " repaired" % storage_type,
8662 def ExpandNames(self):
8663 self.needed_locks = {
8664 locking.LEVEL_NODE: [self.op.node_name],
8667 def _CheckFaultyDisks(self, instance, node_name):
8668 """Ensure faulty disks abort the opcode or at least warn."""
8670 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8672 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8673 " node '%s'" % (instance.name, node_name),
8675 except errors.OpPrereqError, err:
8676 if self.op.ignore_consistency:
8677 self.proc.LogWarning(str(err.args[0]))
8681 def CheckPrereq(self):
8682 """Check prerequisites.
8685 # Check whether any instance on this node has faulty disks
8686 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8687 if not inst.admin_up:
8689 check_nodes = set(inst.all_nodes)
8690 check_nodes.discard(self.op.node_name)
8691 for inst_node_name in check_nodes:
8692 self._CheckFaultyDisks(inst, inst_node_name)
8694 def Exec(self, feedback_fn):
8695 feedback_fn("Repairing storage unit '%s' on %s ..." %
8696 (self.op.name, self.op.node_name))
8698 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8699 result = self.rpc.call_storage_execute(self.op.node_name,
8700 self.op.storage_type, st_args,
8702 constants.SO_FIX_CONSISTENCY)
8703 result.Raise("Failed to repair storage unit '%s' on %s" %
8704 (self.op.name, self.op.node_name))
8707 class LUNodeEvacStrategy(NoHooksLU):
8708 """Computes the node evacuation strategy.
8713 def CheckArguments(self):
8714 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8716 def ExpandNames(self):
8717 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8718 self.needed_locks = locks = {}
8719 if self.op.remote_node is None:
8720 locks[locking.LEVEL_NODE] = locking.ALL_SET
8722 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8723 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8725 def Exec(self, feedback_fn):
8726 if self.op.remote_node is not None:
8728 for node in self.op.nodes:
8729 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8732 if i.primary_node == self.op.remote_node:
8733 raise errors.OpPrereqError("Node %s is the primary node of"
8734 " instance %s, cannot use it as"
8736 (self.op.remote_node, i.name),
8738 result.append([i.name, self.op.remote_node])
8740 ial = IAllocator(self.cfg, self.rpc,
8741 mode=constants.IALLOCATOR_MODE_MEVAC,
8742 evac_nodes=self.op.nodes)
8743 ial.Run(self.op.iallocator, validate=True)
8745 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8751 class LUInstanceGrowDisk(LogicalUnit):
8752 """Grow a disk of an instance.
8756 HTYPE = constants.HTYPE_INSTANCE
8759 def ExpandNames(self):
8760 self._ExpandAndLockInstance()
8761 self.needed_locks[locking.LEVEL_NODE] = []
8762 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8764 def DeclareLocks(self, level):
8765 if level == locking.LEVEL_NODE:
8766 self._LockInstancesNodes()
8768 def BuildHooksEnv(self):
8771 This runs on the master, the primary and all the secondaries.
8775 "DISK": self.op.disk,
8776 "AMOUNT": self.op.amount,
8778 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8779 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8782 def CheckPrereq(self):
8783 """Check prerequisites.
8785 This checks that the instance is in the cluster.
8788 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8789 assert instance is not None, \
8790 "Cannot retrieve locked instance %s" % self.op.instance_name
8791 nodenames = list(instance.all_nodes)
8792 for node in nodenames:
8793 _CheckNodeOnline(self, node)
8795 self.instance = instance
8797 if instance.disk_template not in constants.DTS_GROWABLE:
8798 raise errors.OpPrereqError("Instance's disk layout does not support"
8799 " growing.", errors.ECODE_INVAL)
8801 self.disk = instance.FindDisk(self.op.disk)
8803 if instance.disk_template != constants.DT_FILE:
8804 # TODO: check the free disk space for file, when that feature
8806 _CheckNodesFreeDiskPerVG(self, nodenames,
8807 self.disk.ComputeGrowth(self.op.amount))
8809 def Exec(self, feedback_fn):
8810 """Execute disk grow.
8813 instance = self.instance
8816 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8818 raise errors.OpExecError("Cannot activate block device to grow")
8820 for node in instance.all_nodes:
8821 self.cfg.SetDiskID(disk, node)
8822 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8823 result.Raise("Grow request failed to node %s" % node)
8825 # TODO: Rewrite code to work properly
8826 # DRBD goes into sync mode for a short amount of time after executing the
8827 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8828 # calling "resize" in sync mode fails. Sleeping for a short amount of
8829 # time is a work-around.
8832 disk.RecordGrow(self.op.amount)
8833 self.cfg.Update(instance, feedback_fn)
8834 if self.op.wait_for_sync:
8835 disk_abort = not _WaitForSync(self, instance, disks=[disk])
8837 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8838 " status.\nPlease check the instance.")
8839 if not instance.admin_up:
8840 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8841 elif not instance.admin_up:
8842 self.proc.LogWarning("Not shutting down the disk even if the instance is"
8843 " not supposed to be running because no wait for"
8844 " sync mode was requested.")
8847 class LUInstanceQueryData(NoHooksLU):
8848 """Query runtime instance data.
8853 def ExpandNames(self):
8854 self.needed_locks = {}
8855 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8857 if self.op.instances:
8858 self.wanted_names = []
8859 for name in self.op.instances:
8860 full_name = _ExpandInstanceName(self.cfg, name)
8861 self.wanted_names.append(full_name)
8862 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8864 self.wanted_names = None
8865 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8867 self.needed_locks[locking.LEVEL_NODE] = []
8868 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8870 def DeclareLocks(self, level):
8871 if level == locking.LEVEL_NODE:
8872 self._LockInstancesNodes()
8874 def CheckPrereq(self):
8875 """Check prerequisites.
8877 This only checks the optional instance list against the existing names.
8880 if self.wanted_names is None:
8881 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8883 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8884 in self.wanted_names]
8886 def _ComputeBlockdevStatus(self, node, instance_name, dev):
8887 """Returns the status of a block device
8890 if self.op.static or not node:
8893 self.cfg.SetDiskID(dev, node)
8895 result = self.rpc.call_blockdev_find(node, dev)
8899 result.Raise("Can't compute disk status for %s" % instance_name)
8901 status = result.payload
8905 return (status.dev_path, status.major, status.minor,
8906 status.sync_percent, status.estimated_time,
8907 status.is_degraded, status.ldisk_status)
8909 def _ComputeDiskStatus(self, instance, snode, dev):
8910 """Compute block device status.
8913 if dev.dev_type in constants.LDS_DRBD:
8914 # we change the snode then (otherwise we use the one passed in)
8915 if dev.logical_id[0] == instance.primary_node:
8916 snode = dev.logical_id[1]
8918 snode = dev.logical_id[0]
8920 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8922 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8925 dev_children = [self._ComputeDiskStatus(instance, snode, child)
8926 for child in dev.children]
8931 "iv_name": dev.iv_name,
8932 "dev_type": dev.dev_type,
8933 "logical_id": dev.logical_id,
8934 "physical_id": dev.physical_id,
8935 "pstatus": dev_pstatus,
8936 "sstatus": dev_sstatus,
8937 "children": dev_children,
8944 def Exec(self, feedback_fn):
8945 """Gather and return data"""
8948 cluster = self.cfg.GetClusterInfo()
8950 for instance in self.wanted_instances:
8951 if not self.op.static:
8952 remote_info = self.rpc.call_instance_info(instance.primary_node,
8954 instance.hypervisor)
8955 remote_info.Raise("Error checking node %s" % instance.primary_node)
8956 remote_info = remote_info.payload
8957 if remote_info and "state" in remote_info:
8960 remote_state = "down"
8963 if instance.admin_up:
8966 config_state = "down"
8968 disks = [self._ComputeDiskStatus(instance, None, device)
8969 for device in instance.disks]
8972 "name": instance.name,
8973 "config_state": config_state,
8974 "run_state": remote_state,
8975 "pnode": instance.primary_node,
8976 "snodes": instance.secondary_nodes,
8978 # this happens to be the same format used for hooks
8979 "nics": _NICListToTuple(self, instance.nics),
8980 "disk_template": instance.disk_template,
8982 "hypervisor": instance.hypervisor,
8983 "network_port": instance.network_port,
8984 "hv_instance": instance.hvparams,
8985 "hv_actual": cluster.FillHV(instance, skip_globals=True),
8986 "be_instance": instance.beparams,
8987 "be_actual": cluster.FillBE(instance),
8988 "os_instance": instance.osparams,
8989 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8990 "serial_no": instance.serial_no,
8991 "mtime": instance.mtime,
8992 "ctime": instance.ctime,
8993 "uuid": instance.uuid,
8996 result[instance.name] = idict
9001 class LUInstanceSetParams(LogicalUnit):
9002 """Modifies an instances's parameters.
