4 # Copyright (C) 2006, 2007, 2008, 2009, 2010 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Module implementing the master-side code."""
24 # pylint: disable-msg=W0201,C0302
26 # W0201 since most LU attributes are defined in CheckPrereq or similar
29 # C0302: since we have waaaay to many lines in this module
43 from ganeti import ssh
44 from ganeti import utils
45 from ganeti import errors
46 from ganeti import hypervisor
47 from ganeti import locking
48 from ganeti import constants
49 from ganeti import objects
50 from ganeti import serializer
51 from ganeti import ssconf
52 from ganeti import uidpool
53 from ganeti import compat
54 from ganeti import masterd
55 from ganeti import netutils
57 import ganeti.masterd.instance # pylint: disable-msg=W0611
60 # Modifiable default values; need to define these here before the
64 """Returns an empty list.
71 """Returns an empty dict.
77 #: The without-default default value
81 #: The no-type (value to complex to check it in the type system)
87 """Checks if the given value is not None.
90 return val is not None
94 """Checks if the given value is None.
101 """Checks if the given value is a boolean.
104 return isinstance(val, bool)
108 """Checks if the given value is an integer.
111 return isinstance(val, int)
115 """Checks if the given value is a float.
118 return isinstance(val, float)
122 """Checks if the given value is a string.
125 return isinstance(val, basestring)
129 """Checks if a given value evaluates to a boolean True value.
135 def _TElemOf(target_list):
136 """Builds a function that checks if a given value is a member of a list.
139 return lambda val: val in target_list
144 """Checks if the given value is a list.
147 return isinstance(val, list)
151 """Checks if the given value is a dictionary.
154 return isinstance(val, dict)
159 """Combine multiple functions using an AND operation.
163 return compat.all(t(val) for t in args)
168 """Combine multiple functions using an AND operation.
172 return compat.any(t(val) for t in args)
178 #: a non-empty string
179 _TNonEmptyString = _TAnd(_TString, _TTrue)
182 #: a maybe non-empty string
183 _TMaybeString = _TOr(_TNonEmptyString, _TNone)
186 #: a maybe boolean (bool or none)
187 _TMaybeBool = _TOr(_TBool, _TNone)
190 #: a positive integer
191 _TPositiveInt = _TAnd(_TInt, lambda v: v >= 0)
193 #: a strictly positive integer
194 _TStrictPositiveInt = _TAnd(_TInt, lambda v: v > 0)
197 def _TListOf(my_type):
198 """Checks if a given value is a list with all elements of the same type.
202 lambda lst: compat.all(my_type(v) for v in lst))
205 def _TDictOf(key_type, val_type):
206 """Checks a dict type for the type of its key/values.
210 lambda my_dict: (compat.all(key_type(v) for v in my_dict.keys())
211 and compat.all(val_type(v)
212 for v in my_dict.values())))
215 # Common opcode attributes
217 #: output fields for a query operation
218 _POutputFields = ("output_fields", _NoDefault, _TListOf(_TNonEmptyString))
221 #: the shutdown timeout
222 _PShutdownTimeout = ("shutdown_timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT,
225 #: the force parameter
226 _PForce = ("force", False, _TBool)
228 #: a required instance name (for single-instance LUs)
229 _PInstanceName = ("instance_name", _NoDefault, _TNonEmptyString)
232 #: a required node name (for single-node LUs)
233 _PNodeName = ("node_name", _NoDefault, _TNonEmptyString)
235 #: the migration type (live/non-live)
236 _PMigrationMode = ("mode", None, _TOr(_TNone,
237 _TElemOf(constants.HT_MIGRATION_MODES)))
239 #: the obsolete 'live' mode (boolean)
240 _PMigrationLive = ("live", None, _TMaybeBool)
244 class LogicalUnit(object):
245 """Logical Unit base class.
247 Subclasses must follow these rules:
248 - implement ExpandNames
249 - implement CheckPrereq (except when tasklets are used)
250 - implement Exec (except when tasklets are used)
251 - implement BuildHooksEnv
252 - redefine HPATH and HTYPE
253 - optionally redefine their run requirements:
254 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
256 Note that all commands require root permissions.
258 @ivar dry_run_result: the value (if any) that will be returned to the caller
259 in dry-run mode (signalled by opcode dry_run parameter)
260 @cvar _OP_PARAMS: a list of opcode attributes, their defaults values
261 they should get if not already defined, and types they must match
269 def __init__(self, processor, op, context, rpc):
270 """Constructor for LogicalUnit.
272 This needs to be overridden in derived classes in order to check op
276 self.proc = processor
278 self.cfg = context.cfg
279 self.context = context
281 # Dicts used to declare locking needs to mcpu
282 self.needed_locks = None
283 self.acquired_locks = {}
284 self.share_locks = dict.fromkeys(locking.LEVELS, 0)
286 self.remove_locks = {}
287 # Used to force good behavior when calling helper functions
288 self.recalculate_locks = {}
291 self.Log = processor.Log # pylint: disable-msg=C0103
292 self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
293 self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
294 self.LogStep = processor.LogStep # pylint: disable-msg=C0103
295 # support for dry-run
296 self.dry_run_result = None
297 # support for generic debug attribute
298 if (not hasattr(self.op, "debug_level") or
299 not isinstance(self.op.debug_level, int)):
300 self.op.debug_level = 0
305 # The new kind-of-type-system
306 op_id = self.op.OP_ID
307 for attr_name, aval, test in self._OP_PARAMS:
308 if not hasattr(op, attr_name):
309 if aval == _NoDefault:
310 raise errors.OpPrereqError("Required parameter '%s.%s' missing" %
311 (op_id, attr_name), errors.ECODE_INVAL)
317 setattr(self.op, attr_name, dval)
318 attr_val = getattr(op, attr_name)
322 if not callable(test):
323 raise errors.ProgrammerError("Validation for parameter '%s.%s' failed,"
324 " given type is not a proper type (%s)" %
325 (op_id, attr_name, test))
326 if not test(attr_val):
327 logging.error("OpCode %s, parameter %s, has invalid type %s/value %s",
328 self.op.OP_ID, attr_name, type(attr_val), attr_val)
329 raise errors.OpPrereqError("Parameter '%s.%s' fails validation" %
330 (op_id, attr_name), errors.ECODE_INVAL)
332 self.CheckArguments()
335 """Returns the SshRunner object
339 self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
342 ssh = property(fget=__GetSSH)
344 def CheckArguments(self):
345 """Check syntactic validity for the opcode arguments.
347 This method is for doing a simple syntactic check and ensure
348 validity of opcode parameters, without any cluster-related
349 checks. While the same can be accomplished in ExpandNames and/or
350 CheckPrereq, doing these separate is better because:
352 - ExpandNames is left as as purely a lock-related function
353 - CheckPrereq is run after we have acquired locks (and possible
356 The function is allowed to change the self.op attribute so that
357 later methods can no longer worry about missing parameters.
362 def ExpandNames(self):
363 """Expand names for this LU.
365 This method is called before starting to execute the opcode, and it should
366 update all the parameters of the opcode to their canonical form (e.g. a
367 short node name must be fully expanded after this method has successfully
368 completed). This way locking, hooks, logging, ecc. can work correctly.
370 LUs which implement this method must also populate the self.needed_locks
371 member, as a dict with lock levels as keys, and a list of needed lock names
374 - use an empty dict if you don't need any lock
375 - if you don't need any lock at a particular level omit that level
376 - don't put anything for the BGL level
377 - if you want all locks at a level use locking.ALL_SET as a value
379 If you need to share locks (rather than acquire them exclusively) at one
380 level you can modify self.share_locks, setting a true value (usually 1) for
381 that level. By default locks are not shared.
383 This function can also define a list of tasklets, which then will be
384 executed in order instead of the usual LU-level CheckPrereq and Exec
385 functions, if those are not defined by the LU.
389 # Acquire all nodes and one instance
390 self.needed_locks = {
391 locking.LEVEL_NODE: locking.ALL_SET,
392 locking.LEVEL_INSTANCE: ['instance1.example.com'],
394 # Acquire just two nodes
395 self.needed_locks = {
396 locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
399 self.needed_locks = {} # No, you can't leave it to the default value None
402 # The implementation of this method is mandatory only if the new LU is
403 # concurrent, so that old LUs don't need to be changed all at the same
406 self.needed_locks = {} # Exclusive LUs don't need locks.
408 raise NotImplementedError
410 def DeclareLocks(self, level):
411 """Declare LU locking needs for a level
413 While most LUs can just declare their locking needs at ExpandNames time,
414 sometimes there's the need to calculate some locks after having acquired
415 the ones before. This function is called just before acquiring locks at a
416 particular level, but after acquiring the ones at lower levels, and permits
417 such calculations. It can be used to modify self.needed_locks, and by
418 default it does nothing.
420 This function is only called if you have something already set in
421 self.needed_locks for the level.
423 @param level: Locking level which is going to be locked
424 @type level: member of ganeti.locking.LEVELS
428 def CheckPrereq(self):
429 """Check prerequisites for this LU.
431 This method should check that the prerequisites for the execution
432 of this LU are fulfilled. It can do internode communication, but
433 it should be idempotent - no cluster or system changes are
436 The method should raise errors.OpPrereqError in case something is
437 not fulfilled. Its return value is ignored.
439 This method should also update all the parameters of the opcode to
440 their canonical form if it hasn't been done by ExpandNames before.
443 if self.tasklets is not None:
444 for (idx, tl) in enumerate(self.tasklets):
445 logging.debug("Checking prerequisites for tasklet %s/%s",
446 idx + 1, len(self.tasklets))
451 def Exec(self, feedback_fn):
454 This method should implement the actual work. It should raise
455 errors.OpExecError for failures that are somewhat dealt with in
459 if self.tasklets is not None:
460 for (idx, tl) in enumerate(self.tasklets):
461 logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
464 raise NotImplementedError
466 def BuildHooksEnv(self):
467 """Build hooks environment for this LU.
469 This method should return a three-node tuple consisting of: a dict
470 containing the environment that will be used for running the
471 specific hook for this LU, a list of node names on which the hook
472 should run before the execution, and a list of node names on which
473 the hook should run after the execution.
