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, names, fields, use_locking):
457 """Initializes this class.
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: diction of {secondary-node: list of instances} of all peers
1258 of this node (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 test = nresult.get(constants.NV_NODESETUP,
1401 ["Missing NODESETUP results"])
1402 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1407 def _VerifyNodeTime(self, ninfo, nresult,
1408 nvinfo_starttime, nvinfo_endtime):
1409 """Check the node time.
1411 @type ninfo: L{objects.Node}
1412 @param ninfo: the node to check
1413 @param nresult: the remote results for the node
1414 @param nvinfo_starttime: the start time of the RPC call
1415 @param nvinfo_endtime: the end time of the RPC call
1419 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1421 ntime = nresult.get(constants.NV_TIME, None)
1423 ntime_merged = utils.MergeTime(ntime)
1424 except (ValueError, TypeError):
1425 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1428 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1429 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1430 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1431 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1435 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1436 "Node time diverges by at least %s from master node time",
1439 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1440 """Check the node time.
1442 @type ninfo: L{objects.Node}
1443 @param ninfo: the node to check
1444 @param nresult: the remote results for the node
1445 @param vg_name: the configured VG name
1452 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1454 # checks vg existence and size > 20G
1455 vglist = nresult.get(constants.NV_VGLIST, None)
1457 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1459 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1460 constants.MIN_VG_SIZE)
1461 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1464 pvlist = nresult.get(constants.NV_PVLIST, None)
1465 test = pvlist is None
1466 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1468 # check that ':' is not present in PV names, since it's a
1469 # special character for lvcreate (denotes the range of PEs to
1471 for _, pvname, owner_vg in pvlist:
1472 test = ":" in pvname
1473 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1474 " '%s' of VG '%s'", pvname, owner_vg)
1476 def _VerifyNodeNetwork(self, ninfo, nresult):
1477 """Check the node time.
1479 @type ninfo: L{objects.Node}
1480 @param ninfo: the node to check
1481 @param nresult: the remote results for the node
1485 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1487 test = constants.NV_NODELIST not in nresult
1488 _ErrorIf(test, self.ENODESSH, node,
1489 "node hasn't returned node ssh connectivity data")
1491 if nresult[constants.NV_NODELIST]:
1492 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1493 _ErrorIf(True, self.ENODESSH, node,
1494 "ssh communication with node '%s': %s", a_node, a_msg)
1496 test = constants.NV_NODENETTEST not in nresult
1497 _ErrorIf(test, self.ENODENET, node,
1498 "node hasn't returned node tcp connectivity data")
1500 if nresult[constants.NV_NODENETTEST]:
1501 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1503 _ErrorIf(True, self.ENODENET, node,
1504 "tcp communication with node '%s': %s",
1505 anode, nresult[constants.NV_NODENETTEST][anode])
1507 test = constants.NV_MASTERIP not in nresult
1508 _ErrorIf(test, self.ENODENET, node,
1509 "node hasn't returned node master IP reachability data")
1511 if not nresult[constants.NV_MASTERIP]:
1512 if node == self.master_node:
1513 msg = "the master node cannot reach the master IP (not configured?)"
1515 msg = "cannot reach the master IP"
1516 _ErrorIf(True, self.ENODENET, node, msg)
1518 def _VerifyInstance(self, instance, instanceconfig, node_image,
1520 """Verify an instance.
1522 This function checks to see if the required block devices are
1523 available on the instance's node.
1526 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1527 node_current = instanceconfig.primary_node
1529 node_vol_should = {}
1530 instanceconfig.MapLVsByNode(node_vol_should)
1532 for node in node_vol_should:
1533 n_img = node_image[node]
1534 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1535 # ignore missing volumes on offline or broken nodes
1537 for volume in node_vol_should[node]:
1538 test = volume not in n_img.volumes
1539 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1540 "volume %s missing on node %s", volume, node)
1542 if instanceconfig.admin_up:
1543 pri_img = node_image[node_current]
1544 test = instance not in pri_img.instances and not pri_img.offline
1545 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1546 "instance not running on its primary node %s",
1549 for node, n_img in node_image.items():
1550 if (not node == node_current):
1551 test = instance in n_img.instances
1552 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1553 "instance should not run on node %s", node)
1555 diskdata = [(nname, success, status, idx)
1556 for (nname, disks) in diskstatus.items()
1557 for idx, (success, status) in enumerate(disks)]
1559 for nname, success, bdev_status, idx in diskdata:
1560 _ErrorIf(instanceconfig.admin_up and not success,
1561 self.EINSTANCEFAULTYDISK, instance,
1562 "couldn't retrieve status for disk/%s on %s: %s",
1563 idx, nname, bdev_status)
1564 _ErrorIf((instanceconfig.admin_up and success and
1565 bdev_status.ldisk_status == constants.LDS_FAULTY),
1566 self.EINSTANCEFAULTYDISK, instance,
1567 "disk/%s on %s is faulty", idx, nname)
1569 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1570 """Verify if there are any unknown volumes in the cluster.
1572 The .os, .swap and backup volumes are ignored. All other volumes are
1573 reported as unknown.
1575 @type reserved: L{ganeti.utils.FieldSet}
1576 @param reserved: a FieldSet of reserved volume names
1579 for node, n_img in node_image.items():
1580 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1581 # skip non-healthy nodes
1583 for volume in n_img.volumes:
1584 test = ((node not in node_vol_should or
1585 volume not in node_vol_should[node]) and
1586 not reserved.Matches(volume))
1587 self._ErrorIf(test, self.ENODEORPHANLV, node,
1588 "volume %s is unknown", volume)
1590 def _VerifyOrphanInstances(self, instancelist, node_image):
1591 """Verify the list of running instances.
1593 This checks what instances are running but unknown to the cluster.
1596 for node, n_img in node_image.items():
1597 for o_inst in n_img.instances:
1598 test = o_inst not in instancelist
1599 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1600 "instance %s on node %s should not exist", o_inst, node)
1602 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1603 """Verify N+1 Memory Resilience.
1605 Check that if one single node dies we can still start all the
1606 instances it was primary for.
1609 for node, n_img in node_image.items():
1610 # This code checks that every node which is now listed as
1611 # secondary has enough memory to host all instances it is
1612 # supposed to should a single other node in the cluster fail.
1613 # FIXME: not ready for failover to an arbitrary node
1614 # FIXME: does not support file-backed instances
1615 # WARNING: we currently take into account down instances as well
1616 # as up ones, considering that even if they're down someone
1617 # might want to start them even in the event of a node failure.
1618 for prinode, instances in n_img.sbp.items():
1620 for instance in instances:
1621 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1622 if bep[constants.BE_AUTO_BALANCE]:
1623 needed_mem += bep[constants.BE_MEMORY]
1624 test = n_img.mfree < needed_mem
1625 self._ErrorIf(test, self.ENODEN1, node,
1626 "not enough memory to accomodate instance failovers"
1627 " should node %s fail", prinode)
1629 def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1631 """Verifies and computes the node required file checksums.
1633 @type ninfo: L{objects.Node}
1634 @param ninfo: the node to check
1635 @param nresult: the remote results for the node
1636 @param file_list: required list of files
1637 @param local_cksum: dictionary of local files and their checksums
1638 @param master_files: list of files that only masters should have
1642 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1644 remote_cksum = nresult.get(constants.NV_FILELIST, None)
1645 test = not isinstance(remote_cksum, dict)
1646 _ErrorIf(test, self.ENODEFILECHECK, node,
1647 "node hasn't returned file checksum data")
1651 for file_name in file_list:
1652 node_is_mc = ninfo.master_candidate
1653 must_have = (file_name not in master_files) or node_is_mc
1655 test1 = file_name not in remote_cksum
1657 test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1659 test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1660 _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1661 "file '%s' missing", file_name)
1662 _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1663 "file '%s' has wrong checksum", file_name)
1664 # not candidate and this is not a must-have file
1665 _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1666 "file '%s' should not exist on non master"
1667 " candidates (and the file is outdated)", file_name)
1668 # all good, except non-master/non-must have combination
1669 _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1670 "file '%s' should not exist"
1671 " on non master candidates", file_name)
1673 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1675 """Verifies and the node DRBD status.
1677 @type ninfo: L{objects.Node}
1678 @param ninfo: the node to check
1679 @param nresult: the remote results for the node
1680 @param instanceinfo: the dict of instances
1681 @param drbd_helper: the configured DRBD usermode helper
1682 @param drbd_map: the DRBD map as returned by
1683 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1687 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1690 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1691 test = (helper_result == None)
1692 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1693 "no drbd usermode helper returned")
1695 status, payload = helper_result
1697 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1698 "drbd usermode helper check unsuccessful: %s", payload)
1699 test = status and (payload != drbd_helper)
1700 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1701 "wrong drbd usermode helper: %s", payload)
1703 # compute the DRBD minors
1705 for minor, instance in drbd_map[node].items():
1706 test = instance not in instanceinfo
1707 _ErrorIf(test, self.ECLUSTERCFG, None,
1708 "ghost instance '%s' in temporary DRBD map", instance)
1709 # ghost instance should not be running, but otherwise we
1710 # don't give double warnings (both ghost instance and
1711 # unallocated minor in use)
1713 node_drbd[minor] = (instance, False)
1715 instance = instanceinfo[instance]
1716 node_drbd[minor] = (instance.name, instance.admin_up)
1718 # and now check them
1719 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1720 test = not isinstance(used_minors, (tuple, list))
1721 _ErrorIf(test, self.ENODEDRBD, node,
1722 "cannot parse drbd status file: %s", str(used_minors))
1724 # we cannot check drbd status
1727 for minor, (iname, must_exist) in node_drbd.items():
1728 test = minor not in used_minors and must_exist
1729 _ErrorIf(test, self.ENODEDRBD, node,
1730 "drbd minor %d of instance %s is not active", minor, iname)
1731 for minor in used_minors:
1732 test = minor not in node_drbd
1733 _ErrorIf(test, self.ENODEDRBD, node,
1734 "unallocated drbd minor %d is in use", minor)
1736 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1737 """Builds the node OS structures.
1739 @type ninfo: L{objects.Node}
1740 @param ninfo: the node to check
1741 @param nresult: the remote results for the node
1742 @param nimg: the node image object
1746 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1748 remote_os = nresult.get(constants.NV_OSLIST, None)
1749 test = (not isinstance(remote_os, list) or
1750 not compat.all(isinstance(v, list) and len(v) == 7
1751 for v in remote_os))
1753 _ErrorIf(test, self.ENODEOS, node,
1754 "node hasn't returned valid OS data")
1763 for (name, os_path, status, diagnose,
1764 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1766 if name not in os_dict:
1769 # parameters is a list of lists instead of list of tuples due to
1770 # JSON lacking a real tuple type, fix it:
1771 parameters = [tuple(v) for v in parameters]
1772 os_dict[name].append((os_path, status, diagnose,
1773 set(variants), set(parameters), set(api_ver)))
1775 nimg.oslist = os_dict
1777 def _VerifyNodeOS(self, ninfo, nimg, base):
1778 """Verifies the node OS list.
1780 @type ninfo: L{objects.Node}
1781 @param ninfo: the node to check
1782 @param nimg: the node image object
1783 @param base: the 'template' node we match against (e.g. from the master)
1787 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1789 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1791 for os_name, os_data in nimg.oslist.items():
1792 assert os_data, "Empty OS status for OS %s?!" % os_name
1793 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1794 _ErrorIf(not f_status, self.ENODEOS, node,
1795 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1796 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1797 "OS '%s' has multiple entries (first one shadows the rest): %s",
1798 os_name, utils.CommaJoin([v[0] for v in os_data]))
1799 # this will catched in backend too
1800 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1801 and not f_var, self.ENODEOS, node,
1802 "OS %s with API at least %d does not declare any variant",
1803 os_name, constants.OS_API_V15)
1804 # comparisons with the 'base' image
1805 test = os_name not in base.oslist
1806 _ErrorIf(test, self.ENODEOS, node,
1807 "Extra OS %s not present on reference node (%s)",
1811 assert base.oslist[os_name], "Base node has empty OS status?"
1812 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1814 # base OS is invalid, skipping
1816 for kind, a, b in [("API version", f_api, b_api),
1817 ("variants list", f_var, b_var),
1818 ("parameters", f_param, b_param)]:
1819 _ErrorIf(a != b, self.ENODEOS, node,
1820 "OS %s %s differs from reference node %s: %s vs. %s",
1821 kind, os_name, base.name,
1822 utils.CommaJoin(a), utils.CommaJoin(b))
1824 # check any missing OSes
1825 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1826 _ErrorIf(missing, self.ENODEOS, node,
1827 "OSes present on reference node %s but missing on this node: %s",
1828 base.name, utils.CommaJoin(missing))
1830 def _VerifyOob(self, ninfo, nresult):
1831 """Verifies out of band functionality of a node.
1833 @type ninfo: L{objects.Node}
1834 @param ninfo: the node to check
1835 @param nresult: the remote results for the node
1839 # We just have to verify the paths on master and/or master candidates
1840 # as the oob helper is invoked on the master
1841 if ((ninfo.master_candidate or ninfo.master_capable) and
1842 constants.NV_OOB_PATHS in nresult):
1843 for path_result in nresult[constants.NV_OOB_PATHS]:
1844 self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1846 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1847 """Verifies and updates the node volume data.
1849 This function will update a L{NodeImage}'s internal structures
1850 with data from the remote call.
1852 @type ninfo: L{objects.Node}
1853 @param ninfo: the node to check
1854 @param nresult: the remote results for the node
1855 @param nimg: the node image object
1856 @param vg_name: the configured VG name
1860 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1862 nimg.lvm_fail = True
1863 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1866 elif isinstance(lvdata, basestring):
1867 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1868 utils.SafeEncode(lvdata))
1869 elif not isinstance(lvdata, dict):
1870 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1872 nimg.volumes = lvdata
1873 nimg.lvm_fail = False
1875 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1876 """Verifies and updates the node instance list.
1878 If the listing was successful, then updates this node's instance
1879 list. Otherwise, it marks the RPC call as failed for the instance
1882 @type ninfo: L{objects.Node}
1883 @param ninfo: the node to check
1884 @param nresult: the remote results for the node
1885 @param nimg: the node image object
1888 idata = nresult.get(constants.NV_INSTANCELIST, None)
1889 test = not isinstance(idata, list)
1890 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1891 " (instancelist): %s", utils.SafeEncode(str(idata)))
1893 nimg.hyp_fail = True
1895 nimg.instances = idata
1897 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1898 """Verifies and computes a node information map
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
1904 @param vg_name: the configured VG name
1908 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1910 # try to read free memory (from the hypervisor)
1911 hv_info = nresult.get(constants.NV_HVINFO, None)
1912 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1913 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1916 nimg.mfree = int(hv_info["memory_free"])
1917 except (ValueError, TypeError):
1918 _ErrorIf(True, self.ENODERPC, node,
1919 "node returned invalid nodeinfo, check hypervisor")
1921 # FIXME: devise a free space model for file based instances as well
1922 if vg_name is not None:
1923 test = (constants.NV_VGLIST not in nresult or
1924 vg_name not in nresult[constants.NV_VGLIST])
1925 _ErrorIf(test, self.ENODELVM, node,
1926 "node didn't return data for the volume group '%s'"
1927 " - it is either missing or broken", vg_name)
1930 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1931 except (ValueError, TypeError):
1932 _ErrorIf(True, self.ENODERPC, node,
1933 "node returned invalid LVM info, check LVM status")
1935 def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1936 """Gets per-disk status information for all instances.
1938 @type nodelist: list of strings
1939 @param nodelist: Node names
1940 @type node_image: dict of (name, L{objects.Node})
1941 @param node_image: Node objects
1942 @type instanceinfo: dict of (name, L{objects.Instance})
1943 @param instanceinfo: Instance objects
1944 @rtype: {instance: {node: [(succes, payload)]}}
1945 @return: a dictionary of per-instance dictionaries with nodes as
1946 keys and disk information as values; the disk information is a
1947 list of tuples (success, payload)
1950 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1953 node_disks_devonly = {}
1954 diskless_instances = set()
1955 diskless = constants.DT_DISKLESS
1957 for nname in nodelist:
1958 node_instances = list(itertools.chain(node_image[nname].pinst,
1959 node_image[nname].sinst))
1960 diskless_instances.update(inst for inst in node_instances
1961 if instanceinfo[inst].disk_template == diskless)
1962 disks = [(inst, disk)
1963 for inst in node_instances
1964 for disk in instanceinfo[inst].disks]
1967 # No need to collect data
1970 node_disks[nname] = disks
1972 # Creating copies as SetDiskID below will modify the objects and that can
1973 # lead to incorrect data returned from nodes
1974 devonly = [dev.Copy() for (_, dev) in disks]
1977 self.cfg.SetDiskID(dev, nname)
1979 node_disks_devonly[nname] = devonly
1981 assert len(node_disks) == len(node_disks_devonly)
1983 # Collect data from all nodes with disks
1984 result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
1987 assert len(result) == len(node_disks)
1991 for (nname, nres) in result.items():
1992 disks = node_disks[nname]
1995 # No data from this node
1996 data = len(disks) * [(False, "node offline")]
1999 _ErrorIf(msg, self.ENODERPC, nname,
2000 "while getting disk information: %s", msg)
2002 # No data from this node
2003 data = len(disks) * [(False, msg)]
2006 for idx, i in enumerate(nres.payload):
2007 if isinstance(i, (tuple, list)) and len(i) == 2:
2010 logging.warning("Invalid result from node %s, entry %d: %s",
2012 data.append((False, "Invalid result from the remote node"))
2014 for ((inst, _), status) in zip(disks, data):
2015 instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2017 # Add empty entries for diskless instances.
2018 for inst in diskless_instances:
2019 assert inst not in instdisk
2022 assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2023 len(nnames) <= len(instanceinfo[inst].all_nodes) and
2024 compat.all(isinstance(s, (tuple, list)) and
2025 len(s) == 2 for s in statuses)
2026 for inst, nnames in instdisk.items()
2027 for nname, statuses in nnames.items())
2028 assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2032 def BuildHooksEnv(self):
2035 Cluster-Verify hooks just ran in the post phase and their failure makes
2036 the output be logged in the verify output and the verification to fail.
2039 all_nodes = self.cfg.GetNodeList()
2041 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2043 for node in self.cfg.GetAllNodesInfo().values():
2044 env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2046 return env, [], all_nodes
2048 def Exec(self, feedback_fn):
2049 """Verify integrity of cluster, performing various test on nodes.
2052 # This method has too many local variables. pylint: disable-msg=R0914
2054 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2055 verbose = self.op.verbose
2056 self._feedback_fn = feedback_fn
2057 feedback_fn("* Verifying global settings")
2058 for msg in self.cfg.VerifyConfig():
2059 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2061 # Check the cluster certificates
2062 for cert_filename in constants.ALL_CERT_FILES:
2063 (errcode, msg) = _VerifyCertificate(cert_filename)
2064 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2066 vg_name = self.cfg.GetVGName()
2067 drbd_helper = self.cfg.GetDRBDHelper()
2068 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2069 cluster = self.cfg.GetClusterInfo()
2070 nodelist = utils.NiceSort(self.cfg.GetNodeList())
2071 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2072 nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2073 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2074 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2075 for iname in instancelist)
2076 groupinfo = self.cfg.GetAllNodeGroupsInfo()
2077 i_non_redundant = [] # Non redundant instances
2078 i_non_a_balanced = [] # Non auto-balanced instances
2079 n_offline = 0 # Count of offline nodes
2080 n_drained = 0 # Count of nodes being drained
2081 node_vol_should = {}
2083 # FIXME: verify OS list
2084 # do local checksums
2085 master_files = [constants.CLUSTER_CONF_FILE]
2086 master_node = self.master_node = self.cfg.GetMasterNode()
2087 master_ip = self.cfg.GetMasterIP()
2089 file_names = ssconf.SimpleStore().GetFileList()
2090 file_names.extend(constants.ALL_CERT_FILES)
2091 file_names.extend(master_files)
2092 if cluster.modify_etc_hosts:
2093 file_names.append(constants.ETC_HOSTS)
2095 local_checksums = utils.FingerprintFiles(file_names)
2097 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2098 node_verify_param = {
2099 constants.NV_FILELIST: file_names,
2100 constants.NV_NODELIST: [node.name for node in nodeinfo
2101 if not node.offline],
2102 constants.NV_HYPERVISOR: hypervisors,
2103 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2104 node.secondary_ip) for node in nodeinfo
2105 if not node.offline],
2106 constants.NV_INSTANCELIST: hypervisors,
2107 constants.NV_VERSION: None,
2108 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2109 constants.NV_NODESETUP: None,
2110 constants.NV_TIME: None,
2111 constants.NV_MASTERIP: (master_node, master_ip),
2112 constants.NV_OSLIST: None,
2113 constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2116 if vg_name is not None:
2117 node_verify_param[constants.NV_VGLIST] = None
2118 node_verify_param[constants.NV_LVLIST] = vg_name
2119 node_verify_param[constants.NV_PVLIST] = [vg_name]
2120 node_verify_param[constants.NV_DRBDLIST] = None
2123 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2125 # Build our expected cluster state
2126 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2128 vm_capable=node.vm_capable))
2129 for node in nodeinfo)
2133 for node in nodeinfo:
2134 path = _SupportsOob(self.cfg, node)
2135 if path and path not in oob_paths:
2136 oob_paths.append(path)
2139 node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2141 for instance in instancelist:
2142 inst_config = instanceinfo[instance]
2144 for nname in inst_config.all_nodes:
2145 if nname not in node_image:
2147 gnode = self.NodeImage(name=nname)
2149 node_image[nname] = gnode
2151 inst_config.MapLVsByNode(node_vol_should)
2153 pnode = inst_config.primary_node
2154 node_image[pnode].pinst.append(instance)
2156 for snode in inst_config.secondary_nodes:
2157 nimg = node_image[snode]
2158 nimg.sinst.append(instance)
2159 if pnode not in nimg.sbp:
2160 nimg.sbp[pnode] = []
2161 nimg.sbp[pnode].append(instance)
2163 # At this point, we have the in-memory data structures complete,
2164 # except for the runtime information, which we'll gather next
2166 # Due to the way our RPC system works, exact response times cannot be
2167 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2168 # time before and after executing the request, we can at least have a time
2170 nvinfo_starttime = time.time()
2171 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2172 self.cfg.GetClusterName())
2173 nvinfo_endtime = time.time()
2175 all_drbd_map = self.cfg.ComputeDRBDMap()
2177 feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2178 instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2180 feedback_fn("* Verifying node status")
2184 for node_i in nodeinfo:
2186 nimg = node_image[node]
2190 feedback_fn("* Skipping offline node %s" % (node,))
2194 if node == master_node:
2196 elif node_i.master_candidate:
2197 ntype = "master candidate"
2198 elif node_i.drained:
2204 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2206 msg = all_nvinfo[node].fail_msg
2207 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2209 nimg.rpc_fail = True
2212 nresult = all_nvinfo[node].payload
2214 nimg.call_ok = self._VerifyNode(node_i, nresult)
2215 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2216 self._VerifyNodeNetwork(node_i, nresult)
2217 self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2220 self._VerifyOob(node_i, nresult)
2223 self._VerifyNodeLVM(node_i, nresult, vg_name)
2224 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2227 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2228 self._UpdateNodeInstances(node_i, nresult, nimg)
2229 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2230 self._UpdateNodeOS(node_i, nresult, nimg)
2231 if not nimg.os_fail:
2232 if refos_img is None:
2234 self._VerifyNodeOS(node_i, nimg, refos_img)
2236 feedback_fn("* Verifying instance status")
2237 for instance in instancelist:
2239 feedback_fn("* Verifying instance %s" % instance)
2240 inst_config = instanceinfo[instance]
2241 self._VerifyInstance(instance, inst_config, node_image,
2243 inst_nodes_offline = []
2245 pnode = inst_config.primary_node
2246 pnode_img = node_image[pnode]
2247 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2248 self.ENODERPC, pnode, "instance %s, connection to"
2249 " primary node failed", instance)
2251 if pnode_img.offline:
2252 inst_nodes_offline.append(pnode)
2254 # If the instance is non-redundant we cannot survive losing its primary
2255 # node, so we are not N+1 compliant. On the other hand we have no disk
2256 # templates with more than one secondary so that situation is not well
2258 # FIXME: does not support file-backed instances
2259 if not inst_config.secondary_nodes:
2260 i_non_redundant.append(instance)
2262 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2263 instance, "instance has multiple secondary nodes: %s",
2264 utils.CommaJoin(inst_config.secondary_nodes),
2265 code=self.ETYPE_WARNING)
2267 if inst_config.disk_template in constants.DTS_NET_MIRROR:
2268 pnode = inst_config.primary_node
2269 instance_nodes = utils.NiceSort(inst_config.all_nodes)
2270 instance_groups = {}
2272 for node in instance_nodes:
2273 instance_groups.setdefault(nodeinfo_byname[node].group,
2277 "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2278 # Sort so that we always list the primary node first.
2279 for group, nodes in sorted(instance_groups.items(),
2280 key=lambda (_, nodes): pnode in nodes,
2283 self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2284 instance, "instance has primary and secondary nodes in"
2285 " different groups: %s", utils.CommaJoin(pretty_list),
2286 code=self.ETYPE_WARNING)
2288 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2289 i_non_a_balanced.append(instance)
2291 for snode in inst_config.secondary_nodes:
2292 s_img = node_image[snode]
2293 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2294 "instance %s, connection to secondary node failed", instance)
2297 inst_nodes_offline.append(snode)
2299 # warn that the instance lives on offline nodes
2300 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2301 "instance lives on offline node(s) %s",
2302 utils.CommaJoin(inst_nodes_offline))
2303 # ... or ghost/non-vm_capable nodes
2304 for node in inst_config.all_nodes:
2305 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2306 "instance lives on ghost node %s", node)
2307 _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2308 instance, "instance lives on non-vm_capable node %s", node)
2310 feedback_fn("* Verifying orphan volumes")
2311 reserved = utils.FieldSet(*cluster.reserved_lvs)
2312 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2314 feedback_fn("* Verifying orphan instances")
2315 self._VerifyOrphanInstances(instancelist, node_image)
2317 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2318 feedback_fn("* Verifying N+1 Memory redundancy")
2319 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2321 feedback_fn("* Other Notes")
2323 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2324 % len(i_non_redundant))
2326 if i_non_a_balanced:
2327 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2328 % len(i_non_a_balanced))
2331 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2334 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2338 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2339 """Analyze the post-hooks' result
2341 This method analyses the hook result, handles it, and sends some
2342 nicely-formatted feedback back to the user.
2344 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2345 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2346 @param hooks_results: the results of the multi-node hooks rpc call
2347 @param feedback_fn: function used send feedback back to the caller
2348 @param lu_result: previous Exec result
2349 @return: the new Exec result, based on the previous result
2353 # We only really run POST phase hooks, and are only interested in
2355 if phase == constants.HOOKS_PHASE_POST:
2356 # Used to change hooks' output to proper indentation
2357 feedback_fn("* Hooks Results")
2358 assert hooks_results, "invalid result from hooks"
2360 for node_name in hooks_results:
2361 res = hooks_results[node_name]
2363 test = msg and not res.offline
2364 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2365 "Communication failure in hooks execution: %s", msg)
2366 if res.offline or msg:
2367 # No need to investigate payload if node is offline or gave an error.
2368 # override manually lu_result here as _ErrorIf only
2369 # overrides self.bad
2372 for script, hkr, output in res.payload:
2373 test = hkr == constants.HKR_FAIL
2374 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2375 "Script %s failed, output:", script)
2377 output = self._HOOKS_INDENT_RE.sub(' ', output)
2378 feedback_fn("%s" % output)
2384 class LUClusterVerifyDisks(NoHooksLU):
2385 """Verifies the cluster disks status.
2390 def ExpandNames(self):
2391 self.needed_locks = {
2392 locking.LEVEL_NODE: locking.ALL_SET,
2393 locking.LEVEL_INSTANCE: locking.ALL_SET,
2395 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2397 def Exec(self, feedback_fn):
2398 """Verify integrity of cluster disks.
2400 @rtype: tuple of three items
2401 @return: a tuple of (dict of node-to-node_error, list of instances
2402 which need activate-disks, dict of instance: (node, volume) for
2406 result = res_nodes, res_instances, res_missing = {}, [], {}
2408 nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2409 instances = [self.cfg.GetInstanceInfo(name)
2410 for name in self.cfg.GetInstanceList()]
2413 for inst in instances:
2415 if (not inst.admin_up or
2416 inst.disk_template not in constants.DTS_NET_MIRROR):
2418 inst.MapLVsByNode(inst_lvs)
2419 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2420 for node, vol_list in inst_lvs.iteritems():
2421 for vol in vol_list:
2422 nv_dict[(node, vol)] = inst
2427 vg_names = self.rpc.call_vg_list(nodes)
2429 vg_names[node].Raise("Cannot get list of VGs")
2433 node_res = self.rpc.call_lv_list([node],
2434 vg_names[node].payload.keys())[node]
2435 if node_res.offline:
2437 msg = node_res.fail_msg
2439 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2440 res_nodes[node] = msg
2443 lvs = node_res.payload
2444 for lv_name, (_, _, lv_online) in lvs.items():
2445 inst = nv_dict.pop((node, lv_name), None)
2446 if (not lv_online and inst is not None
2447 and inst.name not in res_instances):
2448 res_instances.append(inst.name)
2450 # any leftover items in nv_dict are missing LVs, let's arrange the
2452 for key, inst in nv_dict.iteritems():
2453 if inst.name not in res_missing:
2454 res_missing[inst.name] = []
2455 res_missing[inst.name].append(key)
2460 class LUClusterRepairDiskSizes(NoHooksLU):
2461 """Verifies the cluster disks sizes.
2466 def ExpandNames(self):
2467 if self.op.instances:
2468 self.wanted_names = []
2469 for name in self.op.instances:
2470 full_name = _ExpandInstanceName(self.cfg, name)
2471 self.wanted_names.append(full_name)
2472 self.needed_locks = {
2473 locking.LEVEL_NODE: [],
2474 locking.LEVEL_INSTANCE: self.wanted_names,
2476 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2478 self.wanted_names = None
2479 self.needed_locks = {
2480 locking.LEVEL_NODE: locking.ALL_SET,
2481 locking.LEVEL_INSTANCE: locking.ALL_SET,
2483 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2485 def DeclareLocks(self, level):
2486 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2487 self._LockInstancesNodes(primary_only=True)
2489 def CheckPrereq(self):
2490 """Check prerequisites.
2492 This only checks the optional instance list against the existing names.
2495 if self.wanted_names is None:
2496 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2498 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2499 in self.wanted_names]
2501 def _EnsureChildSizes(self, disk):
2502 """Ensure children of the disk have the needed disk size.
2504 This is valid mainly for DRBD8 and fixes an issue where the
2505 children have smaller disk size.
2507 @param disk: an L{ganeti.objects.Disk} object
2510 if disk.dev_type == constants.LD_DRBD8:
2511 assert disk.children, "Empty children for DRBD8?"
2512 fchild = disk.children[0]
2513 mismatch = fchild.size < disk.size
2515 self.LogInfo("Child disk has size %d, parent %d, fixing",
2516 fchild.size, disk.size)
2517 fchild.size = disk.size
2519 # and we recurse on this child only, not on the metadev
2520 return self._EnsureChildSizes(fchild) or mismatch
2524 def Exec(self, feedback_fn):
2525 """Verify the size of cluster disks.
2528 # TODO: check child disks too
2529 # TODO: check differences in size between primary/secondary nodes
2531 for instance in self.wanted_instances:
2532 pnode = instance.primary_node
2533 if pnode not in per_node_disks:
2534 per_node_disks[pnode] = []
2535 for idx, disk in enumerate(instance.disks):
2536 per_node_disks[pnode].append((instance, idx, disk))
2539 for node, dskl in per_node_disks.items():
2540 newl = [v[2].Copy() for v in dskl]
2542 self.cfg.SetDiskID(dsk, node)
2543 result = self.rpc.call_blockdev_getsizes(node, newl)
2545 self.LogWarning("Failure in blockdev_getsizes call to node"
2546 " %s, ignoring", node)
2548 if len(result.data) != len(dskl):
2549 self.LogWarning("Invalid result from node %s, ignoring node results",
2552 for ((instance, idx, disk), size) in zip(dskl, result.data):
2554 self.LogWarning("Disk %d of instance %s did not return size"
2555 " information, ignoring", idx, instance.name)
2557 if not isinstance(size, (int, long)):
2558 self.LogWarning("Disk %d of instance %s did not return valid"
2559 " size information, ignoring", idx, instance.name)
2562 if size != disk.size:
2563 self.LogInfo("Disk %d of instance %s has mismatched size,"
2564 " correcting: recorded %d, actual %d", idx,
2565 instance.name, disk.size, size)
2567 self.cfg.Update(instance, feedback_fn)
2568 changed.append((instance.name, idx, size))
2569 if self._EnsureChildSizes(disk):
2570 self.cfg.Update(instance, feedback_fn)
2571 changed.append((instance.name, idx, disk.size))
2575 class LUClusterRename(LogicalUnit):
2576 """Rename the cluster.
2579 HPATH = "cluster-rename"
2580 HTYPE = constants.HTYPE_CLUSTER
2582 def BuildHooksEnv(self):
2587 "OP_TARGET": self.cfg.GetClusterName(),
2588 "NEW_NAME": self.op.name,
2590 mn = self.cfg.GetMasterNode()
2591 all_nodes = self.cfg.GetNodeList()
2592 return env, [mn], all_nodes
2594 def CheckPrereq(self):
2595 """Verify that the passed name is a valid one.
2598 hostname = netutils.GetHostname(name=self.op.name,
2599 family=self.cfg.GetPrimaryIPFamily())
2601 new_name = hostname.name
2602 self.ip = new_ip = hostname.ip
2603 old_name = self.cfg.GetClusterName()
2604 old_ip = self.cfg.GetMasterIP()
2605 if new_name == old_name and new_ip == old_ip:
2606 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2607 " cluster has changed",
2609 if new_ip != old_ip:
2610 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2611 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2612 " reachable on the network" %
2613 new_ip, errors.ECODE_NOTUNIQUE)
2615 self.op.name = new_name
2617 def Exec(self, feedback_fn):
2618 """Rename the cluster.