9005 HPATH = "instance-modify"
9006 HTYPE = constants.HTYPE_INSTANCE
9009 def CheckArguments(self):
9010 if not (self.op.nics or self.op.disks or self.op.disk_template or
9011 self.op.hvparams or self.op.beparams or self.op.os_name):
9012 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
9014 if self.op.hvparams:
9015 _CheckGlobalHvParams(self.op.hvparams)
9019 for disk_op, disk_dict in self.op.disks:
9020 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
9021 if disk_op == constants.DDM_REMOVE:
9024 elif disk_op == constants.DDM_ADD:
9027 if not isinstance(disk_op, int):
9028 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
9029 if not isinstance(disk_dict, dict):
9030 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
9031 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9033 if disk_op == constants.DDM_ADD:
9034 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
9035 if mode not in constants.DISK_ACCESS_SET:
9036 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
9038 size = disk_dict.get('size', None)
9040 raise errors.OpPrereqError("Required disk parameter size missing",
9044 except (TypeError, ValueError), err:
9045 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
9046 str(err), errors.ECODE_INVAL)
9047 disk_dict['size'] = size
9049 # modification of disk
9050 if 'size' in disk_dict:
9051 raise errors.OpPrereqError("Disk size change not possible, use"
9052 " grow-disk", errors.ECODE_INVAL)
9054 if disk_addremove > 1:
9055 raise errors.OpPrereqError("Only one disk add or remove operation"
9056 " supported at a time", errors.ECODE_INVAL)
9058 if self.op.disks and self.op.disk_template is not None:
9059 raise errors.OpPrereqError("Disk template conversion and other disk"
9060 " changes not supported at the same time",
9063 if (self.op.disk_template and
9064 self.op.disk_template in constants.DTS_NET_MIRROR and
9065 self.op.remote_node is None):
9066 raise errors.OpPrereqError("Changing the disk template to a mirrored"
9067 " one requires specifying a secondary node",
9072 for nic_op, nic_dict in self.op.nics:
9073 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
9074 if nic_op == constants.DDM_REMOVE:
9077 elif nic_op == constants.DDM_ADD:
9080 if not isinstance(nic_op, int):
9081 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
9082 if not isinstance(nic_dict, dict):
9083 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
9084 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9086 # nic_dict should be a dict
9087 nic_ip = nic_dict.get('ip', None)
9088 if nic_ip is not None:
9089 if nic_ip.lower() == constants.VALUE_NONE:
9090 nic_dict['ip'] = None
9092 if not netutils.IPAddress.IsValid(nic_ip):
9093 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
9096 nic_bridge = nic_dict.get('bridge', None)
9097 nic_link = nic_dict.get('link', None)
9098 if nic_bridge and nic_link:
9099 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
9100 " at the same time", errors.ECODE_INVAL)
9101 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
9102 nic_dict['bridge'] = None
9103 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
9104 nic_dict['link'] = None
9106 if nic_op == constants.DDM_ADD:
9107 nic_mac = nic_dict.get('mac', None)
9109 nic_dict['mac'] = constants.VALUE_AUTO
9111 if 'mac' in nic_dict:
9112 nic_mac = nic_dict['mac']
9113 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9114 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
9116 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
9117 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
9118 " modifying an existing nic",
9121 if nic_addremove > 1:
9122 raise errors.OpPrereqError("Only one NIC add or remove operation"
9123 " supported at a time", errors.ECODE_INVAL)
9125 def ExpandNames(self):
9126 self._ExpandAndLockInstance()
9127 self.needed_locks[locking.LEVEL_NODE] = []
9128 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9130 def DeclareLocks(self, level):
9131 if level == locking.LEVEL_NODE:
9132 self._LockInstancesNodes()
9133 if self.op.disk_template and self.op.remote_node:
9134 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9135 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
9137 def BuildHooksEnv(self):
9140 This runs on the master, primary and secondaries.
9144 if constants.BE_MEMORY in self.be_new:
9145 args['memory'] = self.be_new[constants.BE_MEMORY]
9146 if constants.BE_VCPUS in self.be_new:
9147 args['vcpus'] = self.be_new[constants.BE_VCPUS]
9148 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
9149 # information at all.
9152 nic_override = dict(self.op.nics)
9153 for idx, nic in enumerate(self.instance.nics):
9154 if idx in nic_override:
9155 this_nic_override = nic_override[idx]
9157 this_nic_override = {}
9158 if 'ip' in this_nic_override:
9159 ip = this_nic_override['ip']
9162 if 'mac' in this_nic_override:
9163 mac = this_nic_override['mac']
9166 if idx in self.nic_pnew:
9167 nicparams = self.nic_pnew[idx]
9169 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
9170 mode = nicparams[constants.NIC_MODE]
9171 link = nicparams[constants.NIC_LINK]
9172 args['nics'].append((ip, mac, mode, link))
9173 if constants.DDM_ADD in nic_override:
9174 ip = nic_override[constants.DDM_ADD].get('ip', None)
9175 mac = nic_override[constants.DDM_ADD]['mac']
9176 nicparams = self.nic_pnew[constants.DDM_ADD]
9177 mode = nicparams[constants.NIC_MODE]
9178 link = nicparams[constants.NIC_LINK]
9179 args['nics'].append((ip, mac, mode, link))
9180 elif constants.DDM_REMOVE in nic_override:
9181 del args['nics'][-1]
9183 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
9184 if self.op.disk_template:
9185 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
9186 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9189 def CheckPrereq(self):
9190 """Check prerequisites.
9192 This only checks the instance list against the existing names.
9195 # checking the new params on the primary/secondary nodes
9197 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9198 cluster = self.cluster = self.cfg.GetClusterInfo()
9199 assert self.instance is not None, \
9200 "Cannot retrieve locked instance %s" % self.op.instance_name
9201 pnode = instance.primary_node
9202 nodelist = list(instance.all_nodes)
9205 if self.op.os_name and not self.op.force:
9206 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
9207 self.op.force_variant)
9208 instance_os = self.op.os_name
9210 instance_os = instance.os
9212 if self.op.disk_template:
9213 if instance.disk_template == self.op.disk_template:
9214 raise errors.OpPrereqError("Instance already has disk template %s" %
9215 instance.disk_template, errors.ECODE_INVAL)
9217 if (instance.disk_template,
9218 self.op.disk_template) not in self._DISK_CONVERSIONS:
9219 raise errors.OpPrereqError("Unsupported disk template conversion from"
9220 " %s to %s" % (instance.disk_template,
9221 self.op.disk_template),
9223 _CheckInstanceDown(self, instance, "cannot change disk template")
9224 if self.op.disk_template in constants.DTS_NET_MIRROR:
9225 if self.op.remote_node == pnode:
9226 raise errors.OpPrereqError("Given new secondary node %s is the same"
9227 " as the primary node of the instance" %
9228 self.op.remote_node, errors.ECODE_STATE)
9229 _CheckNodeOnline(self, self.op.remote_node)
9230 _CheckNodeNotDrained(self, self.op.remote_node)
9231 # FIXME: here we assume that the old instance type is DT_PLAIN
9232 assert instance.disk_template == constants.DT_PLAIN
9233 disks = [{"size": d.size, "vg": d.logical_id[0]}
9234 for d in instance.disks]
9235 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
9236 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
9238 # hvparams processing
9239 if self.op.hvparams:
9240 hv_type = instance.hypervisor
9241 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
9242 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
9243 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
9246 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
9247 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
9248 self.hv_new = hv_new # the new actual values
9249 self.hv_inst = i_hvdict # the new dict (without defaults)
9251 self.hv_new = self.hv_inst = {}
9253 # beparams processing
9254 if self.op.beparams:
9255 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
9257 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
9258 be_new = cluster.SimpleFillBE(i_bedict)
9259 self.be_new = be_new # the new actual values
9260 self.be_inst = i_bedict # the new dict (without defaults)
9262 self.be_new = self.be_inst = {}
9264 # osparams processing
9265 if self.op.osparams:
9266 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
9267 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
9268 self.os_inst = i_osdict # the new dict (without defaults)
9274 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
9275 mem_check_list = [pnode]
9276 if be_new[constants.BE_AUTO_BALANCE]:
9277 # either we changed auto_balance to yes or it was from before
9278 mem_check_list.extend(instance.secondary_nodes)
9279 instance_info = self.rpc.call_instance_info(pnode, instance.name,
9280 instance.hypervisor)
9281 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
9282 instance.hypervisor)
9283 pninfo = nodeinfo[pnode]
9284 msg = pninfo.fail_msg
9286 # Assume the primary node is unreachable and go ahead
9287 self.warn.append("Can't get info from primary node %s: %s" %
9289 elif not isinstance(pninfo.payload.get('memory_free', None), int):
9290 self.warn.append("Node data from primary node %s doesn't contain"
9291 " free memory information" % pnode)
9292 elif instance_info.fail_msg:
9293 self.warn.append("Can't get instance runtime information: %s" %
9294 instance_info.fail_msg)
9296 if instance_info.payload:
9297 current_mem = int(instance_info.payload['memory'])
9299 # Assume instance not running
9300 # (there is a slight race condition here, but it's not very probable,
9301 # and we have no other way to check)
9303 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
9304 pninfo.payload['memory_free'])
9306 raise errors.OpPrereqError("This change will prevent the instance"
9307 " from starting, due to %d MB of memory"
9308 " missing on its primary node" % miss_mem,
9311 if be_new[constants.BE_AUTO_BALANCE]:
9312 for node, nres in nodeinfo.items():
9313 if node not in instance.secondary_nodes:
9317 self.warn.append("Can't get info from secondary node %s: %s" %
9319 elif not isinstance(nres.payload.get('memory_free', None), int):
9320 self.warn.append("Secondary node %s didn't return free"
9321 " memory information" % node)
9322 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
9323 self.warn.append("Not enough memory to failover instance to"
9324 " secondary node %s" % node)
9329 for nic_op, nic_dict in self.op.nics:
9330 if nic_op == constants.DDM_REMOVE:
9331 if not instance.nics:
9332 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
9335 if nic_op != constants.DDM_ADD:
9337 if not instance.nics:
9338 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
9339 " no NICs" % nic_op,
9341 if nic_op < 0 or nic_op >= len(instance.nics):
9342 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
9344 (nic_op, len(instance.nics) - 1),
9346 old_nic_params = instance.nics[nic_op].nicparams
9347 old_nic_ip = instance.nics[nic_op].ip
9352 update_params_dict = dict([(key, nic_dict[key])
9353 for key in constants.NICS_PARAMETERS
9354 if key in nic_dict])
9356 if 'bridge' in nic_dict:
9357 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
9359 new_nic_params = _GetUpdatedParams(old_nic_params,
9361 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9362 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9363 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9364 self.nic_pinst[nic_op] = new_nic_params
9365 self.nic_pnew[nic_op] = new_filled_nic_params
9366 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9368 if new_nic_mode == constants.NIC_MODE_BRIDGED:
9369 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9370 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9372 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9374 self.warn.append(msg)
9376 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9377 if new_nic_mode == constants.NIC_MODE_ROUTED:
9378 if 'ip' in nic_dict:
9379 nic_ip = nic_dict['ip']
9383 raise errors.OpPrereqError('Cannot set the nic ip to None'
9384 ' on a routed nic', errors.ECODE_INVAL)
9385 if 'mac' in nic_dict:
9386 nic_mac = nic_dict['mac']
9388 raise errors.OpPrereqError('Cannot set the nic mac to None',
9390 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9391 # otherwise generate the mac
9392 nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9394 # or validate/reserve the current one
9396 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9397 except errors.ReservationError:
9398 raise errors.OpPrereqError("MAC address %s already in use"
9399 " in cluster" % nic_mac,
9400 errors.ECODE_NOTUNIQUE)
9403 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9404 raise errors.OpPrereqError("Disk operations not supported for"
9405 " diskless instances",
9407 for disk_op, _ in self.op.disks:
9408 if disk_op == constants.DDM_REMOVE:
9409 if len(instance.disks) == 1:
9410 raise errors.OpPrereqError("Cannot remove the last disk of"
9411 " an instance", errors.ECODE_INVAL)
9412 _CheckInstanceDown(self, instance, "cannot remove disks")
9414 if (disk_op == constants.DDM_ADD and
9415 len(instance.disks) >= constants.MAX_DISKS):
9416 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9417 " add more" % constants.MAX_DISKS,
9419 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9421 if disk_op < 0 or disk_op >= len(instance.disks):
9422 raise errors.OpPrereqError("Invalid disk index %s, valid values"
9424 (disk_op, len(instance.disks)),
9429 def _ConvertPlainToDrbd(self, feedback_fn):
9430 """Converts an instance from plain to drbd.
9433 feedback_fn("Converting template to drbd")
9434 instance = self.instance
9435 pnode = instance.primary_node
9436 snode = self.op.remote_node
9438 # create a fake disk info for _GenerateDiskTemplate
9439 disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9440 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9441 instance.name, pnode, [snode],
9442 disk_info, None, None, 0, feedback_fn)
9443 info = _GetInstanceInfoText(instance)
9444 feedback_fn("Creating aditional volumes...")
9445 # first, create the missing data and meta devices
9446 for disk in new_disks:
9447 # unfortunately this is... not too nice
9448 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9450 for child in disk.children:
9451 _CreateSingleBlockDev(self, snode, instance, child, info, True)
9452 # at this stage, all new LVs have been created, we can rename the
9454 feedback_fn("Renaming original volumes...")
9455 rename_list = [(o, n.children[0].logical_id)
9456 for (o, n) in zip(instance.disks, new_disks)]
9457 result = self.rpc.call_blockdev_rename(pnode, rename_list)
9458 result.Raise("Failed to rename original LVs")
9460 feedback_fn("Initializing DRBD devices...")
9461 # all child devices are in place, we can now create the DRBD devices
9462 for disk in new_disks:
9463 for node in [pnode, snode]:
9464 f_create = node == pnode
9465 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9467 # at this point, the instance has been modified
9468 instance.disk_template = constants.DT_DRBD8
9469 instance.disks = new_disks
9470 self.cfg.Update(instance, feedback_fn)
9472 # disks are created, waiting for sync
9473 disk_abort = not _WaitForSync(self, instance)
9475 raise errors.OpExecError("There are some degraded disks for"
9476 " this instance, please cleanup manually")
9478 def _ConvertDrbdToPlain(self, feedback_fn):
9479 """Converts an instance from drbd to plain.
9482 instance = self.instance
9483 assert len(instance.secondary_nodes) == 1
9484 pnode = instance.primary_node
9485 snode = instance.secondary_nodes[0]
9486 feedback_fn("Converting template to plain")
9488 old_disks = instance.disks
9489 new_disks = [d.children[0] for d in old_disks]
9491 # copy over size and mode
9492 for parent, child in zip(old_disks, new_disks):
9493 child.size = parent.size
9494 child.mode = parent.mode
9496 # update instance structure
9497 instance.disks = new_disks
9498 instance.disk_template = constants.DT_PLAIN
9499 self.cfg.Update(instance, feedback_fn)
9501 feedback_fn("Removing volumes on the secondary node...")
9502 for disk in old_disks:
9503 self.cfg.SetDiskID(disk, snode)
9504 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9506 self.LogWarning("Could not remove block device %s on node %s,"
9507 " continuing anyway: %s", disk.iv_name, snode, msg)
9509 feedback_fn("Removing unneeded volumes on the primary node...")
9510 for idx, disk in enumerate(old_disks):
9511 meta = disk.children[1]
9512 self.cfg.SetDiskID(meta, pnode)
9513 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9515 self.LogWarning("Could not remove metadata for disk %d on node %s,"
9516 " continuing anyway: %s", idx, pnode, msg)
9518 def Exec(self, feedback_fn):
9519 """Modifies an instance.
9521 All parameters take effect only at the next restart of the instance.
9524 # Process here the warnings from CheckPrereq, as we don't have a
9525 # feedback_fn there.
9526 for warn in self.warn:
9527 feedback_fn("WARNING: %s" % warn)
9530 instance = self.instance
9532 for disk_op, disk_dict in self.op.disks:
9533 if disk_op == constants.DDM_REMOVE:
9534 # remove the last disk
9535 device = instance.disks.pop()
9536 device_idx = len(instance.disks)
9537 for node, disk in device.ComputeNodeTree(instance.primary_node):
9538 self.cfg.SetDiskID(disk, node)
9539 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9541 self.LogWarning("Could not remove disk/%d on node %s: %s,"
9542 " continuing anyway", device_idx, node, msg)
9543 result.append(("disk/%d" % device_idx, "remove"))
9544 elif disk_op == constants.DDM_ADD:
9546 if instance.disk_template == constants.DT_FILE:
9547 file_driver, file_path = instance.disks[0].logical_id
9548 file_path = os.path.dirname(file_path)
9550 file_driver = file_path = None
9551 disk_idx_base = len(instance.disks)
9552 new_disk = _GenerateDiskTemplate(self,
9553 instance.disk_template,
9554 instance.name, instance.primary_node,
9555 instance.secondary_nodes,
9559 disk_idx_base, feedback_fn)[0]
9560 instance.disks.append(new_disk)
9561 info = _GetInstanceInfoText(instance)
9563 logging.info("Creating volume %s for instance %s",
9564 new_disk.iv_name, instance.name)
9565 # Note: this needs to be kept in sync with _CreateDisks
9567 for node in instance.all_nodes:
9568 f_create = node == instance.primary_node
9570 _CreateBlockDev(self, node, instance, new_disk,
9571 f_create, info, f_create)
9572 except errors.OpExecError, err:
9573 self.LogWarning("Failed to create volume %s (%s) on"
9575 new_disk.iv_name, new_disk, node, err)
9576 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9577 (new_disk.size, new_disk.mode)))
9579 # change a given disk
9580 instance.disks[disk_op].mode = disk_dict['mode']
9581 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9583 if self.op.disk_template:
9584 r_shut = _ShutdownInstanceDisks(self, instance)
9586 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
9587 " proceed with disk template conversion")
9588 mode = (instance.disk_template, self.op.disk_template)
9590 self._DISK_CONVERSIONS[mode](self, feedback_fn)
9592 self.cfg.ReleaseDRBDMinors(instance.name)
9594 result.append(("disk_template", self.op.disk_template))
9597 for nic_op, nic_dict in self.op.nics:
9598 if nic_op == constants.DDM_REMOVE:
9599 # remove the last nic
9600 del instance.nics[-1]
9601 result.append(("nic.%d" % len(instance.nics), "remove"))
9602 elif nic_op == constants.DDM_ADD:
9603 # mac and bridge should be set, by now
9604 mac = nic_dict['mac']
9605 ip = nic_dict.get('ip', None)
9606 nicparams = self.nic_pinst[constants.DDM_ADD]
9607 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9608 instance.nics.append(new_nic)
9609 result.append(("nic.%d" % (len(instance.nics) - 1),
9610 "add:mac=%s,ip=%s,mode=%s,link=%s" %
9611 (new_nic.mac, new_nic.ip,
9612 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9613 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9616 for key in 'mac', 'ip':
9618 setattr(instance.nics[nic_op], key, nic_dict[key])
9619 if nic_op in self.nic_pinst:
9620 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9621 for key, val in nic_dict.iteritems():
9622 result.append(("nic.%s/%d" % (key, nic_op), val))
9625 if self.op.hvparams:
9626 instance.hvparams = self.hv_inst
9627 for key, val in self.op.hvparams.iteritems():
9628 result.append(("hv/%s" % key, val))
9631 if self.op.beparams:
9632 instance.beparams = self.be_inst
9633 for key, val in self.op.beparams.iteritems():
9634 result.append(("be/%s" % key, val))
9638 instance.os = self.op.os_name
9641 if self.op.osparams:
9642 instance.osparams = self.os_inst
9643 for key, val in self.op.osparams.iteritems():
9644 result.append(("os/%s" % key, val))
9646 self.cfg.Update(instance, feedback_fn)
9650 _DISK_CONVERSIONS = {
9651 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9652 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9656 class LUBackupQuery(NoHooksLU):
9657 """Query the exports list
9662 def ExpandNames(self):
9663 self.needed_locks = {}
9664 self.share_locks[locking.LEVEL_NODE] = 1
9665 if not self.op.nodes:
9666 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9668 self.needed_locks[locking.LEVEL_NODE] = \
9669 _GetWantedNodes(self, self.op.nodes)
9671 def Exec(self, feedback_fn):
9672 """Compute the list of all the exported system images.