475 The keys of the dict must not have 'GANETI_' prefixed as this will
476 be handled in the hooks runner. Also note additional keys will be
477 added by the hooks runner. If the LU doesn't define any
478 environment, an empty dict (and not None) should be returned.
480 No nodes should be returned as an empty list (and not None).
482 Note that if the HPATH for a LU class is None, this function will
486 raise NotImplementedError
488 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
489 """Notify the LU about the results of its hooks.
491 This method is called every time a hooks phase is executed, and notifies
492 the Logical Unit about the hooks' result. The LU can then use it to alter
493 its result based on the hooks. By default the method does nothing and the
494 previous result is passed back unchanged but any LU can define it if it
495 wants to use the local cluster hook-scripts somehow.
497 @param phase: one of L{constants.HOOKS_PHASE_POST} or
498 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
499 @param hook_results: the results of the multi-node hooks rpc call
500 @param feedback_fn: function used send feedback back to the caller
501 @param lu_result: the previous Exec result this LU had, or None
503 @return: the new Exec result, based on the previous result
507 # API must be kept, thus we ignore the unused argument and could
508 # be a function warnings
509 # pylint: disable-msg=W0613,R0201
512 def _ExpandAndLockInstance(self):
513 """Helper function to expand and lock an instance.
515 Many LUs that work on an instance take its name in self.op.instance_name
516 and need to expand it and then declare the expanded name for locking. This
517 function does it, and then updates self.op.instance_name to the expanded
518 name. It also initializes needed_locks as a dict, if this hasn't been done
522 if self.needed_locks is None:
523 self.needed_locks = {}
525 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
526 "_ExpandAndLockInstance called with instance-level locks set"
527 self.op.instance_name = _ExpandInstanceName(self.cfg,
528 self.op.instance_name)
529 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
531 def _LockInstancesNodes(self, primary_only=False):
532 """Helper function to declare instances' nodes for locking.
534 This function should be called after locking one or more instances to lock
535 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
536 with all primary or secondary nodes for instances already locked and
537 present in self.needed_locks[locking.LEVEL_INSTANCE].
539 It should be called from DeclareLocks, and for safety only works if
540 self.recalculate_locks[locking.LEVEL_NODE] is set.
542 In the future it may grow parameters to just lock some instance's nodes, or
543 to just lock primaries or secondary nodes, if needed.
545 If should be called in DeclareLocks in a way similar to::
547 if level == locking.LEVEL_NODE:
548 self._LockInstancesNodes()
550 @type primary_only: boolean
551 @param primary_only: only lock primary nodes of locked instances
554 assert locking.LEVEL_NODE in self.recalculate_locks, \
555 "_LockInstancesNodes helper function called with no nodes to recalculate"
557 # TODO: check if we're really been called with the instance locks held
559 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
560 # future we might want to have different behaviors depending on the value
561 # of self.recalculate_locks[locking.LEVEL_NODE]
563 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
564 instance = self.context.cfg.GetInstanceInfo(instance_name)
565 wanted_nodes.append(instance.primary_node)
567 wanted_nodes.extend(instance.secondary_nodes)
569 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
570 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
571 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
572 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
574 del self.recalculate_locks[locking.LEVEL_NODE]
577 class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
578 """Simple LU which runs no hooks.
580 This LU is intended as a parent for other LogicalUnits which will
581 run no hooks, in order to reduce duplicate code.
587 def BuildHooksEnv(self):
588 """Empty BuildHooksEnv for NoHooksLu.
590 This just raises an error.
593 assert False, "BuildHooksEnv called for NoHooksLUs"
597 """Tasklet base class.
599 Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
600 they can mix legacy code with tasklets. Locking needs to be done in the LU,
601 tasklets know nothing about locks.
603 Subclasses must follow these rules:
604 - Implement CheckPrereq
608 def __init__(self, lu):
615 def CheckPrereq(self):
616 """Check prerequisites for this tasklets.
618 This method should check whether the prerequisites for the execution of
619 this tasklet are fulfilled. It can do internode communication, but it
620 should be idempotent - no cluster or system changes are allowed.
622 The method should raise errors.OpPrereqError in case something is not
623 fulfilled. Its return value is ignored.
625 This method should also update all parameters to their canonical form if it
626 hasn't been done before.
631 def Exec(self, feedback_fn):
632 """Execute the tasklet.
634 This method should implement the actual work. It should raise
635 errors.OpExecError for failures that are somewhat dealt with in code, or
639 raise NotImplementedError
642 def _GetWantedNodes(lu, nodes):
643 """Returns list of checked and expanded node names.
645 @type lu: L{LogicalUnit}
646 @param lu: the logical unit on whose behalf we execute
648 @param nodes: list of node names or None for all nodes
650 @return: the list of nodes, sorted
651 @raise errors.ProgrammerError: if the nodes parameter is wrong type
655 raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
656 " non-empty list of nodes whose name is to be expanded.")
658 wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
659 return utils.NiceSort(wanted)
662 def _GetWantedInstances(lu, instances):
663 """Returns list of checked and expanded instance names.
665 @type lu: L{LogicalUnit}
666 @param lu: the logical unit on whose behalf we execute
667 @type instances: list
668 @param instances: list of instance names or None for all instances
670 @return: the list of instances, sorted
671 @raise errors.OpPrereqError: if the instances parameter is wrong type
672 @raise errors.OpPrereqError: if any of the passed instances is not found
676 wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
678 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
682 def _GetUpdatedParams(old_params, update_dict,
683 use_default=True, use_none=False):
684 """Return the new version of a parameter dictionary.
686 @type old_params: dict
687 @param old_params: old parameters
688 @type update_dict: dict
689 @param update_dict: dict containing new parameter values, or
690 constants.VALUE_DEFAULT to reset the parameter to its default
692 @param use_default: boolean
693 @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
694 values as 'to be deleted' values
695 @param use_none: boolean
696 @type use_none: whether to recognise C{None} values as 'to be
699 @return: the new parameter dictionary
702 params_copy = copy.deepcopy(old_params)
703 for key, val in update_dict.iteritems():
704 if ((use_default and val == constants.VALUE_DEFAULT) or
705 (use_none and val is None)):
711 params_copy[key] = val
715 def _CheckOutputFields(static, dynamic, selected):
716 """Checks whether all selected fields are valid.
718 @type static: L{utils.FieldSet}
719 @param static: static fields set
720 @type dynamic: L{utils.FieldSet}
721 @param dynamic: dynamic fields set
728 delta = f.NonMatching(selected)
730 raise errors.OpPrereqError("Unknown output fields selected: %s"
731 % ",".join(delta), errors.ECODE_INVAL)
734 def _CheckGlobalHvParams(params):
735 """Validates that given hypervisor params are not global ones.
737 This will ensure that instances don't get customised versions of
741 used_globals = constants.HVC_GLOBALS.intersection(params)
743 msg = ("The following hypervisor parameters are global and cannot"
744 " be customized at instance level, please modify them at"
745 " cluster level: %s" % utils.CommaJoin(used_globals))
746 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
749 def _CheckNodeOnline(lu, node):
750 """Ensure that a given node is online.
752 @param lu: the LU on behalf of which we make the check
753 @param node: the node to check
754 @raise errors.OpPrereqError: if the node is offline
757 if lu.cfg.GetNodeInfo(node).offline:
758 raise errors.OpPrereqError("Can't use offline node %s" % node,
762 def _CheckNodeNotDrained(lu, node):
763 """Ensure that a given node is not drained.
765 @param lu: the LU on behalf of which we make the check
766 @param node: the node to check
767 @raise errors.OpPrereqError: if the node is drained
770 if lu.cfg.GetNodeInfo(node).drained:
771 raise errors.OpPrereqError("Can't use drained node %s" % node,
775 def _CheckNodeHasOS(lu, node, os_name, force_variant):
776 """Ensure that a node supports a given OS.
778 @param lu: the LU on behalf of which we make the check
779 @param node: the node to check
780 @param os_name: the OS to query about
781 @param force_variant: whether to ignore variant errors
782 @raise errors.OpPrereqError: if the node is not supporting the OS
785 result = lu.rpc.call_os_get(node, os_name)
786 result.Raise("OS '%s' not in supported OS list for node %s" %
788 prereq=True, ecode=errors.ECODE_INVAL)
789 if not force_variant:
790 _CheckOSVariant(result.payload, os_name)
793 def _RequireFileStorage():
794 """Checks that file storage is enabled.
796 @raise errors.OpPrereqError: when file storage is disabled
799 if not constants.ENABLE_FILE_STORAGE:
800 raise errors.OpPrereqError("File storage disabled at configure time",
804 def _CheckDiskTemplate(template):
805 """Ensure a given disk template is valid.
808 if template not in constants.DISK_TEMPLATES:
809 msg = ("Invalid disk template name '%s', valid templates are: %s" %
810 (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
811 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
812 if template == constants.DT_FILE:
813 _RequireFileStorage()
817 def _CheckStorageType(storage_type):
818 """Ensure a given storage type is valid.
821 if storage_type not in constants.VALID_STORAGE_TYPES:
822 raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
824 if storage_type == constants.ST_FILE:
825 _RequireFileStorage()
829 def _GetClusterDomainSecret():
830 """Reads the cluster domain secret.
833 return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
837 def _CheckInstanceDown(lu, instance, reason):
838 """Ensure that an instance is not running."""
839 if instance.admin_up:
840 raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
841 (instance.name, reason), errors.ECODE_STATE)
843 pnode = instance.primary_node
844 ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
845 ins_l.Raise("Can't contact node %s for instance information" % pnode,
846 prereq=True, ecode=errors.ECODE_ENVIRON)
848 if instance.name in ins_l.payload:
849 raise errors.OpPrereqError("Instance %s is running, %s" %
850 (instance.name, reason), errors.ECODE_STATE)
853 def _ExpandItemName(fn, name, kind):
854 """Expand an item name.
856 @param fn: the function to use for expansion
857 @param name: requested item name
858 @param kind: text description ('Node' or 'Instance')
859 @return: the resolved (full) name
860 @raise errors.OpPrereqError: if the item is not found
864 if full_name is None:
865 raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
870 def _ExpandNodeName(cfg, name):
871 """Wrapper over L{_ExpandItemName} for nodes."""
872 return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
875 def _ExpandInstanceName(cfg, name):
876 """Wrapper over L{_ExpandItemName} for instance."""