2621 clustername = self.op.name
2624 # shutdown the master IP
2625 master = self.cfg.GetMasterNode()
2626 result = self.rpc.call_node_stop_master(master, False)
2627 result.Raise("Could not disable the master role")
2630 cluster = self.cfg.GetClusterInfo()
2631 cluster.cluster_name = clustername
2632 cluster.master_ip = ip
2633 self.cfg.Update(cluster, feedback_fn)
2635 # update the known hosts file
2636 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2637 node_list = self.cfg.GetOnlineNodeList()
2639 node_list.remove(master)
2642 _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2644 result = self.rpc.call_node_start_master(master, False, False)
2645 msg = result.fail_msg
2647 self.LogWarning("Could not re-enable the master role on"
2648 " the master, please restart manually: %s", msg)
2653 class LUClusterSetParams(LogicalUnit):
2654 """Change the parameters of the cluster.
2657 HPATH = "cluster-modify"
2658 HTYPE = constants.HTYPE_CLUSTER
2661 def CheckArguments(self):
2665 if self.op.uid_pool:
2666 uidpool.CheckUidPool(self.op.uid_pool)
2668 if self.op.add_uids:
2669 uidpool.CheckUidPool(self.op.add_uids)
2671 if self.op.remove_uids:
2672 uidpool.CheckUidPool(self.op.remove_uids)
2674 def ExpandNames(self):
2675 # FIXME: in the future maybe other cluster params won't require checking on
2676 # all nodes to be modified.
2677 self.needed_locks = {
2678 locking.LEVEL_NODE: locking.ALL_SET,
2680 self.share_locks[locking.LEVEL_NODE] = 1
2682 def BuildHooksEnv(self):
2687 "OP_TARGET": self.cfg.GetClusterName(),
2688 "NEW_VG_NAME": self.op.vg_name,
2690 mn = self.cfg.GetMasterNode()
2691 return env, [mn], [mn]
2693 def CheckPrereq(self):
2694 """Check prerequisites.
2696 This checks whether the given params don't conflict and
2697 if the given volume group is valid.
2700 if self.op.vg_name is not None and not self.op.vg_name:
2701 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2702 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2703 " instances exist", errors.ECODE_INVAL)
2705 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2706 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2707 raise errors.OpPrereqError("Cannot disable drbd helper while"
2708 " drbd-based instances exist",
2711 node_list = self.acquired_locks[locking.LEVEL_NODE]
2713 # if vg_name not None, checks given volume group on all nodes
2715 vglist = self.rpc.call_vg_list(node_list)
2716 for node in node_list:
2717 msg = vglist[node].fail_msg
2719 # ignoring down node
2720 self.LogWarning("Error while gathering data on node %s"
2721 " (ignoring node): %s", node, msg)
2723 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2725 constants.MIN_VG_SIZE)
2727 raise errors.OpPrereqError("Error on node '%s': %s" %
2728 (node, vgstatus), errors.ECODE_ENVIRON)
2730 if self.op.drbd_helper:
2731 # checks given drbd helper on all nodes
2732 helpers = self.rpc.call_drbd_helper(node_list)
2733 for node in node_list:
2734 ninfo = self.cfg.GetNodeInfo(node)
2736 self.LogInfo("Not checking drbd helper on offline node %s", node)
2738 msg = helpers[node].fail_msg
2740 raise errors.OpPrereqError("Error checking drbd helper on node"
2741 " '%s': %s" % (node, msg),
2742 errors.ECODE_ENVIRON)
2743 node_helper = helpers[node].payload
2744 if node_helper != self.op.drbd_helper:
2745 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2746 (node, node_helper), errors.ECODE_ENVIRON)
2748 self.cluster = cluster = self.cfg.GetClusterInfo()
2749 # validate params changes
2750 if self.op.beparams:
2751 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2752 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2754 if self.op.ndparams:
2755 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2756 self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2758 if self.op.nicparams:
2759 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2760 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2761 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2764 # check all instances for consistency
2765 for instance in self.cfg.GetAllInstancesInfo().values():
2766 for nic_idx, nic in enumerate(instance.nics):
2767 params_copy = copy.deepcopy(nic.nicparams)
2768 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2770 # check parameter syntax
2772 objects.NIC.CheckParameterSyntax(params_filled)
2773 except errors.ConfigurationError, err:
2774 nic_errors.append("Instance %s, nic/%d: %s" %
2775 (instance.name, nic_idx, err))
2777 # if we're moving instances to routed, check that they have an ip
2778 target_mode = params_filled[constants.NIC_MODE]
2779 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2780 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2781 (instance.name, nic_idx))
2783 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2784 "\n".join(nic_errors))
2786 # hypervisor list/parameters
2787 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2788 if self.op.hvparams:
2789 for hv_name, hv_dict in self.op.hvparams.items():
2790 if hv_name not in self.new_hvparams:
2791 self.new_hvparams[hv_name] = hv_dict
2793 self.new_hvparams[hv_name].update(hv_dict)
2795 # os hypervisor parameters
2796 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2798 for os_name, hvs in self.op.os_hvp.items():
2799 if os_name not in self.new_os_hvp:
2800 self.new_os_hvp[os_name] = hvs
2802 for hv_name, hv_dict in hvs.items():
2803 if hv_name not in self.new_os_hvp[os_name]:
2804 self.new_os_hvp[os_name][hv_name] = hv_dict
2806 self.new_os_hvp[os_name][hv_name].update(hv_dict)
2809 self.new_osp = objects.FillDict(cluster.osparams, {})
2810 if self.op.osparams:
2811 for os_name, osp in self.op.osparams.items():
2812 if os_name not in self.new_osp:
2813 self.new_osp[os_name] = {}
2815 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2818 if not self.new_osp[os_name]:
2819 # we removed all parameters
2820 del self.new_osp[os_name]
2822 # check the parameter validity (remote check)
2823 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2824 os_name, self.new_osp[os_name])
2826 # changes to the hypervisor list
2827 if self.op.enabled_hypervisors is not None:
2828 self.hv_list = self.op.enabled_hypervisors
2829 for hv in self.hv_list:
2830 # if the hypervisor doesn't already exist in the cluster
2831 # hvparams, we initialize it to empty, and then (in both
2832 # cases) we make sure to fill the defaults, as we might not
2833 # have a complete defaults list if the hypervisor wasn't
2835 if hv not in new_hvp:
2837 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2838 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2840 self.hv_list = cluster.enabled_hypervisors
2842 if self.op.hvparams or self.op.enabled_hypervisors is not None:
2843 # either the enabled list has changed, or the parameters have, validate
2844 for hv_name, hv_params in self.new_hvparams.items():
2845 if ((self.op.hvparams and hv_name in self.op.hvparams) or
2846 (self.op.enabled_hypervisors and
2847 hv_name in self.op.enabled_hypervisors)):
2848 # either this is a new hypervisor, or its parameters have changed
2849 hv_class = hypervisor.GetHypervisor(hv_name)
2850 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2851 hv_class.CheckParameterSyntax(hv_params)
2852 _CheckHVParams(self, node_list, hv_name, hv_params)
2855 # no need to check any newly-enabled hypervisors, since the
2856 # defaults have already been checked in the above code-block
2857 for os_name, os_hvp in self.new_os_hvp.items():
2858 for hv_name, hv_params in os_hvp.items():
2859 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2860 # we need to fill in the new os_hvp on top of the actual hv_p
2861 cluster_defaults = self.new_hvparams.get(hv_name, {})
2862 new_osp = objects.FillDict(cluster_defaults, hv_params)
2863 hv_class = hypervisor.GetHypervisor(hv_name)
2864 hv_class.CheckParameterSyntax(new_osp)
2865 _CheckHVParams(self, node_list, hv_name, new_osp)
2867 if self.op.default_iallocator:
2868 alloc_script = utils.FindFile(self.op.default_iallocator,
2869 constants.IALLOCATOR_SEARCH_PATH,
2871 if alloc_script is None:
2872 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2873 " specified" % self.op.default_iallocator,
2876 def Exec(self, feedback_fn):
2877 """Change the parameters of the cluster.
2880 if self.op.vg_name is not None:
2881 new_volume = self.op.vg_name
2884 if new_volume != self.cfg.GetVGName():
2885 self.cfg.SetVGName(new_volume)
2887 feedback_fn("Cluster LVM configuration already in desired"
2888 " state, not changing")
2889 if self.op.drbd_helper is not None:
2890 new_helper = self.op.drbd_helper
2893 if new_helper != self.cfg.GetDRBDHelper():
2894 self.cfg.SetDRBDHelper(new_helper)
2896 feedback_fn("Cluster DRBD helper already in desired state,"
2898 if self.op.hvparams:
2899 self.cluster.hvparams = self.new_hvparams
2901 self.cluster.os_hvp = self.new_os_hvp
2902 if self.op.enabled_hypervisors is not None:
2903 self.cluster.hvparams = self.new_hvparams
2904 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2905 if self.op.beparams:
2906 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2907 if self.op.nicparams:
2908 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2909 if self.op.osparams:
2910 self.cluster.osparams = self.new_osp
2911 if self.op.ndparams:
2912 self.cluster.ndparams = self.new_ndparams
2914 if self.op.candidate_pool_size is not None:
2915 self.cluster.candidate_pool_size = self.op.candidate_pool_size
2916 # we need to update the pool size here, otherwise the save will fail
2917 _AdjustCandidatePool(self, [])
2919 if self.op.maintain_node_health is not None:
2920 self.cluster.maintain_node_health = self.op.maintain_node_health
2922 if self.op.prealloc_wipe_disks is not None:
2923 self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2925 if self.op.add_uids is not None:
2926 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2928 if self.op.remove_uids is not None:
2929 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2931 if self.op.uid_pool is not None:
2932 self.cluster.uid_pool = self.op.uid_pool
2934 if self.op.default_iallocator is not None:
2935 self.cluster.default_iallocator = self.op.default_iallocator
2937 if self.op.reserved_lvs is not None:
2938 self.cluster.reserved_lvs = self.op.reserved_lvs
2940 def helper_os(aname, mods, desc):
2942 lst = getattr(self.cluster, aname)
2943 for key, val in mods:
2944 if key == constants.DDM_ADD:
2946 feedback_fn("OS %s already in %s, ignoring" % (val, desc))
2949 elif key == constants.DDM_REMOVE:
2953 feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
2955 raise errors.ProgrammerError("Invalid modification '%s'" % key)
2957 if self.op.hidden_os:
2958 helper_os("hidden_os", self.op.hidden_os, "hidden")
2960 if self.op.blacklisted_os:
2961 helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
2963 if self.op.master_netdev:
2964 master = self.cfg.GetMasterNode()
2965 feedback_fn("Shutting down master ip on the current netdev (%s)" %
2966 self.cluster.master_netdev)
2967 result = self.rpc.call_node_stop_master(master, False)
2968 result.Raise("Could not disable the master ip")
2969 feedback_fn("Changing master_netdev from %s to %s" %
2970 (self.cluster.master_netdev, self.op.master_netdev))
2971 self.cluster.master_netdev = self.op.master_netdev
2973 self.cfg.Update(self.cluster, feedback_fn)
2975 if self.op.master_netdev:
2976 feedback_fn("Starting the master ip on the new master netdev (%s)" %
2977 self.op.master_netdev)
2978 result = self.rpc.call_node_start_master(master, False, False)
2980 self.LogWarning("Could not re-enable the master ip on"
2981 " the master, please restart manually: %s",
2985 def _UploadHelper(lu, nodes, fname):
2986 """Helper for uploading a file and showing warnings.
2989 if os.path.exists(fname):
2990 result = lu.rpc.call_upload_file(nodes, fname)
2991 for to_node, to_result in result.items():
2992 msg = to_result.fail_msg
2994 msg = ("Copy of file %s to node %s failed: %s" %
2995 (fname, to_node, msg))
2996 lu.proc.LogWarning(msg)
2999 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3000 """Distribute additional files which are part of the cluster configuration.
3002 ConfigWriter takes care of distributing the config and ssconf files, but
3003 there are more files which should be distributed to all nodes. This function
3004 makes sure those are copied.
3006 @param lu: calling logical unit
3007 @param additional_nodes: list of nodes not in the config to distribute to
3008 @type additional_vm: boolean
3009 @param additional_vm: whether the additional nodes are vm-capable or not
3012 # 1. Gather target nodes
3013 myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3014 dist_nodes = lu.cfg.GetOnlineNodeList()
3015 nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3016 vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3017 if additional_nodes is not None:
3018 dist_nodes.extend(additional_nodes)
3020 vm_nodes.extend(additional_nodes)
3021 if myself.name in dist_nodes:
3022 dist_nodes.remove(myself.name)
3023 if myself.name in vm_nodes:
3024 vm_nodes.remove(myself.name)
3026 # 2. Gather files to distribute
3027 dist_files = set([constants.ETC_HOSTS,
3028 constants.SSH_KNOWN_HOSTS_FILE,
3029 constants.RAPI_CERT_FILE,
3030 constants.RAPI_USERS_FILE,
3031 constants.CONFD_HMAC_KEY,
3032 constants.CLUSTER_DOMAIN_SECRET_FILE,
3036 enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3037 for hv_name in enabled_hypervisors:
3038 hv_class = hypervisor.GetHypervisor(hv_name)
3039 vm_files.update(hv_class.GetAncillaryFiles())
3041 # 3. Perform the files upload
3042 for fname in dist_files:
3043 _UploadHelper(lu, dist_nodes, fname)
3044 for fname in vm_files:
3045 _UploadHelper(lu, vm_nodes, fname)
3048 class LUClusterRedistConf(NoHooksLU):
3049 """Force the redistribution of cluster configuration.
3051 This is a very simple LU.
3056 def ExpandNames(self):
3057 self.needed_locks = {
3058 locking.LEVEL_NODE: locking.ALL_SET,
3060 self.share_locks[locking.LEVEL_NODE] = 1
3062 def Exec(self, feedback_fn):
3063 """Redistribute the configuration.
3066 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3067 _RedistributeAncillaryFiles(self)
3070 def _WaitForSync(lu, instance, disks=None, oneshot=False):
3071 """Sleep and poll for an instance's disk to sync.
3074 if not instance.disks or disks is not None and not disks:
3077 disks = _ExpandCheckDisks(instance, disks)
3080 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3082 node = instance.primary_node
3085 lu.cfg.SetDiskID(dev, node)
3087 # TODO: Convert to utils.Retry
3090 degr_retries = 10 # in seconds, as we sleep 1 second each time
3094 cumul_degraded = False
3095 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3096 msg = rstats.fail_msg
3098 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3101 raise errors.RemoteError("Can't contact node %s for mirror data,"
3102 " aborting." % node)
3105 rstats = rstats.payload
3107 for i, mstat in enumerate(rstats):
3109 lu.LogWarning("Can't compute data for node %s/%s",
3110 node, disks[i].iv_name)
3113 cumul_degraded = (cumul_degraded or
3114 (mstat.is_degraded and mstat.sync_percent is None))
3115 if mstat.sync_percent is not None:
3117 if mstat.estimated_time is not None:
3118 rem_time = ("%s remaining (estimated)" %
3119 utils.FormatSeconds(mstat.estimated_time))
3120 max_time = mstat.estimated_time
3122 rem_time = "no time estimate"
3123 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3124 (disks[i].iv_name, mstat.sync_percent, rem_time))
3126 # if we're done but degraded, let's do a few small retries, to
3127 # make sure we see a stable and not transient situation; therefore
3128 # we force restart of the loop
3129 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3130 logging.info("Degraded disks found, %d retries left", degr_retries)
3138 time.sleep(min(60, max_time))
3141 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3142 return not cumul_degraded
3145 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3146 """Check that mirrors are not degraded.
3148 The ldisk parameter, if True, will change the test from the
3149 is_degraded attribute (which represents overall non-ok status for
3150 the device(s)) to the ldisk (representing the local storage status).
3153 lu.cfg.SetDiskID(dev, node)
3157 if on_primary or dev.AssembleOnSecondary():
3158 rstats = lu.rpc.call_blockdev_find(node, dev)
3159 msg = rstats.fail_msg
3161 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3163 elif not rstats.payload:
3164 lu.LogWarning("Can't find disk on node %s", node)
3168 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3170 result = result and not rstats.payload.is_degraded
3173 for child in dev.children:
3174 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3179 class LUOobCommand(NoHooksLU):
3180 """Logical unit for OOB handling.
3185 def CheckPrereq(self):
3186 """Check prerequisites.
3189 - the node exists in the configuration
3192 Any errors are signaled by raising errors.OpPrereqError.
3195 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3196 node = self.cfg.GetNodeInfo(self.op.node_name)
3199 raise errors.OpPrereqError("Node %s not found" % self.op.node_name)
3201 self.oob_program = _SupportsOob(self.cfg, node)
3203 if not self.oob_program:
3204 raise errors.OpPrereqError("OOB is not supported for node %s" %
3207 if self.op.command == constants.OOB_POWER_OFF and not node.offline:
3208 raise errors.OpPrereqError(("Cannot power off node %s because it is"
3209 " not marked offline") % self.op.node_name)
3213 def ExpandNames(self):
3214 """Gather locks we need.
3217 node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3218 self.needed_locks = {
3219 locking.LEVEL_NODE: [node_name],
3222 def Exec(self, feedback_fn):
3223 """Execute OOB and return result if we expect any.
3226 master_node = self.cfg.GetMasterNode()
3229 logging.info("Executing out-of-band command '%s' using '%s' on %s",
3230 self.op.command, self.oob_program, self.op.node_name)
3231 result = self.rpc.call_run_oob(master_node, self.oob_program,
3232 self.op.command, self.op.node_name,
3235 result.Raise("An error occurred on execution of OOB helper")
3237 self._CheckPayload(result)
3239 if self.op.command == constants.OOB_HEALTH:
3240 # For health we should log important events
3241 for item, status in result.payload:
3242 if status in [constants.OOB_STATUS_WARNING,
3243 constants.OOB_STATUS_CRITICAL]:
3244 logging.warning("On node '%s' item '%s' has status '%s'",
3245 self.op.node_name, item, status)
3247 if self.op.command == constants.OOB_POWER_ON:
3249 elif self.op.command == constants.OOB_POWER_OFF:
3250 node.powered = False
3251 elif self.op.command == constants.OOB_POWER_STATUS:
3252 powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3253 if powered != self.node.powered:
3254 logging.warning(("Recorded power state (%s) of node '%s' does not match"
3255 " actual power state (%s)"), node.powered,
3256 self.op.node_name, powered)
3258 self.cfg.Update(node, feedback_fn)
3260 return result.payload
3262 def _CheckPayload(self, result):
3263 """Checks if the payload is valid.
3265 @param result: RPC result
3266 @raises errors.OpExecError: If payload is not valid
3270 if self.op.command == constants.OOB_HEALTH:
3271 if not isinstance(result.payload, list):
3272 errs.append("command 'health' is expected to return a list but got %s" %
3273 type(result.payload))
3274 for item, status in result.payload:
3275 if status not in constants.OOB_STATUSES:
3276 errs.append("health item '%s' has invalid status '%s'" %
3279 if self.op.command == constants.OOB_POWER_STATUS:
3280 if not isinstance(result.payload, dict):
3281 errs.append("power-status is expected to return a dict but got %s" %
3282 type(result.payload))
3284 if self.op.command in [
3285 constants.OOB_POWER_ON,
3286 constants.OOB_POWER_OFF,
3287 constants.OOB_POWER_CYCLE,
3289 if result.payload is not None:
3290 errs.append("%s is expected to not return payload but got '%s'" %
3291 (self.op.command, result.payload))
3294 raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3295 utils.CommaJoin(errs))
3299 class LUOsDiagnose(NoHooksLU):
3300 """Logical unit for OS diagnose/query.
3305 _BLK = "blacklisted"
3307 _FIELDS_STATIC = utils.FieldSet()
3308 _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3309 "parameters", "api_versions", _HID, _BLK)
3311 def CheckArguments(self):
3313 raise errors.OpPrereqError("Selective OS query not supported",
3316 _CheckOutputFields(static=self._FIELDS_STATIC,
3317 dynamic=self._FIELDS_DYNAMIC,
3318 selected=self.op.output_fields)
3320 def ExpandNames(self):
3321 # Lock all nodes, in shared mode
3322 # Temporary removal of locks, should be reverted later
3323 # TODO: reintroduce locks when they are lighter-weight
3324 self.needed_locks = {}
3325 #self.share_locks[locking.LEVEL_NODE] = 1
3326 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3329 def _DiagnoseByOS(rlist):
3330 """Remaps a per-node return list into an a per-os per-node dictionary
3332 @param rlist: a map with node names as keys and OS objects as values
3335 @return: a dictionary with osnames as keys and as value another
3336 map, with nodes as keys and tuples of (path, status, diagnose,
3337 variants, parameters, api_versions) as values, eg::
3339 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3340 (/srv/..., False, "invalid api")],
3341 "node2": [(/srv/..., True, "", [], [])]}
3346 # we build here the list of nodes that didn't fail the RPC (at RPC
3347 # level), so that nodes with a non-responding node daemon don't
3348 # make all OSes invalid
3349 good_nodes = [node_name for node_name in rlist
3350 if not rlist[node_name].fail_msg]
3351 for node_name, nr in rlist.items():
3352 if nr.fail_msg or not nr.payload:
3354 for (name, path, status, diagnose, variants,
3355 params, api_versions) in nr.payload:
3356 if name not in all_os:
3357 # build a list of nodes for this os containing empty lists
3358 # for each node in node_list
3360 for nname in good_nodes:
3361 all_os[name][nname] = []
3362 # convert params from [name, help] to (name, help)
3363 params = [tuple(v) for v in params]
3364 all_os[name][node_name].append((path, status, diagnose,
3365 variants, params, api_versions))
3368 def Exec(self, feedback_fn):
3369 """Compute the list of OSes.
3372 valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
3373 node_data = self.rpc.call_os_diagnose(valid_nodes)
3374 pol = self._DiagnoseByOS(node_data)
3376 cluster = self.cfg.GetClusterInfo()
3378 for os_name in utils.NiceSort(pol.keys()):
3379 os_data = pol[os_name]
3382 (variants, params, api_versions) = null_state = (set(), set(), set())
3383 for idx, osl in enumerate(os_data.values()):
3384 valid = bool(valid and osl and osl[0][1])
3386 (variants, params, api_versions) = null_state
3388 node_variants, node_params, node_api = osl[0][3:6]
3389 if idx == 0: # first entry
3390 variants = set(node_variants)
3391 params = set(node_params)
3392 api_versions = set(node_api)
3393 else: # keep consistency
3394 variants.intersection_update(node_variants)
3395 params.intersection_update(node_params)
3396 api_versions.intersection_update(node_api)
3398 is_hid = os_name in cluster.hidden_os
3399 is_blk = os_name in cluster.blacklisted_os
3400 if ((self._HID not in self.op.output_fields and is_hid) or
3401 (self._BLK not in self.op.output_fields and is_blk) or
3402 (self._VLD not in self.op.output_fields and not valid)):
3405 for field in self.op.output_fields:
3408 elif field == self._VLD:
3410 elif field == "node_status":
3411 # this is just a copy of the dict
3413 for node_name, nos_list in os_data.items():
3414 val[node_name] = nos_list
3415 elif field == "variants":
3416 val = utils.NiceSort(list(variants))
3417 elif field == "parameters":
3419 elif field == "api_versions":
3420 val = list(api_versions)
3421 elif field == self._HID:
3423 elif field == self._BLK:
3426 raise errors.ParameterError(field)
3433 class LUNodeRemove(LogicalUnit):
3434 """Logical unit for removing a node.
3437 HPATH = "node-remove"
3438 HTYPE = constants.HTYPE_NODE
3440 def BuildHooksEnv(self):
3443 This doesn't run on the target node in the pre phase as a failed
3444 node would then be impossible to remove.
3448 "OP_TARGET": self.op.node_name,
3449 "NODE_NAME": self.op.node_name,
3451 all_nodes = self.cfg.GetNodeList()
3453 all_nodes.remove(self.op.node_name)
3455 logging.warning("Node %s which is about to be removed not found"
3456 " in the all nodes list", self.op.node_name)
3457 return env, all_nodes, all_nodes
3459 def CheckPrereq(self):
3460 """Check prerequisites.
3463 - the node exists in the configuration
3464 - it does not have primary or secondary instances
3465 - it's not the master
3467 Any errors are signaled by raising errors.OpPrereqError.
3470 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3471 node = self.cfg.GetNodeInfo(self.op.node_name)
3472 assert node is not None
3474 instance_list = self.cfg.GetInstanceList()
3476 masternode = self.cfg.GetMasterNode()
3477 if node.name == masternode:
3478 raise errors.OpPrereqError("Node is the master node,"
3479 " you need to failover first.",
3482 for instance_name in instance_list:
3483 instance = self.cfg.GetInstanceInfo(instance_name)
3484 if node.name in instance.all_nodes:
3485 raise errors.OpPrereqError("Instance %s is still running on the node,"
3486 " please remove first." % instance_name,
3488 self.op.node_name = node.name
3491 def Exec(self, feedback_fn):
3492 """Removes the node from the cluster.
3496 logging.info("Stopping the node daemon and removing configs from node %s",
3499 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3501 # Promote nodes to master candidate as needed
3502 _AdjustCandidatePool(self, exceptions=[node.name])
3503 self.context.RemoveNode(node.name)
3505 # Run post hooks on the node before it's removed
3506 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3508 hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3510 # pylint: disable-msg=W0702
3511 self.LogWarning("Errors occurred running hooks on %s" % node.name)
3513 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3514 msg = result.fail_msg
3516 self.LogWarning("Errors encountered on the remote node while leaving"
3517 " the cluster: %s", msg)
3519 # Remove node from our /etc/hosts
3520 if self.cfg.GetClusterInfo().modify_etc_hosts:
3521 master_node = self.cfg.GetMasterNode()
3522 result = self.rpc.call_etc_hosts_modify(master_node,
3523 constants.ETC_HOSTS_REMOVE,
3525 result.Raise("Can't update hosts file with new host data")
3526 _RedistributeAncillaryFiles(self)
3529 class _NodeQuery(_QueryBase):
3530 FIELDS = query.NODE_FIELDS
3532 def ExpandNames(self, lu):
3533 lu.needed_locks = {}
3534 lu.share_locks[locking.LEVEL_NODE] = 1
3537 self.wanted = _GetWantedNodes(lu, self.names)
3539 self.wanted = locking.ALL_SET
3541 self.do_locking = (self.use_locking and
3542 query.NQ_LIVE in self.requested_data)
3545 # if we don't request only static fields, we need to lock the nodes
3546 lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3548 def DeclareLocks(self, lu, level):
3551 def _GetQueryData(self, lu):
3552 """Computes the list of nodes and their attributes.
3555 all_info = lu.cfg.GetAllNodesInfo()
3557 nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3559 # Gather data as requested
3560 if query.NQ_LIVE in self.requested_data:
3561 node_data = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
3562 lu.cfg.GetHypervisorType())
3563 live_data = dict((name, nresult.payload)
3564 for (name, nresult) in node_data.items()
3565 if not nresult.fail_msg and nresult.payload)
3569 if query.NQ_INST in self.requested_data:
3570 node_to_primary = dict([(name, set()) for name in nodenames])
3571 node_to_secondary = dict([(name, set()) for name in nodenames])
3573 inst_data = lu.cfg.GetAllInstancesInfo()
3575 for inst in inst_data.values():
3576 if inst.primary_node in node_to_primary:
3577 node_to_primary[inst.primary_node].add(inst.name)
3578 for secnode in inst.secondary_nodes:
3579 if secnode in node_to_secondary:
3580 node_to_secondary[secnode].add(inst.name)
3582 node_to_primary = None
3583 node_to_secondary = None
3585 if query.NQ_OOB in self.requested_data:
3586 oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3587 for name, node in all_info.iteritems())
3591 if query.NQ_GROUP in self.requested_data:
3592 groups = lu.cfg.GetAllNodeGroupsInfo()
3596 return query.NodeQueryData([all_info[name] for name in nodenames],
3597 live_data, lu.cfg.GetMasterNode(),
3598 node_to_primary, node_to_secondary, groups,
3599 oob_support, lu.cfg.GetClusterInfo())
3602 class LUNodeQuery(NoHooksLU):
3603 """Logical unit for querying nodes.
3606 # pylint: disable-msg=W0142
3609 def CheckArguments(self):
3610 self.nq = _NodeQuery(self.op.names, self.op.output_fields,
3611 self.op.use_locking)
3613 def ExpandNames(self):
3614 self.nq.ExpandNames(self)
3616 def Exec(self, feedback_fn):
3617 return self.nq.OldStyleQuery(self)
3620 class LUNodeQueryvols(NoHooksLU):
3621 """Logical unit for getting volumes on node(s).
3625 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3626 _FIELDS_STATIC = utils.FieldSet("node")
3628 def CheckArguments(self):
3629 _CheckOutputFields(static=self._FIELDS_STATIC,
3630 dynamic=self._FIELDS_DYNAMIC,
3631 selected=self.op.output_fields)
3633 def ExpandNames(self):
3634 self.needed_locks = {}
3635 self.share_locks[locking.LEVEL_NODE] = 1
3636 if not self.op.nodes:
3637 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3639 self.needed_locks[locking.LEVEL_NODE] = \
3640 _GetWantedNodes(self, self.op.nodes)
3642 def Exec(self, feedback_fn):
3643 """Computes the list of nodes and their attributes.
3646 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3647 volumes = self.rpc.call_node_volumes(nodenames)
3649 ilist = [self.cfg.GetInstanceInfo(iname) for iname
3650 in self.cfg.GetInstanceList()]
3652 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3655 for node in nodenames:
3656 nresult = volumes[node]
3659 msg = nresult.fail_msg
3661 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3664 node_vols = nresult.payload[:]
3665 node_vols.sort(key=lambda vol: vol['dev'])
3667 for vol in node_vols:
3669 for field in self.op.output_fields:
3672 elif field == "phys":
3676 elif field == "name":
3678 elif field == "size":
3679 val = int(float(vol['size']))
3680 elif field == "instance":
3682 if node not in lv_by_node[inst]:
3684 if vol['name'] in lv_by_node[inst][node]:
3690 raise errors.ParameterError(field)
3691 node_output.append(str(val))
3693 output.append(node_output)
3698 class LUNodeQueryStorage(NoHooksLU):
3699 """Logical unit for getting information on storage units on node(s).
3702 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3705 def CheckArguments(self):
3706 _CheckOutputFields(static=self._FIELDS_STATIC,
3707 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3708 selected=self.op.output_fields)
3710 def ExpandNames(self):
3711 self.needed_locks = {}
3712 self.share_locks[locking.LEVEL_NODE] = 1
3715 self.needed_locks[locking.LEVEL_NODE] = \
3716 _GetWantedNodes(self, self.op.nodes)
3718 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3720 def Exec(self, feedback_fn):
3721 """Computes the list of nodes and their attributes.
3724 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3726 # Always get name to sort by
3727 if constants.SF_NAME in self.op.output_fields:
3728 fields = self.op.output_fields[:]
3730 fields = [constants.SF_NAME] + self.op.output_fields
3732 # Never ask for node or type as it's only known to the LU
3733 for extra in [constants.SF_NODE, constants.SF_TYPE]:
3734 while extra in fields:
3735 fields.remove(extra)
3737 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3738 name_idx = field_idx[constants.SF_NAME]
3740 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3741 data = self.rpc.call_storage_list(self.nodes,
3742 self.op.storage_type, st_args,
3743 self.op.name, fields)
3747 for node in utils.NiceSort(self.nodes):
3748 nresult = data[node]
3752 msg = nresult.fail_msg
3754 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3757 rows = dict([(row[name_idx], row) for row in nresult.payload])
3759 for name in utils.NiceSort(rows.keys()):
3764 for field in self.op.output_fields:
3765 if field == constants.SF_NODE:
3767 elif field == constants.SF_TYPE:
3768 val = self.op.storage_type
3769 elif field in field_idx:
3770 val = row[field_idx[field]]
3772 raise errors.ParameterError(field)
3781 class _InstanceQuery(_QueryBase):
3782 FIELDS = query.INSTANCE_FIELDS
3784 def ExpandNames(self, lu):
3785 lu.needed_locks = {}
3786 lu.share_locks[locking.LEVEL_INSTANCE] = 1
3787 lu.share_locks[locking.LEVEL_NODE] = 1
3790 self.wanted = _GetWantedInstances(lu, self.names)
3792 self.wanted = locking.ALL_SET
3794 self.do_locking = (self.use_locking and
3795 query.IQ_LIVE in self.requested_data)
3797 lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3798 lu.needed_locks[locking.LEVEL_NODE] = []
3799 lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3801 def DeclareLocks(self, lu, level):
3802 if level == locking.LEVEL_NODE and self.do_locking:
3803 lu._LockInstancesNodes() # pylint: disable-msg=W0212
3805 def _GetQueryData(self, lu):
3806 """Computes the list of instances and their attributes.
3809 all_info = lu.cfg.GetAllInstancesInfo()
3811 instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3813 instance_list = [all_info[name] for name in instance_names]
3814 nodes = frozenset([inst.primary_node for inst in instance_list])
3815 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3819 # Gather data as requested
3820 if query.IQ_LIVE in self.requested_data:
3822 node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3824 result = node_data[name]
3826 # offline nodes will be in both lists
3827 assert result.fail_msg
3828 offline_nodes.append(name)
3830 bad_nodes.append(name)
3831 elif result.payload:
3832 live_data.update(result.payload)
3833 # else no instance is alive
3837 if query.IQ_DISKUSAGE in self.requested_data:
3838 disk_usage = dict((inst.name,
3839 _ComputeDiskSize(inst.disk_template,
3840 [{"size": disk.size}
3841 for disk in inst.disks]))
3842 for inst in instance_list)
3846 return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3847 disk_usage, offline_nodes, bad_nodes,
3851 class LUQuery(NoHooksLU):
3852 """Query for resources/items of a certain kind.