9675 @return: a dictionary with the structure node->(export-list)
9676 where export-list is a list of the instances exported on
9680 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9681 rpcresult = self.rpc.call_export_list(self.nodes)
9683 for node in rpcresult:
9684 if rpcresult[node].fail_msg:
9685 result[node] = False
9687 result[node] = rpcresult[node].payload
9692 class LUBackupPrepare(NoHooksLU):
9693 """Prepares an instance for an export and returns useful information.
9698 def ExpandNames(self):
9699 self._ExpandAndLockInstance()
9701 def CheckPrereq(self):
9702 """Check prerequisites.
9705 instance_name = self.op.instance_name
9707 self.instance = self.cfg.GetInstanceInfo(instance_name)
9708 assert self.instance is not None, \
9709 "Cannot retrieve locked instance %s" % self.op.instance_name
9710 _CheckNodeOnline(self, self.instance.primary_node)
9712 self._cds = _GetClusterDomainSecret()
9714 def Exec(self, feedback_fn):
9715 """Prepares an instance for an export.
9718 instance = self.instance
9720 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9721 salt = utils.GenerateSecret(8)
9723 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9724 result = self.rpc.call_x509_cert_create(instance.primary_node,
9725 constants.RIE_CERT_VALIDITY)
9726 result.Raise("Can't create X509 key and certificate on %s" % result.node)
9728 (name, cert_pem) = result.payload
9730 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9734 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9735 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9737 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9743 class LUBackupExport(LogicalUnit):
9744 """Export an instance to an image in the cluster.
9747 HPATH = "instance-export"
9748 HTYPE = constants.HTYPE_INSTANCE
9751 def CheckArguments(self):
9752 """Check the arguments.
9755 self.x509_key_name = self.op.x509_key_name
9756 self.dest_x509_ca_pem = self.op.destination_x509_ca
9758 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9759 if not self.x509_key_name:
9760 raise errors.OpPrereqError("Missing X509 key name for encryption",
9763 if not self.dest_x509_ca_pem:
9764 raise errors.OpPrereqError("Missing destination X509 CA",
9767 def ExpandNames(self):
9768 self._ExpandAndLockInstance()
9770 # Lock all nodes for local exports
9771 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9772 # FIXME: lock only instance primary and destination node
9774 # Sad but true, for now we have do lock all nodes, as we don't know where
9775 # the previous export might be, and in this LU we search for it and
9776 # remove it from its current node. In the future we could fix this by:
9777 # - making a tasklet to search (share-lock all), then create the
9778 # new one, then one to remove, after
9779 # - removing the removal operation altogether
9780 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9782 def DeclareLocks(self, level):
9783 """Last minute lock declaration."""
9784 # All nodes are locked anyway, so nothing to do here.
9786 def BuildHooksEnv(self):
9789 This will run on the master, primary node and target node.
9793 "EXPORT_MODE": self.op.mode,
9794 "EXPORT_NODE": self.op.target_node,
9795 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9796 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9797 # TODO: Generic function for boolean env variables
9798 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9801 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9803 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9805 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9806 nl.append(self.op.target_node)
9810 def CheckPrereq(self):
9811 """Check prerequisites.
9813 This checks that the instance and node names are valid.
9816 instance_name = self.op.instance_name
9818 self.instance = self.cfg.GetInstanceInfo(instance_name)
9819 assert self.instance is not None, \
9820 "Cannot retrieve locked instance %s" % self.op.instance_name
9821 _CheckNodeOnline(self, self.instance.primary_node)
9823 if (self.op.remove_instance and self.instance.admin_up and
9824 not self.op.shutdown):
9825 raise errors.OpPrereqError("Can not remove instance without shutting it"
9828 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9829 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9830 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9831 assert self.dst_node is not None
9833 _CheckNodeOnline(self, self.dst_node.name)
9834 _CheckNodeNotDrained(self, self.dst_node.name)
9837 self.dest_disk_info = None
9838 self.dest_x509_ca = None
9840 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9841 self.dst_node = None
9843 if len(self.op.target_node) != len(self.instance.disks):
9844 raise errors.OpPrereqError(("Received destination information for %s"
9845 " disks, but instance %s has %s disks") %
9846 (len(self.op.target_node), instance_name,
9847 len(self.instance.disks)),
9850 cds = _GetClusterDomainSecret()
9852 # Check X509 key name
9854 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9855 except (TypeError, ValueError), err:
9856 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9858 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9859 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9862 # Load and verify CA
9864 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9865 except OpenSSL.crypto.Error, err:
9866 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9867 (err, ), errors.ECODE_INVAL)
9869 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9870 if errcode is not None:
9871 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9872 (msg, ), errors.ECODE_INVAL)
9874 self.dest_x509_ca = cert
9876 # Verify target information
9878 for idx, disk_data in enumerate(self.op.target_node):
9880 (host, port, magic) = \
9881 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9882 except errors.GenericError, err:
9883 raise errors.OpPrereqError("Target info for disk %s: %s" %
9884 (idx, err), errors.ECODE_INVAL)
9886 disk_info.append((host, port, magic))
9888 assert len(disk_info) == len(self.op.target_node)
9889 self.dest_disk_info = disk_info
9892 raise errors.ProgrammerError("Unhandled export mode %r" %
9895 # instance disk type verification
9896 # TODO: Implement export support for file-based disks
9897 for disk in self.instance.disks:
9898 if disk.dev_type == constants.LD_FILE:
9899 raise errors.OpPrereqError("Export not supported for instances with"
9900 " file-based disks", errors.ECODE_INVAL)
9902 def _CleanupExports(self, feedback_fn):
9903 """Removes exports of current instance from all other nodes.
9905 If an instance in a cluster with nodes A..D was exported to node C, its
9906 exports will be removed from the nodes A, B and D.
9909 assert self.op.mode != constants.EXPORT_MODE_REMOTE
9911 nodelist = self.cfg.GetNodeList()
9912 nodelist.remove(self.dst_node.name)
9914 # on one-node clusters nodelist will be empty after the removal
9915 # if we proceed the backup would be removed because OpBackupQuery
9916 # substitutes an empty list with the full cluster node list.
9917 iname = self.instance.name
9919 feedback_fn("Removing old exports for instance %s" % iname)
9920 exportlist = self.rpc.call_export_list(nodelist)
9921 for node in exportlist:
9922 if exportlist[node].fail_msg:
9924 if iname in exportlist[node].payload:
9925 msg = self.rpc.call_export_remove(node, iname).fail_msg
9927 self.LogWarning("Could not remove older export for instance %s"
9928 " on node %s: %s", iname, node, msg)
9930 def Exec(self, feedback_fn):
9931 """Export an instance to an image in the cluster.