877 return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
880 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
881 memory, vcpus, nics, disk_template, disks,
882 bep, hvp, hypervisor_name):
883 """Builds instance related env variables for hooks
885 This builds the hook environment from individual variables.
888 @param name: the name of the instance
889 @type primary_node: string
890 @param primary_node: the name of the instance's primary node
891 @type secondary_nodes: list
892 @param secondary_nodes: list of secondary nodes as strings
893 @type os_type: string
894 @param os_type: the name of the instance's OS
895 @type status: boolean
896 @param status: the should_run status of the instance
898 @param memory: the memory size of the instance
900 @param vcpus: the count of VCPUs the instance has
902 @param nics: list of tuples (ip, mac, mode, link) representing
903 the NICs the instance has
904 @type disk_template: string
905 @param disk_template: the disk template of the instance
907 @param disks: the list of (size, mode) pairs
909 @param bep: the backend parameters for the instance
911 @param hvp: the hypervisor parameters for the instance
912 @type hypervisor_name: string
913 @param hypervisor_name: the hypervisor for the instance
915 @return: the hook environment for this instance
924 "INSTANCE_NAME": name,
925 "INSTANCE_PRIMARY": primary_node,
926 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
927 "INSTANCE_OS_TYPE": os_type,
928 "INSTANCE_STATUS": str_status,
929 "INSTANCE_MEMORY": memory,
930 "INSTANCE_VCPUS": vcpus,
931 "INSTANCE_DISK_TEMPLATE": disk_template,
932 "INSTANCE_HYPERVISOR": hypervisor_name,
936 nic_count = len(nics)
937 for idx, (ip, mac, mode, link) in enumerate(nics):
940 env["INSTANCE_NIC%d_IP" % idx] = ip
941 env["INSTANCE_NIC%d_MAC" % idx] = mac
942 env["INSTANCE_NIC%d_MODE" % idx] = mode
943 env["INSTANCE_NIC%d_LINK" % idx] = link
944 if mode == constants.NIC_MODE_BRIDGED:
945 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
949 env["INSTANCE_NIC_COUNT"] = nic_count
952 disk_count = len(disks)
953 for idx, (size, mode) in enumerate(disks):
954 env["INSTANCE_DISK%d_SIZE" % idx] = size
955 env["INSTANCE_DISK%d_MODE" % idx] = mode
959 env["INSTANCE_DISK_COUNT"] = disk_count
961 for source, kind in [(bep, "BE"), (hvp, "HV")]:
962 for key, value in source.items():
963 env["INSTANCE_%s_%s" % (kind, key)] = value
968 def _NICListToTuple(lu, nics):
969 """Build a list of nic information tuples.
971 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
972 value in LUQueryInstanceData.
974 @type lu: L{LogicalUnit}
975 @param lu: the logical unit on whose behalf we execute
976 @type nics: list of L{objects.NIC}
977 @param nics: list of nics to convert to hooks tuples
981 cluster = lu.cfg.GetClusterInfo()
985 filled_params = cluster.SimpleFillNIC(nic.nicparams)
986 mode = filled_params[constants.NIC_MODE]
987 link = filled_params[constants.NIC_LINK]
988 hooks_nics.append((ip, mac, mode, link))
992 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
993 """Builds instance related env variables for hooks from an object.
995 @type lu: L{LogicalUnit}
996 @param lu: the logical unit on whose behalf we execute
997 @type instance: L{objects.Instance}
998 @param instance: the instance for which we should build the
1000 @type override: dict
1001 @param override: dictionary with key/values that will override
1004 @return: the hook environment dictionary
1007 cluster = lu.cfg.GetClusterInfo()
1008 bep = cluster.FillBE(instance)
1009 hvp = cluster.FillHV(instance)
1011 'name': instance.name,
1012 'primary_node': instance.primary_node,
1013 'secondary_nodes': instance.secondary_nodes,
1014 'os_type': instance.os,
1015 'status': instance.admin_up,
1016 'memory': bep[constants.BE_MEMORY],
1017 'vcpus': bep[constants.BE_VCPUS],
1018 'nics': _NICListToTuple(lu, instance.nics),
1019 'disk_template': instance.disk_template,
1020 'disks': [(disk.size, disk.mode) for disk in instance.disks],
1023 'hypervisor_name': instance.hypervisor,
1026 args.update(override)
1027 return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
1030 def _AdjustCandidatePool(lu, exceptions):
1031 """Adjust the candidate pool after node operations.
1034 mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1036 lu.LogInfo("Promoted nodes to master candidate role: %s",
1037 utils.CommaJoin(node.name for node in mod_list))
1038 for name in mod_list:
1039 lu.context.ReaddNode(name)
1040 mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1042 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1046 def _DecideSelfPromotion(lu, exceptions=None):
1047 """Decide whether I should promote myself as a master candidate.
1050 cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1051 mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1052 # the new node will increase mc_max with one, so:
1053 mc_should = min(mc_should + 1, cp_size)
1054 return mc_now < mc_should
1057 def _CheckNicsBridgesExist(lu, target_nics, target_node):
1058 """Check that the brigdes needed by a list of nics exist.
1061 cluster = lu.cfg.GetClusterInfo()
1062 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1063 brlist = [params[constants.NIC_LINK] for params in paramslist
1064 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1066 result = lu.rpc.call_bridges_exist(target_node, brlist)
1067 result.Raise("Error checking bridges on destination node '%s'" %
1068 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1071 def _CheckInstanceBridgesExist(lu, instance, node=None):
1072 """Check that the brigdes needed by an instance exist.
1076 node = instance.primary_node
1077 _CheckNicsBridgesExist(lu, instance.nics, node)
1080 def _CheckOSVariant(os_obj, name):
1081 """Check whether an OS name conforms to the os variants specification.
1083 @type os_obj: L{objects.OS}
1084 @param os_obj: OS object to check
1086 @param name: OS name passed by the user, to check for validity
1089 if not os_obj.supported_variants:
1092 variant = name.split("+", 1)[1]
1094 raise errors.OpPrereqError("OS name must include a variant",
1097 if variant not in os_obj.supported_variants:
1098 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1101 def _GetNodeInstancesInner(cfg, fn):
1102 return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1105 def _GetNodeInstances(cfg, node_name):
1106 """Returns a list of all primary and secondary instances on a node.
1110 return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1113 def _GetNodePrimaryInstances(cfg, node_name):
1114 """Returns primary instances on a node.
1117 return _GetNodeInstancesInner(cfg,
1118 lambda inst: node_name == inst.primary_node)
1121 def _GetNodeSecondaryInstances(cfg, node_name):
1122 """Returns secondary instances on a node.
1125 return _GetNodeInstancesInner(cfg,
1126 lambda inst: node_name in inst.secondary_nodes)
1129 def _GetStorageTypeArgs(cfg, storage_type):
1130 """Returns the arguments for a storage type.
1133 # Special case for file storage
1134 if storage_type == constants.ST_FILE:
1135 # storage.FileStorage wants a list of storage directories
1136 return [[cfg.GetFileStorageDir()]]
1141 def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1144 for dev in instance.disks:
1145 cfg.SetDiskID(dev, node_name)
1147 result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1148 result.Raise("Failed to get disk status from node %s" % node_name,
1149 prereq=prereq, ecode=errors.ECODE_ENVIRON)
1151 for idx, bdev_status in enumerate(result.payload):
1152 if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1158 def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1159 """Check the sanity of iallocator and node arguments and use the
1160 cluster-wide iallocator if appropriate.
1162 Check that at most one of (iallocator, node) is specified. If none is
1163 specified, then the LU's opcode's iallocator slot is filled with the
1164 cluster-wide default iallocator.
1166 @type iallocator_slot: string
1167 @param iallocator_slot: the name of the opcode iallocator slot
1168 @type node_slot: string
1169 @param node_slot: the name of the opcode target node slot
1172 node = getattr(lu.op, node_slot, None)
1173 iallocator = getattr(lu.op, iallocator_slot, None)
1175 if node is not None and iallocator is not None:
1176 raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1178 elif node is None and iallocator is None:
1179 default_iallocator = lu.cfg.GetDefaultIAllocator()
1180 if default_iallocator:
1181 setattr(lu.op, iallocator_slot, default_iallocator)
1183 raise errors.OpPrereqError("No iallocator or node given and no"
1184 " cluster-wide default iallocator found."
1185 " Please specify either an iallocator or a"
1186 " node, or set a cluster-wide default"
1190 class LUPostInitCluster(LogicalUnit):
1191 """Logical unit for running hooks after cluster initialization.
1194 HPATH = "cluster-init"
1195 HTYPE = constants.HTYPE_CLUSTER
1197 def BuildHooksEnv(self):
1201 env = {"OP_TARGET": self.cfg.GetClusterName()}
1202 mn = self.cfg.GetMasterNode()
1203 return env, [], [mn]
1205 def Exec(self, feedback_fn):
1212 class LUDestroyCluster(LogicalUnit):
1213 """Logical unit for destroying the cluster.
1216 HPATH = "cluster-destroy"
1217 HTYPE = constants.HTYPE_CLUSTER
1219 def BuildHooksEnv(self):
1223 env = {"OP_TARGET": self.cfg.GetClusterName()}
1226 def CheckPrereq(self):
1227 """Check prerequisites.
1229 This checks whether the cluster is empty.
1231 Any errors are signaled by raising errors.OpPrereqError.
1234 master = self.cfg.GetMasterNode()
1236 nodelist = self.cfg.GetNodeList()
1237 if len(nodelist) != 1 or nodelist[0] != master:
1238 raise errors.OpPrereqError("There are still %d node(s) in"
1239 " this cluster." % (len(nodelist) - 1),
1241 instancelist = self.cfg.GetInstanceList()
1243 raise errors.OpPrereqError("There are still %d instance(s) in"
1244 " this cluster." % len(instancelist),
1247 def Exec(self, feedback_fn):
1248 """Destroys the cluster.
1251 master = self.cfg.GetMasterNode()
1252 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
1254 # Run post hooks on master node before it's removed
1255 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1257 hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1259 # pylint: disable-msg=W0702
1260 self.LogWarning("Errors occurred running hooks on %s" % master)
1262 result = self.rpc.call_node_stop_master(master, False)
1263 result.Raise("Could not disable the master role")
1265 if modify_ssh_setup:
1266 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
1267 utils.CreateBackup(priv_key)
1268 utils.CreateBackup(pub_key)
1273 def _VerifyCertificate(filename):
1274 """Verifies a certificate for LUVerifyCluster.