3855 # pylint: disable-msg=W0142
3858 def CheckArguments(self):
3859 qcls = _GetQueryImplementation(self.op.what)
3860 names = qlang.ReadSimpleFilter("name", self.op.filter)
3862 self.impl = qcls(names, self.op.fields, False)
3864 def ExpandNames(self):
3865 self.impl.ExpandNames(self)
3867 def DeclareLocks(self, level):
3868 self.impl.DeclareLocks(self, level)
3870 def Exec(self, feedback_fn):
3871 return self.impl.NewStyleQuery(self)
3874 class LUQueryFields(NoHooksLU):
3875 """Query for resources/items of a certain kind.
3878 # pylint: disable-msg=W0142
3881 def CheckArguments(self):
3882 self.qcls = _GetQueryImplementation(self.op.what)
3884 def ExpandNames(self):
3885 self.needed_locks = {}
3887 def Exec(self, feedback_fn):
3888 return self.qcls.FieldsQuery(self.op.fields)
3891 class LUNodeModifyStorage(NoHooksLU):
3892 """Logical unit for modifying a storage volume on a node.
3897 def CheckArguments(self):
3898 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3900 storage_type = self.op.storage_type
3903 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3905 raise errors.OpPrereqError("Storage units of type '%s' can not be"
3906 " modified" % storage_type,
3909 diff = set(self.op.changes.keys()) - modifiable
3911 raise errors.OpPrereqError("The following fields can not be modified for"
3912 " storage units of type '%s': %r" %
3913 (storage_type, list(diff)),
3916 def ExpandNames(self):
3917 self.needed_locks = {
3918 locking.LEVEL_NODE: self.op.node_name,
3921 def Exec(self, feedback_fn):
3922 """Computes the list of nodes and their attributes.
3925 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3926 result = self.rpc.call_storage_modify(self.op.node_name,
3927 self.op.storage_type, st_args,
3928 self.op.name, self.op.changes)
3929 result.Raise("Failed to modify storage unit '%s' on %s" %
3930 (self.op.name, self.op.node_name))
3933 class LUNodeAdd(LogicalUnit):
3934 """Logical unit for adding node to the cluster.
3938 HTYPE = constants.HTYPE_NODE
3939 _NFLAGS = ["master_capable", "vm_capable"]
3941 def CheckArguments(self):
3942 self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
3943 # validate/normalize the node name
3944 self.hostname = netutils.GetHostname(name=self.op.node_name,
3945 family=self.primary_ip_family)
3946 self.op.node_name = self.hostname.name
3947 if self.op.readd and self.op.group:
3948 raise errors.OpPrereqError("Cannot pass a node group when a node is"
3949 " being readded", errors.ECODE_INVAL)
3951 def BuildHooksEnv(self):
3954 This will run on all nodes before, and on all nodes + the new node after.
3958 "OP_TARGET": self.op.node_name,
3959 "NODE_NAME": self.op.node_name,
3960 "NODE_PIP": self.op.primary_ip,
3961 "NODE_SIP": self.op.secondary_ip,
3962 "MASTER_CAPABLE": str(self.op.master_capable),
3963 "VM_CAPABLE": str(self.op.vm_capable),
3965 nodes_0 = self.cfg.GetNodeList()
3966 nodes_1 = nodes_0 + [self.op.node_name, ]
3967 return env, nodes_0, nodes_1
3969 def CheckPrereq(self):
3970 """Check prerequisites.
3973 - the new node is not already in the config
3975 - its parameters (single/dual homed) matches the cluster
3977 Any errors are signaled by raising errors.OpPrereqError.
3981 hostname = self.hostname
3982 node = hostname.name
3983 primary_ip = self.op.primary_ip = hostname.ip
3984 if self.op.secondary_ip is None:
3985 if self.primary_ip_family == netutils.IP6Address.family:
3986 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
3987 " IPv4 address must be given as secondary",
3989 self.op.secondary_ip = primary_ip
3991 secondary_ip = self.op.secondary_ip
3992 if not netutils.IP4Address.IsValid(secondary_ip):
3993 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
3994 " address" % secondary_ip, errors.ECODE_INVAL)
3996 node_list = cfg.GetNodeList()
3997 if not self.op.readd and node in node_list:
3998 raise errors.OpPrereqError("Node %s is already in the configuration" %
3999 node, errors.ECODE_EXISTS)
4000 elif self.op.readd and node not in node_list:
4001 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4004 self.changed_primary_ip = False
4006 for existing_node_name in node_list:
4007 existing_node = cfg.GetNodeInfo(existing_node_name)
4009 if self.op.readd and node == existing_node_name:
4010 if existing_node.secondary_ip != secondary_ip:
4011 raise errors.OpPrereqError("Readded node doesn't have the same IP"
4012 " address configuration as before",
4014 if existing_node.primary_ip != primary_ip:
4015 self.changed_primary_ip = True
4019 if (existing_node.primary_ip == primary_ip or
4020 existing_node.secondary_ip == primary_ip or
4021 existing_node.primary_ip == secondary_ip or
4022 existing_node.secondary_ip == secondary_ip):
4023 raise errors.OpPrereqError("New node ip address(es) conflict with"
4024 " existing node %s" % existing_node.name,
4025 errors.ECODE_NOTUNIQUE)
4027 # After this 'if' block, None is no longer a valid value for the
4028 # _capable op attributes
4030 old_node = self.cfg.GetNodeInfo(node)
4031 assert old_node is not None, "Can't retrieve locked node %s" % node
4032 for attr in self._NFLAGS:
4033 if getattr(self.op, attr) is None:
4034 setattr(self.op, attr, getattr(old_node, attr))
4036 for attr in self._NFLAGS:
4037 if getattr(self.op, attr) is None:
4038 setattr(self.op, attr, True)
4040 if self.op.readd and not self.op.vm_capable:
4041 pri, sec = cfg.GetNodeInstances(node)
4043 raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4044 " flag set to false, but it already holds"
4045 " instances" % node,
4048 # check that the type of the node (single versus dual homed) is the
4049 # same as for the master
4050 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4051 master_singlehomed = myself.secondary_ip == myself.primary_ip
4052 newbie_singlehomed = secondary_ip == primary_ip
4053 if master_singlehomed != newbie_singlehomed:
4054 if master_singlehomed:
4055 raise errors.OpPrereqError("The master has no secondary ip but the"
4056 " new node has one",
4059 raise errors.OpPrereqError("The master has a secondary ip but the"
4060 " new node doesn't have one",
4063 # checks reachability
4064 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
4065 raise errors.OpPrereqError("Node not reachable by ping",
4066 errors.ECODE_ENVIRON)
4068 if not newbie_singlehomed:
4069 # check reachability from my secondary ip to newbie's secondary ip
4070 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
4071 source=myself.secondary_ip):
4072 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4073 " based ping to node daemon port",
4074 errors.ECODE_ENVIRON)
4081 if self.op.master_capable:
4082 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
4084 self.master_candidate = False
4087 self.new_node = old_node
4089 node_group = cfg.LookupNodeGroup(self.op.group)
4090 self.new_node = objects.Node(name=node,
4091 primary_ip=primary_ip,
4092 secondary_ip=secondary_ip,
4093 master_candidate=self.master_candidate,
4094 offline=False, drained=False,
4097 if self.op.ndparams:
4098 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
4100 def Exec(self, feedback_fn):
4101 """Adds the new node to the cluster.
4104 new_node = self.new_node
4105 node = new_node.name
4107 # We adding a new node so we assume it's powered
4108 new_node.powered = True
4110 # for re-adds, reset the offline/drained/master-candidate flags;
4111 # we need to reset here, otherwise offline would prevent RPC calls
4112 # later in the procedure; this also means that if the re-add
4113 # fails, we are left with a non-offlined, broken node
4115 new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
4116 self.LogInfo("Readding a node, the offline/drained flags were reset")
4117 # if we demote the node, we do cleanup later in the procedure
4118 new_node.master_candidate = self.master_candidate
4119 if self.changed_primary_ip:
4120 new_node.primary_ip = self.op.primary_ip
4122 # copy the master/vm_capable flags
4123 for attr in self._NFLAGS:
4124 setattr(new_node, attr, getattr(self.op, attr))
4126 # notify the user about any possible mc promotion
4127 if new_node.master_candidate:
4128 self.LogInfo("Node will be a master candidate")
4130 if self.op.ndparams:
4131 new_node.ndparams = self.op.ndparams
4133 new_node.ndparams = {}
4135 # check connectivity
4136 result = self.rpc.call_version([node])[node]
4137 result.Raise("Can't get version information from node %s" % node)
4138 if constants.PROTOCOL_VERSION == result.payload:
4139 logging.info("Communication to node %s fine, sw version %s match",
4140 node, result.payload)
4142 raise errors.OpExecError("Version mismatch master version %s,"
4143 " node version %s" %
4144 (constants.PROTOCOL_VERSION, result.payload))
4146 # Add node to our /etc/hosts, and add key to known_hosts
4147 if self.cfg.GetClusterInfo().modify_etc_hosts:
4148 master_node = self.cfg.GetMasterNode()
4149 result = self.rpc.call_etc_hosts_modify(master_node,
4150 constants.ETC_HOSTS_ADD,
4153 result.Raise("Can't update hosts file with new host data")
4155 if new_node.secondary_ip != new_node.primary_ip:
4156 _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
4159 node_verify_list = [self.cfg.GetMasterNode()]
4160 node_verify_param = {
4161 constants.NV_NODELIST: [node],
4162 # TODO: do a node-net-test as well?
4165 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
4166 self.cfg.GetClusterName())
4167 for verifier in node_verify_list:
4168 result[verifier].Raise("Cannot communicate with node %s" % verifier)
4169 nl_payload = result[verifier].payload[constants.NV_NODELIST]
4171 for failed in nl_payload:
4172 feedback_fn("ssh/hostname verification failed"
4173 " (checking from %s): %s" %
4174 (verifier, nl_payload[failed]))
4175 raise errors.OpExecError("ssh/hostname verification failed.")
4178 _RedistributeAncillaryFiles(self)
4179 self.context.ReaddNode(new_node)
4180 # make sure we redistribute the config
4181 self.cfg.Update(new_node, feedback_fn)
4182 # and make sure the new node will not have old files around
4183 if not new_node.master_candidate:
4184 result = self.rpc.call_node_demote_from_mc(new_node.name)
4185 msg = result.fail_msg
4187 self.LogWarning("Node failed to demote itself from master"
4188 " candidate status: %s" % msg)
4190 _RedistributeAncillaryFiles(self, additional_nodes=[node],
4191 additional_vm=self.op.vm_capable)
4192 self.context.AddNode(new_node, self.proc.GetECId())
4195 class LUNodeSetParams(LogicalUnit):
4196 """Modifies the parameters of a node.
4198 @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
4199 to the node role (as _ROLE_*)
4200 @cvar _R2F: a dictionary from node role to tuples of flags
4201 @cvar _FLAGS: a list of attribute names corresponding to the flags
4204 HPATH = "node-modify"
4205 HTYPE = constants.HTYPE_NODE
4207 (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
4209 (True, False, False): _ROLE_CANDIDATE,
4210 (False, True, False): _ROLE_DRAINED,
4211 (False, False, True): _ROLE_OFFLINE,
4212 (False, False, False): _ROLE_REGULAR,
4214 _R2F = dict((v, k) for k, v in _F2R.items())
4215 _FLAGS = ["master_candidate", "drained", "offline"]
4217 def CheckArguments(self):
4218 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4219 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
4220 self.op.master_capable, self.op.vm_capable,
4221 self.op.secondary_ip, self.op.ndparams]
4222 if all_mods.count(None) == len(all_mods):
4223 raise errors.OpPrereqError("Please pass at least one modification",
4225 if all_mods.count(True) > 1:
4226 raise errors.OpPrereqError("Can't set the node into more than one"
4227 " state at the same time",
4230 # Boolean value that tells us whether we might be demoting from MC
4231 self.might_demote = (self.op.master_candidate == False or
4232 self.op.offline == True or
4233 self.op.drained == True or
4234 self.op.master_capable == False)
4236 if self.op.secondary_ip:
4237 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
4238 raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4239 " address" % self.op.secondary_ip,
4242 self.lock_all = self.op.auto_promote and self.might_demote
4243 self.lock_instances = self.op.secondary_ip is not None
4245 def ExpandNames(self):
4247 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
4249 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
4251 if self.lock_instances:
4252 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4254 def DeclareLocks(self, level):
4255 # If we have locked all instances, before waiting to lock nodes, release
4256 # all the ones living on nodes unrelated to the current operation.
4257 if level == locking.LEVEL_NODE and self.lock_instances:
4258 instances_release = []
4260 self.affected_instances = []
4261 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4262 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
4263 instance = self.context.cfg.GetInstanceInfo(instance_name)
4264 i_mirrored = instance.disk_template in constants.DTS_NET_MIRROR
4265 if i_mirrored and self.op.node_name in instance.all_nodes:
4266 instances_keep.append(instance_name)
4267 self.affected_instances.append(instance)
4269 instances_release.append(instance_name)
4270 if instances_release:
4271 self.context.glm.release(locking.LEVEL_INSTANCE, instances_release)
4272 self.acquired_locks[locking.LEVEL_INSTANCE] = instances_keep
4274 def BuildHooksEnv(self):
4277 This runs on the master node.
4281 "OP_TARGET": self.op.node_name,
4282 "MASTER_CANDIDATE": str(self.op.master_candidate),
4283 "OFFLINE": str(self.op.offline),
4284 "DRAINED": str(self.op.drained),
4285 "MASTER_CAPABLE": str(self.op.master_capable),
4286 "VM_CAPABLE": str(self.op.vm_capable),
4288 nl = [self.cfg.GetMasterNode(),
4292 def CheckPrereq(self):
4293 """Check prerequisites.
4295 This only checks the instance list against the existing names.
4298 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
4300 if (self.op.master_candidate is not None or
4301 self.op.drained is not None or
4302 self.op.offline is not None):
4303 # we can't change the master's node flags
4304 if self.op.node_name == self.cfg.GetMasterNode():
4305 raise errors.OpPrereqError("The master role can be changed"
4306 " only via master-failover",
4309 if self.op.master_candidate and not node.master_capable:
4310 raise errors.OpPrereqError("Node %s is not master capable, cannot make"
4311 " it a master candidate" % node.name,
4314 if self.op.vm_capable == False:
4315 (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
4317 raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
4318 " the vm_capable flag" % node.name,
4321 if node.master_candidate and self.might_demote and not self.lock_all:
4322 assert not self.op.auto_promote, "auto-promote set but lock_all not"
4323 # check if after removing the current node, we're missing master
4325 (mc_remaining, mc_should, _) = \
4326 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
4327 if mc_remaining < mc_should:
4328 raise errors.OpPrereqError("Not enough master candidates, please"
4329 " pass auto_promote to allow promotion",
4332 self.old_flags = old_flags = (node.master_candidate,
4333 node.drained, node.offline)
4334 assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags)
4335 self.old_role = old_role = self._F2R[old_flags]
4337 # Check for ineffective changes
4338 for attr in self._FLAGS:
4339 if (getattr(self.op, attr) == False and getattr(node, attr) == False):
4340 self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
4341 setattr(self.op, attr, None)
4343 # Past this point, any flag change to False means a transition
4344 # away from the respective state, as only real changes are kept
4346 # TODO: We might query the real power state if it supports OOB
4347 if _SupportsOob(self.cfg, node):
4348 if self.op.offline is False and not (node.powered or
4349 self.op.powered == True):
4350 raise errors.OpPrereqError(("Please power on node %s first before you"
4351 " can reset offline state") %
4353 elif self.op.powered is not None:
4354 raise errors.OpPrereqError(("Unable to change powered state for node %s"
4355 " which does not support out-of-band"
4356 " handling") % self.op.node_name)
4358 # If we're being deofflined/drained, we'll MC ourself if needed
4359 if (self.op.drained == False or self.op.offline == False or
4360 (self.op.master_capable and not node.master_capable)):
4361 if _DecideSelfPromotion(self):
4362 self.op.master_candidate = True
4363 self.LogInfo("Auto-promoting node to master candidate")
4365 # If we're no longer master capable, we'll demote ourselves from MC
4366 if self.op.master_capable == False and node.master_candidate:
4367 self.LogInfo("Demoting from master candidate")
4368 self.op.master_candidate = False
4371 assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
4372 if self.op.master_candidate:
4373 new_role = self._ROLE_CANDIDATE
4374 elif self.op.drained:
4375 new_role = self._ROLE_DRAINED
4376 elif self.op.offline:
4377 new_role = self._ROLE_OFFLINE
4378 elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
4379 # False is still in new flags, which means we're un-setting (the
4381 new_role = self._ROLE_REGULAR
4382 else: # no new flags, nothing, keep old role
4385 self.new_role = new_role
4387 if old_role == self._ROLE_OFFLINE and new_role != old_role:
4388 # Trying to transition out of offline status
4389 result = self.rpc.call_version([node.name])[node.name]
4391 raise errors.OpPrereqError("Node %s is being de-offlined but fails"
4392 " to report its version: %s" %
4393 (node.name, result.fail_msg),
4396 self.LogWarning("Transitioning node from offline to online state"
4397 " without using re-add. Please make sure the node"
4400 if self.op.secondary_ip:
4401 # Ok even without locking, because this can't be changed by any LU
4402 master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
4403 master_singlehomed = master.secondary_ip == master.primary_ip
4404 if master_singlehomed and self.op.secondary_ip:
4405 raise errors.OpPrereqError("Cannot change the secondary ip on a single"
4406 " homed cluster", errors.ECODE_INVAL)
4409 if self.affected_instances:
4410 raise errors.OpPrereqError("Cannot change secondary ip: offline"
4411 " node has instances (%s) configured"
4412 " to use it" % self.affected_instances)
4414 # On online nodes, check that no instances are running, and that
4415 # the node has the new ip and we can reach it.
4416 for instance in self.affected_instances:
4417 _CheckInstanceDown(self, instance, "cannot change secondary ip")
4419 _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
4420 if master.name != node.name:
4421 # check reachability from master secondary ip to new secondary ip
4422 if not netutils.TcpPing(self.op.secondary_ip,
4423 constants.DEFAULT_NODED_PORT,
4424 source=master.secondary_ip):
4425 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4426 " based ping to node daemon port",
4427 errors.ECODE_ENVIRON)
4429 if self.op.ndparams:
4430 new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
4431 utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
4432 self.new_ndparams = new_ndparams
4434 def Exec(self, feedback_fn):
4439 old_role = self.old_role
4440 new_role = self.new_role
4444 if self.op.ndparams:
4445 node.ndparams = self.new_ndparams
4447 if self.op.powered is not None:
4448 node.powered = self.op.powered
4450 for attr in ["master_capable", "vm_capable"]:
4451 val = getattr(self.op, attr)
4453 setattr(node, attr, val)
4454 result.append((attr, str(val)))
4456 if new_role != old_role:
4457 # Tell the node to demote itself, if no longer MC and not offline
4458 if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
4459 msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
4461 self.LogWarning("Node failed to demote itself: %s", msg)
4463 new_flags = self._R2F[new_role]
4464 for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
4466 result.append((desc, str(nf)))
4467 (node.master_candidate, node.drained, node.offline) = new_flags
4469 # we locked all nodes, we adjust the CP before updating this node
4471 _AdjustCandidatePool(self, [node.name])
4473 if self.op.secondary_ip:
4474 node.secondary_ip = self.op.secondary_ip
4475 result.append(("secondary_ip", self.op.secondary_ip))
4477 # this will trigger configuration file update, if needed
4478 self.cfg.Update(node, feedback_fn)
4480 # this will trigger job queue propagation or cleanup if the mc
4482 if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
4483 self.context.ReaddNode(node)
4488 class LUNodePowercycle(NoHooksLU):
4489 """Powercycles a node.
4494 def CheckArguments(self):
4495 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4496 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4497 raise errors.OpPrereqError("The node is the master and the force"
4498 " parameter was not set",
4501 def ExpandNames(self):
4502 """Locking for PowercycleNode.
4504 This is a last-resort option and shouldn't block on other
4505 jobs. Therefore, we grab no locks.
4508 self.needed_locks = {}
4510 def Exec(self, feedback_fn):
4514 result = self.rpc.call_node_powercycle(self.op.node_name,
4515 self.cfg.GetHypervisorType())
4516 result.Raise("Failed to schedule the reboot")
4517 return result.payload
4520 class LUClusterQuery(NoHooksLU):
4521 """Query cluster configuration.
4526 def ExpandNames(self):
4527 self.needed_locks = {}
4529 def Exec(self, feedback_fn):
4530 """Return cluster config.
4533 cluster = self.cfg.GetClusterInfo()
4536 # Filter just for enabled hypervisors
4537 for os_name, hv_dict in cluster.os_hvp.items():
4538 os_hvp[os_name] = {}
4539 for hv_name, hv_params in hv_dict.items():
4540 if hv_name in cluster.enabled_hypervisors:
4541 os_hvp[os_name][hv_name] = hv_params
4543 # Convert ip_family to ip_version
4544 primary_ip_version = constants.IP4_VERSION
4545 if cluster.primary_ip_family == netutils.IP6Address.family:
4546 primary_ip_version = constants.IP6_VERSION
4549 "software_version": constants.RELEASE_VERSION,
4550 "protocol_version": constants.PROTOCOL_VERSION,
4551 "config_version": constants.CONFIG_VERSION,
4552 "os_api_version": max(constants.OS_API_VERSIONS),
4553 "export_version": constants.EXPORT_VERSION,
4554 "architecture": (platform.architecture()[0], platform.machine()),
4555 "name": cluster.cluster_name,
4556 "master": cluster.master_node,
4557 "default_hypervisor": cluster.enabled_hypervisors[0],
4558 "enabled_hypervisors": cluster.enabled_hypervisors,
4559 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4560 for hypervisor_name in cluster.enabled_hypervisors]),
4562 "beparams": cluster.beparams,
4563 "osparams": cluster.osparams,
4564 "nicparams": cluster.nicparams,
4565 "ndparams": cluster.ndparams,
4566 "candidate_pool_size": cluster.candidate_pool_size,
4567 "master_netdev": cluster.master_netdev,
4568 "volume_group_name": cluster.volume_group_name,
4569 "drbd_usermode_helper": cluster.drbd_usermode_helper,
4570 "file_storage_dir": cluster.file_storage_dir,
4571 "maintain_node_health": cluster.maintain_node_health,
4572 "ctime": cluster.ctime,
4573 "mtime": cluster.mtime,
4574 "uuid": cluster.uuid,
4575 "tags": list(cluster.GetTags()),
4576 "uid_pool": cluster.uid_pool,
4577 "default_iallocator": cluster.default_iallocator,
4578 "reserved_lvs": cluster.reserved_lvs,
4579 "primary_ip_version": primary_ip_version,
4580 "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4586 class LUClusterConfigQuery(NoHooksLU):
4587 """Return configuration values.
4591 _FIELDS_DYNAMIC = utils.FieldSet()
4592 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4593 "watcher_pause", "volume_group_name")
4595 def CheckArguments(self):
4596 _CheckOutputFields(static=self._FIELDS_STATIC,
4597 dynamic=self._FIELDS_DYNAMIC,
4598 selected=self.op.output_fields)
4600 def ExpandNames(self):
4601 self.needed_locks = {}
4603 def Exec(self, feedback_fn):
4604 """Dump a representation of the cluster config to the standard output.
4608 for field in self.op.output_fields:
4609 if field == "cluster_name":
4610 entry = self.cfg.GetClusterName()
4611 elif field == "master_node":
4612 entry = self.cfg.GetMasterNode()
4613 elif field == "drain_flag":
4614 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4615 elif field == "watcher_pause":
4616 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4617 elif field == "volume_group_name":
4618 entry = self.cfg.GetVGName()
4620 raise errors.ParameterError(field)
4621 values.append(entry)
4625 class LUInstanceActivateDisks(NoHooksLU):
4626 """Bring up an instance's disks.
4631 def ExpandNames(self):
4632 self._ExpandAndLockInstance()
4633 self.needed_locks[locking.LEVEL_NODE] = []
4634 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4636 def DeclareLocks(self, level):
4637 if level == locking.LEVEL_NODE:
4638 self._LockInstancesNodes()
4640 def CheckPrereq(self):
4641 """Check prerequisites.
4643 This checks that the instance is in the cluster.
4646 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4647 assert self.instance is not None, \
4648 "Cannot retrieve locked instance %s" % self.op.instance_name
4649 _CheckNodeOnline(self, self.instance.primary_node)
4651 def Exec(self, feedback_fn):
4652 """Activate the disks.
4655 disks_ok, disks_info = \
4656 _AssembleInstanceDisks(self, self.instance,
4657 ignore_size=self.op.ignore_size)
4659 raise errors.OpExecError("Cannot activate block devices")
4664 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4666 """Prepare the block devices for an instance.
4668 This sets up the block devices on all nodes.
4670 @type lu: L{LogicalUnit}
4671 @param lu: the logical unit on whose behalf we execute
4672 @type instance: L{objects.Instance}
4673 @param instance: the instance for whose disks we assemble
4674 @type disks: list of L{objects.Disk} or None
4675 @param disks: which disks to assemble (or all, if None)
4676 @type ignore_secondaries: boolean
4677 @param ignore_secondaries: if true, errors on secondary nodes
4678 won't result in an error return from the function
4679 @type ignore_size: boolean
4680 @param ignore_size: if true, the current known size of the disk
4681 will not be used during the disk activation, useful for cases
4682 when the size is wrong
4683 @return: False if the operation failed, otherwise a list of
4684 (host, instance_visible_name, node_visible_name)
4685 with the mapping from node devices to instance devices
4690 iname = instance.name
4691 disks = _ExpandCheckDisks(instance, disks)
4693 # With the two passes mechanism we try to reduce the window of
4694 # opportunity for the race condition of switching DRBD to primary
4695 # before handshaking occured, but we do not eliminate it
4697 # The proper fix would be to wait (with some limits) until the
4698 # connection has been made and drbd transitions from WFConnection
4699 # into any other network-connected state (Connected, SyncTarget,
4702 # 1st pass, assemble on all nodes in secondary mode
4703 for inst_disk in disks:
4704 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4706 node_disk = node_disk.Copy()
4707 node_disk.UnsetSize()
4708 lu.cfg.SetDiskID(node_disk, node)
4709 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4710 msg = result.fail_msg
4712 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4713 " (is_primary=False, pass=1): %s",
4714 inst_disk.iv_name, node, msg)
4715 if not ignore_secondaries:
4718 # FIXME: race condition on drbd migration to primary
4720 # 2nd pass, do only the primary node
4721 for inst_disk in disks:
4724 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4725 if node != instance.primary_node:
4728 node_disk = node_disk.Copy()
4729 node_disk.UnsetSize()
4730 lu.cfg.SetDiskID(node_disk, node)
4731 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4732 msg = result.fail_msg
4734 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4735 " (is_primary=True, pass=2): %s",
4736 inst_disk.iv_name, node, msg)
4739 dev_path = result.payload
4741 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4743 # leave the disks configured for the primary node
4744 # this is a workaround that would be fixed better by
4745 # improving the logical/physical id handling
4747 lu.cfg.SetDiskID(disk, instance.primary_node)
4749 return disks_ok, device_info
4752 def _StartInstanceDisks(lu, instance, force):
4753 """Start the disks of an instance.
4756 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4757 ignore_secondaries=force)
4759 _ShutdownInstanceDisks(lu, instance)
4760 if force is not None and not force:
4761 lu.proc.LogWarning("", hint="If the message above refers to a"
4763 " you can retry the operation using '--force'.")
4764 raise errors.OpExecError("Disk consistency error")
4767 class LUInstanceDeactivateDisks(NoHooksLU):
4768 """Shutdown an instance's disks.
4773 def ExpandNames(self):
4774 self._ExpandAndLockInstance()
4775 self.needed_locks[locking.LEVEL_NODE] = []
4776 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4778 def DeclareLocks(self, level):
4779 if level == locking.LEVEL_NODE:
4780 self._LockInstancesNodes()
4782 def CheckPrereq(self):
4783 """Check prerequisites.
4785 This checks that the instance is in the cluster.
4788 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4789 assert self.instance is not None, \
4790 "Cannot retrieve locked instance %s" % self.op.instance_name
4792 def Exec(self, feedback_fn):
4793 """Deactivate the disks
4796 instance = self.instance
4797 _SafeShutdownInstanceDisks(self, instance)
4800 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4801 """Shutdown block devices of an instance.
4803 This function checks if an instance is running, before calling
4804 _ShutdownInstanceDisks.
4807 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4808 _ShutdownInstanceDisks(lu, instance, disks=disks)
4811 def _ExpandCheckDisks(instance, disks):
4812 """Return the instance disks selected by the disks list
4814 @type disks: list of L{objects.Disk} or None
4815 @param disks: selected disks
4816 @rtype: list of L{objects.Disk}
4817 @return: selected instance disks to act on
4821 return instance.disks
4823 if not set(disks).issubset(instance.disks):
4824 raise errors.ProgrammerError("Can only act on disks belonging to the"
4829 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4830 """Shutdown block devices of an instance.
4832 This does the shutdown on all nodes of the instance.
4834 If the ignore_primary is false, errors on the primary node are
4839 disks = _ExpandCheckDisks(instance, disks)
4842 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4843 lu.cfg.SetDiskID(top_disk, node)
4844 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4845 msg = result.fail_msg
4847 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4848 disk.iv_name, node, msg)
4849 if ((node == instance.primary_node and not ignore_primary) or
4850 (node != instance.primary_node and not result.offline)):
4855 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4856 """Checks if a node has enough free memory.
4858 This function check if a given node has the needed amount of free
4859 memory. In case the node has less memory or we cannot get the
4860 information from the node, this function raise an OpPrereqError
4863 @type lu: C{LogicalUnit}
4864 @param lu: a logical unit from which we get configuration data
4866 @param node: the node to check
4867 @type reason: C{str}
4868 @param reason: string to use in the error message
4869 @type requested: C{int}
4870 @param requested: the amount of memory in MiB to check for
4871 @type hypervisor_name: C{str}
4872 @param hypervisor_name: the hypervisor to ask for memory stats
4873 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4874 we cannot check the node
4877 nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
4878 nodeinfo[node].Raise("Can't get data from node %s" % node,
4879 prereq=True, ecode=errors.ECODE_ENVIRON)
4880 free_mem = nodeinfo[node].payload.get('memory_free', None)
4881 if not isinstance(free_mem, int):
4882 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4883 " was '%s'" % (node, free_mem),
4884 errors.ECODE_ENVIRON)
4885 if requested > free_mem:
4886 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4887 " needed %s MiB, available %s MiB" %
4888 (node, reason, requested, free_mem),
4892 def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
4893 """Checks if nodes have enough free disk space in the all VGs.
4895 This function check if all given nodes have the needed amount of
4896 free disk. In case any node has less disk or we cannot get the
4897 information from the node, this function raise an OpPrereqError
4900 @type lu: C{LogicalUnit}
4901 @param lu: a logical unit from which we get configuration data
4902 @type nodenames: C{list}
4903 @param nodenames: the list of node names to check
4904 @type req_sizes: C{dict}
4905 @param req_sizes: the hash of vg and corresponding amount of disk in
4907 @raise errors.OpPrereqError: if the node doesn't have enough disk,
4908 or we cannot check the node
4911 for vg, req_size in req_sizes.items():
4912 _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
4915 def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
4916 """Checks if nodes have enough free disk space in the specified VG.
4918 This function check if all given nodes have the needed amount of
4919 free disk. In case any node has less disk or we cannot get the
4920 information from the node, this function raise an OpPrereqError
4923 @type lu: C{LogicalUnit}
4924 @param lu: a logical unit from which we get configuration data
4925 @type nodenames: C{list}
4926 @param nodenames: the list of node names to check
4928 @param vg: the volume group to check
4929 @type requested: C{int}
4930 @param requested: the amount of disk in MiB to check for
4931 @raise errors.OpPrereqError: if the node doesn't have enough disk,
4932 or we cannot check the node
4935 nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
4936 for node in nodenames:
4937 info = nodeinfo[node]
4938 info.Raise("Cannot get current information from node %s" % node,
4939 prereq=True, ecode=errors.ECODE_ENVIRON)
4940 vg_free = info.payload.get("vg_free", None)
4941 if not isinstance(vg_free, int):
4942 raise errors.OpPrereqError("Can't compute free disk space on node"
4943 " %s for vg %s, result was '%s'" %
4944 (node, vg, vg_free), errors.ECODE_ENVIRON)
4945 if requested > vg_free:
4946 raise errors.OpPrereqError("Not enough disk space on target node %s"
4947 " vg %s: required %d MiB, available %d MiB" %
4948 (node, vg, requested, vg_free),
4952 class LUInstanceStartup(LogicalUnit):
4953 """Starts an instance.
4956 HPATH = "instance-start"
4957 HTYPE = constants.HTYPE_INSTANCE
4960 def CheckArguments(self):
4962 if self.op.beparams:
4963 # fill the beparams dict
4964 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4966 def ExpandNames(self):
4967 self._ExpandAndLockInstance()
4969 def BuildHooksEnv(self):
4972 This runs on master, primary and secondary nodes of the instance.
4976 "FORCE": self.op.force,
4978 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4979 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4982 def CheckPrereq(self):
4983 """Check prerequisites.
4985 This checks that the instance is in the cluster.
4988 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4989 assert self.instance is not None, \
4990 "Cannot retrieve locked instance %s" % self.op.instance_name
4993 if self.op.hvparams:
4994 # check hypervisor parameter syntax (locally)
4995 cluster = self.cfg.GetClusterInfo()
4996 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4997 filled_hvp = cluster.FillHV(instance)
4998 filled_hvp.update(self.op.hvparams)
4999 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5000 hv_type.CheckParameterSyntax(filled_hvp)
5001 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5003 self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5005 if self.primary_offline and self.op.ignore_offline_nodes:
5006 self.proc.LogWarning("Ignoring offline primary node")
5008 if self.op.hvparams or self.op.beparams:
5009 self.proc.LogWarning("Overridden parameters are ignored")
5011 _CheckNodeOnline(self, instance.primary_node)
5013 bep = self.cfg.GetClusterInfo().FillBE(instance)
5015 # check bridges existence
5016 _CheckInstanceBridgesExist(self, instance)
5018 remote_info = self.rpc.call_instance_info(instance.primary_node,
5020 instance.hypervisor)
5021 remote_info.Raise("Error checking node %s" % instance.primary_node,
5022 prereq=True, ecode=errors.ECODE_ENVIRON)
5023 if not remote_info.payload: # not running already
5024 _CheckNodeFreeMemory(self, instance.primary_node,
5025 "starting instance %s" % instance.name,
5026 bep[constants.BE_MEMORY], instance.hypervisor)
5028 def Exec(self, feedback_fn):
5029 """Start the instance.