9934 assert self.op.mode in constants.EXPORT_MODES
9936 instance = self.instance
9937 src_node = instance.primary_node
9939 if self.op.shutdown:
9940 # shutdown the instance, but not the disks
9941 feedback_fn("Shutting down instance %s" % instance.name)
9942 result = self.rpc.call_instance_shutdown(src_node, instance,
9943 self.op.shutdown_timeout)
9944 # TODO: Maybe ignore failures if ignore_remove_failures is set
9945 result.Raise("Could not shutdown instance %s on"
9946 " node %s" % (instance.name, src_node))
9948 # set the disks ID correctly since call_instance_start needs the
9949 # correct drbd minor to create the symlinks
9950 for disk in instance.disks:
9951 self.cfg.SetDiskID(disk, src_node)
9953 activate_disks = (not instance.admin_up)
9956 # Activate the instance disks if we'exporting a stopped instance
9957 feedback_fn("Activating disks for %s" % instance.name)
9958 _StartInstanceDisks(self, instance, None)
9961 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9964 helper.CreateSnapshots()
9966 if (self.op.shutdown and instance.admin_up and
9967 not self.op.remove_instance):
9968 assert not activate_disks
9969 feedback_fn("Starting instance %s" % instance.name)
9970 result = self.rpc.call_instance_start(src_node, instance, None, None)
9971 msg = result.fail_msg
9973 feedback_fn("Failed to start instance: %s" % msg)
9974 _ShutdownInstanceDisks(self, instance)
9975 raise errors.OpExecError("Could not start instance: %s" % msg)
9977 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9978 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9979 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9980 connect_timeout = constants.RIE_CONNECT_TIMEOUT
9981 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9983 (key_name, _, _) = self.x509_key_name
9986 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9989 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9990 key_name, dest_ca_pem,
9995 # Check for backwards compatibility
9996 assert len(dresults) == len(instance.disks)
9997 assert compat.all(isinstance(i, bool) for i in dresults), \
9998 "Not all results are boolean: %r" % dresults
10002 feedback_fn("Deactivating disks for %s" % instance.name)
10003 _ShutdownInstanceDisks(self, instance)
10005 if not (compat.all(dresults) and fin_resu):
10008 failures.append("export finalization")
10009 if not compat.all(dresults):
10010 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
10012 failures.append("disk export: disk(s) %s" % fdsk)
10014 raise errors.OpExecError("Export failed, errors in %s" %
10015 utils.CommaJoin(failures))
10017 # At this point, the export was successful, we can cleanup/finish
10019 # Remove instance if requested
10020 if self.op.remove_instance:
10021 feedback_fn("Removing instance %s" % instance.name)
10022 _RemoveInstance(self, feedback_fn, instance,
10023 self.op.ignore_remove_failures)
10025 if self.op.mode == constants.EXPORT_MODE_LOCAL:
10026 self._CleanupExports(feedback_fn)
10028 return fin_resu, dresults
10031 class LUBackupRemove(NoHooksLU):
10032 """Remove exports related to the named instance.
10037 def ExpandNames(self):
10038 self.needed_locks = {}
10039 # We need all nodes to be locked in order for RemoveExport to work, but we
10040 # don't need to lock the instance itself, as nothing will happen to it (and
10041 # we can remove exports also for a removed instance)
10042 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10044 def Exec(self, feedback_fn):
10045 """Remove any export.
10048 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
10049 # If the instance was not found we'll try with the name that was passed in.
10050 # This will only work if it was an FQDN, though.
10052 if not instance_name:
10054 instance_name = self.op.instance_name
10056 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
10057 exportlist = self.rpc.call_export_list(locked_nodes)
10059 for node in exportlist:
10060 msg = exportlist[node].fail_msg
10062 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
10064 if instance_name in exportlist[node].payload:
10066 result = self.rpc.call_export_remove(node, instance_name)
10067 msg = result.fail_msg
10069 logging.error("Could not remove export for instance %s"
10070 " on node %s: %s", instance_name, node, msg)
10072 if fqdn_warn and not found:
10073 feedback_fn("Export not found. If trying to remove an export belonging"
10074 " to a deleted instance please use its Fully Qualified"
10078 class LUGroupAdd(LogicalUnit):
10079 """Logical unit for creating node groups.
10082 HPATH = "group-add"
10083 HTYPE = constants.HTYPE_GROUP
10086 def ExpandNames(self):
10087 # We need the new group's UUID here so that we can create and acquire the
10088 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
10089 # that it should not check whether the UUID exists in the configuration.
10090 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
10091 self.needed_locks = {}
10092 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10094 def CheckPrereq(self):
10095 """Check prerequisites.
10097 This checks that the given group name is not an existing node group
10102 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10103 except errors.OpPrereqError:
10106 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
10107 " node group (UUID: %s)" %
10108 (self.op.group_name, existing_uuid),
10109 errors.ECODE_EXISTS)
10111 if self.op.ndparams:
10112 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10114 def BuildHooksEnv(self):
10115 """Build hooks env.
10119 "GROUP_NAME": self.op.group_name,
10121 mn = self.cfg.GetMasterNode()
10122 return env, [mn], [mn]
10124 def Exec(self, feedback_fn):
10125 """Add the node group to the cluster.
10128 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
10129 uuid=self.group_uuid,
10130 alloc_policy=self.op.alloc_policy,
10131 ndparams=self.op.ndparams)
10133 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
10134 del self.remove_locks[locking.LEVEL_NODEGROUP]
10137 class LUGroupAssignNodes(NoHooksLU):
10138 """Logical unit for assigning nodes to groups.
10143 def ExpandNames(self):
10144 # These raise errors.OpPrereqError on their own:
10145 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10146 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
10148 # We want to lock all the affected nodes and groups. We have readily
10149 # available the list of nodes, and the *destination* group. To gather the
10150 # list of "source" groups, we need to fetch node information.
10151 self.node_data = self.cfg.GetAllNodesInfo()
10152 affected_groups = set(self.node_data[node].group for node in self.op.nodes)
10153 affected_groups.add(self.group_uuid)
10155 self.needed_locks = {
10156 locking.LEVEL_NODEGROUP: list(affected_groups),
10157 locking.LEVEL_NODE: self.op.nodes,
10160 def CheckPrereq(self):
10161 """Check prerequisites.
10164 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10165 instance_data = self.cfg.GetAllInstancesInfo()
10167 if self.group is None:
10168 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10169 (self.op.group_name, self.group_uuid))
10171 (new_splits, previous_splits) = \
10172 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
10173 for node in self.op.nodes],
10174 self.node_data, instance_data)
10177 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
10179 if not self.op.force:
10180 raise errors.OpExecError("The following instances get split by this"
10181 " change and --force was not given: %s" %
10184 self.LogWarning("This operation will split the following instances: %s",
10187 if previous_splits:
10188 self.LogWarning("In addition, these already-split instances continue"
10189 " to be spit across groups: %s",
10190 utils.CommaJoin(utils.NiceSort(previous_splits)))
10192 def Exec(self, feedback_fn):
10193 """Assign nodes to a new group.
10196 for node in self.op.nodes:
10197 self.node_data[node].group = self.group_uuid
10199 self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
10202 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
10203 """Check for split instances after a node assignment.
10205 This method considers a series of node assignments as an atomic operation,
10206 and returns information about split instances after applying the set of
10209 In particular, it returns information about newly split instances, and
10210 instances that were already split, and remain so after the change.
10212 Only instances whose disk template is listed in constants.DTS_NET_MIRROR are
10215 @type changes: list of (node_name, new_group_uuid) pairs.
10216 @param changes: list of node assignments to consider.
10217 @param node_data: a dict with data for all nodes
10218 @param instance_data: a dict with all instances to consider
10219 @rtype: a two-tuple
10220 @return: a list of instances that were previously okay and result split as a
10221 consequence of this change, and a list of instances that were previously
10222 split and this change does not fix.
10225 changed_nodes = dict((node, group) for node, group in changes
10226 if node_data[node].group != group)
10228 all_split_instances = set()
10229 previously_split_instances = set()
10231 def InstanceNodes(instance):
10232 return [instance.primary_node] + list(instance.secondary_nodes)
10234 for inst in instance_data.values():
10235 if inst.disk_template not in constants.DTS_NET_MIRROR:
10238 instance_nodes = InstanceNodes(inst)
10240 if len(set(node_data[node].group for node in instance_nodes)) > 1:
10241 previously_split_instances.add(inst.name)
10243 if len(set(changed_nodes.get(node, node_data[node].group)
10244 for node in instance_nodes)) > 1:
10245 all_split_instances.add(inst.name)
10247 return (list(all_split_instances - previously_split_instances),
10248 list(previously_split_instances & all_split_instances))
10251 class _GroupQuery(_QueryBase):
10252 FIELDS = query.GROUP_FIELDS
10254 def ExpandNames(self, lu):
10255 lu.needed_locks = {}
10257 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
10258 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
10261 self.wanted = [name_to_uuid[name]
10262 for name in utils.NiceSort(name_to_uuid.keys())]
10264 # Accept names to be either names or UUIDs.
10267 all_uuid = frozenset(self._all_groups.keys())
10269 for name in self.names:
10270 if name in all_uuid:
10271 self.wanted.append(name)
10272 elif name in name_to_uuid:
10273 self.wanted.append(name_to_uuid[name])
10275 missing.append(name)
10278 raise errors.OpPrereqError("Some groups do not exist: %s" % missing,
10279 errors.ECODE_NOENT)
10281 def DeclareLocks(self, lu, level):
10284 def _GetQueryData(self, lu):
10285 """Computes the list of node groups and their attributes.
10288 do_nodes = query.GQ_NODE in self.requested_data
10289 do_instances = query.GQ_INST in self.requested_data
10291 group_to_nodes = None
10292 group_to_instances = None
10294 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
10295 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
10296 # latter GetAllInstancesInfo() is not enough, for we have to go through
10297 # instance->node. Hence, we will need to process nodes even if we only need
10298 # instance information.