1276 @type filename: string
1277 @param filename: Path to PEM file
1281 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1282 utils.ReadFile(filename))
1283 except Exception, err: # pylint: disable-msg=W0703
1284 return (LUVerifyCluster.ETYPE_ERROR,
1285 "Failed to load X509 certificate %s: %s" % (filename, err))
1288 utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1289 constants.SSL_CERT_EXPIRATION_ERROR)
1292 fnamemsg = "While verifying %s: %s" % (filename, msg)
1297 return (None, fnamemsg)
1298 elif errcode == utils.CERT_WARNING:
1299 return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
1300 elif errcode == utils.CERT_ERROR:
1301 return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
1303 raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1306 class LUVerifyCluster(LogicalUnit):
1307 """Verifies the cluster status.
1310 HPATH = "cluster-verify"
1311 HTYPE = constants.HTYPE_CLUSTER
1313 ("skip_checks", _EmptyList,
1314 _TListOf(_TElemOf(constants.VERIFY_OPTIONAL_CHECKS))),
1315 ("verbose", False, _TBool),
1316 ("error_codes", False, _TBool),
1317 ("debug_simulate_errors", False, _TBool),
1321 TCLUSTER = "cluster"
1323 TINSTANCE = "instance"
1325 ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1326 ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1327 EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1328 EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1329 EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1330 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1331 EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1332 EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1333 ENODEDRBD = (TNODE, "ENODEDRBD")
1334 ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1335 ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1336 ENODEHOOKS = (TNODE, "ENODEHOOKS")
1337 ENODEHV = (TNODE, "ENODEHV")
1338 ENODELVM = (TNODE, "ENODELVM")
1339 ENODEN1 = (TNODE, "ENODEN1")
1340 ENODENET = (TNODE, "ENODENET")
1341 ENODEOS = (TNODE, "ENODEOS")
1342 ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1343 ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1344 ENODERPC = (TNODE, "ENODERPC")
1345 ENODESSH = (TNODE, "ENODESSH")
1346 ENODEVERSION = (TNODE, "ENODEVERSION")
1347 ENODESETUP = (TNODE, "ENODESETUP")
1348 ENODETIME = (TNODE, "ENODETIME")
1350 ETYPE_FIELD = "code"
1351 ETYPE_ERROR = "ERROR"
1352 ETYPE_WARNING = "WARNING"
1354 class NodeImage(object):
1355 """A class representing the logical and physical status of a node.
1358 @ivar name: the node name to which this object refers
1359 @ivar volumes: a structure as returned from
1360 L{ganeti.backend.GetVolumeList} (runtime)
1361 @ivar instances: a list of running instances (runtime)
1362 @ivar pinst: list of configured primary instances (config)
1363 @ivar sinst: list of configured secondary instances (config)
1364 @ivar sbp: diction of {secondary-node: list of instances} of all peers
1365 of this node (config)
1366 @ivar mfree: free memory, as reported by hypervisor (runtime)
1367 @ivar dfree: free disk, as reported by the node (runtime)
1368 @ivar offline: the offline status (config)
1369 @type rpc_fail: boolean
1370 @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1371 not whether the individual keys were correct) (runtime)
1372 @type lvm_fail: boolean
1373 @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1374 @type hyp_fail: boolean
1375 @ivar hyp_fail: whether the RPC call didn't return the instance list
1376 @type ghost: boolean
1377 @ivar ghost: whether this is a known node or not (config)
1378 @type os_fail: boolean
1379 @ivar os_fail: whether the RPC call didn't return valid OS data
1381 @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1384 def __init__(self, offline=False, name=None):
1393 self.offline = offline
1394 self.rpc_fail = False
1395 self.lvm_fail = False
1396 self.hyp_fail = False
1398 self.os_fail = False
1401 def ExpandNames(self):
1402 self.needed_locks = {
1403 locking.LEVEL_NODE: locking.ALL_SET,
1404 locking.LEVEL_INSTANCE: locking.ALL_SET,
1406 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1408 def _Error(self, ecode, item, msg, *args, **kwargs):
1409 """Format an error message.
1411 Based on the opcode's error_codes parameter, either format a
1412 parseable error code, or a simpler error string.
1414 This must be called only from Exec and functions called from Exec.
1417 ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1419 # first complete the msg
1422 # then format the whole message
1423 if self.op.error_codes:
1424 msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1430 msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1431 # and finally report it via the feedback_fn
1432 self._feedback_fn(" - %s" % msg)
1434 def _ErrorIf(self, cond, *args, **kwargs):
1435 """Log an error message if the passed condition is True.
1438 cond = bool(cond) or self.op.debug_simulate_errors
1440 self._Error(*args, **kwargs)
1441 # do not mark the operation as failed for WARN cases only
1442 if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1443 self.bad = self.bad or cond
1445 def _VerifyNode(self, ninfo, nresult):
1446 """Perform some basic validation on data returned from a node.
1448 - check the result data structure is well formed and has all the
1450 - check ganeti version
1452 @type ninfo: L{objects.Node}
1453 @param ninfo: the node to check
1454 @param nresult: the results from the node
1456 @return: whether overall this call was successful (and we can expect
1457 reasonable values in the respose)
1461 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1463 # main result, nresult should be a non-empty dict
1464 test = not nresult or not isinstance(nresult, dict)
1465 _ErrorIf(test, self.ENODERPC, node,
1466 "unable to verify node: no data returned")
1470 # compares ganeti version
1471 local_version = constants.PROTOCOL_VERSION
1472 remote_version = nresult.get("version", None)
1473 test = not (remote_version and
1474 isinstance(remote_version, (list, tuple)) and
1475 len(remote_version) == 2)
1476 _ErrorIf(test, self.ENODERPC, node,
1477 "connection to node returned invalid data")
1481 test = local_version != remote_version[0]
1482 _ErrorIf(test, self.ENODEVERSION, node,
1483 "incompatible protocol versions: master %s,"
1484 " node %s", local_version, remote_version[0])
1488 # node seems compatible, we can actually try to look into its results
1490 # full package version
1491 self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1492 self.ENODEVERSION, node,
1493 "software version mismatch: master %s, node %s",
1494 constants.RELEASE_VERSION, remote_version[1],
1495 code=self.ETYPE_WARNING)
1497 hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1498 if isinstance(hyp_result, dict):
1499 for hv_name, hv_result in hyp_result.iteritems():
1500 test = hv_result is not None
1501 _ErrorIf(test, self.ENODEHV, node,
1502 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1505 test = nresult.get(constants.NV_NODESETUP,
1506 ["Missing NODESETUP results"])
1507 _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1512 def _VerifyNodeTime(self, ninfo, nresult,
1513 nvinfo_starttime, nvinfo_endtime):
1514 """Check the node time.
1516 @type ninfo: L{objects.Node}
1517 @param ninfo: the node to check
1518 @param nresult: the remote results for the node
1519 @param nvinfo_starttime: the start time of the RPC call
1520 @param nvinfo_endtime: the end time of the RPC call
1524 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1526 ntime = nresult.get(constants.NV_TIME, None)
1528 ntime_merged = utils.MergeTime(ntime)
1529 except (ValueError, TypeError):
1530 _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1533 if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1534 ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1535 elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1536 ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1540 _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1541 "Node time diverges by at least %s from master node time",
1544 def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1545 """Check the node time.
1547 @type ninfo: L{objects.Node}
1548 @param ninfo: the node to check
1549 @param nresult: the remote results for the node
1550 @param vg_name: the configured VG name
1557 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1559 # checks vg existence and size > 20G
1560 vglist = nresult.get(constants.NV_VGLIST, None)
1562 _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1564 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1565 constants.MIN_VG_SIZE)
1566 _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1569 pvlist = nresult.get(constants.NV_PVLIST, None)
1570 test = pvlist is None
1571 _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1573 # check that ':' is not present in PV names, since it's a
1574 # special character for lvcreate (denotes the range of PEs to
1576 for _, pvname, owner_vg in pvlist:
1577 test = ":" in pvname
1578 _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1579 " '%s' of VG '%s'", pvname, owner_vg)
1581 def _VerifyNodeNetwork(self, ninfo, nresult):
1582 """Check the node time.
1584 @type ninfo: L{objects.Node}
1585 @param ninfo: the node to check
1586 @param nresult: the remote results for the node
1590 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1592 test = constants.NV_NODELIST not in nresult
1593 _ErrorIf(test, self.ENODESSH, node,
1594 "node hasn't returned node ssh connectivity data")
1596 if nresult[constants.NV_NODELIST]:
1597 for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1598 _ErrorIf(True, self.ENODESSH, node,
1599 "ssh communication with node '%s': %s", a_node, a_msg)
1601 test = constants.NV_NODENETTEST not in nresult
1602 _ErrorIf(test, self.ENODENET, node,
1603 "node hasn't returned node tcp connectivity data")
1605 if nresult[constants.NV_NODENETTEST]:
1606 nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1608 _ErrorIf(True, self.ENODENET, node,
1609 "tcp communication with node '%s': %s",
1610 anode, nresult[constants.NV_NODENETTEST][anode])
1612 test = constants.NV_MASTERIP not in nresult
1613 _ErrorIf(test, self.ENODENET, node,
1614 "node hasn't returned node master IP reachability data")
1616 if not nresult[constants.NV_MASTERIP]:
1617 if node == self.master_node:
1618 msg = "the master node cannot reach the master IP (not configured?)"
1620 msg = "cannot reach the master IP"
1621 _ErrorIf(True, self.ENODENET, node, msg)
1624 def _VerifyInstance(self, instance, instanceconfig, node_image):
1625 """Verify an instance.
1627 This function checks to see if the required block devices are
1628 available on the instance's node.