5032 instance = self.instance
5033 force = self.op.force
5035 self.cfg.MarkInstanceUp(instance.name)
5037 if self.primary_offline:
5038 assert self.op.ignore_offline_nodes
5039 self.proc.LogInfo("Primary node offline, marked instance as started")
5041 node_current = instance.primary_node
5043 _StartInstanceDisks(self, instance, force)
5045 result = self.rpc.call_instance_start(node_current, instance,
5046 self.op.hvparams, self.op.beparams)
5047 msg = result.fail_msg
5049 _ShutdownInstanceDisks(self, instance)
5050 raise errors.OpExecError("Could not start instance: %s" % msg)
5053 class LUInstanceReboot(LogicalUnit):
5054 """Reboot an instance.
5057 HPATH = "instance-reboot"
5058 HTYPE = constants.HTYPE_INSTANCE
5061 def ExpandNames(self):
5062 self._ExpandAndLockInstance()
5064 def BuildHooksEnv(self):
5067 This runs on master, primary and secondary nodes of the instance.
5071 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
5072 "REBOOT_TYPE": self.op.reboot_type,
5073 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5075 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5076 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5079 def CheckPrereq(self):
5080 """Check prerequisites.
5082 This checks that the instance is in the cluster.
5085 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5086 assert self.instance is not None, \
5087 "Cannot retrieve locked instance %s" % self.op.instance_name
5089 _CheckNodeOnline(self, instance.primary_node)
5091 # check bridges existence
5092 _CheckInstanceBridgesExist(self, instance)
5094 def Exec(self, feedback_fn):
5095 """Reboot the instance.
5098 instance = self.instance
5099 ignore_secondaries = self.op.ignore_secondaries
5100 reboot_type = self.op.reboot_type
5102 node_current = instance.primary_node
5104 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
5105 constants.INSTANCE_REBOOT_HARD]:
5106 for disk in instance.disks:
5107 self.cfg.SetDiskID(disk, node_current)
5108 result = self.rpc.call_instance_reboot(node_current, instance,
5110 self.op.shutdown_timeout)
5111 result.Raise("Could not reboot instance")
5113 result = self.rpc.call_instance_shutdown(node_current, instance,
5114 self.op.shutdown_timeout)
5115 result.Raise("Could not shutdown instance for full reboot")
5116 _ShutdownInstanceDisks(self, instance)
5117 _StartInstanceDisks(self, instance, ignore_secondaries)
5118 result = self.rpc.call_instance_start(node_current, instance, None, None)
5119 msg = result.fail_msg
5121 _ShutdownInstanceDisks(self, instance)
5122 raise errors.OpExecError("Could not start instance for"
5123 " full reboot: %s" % msg)
5125 self.cfg.MarkInstanceUp(instance.name)
5128 class LUInstanceShutdown(LogicalUnit):
5129 """Shutdown an instance.
5132 HPATH = "instance-stop"
5133 HTYPE = constants.HTYPE_INSTANCE
5136 def ExpandNames(self):
5137 self._ExpandAndLockInstance()
5139 def BuildHooksEnv(self):
5142 This runs on master, primary and secondary nodes of the instance.
5145 env = _BuildInstanceHookEnvByObject(self, self.instance)
5146 env["TIMEOUT"] = self.op.timeout
5147 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5150 def CheckPrereq(self):
5151 """Check prerequisites.
5153 This checks that the instance is in the cluster.
5156 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5157 assert self.instance is not None, \
5158 "Cannot retrieve locked instance %s" % self.op.instance_name
5160 self.primary_offline = \
5161 self.cfg.GetNodeInfo(self.instance.primary_node).offline
5163 if self.primary_offline and self.op.ignore_offline_nodes:
5164 self.proc.LogWarning("Ignoring offline primary node")
5166 _CheckNodeOnline(self, self.instance.primary_node)
5168 def Exec(self, feedback_fn):
5169 """Shutdown the instance.
5172 instance = self.instance
5173 node_current = instance.primary_node
5174 timeout = self.op.timeout
5176 self.cfg.MarkInstanceDown(instance.name)
5178 if self.primary_offline:
5179 assert self.op.ignore_offline_nodes
5180 self.proc.LogInfo("Primary node offline, marked instance as stopped")
5182 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
5183 msg = result.fail_msg
5185 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
5187 _ShutdownInstanceDisks(self, instance)
5190 class LUInstanceReinstall(LogicalUnit):
5191 """Reinstall an instance.
5194 HPATH = "instance-reinstall"
5195 HTYPE = constants.HTYPE_INSTANCE
5198 def ExpandNames(self):
5199 self._ExpandAndLockInstance()
5201 def BuildHooksEnv(self):
5204 This runs on master, primary and secondary nodes of the instance.
5207 env = _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 and is not running.
5217 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5218 assert instance is not None, \
5219 "Cannot retrieve locked instance %s" % self.op.instance_name
5220 _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
5221 " offline, cannot reinstall")
5222 for node in instance.secondary_nodes:
5223 _CheckNodeOnline(self, node, "Instance secondary node offline,"
5224 " cannot reinstall")
5226 if instance.disk_template == constants.DT_DISKLESS:
5227 raise errors.OpPrereqError("Instance '%s' has no disks" %
5228 self.op.instance_name,
5230 _CheckInstanceDown(self, instance, "cannot reinstall")
5232 if self.op.os_type is not None:
5234 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
5235 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
5236 instance_os = self.op.os_type
5238 instance_os = instance.os
5240 nodelist = list(instance.all_nodes)
5242 if self.op.osparams:
5243 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
5244 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
5245 self.os_inst = i_osdict # the new dict (without defaults)
5249 self.instance = instance
5251 def Exec(self, feedback_fn):
5252 """Reinstall the instance.
5255 inst = self.instance
5257 if self.op.os_type is not None:
5258 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
5259 inst.os = self.op.os_type
5260 # Write to configuration
5261 self.cfg.Update(inst, feedback_fn)
5263 _StartInstanceDisks(self, inst, None)
5265 feedback_fn("Running the instance OS create scripts...")
5266 # FIXME: pass debug option from opcode to backend
5267 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
5268 self.op.debug_level,
5269 osparams=self.os_inst)
5270 result.Raise("Could not install OS for instance %s on node %s" %
5271 (inst.name, inst.primary_node))
5273 _ShutdownInstanceDisks(self, inst)
5276 class LUInstanceRecreateDisks(LogicalUnit):
5277 """Recreate an instance's missing disks.
5280 HPATH = "instance-recreate-disks"
5281 HTYPE = constants.HTYPE_INSTANCE
5284 def ExpandNames(self):
5285 self._ExpandAndLockInstance()
5287 def BuildHooksEnv(self):
5290 This runs on master, primary and secondary nodes of the instance.
5293 env = _BuildInstanceHookEnvByObject(self, self.instance)
5294 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5297 def CheckPrereq(self):
5298 """Check prerequisites.
5300 This checks that the instance is in the cluster and is not running.
5303 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5304 assert instance is not None, \
5305 "Cannot retrieve locked instance %s" % self.op.instance_name
5306 _CheckNodeOnline(self, instance.primary_node)
5308 if instance.disk_template == constants.DT_DISKLESS:
5309 raise errors.OpPrereqError("Instance '%s' has no disks" %
5310 self.op.instance_name, errors.ECODE_INVAL)
5311 _CheckInstanceDown(self, instance, "cannot recreate disks")
5313 if not self.op.disks:
5314 self.op.disks = range(len(instance.disks))
5316 for idx in self.op.disks:
5317 if idx >= len(instance.disks):
5318 raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
5321 self.instance = instance
5323 def Exec(self, feedback_fn):
5324 """Recreate the disks.
5328 for idx, _ in enumerate(self.instance.disks):
5329 if idx not in self.op.disks: # disk idx has not been passed in
5333 _CreateDisks(self, self.instance, to_skip=to_skip)
5336 class LUInstanceRename(LogicalUnit):
5337 """Rename an instance.
5340 HPATH = "instance-rename"
5341 HTYPE = constants.HTYPE_INSTANCE
5343 def CheckArguments(self):
5347 if self.op.ip_check and not self.op.name_check:
5348 # TODO: make the ip check more flexible and not depend on the name check
5349 raise errors.OpPrereqError("Cannot do ip check without a name check",
5352 def BuildHooksEnv(self):
5355 This runs on master, primary and secondary nodes of the instance.
5358 env = _BuildInstanceHookEnvByObject(self, self.instance)
5359 env["INSTANCE_NEW_NAME"] = self.op.new_name
5360 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5363 def CheckPrereq(self):
5364 """Check prerequisites.
5366 This checks that the instance is in the cluster and is not running.
5369 self.op.instance_name = _ExpandInstanceName(self.cfg,
5370 self.op.instance_name)
5371 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5372 assert instance is not None
5373 _CheckNodeOnline(self, instance.primary_node)
5374 _CheckInstanceDown(self, instance, "cannot rename")
5375 self.instance = instance
5377 new_name = self.op.new_name
5378 if self.op.name_check:
5379 hostname = netutils.GetHostname(name=new_name)
5380 self.LogInfo("Resolved given name '%s' to '%s'", new_name,
5382 new_name = self.op.new_name = hostname.name
5383 if (self.op.ip_check and
5384 netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
5385 raise errors.OpPrereqError("IP %s of instance %s already in use" %
5386 (hostname.ip, new_name),
5387 errors.ECODE_NOTUNIQUE)
5389 instance_list = self.cfg.GetInstanceList()
5390 if new_name in instance_list and new_name != instance.name:
5391 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5392 new_name, errors.ECODE_EXISTS)
5394 def Exec(self, feedback_fn):
5395 """Rename the instance.
5398 inst = self.instance
5399 old_name = inst.name
5401 rename_file_storage = False
5402 if (inst.disk_template == constants.DT_FILE and
5403 self.op.new_name != inst.name):
5404 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5405 rename_file_storage = True
5407 self.cfg.RenameInstance(inst.name, self.op.new_name)
5408 # Change the instance lock. This is definitely safe while we hold the BGL
5409 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5410 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5412 # re-read the instance from the configuration after rename
5413 inst = self.cfg.GetInstanceInfo(self.op.new_name)
5415 if rename_file_storage:
5416 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5417 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5418 old_file_storage_dir,
5419 new_file_storage_dir)
5420 result.Raise("Could not rename on node %s directory '%s' to '%s'"
5421 " (but the instance has been renamed in Ganeti)" %
5422 (inst.primary_node, old_file_storage_dir,
5423 new_file_storage_dir))
5425 _StartInstanceDisks(self, inst, None)
5427 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5428 old_name, self.op.debug_level)
5429 msg = result.fail_msg
5431 msg = ("Could not run OS rename script for instance %s on node %s"
5432 " (but the instance has been renamed in Ganeti): %s" %
5433 (inst.name, inst.primary_node, msg))
5434 self.proc.LogWarning(msg)
5436 _ShutdownInstanceDisks(self, inst)
5441 class LUInstanceRemove(LogicalUnit):
5442 """Remove an instance.
5445 HPATH = "instance-remove"
5446 HTYPE = constants.HTYPE_INSTANCE
5449 def ExpandNames(self):
5450 self._ExpandAndLockInstance()
5451 self.needed_locks[locking.LEVEL_NODE] = []
5452 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5454 def DeclareLocks(self, level):
5455 if level == locking.LEVEL_NODE:
5456 self._LockInstancesNodes()
5458 def BuildHooksEnv(self):
5461 This runs on master, primary and secondary nodes of the instance.
5464 env = _BuildInstanceHookEnvByObject(self, self.instance)
5465 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5466 nl = [self.cfg.GetMasterNode()]
5467 nl_post = list(self.instance.all_nodes) + nl
5468 return env, nl, nl_post
5470 def CheckPrereq(self):
5471 """Check prerequisites.
5473 This checks that the instance is in the cluster.
5476 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5477 assert self.instance is not None, \
5478 "Cannot retrieve locked instance %s" % self.op.instance_name
5480 def Exec(self, feedback_fn):
5481 """Remove the instance.
5484 instance = self.instance
5485 logging.info("Shutting down instance %s on node %s",
5486 instance.name, instance.primary_node)
5488 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5489 self.op.shutdown_timeout)
5490 msg = result.fail_msg
5492 if self.op.ignore_failures:
5493 feedback_fn("Warning: can't shutdown instance: %s" % msg)
5495 raise errors.OpExecError("Could not shutdown instance %s on"
5497 (instance.name, instance.primary_node, msg))
5499 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5502 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5503 """Utility function to remove an instance.
5506 logging.info("Removing block devices for instance %s", instance.name)
5508 if not _RemoveDisks(lu, instance):
5509 if not ignore_failures:
5510 raise errors.OpExecError("Can't remove instance's disks")
5511 feedback_fn("Warning: can't remove instance's disks")
5513 logging.info("Removing instance %s out of cluster config", instance.name)
5515 lu.cfg.RemoveInstance(instance.name)
5517 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5518 "Instance lock removal conflict"
5520 # Remove lock for the instance
5521 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5524 class LUInstanceQuery(NoHooksLU):
5525 """Logical unit for querying instances.
5528 # pylint: disable-msg=W0142
5531 def CheckArguments(self):
5532 self.iq = _InstanceQuery(self.op.names, self.op.output_fields,
5533 self.op.use_locking)
5535 def ExpandNames(self):
5536 self.iq.ExpandNames(self)
5538 def DeclareLocks(self, level):
5539 self.iq.DeclareLocks(self, level)
5541 def Exec(self, feedback_fn):
5542 return self.iq.OldStyleQuery(self)
5545 class LUInstanceFailover(LogicalUnit):
5546 """Failover an instance.
5549 HPATH = "instance-failover"
5550 HTYPE = constants.HTYPE_INSTANCE
5553 def ExpandNames(self):
5554 self._ExpandAndLockInstance()
5555 self.needed_locks[locking.LEVEL_NODE] = []
5556 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5558 def DeclareLocks(self, level):
5559 if level == locking.LEVEL_NODE:
5560 self._LockInstancesNodes()
5562 def BuildHooksEnv(self):
5565 This runs on master, primary and secondary nodes of the instance.
5568 instance = self.instance
5569 source_node = instance.primary_node
5570 target_node = instance.secondary_nodes[0]
5572 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5573 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5574 "OLD_PRIMARY": source_node,
5575 "OLD_SECONDARY": target_node,
5576 "NEW_PRIMARY": target_node,
5577 "NEW_SECONDARY": source_node,
5579 env.update(_BuildInstanceHookEnvByObject(self, instance))
5580 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5582 nl_post.append(source_node)
5583 return env, nl, nl_post
5585 def CheckPrereq(self):
5586 """Check prerequisites.
5588 This checks that the instance is in the cluster.
5591 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5592 assert self.instance is not None, \
5593 "Cannot retrieve locked instance %s" % self.op.instance_name
5595 bep = self.cfg.GetClusterInfo().FillBE(instance)
5596 if instance.disk_template not in constants.DTS_NET_MIRROR:
5597 raise errors.OpPrereqError("Instance's disk layout is not"
5598 " network mirrored, cannot failover.",
5601 secondary_nodes = instance.secondary_nodes
5602 if not secondary_nodes:
5603 raise errors.ProgrammerError("no secondary node but using "
5604 "a mirrored disk template")
5606 target_node = secondary_nodes[0]
5607 _CheckNodeOnline(self, target_node)
5608 _CheckNodeNotDrained(self, target_node)
5609 if instance.admin_up:
5610 # check memory requirements on the secondary node
5611 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5612 instance.name, bep[constants.BE_MEMORY],
5613 instance.hypervisor)
5615 self.LogInfo("Not checking memory on the secondary node as"
5616 " instance will not be started")
5618 # check bridge existance
5619 _CheckInstanceBridgesExist(self, instance, node=target_node)
5621 def Exec(self, feedback_fn):
5622 """Failover an instance.
5624 The failover is done by shutting it down on its present node and
5625 starting it on the secondary.
5628 instance = self.instance
5629 primary_node = self.cfg.GetNodeInfo(instance.primary_node)
5631 source_node = instance.primary_node
5632 target_node = instance.secondary_nodes[0]
5634 if instance.admin_up:
5635 feedback_fn("* checking disk consistency between source and target")
5636 for dev in instance.disks:
5637 # for drbd, these are drbd over lvm
5638 if not _CheckDiskConsistency(self, dev, target_node, False):
5639 if not self.op.ignore_consistency:
5640 raise errors.OpExecError("Disk %s is degraded on target node,"
5641 " aborting failover." % dev.iv_name)
5643 feedback_fn("* not checking disk consistency as instance is not running")
5645 feedback_fn("* shutting down instance on source node")
5646 logging.info("Shutting down instance %s on node %s",
5647 instance.name, source_node)
5649 result = self.rpc.call_instance_shutdown(source_node, instance,
5650 self.op.shutdown_timeout)
5651 msg = result.fail_msg
5653 if self.op.ignore_consistency or primary_node.offline:
5654 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5655 " Proceeding anyway. Please make sure node"
5656 " %s is down. Error details: %s",
5657 instance.name, source_node, source_node, msg)
5659 raise errors.OpExecError("Could not shutdown instance %s on"
5661 (instance.name, source_node, msg))
5663 feedback_fn("* deactivating the instance's disks on source node")
5664 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5665 raise errors.OpExecError("Can't shut down the instance's disks.")
5667 instance.primary_node = target_node
5668 # distribute new instance config to the other nodes
5669 self.cfg.Update(instance, feedback_fn)
5671 # Only start the instance if it's marked as up
5672 if instance.admin_up:
5673 feedback_fn("* activating the instance's disks on target node")
5674 logging.info("Starting instance %s on node %s",
5675 instance.name, target_node)
5677 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5678 ignore_secondaries=True)
5680 _ShutdownInstanceDisks(self, instance)
5681 raise errors.OpExecError("Can't activate the instance's disks")
5683 feedback_fn("* starting the instance on the target node")
5684 result = self.rpc.call_instance_start(target_node, instance, None, None)
5685 msg = result.fail_msg
5687 _ShutdownInstanceDisks(self, instance)
5688 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5689 (instance.name, target_node, msg))
5692 class LUInstanceMigrate(LogicalUnit):
5693 """Migrate an instance.
5695 This is migration without shutting down, compared to the failover,
5696 which is done with shutdown.
5699 HPATH = "instance-migrate"
5700 HTYPE = constants.HTYPE_INSTANCE
5703 def ExpandNames(self):
5704 self._ExpandAndLockInstance()
5706 self.needed_locks[locking.LEVEL_NODE] = []
5707 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5709 self._migrater = TLMigrateInstance(self, self.op.instance_name,
5711 self.tasklets = [self._migrater]
5713 def DeclareLocks(self, level):
5714 if level == locking.LEVEL_NODE:
5715 self._LockInstancesNodes()
5717 def BuildHooksEnv(self):
5720 This runs on master, primary and secondary nodes of the instance.
5723 instance = self._migrater.instance
5724 source_node = instance.primary_node
5725 target_node = instance.secondary_nodes[0]
5726 env = _BuildInstanceHookEnvByObject(self, instance)
5727 env["MIGRATE_LIVE"] = self._migrater.live
5728 env["MIGRATE_CLEANUP"] = self.op.cleanup
5730 "OLD_PRIMARY": source_node,
5731 "OLD_SECONDARY": target_node,
5732 "NEW_PRIMARY": target_node,
5733 "NEW_SECONDARY": source_node,
5735 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5737 nl_post.append(source_node)
5738 return env, nl, nl_post
5741 class LUInstanceMove(LogicalUnit):
5742 """Move an instance by data-copying.
5745 HPATH = "instance-move"
5746 HTYPE = constants.HTYPE_INSTANCE
5749 def ExpandNames(self):
5750 self._ExpandAndLockInstance()
5751 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5752 self.op.target_node = target_node
5753 self.needed_locks[locking.LEVEL_NODE] = [target_node]
5754 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5756 def DeclareLocks(self, level):
5757 if level == locking.LEVEL_NODE:
5758 self._LockInstancesNodes(primary_only=True)
5760 def BuildHooksEnv(self):
5763 This runs on master, primary and secondary nodes of the instance.
5767 "TARGET_NODE": self.op.target_node,
5768 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5770 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5771 nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5772 self.op.target_node]
5775 def CheckPrereq(self):
5776 """Check prerequisites.
5778 This checks that the instance is in the cluster.
5781 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5782 assert self.instance is not None, \
5783 "Cannot retrieve locked instance %s" % self.op.instance_name
5785 node = self.cfg.GetNodeInfo(self.op.target_node)
5786 assert node is not None, \
5787 "Cannot retrieve locked node %s" % self.op.target_node
5789 self.target_node = target_node = node.name
5791 if target_node == instance.primary_node:
5792 raise errors.OpPrereqError("Instance %s is already on the node %s" %
5793 (instance.name, target_node),
5796 bep = self.cfg.GetClusterInfo().FillBE(instance)
5798 for idx, dsk in enumerate(instance.disks):
5799 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5800 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5801 " cannot copy" % idx, errors.ECODE_STATE)
5803 _CheckNodeOnline(self, target_node)
5804 _CheckNodeNotDrained(self, target_node)
5805 _CheckNodeVmCapable(self, target_node)
5807 if instance.admin_up:
5808 # check memory requirements on the secondary node
5809 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5810 instance.name, bep[constants.BE_MEMORY],
5811 instance.hypervisor)
5813 self.LogInfo("Not checking memory on the secondary node as"
5814 " instance will not be started")
5816 # check bridge existance
5817 _CheckInstanceBridgesExist(self, instance, node=target_node)
5819 def Exec(self, feedback_fn):
5820 """Move an instance.
5822 The move is done by shutting it down on its present node, copying
5823 the data over (slow) and starting it on the new node.
5826 instance = self.instance
5828 source_node = instance.primary_node
5829 target_node = self.target_node
5831 self.LogInfo("Shutting down instance %s on source node %s",
5832 instance.name, source_node)
5834 result = self.rpc.call_instance_shutdown(source_node, instance,
5835 self.op.shutdown_timeout)
5836 msg = result.fail_msg
5838 if self.op.ignore_consistency:
5839 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5840 " Proceeding anyway. Please make sure node"
5841 " %s is down. Error details: %s",
5842 instance.name, source_node, source_node, msg)
5844 raise errors.OpExecError("Could not shutdown instance %s on"
5846 (instance.name, source_node, msg))
5848 # create the target disks
5850 _CreateDisks(self, instance, target_node=target_node)
5851 except errors.OpExecError:
5852 self.LogWarning("Device creation failed, reverting...")
5854 _RemoveDisks(self, instance, target_node=target_node)
5856 self.cfg.ReleaseDRBDMinors(instance.name)
5859 cluster_name = self.cfg.GetClusterInfo().cluster_name
5862 # activate, get path, copy the data over
5863 for idx, disk in enumerate(instance.disks):
5864 self.LogInfo("Copying data for disk %d", idx)
5865 result = self.rpc.call_blockdev_assemble(target_node, disk,
5866 instance.name, True)
5868 self.LogWarning("Can't assemble newly created disk %d: %s",
5869 idx, result.fail_msg)
5870 errs.append(result.fail_msg)
5872 dev_path = result.payload
5873 result = self.rpc.call_blockdev_export(source_node, disk,
5874 target_node, dev_path,
5877 self.LogWarning("Can't copy data over for disk %d: %s",
5878 idx, result.fail_msg)
5879 errs.append(result.fail_msg)
5883 self.LogWarning("Some disks failed to copy, aborting")
5885 _RemoveDisks(self, instance, target_node=target_node)
5887 self.cfg.ReleaseDRBDMinors(instance.name)
5888 raise errors.OpExecError("Errors during disk copy: %s" %
5891 instance.primary_node = target_node
5892 self.cfg.Update(instance, feedback_fn)
5894 self.LogInfo("Removing the disks on the original node")
5895 _RemoveDisks(self, instance, target_node=source_node)
5897 # Only start the instance if it's marked as up
5898 if instance.admin_up:
5899 self.LogInfo("Starting instance %s on node %s",
5900 instance.name, target_node)
5902 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5903 ignore_secondaries=True)
5905 _ShutdownInstanceDisks(self, instance)
5906 raise errors.OpExecError("Can't activate the instance's disks")
5908 result = self.rpc.call_instance_start(target_node, instance, None, None)
5909 msg = result.fail_msg
5911 _ShutdownInstanceDisks(self, instance)
5912 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5913 (instance.name, target_node, msg))
5916 class LUNodeMigrate(LogicalUnit):
5917 """Migrate all instances from a node.
5920 HPATH = "node-migrate"
5921 HTYPE = constants.HTYPE_NODE
5924 def ExpandNames(self):
5925 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5927 self.needed_locks = {
5928 locking.LEVEL_NODE: [self.op.node_name],
5931 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5933 # Create tasklets for migrating instances for all instances on this node
5937 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5938 logging.debug("Migrating instance %s", inst.name)
5939 names.append(inst.name)
5941 tasklets.append(TLMigrateInstance(self, inst.name, False))
5943 self.tasklets = tasklets
5945 # Declare instance locks
5946 self.needed_locks[locking.LEVEL_INSTANCE] = names
5948 def DeclareLocks(self, level):
5949 if level == locking.LEVEL_NODE:
5950 self._LockInstancesNodes()
5952 def BuildHooksEnv(self):
5955 This runs on the master, the primary and all the secondaries.
5959 "NODE_NAME": self.op.node_name,
5962 nl = [self.cfg.GetMasterNode()]
5964 return (env, nl, nl)
5967 class TLMigrateInstance(Tasklet):
5968 """Tasklet class for instance migration.
5971 @ivar live: whether the migration will be done live or non-live;
5972 this variable is initalized only after CheckPrereq has run
5975 def __init__(self, lu, instance_name, cleanup):
5976 """Initializes this class.
5979 Tasklet.__init__(self, lu)
5982 self.instance_name = instance_name
5983 self.cleanup = cleanup
5984 self.live = False # will be overridden later
5986 def CheckPrereq(self):
5987 """Check prerequisites.
5989 This checks that the instance is in the cluster.
5992 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5993 instance = self.cfg.GetInstanceInfo(instance_name)
5994 assert instance is not None
5996 if instance.disk_template != constants.DT_DRBD8:
5997 raise errors.OpPrereqError("Instance's disk layout is not"
5998 " drbd8, cannot migrate.", errors.ECODE_STATE)
6000 secondary_nodes = instance.secondary_nodes
6001 if not secondary_nodes:
6002 raise errors.ConfigurationError("No secondary node but using"
6003 " drbd8 disk template")
6005 i_be = self.cfg.GetClusterInfo().FillBE(instance)
6007 target_node = secondary_nodes[0]
6008 # check memory requirements on the secondary node
6009 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
6010 instance.name, i_be[constants.BE_MEMORY],
6011 instance.hypervisor)
6013 # check bridge existance
6014 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
6016 if not self.cleanup:
6017 _CheckNodeNotDrained(self.lu, target_node)
6018 result = self.rpc.call_instance_migratable(instance.primary_node,
6020 result.Raise("Can't migrate, please use failover",
6021 prereq=True, ecode=errors.ECODE_STATE)
6023 self.instance = instance
6025 if self.lu.op.live is not None and self.lu.op.mode is not None:
6026 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
6027 " parameters are accepted",
6029 if self.lu.op.live is not None:
6031 self.lu.op.mode = constants.HT_MIGRATION_LIVE
6033 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
6034 # reset the 'live' parameter to None so that repeated
6035 # invocations of CheckPrereq do not raise an exception
6036 self.lu.op.live = None
6037 elif self.lu.op.mode is None:
6038 # read the default value from the hypervisor
6039 i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
6040 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
6042 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
6044 def _WaitUntilSync(self):
6045 """Poll with custom rpc for disk sync.
6047 This uses our own step-based rpc call.
6050 self.feedback_fn("* wait until resync is done")
6054 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
6056 self.instance.disks)
6058 for node, nres in result.items():
6059 nres.Raise("Cannot resync disks on node %s" % node)
6060 node_done, node_percent = nres.payload
6061 all_done = all_done and node_done
6062 if node_percent is not None:
6063 min_percent = min(min_percent, node_percent)
6065 if min_percent < 100:
6066 self.feedback_fn(" - progress: %.1f%%" % min_percent)
6069 def _EnsureSecondary(self, node):
6070 """Demote a node to secondary.
6073 self.feedback_fn("* switching node %s to secondary mode" % node)
6075 for dev in self.instance.disks:
6076 self.cfg.SetDiskID(dev, node)
6078 result = self.rpc.call_blockdev_close(node, self.instance.name,
6079 self.instance.disks)
6080 result.Raise("Cannot change disk to secondary on node %s" % node)
6082 def _GoStandalone(self):
6083 """Disconnect from the network.
6086 self.feedback_fn("* changing into standalone mode")
6087 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
6088 self.instance.disks)
6089 for node, nres in result.items():
6090 nres.Raise("Cannot disconnect disks node %s" % node)
6092 def _GoReconnect(self, multimaster):
6093 """Reconnect to the network.
6099 msg = "single-master"
6100 self.feedback_fn("* changing disks into %s mode" % msg)
6101 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
6102 self.instance.disks,
6103 self.instance.name, multimaster)
6104 for node, nres in result.items():
6105 nres.Raise("Cannot change disks config on node %s" % node)
6107 def _ExecCleanup(self):
6108 """Try to cleanup after a failed migration.
6110 The cleanup is done by:
6111 - check that the instance is running only on one node
6112 (and update the config if needed)
6113 - change disks on its secondary node to secondary
6114 - wait until disks are fully synchronized
6115 - disconnect from the network
6116 - change disks into single-master mode
6117 - wait again until disks are fully synchronized
6120 instance = self.instance
6121 target_node = self.target_node
6122 source_node = self.source_node
6124 # check running on only one node
6125 self.feedback_fn("* checking where the instance actually runs"
6126 " (if this hangs, the hypervisor might be in"
6128 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6129 for node, result in ins_l.items():
6130 result.Raise("Can't contact node %s" % node)
6132 runningon_source = instance.name in ins_l[source_node].payload
6133 runningon_target = instance.name in ins_l[target_node].payload
6135 if runningon_source and runningon_target:
6136 raise errors.OpExecError("Instance seems to be running on two nodes,"
6137 " or the hypervisor is confused. You will have"
6138 " to ensure manually that it runs only on one"
6139 " and restart this operation.")
6141 if not (runningon_source or runningon_target):
6142 raise errors.OpExecError("Instance does not seem to be running at all."
6143 " In this case, it's safer to repair by"
6144 " running 'gnt-instance stop' to ensure disk"
6145 " shutdown, and then restarting it.")
6147 if runningon_target:
6148 # the migration has actually succeeded, we need to update the config
6149 self.feedback_fn("* instance running on secondary node (%s),"
6150 " updating config" % target_node)
6151 instance.primary_node = target_node
6152 self.cfg.Update(instance, self.feedback_fn)
6153 demoted_node = source_node
6155 self.feedback_fn("* instance confirmed to be running on its"
6156 " primary node (%s)" % source_node)
6157 demoted_node = target_node
6159 self._EnsureSecondary(demoted_node)
6161 self._WaitUntilSync()
6162 except errors.OpExecError:
6163 # we ignore here errors, since if the device is standalone, it
6164 # won't be able to sync
6166 self._GoStandalone()
6167 self._GoReconnect(False)
6168 self._WaitUntilSync()
6170 self.feedback_fn("* done")
6172 def _RevertDiskStatus(self):
6173 """Try to revert the disk status after a failed migration.
6176 target_node = self.target_node
6178 self._EnsureSecondary(target_node)
6179 self._GoStandalone()
6180 self._GoReconnect(False)
6181 self._WaitUntilSync()
6182 except errors.OpExecError, err:
6183 self.lu.LogWarning("Migration failed and I can't reconnect the"
6184 " drives: error '%s'\n"
6185 "Please look and recover the instance status" %
6188 def _AbortMigration(self):
6189 """Call the hypervisor code to abort a started migration.
6192 instance = self.instance
6193 target_node = self.target_node
6194 migration_info = self.migration_info
6196 abort_result = self.rpc.call_finalize_migration(target_node,
6200 abort_msg = abort_result.fail_msg
6202 logging.error("Aborting migration failed on target node %s: %s",
6203 target_node, abort_msg)
6204 # Don't raise an exception here, as we stil have to try to revert the
6205 # disk status, even if this step failed.
6207 def _ExecMigration(self):
6208 """Migrate an instance.
6210 The migrate is done by:
6211 - change the disks into dual-master mode
6212 - wait until disks are fully synchronized again
6213 - migrate the instance
6214 - change disks on the new secondary node (the old primary) to secondary
6215 - wait until disks are fully synchronized
6216 - change disks into single-master mode
6219 instance = self.instance
6220 target_node = self.target_node
6221 source_node = self.source_node
6223 self.feedback_fn("* checking disk consistency between source and target")
6224 for dev in instance.disks:
6225 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6226 raise errors.OpExecError("Disk %s is degraded or not fully"
6227 " synchronized on target node,"
6228 " aborting migrate." % dev.iv_name)
6230 # First get the migration information from the remote node
6231 result = self.rpc.call_migration_info(source_node, instance)
6232 msg = result.fail_msg
6234 log_err = ("Failed fetching source migration information from %s: %s" %
6236 logging.error(log_err)
6237 raise errors.OpExecError(log_err)
6239 self.migration_info = migration_info = result.payload
6241 # Then switch the disks to master/master mode
6242 self._EnsureSecondary(target_node)
6243 self._GoStandalone()
6244 self._GoReconnect(True)
6245 self._WaitUntilSync()
6247 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6248 result = self.rpc.call_accept_instance(target_node,
6251 self.nodes_ip[target_node])
6253 msg = result.fail_msg
6255 logging.error("Instance pre-migration failed, trying to revert"
6256 " disk status: %s", msg)
6257 self.feedback_fn("Pre-migration failed, aborting")
6258 self._AbortMigration()
6259 self._RevertDiskStatus()
6260 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6261 (instance.name, msg))
6263 self.feedback_fn("* migrating instance to %s" % target_node)
6265 result = self.rpc.call_instance_migrate(source_node, instance,
6266 self.nodes_ip[target_node],
6268 msg = result.fail_msg
6270 logging.error("Instance migration failed, trying to revert"
6271 " disk status: %s", msg)
6272 self.feedback_fn("Migration failed, aborting")
6273 self._AbortMigration()
6274 self._RevertDiskStatus()
6275 raise errors.OpExecError("Could not migrate instance %s: %s" %
6276 (instance.name, msg))
6279 instance.primary_node = target_node
6280 # distribute new instance config to the other nodes
6281 self.cfg.Update(instance, self.feedback_fn)
6283 result = self.rpc.call_finalize_migration(target_node,
6287 msg = result.fail_msg
6289 logging.error("Instance migration succeeded, but finalization failed:"
6291 raise errors.OpExecError("Could not finalize instance migration: %s" %
6294 self._EnsureSecondary(source_node)
6295 self._WaitUntilSync()
6296 self._GoStandalone()
6297 self._GoReconnect(False)
6298 self._WaitUntilSync()
6300 self.feedback_fn("* done")
6302 def Exec(self, feedback_fn):
6303 """Perform the migration.