10299 if do_nodes or do_instances:
10300 all_nodes = lu.cfg.GetAllNodesInfo()
10301 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
10304 for node in all_nodes.values():
10305 if node.group in group_to_nodes:
10306 group_to_nodes[node.group].append(node.name)
10307 node_to_group[node.name] = node.group
10310 all_instances = lu.cfg.GetAllInstancesInfo()
10311 group_to_instances = dict((uuid, []) for uuid in self.wanted)
10313 for instance in all_instances.values():
10314 node = instance.primary_node
10315 if node in node_to_group:
10316 group_to_instances[node_to_group[node]].append(instance.name)
10319 # Do not pass on node information if it was not requested.
10320 group_to_nodes = None
10322 return query.GroupQueryData([self._all_groups[uuid]
10323 for uuid in self.wanted],
10324 group_to_nodes, group_to_instances)
10327 class LUGroupQuery(NoHooksLU):
10328 """Logical unit for querying node groups.
10333 def CheckArguments(self):
10334 self.gq = _GroupQuery(qlang.MakeSimpleFilter("name", self.op.names),
10335 self.op.output_fields, False)
10337 def ExpandNames(self):
10338 self.gq.ExpandNames(self)
10340 def Exec(self, feedback_fn):
10341 return self.gq.OldStyleQuery(self)
10344 class LUGroupSetParams(LogicalUnit):
10345 """Modifies the parameters of a node group.
10348 HPATH = "group-modify"
10349 HTYPE = constants.HTYPE_GROUP
10352 def CheckArguments(self):
10355 self.op.alloc_policy,
10358 if all_changes.count(None) == len(all_changes):
10359 raise errors.OpPrereqError("Please pass at least one modification",
10360 errors.ECODE_INVAL)
10362 def ExpandNames(self):
10363 # This raises errors.OpPrereqError on its own:
10364 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10366 self.needed_locks = {
10367 locking.LEVEL_NODEGROUP: [self.group_uuid],
10370 def CheckPrereq(self):
10371 """Check prerequisites.
10374 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10376 if self.group is None:
10377 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10378 (self.op.group_name, self.group_uuid))
10380 if self.op.ndparams:
10381 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
10382 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10383 self.new_ndparams = new_ndparams
10385 def BuildHooksEnv(self):
10386 """Build hooks env.
10390 "GROUP_NAME": self.op.group_name,
10391 "NEW_ALLOC_POLICY": self.op.alloc_policy,
10393 mn = self.cfg.GetMasterNode()
10394 return env, [mn], [mn]
10396 def Exec(self, feedback_fn):
10397 """Modifies the node group.
10402 if self.op.ndparams:
10403 self.group.ndparams = self.new_ndparams
10404 result.append(("ndparams", str(self.group.ndparams)))
10406 if self.op.alloc_policy:
10407 self.group.alloc_policy = self.op.alloc_policy
10409 self.cfg.Update(self.group, feedback_fn)
10414 class LUGroupRemove(LogicalUnit):
10415 HPATH = "group-remove"
10416 HTYPE = constants.HTYPE_GROUP
10419 def ExpandNames(self):
10420 # This will raises errors.OpPrereqError on its own:
10421 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10422 self.needed_locks = {
10423 locking.LEVEL_NODEGROUP: [self.group_uuid],
10426 def CheckPrereq(self):
10427 """Check prerequisites.
10429 This checks that the given group name exists as a node group, that is
10430 empty (i.e., contains no nodes), and that is not the last group of the
10434 # Verify that the group is empty.
10435 group_nodes = [node.name
10436 for node in self.cfg.GetAllNodesInfo().values()
10437 if node.group == self.group_uuid]
10440 raise errors.OpPrereqError("Group '%s' not empty, has the following"
10442 (self.op.group_name,
10443 utils.CommaJoin(utils.NiceSort(group_nodes))),
10444 errors.ECODE_STATE)
10446 # Verify the cluster would not be left group-less.
10447 if len(self.cfg.GetNodeGroupList()) == 1:
10448 raise errors.OpPrereqError("Group '%s' is the only group,"
10449 " cannot be removed" %
10450 self.op.group_name,
10451 errors.ECODE_STATE)
10453 def BuildHooksEnv(self):
10454 """Build hooks env.
10458 "GROUP_NAME": self.op.group_name,
10460 mn = self.cfg.GetMasterNode()
10461 return env, [mn], [mn]
10463 def Exec(self, feedback_fn):
10464 """Remove the node group.
10468 self.cfg.RemoveNodeGroup(self.group_uuid)
10469 except errors.ConfigurationError:
10470 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
10471 (self.op.group_name, self.group_uuid))
10473 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10476 class LUGroupRename(LogicalUnit):
10477 HPATH = "group-rename"
10478 HTYPE = constants.HTYPE_GROUP
10481 def ExpandNames(self):
10482 # This raises errors.OpPrereqError on its own:
10483 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10485 self.needed_locks = {
10486 locking.LEVEL_NODEGROUP: [self.group_uuid],
10489 def CheckPrereq(self):
10490 """Check prerequisites.
10492 Ensures requested new name is not yet used.
10496 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
10497 except errors.OpPrereqError:
10500 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
10501 " node group (UUID: %s)" %
10502 (self.op.new_name, new_name_uuid),
10503 errors.ECODE_EXISTS)
10505 def BuildHooksEnv(self):
10506 """Build hooks env.
10510 "OLD_NAME": self.op.group_name,
10511 "NEW_NAME": self.op.new_name,
10514 mn = self.cfg.GetMasterNode()
10515 all_nodes = self.cfg.GetAllNodesInfo()
10517 all_nodes.pop(mn, None)
10519 for node in all_nodes.values():
10520 if node.group == self.group_uuid:
10521 run_nodes.append(node.name)
10523 return env, run_nodes, run_nodes
10525 def Exec(self, feedback_fn):
10526 """Rename the node group.
10529 group = self.cfg.GetNodeGroup(self.group_uuid)
10532 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10533 (self.op.group_name, self.group_uuid))
10535 group.name = self.op.new_name
10536 self.cfg.Update(group, feedback_fn)
10538 return self.op.new_name
10541 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
10542 """Generic tags LU.
10544 This is an abstract class which is the parent of all the other tags LUs.
10548 def ExpandNames(self):
10549 self.needed_locks = {}
10550 if self.op.kind == constants.TAG_NODE:
10551 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
10552 self.needed_locks[locking.LEVEL_NODE] = self.op.name
10553 elif self.op.kind == constants.TAG_INSTANCE:
10554 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
10555 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
10557 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
10558 # not possible to acquire the BGL based on opcode parameters)
10560 def CheckPrereq(self):
10561 """Check prerequisites.
10564 if self.op.kind == constants.TAG_CLUSTER:
10565 self.target = self.cfg.GetClusterInfo()
10566 elif self.op.kind == constants.TAG_NODE:
10567 self.target = self.cfg.GetNodeInfo(self.op.name)
10568 elif self.op.kind == constants.TAG_INSTANCE:
10569 self.target = self.cfg.GetInstanceInfo(self.op.name)
10571 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
10572 str(self.op.kind), errors.ECODE_INVAL)
10575 class LUTagsGet(TagsLU):
10576 """Returns the tags of a given object.
10581 def ExpandNames(self):
10582 TagsLU.ExpandNames(self)
10584 # Share locks as this is only a read operation
10585 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
10587 def Exec(self, feedback_fn):
10588 """Returns the tag list.
10591 return list(self.target.GetTags())
10594 class LUTagsSearch(NoHooksLU):
10595 """Searches the tags for a given pattern.
10600 def ExpandNames(self):
10601 self.needed_locks = {}
10603 def CheckPrereq(self):
10604 """Check prerequisites.
10606 This checks the pattern passed for validity by compiling it.
10610 self.re = re.compile(self.op.pattern)
10611 except re.error, err:
10612 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
10613 (self.op.pattern, err), errors.ECODE_INVAL)
10615 def Exec(self, feedback_fn):
10616 """Returns the tag list.
10620 tgts = [("/cluster", cfg.GetClusterInfo())]
10621 ilist = cfg.GetAllInstancesInfo().values()
10622 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
10623 nlist = cfg.GetAllNodesInfo().values()
10624 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
10626 for path, target in tgts:
10627 for tag in target.GetTags():
10628 if self.re.search(tag):
10629 results.append((path, tag))
10633 class LUTagsSet(TagsLU):
10634 """Sets a tag on a given object.
10639 def CheckPrereq(self):
10640 """Check prerequisites.
10642 This checks the type and length of the tag name and value.
10645 TagsLU.CheckPrereq(self)
10646 for tag in self.op.tags:
10647 objects.TaggableObject.ValidateTag(tag)
10649 def Exec(self, feedback_fn):
10654 for tag in self.op.tags:
10655 self.target.AddTag(tag)
10656 except errors.TagError, err:
10657 raise errors.OpExecError("Error while setting tag: %s" % str(err))
10658 self.cfg.Update(self.target, feedback_fn)
10661 class LUTagsDel(TagsLU):
10662 """Delete a list of tags from a given object.
10667 def CheckPrereq(self):
10668 """Check prerequisites.