1631 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1632 node_current = instanceconfig.primary_node
1634 node_vol_should = {}
1635 instanceconfig.MapLVsByNode(node_vol_should)
1637 for node in node_vol_should:
1638 n_img = node_image[node]
1639 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1640 # ignore missing volumes on offline or broken nodes
1642 for volume in node_vol_should[node]:
1643 test = volume not in n_img.volumes
1644 _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1645 "volume %s missing on node %s", volume, node)
1647 if instanceconfig.admin_up:
1648 pri_img = node_image[node_current]
1649 test = instance not in pri_img.instances and not pri_img.offline
1650 _ErrorIf(test, self.EINSTANCEDOWN, instance,
1651 "instance not running on its primary node %s",
1654 for node, n_img in node_image.items():
1655 if (not node == node_current):
1656 test = instance in n_img.instances
1657 _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1658 "instance should not run on node %s", node)
1660 def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1661 """Verify if there are any unknown volumes in the cluster.
1663 The .os, .swap and backup volumes are ignored. All other volumes are
1664 reported as unknown.
1666 @type reserved: L{ganeti.utils.FieldSet}
1667 @param reserved: a FieldSet of reserved volume names
1670 for node, n_img in node_image.items():
1671 if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1672 # skip non-healthy nodes
1674 for volume in n_img.volumes:
1675 test = ((node not in node_vol_should or
1676 volume not in node_vol_should[node]) and
1677 not reserved.Matches(volume))
1678 self._ErrorIf(test, self.ENODEORPHANLV, node,
1679 "volume %s is unknown", volume)
1681 def _VerifyOrphanInstances(self, instancelist, node_image):
1682 """Verify the list of running instances.
1684 This checks what instances are running but unknown to the cluster.
1687 for node, n_img in node_image.items():
1688 for o_inst in n_img.instances:
1689 test = o_inst not in instancelist
1690 self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1691 "instance %s on node %s should not exist", o_inst, node)
1693 def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1694 """Verify N+1 Memory Resilience.
1696 Check that if one single node dies we can still start all the
1697 instances it was primary for.
1700 for node, n_img in node_image.items():
1701 # This code checks that every node which is now listed as
1702 # secondary has enough memory to host all instances it is
1703 # supposed to should a single other node in the cluster fail.
1704 # FIXME: not ready for failover to an arbitrary node
1705 # FIXME: does not support file-backed instances
1706 # WARNING: we currently take into account down instances as well
1707 # as up ones, considering that even if they're down someone
1708 # might want to start them even in the event of a node failure.
1709 for prinode, instances in n_img.sbp.items():
1711 for instance in instances:
1712 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1713 if bep[constants.BE_AUTO_BALANCE]:
1714 needed_mem += bep[constants.BE_MEMORY]
1715 test = n_img.mfree < needed_mem
1716 self._ErrorIf(test, self.ENODEN1, node,
1717 "not enough memory on to accommodate"
1718 " failovers should peer node %s fail", prinode)
1720 def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1722 """Verifies and computes the node required file checksums.
1724 @type ninfo: L{objects.Node}
1725 @param ninfo: the node to check
1726 @param nresult: the remote results for the node
1727 @param file_list: required list of files
1728 @param local_cksum: dictionary of local files and their checksums
1729 @param master_files: list of files that only masters should have
1733 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1735 remote_cksum = nresult.get(constants.NV_FILELIST, None)
1736 test = not isinstance(remote_cksum, dict)
1737 _ErrorIf(test, self.ENODEFILECHECK, node,
1738 "node hasn't returned file checksum data")
1742 for file_name in file_list:
1743 node_is_mc = ninfo.master_candidate
1744 must_have = (file_name not in master_files) or node_is_mc
1746 test1 = file_name not in remote_cksum
1748 test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1750 test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1751 _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1752 "file '%s' missing", file_name)
1753 _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1754 "file '%s' has wrong checksum", file_name)
1755 # not candidate and this is not a must-have file
1756 _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1757 "file '%s' should not exist on non master"
1758 " candidates (and the file is outdated)", file_name)
1759 # all good, except non-master/non-must have combination
1760 _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1761 "file '%s' should not exist"
1762 " on non master candidates", file_name)
1764 def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1766 """Verifies and the node DRBD status.
1768 @type ninfo: L{objects.Node}
1769 @param ninfo: the node to check
1770 @param nresult: the remote results for the node
1771 @param instanceinfo: the dict of instances
1772 @param drbd_helper: the configured DRBD usermode helper
1773 @param drbd_map: the DRBD map as returned by
1774 L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1778 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1781 helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1782 test = (helper_result == None)
1783 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1784 "no drbd usermode helper returned")
1786 status, payload = helper_result
1788 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1789 "drbd usermode helper check unsuccessful: %s", payload)
1790 test = status and (payload != drbd_helper)
1791 _ErrorIf(test, self.ENODEDRBDHELPER, node,
1792 "wrong drbd usermode helper: %s", payload)
1794 # compute the DRBD minors
1796 for minor, instance in drbd_map[node].items():
1797 test = instance not in instanceinfo
1798 _ErrorIf(test, self.ECLUSTERCFG, None,
1799 "ghost instance '%s' in temporary DRBD map", instance)
1800 # ghost instance should not be running, but otherwise we
1801 # don't give double warnings (both ghost instance and
1802 # unallocated minor in use)
1804 node_drbd[minor] = (instance, False)
1806 instance = instanceinfo[instance]
1807 node_drbd[minor] = (instance.name, instance.admin_up)
1809 # and now check them
1810 used_minors = nresult.get(constants.NV_DRBDLIST, [])
1811 test = not isinstance(used_minors, (tuple, list))
1812 _ErrorIf(test, self.ENODEDRBD, node,
1813 "cannot parse drbd status file: %s", str(used_minors))
1815 # we cannot check drbd status
1818 for minor, (iname, must_exist) in node_drbd.items():
1819 test = minor not in used_minors and must_exist
1820 _ErrorIf(test, self.ENODEDRBD, node,
1821 "drbd minor %d of instance %s is not active", minor, iname)
1822 for minor in used_minors:
1823 test = minor not in node_drbd
1824 _ErrorIf(test, self.ENODEDRBD, node,
1825 "unallocated drbd minor %d is in use", minor)
1827 def _UpdateNodeOS(self, ninfo, nresult, nimg):
1828 """Builds the node OS structures.
1830 @type ninfo: L{objects.Node}
1831 @param ninfo: the node to check
1832 @param nresult: the remote results for the node
1833 @param nimg: the node image object
1837 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1839 remote_os = nresult.get(constants.NV_OSLIST, None)
1840 test = (not isinstance(remote_os, list) or
1841 not compat.all(isinstance(v, list) and len(v) == 7
1842 for v in remote_os))
1844 _ErrorIf(test, self.ENODEOS, node,
1845 "node hasn't returned valid OS data")
1854 for (name, os_path, status, diagnose,
1855 variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1857 if name not in os_dict:
1860 # parameters is a list of lists instead of list of tuples due to
1861 # JSON lacking a real tuple type, fix it:
1862 parameters = [tuple(v) for v in parameters]
1863 os_dict[name].append((os_path, status, diagnose,
1864 set(variants), set(parameters), set(api_ver)))
1866 nimg.oslist = os_dict
1868 def _VerifyNodeOS(self, ninfo, nimg, base):
1869 """Verifies the node OS list.
1871 @type ninfo: L{objects.Node}
1872 @param ninfo: the node to check
1873 @param nimg: the node image object
1874 @param base: the 'template' node we match against (e.g. from the master)
1878 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1880 assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1882 for os_name, os_data in nimg.oslist.items():
1883 assert os_data, "Empty OS status for OS %s?!" % os_name
1884 f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1885 _ErrorIf(not f_status, self.ENODEOS, node,
1886 "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1887 _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1888 "OS '%s' has multiple entries (first one shadows the rest): %s",
1889 os_name, utils.CommaJoin([v[0] for v in os_data]))
1890 # this will catched in backend too
1891 _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1892 and not f_var, self.ENODEOS, node,
1893 "OS %s with API at least %d does not declare any variant",
1894 os_name, constants.OS_API_V15)
1895 # comparisons with the 'base' image
1896 test = os_name not in base.oslist
1897 _ErrorIf(test, self.ENODEOS, node,
1898 "Extra OS %s not present on reference node (%s)",
1902 assert base.oslist[os_name], "Base node has empty OS status?"
1903 _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1905 # base OS is invalid, skipping
1907 for kind, a, b in [("API version", f_api, b_api),
1908 ("variants list", f_var, b_var),
1909 ("parameters", f_param, b_param)]:
1910 _ErrorIf(a != b, self.ENODEOS, node,
1911 "OS %s %s differs from reference node %s: %s vs. %s",
1912 kind, os_name, base.name,
1913 utils.CommaJoin(a), utils.CommaJoin(b))
1915 # check any missing OSes
1916 missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1917 _ErrorIf(missing, self.ENODEOS, node,
1918 "OSes present on reference node %s but missing on this node: %s",
1919 base.name, utils.CommaJoin(missing))
1921 def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1922 """Verifies and updates the node volume data.
1924 This function will update a L{NodeImage}'s internal structures
1925 with data from the remote call.
1927 @type ninfo: L{objects.Node}
1928 @param ninfo: the node to check
1929 @param nresult: the remote results for the node
1930 @param nimg: the node image object
1931 @param vg_name: the configured VG name
1935 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1937 nimg.lvm_fail = True
1938 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1941 elif isinstance(lvdata, basestring):
1942 _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1943 utils.SafeEncode(lvdata))
1944 elif not isinstance(lvdata, dict):
1945 _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1947 nimg.volumes = lvdata
1948 nimg.lvm_fail = False
1950 def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1951 """Verifies and updates the node instance list.