6306 feedback_fn("Migrating instance %s" % self.instance.name)
6308 self.feedback_fn = feedback_fn
6310 self.source_node = self.instance.primary_node
6311 self.target_node = self.instance.secondary_nodes[0]
6312 self.all_nodes = [self.source_node, self.target_node]
6314 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6315 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6319 return self._ExecCleanup()
6321 return self._ExecMigration()
6324 def _CreateBlockDev(lu, node, instance, device, force_create,
6326 """Create a tree of block devices on a given node.
6328 If this device type has to be created on secondaries, create it and
6331 If not, just recurse to children keeping the same 'force' value.
6333 @param lu: the lu on whose behalf we execute
6334 @param node: the node on which to create the device
6335 @type instance: L{objects.Instance}
6336 @param instance: the instance which owns the device
6337 @type device: L{objects.Disk}
6338 @param device: the device to create
6339 @type force_create: boolean
6340 @param force_create: whether to force creation of this device; this
6341 will be change to True whenever we find a device which has
6342 CreateOnSecondary() attribute
6343 @param info: the extra 'metadata' we should attach to the device
6344 (this will be represented as a LVM tag)
6345 @type force_open: boolean
6346 @param force_open: this parameter will be passes to the
6347 L{backend.BlockdevCreate} function where it specifies
6348 whether we run on primary or not, and it affects both
6349 the child assembly and the device own Open() execution
6352 if device.CreateOnSecondary():
6356 for child in device.children:
6357 _CreateBlockDev(lu, node, instance, child, force_create,
6360 if not force_create:
6363 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6366 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6367 """Create a single block device on a given node.
6369 This will not recurse over children of the device, so they must be
6372 @param lu: the lu on whose behalf we execute
6373 @param node: the node on which to create the device
6374 @type instance: L{objects.Instance}
6375 @param instance: the instance which owns the device
6376 @type device: L{objects.Disk}
6377 @param device: the device to create
6378 @param info: the extra 'metadata' we should attach to the device
6379 (this will be represented as a LVM tag)
6380 @type force_open: boolean
6381 @param force_open: this parameter will be passes to the
6382 L{backend.BlockdevCreate} function where it specifies
6383 whether we run on primary or not, and it affects both
6384 the child assembly and the device own Open() execution
6387 lu.cfg.SetDiskID(device, node)
6388 result = lu.rpc.call_blockdev_create(node, device, device.size,
6389 instance.name, force_open, info)
6390 result.Raise("Can't create block device %s on"
6391 " node %s for instance %s" % (device, node, instance.name))
6392 if device.physical_id is None:
6393 device.physical_id = result.payload
6396 def _GenerateUniqueNames(lu, exts):
6397 """Generate a suitable LV name.
6399 This will generate a logical volume name for the given instance.
6404 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6405 results.append("%s%s" % (new_id, val))
6409 def _GenerateDRBD8Branch(lu, primary, secondary, size, vgname, names, iv_name,
6411 """Generate a drbd8 device complete with its children.
6414 port = lu.cfg.AllocatePort()
6415 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6416 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6417 logical_id=(vgname, names[0]))
6418 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6419 logical_id=(vgname, names[1]))
6420 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6421 logical_id=(primary, secondary, port,
6424 children=[dev_data, dev_meta],
6429 def _GenerateDiskTemplate(lu, template_name,
6430 instance_name, primary_node,
6431 secondary_nodes, disk_info,
6432 file_storage_dir, file_driver,
6433 base_index, feedback_fn):
6434 """Generate the entire disk layout for a given template type.
6437 #TODO: compute space requirements
6439 vgname = lu.cfg.GetVGName()
6440 disk_count = len(disk_info)
6442 if template_name == constants.DT_DISKLESS:
6444 elif template_name == constants.DT_PLAIN:
6445 if len(secondary_nodes) != 0:
6446 raise errors.ProgrammerError("Wrong template configuration")
6448 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6449 for i in range(disk_count)])
6450 for idx, disk in enumerate(disk_info):
6451 disk_index = idx + base_index
6452 vg = disk.get("vg", vgname)
6453 feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
6454 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6455 logical_id=(vg, names[idx]),
6456 iv_name="disk/%d" % disk_index,
6458 disks.append(disk_dev)
6459 elif template_name == constants.DT_DRBD8:
6460 if len(secondary_nodes) != 1:
6461 raise errors.ProgrammerError("Wrong template configuration")
6462 remote_node = secondary_nodes[0]
6463 minors = lu.cfg.AllocateDRBDMinor(
6464 [primary_node, remote_node] * len(disk_info), instance_name)
6467 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6468 for i in range(disk_count)]):
6469 names.append(lv_prefix + "_data")
6470 names.append(lv_prefix + "_meta")
6471 for idx, disk in enumerate(disk_info):
6472 disk_index = idx + base_index
6473 vg = disk.get("vg", vgname)
6474 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6475 disk["size"], vg, names[idx*2:idx*2+2],
6476 "disk/%d" % disk_index,
6477 minors[idx*2], minors[idx*2+1])
6478 disk_dev.mode = disk["mode"]
6479 disks.append(disk_dev)
6480 elif template_name == constants.DT_FILE:
6481 if len(secondary_nodes) != 0:
6482 raise errors.ProgrammerError("Wrong template configuration")
6484 opcodes.RequireFileStorage()
6486 for idx, disk in enumerate(disk_info):
6487 disk_index = idx + base_index
6488 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6489 iv_name="disk/%d" % disk_index,
6490 logical_id=(file_driver,
6491 "%s/disk%d" % (file_storage_dir,
6494 disks.append(disk_dev)
6496 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6500 def _GetInstanceInfoText(instance):
6501 """Compute that text that should be added to the disk's metadata.
6504 return "originstname+%s" % instance.name
6507 def _CalcEta(time_taken, written, total_size):
6508 """Calculates the ETA based on size written and total size.
6510 @param time_taken: The time taken so far
6511 @param written: amount written so far
6512 @param total_size: The total size of data to be written
6513 @return: The remaining time in seconds
6516 avg_time = time_taken / float(written)
6517 return (total_size - written) * avg_time
6520 def _WipeDisks(lu, instance):
6521 """Wipes instance disks.
6523 @type lu: L{LogicalUnit}
6524 @param lu: the logical unit on whose behalf we execute
6525 @type instance: L{objects.Instance}
6526 @param instance: the instance whose disks we should create
6527 @return: the success of the wipe
6530 node = instance.primary_node
6531 logging.info("Pause sync of instance %s disks", instance.name)
6532 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
6534 for idx, success in enumerate(result.payload):
6536 logging.warn("pause-sync of instance %s for disks %d failed",
6540 for idx, device in enumerate(instance.disks):
6541 lu.LogInfo("* Wiping disk %d", idx)
6542 logging.info("Wiping disk %d for instance %s", idx, instance.name)
6544 # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
6545 # MAX_WIPE_CHUNK at max
6546 wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
6547 constants.MIN_WIPE_CHUNK_PERCENT)
6552 start_time = time.time()
6554 while offset < size:
6555 wipe_size = min(wipe_chunk_size, size - offset)
6556 result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
6557 result.Raise("Could not wipe disk %d at offset %d for size %d" %
6558 (idx, offset, wipe_size))
6561 if now - last_output >= 60:
6562 eta = _CalcEta(now - start_time, offset, size)
6563 lu.LogInfo(" - done: %.1f%% ETA: %s" %
6564 (offset / float(size) * 100, utils.FormatSeconds(eta)))
6567 logging.info("Resume sync of instance %s disks", instance.name)
6569 result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
6571 for idx, success in enumerate(result.payload):
6573 lu.LogWarning("Warning: Resume sync of disk %d failed. Please have a"
6574 " look at the status and troubleshoot the issue.", idx)
6575 logging.warn("resume-sync of instance %s for disks %d failed",
6579 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6580 """Create all disks for an instance.
6582 This abstracts away some work from AddInstance.
6584 @type lu: L{LogicalUnit}
6585 @param lu: the logical unit on whose behalf we execute
6586 @type instance: L{objects.Instance}
6587 @param instance: the instance whose disks we should create
6589 @param to_skip: list of indices to skip
6590 @type target_node: string
6591 @param target_node: if passed, overrides the target node for creation
6593 @return: the success of the creation
6596 info = _GetInstanceInfoText(instance)
6597 if target_node is None:
6598 pnode = instance.primary_node
6599 all_nodes = instance.all_nodes
6604 if instance.disk_template == constants.DT_FILE:
6605 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6606 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6608 result.Raise("Failed to create directory '%s' on"
6609 " node %s" % (file_storage_dir, pnode))
6611 # Note: this needs to be kept in sync with adding of disks in
6612 # LUInstanceSetParams
6613 for idx, device in enumerate(instance.disks):
6614 if to_skip and idx in to_skip:
6616 logging.info("Creating volume %s for instance %s",
6617 device.iv_name, instance.name)
6619 for node in all_nodes:
6620 f_create = node == pnode
6621 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6624 def _RemoveDisks(lu, instance, target_node=None):
6625 """Remove all disks for an instance.
6627 This abstracts away some work from `AddInstance()` and
6628 `RemoveInstance()`. Note that in case some of the devices couldn't
6629 be removed, the removal will continue with the other ones (compare
6630 with `_CreateDisks()`).
6632 @type lu: L{LogicalUnit}
6633 @param lu: the logical unit on whose behalf we execute
6634 @type instance: L{objects.Instance}
6635 @param instance: the instance whose disks we should remove
6636 @type target_node: string
6637 @param target_node: used to override the node on which to remove the disks
6639 @return: the success of the removal
6642 logging.info("Removing block devices for instance %s", instance.name)
6645 for device in instance.disks:
6647 edata = [(target_node, device)]
6649 edata = device.ComputeNodeTree(instance.primary_node)
6650 for node, disk in edata:
6651 lu.cfg.SetDiskID(disk, node)
6652 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6654 lu.LogWarning("Could not remove block device %s on node %s,"
6655 " continuing anyway: %s", device.iv_name, node, msg)
6658 if instance.disk_template == constants.DT_FILE:
6659 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6663 tgt = instance.primary_node
6664 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6666 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6667 file_storage_dir, instance.primary_node, result.fail_msg)
6673 def _ComputeDiskSizePerVG(disk_template, disks):
6674 """Compute disk size requirements in the volume group
6677 def _compute(disks, payload):
6678 """Universal algorithm
6683 vgs[disk["vg"]] = vgs.get("vg", 0) + disk["size"] + payload
6687 # Required free disk space as a function of disk and swap space
6689 constants.DT_DISKLESS: {},
6690 constants.DT_PLAIN: _compute(disks, 0),
6691 # 128 MB are added for drbd metadata for each disk
6692 constants.DT_DRBD8: _compute(disks, 128),
6693 constants.DT_FILE: {},
6696 if disk_template not in req_size_dict:
6697 raise errors.ProgrammerError("Disk template '%s' size requirement"
6698 " is unknown" % disk_template)
6700 return req_size_dict[disk_template]
6703 def _ComputeDiskSize(disk_template, disks):
6704 """Compute disk size requirements in the volume group
6707 # Required free disk space as a function of disk and swap space
6709 constants.DT_DISKLESS: None,
6710 constants.DT_PLAIN: sum(d["size"] for d in disks),
6711 # 128 MB are added for drbd metadata for each disk
6712 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6713 constants.DT_FILE: None,
6716 if disk_template not in req_size_dict:
6717 raise errors.ProgrammerError("Disk template '%s' size requirement"
6718 " is unknown" % disk_template)
6720 return req_size_dict[disk_template]
6723 def _CheckHVParams(lu, nodenames, hvname, hvparams):
6724 """Hypervisor parameter validation.
6726 This function abstract the hypervisor parameter validation to be
6727 used in both instance create and instance modify.
6729 @type lu: L{LogicalUnit}
6730 @param lu: the logical unit for which we check
6731 @type nodenames: list
6732 @param nodenames: the list of nodes on which we should check
6733 @type hvname: string
6734 @param hvname: the name of the hypervisor we should use
6735 @type hvparams: dict
6736 @param hvparams: the parameters which we need to check
6737 @raise errors.OpPrereqError: if the parameters are not valid
6740 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6743 for node in nodenames:
6747 info.Raise("Hypervisor parameter validation failed on node %s" % node)
6750 def _CheckOSParams(lu, required, nodenames, osname, osparams):
6751 """OS parameters validation.
6753 @type lu: L{LogicalUnit}
6754 @param lu: the logical unit for which we check
6755 @type required: boolean
6756 @param required: whether the validation should fail if the OS is not
6758 @type nodenames: list
6759 @param nodenames: the list of nodes on which we should check
6760 @type osname: string
6761 @param osname: the name of the hypervisor we should use
6762 @type osparams: dict
6763 @param osparams: the parameters which we need to check
6764 @raise errors.OpPrereqError: if the parameters are not valid
6767 result = lu.rpc.call_os_validate(required, nodenames, osname,
6768 [constants.OS_VALIDATE_PARAMETERS],
6770 for node, nres in result.items():
6771 # we don't check for offline cases since this should be run only
6772 # against the master node and/or an instance's nodes
6773 nres.Raise("OS Parameters validation failed on node %s" % node)
6774 if not nres.payload:
6775 lu.LogInfo("OS %s not found on node %s, validation skipped",
6779 class LUInstanceCreate(LogicalUnit):
6780 """Create an instance.
6783 HPATH = "instance-add"
6784 HTYPE = constants.HTYPE_INSTANCE
6787 def CheckArguments(self):
6791 # do not require name_check to ease forward/backward compatibility
6793 if self.op.no_install and self.op.start:
6794 self.LogInfo("No-installation mode selected, disabling startup")
6795 self.op.start = False
6796 # validate/normalize the instance name
6797 self.op.instance_name = \
6798 netutils.Hostname.GetNormalizedName(self.op.instance_name)
6800 if self.op.ip_check and not self.op.name_check:
6801 # TODO: make the ip check more flexible and not depend on the name check
6802 raise errors.OpPrereqError("Cannot do ip check without a name check",
6805 # check nics' parameter names
6806 for nic in self.op.nics:
6807 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6809 # check disks. parameter names and consistent adopt/no-adopt strategy
6810 has_adopt = has_no_adopt = False
6811 for disk in self.op.disks:
6812 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6817 if has_adopt and has_no_adopt:
6818 raise errors.OpPrereqError("Either all disks are adopted or none is",
6821 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6822 raise errors.OpPrereqError("Disk adoption is not supported for the"
6823 " '%s' disk template" %
6824 self.op.disk_template,
6826 if self.op.iallocator is not None:
6827 raise errors.OpPrereqError("Disk adoption not allowed with an"
6828 " iallocator script", errors.ECODE_INVAL)
6829 if self.op.mode == constants.INSTANCE_IMPORT:
6830 raise errors.OpPrereqError("Disk adoption not allowed for"
6831 " instance import", errors.ECODE_INVAL)
6833 self.adopt_disks = has_adopt
6835 # instance name verification
6836 if self.op.name_check:
6837 self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
6838 self.op.instance_name = self.hostname1.name
6839 # used in CheckPrereq for ip ping check
6840 self.check_ip = self.hostname1.ip
6842 self.check_ip = None
6844 # file storage checks
6845 if (self.op.file_driver and
6846 not self.op.file_driver in constants.FILE_DRIVER):
6847 raise errors.OpPrereqError("Invalid file driver name '%s'" %
6848 self.op.file_driver, errors.ECODE_INVAL)
6850 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6851 raise errors.OpPrereqError("File storage directory path not absolute",
6854 ### Node/iallocator related checks
6855 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6857 if self.op.pnode is not None:
6858 if self.op.disk_template in constants.DTS_NET_MIRROR:
6859 if self.op.snode is None:
6860 raise errors.OpPrereqError("The networked disk templates need"
6861 " a mirror node", errors.ECODE_INVAL)
6863 self.LogWarning("Secondary node will be ignored on non-mirrored disk"
6865 self.op.snode = None
6867 self._cds = _GetClusterDomainSecret()
6869 if self.op.mode == constants.INSTANCE_IMPORT:
6870 # On import force_variant must be True, because if we forced it at
6871 # initial install, our only chance when importing it back is that it
6873 self.op.force_variant = True
6875 if self.op.no_install:
6876 self.LogInfo("No-installation mode has no effect during import")
6878 elif self.op.mode == constants.INSTANCE_CREATE:
6879 if self.op.os_type is None:
6880 raise errors.OpPrereqError("No guest OS specified",
6882 if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
6883 raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
6884 " installation" % self.op.os_type,
6886 if self.op.disk_template is None:
6887 raise errors.OpPrereqError("No disk template specified",
6890 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6891 # Check handshake to ensure both clusters have the same domain secret
6892 src_handshake = self.op.source_handshake
6893 if not src_handshake:
6894 raise errors.OpPrereqError("Missing source handshake",
6897 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6900 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6903 # Load and check source CA
6904 self.source_x509_ca_pem = self.op.source_x509_ca
6905 if not self.source_x509_ca_pem:
6906 raise errors.OpPrereqError("Missing source X509 CA",
6910 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6912 except OpenSSL.crypto.Error, err:
6913 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6914 (err, ), errors.ECODE_INVAL)
6916 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6917 if errcode is not None:
6918 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6921 self.source_x509_ca = cert
6923 src_instance_name = self.op.source_instance_name
6924 if not src_instance_name:
6925 raise errors.OpPrereqError("Missing source instance name",
6928 self.source_instance_name = \
6929 netutils.GetHostname(name=src_instance_name).name
6932 raise errors.OpPrereqError("Invalid instance creation mode %r" %
6933 self.op.mode, errors.ECODE_INVAL)
6935 def ExpandNames(self):
6936 """ExpandNames for CreateInstance.
6938 Figure out the right locks for instance creation.
6941 self.needed_locks = {}
6943 instance_name = self.op.instance_name
6944 # this is just a preventive check, but someone might still add this
6945 # instance in the meantime, and creation will fail at lock-add time
6946 if instance_name in self.cfg.GetInstanceList():
6947 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6948 instance_name, errors.ECODE_EXISTS)
6950 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6952 if self.op.iallocator:
6953 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6955 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6956 nodelist = [self.op.pnode]
6957 if self.op.snode is not None:
6958 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6959 nodelist.append(self.op.snode)
6960 self.needed_locks[locking.LEVEL_NODE] = nodelist
6962 # in case of import lock the source node too
6963 if self.op.mode == constants.INSTANCE_IMPORT:
6964 src_node = self.op.src_node
6965 src_path = self.op.src_path
6967 if src_path is None:
6968 self.op.src_path = src_path = self.op.instance_name
6970 if src_node is None:
6971 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6972 self.op.src_node = None
6973 if os.path.isabs(src_path):
6974 raise errors.OpPrereqError("Importing an instance from an absolute"
6975 " path requires a source node option.",
6978 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6979 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6980 self.needed_locks[locking.LEVEL_NODE].append(src_node)
6981 if not os.path.isabs(src_path):
6982 self.op.src_path = src_path = \
6983 utils.PathJoin(constants.EXPORT_DIR, src_path)
6985 def _RunAllocator(self):
6986 """Run the allocator based on input opcode.
6989 nics = [n.ToDict() for n in self.nics]
6990 ial = IAllocator(self.cfg, self.rpc,
6991 mode=constants.IALLOCATOR_MODE_ALLOC,
6992 name=self.op.instance_name,
6993 disk_template=self.op.disk_template,
6996 vcpus=self.be_full[constants.BE_VCPUS],
6997 mem_size=self.be_full[constants.BE_MEMORY],
7000 hypervisor=self.op.hypervisor,
7003 ial.Run(self.op.iallocator)
7006 raise errors.OpPrereqError("Can't compute nodes using"
7007 " iallocator '%s': %s" %
7008 (self.op.iallocator, ial.info),
7010 if len(ial.result) != ial.required_nodes:
7011 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7012 " of nodes (%s), required %s" %
7013 (self.op.iallocator, len(ial.result),
7014 ial.required_nodes), errors.ECODE_FAULT)
7015 self.op.pnode = ial.result[0]
7016 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7017 self.op.instance_name, self.op.iallocator,
7018 utils.CommaJoin(ial.result))
7019 if ial.required_nodes == 2:
7020 self.op.snode = ial.result[1]
7022 def BuildHooksEnv(self):
7025 This runs on master, primary and secondary nodes of the instance.
7029 "ADD_MODE": self.op.mode,
7031 if self.op.mode == constants.INSTANCE_IMPORT:
7032 env["SRC_NODE"] = self.op.src_node
7033 env["SRC_PATH"] = self.op.src_path
7034 env["SRC_IMAGES"] = self.src_images
7036 env.update(_BuildInstanceHookEnv(
7037 name=self.op.instance_name,
7038 primary_node=self.op.pnode,
7039 secondary_nodes=self.secondaries,
7040 status=self.op.start,
7041 os_type=self.op.os_type,
7042 memory=self.be_full[constants.BE_MEMORY],
7043 vcpus=self.be_full[constants.BE_VCPUS],
7044 nics=_NICListToTuple(self, self.nics),
7045 disk_template=self.op.disk_template,
7046 disks=[(d["size"], d["mode"]) for d in self.disks],
7049 hypervisor_name=self.op.hypervisor,
7052 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
7056 def _ReadExportInfo(self):
7057 """Reads the export information from disk.
7059 It will override the opcode source node and path with the actual
7060 information, if these two were not specified before.
7062 @return: the export information
7065 assert self.op.mode == constants.INSTANCE_IMPORT
7067 src_node = self.op.src_node
7068 src_path = self.op.src_path
7070 if src_node is None:
7071 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
7072 exp_list = self.rpc.call_export_list(locked_nodes)
7074 for node in exp_list:
7075 if exp_list[node].fail_msg:
7077 if src_path in exp_list[node].payload:
7079 self.op.src_node = src_node = node
7080 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
7084 raise errors.OpPrereqError("No export found for relative path %s" %
7085 src_path, errors.ECODE_INVAL)
7087 _CheckNodeOnline(self, src_node)
7088 result = self.rpc.call_export_info(src_node, src_path)
7089 result.Raise("No export or invalid export found in dir %s" % src_path)
7091 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
7092 if not export_info.has_section(constants.INISECT_EXP):
7093 raise errors.ProgrammerError("Corrupted export config",
7094 errors.ECODE_ENVIRON)
7096 ei_version = export_info.get(constants.INISECT_EXP, "version")
7097 if (int(ei_version) != constants.EXPORT_VERSION):
7098 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
7099 (ei_version, constants.EXPORT_VERSION),
7100 errors.ECODE_ENVIRON)
7103 def _ReadExportParams(self, einfo):
7104 """Use export parameters as defaults.
7106 In case the opcode doesn't specify (as in override) some instance
7107 parameters, then try to use them from the export information, if
7111 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
7113 if self.op.disk_template is None:
7114 if einfo.has_option(constants.INISECT_INS, "disk_template"):
7115 self.op.disk_template = einfo.get(constants.INISECT_INS,
7118 raise errors.OpPrereqError("No disk template specified and the export"
7119 " is missing the disk_template information",
7122 if not self.op.disks:
7123 if einfo.has_option(constants.INISECT_INS, "disk_count"):
7125 # TODO: import the disk iv_name too
7126 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
7127 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
7128 disks.append({"size": disk_sz})
7129 self.op.disks = disks
7131 raise errors.OpPrereqError("No disk info specified and the export"
7132 " is missing the disk information",
7135 if (not self.op.nics and
7136 einfo.has_option(constants.INISECT_INS, "nic_count")):
7138 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
7140 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
7141 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
7146 if (self.op.hypervisor is None and
7147 einfo.has_option(constants.INISECT_INS, "hypervisor")):
7148 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
7149 if einfo.has_section(constants.INISECT_HYP):
7150 # use the export parameters but do not override the ones
7151 # specified by the user
7152 for name, value in einfo.items(constants.INISECT_HYP):
7153 if name not in self.op.hvparams:
7154 self.op.hvparams[name] = value
7156 if einfo.has_section(constants.INISECT_BEP):
7157 # use the parameters, without overriding
7158 for name, value in einfo.items(constants.INISECT_BEP):
7159 if name not in self.op.beparams:
7160 self.op.beparams[name] = value
7162 # try to read the parameters old style, from the main section
7163 for name in constants.BES_PARAMETERS:
7164 if (name not in self.op.beparams and
7165 einfo.has_option(constants.INISECT_INS, name)):
7166 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
7168 if einfo.has_section(constants.INISECT_OSP):
7169 # use the parameters, without overriding
7170 for name, value in einfo.items(constants.INISECT_OSP):
7171 if name not in self.op.osparams:
7172 self.op.osparams[name] = value
7174 def _RevertToDefaults(self, cluster):
7175 """Revert the instance parameters to the default values.
7179 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7180 for name in self.op.hvparams.keys():
7181 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7182 del self.op.hvparams[name]
7184 be_defs = cluster.SimpleFillBE({})
7185 for name in self.op.beparams.keys():
7186 if name in be_defs and be_defs[name] == self.op.beparams[name]:
7187 del self.op.beparams[name]
7189 nic_defs = cluster.SimpleFillNIC({})
7190 for nic in self.op.nics:
7191 for name in constants.NICS_PARAMETERS:
7192 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7195 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7196 for name in self.op.osparams.keys():
7197 if name in os_defs and os_defs[name] == self.op.osparams[name]:
7198 del self.op.osparams[name]
7200 def CheckPrereq(self):
7201 """Check prerequisites.
7204 if self.op.mode == constants.INSTANCE_IMPORT:
7205 export_info = self._ReadExportInfo()
7206 self._ReadExportParams(export_info)
7208 if (not self.cfg.GetVGName() and
7209 self.op.disk_template not in constants.DTS_NOT_LVM):
7210 raise errors.OpPrereqError("Cluster does not support lvm-based"
7211 " instances", errors.ECODE_STATE)
7213 if self.op.hypervisor is None:
7214 self.op.hypervisor = self.cfg.GetHypervisorType()
7216 cluster = self.cfg.GetClusterInfo()
7217 enabled_hvs = cluster.enabled_hypervisors
7218 if self.op.hypervisor not in enabled_hvs:
7219 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7220 " cluster (%s)" % (self.op.hypervisor,
7221 ",".join(enabled_hvs)),
7224 # check hypervisor parameter syntax (locally)
7225 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7226 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7228 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7229 hv_type.CheckParameterSyntax(filled_hvp)
7230 self.hv_full = filled_hvp
7231 # check that we don't specify global parameters on an instance
7232 _CheckGlobalHvParams(self.op.hvparams)
7234 # fill and remember the beparams dict
7235 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7236 self.be_full = cluster.SimpleFillBE(self.op.beparams)
7238 # build os parameters
7239 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7241 # now that hvp/bep are in final format, let's reset to defaults,
7243 if self.op.identify_defaults:
7244 self._RevertToDefaults(cluster)
7248 for idx, nic in enumerate(self.op.nics):
7249 nic_mode_req = nic.get("mode", None)
7250 nic_mode = nic_mode_req
7251 if nic_mode is None:
7252 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7254 # in routed mode, for the first nic, the default ip is 'auto'
7255 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7256 default_ip_mode = constants.VALUE_AUTO
7258 default_ip_mode = constants.VALUE_NONE
7260 # ip validity checks
7261 ip = nic.get("ip", default_ip_mode)
7262 if ip is None or ip.lower() == constants.VALUE_NONE:
7264 elif ip.lower() == constants.VALUE_AUTO:
7265 if not self.op.name_check:
7266 raise errors.OpPrereqError("IP address set to auto but name checks"
7267 " have been skipped",
7269 nic_ip = self.hostname1.ip
7271 if not netutils.IPAddress.IsValid(ip):
7272 raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7276 # TODO: check the ip address for uniqueness
7277 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7278 raise errors.OpPrereqError("Routed nic mode requires an ip address",
7281 # MAC address verification
7282 mac = nic.get("mac", constants.VALUE_AUTO)
7283 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7284 mac = utils.NormalizeAndValidateMac(mac)
7287 self.cfg.ReserveMAC(mac, self.proc.GetECId())
7288 except errors.ReservationError:
7289 raise errors.OpPrereqError("MAC address %s already in use"
7290 " in cluster" % mac,
7291 errors.ECODE_NOTUNIQUE)
7293 # bridge verification
7294 bridge = nic.get("bridge", None)
7295 link = nic.get("link", None)
7297 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7298 " at the same time", errors.ECODE_INVAL)
7299 elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7300 raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7307 nicparams[constants.NIC_MODE] = nic_mode_req
7309 nicparams[constants.NIC_LINK] = link
7311 check_params = cluster.SimpleFillNIC(nicparams)
7312 objects.NIC.CheckParameterSyntax(check_params)
7313 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7315 # disk checks/pre-build
7317 for disk in self.op.disks:
7318 mode = disk.get("mode", constants.DISK_RDWR)
7319 if mode not in constants.DISK_ACCESS_SET:
7320 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7321 mode, errors.ECODE_INVAL)
7322 size = disk.get("size", None)
7324 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7327 except (TypeError, ValueError):
7328 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7330 vg = disk.get("vg", self.cfg.GetVGName())
7331 new_disk = {"size": size, "mode": mode, "vg": vg}
7333 new_disk["adopt"] = disk["adopt"]
7334 self.disks.append(new_disk)
7336 if self.op.mode == constants.INSTANCE_IMPORT:
7338 # Check that the new instance doesn't have less disks than the export
7339 instance_disks = len(self.disks)
7340 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7341 if instance_disks < export_disks:
7342 raise errors.OpPrereqError("Not enough disks to import."
7343 " (instance: %d, export: %d)" %
7344 (instance_disks, export_disks),
7348 for idx in range(export_disks):
7349 option = 'disk%d_dump' % idx
7350 if export_info.has_option(constants.INISECT_INS, option):
7351 # FIXME: are the old os-es, disk sizes, etc. useful?
7352 export_name = export_info.get(constants.INISECT_INS, option)
7353 image = utils.PathJoin(self.op.src_path, export_name)
7354 disk_images.append(image)
7356 disk_images.append(False)
7358 self.src_images = disk_images
7360 old_name = export_info.get(constants.INISECT_INS, 'name')
7362 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7363 except (TypeError, ValueError), err:
7364 raise errors.OpPrereqError("Invalid export file, nic_count is not"
7365 " an integer: %s" % str(err),
7367 if self.op.instance_name == old_name:
7368 for idx, nic in enumerate(self.nics):
7369 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7370 nic_mac_ini = 'nic%d_mac' % idx
7371 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7373 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7375 # ip ping checks (we use the same ip that was resolved in ExpandNames)
7376 if self.op.ip_check:
7377 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7378 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7379 (self.check_ip, self.op.instance_name),
7380 errors.ECODE_NOTUNIQUE)
7382 #### mac address generation
7383 # By generating here the mac address both the allocator and the hooks get
7384 # the real final mac address rather than the 'auto' or 'generate' value.
7385 # There is a race condition between the generation and the instance object
7386 # creation, which means that we know the mac is valid now, but we're not
7387 # sure it will be when we actually add the instance. If things go bad
7388 # adding the instance will abort because of a duplicate mac, and the
7389 # creation job will fail.
7390 for nic in self.nics:
7391 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7392 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7396 if self.op.iallocator is not None:
7397 self._RunAllocator()
7399 #### node related checks
7401 # check primary node
7402 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7403 assert self.pnode is not None, \
7404 "Cannot retrieve locked node %s" % self.op.pnode
7406 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7407 pnode.name, errors.ECODE_STATE)
7409 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7410 pnode.name, errors.ECODE_STATE)
7411 if not pnode.vm_capable:
7412 raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
7413 " '%s'" % pnode.name, errors.ECODE_STATE)
7415 self.secondaries = []
7417 # mirror node verification
7418 if self.op.disk_template in constants.DTS_NET_MIRROR:
7419 if self.op.snode == pnode.name:
7420 raise errors.OpPrereqError("The secondary node cannot be the"
7421 " primary node.", errors.ECODE_INVAL)
7422 _CheckNodeOnline(self, self.op.snode)
7423 _CheckNodeNotDrained(self, self.op.snode)
7424 _CheckNodeVmCapable(self, self.op.snode)
7425 self.secondaries.append(self.op.snode)
7427 nodenames = [pnode.name] + self.secondaries
7429 if not self.adopt_disks:
7430 # Check lv size requirements, if not adopting
7431 req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
7432 _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
7434 else: # instead, we must check the adoption data
7435 all_lvs = set([i["vg"] + "/" + i["adopt"] for i in self.disks])
7436 if len(all_lvs) != len(self.disks):
7437 raise errors.OpPrereqError("Duplicate volume names given for adoption",
7439 for lv_name in all_lvs:
7441 # FIXME: lv_name here is "vg/lv" need to ensure that other calls
7442 # to ReserveLV uses the same syntax
7443 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7444 except errors.ReservationError:
7445 raise errors.OpPrereqError("LV named %s used by another instance" %
7446 lv_name, errors.ECODE_NOTUNIQUE)
7448 vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
7449 vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
7451 node_lvs = self.rpc.call_lv_list([pnode.name],
7452 vg_names.payload.keys())[pnode.name]
7453 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7454 node_lvs = node_lvs.payload
7456 delta = all_lvs.difference(node_lvs.keys())
7458 raise errors.OpPrereqError("Missing logical volume(s): %s" %
7459 utils.CommaJoin(delta),
7461 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7463 raise errors.OpPrereqError("Online logical volumes found, cannot"
7464 " adopt: %s" % utils.CommaJoin(online_lvs),
7466 # update the size of disk based on what is found
7467 for dsk in self.disks:
7468 dsk["size"] = int(float(node_lvs[dsk["vg"] + "/" + dsk["adopt"]][0]))
7470 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7472 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7473 # check OS parameters (remotely)
7474 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7476 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7478 # memory check on primary node
7480 _CheckNodeFreeMemory(self, self.pnode.name,
7481 "creating instance %s" % self.op.instance_name,
7482 self.be_full[constants.BE_MEMORY],
7485 self.dry_run_result = list(nodenames)
7487 def Exec(self, feedback_fn):
7488 """Create and add the instance to the cluster.