10670 This checks that we have the given tag.
10673 TagsLU.CheckPrereq(self)
10674 for tag in self.op.tags:
10675 objects.TaggableObject.ValidateTag(tag)
10676 del_tags = frozenset(self.op.tags)
10677 cur_tags = self.target.GetTags()
10679 diff_tags = del_tags - cur_tags
10681 diff_names = ("'%s'" % i for i in sorted(diff_tags))
10682 raise errors.OpPrereqError("Tag(s) %s not found" %
10683 (utils.CommaJoin(diff_names), ),
10684 errors.ECODE_NOENT)
10686 def Exec(self, feedback_fn):
10687 """Remove the tag from the object.
10690 for tag in self.op.tags:
10691 self.target.RemoveTag(tag)
10692 self.cfg.Update(self.target, feedback_fn)
10695 class LUTestDelay(NoHooksLU):
10696 """Sleep for a specified amount of time.
10698 This LU sleeps on the master and/or nodes for a specified amount of
10704 def ExpandNames(self):
10705 """Expand names and set required locks.
10707 This expands the node list, if any.
10710 self.needed_locks = {}
10711 if self.op.on_nodes:
10712 # _GetWantedNodes can be used here, but is not always appropriate to use
10713 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
10714 # more information.
10715 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
10716 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
10718 def _TestDelay(self):
10719 """Do the actual sleep.
10722 if self.op.on_master:
10723 if not utils.TestDelay(self.op.duration):
10724 raise errors.OpExecError("Error during master delay test")
10725 if self.op.on_nodes:
10726 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
10727 for node, node_result in result.items():
10728 node_result.Raise("Failure during rpc call to node %s" % node)
10730 def Exec(self, feedback_fn):
10731 """Execute the test delay opcode, with the wanted repetitions.
10734 if self.op.repeat == 0:
10737 top_value = self.op.repeat - 1
10738 for i in range(self.op.repeat):
10739 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
10743 class LUTestJqueue(NoHooksLU):
10744 """Utility LU to test some aspects of the job queue.
10749 # Must be lower than default timeout for WaitForJobChange to see whether it
10750 # notices changed jobs
10751 _CLIENT_CONNECT_TIMEOUT = 20.0
10752 _CLIENT_CONFIRM_TIMEOUT = 60.0
10755 def _NotifyUsingSocket(cls, cb, errcls):
10756 """Opens a Unix socket and waits for another program to connect.
10759 @param cb: Callback to send socket name to client
10760 @type errcls: class
10761 @param errcls: Exception class to use for errors
10764 # Using a temporary directory as there's no easy way to create temporary
10765 # sockets without writing a custom loop around tempfile.mktemp and
10767 tmpdir = tempfile.mkdtemp()
10769 tmpsock = utils.PathJoin(tmpdir, "sock")
10771 logging.debug("Creating temporary socket at %s", tmpsock)
10772 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10777 # Send details to client
10780 # Wait for client to connect before continuing
10781 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10783 (conn, _) = sock.accept()
10784 except socket.error, err:
10785 raise errcls("Client didn't connect in time (%s)" % err)
10789 # Remove as soon as client is connected
10790 shutil.rmtree(tmpdir)
10792 # Wait for client to close
10795 # pylint: disable-msg=E1101
10796 # Instance of '_socketobject' has no ... member
10797 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10799 except socket.error, err:
10800 raise errcls("Client failed to confirm notification (%s)" % err)
10804 def _SendNotification(self, test, arg, sockname):
10805 """Sends a notification to the client.
10808 @param test: Test name
10809 @param arg: Test argument (depends on test)
10810 @type sockname: string
10811 @param sockname: Socket path
10814 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
10816 def _Notify(self, prereq, test, arg):
10817 """Notifies the client of a test.
10820 @param prereq: Whether this is a prereq-phase test
10822 @param test: Test name
10823 @param arg: Test argument (depends on test)
10827 errcls = errors.OpPrereqError
10829 errcls = errors.OpExecError
10831 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10835 def CheckArguments(self):
10836 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10837 self.expandnames_calls = 0
10839 def ExpandNames(self):
10840 checkargs_calls = getattr(self, "checkargs_calls", 0)
10841 if checkargs_calls < 1:
10842 raise errors.ProgrammerError("CheckArguments was not called")
10844 self.expandnames_calls += 1
10846 if self.op.notify_waitlock:
10847 self._Notify(True, constants.JQT_EXPANDNAMES, None)
10849 self.LogInfo("Expanding names")
10851 # Get lock on master node (just to get a lock, not for a particular reason)
10852 self.needed_locks = {
10853 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10856 def Exec(self, feedback_fn):
10857 if self.expandnames_calls < 1:
10858 raise errors.ProgrammerError("ExpandNames was not called")
10860 if self.op.notify_exec:
10861 self._Notify(False, constants.JQT_EXEC, None)
10863 self.LogInfo("Executing")
10865 if self.op.log_messages:
10866 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10867 for idx, msg in enumerate(self.op.log_messages):
10868 self.LogInfo("Sending log message %s", idx + 1)
10869 feedback_fn(constants.JQT_MSGPREFIX + msg)
10870 # Report how many test messages have been sent
10871 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10874 raise errors.OpExecError("Opcode failure was requested")
10879 class IAllocator(object):
10880 """IAllocator framework.
10882 An IAllocator instance has three sets of attributes:
10883 - cfg that is needed to query the cluster
10884 - input data (all members of the _KEYS class attribute are required)
10885 - four buffer attributes (in|out_data|text), that represent the
10886 input (to the external script) in text and data structure format,
10887 and the output from it, again in two formats
10888 - the result variables from the script (success, info, nodes) for
10892 # pylint: disable-msg=R0902
10893 # lots of instance attributes
10895 "name", "mem_size", "disks", "disk_template",
10896 "os", "tags", "nics", "vcpus", "hypervisor",
10899 "name", "relocate_from",
10905 def __init__(self, cfg, rpc, mode, **kwargs):
10908 # init buffer variables
10909 self.in_text = self.out_text = self.in_data = self.out_data = None
10910 # init all input fields so that pylint is happy
10912 self.mem_size = self.disks = self.disk_template = None
10913 self.os = self.tags = self.nics = self.vcpus = None
10914 self.hypervisor = None
10915 self.relocate_from = None
10917 self.evac_nodes = None
10919 self.required_nodes = None
10920 # init result fields
10921 self.success = self.info = self.result = None
10922 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10923 keyset = self._ALLO_KEYS
10924 fn = self._AddNewInstance
10925 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10926 keyset = self._RELO_KEYS
10927 fn = self._AddRelocateInstance
10928 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10929 keyset = self._EVAC_KEYS
10930 fn = self._AddEvacuateNodes
10932 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10933 " IAllocator" % self.mode)
10935 if key not in keyset:
10936 raise errors.ProgrammerError("Invalid input parameter '%s' to"
10937 " IAllocator" % key)
10938 setattr(self, key, kwargs[key])
10941 if key not in kwargs:
10942 raise errors.ProgrammerError("Missing input parameter '%s' to"
10943 " IAllocator" % key)
10944 self._BuildInputData(fn)
10946 def _ComputeClusterData(self):
10947 """Compute the generic allocator input data.
10949 This is the data that is independent of the actual operation.
10953 cluster_info = cfg.GetClusterInfo()
10956 "version": constants.IALLOCATOR_VERSION,
10957 "cluster_name": cfg.GetClusterName(),
10958 "cluster_tags": list(cluster_info.GetTags()),
10959 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10960 # we don't have job IDs
10962 ninfo = cfg.GetAllNodesInfo()
10963 iinfo = cfg.GetAllInstancesInfo().values()
10964 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10967 node_list = [n.name for n in ninfo.values() if n.vm_capable]
10969 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10970 hypervisor_name = self.hypervisor
10971 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10972 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10973 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10974 hypervisor_name = cluster_info.enabled_hypervisors[0]
10976 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10979 self.rpc.call_all_instances_info(node_list,
10980 cluster_info.enabled_hypervisors)
10982 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
10984 config_ndata = self._ComputeBasicNodeData(ninfo)
10985 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
10986 i_list, config_ndata)
10987 assert len(data["nodes"]) == len(ninfo), \
10988 "Incomplete node data computed"
10990 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
10992 self.in_data = data
10995 def _ComputeNodeGroupData(cfg):
10996 """Compute node groups data.
11000 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
11002 "name": gdata.name,
11003 "alloc_policy": gdata.alloc_policy,
11008 def _ComputeBasicNodeData(node_cfg):
11009 """Compute global node data.
11012 @returns: a dict of name: (node dict, node config)
11016 for ninfo in node_cfg.values():
11017 # fill in static (config-based) values
11019 "tags": list(ninfo.GetTags()),
11020 "primary_ip": ninfo.primary_ip,
11021 "secondary_ip": ninfo.secondary_ip,
11022 "offline": ninfo.offline,
11023 "drained": ninfo.drained,
11024 "master_candidate": ninfo.master_candidate,
11025 "group": ninfo.group,
11026 "master_capable": ninfo.master_capable,
11027 "vm_capable": ninfo.vm_capable,
11030 node_results[ninfo.name] = pnr
11032 return node_results
11035 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
11037 """Compute global node data.