1953 If the listing was successful, then updates this node's instance
1954 list. Otherwise, it marks the RPC call as failed for the instance
1957 @type ninfo: L{objects.Node}
1958 @param ninfo: the node to check
1959 @param nresult: the remote results for the node
1960 @param nimg: the node image object
1963 idata = nresult.get(constants.NV_INSTANCELIST, None)
1964 test = not isinstance(idata, list)
1965 self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1966 " (instancelist): %s", utils.SafeEncode(str(idata)))
1968 nimg.hyp_fail = True
1970 nimg.instances = idata
1972 def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1973 """Verifies and computes a node information map
1975 @type ninfo: L{objects.Node}
1976 @param ninfo: the node to check
1977 @param nresult: the remote results for the node
1978 @param nimg: the node image object
1979 @param vg_name: the configured VG name
1983 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1985 # try to read free memory (from the hypervisor)
1986 hv_info = nresult.get(constants.NV_HVINFO, None)
1987 test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1988 _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1991 nimg.mfree = int(hv_info["memory_free"])
1992 except (ValueError, TypeError):
1993 _ErrorIf(True, self.ENODERPC, node,
1994 "node returned invalid nodeinfo, check hypervisor")
1996 # FIXME: devise a free space model for file based instances as well
1997 if vg_name is not None:
1998 test = (constants.NV_VGLIST not in nresult or
1999 vg_name not in nresult[constants.NV_VGLIST])
2000 _ErrorIf(test, self.ENODELVM, node,
2001 "node didn't return data for the volume group '%s'"
2002 " - it is either missing or broken", vg_name)
2005 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2006 except (ValueError, TypeError):
2007 _ErrorIf(True, self.ENODERPC, node,
2008 "node returned invalid LVM info, check LVM status")
2010 def BuildHooksEnv(self):
2013 Cluster-Verify hooks just ran in the post phase and their failure makes
2014 the output be logged in the verify output and the verification to fail.
2017 all_nodes = self.cfg.GetNodeList()
2019 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2021 for node in self.cfg.GetAllNodesInfo().values():
2022 env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2024 return env, [], all_nodes
2026 def Exec(self, feedback_fn):
2027 """Verify integrity of cluster, performing various test on nodes.
2031 _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2032 verbose = self.op.verbose
2033 self._feedback_fn = feedback_fn
2034 feedback_fn("* Verifying global settings")
2035 for msg in self.cfg.VerifyConfig():
2036 _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2038 # Check the cluster certificates
2039 for cert_filename in constants.ALL_CERT_FILES:
2040 (errcode, msg) = _VerifyCertificate(cert_filename)
2041 _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2043 vg_name = self.cfg.GetVGName()
2044 drbd_helper = self.cfg.GetDRBDHelper()
2045 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2046 cluster = self.cfg.GetClusterInfo()
2047 nodelist = utils.NiceSort(self.cfg.GetNodeList())
2048 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2049 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2050 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2051 for iname in instancelist)
2052 i_non_redundant = [] # Non redundant instances
2053 i_non_a_balanced = [] # Non auto-balanced instances
2054 n_offline = 0 # Count of offline nodes
2055 n_drained = 0 # Count of nodes being drained
2056 node_vol_should = {}
2058 # FIXME: verify OS list
2059 # do local checksums
2060 master_files = [constants.CLUSTER_CONF_FILE]
2061 master_node = self.master_node = self.cfg.GetMasterNode()
2062 master_ip = self.cfg.GetMasterIP()
2064 file_names = ssconf.SimpleStore().GetFileList()
2065 file_names.extend(constants.ALL_CERT_FILES)
2066 file_names.extend(master_files)
2067 if cluster.modify_etc_hosts:
2068 file_names.append(constants.ETC_HOSTS)
2070 local_checksums = utils.FingerprintFiles(file_names)
2072 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2073 node_verify_param = {
2074 constants.NV_FILELIST: file_names,
2075 constants.NV_NODELIST: [node.name for node in nodeinfo
2076 if not node.offline],
2077 constants.NV_HYPERVISOR: hypervisors,
2078 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2079 node.secondary_ip) for node in nodeinfo
2080 if not node.offline],
2081 constants.NV_INSTANCELIST: hypervisors,
2082 constants.NV_VERSION: None,
2083 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2084 constants.NV_NODESETUP: None,
2085 constants.NV_TIME: None,
2086 constants.NV_MASTERIP: (master_node, master_ip),
2087 constants.NV_OSLIST: None,
2090 if vg_name is not None:
2091 node_verify_param[constants.NV_VGLIST] = None
2092 node_verify_param[constants.NV_LVLIST] = vg_name
2093 node_verify_param[constants.NV_PVLIST] = [vg_name]
2094 node_verify_param[constants.NV_DRBDLIST] = None
2097 node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2099 # Build our expected cluster state
2100 node_image = dict((node.name, self.NodeImage(offline=node.offline,
2102 for node in nodeinfo)
2104 for instance in instancelist:
2105 inst_config = instanceinfo[instance]
2107 for nname in inst_config.all_nodes:
2108 if nname not in node_image:
2110 gnode = self.NodeImage(name=nname)
2112 node_image[nname] = gnode
2114 inst_config.MapLVsByNode(node_vol_should)
2116 pnode = inst_config.primary_node
2117 node_image[pnode].pinst.append(instance)
2119 for snode in inst_config.secondary_nodes:
2120 nimg = node_image[snode]
2121 nimg.sinst.append(instance)
2122 if pnode not in nimg.sbp:
2123 nimg.sbp[pnode] = []
2124 nimg.sbp[pnode].append(instance)
2126 # At this point, we have the in-memory data structures complete,
2127 # except for the runtime information, which we'll gather next
2129 # Due to the way our RPC system works, exact response times cannot be
2130 # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2131 # time before and after executing the request, we can at least have a time
2133 nvinfo_starttime = time.time()
2134 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2135 self.cfg.GetClusterName())
2136 nvinfo_endtime = time.time()
2138 all_drbd_map = self.cfg.ComputeDRBDMap()
2140 feedback_fn("* Verifying node status")
2144 for node_i in nodeinfo:
2146 nimg = node_image[node]
2150 feedback_fn("* Skipping offline node %s" % (node,))
2154 if node == master_node:
2156 elif node_i.master_candidate:
2157 ntype = "master candidate"
2158 elif node_i.drained:
2164 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2166 msg = all_nvinfo[node].fail_msg
2167 _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2169 nimg.rpc_fail = True
2172 nresult = all_nvinfo[node].payload
2174 nimg.call_ok = self._VerifyNode(node_i, nresult)
2175 self._VerifyNodeNetwork(node_i, nresult)
2176 self._VerifyNodeLVM(node_i, nresult, vg_name)
2177 self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2179 self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2181 self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2183 self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2184 self._UpdateNodeInstances(node_i, nresult, nimg)
2185 self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2186 self._UpdateNodeOS(node_i, nresult, nimg)
2187 if not nimg.os_fail:
2188 if refos_img is None:
2190 self._VerifyNodeOS(node_i, nimg, refos_img)
2192 feedback_fn("* Verifying instance status")
2193 for instance in instancelist:
2195 feedback_fn("* Verifying instance %s" % instance)
2196 inst_config = instanceinfo[instance]
2197 self._VerifyInstance(instance, inst_config, node_image)
2198 inst_nodes_offline = []
2200 pnode = inst_config.primary_node
2201 pnode_img = node_image[pnode]
2202 _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2203 self.ENODERPC, pnode, "instance %s, connection to"
2204 " primary node failed", instance)
2206 if pnode_img.offline:
2207 inst_nodes_offline.append(pnode)
2209 # If the instance is non-redundant we cannot survive losing its primary
2210 # node, so we are not N+1 compliant. On the other hand we have no disk
2211 # templates with more than one secondary so that situation is not well
2213 # FIXME: does not support file-backed instances
2214 if not inst_config.secondary_nodes:
2215 i_non_redundant.append(instance)
2216 _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2217 instance, "instance has multiple secondary nodes: %s",
2218 utils.CommaJoin(inst_config.secondary_nodes),
2219 code=self.ETYPE_WARNING)
2221 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2222 i_non_a_balanced.append(instance)
2224 for snode in inst_config.secondary_nodes:
2225 s_img = node_image[snode]
2226 _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2227 "instance %s, connection to secondary node failed", instance)
2230 inst_nodes_offline.append(snode)
2232 # warn that the instance lives on offline nodes
2233 _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2234 "instance lives on offline node(s) %s",
2235 utils.CommaJoin(inst_nodes_offline))
2236 # ... or ghost nodes
2237 for node in inst_config.all_nodes:
2238 _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2239 "instance lives on ghost node %s", node)
2241 feedback_fn("* Verifying orphan volumes")
2242 reserved = utils.FieldSet(*cluster.reserved_lvs)
2243 self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2245 feedback_fn("* Verifying orphan instances")
2246 self._VerifyOrphanInstances(instancelist, node_image)
2248 if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2249 feedback_fn("* Verifying N+1 Memory redundancy")
2250 self._VerifyNPlusOneMemory(node_image, instanceinfo)
2252 feedback_fn("* Other Notes")
2254 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
2255 % len(i_non_redundant))
2257 if i_non_a_balanced:
2258 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
2259 % len(i_non_a_balanced))
2262 feedback_fn(" - NOTICE: %d offline node(s) found." % n_offline)
2265 feedback_fn(" - NOTICE: %d drained node(s) found." % n_drained)
2269 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2270 """Analyze the post-hooks' result
2272 This method analyses the hook result, handles it, and sends some
2273 nicely-formatted feedback back to the user.
2275 @param phase: one of L{constants.HOOKS_PHASE_POST} or
2276 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2277 @param hooks_results: the results of the multi-node hooks rpc call
2278 @param feedback_fn: function used send feedback back to the caller
2279 @param lu_result: previous Exec result
2280 @return: the new Exec result, based on the previous result
2284 # We only really run POST phase hooks, and are only interested in
2286 if phase == constants.HOOKS_PHASE_POST:
2287 # Used to change hooks' output to proper indentation
2288 indent_re = re.compile('^', re.M)
2289 feedback_fn("* Hooks Results")
2290 assert hooks_results, "invalid result from hooks"
2292 for node_name in hooks_results:
2293 res = hooks_results[node_name]
2295 test = msg and not res.offline
2296 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2297 "Communication failure in hooks execution: %s", msg)
2298 if res.offline or msg:
2299 # No need to investigate payload if node is offline or gave an error.
2300 # override manually lu_result here as _ErrorIf only
2301 # overrides self.bad
2304 for script, hkr, output in res.payload:
2305 test = hkr == constants.HKR_FAIL
2306 self._ErrorIf(test, self.ENODEHOOKS, node_name,
2307 "Script %s failed, output:", script)
2309 output = indent_re.sub(' ', output)
2310 feedback_fn("%s" % output)
2316 class LUVerifyDisks(NoHooksLU):
2317 """Verifies the cluster disks status.