7491 instance = self.op.instance_name
7492 pnode_name = self.pnode.name
7494 ht_kind = self.op.hypervisor
7495 if ht_kind in constants.HTS_REQ_PORT:
7496 network_port = self.cfg.AllocatePort()
7500 if constants.ENABLE_FILE_STORAGE:
7501 # this is needed because os.path.join does not accept None arguments
7502 if self.op.file_storage_dir is None:
7503 string_file_storage_dir = ""
7505 string_file_storage_dir = self.op.file_storage_dir
7507 # build the full file storage dir path
7508 file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7509 string_file_storage_dir, instance)
7511 file_storage_dir = ""
7513 disks = _GenerateDiskTemplate(self,
7514 self.op.disk_template,
7515 instance, pnode_name,
7519 self.op.file_driver,
7523 iobj = objects.Instance(name=instance, os=self.op.os_type,
7524 primary_node=pnode_name,
7525 nics=self.nics, disks=disks,
7526 disk_template=self.op.disk_template,
7528 network_port=network_port,
7529 beparams=self.op.beparams,
7530 hvparams=self.op.hvparams,
7531 hypervisor=self.op.hypervisor,
7532 osparams=self.op.osparams,
7535 if self.adopt_disks:
7536 # rename LVs to the newly-generated names; we need to construct
7537 # 'fake' LV disks with the old data, plus the new unique_id
7538 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7540 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7541 rename_to.append(t_dsk.logical_id)
7542 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7543 self.cfg.SetDiskID(t_dsk, pnode_name)
7544 result = self.rpc.call_blockdev_rename(pnode_name,
7545 zip(tmp_disks, rename_to))
7546 result.Raise("Failed to rename adoped LVs")
7548 feedback_fn("* creating instance disks...")
7550 _CreateDisks(self, iobj)
7551 except errors.OpExecError:
7552 self.LogWarning("Device creation failed, reverting...")
7554 _RemoveDisks(self, iobj)
7556 self.cfg.ReleaseDRBDMinors(instance)
7559 if self.cfg.GetClusterInfo().prealloc_wipe_disks:
7560 feedback_fn("* wiping instance disks...")
7562 _WipeDisks(self, iobj)
7563 except errors.OpExecError:
7564 self.LogWarning("Device wiping failed, reverting...")
7566 _RemoveDisks(self, iobj)
7568 self.cfg.ReleaseDRBDMinors(instance)
7571 feedback_fn("adding instance %s to cluster config" % instance)
7573 self.cfg.AddInstance(iobj, self.proc.GetECId())
7575 # Declare that we don't want to remove the instance lock anymore, as we've
7576 # added the instance to the config
7577 del self.remove_locks[locking.LEVEL_INSTANCE]
7578 # Unlock all the nodes
7579 if self.op.mode == constants.INSTANCE_IMPORT:
7580 nodes_keep = [self.op.src_node]
7581 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7582 if node != self.op.src_node]
7583 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7584 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7586 self.context.glm.release(locking.LEVEL_NODE)
7587 del self.acquired_locks[locking.LEVEL_NODE]
7589 if self.op.wait_for_sync:
7590 disk_abort = not _WaitForSync(self, iobj)
7591 elif iobj.disk_template in constants.DTS_NET_MIRROR:
7592 # make sure the disks are not degraded (still sync-ing is ok)
7594 feedback_fn("* checking mirrors status")
7595 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7600 _RemoveDisks(self, iobj)
7601 self.cfg.RemoveInstance(iobj.name)
7602 # Make sure the instance lock gets removed
7603 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7604 raise errors.OpExecError("There are some degraded disks for"
7607 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7608 if self.op.mode == constants.INSTANCE_CREATE:
7609 if not self.op.no_install:
7610 feedback_fn("* running the instance OS create scripts...")
7611 # FIXME: pass debug option from opcode to backend
7612 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7613 self.op.debug_level)
7614 result.Raise("Could not add os for instance %s"
7615 " on node %s" % (instance, pnode_name))
7617 elif self.op.mode == constants.INSTANCE_IMPORT:
7618 feedback_fn("* running the instance OS import scripts...")
7622 for idx, image in enumerate(self.src_images):
7626 # FIXME: pass debug option from opcode to backend
7627 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7628 constants.IEIO_FILE, (image, ),
7629 constants.IEIO_SCRIPT,
7630 (iobj.disks[idx], idx),
7632 transfers.append(dt)
7635 masterd.instance.TransferInstanceData(self, feedback_fn,
7636 self.op.src_node, pnode_name,
7637 self.pnode.secondary_ip,
7639 if not compat.all(import_result):
7640 self.LogWarning("Some disks for instance %s on node %s were not"
7641 " imported successfully" % (instance, pnode_name))
7643 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7644 feedback_fn("* preparing remote import...")
7645 # The source cluster will stop the instance before attempting to make a
7646 # connection. In some cases stopping an instance can take a long time,
7647 # hence the shutdown timeout is added to the connection timeout.
7648 connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
7649 self.op.source_shutdown_timeout)
7650 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7652 assert iobj.primary_node == self.pnode.name
7654 masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
7655 self.source_x509_ca,
7656 self._cds, timeouts)
7657 if not compat.all(disk_results):
7658 # TODO: Should the instance still be started, even if some disks
7659 # failed to import (valid for local imports, too)?
7660 self.LogWarning("Some disks for instance %s on node %s were not"
7661 " imported successfully" % (instance, pnode_name))
7663 # Run rename script on newly imported instance
7664 assert iobj.name == instance
7665 feedback_fn("Running rename script for %s" % instance)
7666 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7667 self.source_instance_name,
7668 self.op.debug_level)
7670 self.LogWarning("Failed to run rename script for %s on node"
7671 " %s: %s" % (instance, pnode_name, result.fail_msg))
7674 # also checked in the prereq part
7675 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7679 iobj.admin_up = True
7680 self.cfg.Update(iobj, feedback_fn)
7681 logging.info("Starting instance %s on node %s", instance, pnode_name)
7682 feedback_fn("* starting instance...")
7683 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7684 result.Raise("Could not start instance")
7686 return list(iobj.all_nodes)
7689 class LUInstanceConsole(NoHooksLU):
7690 """Connect to an instance's console.
7692 This is somewhat special in that it returns the command line that
7693 you need to run on the master node in order to connect to the
7699 def ExpandNames(self):
7700 self._ExpandAndLockInstance()
7702 def CheckPrereq(self):
7703 """Check prerequisites.
7705 This checks that the instance is in the cluster.
7708 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7709 assert self.instance is not None, \
7710 "Cannot retrieve locked instance %s" % self.op.instance_name
7711 _CheckNodeOnline(self, self.instance.primary_node)
7713 def Exec(self, feedback_fn):
7714 """Connect to the console of an instance
7717 instance = self.instance
7718 node = instance.primary_node
7720 node_insts = self.rpc.call_instance_list([node],
7721 [instance.hypervisor])[node]
7722 node_insts.Raise("Can't get node information from %s" % node)
7724 if instance.name not in node_insts.payload:
7725 if instance.admin_up:
7726 state = "ERROR_down"
7728 state = "ADMIN_down"
7729 raise errors.OpExecError("Instance %s is not running (state %s)" %
7730 (instance.name, state))
7732 logging.debug("Connecting to console of %s on %s", instance.name, node)
7734 hyper = hypervisor.GetHypervisor(instance.hypervisor)
7735 cluster = self.cfg.GetClusterInfo()
7736 # beparams and hvparams are passed separately, to avoid editing the
7737 # instance and then saving the defaults in the instance itself.
7738 hvparams = cluster.FillHV(instance)
7739 beparams = cluster.FillBE(instance)
7740 console = hyper.GetInstanceConsole(instance, hvparams, beparams)
7742 assert console.instance == instance.name
7743 assert console.Validate()
7745 return console.ToDict()
7748 class LUInstanceReplaceDisks(LogicalUnit):
7749 """Replace the disks of an instance.
7752 HPATH = "mirrors-replace"
7753 HTYPE = constants.HTYPE_INSTANCE
7756 def CheckArguments(self):
7757 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7760 def ExpandNames(self):
7761 self._ExpandAndLockInstance()
7763 if self.op.iallocator is not None:
7764 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7766 elif self.op.remote_node is not None:
7767 remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7768 self.op.remote_node = remote_node
7770 # Warning: do not remove the locking of the new secondary here
7771 # unless DRBD8.AddChildren is changed to work in parallel;
7772 # currently it doesn't since parallel invocations of
7773 # FindUnusedMinor will conflict
7774 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7775 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7778 self.needed_locks[locking.LEVEL_NODE] = []
7779 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7781 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7782 self.op.iallocator, self.op.remote_node,
7783 self.op.disks, False, self.op.early_release)
7785 self.tasklets = [self.replacer]
7787 def DeclareLocks(self, level):
7788 # If we're not already locking all nodes in the set we have to declare the
7789 # instance's primary/secondary nodes.
7790 if (level == locking.LEVEL_NODE and
7791 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7792 self._LockInstancesNodes()
7794 def BuildHooksEnv(self):
7797 This runs on the master, the primary and all the secondaries.
7800 instance = self.replacer.instance
7802 "MODE": self.op.mode,
7803 "NEW_SECONDARY": self.op.remote_node,
7804 "OLD_SECONDARY": instance.secondary_nodes[0],
7806 env.update(_BuildInstanceHookEnvByObject(self, instance))
7808 self.cfg.GetMasterNode(),
7809 instance.primary_node,
7811 if self.op.remote_node is not None:
7812 nl.append(self.op.remote_node)
7816 class TLReplaceDisks(Tasklet):
7817 """Replaces disks for an instance.
7819 Note: Locking is not within the scope of this class.
7822 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7823 disks, delay_iallocator, early_release):
7824 """Initializes this class.
7827 Tasklet.__init__(self, lu)
7830 self.instance_name = instance_name
7832 self.iallocator_name = iallocator_name
7833 self.remote_node = remote_node
7835 self.delay_iallocator = delay_iallocator
7836 self.early_release = early_release
7839 self.instance = None
7840 self.new_node = None
7841 self.target_node = None
7842 self.other_node = None
7843 self.remote_node_info = None
7844 self.node_secondary_ip = None
7847 def CheckArguments(mode, remote_node, iallocator):
7848 """Helper function for users of this class.
7851 # check for valid parameter combination
7852 if mode == constants.REPLACE_DISK_CHG:
7853 if remote_node is None and iallocator is None:
7854 raise errors.OpPrereqError("When changing the secondary either an"
7855 " iallocator script must be used or the"
7856 " new node given", errors.ECODE_INVAL)
7858 if remote_node is not None and iallocator is not None:
7859 raise errors.OpPrereqError("Give either the iallocator or the new"
7860 " secondary, not both", errors.ECODE_INVAL)
7862 elif remote_node is not None or iallocator is not None:
7863 # Not replacing the secondary
7864 raise errors.OpPrereqError("The iallocator and new node options can"
7865 " only be used when changing the"
7866 " secondary node", errors.ECODE_INVAL)
7869 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7870 """Compute a new secondary node using an IAllocator.
7873 ial = IAllocator(lu.cfg, lu.rpc,
7874 mode=constants.IALLOCATOR_MODE_RELOC,
7876 relocate_from=relocate_from)
7878 ial.Run(iallocator_name)
7881 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7882 " %s" % (iallocator_name, ial.info),
7885 if len(ial.result) != ial.required_nodes:
7886 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7887 " of nodes (%s), required %s" %
7889 len(ial.result), ial.required_nodes),
7892 remote_node_name = ial.result[0]
7894 lu.LogInfo("Selected new secondary for instance '%s': %s",
7895 instance_name, remote_node_name)
7897 return remote_node_name
7899 def _FindFaultyDisks(self, node_name):
7900 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7903 def CheckPrereq(self):
7904 """Check prerequisites.
7906 This checks that the instance is in the cluster.
7909 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7910 assert instance is not None, \
7911 "Cannot retrieve locked instance %s" % self.instance_name
7913 if instance.disk_template != constants.DT_DRBD8:
7914 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7915 " instances", errors.ECODE_INVAL)
7917 if len(instance.secondary_nodes) != 1:
7918 raise errors.OpPrereqError("The instance has a strange layout,"
7919 " expected one secondary but found %d" %
7920 len(instance.secondary_nodes),
7923 if not self.delay_iallocator:
7924 self._CheckPrereq2()
7926 def _CheckPrereq2(self):
7927 """Check prerequisites, second part.
7929 This function should always be part of CheckPrereq. It was separated and is
7930 now called from Exec because during node evacuation iallocator was only
7931 called with an unmodified cluster model, not taking planned changes into
7935 instance = self.instance
7936 secondary_node = instance.secondary_nodes[0]
7938 if self.iallocator_name is None:
7939 remote_node = self.remote_node
7941 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7942 instance.name, instance.secondary_nodes)
7944 if remote_node is not None:
7945 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7946 assert self.remote_node_info is not None, \
7947 "Cannot retrieve locked node %s" % remote_node
7949 self.remote_node_info = None
7951 if remote_node == self.instance.primary_node:
7952 raise errors.OpPrereqError("The specified node is the primary node of"
7953 " the instance.", errors.ECODE_INVAL)
7955 if remote_node == secondary_node:
7956 raise errors.OpPrereqError("The specified node is already the"
7957 " secondary node of the instance.",
7960 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7961 constants.REPLACE_DISK_CHG):
7962 raise errors.OpPrereqError("Cannot specify disks to be replaced",
7965 if self.mode == constants.REPLACE_DISK_AUTO:
7966 faulty_primary = self._FindFaultyDisks(instance.primary_node)
7967 faulty_secondary = self._FindFaultyDisks(secondary_node)
7969 if faulty_primary and faulty_secondary:
7970 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7971 " one node and can not be repaired"
7972 " automatically" % self.instance_name,
7976 self.disks = faulty_primary
7977 self.target_node = instance.primary_node
7978 self.other_node = secondary_node
7979 check_nodes = [self.target_node, self.other_node]
7980 elif faulty_secondary:
7981 self.disks = faulty_secondary
7982 self.target_node = secondary_node
7983 self.other_node = instance.primary_node
7984 check_nodes = [self.target_node, self.other_node]
7990 # Non-automatic modes
7991 if self.mode == constants.REPLACE_DISK_PRI:
7992 self.target_node = instance.primary_node
7993 self.other_node = secondary_node
7994 check_nodes = [self.target_node, self.other_node]
7996 elif self.mode == constants.REPLACE_DISK_SEC:
7997 self.target_node = secondary_node
7998 self.other_node = instance.primary_node
7999 check_nodes = [self.target_node, self.other_node]
8001 elif self.mode == constants.REPLACE_DISK_CHG:
8002 self.new_node = remote_node
8003 self.other_node = instance.primary_node
8004 self.target_node = secondary_node
8005 check_nodes = [self.new_node, self.other_node]
8007 _CheckNodeNotDrained(self.lu, remote_node)
8008 _CheckNodeVmCapable(self.lu, remote_node)
8010 old_node_info = self.cfg.GetNodeInfo(secondary_node)
8011 assert old_node_info is not None
8012 if old_node_info.offline and not self.early_release:
8013 # doesn't make sense to delay the release
8014 self.early_release = True
8015 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
8016 " early-release mode", secondary_node)
8019 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
8022 # If not specified all disks should be replaced
8024 self.disks = range(len(self.instance.disks))
8026 for node in check_nodes:
8027 _CheckNodeOnline(self.lu, node)
8029 # Check whether disks are valid
8030 for disk_idx in self.disks:
8031 instance.FindDisk(disk_idx)
8033 # Get secondary node IP addresses
8036 for node_name in [self.target_node, self.other_node, self.new_node]:
8037 if node_name is not None:
8038 node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
8040 self.node_secondary_ip = node_2nd_ip
8042 def Exec(self, feedback_fn):
8043 """Execute disk replacement.
8045 This dispatches the disk replacement to the appropriate handler.
8048 if self.delay_iallocator:
8049 self._CheckPrereq2()
8052 feedback_fn("No disks need replacement")
8055 feedback_fn("Replacing disk(s) %s for %s" %
8056 (utils.CommaJoin(self.disks), self.instance.name))
8058 activate_disks = (not self.instance.admin_up)
8060 # Activate the instance disks if we're replacing them on a down instance
8062 _StartInstanceDisks(self.lu, self.instance, True)
8065 # Should we replace the secondary node?
8066 if self.new_node is not None:
8067 fn = self._ExecDrbd8Secondary
8069 fn = self._ExecDrbd8DiskOnly
8071 return fn(feedback_fn)
8074 # Deactivate the instance disks if we're replacing them on a
8077 _SafeShutdownInstanceDisks(self.lu, self.instance)
8079 def _CheckVolumeGroup(self, nodes):
8080 self.lu.LogInfo("Checking volume groups")
8082 vgname = self.cfg.GetVGName()
8084 # Make sure volume group exists on all involved nodes
8085 results = self.rpc.call_vg_list(nodes)
8087 raise errors.OpExecError("Can't list volume groups on the nodes")
8091 res.Raise("Error checking node %s" % node)
8092 if vgname not in res.payload:
8093 raise errors.OpExecError("Volume group '%s' not found on node %s" %
8096 def _CheckDisksExistence(self, nodes):
8097 # Check disk existence
8098 for idx, dev in enumerate(self.instance.disks):
8099 if idx not in self.disks:
8103 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
8104 self.cfg.SetDiskID(dev, node)
8106 result = self.rpc.call_blockdev_find(node, dev)
8108 msg = result.fail_msg
8109 if msg or not result.payload:
8111 msg = "disk not found"
8112 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
8115 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
8116 for idx, dev in enumerate(self.instance.disks):
8117 if idx not in self.disks:
8120 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
8123 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
8125 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
8126 " replace disks for instance %s" %
8127 (node_name, self.instance.name))
8129 def _CreateNewStorage(self, node_name):
8130 vgname = self.cfg.GetVGName()
8133 for idx, dev in enumerate(self.instance.disks):
8134 if idx not in self.disks:
8137 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
8139 self.cfg.SetDiskID(dev, node_name)
8141 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
8142 names = _GenerateUniqueNames(self.lu, lv_names)
8144 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
8145 logical_id=(vgname, names[0]))
8146 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
8147 logical_id=(vgname, names[1]))
8149 new_lvs = [lv_data, lv_meta]
8150 old_lvs = dev.children
8151 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
8153 # we pass force_create=True to force the LVM creation
8154 for new_lv in new_lvs:
8155 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
8156 _GetInstanceInfoText(self.instance), False)
8160 def _CheckDevices(self, node_name, iv_names):
8161 for name, (dev, _, _) in iv_names.iteritems():
8162 self.cfg.SetDiskID(dev, node_name)
8164 result = self.rpc.call_blockdev_find(node_name, dev)
8166 msg = result.fail_msg
8167 if msg or not result.payload:
8169 msg = "disk not found"
8170 raise errors.OpExecError("Can't find DRBD device %s: %s" %
8173 if result.payload.is_degraded:
8174 raise errors.OpExecError("DRBD device %s is degraded!" % name)
8176 def _RemoveOldStorage(self, node_name, iv_names):
8177 for name, (_, old_lvs, _) in iv_names.iteritems():
8178 self.lu.LogInfo("Remove logical volumes for %s" % name)
8181 self.cfg.SetDiskID(lv, node_name)
8183 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
8185 self.lu.LogWarning("Can't remove old LV: %s" % msg,
8186 hint="remove unused LVs manually")
8188 def _ReleaseNodeLock(self, node_name):
8189 """Releases the lock for a given node."""
8190 self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
8192 def _ExecDrbd8DiskOnly(self, feedback_fn):
8193 """Replace a disk on the primary or secondary for DRBD 8.
8195 The algorithm for replace is quite complicated:
8197 1. for each disk to be replaced:
8199 1. create new LVs on the target node with unique names
8200 1. detach old LVs from the drbd device
8201 1. rename old LVs to name_replaced.<time_t>
8202 1. rename new LVs to old LVs
8203 1. attach the new LVs (with the old names now) to the drbd device
8205 1. wait for sync across all devices
8207 1. for each modified disk:
8209 1. remove old LVs (which have the name name_replaces.<time_t>)
8211 Failures are not very well handled.
8216 # Step: check device activation
8217 self.lu.LogStep(1, steps_total, "Check device existence")
8218 self._CheckDisksExistence([self.other_node, self.target_node])
8219 self._CheckVolumeGroup([self.target_node, self.other_node])
8221 # Step: check other node consistency
8222 self.lu.LogStep(2, steps_total, "Check peer consistency")
8223 self._CheckDisksConsistency(self.other_node,
8224 self.other_node == self.instance.primary_node,
8227 # Step: create new storage
8228 self.lu.LogStep(3, steps_total, "Allocate new storage")
8229 iv_names = self._CreateNewStorage(self.target_node)
8231 # Step: for each lv, detach+rename*2+attach
8232 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8233 for dev, old_lvs, new_lvs in iv_names.itervalues():
8234 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8236 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8238 result.Raise("Can't detach drbd from local storage on node"
8239 " %s for device %s" % (self.target_node, dev.iv_name))
8241 #cfg.Update(instance)
8243 # ok, we created the new LVs, so now we know we have the needed
8244 # storage; as such, we proceed on the target node to rename
8245 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8246 # using the assumption that logical_id == physical_id (which in
8247 # turn is the unique_id on that node)
8249 # FIXME(iustin): use a better name for the replaced LVs
8250 temp_suffix = int(time.time())
8251 ren_fn = lambda d, suff: (d.physical_id[0],
8252 d.physical_id[1] + "_replaced-%s" % suff)
8254 # Build the rename list based on what LVs exist on the node
8255 rename_old_to_new = []
8256 for to_ren in old_lvs:
8257 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8258 if not result.fail_msg and result.payload:
8260 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8262 self.lu.LogInfo("Renaming the old LVs on the target node")
8263 result = self.rpc.call_blockdev_rename(self.target_node,
8265 result.Raise("Can't rename old LVs on node %s" % self.target_node)
8267 # Now we rename the new LVs to the old LVs
8268 self.lu.LogInfo("Renaming the new LVs on the target node")
8269 rename_new_to_old = [(new, old.physical_id)
8270 for old, new in zip(old_lvs, new_lvs)]
8271 result = self.rpc.call_blockdev_rename(self.target_node,
8273 result.Raise("Can't rename new LVs on node %s" % self.target_node)
8275 for old, new in zip(old_lvs, new_lvs):
8276 new.logical_id = old.logical_id
8277 self.cfg.SetDiskID(new, self.target_node)
8279 for disk in old_lvs:
8280 disk.logical_id = ren_fn(disk, temp_suffix)
8281 self.cfg.SetDiskID(disk, self.target_node)
8283 # Now that the new lvs have the old name, we can add them to the device
8284 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8285 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8287 msg = result.fail_msg
8289 for new_lv in new_lvs:
8290 msg2 = self.rpc.call_blockdev_remove(self.target_node,
8293 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8294 hint=("cleanup manually the unused logical"
8296 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8298 dev.children = new_lvs
8300 self.cfg.Update(self.instance, feedback_fn)
8303 if self.early_release:
8304 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8306 self._RemoveOldStorage(self.target_node, iv_names)
8307 # WARNING: we release both node locks here, do not do other RPCs
8308 # than WaitForSync to the primary node
8309 self._ReleaseNodeLock([self.target_node, self.other_node])
8312 # This can fail as the old devices are degraded and _WaitForSync
8313 # does a combined result over all disks, so we don't check its return value
8314 self.lu.LogStep(cstep, steps_total, "Sync devices")
8316 _WaitForSync(self.lu, self.instance)
8318 # Check all devices manually
8319 self._CheckDevices(self.instance.primary_node, iv_names)
8321 # Step: remove old storage
8322 if not self.early_release:
8323 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8325 self._RemoveOldStorage(self.target_node, iv_names)
8327 def _ExecDrbd8Secondary(self, feedback_fn):
8328 """Replace the secondary node for DRBD 8.
8330 The algorithm for replace is quite complicated:
8331 - for all disks of the instance:
8332 - create new LVs on the new node with same names
8333 - shutdown the drbd device on the old secondary
8334 - disconnect the drbd network on the primary
8335 - create the drbd device on the new secondary
8336 - network attach the drbd on the primary, using an artifice:
8337 the drbd code for Attach() will connect to the network if it
8338 finds a device which is connected to the good local disks but
8340 - wait for sync across all devices
8341 - remove all disks from the old secondary
8343 Failures are not very well handled.
8348 # Step: check device activation
8349 self.lu.LogStep(1, steps_total, "Check device existence")
8350 self._CheckDisksExistence([self.instance.primary_node])
8351 self._CheckVolumeGroup([self.instance.primary_node])
8353 # Step: check other node consistency
8354 self.lu.LogStep(2, steps_total, "Check peer consistency")
8355 self._CheckDisksConsistency(self.instance.primary_node, True, True)
8357 # Step: create new storage
8358 self.lu.LogStep(3, steps_total, "Allocate new storage")
8359 for idx, dev in enumerate(self.instance.disks):
8360 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8361 (self.new_node, idx))
8362 # we pass force_create=True to force LVM creation
8363 for new_lv in dev.children:
8364 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8365 _GetInstanceInfoText(self.instance), False)
8367 # Step 4: dbrd minors and drbd setups changes
8368 # after this, we must manually remove the drbd minors on both the
8369 # error and the success paths
8370 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8371 minors = self.cfg.AllocateDRBDMinor([self.new_node
8372 for dev in self.instance.disks],
8374 logging.debug("Allocated minors %r", minors)
8377 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8378 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8379 (self.new_node, idx))
8380 # create new devices on new_node; note that we create two IDs:
8381 # one without port, so the drbd will be activated without
8382 # networking information on the new node at this stage, and one
8383 # with network, for the latter activation in step 4
8384 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8385 if self.instance.primary_node == o_node1:
8388 assert self.instance.primary_node == o_node2, "Three-node instance?"
8391 new_alone_id = (self.instance.primary_node, self.new_node, None,
8392 p_minor, new_minor, o_secret)
8393 new_net_id = (self.instance.primary_node, self.new_node, o_port,
8394 p_minor, new_minor, o_secret)
8396 iv_names[idx] = (dev, dev.children, new_net_id)
8397 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8399 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8400 logical_id=new_alone_id,
8401 children=dev.children,
8404 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8405 _GetInstanceInfoText(self.instance), False)
8406 except errors.GenericError:
8407 self.cfg.ReleaseDRBDMinors(self.instance.name)
8410 # We have new devices, shutdown the drbd on the old secondary
8411 for idx, dev in enumerate(self.instance.disks):
8412 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8413 self.cfg.SetDiskID(dev, self.target_node)
8414 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8416 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8417 "node: %s" % (idx, msg),
8418 hint=("Please cleanup this device manually as"
8419 " soon as possible"))
8421 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8422 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8423 self.node_secondary_ip,
8424 self.instance.disks)\
8425 [self.instance.primary_node]
8427 msg = result.fail_msg
8429 # detaches didn't succeed (unlikely)
8430 self.cfg.ReleaseDRBDMinors(self.instance.name)
8431 raise errors.OpExecError("Can't detach the disks from the network on"
8432 " old node: %s" % (msg,))
8434 # if we managed to detach at least one, we update all the disks of
8435 # the instance to point to the new secondary
8436 self.lu.LogInfo("Updating instance configuration")
8437 for dev, _, new_logical_id in iv_names.itervalues():
8438 dev.logical_id = new_logical_id
8439 self.cfg.SetDiskID(dev, self.instance.primary_node)
8441 self.cfg.Update(self.instance, feedback_fn)
8443 # and now perform the drbd attach
8444 self.lu.LogInfo("Attaching primary drbds to new secondary"
8445 " (standalone => connected)")
8446 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8448 self.node_secondary_ip,
8449 self.instance.disks,
8452 for to_node, to_result in result.items():
8453 msg = to_result.fail_msg
8455 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8457 hint=("please do a gnt-instance info to see the"
8458 " status of disks"))
8460 if self.early_release:
8461 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8463 self._RemoveOldStorage(self.target_node, iv_names)
8464 # WARNING: we release all node locks here, do not do other RPCs
8465 # than WaitForSync to the primary node
8466 self._ReleaseNodeLock([self.instance.primary_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")
8483 self._RemoveOldStorage(self.target_node, iv_names)
8486 class LURepairNodeStorage(NoHooksLU):
8487 """Repairs the volume group on a node.
8492 def CheckArguments(self):
8493 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8495 storage_type = self.op.storage_type
8497 if (constants.SO_FIX_CONSISTENCY not in
8498 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8499 raise errors.OpPrereqError("Storage units of type '%s' can not be"
8500 " repaired" % storage_type,
8503 def ExpandNames(self):
8504 self.needed_locks = {
8505 locking.LEVEL_NODE: [self.op.node_name],
8508 def _CheckFaultyDisks(self, instance, node_name):
8509 """Ensure faulty disks abort the opcode or at least warn."""
8511 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8513 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8514 " node '%s'" % (instance.name, node_name),
8516 except errors.OpPrereqError, err:
8517 if self.op.ignore_consistency:
8518 self.proc.LogWarning(str(err.args[0]))
8522 def CheckPrereq(self):
8523 """Check prerequisites.
8526 # Check whether any instance on this node has faulty disks
8527 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8528 if not inst.admin_up:
8530 check_nodes = set(inst.all_nodes)
8531 check_nodes.discard(self.op.node_name)
8532 for inst_node_name in check_nodes:
8533 self._CheckFaultyDisks(inst, inst_node_name)
8535 def Exec(self, feedback_fn):
8536 feedback_fn("Repairing storage unit '%s' on %s ..." %
8537 (self.op.name, self.op.node_name))
8539 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8540 result = self.rpc.call_storage_execute(self.op.node_name,
8541 self.op.storage_type, st_args,
8543 constants.SO_FIX_CONSISTENCY)
8544 result.Raise("Failed to repair storage unit '%s' on %s" %
8545 (self.op.name, self.op.node_name))
8548 class LUNodeEvacStrategy(NoHooksLU):
8549 """Computes the node evacuation strategy.
8554 def CheckArguments(self):
8555 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8557 def ExpandNames(self):
8558 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8559 self.needed_locks = locks = {}
8560 if self.op.remote_node is None:
8561 locks[locking.LEVEL_NODE] = locking.ALL_SET
8563 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8564 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8566 def Exec(self, feedback_fn):
8567 if self.op.remote_node is not None:
8569 for node in self.op.nodes:
8570 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8573 if i.primary_node == self.op.remote_node:
8574 raise errors.OpPrereqError("Node %s is the primary node of"
8575 " instance %s, cannot use it as"
8577 (self.op.remote_node, i.name),
8579 result.append([i.name, self.op.remote_node])
8581 ial = IAllocator(self.cfg, self.rpc,
8582 mode=constants.IALLOCATOR_MODE_MEVAC,
8583 evac_nodes=self.op.nodes)
8584 ial.Run(self.op.iallocator, validate=True)
8586 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8592 class LUInstanceGrowDisk(LogicalUnit):
8593 """Grow a disk of an instance.
8597 HTYPE = constants.HTYPE_INSTANCE
8600 def ExpandNames(self):
8601 self._ExpandAndLockInstance()
8602 self.needed_locks[locking.LEVEL_NODE] = []
8603 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8605 def DeclareLocks(self, level):
8606 if level == locking.LEVEL_NODE:
8607 self._LockInstancesNodes()
8609 def BuildHooksEnv(self):
8612 This runs on the master, the primary and all the secondaries.
8616 "DISK": self.op.disk,
8617 "AMOUNT": self.op.amount,
8619 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8620 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8623 def CheckPrereq(self):
8624 """Check prerequisites.
8626 This checks that the instance is in the cluster.
8629 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8630 assert instance is not None, \
8631 "Cannot retrieve locked instance %s" % self.op.instance_name
8632 nodenames = list(instance.all_nodes)
8633 for node in nodenames:
8634 _CheckNodeOnline(self, node)
8636 self.instance = instance
8638 if instance.disk_template not in constants.DTS_GROWABLE:
8639 raise errors.OpPrereqError("Instance's disk layout does not support"
8640 " growing.", errors.ECODE_INVAL)
8642 self.disk = instance.FindDisk(self.op.disk)
8644 if instance.disk_template != constants.DT_FILE:
8645 # TODO: check the free disk space for file, when that feature
8647 _CheckNodesFreeDiskPerVG(self, nodenames,
8648 self.disk.ComputeGrowth(self.op.amount))
8650 def Exec(self, feedback_fn):
8651 """Execute disk grow.
8654 instance = self.instance
8657 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8659 raise errors.OpExecError("Cannot activate block device to grow")
8661 for node in instance.all_nodes:
8662 self.cfg.SetDiskID(disk, node)
8663 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8664 result.Raise("Grow request failed to node %s" % node)
8666 # TODO: Rewrite code to work properly
8667 # DRBD goes into sync mode for a short amount of time after executing the
8668 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8669 # calling "resize" in sync mode fails. Sleeping for a short amount of
8670 # time is a work-around.