11039 @param node_results: the basic node structures as filled from the config
11042 # make a copy of the current dict
11043 node_results = dict(node_results)
11044 for nname, nresult in node_data.items():
11045 assert nname in node_results, "Missing basic data for node %s" % nname
11046 ninfo = node_cfg[nname]
11048 if not (ninfo.offline or ninfo.drained):
11049 nresult.Raise("Can't get data for node %s" % nname)
11050 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
11052 remote_info = nresult.payload
11054 for attr in ['memory_total', 'memory_free', 'memory_dom0',
11055 'vg_size', 'vg_free', 'cpu_total']:
11056 if attr not in remote_info:
11057 raise errors.OpExecError("Node '%s' didn't return attribute"
11058 " '%s'" % (nname, attr))
11059 if not isinstance(remote_info[attr], int):
11060 raise errors.OpExecError("Node '%s' returned invalid value"
11062 (nname, attr, remote_info[attr]))
11063 # compute memory used by primary instances
11064 i_p_mem = i_p_up_mem = 0
11065 for iinfo, beinfo in i_list:
11066 if iinfo.primary_node == nname:
11067 i_p_mem += beinfo[constants.BE_MEMORY]
11068 if iinfo.name not in node_iinfo[nname].payload:
11071 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
11072 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
11073 remote_info['memory_free'] -= max(0, i_mem_diff)
11076 i_p_up_mem += beinfo[constants.BE_MEMORY]
11078 # compute memory used by instances
11080 "total_memory": remote_info['memory_total'],
11081 "reserved_memory": remote_info['memory_dom0'],
11082 "free_memory": remote_info['memory_free'],
11083 "total_disk": remote_info['vg_size'],
11084 "free_disk": remote_info['vg_free'],
11085 "total_cpus": remote_info['cpu_total'],
11086 "i_pri_memory": i_p_mem,
11087 "i_pri_up_memory": i_p_up_mem,
11089 pnr_dyn.update(node_results[nname])
11090 node_results[nname] = pnr_dyn
11092 return node_results
11095 def _ComputeInstanceData(cluster_info, i_list):
11096 """Compute global instance data.
11100 for iinfo, beinfo in i_list:
11102 for nic in iinfo.nics:
11103 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
11104 nic_dict = {"mac": nic.mac,
11106 "mode": filled_params[constants.NIC_MODE],
11107 "link": filled_params[constants.NIC_LINK],
11109 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
11110 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
11111 nic_data.append(nic_dict)
11113 "tags": list(iinfo.GetTags()),
11114 "admin_up": iinfo.admin_up,
11115 "vcpus": beinfo[constants.BE_VCPUS],
11116 "memory": beinfo[constants.BE_MEMORY],
11118 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
11120 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
11121 "disk_template": iinfo.disk_template,
11122 "hypervisor": iinfo.hypervisor,
11124 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
11126 instance_data[iinfo.name] = pir
11128 return instance_data
11130 def _AddNewInstance(self):
11131 """Add new instance data to allocator structure.
11133 This in combination with _AllocatorGetClusterData will create the
11134 correct structure needed as input for the allocator.
11136 The checks for the completeness of the opcode must have already been
11140 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
11142 if self.disk_template in constants.DTS_NET_MIRROR:
11143 self.required_nodes = 2
11145 self.required_nodes = 1
11148 "disk_template": self.disk_template,
11151 "vcpus": self.vcpus,
11152 "memory": self.mem_size,
11153 "disks": self.disks,
11154 "disk_space_total": disk_space,
11156 "required_nodes": self.required_nodes,
11160 def _AddRelocateInstance(self):
11161 """Add relocate instance data to allocator structure.
11163 This in combination with _IAllocatorGetClusterData will create the
11164 correct structure needed as input for the allocator.
11166 The checks for the completeness of the opcode must have already been
11170 instance = self.cfg.GetInstanceInfo(self.name)
11171 if instance is None:
11172 raise errors.ProgrammerError("Unknown instance '%s' passed to"
11173 " IAllocator" % self.name)
11175 if instance.disk_template not in constants.DTS_NET_MIRROR:
11176 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
11177 errors.ECODE_INVAL)
11179 if len(instance.secondary_nodes) != 1:
11180 raise errors.OpPrereqError("Instance has not exactly one secondary node",
11181 errors.ECODE_STATE)
11183 self.required_nodes = 1
11184 disk_sizes = [{'size': disk.size} for disk in instance.disks]
11185 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
11189 "disk_space_total": disk_space,
11190 "required_nodes": self.required_nodes,
11191 "relocate_from": self.relocate_from,
11195 def _AddEvacuateNodes(self):
11196 """Add evacuate nodes data to allocator structure.
11200 "evac_nodes": self.evac_nodes
11204 def _BuildInputData(self, fn):
11205 """Build input data structures.
11208 self._ComputeClusterData()
11211 request["type"] = self.mode
11212 self.in_data["request"] = request
11214 self.in_text = serializer.Dump(self.in_data)
11216 def Run(self, name, validate=True, call_fn=None):
11217 """Run an instance allocator and return the results.
11220 if call_fn is None:
11221 call_fn = self.rpc.call_iallocator_runner
11223 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
11224 result.Raise("Failure while running the iallocator script")
11226 self.out_text = result.payload
11228 self._ValidateResult()
11230 def _ValidateResult(self):
11231 """Process the allocator results.
11233 This will process and if successful save the result in
11234 self.out_data and the other parameters.
11238 rdict = serializer.Load(self.out_text)
11239 except Exception, err:
11240 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
11242 if not isinstance(rdict, dict):
11243 raise errors.OpExecError("Can't parse iallocator results: not a dict")
11245 # TODO: remove backwards compatiblity in later versions
11246 if "nodes" in rdict and "result" not in rdict:
11247 rdict["result"] = rdict["nodes"]
11250 for key in "success", "info", "result":
11251 if key not in rdict:
11252 raise errors.OpExecError("Can't parse iallocator results:"
11253 " missing key '%s'" % key)
11254 setattr(self, key, rdict[key])
11256 if not isinstance(rdict["result"], list):
11257 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
11259 self.out_data = rdict
11262 class LUTestAllocator(NoHooksLU):
11263 """Run allocator tests.
11265 This LU runs the allocator tests
11268 def CheckPrereq(self):
11269 """Check prerequisites.
11271 This checks the opcode parameters depending on the director and mode test.
11274 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11275 for attr in ["mem_size", "disks", "disk_template",
11276 "os", "tags", "nics", "vcpus"]:
11277 if not hasattr(self.op, attr):
11278 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
11279 attr, errors.ECODE_INVAL)
11280 iname = self.cfg.ExpandInstanceName(self.op.name)
11281 if iname is not None:
11282 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
11283 iname, errors.ECODE_EXISTS)
11284 if not isinstance(self.op.nics, list):
11285 raise errors.OpPrereqError("Invalid parameter 'nics'",
11286 errors.ECODE_INVAL)
11287 if not isinstance(self.op.disks, list):
11288 raise errors.OpPrereqError("Invalid parameter 'disks'",
11289 errors.ECODE_INVAL)
11290 for row in self.op.disks:
11291 if (not isinstance(row, dict) or
11292 "size" not in row or
11293 not isinstance(row["size"], int) or
11294 "mode" not in row or
11295 row["mode"] not in ['r', 'w']):
11296 raise errors.OpPrereqError("Invalid contents of the 'disks'"
11297 " parameter", errors.ECODE_INVAL)
11298 if self.op.hypervisor is None:
11299 self.op.hypervisor = self.cfg.GetHypervisorType()
11300 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11301 fname = _ExpandInstanceName(self.cfg, self.op.name)
11302 self.op.name = fname
11303 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
11304 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11305 if not hasattr(self.op, "evac_nodes"):
11306 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
11307 " opcode input", errors.ECODE_INVAL)
11309 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
11310 self.op.mode, errors.ECODE_INVAL)
11312 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
11313 if self.op.allocator is None:
11314 raise errors.OpPrereqError("Missing allocator name",
11315 errors.ECODE_INVAL)
11316 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
11317 raise errors.OpPrereqError("Wrong allocator test '%s'" %
11318 self.op.direction, errors.ECODE_INVAL)
11320 def Exec(self, feedback_fn):
11321 """Run the allocator test.
11324 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11325 ial = IAllocator(self.cfg, self.rpc,
11328 mem_size=self.op.mem_size,
11329 disks=self.op.disks,
11330 disk_template=self.op.disk_template,
11334 vcpus=self.op.vcpus,
11335 hypervisor=self.op.hypervisor,
11337 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11338 ial = IAllocator(self.cfg, self.rpc,
11341 relocate_from=list(self.relocate_from),
11343 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11344 ial = IAllocator(self.cfg, self.rpc,
11346 evac_nodes=self.op.evac_nodes)
11348 raise errors.ProgrammerError("Uncatched mode %s in"
11349 " LUTestAllocator.Exec", self.op.mode)
11351 if self.op.direction == constants.IALLOCATOR_DIR_IN:
11352 result = ial.in_text
11354 ial.Run(self.op.allocator, validate=False)
11355 result = ial.out_text
11359 #: Query type implementations
11361 constants.QR_INSTANCE: _InstanceQuery,
11362 constants.QR_NODE: _NodeQuery,
11363 constants.QR_GROUP: _GroupQuery,
11367 def _GetQueryImplementation(name):
11368 """Returns the implemtnation for a query type.
11370 @param name: Query type, must be one of L{constants.QR_OP_QUERY}
11374 return _QUERY_IMPL[name]
11376 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
11377 errors.ECODE_INVAL)