2322 def ExpandNames(self):
2323 self.needed_locks = {
2324 locking.LEVEL_NODE: locking.ALL_SET,
2325 locking.LEVEL_INSTANCE: locking.ALL_SET,
2327 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2329 def Exec(self, feedback_fn):
2330 """Verify integrity of cluster disks.
2332 @rtype: tuple of three items
2333 @return: a tuple of (dict of node-to-node_error, list of instances
2334 which need activate-disks, dict of instance: (node, volume) for
2338 result = res_nodes, res_instances, res_missing = {}, [], {}
2340 vg_name = self.cfg.GetVGName()
2341 nodes = utils.NiceSort(self.cfg.GetNodeList())
2342 instances = [self.cfg.GetInstanceInfo(name)
2343 for name in self.cfg.GetInstanceList()]
2346 for inst in instances:
2348 if (not inst.admin_up or
2349 inst.disk_template not in constants.DTS_NET_MIRROR):
2351 inst.MapLVsByNode(inst_lvs)
2352 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2353 for node, vol_list in inst_lvs.iteritems():
2354 for vol in vol_list:
2355 nv_dict[(node, vol)] = inst
2360 node_lvs = self.rpc.call_lv_list(nodes, vg_name)
2364 node_res = node_lvs[node]
2365 if node_res.offline:
2367 msg = node_res.fail_msg
2369 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2370 res_nodes[node] = msg
2373 lvs = node_res.payload
2374 for lv_name, (_, _, lv_online) in lvs.items():
2375 inst = nv_dict.pop((node, lv_name), None)
2376 if (not lv_online and inst is not None
2377 and inst.name not in res_instances):
2378 res_instances.append(inst.name)
2380 # any leftover items in nv_dict are missing LVs, let's arrange the
2382 for key, inst in nv_dict.iteritems():
2383 if inst.name not in res_missing:
2384 res_missing[inst.name] = []
2385 res_missing[inst.name].append(key)
2390 class LURepairDiskSizes(NoHooksLU):
2391 """Verifies the cluster disks sizes.
2394 _OP_PARAMS = [("instances", _EmptyList, _TListOf(_TNonEmptyString))]
2397 def ExpandNames(self):
2398 if self.op.instances:
2399 self.wanted_names = []
2400 for name in self.op.instances:
2401 full_name = _ExpandInstanceName(self.cfg, name)
2402 self.wanted_names.append(full_name)
2403 self.needed_locks = {
2404 locking.LEVEL_NODE: [],
2405 locking.LEVEL_INSTANCE: self.wanted_names,
2407 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2409 self.wanted_names = None
2410 self.needed_locks = {
2411 locking.LEVEL_NODE: locking.ALL_SET,
2412 locking.LEVEL_INSTANCE: locking.ALL_SET,
2414 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2416 def DeclareLocks(self, level):
2417 if level == locking.LEVEL_NODE and self.wanted_names is not None:
2418 self._LockInstancesNodes(primary_only=True)
2420 def CheckPrereq(self):
2421 """Check prerequisites.
2423 This only checks the optional instance list against the existing names.
2426 if self.wanted_names is None:
2427 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2429 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2430 in self.wanted_names]
2432 def _EnsureChildSizes(self, disk):
2433 """Ensure children of the disk have the needed disk size.
2435 This is valid mainly for DRBD8 and fixes an issue where the
2436 children have smaller disk size.
2438 @param disk: an L{ganeti.objects.Disk} object
2441 if disk.dev_type == constants.LD_DRBD8:
2442 assert disk.children, "Empty children for DRBD8?"
2443 fchild = disk.children[0]
2444 mismatch = fchild.size < disk.size
2446 self.LogInfo("Child disk has size %d, parent %d, fixing",
2447 fchild.size, disk.size)
2448 fchild.size = disk.size
2450 # and we recurse on this child only, not on the metadev
2451 return self._EnsureChildSizes(fchild) or mismatch
2455 def Exec(self, feedback_fn):
2456 """Verify the size of cluster disks.
2459 # TODO: check child disks too
2460 # TODO: check differences in size between primary/secondary nodes
2462 for instance in self.wanted_instances:
2463 pnode = instance.primary_node
2464 if pnode not in per_node_disks:
2465 per_node_disks[pnode] = []
2466 for idx, disk in enumerate(instance.disks):
2467 per_node_disks[pnode].append((instance, idx, disk))
2470 for node, dskl in per_node_disks.items():
2471 newl = [v[2].Copy() for v in dskl]
2473 self.cfg.SetDiskID(dsk, node)
2474 result = self.rpc.call_blockdev_getsizes(node, newl)
2476 self.LogWarning("Failure in blockdev_getsizes call to node"
2477 " %s, ignoring", node)
2479 if len(result.data) != len(dskl):
2480 self.LogWarning("Invalid result from node %s, ignoring node results",
2483 for ((instance, idx, disk), size) in zip(dskl, result.data):
2485 self.LogWarning("Disk %d of instance %s did not return size"
2486 " information, ignoring", idx, instance.name)
2488 if not isinstance(size, (int, long)):
2489 self.LogWarning("Disk %d of instance %s did not return valid"
2490 " size information, ignoring", idx, instance.name)
2493 if size != disk.size:
2494 self.LogInfo("Disk %d of instance %s has mismatched size,"
2495 " correcting: recorded %d, actual %d", idx,
2496 instance.name, disk.size, size)
2498 self.cfg.Update(instance, feedback_fn)
2499 changed.append((instance.name, idx, size))
2500 if self._EnsureChildSizes(disk):
2501 self.cfg.Update(instance, feedback_fn)
2502 changed.append((instance.name, idx, disk.size))
2506 class LURenameCluster(LogicalUnit):
2507 """Rename the cluster.
2510 HPATH = "cluster-rename"
2511 HTYPE = constants.HTYPE_CLUSTER
2512 _OP_PARAMS = [("name", _NoDefault, _TNonEmptyString)]
2514 def BuildHooksEnv(self):
2519 "OP_TARGET": self.cfg.GetClusterName(),
2520 "NEW_NAME": self.op.name,
2522 mn = self.cfg.GetMasterNode()
2523 all_nodes = self.cfg.GetNodeList()
2524 return env, [mn], all_nodes
2526 def CheckPrereq(self):
2527 """Verify that the passed name is a valid one.
2530 hostname = netutils.GetHostInfo(self.op.name)
2532 new_name = hostname.name
2533 self.ip = new_ip = hostname.ip
2534 old_name = self.cfg.GetClusterName()
2535 old_ip = self.cfg.GetMasterIP()
2536 if new_name == old_name and new_ip == old_ip:
2537 raise errors.OpPrereqError("Neither the name nor the IP address of the"
2538 " cluster has changed",
2540 if new_ip != old_ip:
2541 if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2542 raise errors.OpPrereqError("The given cluster IP address (%s) is"
2543 " reachable on the network. Aborting." %
2544 new_ip, errors.ECODE_NOTUNIQUE)
2546 self.op.name = new_name
2548 def Exec(self, feedback_fn):
2549 """Rename the cluster.
2552 clustername = self.op.name
2555 # shutdown the master IP
2556 master = self.cfg.GetMasterNode()
2557 result = self.rpc.call_node_stop_master(master, False)
2558 result.Raise("Could not disable the master role")
2561 cluster = self.cfg.GetClusterInfo()
2562 cluster.cluster_name = clustername
2563 cluster.master_ip = ip
2564 self.cfg.Update(cluster, feedback_fn)
2566 # update the known hosts file
2567 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2568 node_list = self.cfg.GetNodeList()
2570 node_list.remove(master)
2573 result = self.rpc.call_upload_file(node_list,
2574 constants.SSH_KNOWN_HOSTS_FILE)
2575 for to_node, to_result in result.iteritems():
2576 msg = to_result.fail_msg
2578 msg = ("Copy of file %s to node %s failed: %s" %
2579 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
2580 self.proc.LogWarning(msg)
2583 result = self.rpc.call_node_start_master(master, False, False)
2584 msg = result.fail_msg
2586 self.LogWarning("Could not re-enable the master role on"
2587 " the master, please restart manually: %s", msg)
2592 class LUSetClusterParams(LogicalUnit):
2593 """Change the parameters of the cluster.
2596 HPATH = "cluster-modify"
2597 HTYPE = constants.HTYPE_CLUSTER
2599 ("vg_name", None, _TMaybeString),
2600 ("enabled_hypervisors", None,
2601 _TOr(_TAnd(_TListOf(_TElemOf(constants.HYPER_TYPES)), _TTrue), _TNone)),
2602 ("hvparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2603 ("beparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2604 ("os_hvp", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2605 ("osparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2606 ("candidate_pool_size", None, _TOr(_TStrictPositiveInt, _TNone)),
2607 ("uid_pool", None, _NoType),
2608 ("add_uids", None, _NoType),
2609 ("remove_uids", None, _NoType),
2610 ("maintain_node_health", None, _TMaybeBool),
2611 ("nicparams", None, _TOr(_TDict, _TNone)),
2612 ("drbd_helper", None, _TOr(_TString, _TNone)),
2613 ("default_iallocator", None, _TMaybeString),
2614 ("reserved_lvs", None, _TOr(_TListOf(_TNonEmptyString), _TNone)),
2618 def CheckArguments(self):
2622 if self.op.uid_pool:
2623 uidpool.CheckUidPool(self.op.uid_pool)
2625 if self.op.add_uids:
2626 uidpool.CheckUidPool(self.op.add_uids)
2628 if self.op.remove_uids:
2629 uidpool.CheckUidPool(self.op.remove_uids)
2631 def ExpandNames(self):
2632 # FIXME: in the future maybe other cluster params won't require checking on
2633 # all nodes to be modified.
2634 self.needed_locks = {
2635 locking.LEVEL_NODE: locking.ALL_SET,
2637 self.share_locks[locking.LEVEL_NODE] = 1
2639 def BuildHooksEnv(self):
2644 "OP_TARGET": self.cfg.GetClusterName(),
2645 "NEW_VG_NAME": self.op.vg_name,
2647 mn = self.cfg.GetMasterNode()
2648 return env, [mn], [mn]
2650 def CheckPrereq(self):
2651 """Check prerequisites.
2653 This checks whether the given params don't conflict and
2654 if the given volume group is valid.