8673 disk.RecordGrow(self.op.amount)
8674 self.cfg.Update(instance, feedback_fn)
8675 if self.op.wait_for_sync:
8676 disk_abort = not _WaitForSync(self, instance, disks=[disk])
8678 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8679 " status.\nPlease check the instance.")
8680 if not instance.admin_up:
8681 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8682 elif not instance.admin_up:
8683 self.proc.LogWarning("Not shutting down the disk even if the instance is"
8684 " not supposed to be running because no wait for"
8685 " sync mode was requested.")
8688 class LUInstanceQueryData(NoHooksLU):
8689 """Query runtime instance data.
8694 def ExpandNames(self):
8695 self.needed_locks = {}
8696 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8698 if self.op.instances:
8699 self.wanted_names = []
8700 for name in self.op.instances:
8701 full_name = _ExpandInstanceName(self.cfg, name)
8702 self.wanted_names.append(full_name)
8703 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8705 self.wanted_names = None
8706 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8708 self.needed_locks[locking.LEVEL_NODE] = []
8709 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8711 def DeclareLocks(self, level):
8712 if level == locking.LEVEL_NODE:
8713 self._LockInstancesNodes()
8715 def CheckPrereq(self):
8716 """Check prerequisites.
8718 This only checks the optional instance list against the existing names.
8721 if self.wanted_names is None:
8722 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8724 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8725 in self.wanted_names]
8727 def _ComputeBlockdevStatus(self, node, instance_name, dev):
8728 """Returns the status of a block device
8731 if self.op.static or not node:
8734 self.cfg.SetDiskID(dev, node)
8736 result = self.rpc.call_blockdev_find(node, dev)
8740 result.Raise("Can't compute disk status for %s" % instance_name)
8742 status = result.payload
8746 return (status.dev_path, status.major, status.minor,
8747 status.sync_percent, status.estimated_time,
8748 status.is_degraded, status.ldisk_status)
8750 def _ComputeDiskStatus(self, instance, snode, dev):
8751 """Compute block device status.
8754 if dev.dev_type in constants.LDS_DRBD:
8755 # we change the snode then (otherwise we use the one passed in)
8756 if dev.logical_id[0] == instance.primary_node:
8757 snode = dev.logical_id[1]
8759 snode = dev.logical_id[0]
8761 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8763 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8766 dev_children = [self._ComputeDiskStatus(instance, snode, child)
8767 for child in dev.children]
8772 "iv_name": dev.iv_name,
8773 "dev_type": dev.dev_type,
8774 "logical_id": dev.logical_id,
8775 "physical_id": dev.physical_id,
8776 "pstatus": dev_pstatus,
8777 "sstatus": dev_sstatus,
8778 "children": dev_children,
8785 def Exec(self, feedback_fn):
8786 """Gather and return data"""
8789 cluster = self.cfg.GetClusterInfo()
8791 for instance in self.wanted_instances:
8792 if not self.op.static:
8793 remote_info = self.rpc.call_instance_info(instance.primary_node,
8795 instance.hypervisor)
8796 remote_info.Raise("Error checking node %s" % instance.primary_node)
8797 remote_info = remote_info.payload
8798 if remote_info and "state" in remote_info:
8801 remote_state = "down"
8804 if instance.admin_up:
8807 config_state = "down"
8809 disks = [self._ComputeDiskStatus(instance, None, device)
8810 for device in instance.disks]
8813 "name": instance.name,
8814 "config_state": config_state,
8815 "run_state": remote_state,
8816 "pnode": instance.primary_node,
8817 "snodes": instance.secondary_nodes,
8819 # this happens to be the same format used for hooks
8820 "nics": _NICListToTuple(self, instance.nics),
8821 "disk_template": instance.disk_template,
8823 "hypervisor": instance.hypervisor,
8824 "network_port": instance.network_port,
8825 "hv_instance": instance.hvparams,
8826 "hv_actual": cluster.FillHV(instance, skip_globals=True),
8827 "be_instance": instance.beparams,
8828 "be_actual": cluster.FillBE(instance),
8829 "os_instance": instance.osparams,
8830 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8831 "serial_no": instance.serial_no,
8832 "mtime": instance.mtime,
8833 "ctime": instance.ctime,
8834 "uuid": instance.uuid,
8837 result[instance.name] = idict
8842 class LUInstanceSetParams(LogicalUnit):
8843 """Modifies an instances's parameters.
8846 HPATH = "instance-modify"
8847 HTYPE = constants.HTYPE_INSTANCE
8850 def CheckArguments(self):
8851 if not (self.op.nics or self.op.disks or self.op.disk_template or
8852 self.op.hvparams or self.op.beparams or self.op.os_name):
8853 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8855 if self.op.hvparams:
8856 _CheckGlobalHvParams(self.op.hvparams)
8860 for disk_op, disk_dict in self.op.disks:
8861 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8862 if disk_op == constants.DDM_REMOVE:
8865 elif disk_op == constants.DDM_ADD:
8868 if not isinstance(disk_op, int):
8869 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8870 if not isinstance(disk_dict, dict):
8871 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8872 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8874 if disk_op == constants.DDM_ADD:
8875 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8876 if mode not in constants.DISK_ACCESS_SET:
8877 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8879 size = disk_dict.get('size', None)
8881 raise errors.OpPrereqError("Required disk parameter size missing",
8885 except (TypeError, ValueError), err:
8886 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8887 str(err), errors.ECODE_INVAL)
8888 disk_dict['size'] = size
8890 # modification of disk
8891 if 'size' in disk_dict:
8892 raise errors.OpPrereqError("Disk size change not possible, use"
8893 " grow-disk", errors.ECODE_INVAL)
8895 if disk_addremove > 1:
8896 raise errors.OpPrereqError("Only one disk add or remove operation"
8897 " supported at a time", errors.ECODE_INVAL)
8899 if self.op.disks and self.op.disk_template is not None:
8900 raise errors.OpPrereqError("Disk template conversion and other disk"
8901 " changes not supported at the same time",
8904 if (self.op.disk_template and
8905 self.op.disk_template in constants.DTS_NET_MIRROR and
8906 self.op.remote_node is None):
8907 raise errors.OpPrereqError("Changing the disk template to a mirrored"
8908 " one requires specifying a secondary node",
8913 for nic_op, nic_dict in self.op.nics:
8914 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8915 if nic_op == constants.DDM_REMOVE:
8918 elif nic_op == constants.DDM_ADD:
8921 if not isinstance(nic_op, int):
8922 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8923 if not isinstance(nic_dict, dict):
8924 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8925 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8927 # nic_dict should be a dict
8928 nic_ip = nic_dict.get('ip', None)
8929 if nic_ip is not None:
8930 if nic_ip.lower() == constants.VALUE_NONE:
8931 nic_dict['ip'] = None
8933 if not netutils.IPAddress.IsValid(nic_ip):
8934 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8937 nic_bridge = nic_dict.get('bridge', None)
8938 nic_link = nic_dict.get('link', None)
8939 if nic_bridge and nic_link:
8940 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8941 " at the same time", errors.ECODE_INVAL)
8942 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8943 nic_dict['bridge'] = None
8944 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8945 nic_dict['link'] = None
8947 if nic_op == constants.DDM_ADD:
8948 nic_mac = nic_dict.get('mac', None)
8950 nic_dict['mac'] = constants.VALUE_AUTO
8952 if 'mac' in nic_dict:
8953 nic_mac = nic_dict['mac']
8954 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8955 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8957 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8958 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8959 " modifying an existing nic",
8962 if nic_addremove > 1:
8963 raise errors.OpPrereqError("Only one NIC add or remove operation"
8964 " supported at a time", errors.ECODE_INVAL)
8966 def ExpandNames(self):
8967 self._ExpandAndLockInstance()
8968 self.needed_locks[locking.LEVEL_NODE] = []
8969 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8971 def DeclareLocks(self, level):
8972 if level == locking.LEVEL_NODE:
8973 self._LockInstancesNodes()
8974 if self.op.disk_template and self.op.remote_node:
8975 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8976 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8978 def BuildHooksEnv(self):
8981 This runs on the master, primary and secondaries.
8985 if constants.BE_MEMORY in self.be_new:
8986 args['memory'] = self.be_new[constants.BE_MEMORY]
8987 if constants.BE_VCPUS in self.be_new:
8988 args['vcpus'] = self.be_new[constants.BE_VCPUS]
8989 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8990 # information at all.
8993 nic_override = dict(self.op.nics)
8994 for idx, nic in enumerate(self.instance.nics):
8995 if idx in nic_override:
8996 this_nic_override = nic_override[idx]
8998 this_nic_override = {}
8999 if 'ip' in this_nic_override:
9000 ip = this_nic_override['ip']
9003 if 'mac' in this_nic_override:
9004 mac = this_nic_override['mac']
9007 if idx in self.nic_pnew:
9008 nicparams = self.nic_pnew[idx]
9010 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
9011 mode = nicparams[constants.NIC_MODE]
9012 link = nicparams[constants.NIC_LINK]
9013 args['nics'].append((ip, mac, mode, link))
9014 if constants.DDM_ADD in nic_override:
9015 ip = nic_override[constants.DDM_ADD].get('ip', None)
9016 mac = nic_override[constants.DDM_ADD]['mac']
9017 nicparams = self.nic_pnew[constants.DDM_ADD]
9018 mode = nicparams[constants.NIC_MODE]
9019 link = nicparams[constants.NIC_LINK]
9020 args['nics'].append((ip, mac, mode, link))
9021 elif constants.DDM_REMOVE in nic_override:
9022 del args['nics'][-1]
9024 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
9025 if self.op.disk_template:
9026 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
9027 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9030 def CheckPrereq(self):
9031 """Check prerequisites.
9033 This only checks the instance list against the existing names.
9036 # checking the new params on the primary/secondary nodes
9038 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9039 cluster = self.cluster = self.cfg.GetClusterInfo()
9040 assert self.instance is not None, \
9041 "Cannot retrieve locked instance %s" % self.op.instance_name
9042 pnode = instance.primary_node
9043 nodelist = list(instance.all_nodes)
9046 if self.op.os_name and not self.op.force:
9047 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
9048 self.op.force_variant)
9049 instance_os = self.op.os_name
9051 instance_os = instance.os
9053 if self.op.disk_template:
9054 if instance.disk_template == self.op.disk_template:
9055 raise errors.OpPrereqError("Instance already has disk template %s" %
9056 instance.disk_template, errors.ECODE_INVAL)
9058 if (instance.disk_template,
9059 self.op.disk_template) not in self._DISK_CONVERSIONS:
9060 raise errors.OpPrereqError("Unsupported disk template conversion from"
9061 " %s to %s" % (instance.disk_template,
9062 self.op.disk_template),
9064 _CheckInstanceDown(self, instance, "cannot change disk template")
9065 if self.op.disk_template in constants.DTS_NET_MIRROR:
9066 if self.op.remote_node == pnode:
9067 raise errors.OpPrereqError("Given new secondary node %s is the same"
9068 " as the primary node of the instance" %
9069 self.op.remote_node, errors.ECODE_STATE)
9070 _CheckNodeOnline(self, self.op.remote_node)
9071 _CheckNodeNotDrained(self, self.op.remote_node)
9072 # FIXME: here we assume that the old instance type is DT_PLAIN
9073 assert instance.disk_template == constants.DT_PLAIN
9074 disks = [{"size": d.size, "vg": d.logical_id[0]}
9075 for d in instance.disks]
9076 required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
9077 _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
9079 # hvparams processing
9080 if self.op.hvparams:
9081 hv_type = instance.hypervisor
9082 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
9083 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
9084 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
9087 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
9088 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
9089 self.hv_new = hv_new # the new actual values
9090 self.hv_inst = i_hvdict # the new dict (without defaults)
9092 self.hv_new = self.hv_inst = {}
9094 # beparams processing
9095 if self.op.beparams:
9096 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
9098 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
9099 be_new = cluster.SimpleFillBE(i_bedict)
9100 self.be_new = be_new # the new actual values
9101 self.be_inst = i_bedict # the new dict (without defaults)
9103 self.be_new = self.be_inst = {}
9105 # osparams processing
9106 if self.op.osparams:
9107 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
9108 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
9109 self.os_inst = i_osdict # the new dict (without defaults)
9115 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
9116 mem_check_list = [pnode]
9117 if be_new[constants.BE_AUTO_BALANCE]:
9118 # either we changed auto_balance to yes or it was from before
9119 mem_check_list.extend(instance.secondary_nodes)
9120 instance_info = self.rpc.call_instance_info(pnode, instance.name,
9121 instance.hypervisor)
9122 nodeinfo = self.rpc.call_node_info(mem_check_list, None,
9123 instance.hypervisor)
9124 pninfo = nodeinfo[pnode]
9125 msg = pninfo.fail_msg
9127 # Assume the primary node is unreachable and go ahead
9128 self.warn.append("Can't get info from primary node %s: %s" %
9130 elif not isinstance(pninfo.payload.get('memory_free', None), int):
9131 self.warn.append("Node data from primary node %s doesn't contain"
9132 " free memory information" % pnode)
9133 elif instance_info.fail_msg:
9134 self.warn.append("Can't get instance runtime information: %s" %
9135 instance_info.fail_msg)
9137 if instance_info.payload:
9138 current_mem = int(instance_info.payload['memory'])
9140 # Assume instance not running
9141 # (there is a slight race condition here, but it's not very probable,
9142 # and we have no other way to check)
9144 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
9145 pninfo.payload['memory_free'])
9147 raise errors.OpPrereqError("This change will prevent the instance"
9148 " from starting, due to %d MB of memory"
9149 " missing on its primary node" % miss_mem,
9152 if be_new[constants.BE_AUTO_BALANCE]:
9153 for node, nres in nodeinfo.items():
9154 if node not in instance.secondary_nodes:
9158 self.warn.append("Can't get info from secondary node %s: %s" %
9160 elif not isinstance(nres.payload.get('memory_free', None), int):
9161 self.warn.append("Secondary node %s didn't return free"
9162 " memory information" % node)
9163 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
9164 self.warn.append("Not enough memory to failover instance to"
9165 " secondary node %s" % node)
9170 for nic_op, nic_dict in self.op.nics:
9171 if nic_op == constants.DDM_REMOVE:
9172 if not instance.nics:
9173 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
9176 if nic_op != constants.DDM_ADD:
9178 if not instance.nics:
9179 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
9180 " no NICs" % nic_op,
9182 if nic_op < 0 or nic_op >= len(instance.nics):
9183 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
9185 (nic_op, len(instance.nics) - 1),
9187 old_nic_params = instance.nics[nic_op].nicparams
9188 old_nic_ip = instance.nics[nic_op].ip
9193 update_params_dict = dict([(key, nic_dict[key])
9194 for key in constants.NICS_PARAMETERS
9195 if key in nic_dict])
9197 if 'bridge' in nic_dict:
9198 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
9200 new_nic_params = _GetUpdatedParams(old_nic_params,
9202 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9203 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9204 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9205 self.nic_pinst[nic_op] = new_nic_params
9206 self.nic_pnew[nic_op] = new_filled_nic_params
9207 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9209 if new_nic_mode == constants.NIC_MODE_BRIDGED:
9210 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9211 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9213 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9215 self.warn.append(msg)
9217 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9218 if new_nic_mode == constants.NIC_MODE_ROUTED:
9219 if 'ip' in nic_dict:
9220 nic_ip = nic_dict['ip']
9224 raise errors.OpPrereqError('Cannot set the nic ip to None'
9225 ' on a routed nic', errors.ECODE_INVAL)
9226 if 'mac' in nic_dict:
9227 nic_mac = nic_dict['mac']
9229 raise errors.OpPrereqError('Cannot set the nic mac to None',
9231 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9232 # otherwise generate the mac
9233 nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9235 # or validate/reserve the current one
9237 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9238 except errors.ReservationError:
9239 raise errors.OpPrereqError("MAC address %s already in use"
9240 " in cluster" % nic_mac,
9241 errors.ECODE_NOTUNIQUE)
9244 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9245 raise errors.OpPrereqError("Disk operations not supported for"
9246 " diskless instances",
9248 for disk_op, _ in self.op.disks:
9249 if disk_op == constants.DDM_REMOVE:
9250 if len(instance.disks) == 1:
9251 raise errors.OpPrereqError("Cannot remove the last disk of"
9252 " an instance", errors.ECODE_INVAL)
9253 _CheckInstanceDown(self, instance, "cannot remove disks")
9255 if (disk_op == constants.DDM_ADD and
9256 len(instance.disks) >= constants.MAX_DISKS):
9257 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9258 " add more" % constants.MAX_DISKS,
9260 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9262 if disk_op < 0 or disk_op >= len(instance.disks):
9263 raise errors.OpPrereqError("Invalid disk index %s, valid values"
9265 (disk_op, len(instance.disks)),
9270 def _ConvertPlainToDrbd(self, feedback_fn):
9271 """Converts an instance from plain to drbd.
9274 feedback_fn("Converting template to drbd")
9275 instance = self.instance
9276 pnode = instance.primary_node
9277 snode = self.op.remote_node
9279 # create a fake disk info for _GenerateDiskTemplate
9280 disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9281 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9282 instance.name, pnode, [snode],
9283 disk_info, None, None, 0, feedback_fn)
9284 info = _GetInstanceInfoText(instance)
9285 feedback_fn("Creating aditional volumes...")
9286 # first, create the missing data and meta devices
9287 for disk in new_disks:
9288 # unfortunately this is... not too nice
9289 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9291 for child in disk.children:
9292 _CreateSingleBlockDev(self, snode, instance, child, info, True)
9293 # at this stage, all new LVs have been created, we can rename the
9295 feedback_fn("Renaming original volumes...")
9296 rename_list = [(o, n.children[0].logical_id)
9297 for (o, n) in zip(instance.disks, new_disks)]
9298 result = self.rpc.call_blockdev_rename(pnode, rename_list)
9299 result.Raise("Failed to rename original LVs")
9301 feedback_fn("Initializing DRBD devices...")
9302 # all child devices are in place, we can now create the DRBD devices
9303 for disk in new_disks:
9304 for node in [pnode, snode]:
9305 f_create = node == pnode
9306 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9308 # at this point, the instance has been modified
9309 instance.disk_template = constants.DT_DRBD8
9310 instance.disks = new_disks
9311 self.cfg.Update(instance, feedback_fn)
9313 # disks are created, waiting for sync
9314 disk_abort = not _WaitForSync(self, instance)
9316 raise errors.OpExecError("There are some degraded disks for"
9317 " this instance, please cleanup manually")
9319 def _ConvertDrbdToPlain(self, feedback_fn):
9320 """Converts an instance from drbd to plain.
9323 instance = self.instance
9324 assert len(instance.secondary_nodes) == 1
9325 pnode = instance.primary_node
9326 snode = instance.secondary_nodes[0]
9327 feedback_fn("Converting template to plain")
9329 old_disks = instance.disks
9330 new_disks = [d.children[0] for d in old_disks]
9332 # copy over size and mode
9333 for parent, child in zip(old_disks, new_disks):
9334 child.size = parent.size
9335 child.mode = parent.mode
9337 # update instance structure
9338 instance.disks = new_disks
9339 instance.disk_template = constants.DT_PLAIN
9340 self.cfg.Update(instance, feedback_fn)
9342 feedback_fn("Removing volumes on the secondary node...")
9343 for disk in old_disks:
9344 self.cfg.SetDiskID(disk, snode)
9345 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9347 self.LogWarning("Could not remove block device %s on node %s,"
9348 " continuing anyway: %s", disk.iv_name, snode, msg)
9350 feedback_fn("Removing unneeded volumes on the primary node...")
9351 for idx, disk in enumerate(old_disks):
9352 meta = disk.children[1]
9353 self.cfg.SetDiskID(meta, pnode)
9354 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9356 self.LogWarning("Could not remove metadata for disk %d on node %s,"
9357 " continuing anyway: %s", idx, pnode, msg)
9359 def Exec(self, feedback_fn):
9360 """Modifies an instance.
9362 All parameters take effect only at the next restart of the instance.
9365 # Process here the warnings from CheckPrereq, as we don't have a
9366 # feedback_fn there.
9367 for warn in self.warn:
9368 feedback_fn("WARNING: %s" % warn)
9371 instance = self.instance
9373 for disk_op, disk_dict in self.op.disks:
9374 if disk_op == constants.DDM_REMOVE:
9375 # remove the last disk
9376 device = instance.disks.pop()
9377 device_idx = len(instance.disks)
9378 for node, disk in device.ComputeNodeTree(instance.primary_node):
9379 self.cfg.SetDiskID(disk, node)
9380 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9382 self.LogWarning("Could not remove disk/%d on node %s: %s,"
9383 " continuing anyway", device_idx, node, msg)
9384 result.append(("disk/%d" % device_idx, "remove"))
9385 elif disk_op == constants.DDM_ADD:
9387 if instance.disk_template == constants.DT_FILE:
9388 file_driver, file_path = instance.disks[0].logical_id
9389 file_path = os.path.dirname(file_path)
9391 file_driver = file_path = None
9392 disk_idx_base = len(instance.disks)
9393 new_disk = _GenerateDiskTemplate(self,
9394 instance.disk_template,
9395 instance.name, instance.primary_node,
9396 instance.secondary_nodes,
9400 disk_idx_base, feedback_fn)[0]
9401 instance.disks.append(new_disk)
9402 info = _GetInstanceInfoText(instance)
9404 logging.info("Creating volume %s for instance %s",
9405 new_disk.iv_name, instance.name)
9406 # Note: this needs to be kept in sync with _CreateDisks
9408 for node in instance.all_nodes:
9409 f_create = node == instance.primary_node
9411 _CreateBlockDev(self, node, instance, new_disk,
9412 f_create, info, f_create)
9413 except errors.OpExecError, err:
9414 self.LogWarning("Failed to create volume %s (%s) on"
9416 new_disk.iv_name, new_disk, node, err)
9417 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9418 (new_disk.size, new_disk.mode)))
9420 # change a given disk
9421 instance.disks[disk_op].mode = disk_dict['mode']
9422 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9424 if self.op.disk_template:
9425 r_shut = _ShutdownInstanceDisks(self, instance)
9427 raise errors.OpExecError("Cannot shutdown instance disks, unable to"
9428 " proceed with disk template conversion")
9429 mode = (instance.disk_template, self.op.disk_template)
9431 self._DISK_CONVERSIONS[mode](self, feedback_fn)
9433 self.cfg.ReleaseDRBDMinors(instance.name)
9435 result.append(("disk_template", self.op.disk_template))
9438 for nic_op, nic_dict in self.op.nics:
9439 if nic_op == constants.DDM_REMOVE:
9440 # remove the last nic
9441 del instance.nics[-1]
9442 result.append(("nic.%d" % len(instance.nics), "remove"))
9443 elif nic_op == constants.DDM_ADD:
9444 # mac and bridge should be set, by now
9445 mac = nic_dict['mac']
9446 ip = nic_dict.get('ip', None)
9447 nicparams = self.nic_pinst[constants.DDM_ADD]
9448 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9449 instance.nics.append(new_nic)
9450 result.append(("nic.%d" % (len(instance.nics) - 1),
9451 "add:mac=%s,ip=%s,mode=%s,link=%s" %
9452 (new_nic.mac, new_nic.ip,
9453 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9454 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9457 for key in 'mac', 'ip':
9459 setattr(instance.nics[nic_op], key, nic_dict[key])
9460 if nic_op in self.nic_pinst:
9461 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9462 for key, val in nic_dict.iteritems():
9463 result.append(("nic.%s/%d" % (key, nic_op), val))
9466 if self.op.hvparams:
9467 instance.hvparams = self.hv_inst
9468 for key, val in self.op.hvparams.iteritems():
9469 result.append(("hv/%s" % key, val))
9472 if self.op.beparams:
9473 instance.beparams = self.be_inst
9474 for key, val in self.op.beparams.iteritems():
9475 result.append(("be/%s" % key, val))
9479 instance.os = self.op.os_name
9482 if self.op.osparams:
9483 instance.osparams = self.os_inst
9484 for key, val in self.op.osparams.iteritems():
9485 result.append(("os/%s" % key, val))
9487 self.cfg.Update(instance, feedback_fn)
9491 _DISK_CONVERSIONS = {
9492 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9493 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9497 class LUBackupQuery(NoHooksLU):
9498 """Query the exports list
9503 def ExpandNames(self):
9504 self.needed_locks = {}
9505 self.share_locks[locking.LEVEL_NODE] = 1
9506 if not self.op.nodes:
9507 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9509 self.needed_locks[locking.LEVEL_NODE] = \
9510 _GetWantedNodes(self, self.op.nodes)
9512 def Exec(self, feedback_fn):
9513 """Compute the list of all the exported system images.
9516 @return: a dictionary with the structure node->(export-list)
9517 where export-list is a list of the instances exported on
9521 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9522 rpcresult = self.rpc.call_export_list(self.nodes)
9524 for node in rpcresult:
9525 if rpcresult[node].fail_msg:
9526 result[node] = False
9528 result[node] = rpcresult[node].payload
9533 class LUBackupPrepare(NoHooksLU):
9534 """Prepares an instance for an export and returns useful information.
9539 def ExpandNames(self):
9540 self._ExpandAndLockInstance()
9542 def CheckPrereq(self):
9543 """Check prerequisites.
9546 instance_name = self.op.instance_name
9548 self.instance = self.cfg.GetInstanceInfo(instance_name)
9549 assert self.instance is not None, \
9550 "Cannot retrieve locked instance %s" % self.op.instance_name
9551 _CheckNodeOnline(self, self.instance.primary_node)
9553 self._cds = _GetClusterDomainSecret()
9555 def Exec(self, feedback_fn):
9556 """Prepares an instance for an export.
9559 instance = self.instance
9561 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9562 salt = utils.GenerateSecret(8)
9564 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9565 result = self.rpc.call_x509_cert_create(instance.primary_node,
9566 constants.RIE_CERT_VALIDITY)
9567 result.Raise("Can't create X509 key and certificate on %s" % result.node)
9569 (name, cert_pem) = result.payload
9571 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9575 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9576 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9578 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9584 class LUBackupExport(LogicalUnit):
9585 """Export an instance to an image in the cluster.
9588 HPATH = "instance-export"
9589 HTYPE = constants.HTYPE_INSTANCE
9592 def CheckArguments(self):
9593 """Check the arguments.
9596 self.x509_key_name = self.op.x509_key_name
9597 self.dest_x509_ca_pem = self.op.destination_x509_ca
9599 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9600 if not self.x509_key_name:
9601 raise errors.OpPrereqError("Missing X509 key name for encryption",
9604 if not self.dest_x509_ca_pem:
9605 raise errors.OpPrereqError("Missing destination X509 CA",
9608 def ExpandNames(self):
9609 self._ExpandAndLockInstance()
9611 # Lock all nodes for local exports
9612 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9613 # FIXME: lock only instance primary and destination node
9615 # Sad but true, for now we have do lock all nodes, as we don't know where
9616 # the previous export might be, and in this LU we search for it and
9617 # remove it from its current node. In the future we could fix this by:
9618 # - making a tasklet to search (share-lock all), then create the
9619 # new one, then one to remove, after
9620 # - removing the removal operation altogether
9621 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9623 def DeclareLocks(self, level):
9624 """Last minute lock declaration."""
9625 # All nodes are locked anyway, so nothing to do here.
9627 def BuildHooksEnv(self):
9630 This will run on the master, primary node and target node.
9634 "EXPORT_MODE": self.op.mode,
9635 "EXPORT_NODE": self.op.target_node,
9636 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9637 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9638 # TODO: Generic function for boolean env variables
9639 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9642 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9644 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9646 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9647 nl.append(self.op.target_node)
9651 def CheckPrereq(self):
9652 """Check prerequisites.
9654 This checks that the instance and node names are valid.
9657 instance_name = self.op.instance_name
9659 self.instance = self.cfg.GetInstanceInfo(instance_name)
9660 assert self.instance is not None, \
9661 "Cannot retrieve locked instance %s" % self.op.instance_name
9662 _CheckNodeOnline(self, self.instance.primary_node)
9664 if (self.op.remove_instance and self.instance.admin_up and
9665 not self.op.shutdown):
9666 raise errors.OpPrereqError("Can not remove instance without shutting it"
9669 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9670 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9671 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9672 assert self.dst_node is not None
9674 _CheckNodeOnline(self, self.dst_node.name)
9675 _CheckNodeNotDrained(self, self.dst_node.name)
9678 self.dest_disk_info = None
9679 self.dest_x509_ca = None
9681 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9682 self.dst_node = None
9684 if len(self.op.target_node) != len(self.instance.disks):
9685 raise errors.OpPrereqError(("Received destination information for %s"
9686 " disks, but instance %s has %s disks") %
9687 (len(self.op.target_node), instance_name,
9688 len(self.instance.disks)),
9691 cds = _GetClusterDomainSecret()
9693 # Check X509 key name
9695 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9696 except (TypeError, ValueError), err:
9697 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9699 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9700 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9703 # Load and verify CA
9705 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9706 except OpenSSL.crypto.Error, err:
9707 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9708 (err, ), errors.ECODE_INVAL)
9710 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9711 if errcode is not None:
9712 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9713 (msg, ), errors.ECODE_INVAL)
9715 self.dest_x509_ca = cert
9717 # Verify target information
9719 for idx, disk_data in enumerate(self.op.target_node):
9721 (host, port, magic) = \
9722 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9723 except errors.GenericError, err:
9724 raise errors.OpPrereqError("Target info for disk %s: %s" %
9725 (idx, err), errors.ECODE_INVAL)
9727 disk_info.append((host, port, magic))
9729 assert len(disk_info) == len(self.op.target_node)
9730 self.dest_disk_info = disk_info
9733 raise errors.ProgrammerError("Unhandled export mode %r" %
9736 # instance disk type verification
9737 # TODO: Implement export support for file-based disks
9738 for disk in self.instance.disks:
9739 if disk.dev_type == constants.LD_FILE:
9740 raise errors.OpPrereqError("Export not supported for instances with"
9741 " file-based disks", errors.ECODE_INVAL)
9743 def _CleanupExports(self, feedback_fn):
9744 """Removes exports of current instance from all other nodes.
9746 If an instance in a cluster with nodes A..D was exported to node C, its
9747 exports will be removed from the nodes A, B and D.
9750 assert self.op.mode != constants.EXPORT_MODE_REMOTE
9752 nodelist = self.cfg.GetNodeList()
9753 nodelist.remove(self.dst_node.name)
9755 # on one-node clusters nodelist will be empty after the removal
9756 # if we proceed the backup would be removed because OpBackupQuery
9757 # substitutes an empty list with the full cluster node list.
9758 iname = self.instance.name
9760 feedback_fn("Removing old exports for instance %s" % iname)
9761 exportlist = self.rpc.call_export_list(nodelist)
9762 for node in exportlist:
9763 if exportlist[node].fail_msg:
9765 if iname in exportlist[node].payload:
9766 msg = self.rpc.call_export_remove(node, iname).fail_msg
9768 self.LogWarning("Could not remove older export for instance %s"
9769 " on node %s: %s", iname, node, msg)
9771 def Exec(self, feedback_fn):
9772 """Export an instance to an image in the cluster.
9775 assert self.op.mode in constants.EXPORT_MODES
9777 instance = self.instance
9778 src_node = instance.primary_node
9780 if self.op.shutdown:
9781 # shutdown the instance, but not the disks
9782 feedback_fn("Shutting down instance %s" % instance.name)
9783 result = self.rpc.call_instance_shutdown(src_node, instance,
9784 self.op.shutdown_timeout)
9785 # TODO: Maybe ignore failures if ignore_remove_failures is set
9786 result.Raise("Could not shutdown instance %s on"
9787 " node %s" % (instance.name, src_node))
9789 # set the disks ID correctly since call_instance_start needs the
9790 # correct drbd minor to create the symlinks
9791 for disk in instance.disks:
9792 self.cfg.SetDiskID(disk, src_node)
9794 activate_disks = (not instance.admin_up)
9797 # Activate the instance disks if we'exporting a stopped instance
9798 feedback_fn("Activating disks for %s" % instance.name)
9799 _StartInstanceDisks(self, instance, None)
9802 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9805 helper.CreateSnapshots()
9807 if (self.op.shutdown and instance.admin_up and
9808 not self.op.remove_instance):
9809 assert not activate_disks
9810 feedback_fn("Starting instance %s" % instance.name)
9811 result = self.rpc.call_instance_start(src_node, instance, None, None)
9812 msg = result.fail_msg
9814 feedback_fn("Failed to start instance: %s" % msg)
9815 _ShutdownInstanceDisks(self, instance)
9816 raise errors.OpExecError("Could not start instance: %s" % msg)
9818 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9819 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9820 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9821 connect_timeout = constants.RIE_CONNECT_TIMEOUT
9822 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9824 (key_name, _, _) = self.x509_key_name
9827 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9830 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9831 key_name, dest_ca_pem,
9836 # Check for backwards compatibility
9837 assert len(dresults) == len(instance.disks)
9838 assert compat.all(isinstance(i, bool) for i in dresults), \
9839 "Not all results are boolean: %r" % dresults
9843 feedback_fn("Deactivating disks for %s" % instance.name)
9844 _ShutdownInstanceDisks(self, instance)
9846 if not (compat.all(dresults) and fin_resu):
9849 failures.append("export finalization")
9850 if not compat.all(dresults):
9851 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9853 failures.append("disk export: disk(s) %s" % fdsk)
9855 raise errors.OpExecError("Export failed, errors in %s" %
9856 utils.CommaJoin(failures))
9858 # At this point, the export was successful, we can cleanup/finish
9860 # Remove instance if requested
9861 if self.op.remove_instance:
9862 feedback_fn("Removing instance %s" % instance.name)
9863 _RemoveInstance(self, feedback_fn, instance,
9864 self.op.ignore_remove_failures)
9866 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9867 self._CleanupExports(feedback_fn)
9869 return fin_resu, dresults
9872 class LUBackupRemove(NoHooksLU):
9873 """Remove exports related to the named instance.
9878 def ExpandNames(self):
9879 self.needed_locks = {}
9880 # We need all nodes to be locked in order for RemoveExport to work, but we
9881 # don't need to lock the instance itself, as nothing will happen to it (and
9882 # we can remove exports also for a removed instance)
9883 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9885 def Exec(self, feedback_fn):
9886 """Remove any export.
9889 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9890 # If the instance was not found we'll try with the name that was passed in.
9891 # This will only work if it was an FQDN, though.
9893 if not instance_name:
9895 instance_name = self.op.instance_name
9897 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9898 exportlist = self.rpc.call_export_list(locked_nodes)
9900 for node in exportlist:
9901 msg = exportlist[node].fail_msg
9903 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9905 if instance_name in exportlist[node].payload:
9907 result = self.rpc.call_export_remove(node, instance_name)
9908 msg = result.fail_msg
9910 logging.error("Could not remove export for instance %s"
9911 " on node %s: %s", instance_name, node, msg)
9913 if fqdn_warn and not found:
9914 feedback_fn("Export not found. If trying to remove an export belonging"
9915 " to a deleted instance please use its Fully Qualified"
9919 class LUGroupAdd(LogicalUnit):
9920 """Logical unit for creating node groups.
9924 HTYPE = constants.HTYPE_GROUP
9927 def ExpandNames(self):
9928 # We need the new group's UUID here so that we can create and acquire the
9929 # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
9930 # that it should not check whether the UUID exists in the configuration.
9931 self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
9932 self.needed_locks = {}
9933 self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
9935 def CheckPrereq(self):
9936 """Check prerequisites.
9938 This checks that the given group name is not an existing node group
9943 existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
9944 except errors.OpPrereqError:
9947 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
9948 " node group (UUID: %s)" %
9949 (self.op.group_name, existing_uuid),
9950 errors.ECODE_EXISTS)
9952 if self.op.ndparams:
9953 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
9955 def BuildHooksEnv(self):
9960 "GROUP_NAME": self.op.group_name,
9962 mn = self.cfg.GetMasterNode()
9963 return env, [mn], [mn]
9965 def Exec(self, feedback_fn):
9966 """Add the node group to the cluster.
9969 group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
9970 uuid=self.group_uuid,
9971 alloc_policy=self.op.alloc_policy,
9972 ndparams=self.op.ndparams)
9974 self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
9975 del self.remove_locks[locking.LEVEL_NODEGROUP]
9978 class LUGroupAssignNodes(NoHooksLU):
9979 """Logical unit for assigning nodes to groups.
9984 def ExpandNames(self):
9985 # These raise errors.OpPrereqError on their own:
9986 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
9987 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
9989 # We want to lock all the affected nodes and groups. We have readily
9990 # available the list of nodes, and the *destination* group. To gather the
9991 # list of "source" groups, we need to fetch node information.
9992 self.node_data = self.cfg.GetAllNodesInfo()
9993 affected_groups = set(self.node_data[node].group for node in self.op.nodes)
9994 affected_groups.add(self.group_uuid)
9996 self.needed_locks = {
9997 locking.LEVEL_NODEGROUP: list(affected_groups),
9998 locking.LEVEL_NODE: self.op.nodes,
10001 def CheckPrereq(self):
10002 """Check prerequisites.
10005 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10006 instance_data = self.cfg.GetAllInstancesInfo()
10008 if self.group is None:
10009 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10010 (self.op.group_name, self.group_uuid))
10012 (new_splits, previous_splits) = \
10013 self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
10014 for node in self.op.nodes],
10015 self.node_data, instance_data)
10018 fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
10020 if not self.op.force:
10021 raise errors.OpExecError("The following instances get split by this"
10022 " change and --force was not given: %s" %
10025 self.LogWarning("This operation will split the following instances: %s",
10028 if previous_splits:
10029 self.LogWarning("In addition, these already-split instances continue"
10030 " to be spit across groups: %s",
10031 utils.CommaJoin(utils.NiceSort(previous_splits)))
10033 def Exec(self, feedback_fn):
10034 """Assign nodes to a new group.
10037 for node in self.op.nodes:
10038 self.node_data[node].group = self.group_uuid
10040 self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
10043 def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
10044 """Check for split instances after a node assignment.
10046 This method considers a series of node assignments as an atomic operation,
10047 and returns information about split instances after applying the set of
10050 In particular, it returns information about newly split instances, and
10051 instances that were already split, and remain so after the change.
10053 Only instances whose disk template is listed in constants.DTS_NET_MIRROR are
10056 @type changes: list of (node_name, new_group_uuid) pairs.
10057 @param changes: list of node assignments to consider.
10058 @param node_data: a dict with data for all nodes
10059 @param instance_data: a dict with all instances to consider
10060 @rtype: a two-tuple
10061 @return: a list of instances that were previously okay and result split as a
10062 consequence of this change, and a list of instances that were previously
10063 split and this change does not fix.
10066 changed_nodes = dict((node, group) for node, group in changes
10067 if node_data[node].group != group)
10069 all_split_instances = set()
10070 previously_split_instances = set()
10072 def InstanceNodes(instance):
10073 return [instance.primary_node] + list(instance.secondary_nodes)
10075 for inst in instance_data.values():
10076 if inst.disk_template not in constants.DTS_NET_MIRROR:
10079 instance_nodes = InstanceNodes(inst)
10081 if len(set(node_data[node].group for node in instance_nodes)) > 1:
10082 previously_split_instances.add(inst.name)
10084 if len(set(changed_nodes.get(node, node_data[node].group)
10085 for node in instance_nodes)) > 1:
10086 all_split_instances.add(inst.name)
10088 return (list(all_split_instances - previously_split_instances),
10089 list(previously_split_instances & all_split_instances))
10092 class _GroupQuery(_QueryBase):
10094 FIELDS = query.GROUP_FIELDS
10096 def ExpandNames(self, lu):
10097 lu.needed_locks = {}
10099 self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
10100 name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
10103 self.wanted = [name_to_uuid[name]
10104 for name in utils.NiceSort(name_to_uuid.keys())]
10106 # Accept names to be either names or UUIDs.
10109 all_uuid = frozenset(self._all_groups.keys())
10111 for name in self.names:
10112 if name in all_uuid:
10113 self.wanted.append(name)
10114 elif name in name_to_uuid:
10115 self.wanted.append(name_to_uuid[name])
10117 missing.append(name)
10120 raise errors.OpPrereqError("Some groups do not exist: %s" % missing,
10121 errors.ECODE_NOENT)
10123 def DeclareLocks(self, lu, level):
10126 def _GetQueryData(self, lu):
10127 """Computes the list of node groups and their attributes.
10130 do_nodes = query.GQ_NODE in self.requested_data
10131 do_instances = query.GQ_INST in self.requested_data
10133 group_to_nodes = None
10134 group_to_instances = None
10136 # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
10137 # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
10138 # latter GetAllInstancesInfo() is not enough, for we have to go through
10139 # instance->node. Hence, we will need to process nodes even if we only need
10140 # instance information.
10141 if do_nodes or do_instances:
10142 all_nodes = lu.cfg.GetAllNodesInfo()
10143 group_to_nodes = dict((uuid, []) for uuid in self.wanted)
10146 for node in all_nodes.values():
10147 if node.group in group_to_nodes:
10148 group_to_nodes[node.group].append(node.name)
10149 node_to_group[node.name] = node.group
10152 all_instances = lu.cfg.GetAllInstancesInfo()
10153 group_to_instances = dict((uuid, []) for uuid in self.wanted)
10155 for instance in all_instances.values():
10156 node = instance.primary_node
10157 if node in node_to_group:
10158 group_to_instances[node_to_group[node]].append(instance.name)
10161 # Do not pass on node information if it was not requested.
10162 group_to_nodes = None
10164 return query.GroupQueryData([self._all_groups[uuid]
10165 for uuid in self.wanted],
10166 group_to_nodes, group_to_instances)
10169 class LUGroupQuery(NoHooksLU):
10170 """Logical unit for querying node groups.
10175 def CheckArguments(self):
10176 self.gq = _GroupQuery(self.op.names, self.op.output_fields, False)
10178 def ExpandNames(self):
10179 self.gq.ExpandNames(self)
10181 def Exec(self, feedback_fn):
10182 return self.gq.OldStyleQuery(self)
10185 class LUGroupSetParams(LogicalUnit):
10186 """Modifies the parameters of a node group.
10189 HPATH = "group-modify"
10190 HTYPE = constants.HTYPE_GROUP
10193 def CheckArguments(self):
10196 self.op.alloc_policy,
10199 if all_changes.count(None) == len(all_changes):
10200 raise errors.OpPrereqError("Please pass at least one modification",
10201 errors.ECODE_INVAL)
10203 def ExpandNames(self):
10204 # This raises errors.OpPrereqError on its own:
10205 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10207 self.needed_locks = {
10208 locking.LEVEL_NODEGROUP: [self.group_uuid],
10211 def CheckPrereq(self):
10212 """Check prerequisites.
10215 self.group = self.cfg.GetNodeGroup(self.group_uuid)
10217 if self.group is None:
10218 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10219 (self.op.group_name, self.group_uuid))
10221 if self.op.ndparams:
10222 new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
10223 utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10224 self.new_ndparams = new_ndparams
10226 def BuildHooksEnv(self):
10227 """Build hooks env.
10231 "GROUP_NAME": self.op.group_name,
10232 "NEW_ALLOC_POLICY": self.op.alloc_policy,
10234 mn = self.cfg.GetMasterNode()
10235 return env, [mn], [mn]
10237 def Exec(self, feedback_fn):
10238 """Modifies the node group.
10243 if self.op.ndparams:
10244 self.group.ndparams = self.new_ndparams
10245 result.append(("ndparams", str(self.group.ndparams)))
10247 if self.op.alloc_policy:
10248 self.group.alloc_policy = self.op.alloc_policy
10250 self.cfg.Update(self.group, feedback_fn)
10255 class LUGroupRemove(LogicalUnit):
10256 HPATH = "group-remove"
10257 HTYPE = constants.HTYPE_GROUP
10260 def ExpandNames(self):
10261 # This will raises errors.OpPrereqError on its own:
10262 self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10263 self.needed_locks = {
10264 locking.LEVEL_NODEGROUP: [self.group_uuid],
10267 def CheckPrereq(self):
10268 """Check prerequisites.
10270 This checks that the given group name exists as a node group, that is
10271 empty (i.e., contains no nodes), and that is not the last group of the
10275 # Verify that the group is empty.
10276 group_nodes = [node.name
10277 for node in self.cfg.GetAllNodesInfo().values()
10278 if node.group == self.group_uuid]
10281 raise errors.OpPrereqError("Group '%s' not empty, has the following"
10283 (self.op.group_name,
10284 utils.CommaJoin(utils.NiceSort(group_nodes))),
10285 errors.ECODE_STATE)
10287 # Verify the cluster would not be left group-less.
10288 if len(self.cfg.GetNodeGroupList()) == 1:
10289 raise errors.OpPrereqError("Group '%s' is the last group in the cluster,"
10290 " which cannot be left without at least one"
10291 " group" % self.op.group_name,
10292 errors.ECODE_STATE)
10294 def BuildHooksEnv(self):
10295 """Build hooks env.
10299 "GROUP_NAME": self.op.group_name,
10301 mn = self.cfg.GetMasterNode()
10302 return env, [mn], [mn]
10304 def Exec(self, feedback_fn):
10305 """Remove the node group.
10309 self.cfg.RemoveNodeGroup(self.group_uuid)
10310 except errors.ConfigurationError:
10311 raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
10312 (self.op.group_name, self.group_uuid))
10314 self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10317 class LUGroupRename(LogicalUnit):
10318 HPATH = "group-rename"
10319 HTYPE = constants.HTYPE_GROUP
10322 def ExpandNames(self):
10323 # This raises errors.OpPrereqError on its own:
10324 self.group_uuid = self.cfg.LookupNodeGroup(self.op.old_name)
10326 self.needed_locks = {
10327 locking.LEVEL_NODEGROUP: [self.group_uuid],
10330 def CheckPrereq(self):
10331 """Check prerequisites.
10333 This checks that the given old_name exists as a node group, and that
10338 new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
10339 except errors.OpPrereqError:
10342 raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
10343 " node group (UUID: %s)" %
10344 (self.op.new_name, new_name_uuid),
10345 errors.ECODE_EXISTS)
10347 def BuildHooksEnv(self):
10348 """Build hooks env.
10352 "OLD_NAME": self.op.old_name,
10353 "NEW_NAME": self.op.new_name,
10356 mn = self.cfg.GetMasterNode()
10357 all_nodes = self.cfg.GetAllNodesInfo()
10359 all_nodes.pop(mn, None)
10361 for node in all_nodes.values():
10362 if node.group == self.group_uuid:
10363 run_nodes.append(node.name)
10365 return env, run_nodes, run_nodes
10367 def Exec(self, feedback_fn):
10368 """Rename the node group.
10371 group = self.cfg.GetNodeGroup(self.group_uuid)
10374 raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10375 (self.op.old_name, self.group_uuid))
10377 group.name = self.op.new_name
10378 self.cfg.Update(group, feedback_fn)
10380 return self.op.new_name
10383 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
10384 """Generic tags LU.
10386 This is an abstract class which is the parent of all the other tags LUs.
10390 def ExpandNames(self):
10391 self.needed_locks = {}
10392 if self.op.kind == constants.TAG_NODE:
10393 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
10394 self.needed_locks[locking.LEVEL_NODE] = self.op.name
10395 elif self.op.kind == constants.TAG_INSTANCE:
10396 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
10397 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
10399 # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
10400 # not possible to acquire the BGL based on opcode parameters)
10402 def CheckPrereq(self):
10403 """Check prerequisites.
10406 if self.op.kind == constants.TAG_CLUSTER:
10407 self.target = self.cfg.GetClusterInfo()
10408 elif self.op.kind == constants.TAG_NODE:
10409 self.target = self.cfg.GetNodeInfo(self.op.name)
10410 elif self.op.kind == constants.TAG_INSTANCE:
10411 self.target = self.cfg.GetInstanceInfo(self.op.name)
10413 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
10414 str(self.op.kind), errors.ECODE_INVAL)
10417 class LUTagsGet(TagsLU):
10418 """Returns the tags of a given object.
10423 def ExpandNames(self):
10424 TagsLU.ExpandNames(self)
10426 # Share locks as this is only a read operation
10427 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
10429 def Exec(self, feedback_fn):
10430 """Returns the tag list.
10433 return list(self.target.GetTags())
10436 class LUTagsSearch(NoHooksLU):
10437 """Searches the tags for a given pattern.
10442 def ExpandNames(self):
10443 self.needed_locks = {}
10445 def CheckPrereq(self):
10446 """Check prerequisites.
10448 This checks the pattern passed for validity by compiling it.
10452 self.re = re.compile(self.op.pattern)
10453 except re.error, err:
10454 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
10455 (self.op.pattern, err), errors.ECODE_INVAL)
10457 def Exec(self, feedback_fn):
10458 """Returns the tag list.
10462 tgts = [("/cluster", cfg.GetClusterInfo())]
10463 ilist = cfg.GetAllInstancesInfo().values()
10464 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
10465 nlist = cfg.GetAllNodesInfo().values()
10466 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
10468 for path, target in tgts:
10469 for tag in target.GetTags():
10470 if self.re.search(tag):
10471 results.append((path, tag))
10475 class LUTagsSet(TagsLU):
10476 """Sets a tag on a given object.
10481 def CheckPrereq(self):
10482 """Check prerequisites.
10484 This checks the type and length of the tag name and value.
10487 TagsLU.CheckPrereq(self)
10488 for tag in self.op.tags:
10489 objects.TaggableObject.ValidateTag(tag)
10491 def Exec(self, feedback_fn):
10496 for tag in self.op.tags:
10497 self.target.AddTag(tag)
10498 except errors.TagError, err:
10499 raise errors.OpExecError("Error while setting tag: %s" % str(err))
10500 self.cfg.Update(self.target, feedback_fn)
10503 class LUTagsDel(TagsLU):
10504 """Delete a list of tags from a given object.
10509 def CheckPrereq(self):
10510 """Check prerequisites.
10512 This checks that we have the given tag.
10515 TagsLU.CheckPrereq(self)
10516 for tag in self.op.tags:
10517 objects.TaggableObject.ValidateTag(tag)
10518 del_tags = frozenset(self.op.tags)
10519 cur_tags = self.target.GetTags()
10521 diff_tags = del_tags - cur_tags
10523 diff_names = ("'%s'" % i for i in sorted(diff_tags))
10524 raise errors.OpPrereqError("Tag(s) %s not found" %
10525 (utils.CommaJoin(diff_names), ),
10526 errors.ECODE_NOENT)
10528 def Exec(self, feedback_fn):
10529 """Remove the tag from the object.
10532 for tag in self.op.tags:
10533 self.target.RemoveTag(tag)
10534 self.cfg.Update(self.target, feedback_fn)
10537 class LUTestDelay(NoHooksLU):
10538 """Sleep for a specified amount of time.
10540 This LU sleeps on the master and/or nodes for a specified amount of
10546 def ExpandNames(self):
10547 """Expand names and set required locks.
10549 This expands the node list, if any.
10552 self.needed_locks = {}
10553 if self.op.on_nodes:
10554 # _GetWantedNodes can be used here, but is not always appropriate to use
10555 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
10556 # more information.
10557 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
10558 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
10560 def _TestDelay(self):
10561 """Do the actual sleep.
10564 if self.op.on_master:
10565 if not utils.TestDelay(self.op.duration):
10566 raise errors.OpExecError("Error during master delay test")
10567 if self.op.on_nodes:
10568 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
10569 for node, node_result in result.items():
10570 node_result.Raise("Failure during rpc call to node %s" % node)
10572 def Exec(self, feedback_fn):
10573 """Execute the test delay opcode, with the wanted repetitions.
10576 if self.op.repeat == 0:
10579 top_value = self.op.repeat - 1
10580 for i in range(self.op.repeat):
10581 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
10585 class LUTestJqueue(NoHooksLU):
10586 """Utility LU to test some aspects of the job queue.
10591 # Must be lower than default timeout for WaitForJobChange to see whether it
10592 # notices changed jobs
10593 _CLIENT_CONNECT_TIMEOUT = 20.0
10594 _CLIENT_CONFIRM_TIMEOUT = 60.0
10597 def _NotifyUsingSocket(cls, cb, errcls):
10598 """Opens a Unix socket and waits for another program to connect.
10601 @param cb: Callback to send socket name to client
10602 @type errcls: class
10603 @param errcls: Exception class to use for errors
10606 # Using a temporary directory as there's no easy way to create temporary
10607 # sockets without writing a custom loop around tempfile.mktemp and
10609 tmpdir = tempfile.mkdtemp()
10611 tmpsock = utils.PathJoin(tmpdir, "sock")
10613 logging.debug("Creating temporary socket at %s", tmpsock)
10614 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10619 # Send details to client
10622 # Wait for client to connect before continuing
10623 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10625 (conn, _) = sock.accept()
10626 except socket.error, err:
10627 raise errcls("Client didn't connect in time (%s)" % err)
10631 # Remove as soon as client is connected
10632 shutil.rmtree(tmpdir)
10634 # Wait for client to close
10637 # pylint: disable-msg=E1101
10638 # Instance of '_socketobject' has no ... member
10639 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10641 except socket.error, err:
10642 raise errcls("Client failed to confirm notification (%s)" % err)
10646 def _SendNotification(self, test, arg, sockname):
10647 """Sends a notification to the client.
10650 @param test: Test name
10651 @param arg: Test argument (depends on test)
10652 @type sockname: string
10653 @param sockname: Socket path
10656 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
10658 def _Notify(self, prereq, test, arg):
10659 """Notifies the client of a test.
10662 @param prereq: Whether this is a prereq-phase test
10664 @param test: Test name
10665 @param arg: Test argument (depends on test)
10669 errcls = errors.OpPrereqError
10671 errcls = errors.OpExecError
10673 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10677 def CheckArguments(self):
10678 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10679 self.expandnames_calls = 0
10681 def ExpandNames(self):
10682 checkargs_calls = getattr(self, "checkargs_calls", 0)
10683 if checkargs_calls < 1:
10684 raise errors.ProgrammerError("CheckArguments was not called")
10686 self.expandnames_calls += 1
10688 if self.op.notify_waitlock:
10689 self._Notify(True, constants.JQT_EXPANDNAMES, None)
10691 self.LogInfo("Expanding names")
10693 # Get lock on master node (just to get a lock, not for a particular reason)
10694 self.needed_locks = {
10695 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10698 def Exec(self, feedback_fn):
10699 if self.expandnames_calls < 1:
10700 raise errors.ProgrammerError("ExpandNames was not called")
10702 if self.op.notify_exec:
10703 self._Notify(False, constants.JQT_EXEC, None)
10705 self.LogInfo("Executing")
10707 if self.op.log_messages:
10708 self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10709 for idx, msg in enumerate(self.op.log_messages):
10710 self.LogInfo("Sending log message %s", idx + 1)
10711 feedback_fn(constants.JQT_MSGPREFIX + msg)
10712 # Report how many test messages have been sent
10713 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10716 raise errors.OpExecError("Opcode failure was requested")
10721 class IAllocator(object):
10722 """IAllocator framework.
10724 An IAllocator instance has three sets of attributes:
10725 - cfg that is needed to query the cluster
10726 - input data (all members of the _KEYS class attribute are required)
10727 - four buffer attributes (in|out_data|text), that represent the
10728 input (to the external script) in text and data structure format,
10729 and the output from it, again in two formats
10730 - the result variables from the script (success, info, nodes) for
10734 # pylint: disable-msg=R0902
10735 # lots of instance attributes
10737 "name", "mem_size", "disks", "disk_template",
10738 "os", "tags", "nics", "vcpus", "hypervisor",
10741 "name", "relocate_from",
10747 def __init__(self, cfg, rpc, mode, **kwargs):
10750 # init buffer variables
10751 self.in_text = self.out_text = self.in_data = self.out_data = None
10752 # init all input fields so that pylint is happy
10754 self.mem_size = self.disks = self.disk_template = None
10755 self.os = self.tags = self.nics = self.vcpus = None
10756 self.hypervisor = None
10757 self.relocate_from = None
10759 self.evac_nodes = None
10761 self.required_nodes = None
10762 # init result fields
10763 self.success = self.info = self.result = None
10764 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10765 keyset = self._ALLO_KEYS
10766 fn = self._AddNewInstance
10767 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10768 keyset = self._RELO_KEYS
10769 fn = self._AddRelocateInstance
10770 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10771 keyset = self._EVAC_KEYS
10772 fn = self._AddEvacuateNodes
10774 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10775 " IAllocator" % self.mode)
10777 if key not in keyset:
10778 raise errors.ProgrammerError("Invalid input parameter '%s' to"
10779 " IAllocator" % key)
10780 setattr(self, key, kwargs[key])
10783 if key not in kwargs:
10784 raise errors.ProgrammerError("Missing input parameter '%s' to"
10785 " IAllocator" % key)
10786 self._BuildInputData(fn)
10788 def _ComputeClusterData(self):
10789 """Compute the generic allocator input data.
10791 This is the data that is independent of the actual operation.
10795 cluster_info = cfg.GetClusterInfo()
10798 "version": constants.IALLOCATOR_VERSION,
10799 "cluster_name": cfg.GetClusterName(),
10800 "cluster_tags": list(cluster_info.GetTags()),
10801 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10802 # we don't have job IDs
10804 ninfo = cfg.GetAllNodesInfo()
10805 iinfo = cfg.GetAllInstancesInfo().values()
10806 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10809 node_list = [n.name for n in ninfo.values() if n.vm_capable]
10811 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10812 hypervisor_name = self.hypervisor
10813 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10814 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10815 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10816 hypervisor_name = cluster_info.enabled_hypervisors[0]
10818 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10821 self.rpc.call_all_instances_info(node_list,
10822 cluster_info.enabled_hypervisors)
10824 data["nodegroups"] = self._ComputeNodeGroupData(cfg)
10826 config_ndata = self._ComputeBasicNodeData(ninfo)
10827 data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
10828 i_list, config_ndata)
10829 assert len(data["nodes"]) == len(ninfo), \
10830 "Incomplete node data computed"
10832 data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
10834 self.in_data = data
10837 def _ComputeNodeGroupData(cfg):
10838 """Compute node groups data.
10842 for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
10844 "name": gdata.name,
10845 "alloc_policy": gdata.alloc_policy,
10850 def _ComputeBasicNodeData(node_cfg):
10851 """Compute global node data.
10854 @returns: a dict of name: (node dict, node config)
10858 for ninfo in node_cfg.values():
10859 # fill in static (config-based) values
10861 "tags": list(ninfo.GetTags()),
10862 "primary_ip": ninfo.primary_ip,
10863 "secondary_ip": ninfo.secondary_ip,
10864 "offline": ninfo.offline,
10865 "drained": ninfo.drained,
10866 "master_candidate": ninfo.master_candidate,
10867 "group": ninfo.group,
10868 "master_capable": ninfo.master_capable,
10869 "vm_capable": ninfo.vm_capable,
10872 node_results[ninfo.name] = pnr
10874 return node_results
10877 def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
10879 """Compute global node data.
10881 @param node_results: the basic node structures as filled from the config
10884 # make a copy of the current dict
10885 node_results = dict(node_results)
10886 for nname, nresult in node_data.items():
10887 assert nname in node_results, "Missing basic data for node %s" % nname
10888 ninfo = node_cfg[nname]
10890 if not (ninfo.offline or ninfo.drained):
10891 nresult.Raise("Can't get data for node %s" % nname)
10892 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10894 remote_info = nresult.payload
10896 for attr in ['memory_total', 'memory_free', 'memory_dom0',
10897 'vg_size', 'vg_free', 'cpu_total']:
10898 if attr not in remote_info:
10899 raise errors.OpExecError("Node '%s' didn't return attribute"
10900 " '%s'" % (nname, attr))
10901 if not isinstance(remote_info[attr], int):
10902 raise errors.OpExecError("Node '%s' returned invalid value"
10904 (nname, attr, remote_info[attr]))
10905 # compute memory used by primary instances
10906 i_p_mem = i_p_up_mem = 0
10907 for iinfo, beinfo in i_list:
10908 if iinfo.primary_node == nname:
10909 i_p_mem += beinfo[constants.BE_MEMORY]
10910 if iinfo.name not in node_iinfo[nname].payload:
10913 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10914 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10915 remote_info['memory_free'] -= max(0, i_mem_diff)
10918 i_p_up_mem += beinfo[constants.BE_MEMORY]
10920 # compute memory used by instances
10922 "total_memory": remote_info['memory_total'],
10923 "reserved_memory": remote_info['memory_dom0'],
10924 "free_memory": remote_info['memory_free'],
10925 "total_disk": remote_info['vg_size'],
10926 "free_disk": remote_info['vg_free'],
10927 "total_cpus": remote_info['cpu_total'],
10928 "i_pri_memory": i_p_mem,
10929 "i_pri_up_memory": i_p_up_mem,
10931 pnr_dyn.update(node_results[nname])
10933 node_results[nname] = pnr_dyn
10935 return node_results
10938 def _ComputeInstanceData(cluster_info, i_list):
10939 """Compute global instance data.
10943 for iinfo, beinfo in i_list:
10945 for nic in iinfo.nics:
10946 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10947 nic_dict = {"mac": nic.mac,
10949 "mode": filled_params[constants.NIC_MODE],
10950 "link": filled_params[constants.NIC_LINK],
10952 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10953 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10954 nic_data.append(nic_dict)
10956 "tags": list(iinfo.GetTags()),
10957 "admin_up": iinfo.admin_up,
10958 "vcpus": beinfo[constants.BE_VCPUS],
10959 "memory": beinfo[constants.BE_MEMORY],
10961 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10963 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10964 "disk_template": iinfo.disk_template,
10965 "hypervisor": iinfo.hypervisor,
10967 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10969 instance_data[iinfo.name] = pir
10971 return instance_data
10973 def _AddNewInstance(self):
10974 """Add new instance data to allocator structure.
10976 This in combination with _AllocatorGetClusterData will create the
10977 correct structure needed as input for the allocator.
10979 The checks for the completeness of the opcode must have already been
10983 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10985 if self.disk_template in constants.DTS_NET_MIRROR:
10986 self.required_nodes = 2
10988 self.required_nodes = 1
10991 "disk_template": self.disk_template,
10994 "vcpus": self.vcpus,
10995 "memory": self.mem_size,
10996 "disks": self.disks,
10997 "disk_space_total": disk_space,
10999 "required_nodes": self.required_nodes,
11003 def _AddRelocateInstance(self):
11004 """Add relocate instance data to allocator structure.
11006 This in combination with _IAllocatorGetClusterData will create the
11007 correct structure needed as input for the allocator.
11009 The checks for the completeness of the opcode must have already been
11013 instance = self.cfg.GetInstanceInfo(self.name)
11014 if instance is None:
11015 raise errors.ProgrammerError("Unknown instance '%s' passed to"
11016 " IAllocator" % self.name)
11018 if instance.disk_template not in constants.DTS_NET_MIRROR:
11019 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
11020 errors.ECODE_INVAL)
11022 if len(instance.secondary_nodes) != 1:
11023 raise errors.OpPrereqError("Instance has not exactly one secondary node",
11024 errors.ECODE_STATE)
11026 self.required_nodes = 1
11027 disk_sizes = [{'size': disk.size} for disk in instance.disks]
11028 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
11032 "disk_space_total": disk_space,
11033 "required_nodes": self.required_nodes,
11034 "relocate_from": self.relocate_from,
11038 def _AddEvacuateNodes(self):
11039 """Add evacuate nodes data to allocator structure.
11043 "evac_nodes": self.evac_nodes
11047 def _BuildInputData(self, fn):
11048 """Build input data structures.
11051 self._ComputeClusterData()
11054 request["type"] = self.mode
11055 self.in_data["request"] = request
11057 self.in_text = serializer.Dump(self.in_data)
11059 def Run(self, name, validate=True, call_fn=None):
11060 """Run an instance allocator and return the results.
11063 if call_fn is None:
11064 call_fn = self.rpc.call_iallocator_runner
11066 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
11067 result.Raise("Failure while running the iallocator script")
11069 self.out_text = result.payload
11071 self._ValidateResult()
11073 def _ValidateResult(self):
11074 """Process the allocator results.
11076 This will process and if successful save the result in
11077 self.out_data and the other parameters.
11081 rdict = serializer.Load(self.out_text)
11082 except Exception, err:
11083 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
11085 if not isinstance(rdict, dict):
11086 raise errors.OpExecError("Can't parse iallocator results: not a dict")
11088 # TODO: remove backwards compatiblity in later versions
11089 if "nodes" in rdict and "result" not in rdict:
11090 rdict["result"] = rdict["nodes"]
11093 for key in "success", "info", "result":
11094 if key not in rdict:
11095 raise errors.OpExecError("Can't parse iallocator results:"
11096 " missing key '%s'" % key)
11097 setattr(self, key, rdict[key])
11099 if not isinstance(rdict["result"], list):
11100 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
11102 self.out_data = rdict
11105 class LUTestAllocator(NoHooksLU):
11106 """Run allocator tests.
11108 This LU runs the allocator tests
11111 def CheckPrereq(self):
11112 """Check prerequisites.
11114 This checks the opcode parameters depending on the director and mode test.
11117 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11118 for attr in ["mem_size", "disks", "disk_template",
11119 "os", "tags", "nics", "vcpus"]:
11120 if not hasattr(self.op, attr):
11121 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
11122 attr, errors.ECODE_INVAL)
11123 iname = self.cfg.ExpandInstanceName(self.op.name)
11124 if iname is not None:
11125 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
11126 iname, errors.ECODE_EXISTS)
11127 if not isinstance(self.op.nics, list):
11128 raise errors.OpPrereqError("Invalid parameter 'nics'",
11129 errors.ECODE_INVAL)
11130 if not isinstance(self.op.disks, list):
11131 raise errors.OpPrereqError("Invalid parameter 'disks'",
11132 errors.ECODE_INVAL)
11133 for row in self.op.disks:
11134 if (not isinstance(row, dict) or
11135 "size" not in row or
11136 not isinstance(row["size"], int) or
11137 "mode" not in row or
11138 row["mode"] not in ['r', 'w']):
11139 raise errors.OpPrereqError("Invalid contents of the 'disks'"
11140 " parameter", errors.ECODE_INVAL)
11141 if self.op.hypervisor is None:
11142 self.op.hypervisor = self.cfg.GetHypervisorType()
11143 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11144 fname = _ExpandInstanceName(self.cfg, self.op.name)
11145 self.op.name = fname
11146 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
11147 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11148 if not hasattr(self.op, "evac_nodes"):
11149 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
11150 " opcode input", errors.ECODE_INVAL)
11152 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
11153 self.op.mode, errors.ECODE_INVAL)
11155 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
11156 if self.op.allocator is None:
11157 raise errors.OpPrereqError("Missing allocator name",
11158 errors.ECODE_INVAL)
11159 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
11160 raise errors.OpPrereqError("Wrong allocator test '%s'" %
11161 self.op.direction, errors.ECODE_INVAL)
11163 def Exec(self, feedback_fn):
11164 """Run the allocator test.
11167 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11168 ial = IAllocator(self.cfg, self.rpc,
11171 mem_size=self.op.mem_size,
11172 disks=self.op.disks,
11173 disk_template=self.op.disk_template,
11177 vcpus=self.op.vcpus,
11178 hypervisor=self.op.hypervisor,
11180 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11181 ial = IAllocator(self.cfg, self.rpc,
11184 relocate_from=list(self.relocate_from),
11186 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11187 ial = IAllocator(self.cfg, self.rpc,
11189 evac_nodes=self.op.evac_nodes)
11191 raise errors.ProgrammerError("Uncatched mode %s in"
11192 " LUTestAllocator.Exec", self.op.mode)
11194 if self.op.direction == constants.IALLOCATOR_DIR_IN:
11195 result = ial.in_text
11197 ial.Run(self.op.allocator, validate=False)
11198 result = ial.out_text
11202 #: Query type implementations
11204 constants.QR_INSTANCE: _InstanceQuery,
11205 constants.QR_NODE: _NodeQuery,
11206 constants.QR_GROUP: _GroupQuery,
11210 def _GetQueryImplementation(name):
11211 """Returns the implemtnation for a query type.
11213 @param name: Query type, must be one of L{constants.QR_OP_QUERY}
11217 return _QUERY_IMPL[name]
11219 raise errors.OpPrereqError("Unknown query resource '%s'" % name,
11220 errors.ECODE_INVAL)