2657 if self.op.vg_name is not None and not self.op.vg_name:
2658 if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2659 raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2660 " instances exist", errors.ECODE_INVAL)
2662 if self.op.drbd_helper is not None and not self.op.drbd_helper:
2663 if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2664 raise errors.OpPrereqError("Cannot disable drbd helper while"
2665 " drbd-based instances exist",
2668 node_list = self.acquired_locks[locking.LEVEL_NODE]
2670 # if vg_name not None, checks given volume group on all nodes
2672 vglist = self.rpc.call_vg_list(node_list)
2673 for node in node_list:
2674 msg = vglist[node].fail_msg
2676 # ignoring down node
2677 self.LogWarning("Error while gathering data on node %s"
2678 " (ignoring node): %s", node, msg)
2680 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2682 constants.MIN_VG_SIZE)
2684 raise errors.OpPrereqError("Error on node '%s': %s" %
2685 (node, vgstatus), errors.ECODE_ENVIRON)
2687 if self.op.drbd_helper:
2688 # checks given drbd helper on all nodes
2689 helpers = self.rpc.call_drbd_helper(node_list)
2690 for node in node_list:
2691 ninfo = self.cfg.GetNodeInfo(node)
2693 self.LogInfo("Not checking drbd helper on offline node %s", node)
2695 msg = helpers[node].fail_msg
2697 raise errors.OpPrereqError("Error checking drbd helper on node"
2698 " '%s': %s" % (node, msg),
2699 errors.ECODE_ENVIRON)
2700 node_helper = helpers[node].payload
2701 if node_helper != self.op.drbd_helper:
2702 raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2703 (node, node_helper), errors.ECODE_ENVIRON)
2705 self.cluster = cluster = self.cfg.GetClusterInfo()
2706 # validate params changes
2707 if self.op.beparams:
2708 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2709 self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2711 if self.op.nicparams:
2712 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2713 self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2714 objects.NIC.CheckParameterSyntax(self.new_nicparams)
2717 # check all instances for consistency
2718 for instance in self.cfg.GetAllInstancesInfo().values():
2719 for nic_idx, nic in enumerate(instance.nics):
2720 params_copy = copy.deepcopy(nic.nicparams)
2721 params_filled = objects.FillDict(self.new_nicparams, params_copy)
2723 # check parameter syntax
2725 objects.NIC.CheckParameterSyntax(params_filled)
2726 except errors.ConfigurationError, err:
2727 nic_errors.append("Instance %s, nic/%d: %s" %
2728 (instance.name, nic_idx, err))
2730 # if we're moving instances to routed, check that they have an ip
2731 target_mode = params_filled[constants.NIC_MODE]
2732 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2733 nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2734 (instance.name, nic_idx))
2736 raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2737 "\n".join(nic_errors))
2739 # hypervisor list/parameters
2740 self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2741 if self.op.hvparams:
2742 for hv_name, hv_dict in self.op.hvparams.items():
2743 if hv_name not in self.new_hvparams:
2744 self.new_hvparams[hv_name] = hv_dict
2746 self.new_hvparams[hv_name].update(hv_dict)
2748 # os hypervisor parameters
2749 self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2751 for os_name, hvs in self.op.os_hvp.items():
2752 if os_name not in self.new_os_hvp:
2753 self.new_os_hvp[os_name] = hvs
2755 for hv_name, hv_dict in hvs.items():
2756 if hv_name not in self.new_os_hvp[os_name]:
2757 self.new_os_hvp[os_name][hv_name] = hv_dict
2759 self.new_os_hvp[os_name][hv_name].update(hv_dict)
2762 self.new_osp = objects.FillDict(cluster.osparams, {})
2763 if self.op.osparams:
2764 for os_name, osp in self.op.osparams.items():
2765 if os_name not in self.new_osp:
2766 self.new_osp[os_name] = {}
2768 self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2771 if not self.new_osp[os_name]:
2772 # we removed all parameters
2773 del self.new_osp[os_name]
2775 # check the parameter validity (remote check)
2776 _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2777 os_name, self.new_osp[os_name])
2779 # changes to the hypervisor list
2780 if self.op.enabled_hypervisors is not None:
2781 self.hv_list = self.op.enabled_hypervisors
2782 for hv in self.hv_list:
2783 # if the hypervisor doesn't already exist in the cluster
2784 # hvparams, we initialize it to empty, and then (in both
2785 # cases) we make sure to fill the defaults, as we might not
2786 # have a complete defaults list if the hypervisor wasn't
2788 if hv not in new_hvp:
2790 new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2791 utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2793 self.hv_list = cluster.enabled_hypervisors
2795 if self.op.hvparams or self.op.enabled_hypervisors is not None:
2796 # either the enabled list has changed, or the parameters have, validate
2797 for hv_name, hv_params in self.new_hvparams.items():
2798 if ((self.op.hvparams and hv_name in self.op.hvparams) or
2799 (self.op.enabled_hypervisors and
2800 hv_name in self.op.enabled_hypervisors)):
2801 # either this is a new hypervisor, or its parameters have changed
2802 hv_class = hypervisor.GetHypervisor(hv_name)
2803 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2804 hv_class.CheckParameterSyntax(hv_params)
2805 _CheckHVParams(self, node_list, hv_name, hv_params)
2808 # no need to check any newly-enabled hypervisors, since the
2809 # defaults have already been checked in the above code-block
2810 for os_name, os_hvp in self.new_os_hvp.items():
2811 for hv_name, hv_params in os_hvp.items():
2812 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2813 # we need to fill in the new os_hvp on top of the actual hv_p
2814 cluster_defaults = self.new_hvparams.get(hv_name, {})
2815 new_osp = objects.FillDict(cluster_defaults, hv_params)
2816 hv_class = hypervisor.GetHypervisor(hv_name)
2817 hv_class.CheckParameterSyntax(new_osp)
2818 _CheckHVParams(self, node_list, hv_name, new_osp)
2820 if self.op.default_iallocator:
2821 alloc_script = utils.FindFile(self.op.default_iallocator,
2822 constants.IALLOCATOR_SEARCH_PATH,
2824 if alloc_script is None:
2825 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2826 " specified" % self.op.default_iallocator,
2829 def Exec(self, feedback_fn):
2830 """Change the parameters of the cluster.
2833 if self.op.vg_name is not None:
2834 new_volume = self.op.vg_name
2837 if new_volume != self.cfg.GetVGName():
2838 self.cfg.SetVGName(new_volume)
2840 feedback_fn("Cluster LVM configuration already in desired"
2841 " state, not changing")
2842 if self.op.drbd_helper is not None:
2843 new_helper = self.op.drbd_helper
2846 if new_helper != self.cfg.GetDRBDHelper():
2847 self.cfg.SetDRBDHelper(new_helper)
2849 feedback_fn("Cluster DRBD helper already in desired state,"
2851 if self.op.hvparams:
2852 self.cluster.hvparams = self.new_hvparams
2854 self.cluster.os_hvp = self.new_os_hvp
2855 if self.op.enabled_hypervisors is not None:
2856 self.cluster.hvparams = self.new_hvparams
2857 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2858 if self.op.beparams:
2859 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2860 if self.op.nicparams:
2861 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2862 if self.op.osparams:
2863 self.cluster.osparams = self.new_osp
2865 if self.op.candidate_pool_size is not None:
2866 self.cluster.candidate_pool_size = self.op.candidate_pool_size
2867 # we need to update the pool size here, otherwise the save will fail
2868 _AdjustCandidatePool(self, [])
2870 if self.op.maintain_node_health is not None:
2871 self.cluster.maintain_node_health = self.op.maintain_node_health
2873 if self.op.add_uids is not None:
2874 uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2876 if self.op.remove_uids is not None:
2877 uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2879 if self.op.uid_pool is not None:
2880 self.cluster.uid_pool = self.op.uid_pool
2882 if self.op.default_iallocator is not None:
2883 self.cluster.default_iallocator = self.op.default_iallocator
2885 if self.op.reserved_lvs is not None:
2886 self.cluster.reserved_lvs = self.op.reserved_lvs
2888 self.cfg.Update(self.cluster, feedback_fn)
2891 def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2892 """Distribute additional files which are part of the cluster configuration.
2894 ConfigWriter takes care of distributing the config and ssconf files, but
2895 there are more files which should be distributed to all nodes. This function
2896 makes sure those are copied.
2898 @param lu: calling logical unit
2899 @param additional_nodes: list of nodes not in the config to distribute to
2902 # 1. Gather target nodes
2903 myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2904 dist_nodes = lu.cfg.GetOnlineNodeList()
2905 if additional_nodes is not None:
2906 dist_nodes.extend(additional_nodes)
2907 if myself.name in dist_nodes:
2908 dist_nodes.remove(myself.name)
2910 # 2. Gather files to distribute
2911 dist_files = set([constants.ETC_HOSTS,
2912 constants.SSH_KNOWN_HOSTS_FILE,
2913 constants.RAPI_CERT_FILE,
2914 constants.RAPI_USERS_FILE,
2915 constants.CONFD_HMAC_KEY,
2916 constants.CLUSTER_DOMAIN_SECRET_FILE,
2919 enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2920 for hv_name in enabled_hypervisors:
2921 hv_class = hypervisor.GetHypervisor(hv_name)
2922 dist_files.update(hv_class.GetAncillaryFiles())
2924 # 3. Perform the files upload
2925 for fname in dist_files:
2926 if os.path.exists(fname):
2927 result = lu.rpc.call_upload_file(dist_nodes, fname)
2928 for to_node, to_result in result.items():
2929 msg = to_result.fail_msg
2931 msg = ("Copy of file %s to node %s failed: %s" %
2932 (fname, to_node, msg))
2933 lu.proc.LogWarning(msg)
2936 class LURedistributeConfig(NoHooksLU):
2937 """Force the redistribution of cluster configuration.
2939 This is a very simple LU.
2944 def ExpandNames(self):
2945 self.needed_locks = {
2946 locking.LEVEL_NODE: locking.ALL_SET,
2948 self.share_locks[locking.LEVEL_NODE] = 1
2950 def Exec(self, feedback_fn):
2951 """Redistribute the configuration.
2954 self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2955 _RedistributeAncillaryFiles(self)
2958 def _WaitForSync(lu, instance, disks=None, oneshot=False):
2959 """Sleep and poll for an instance's disk to sync.
2962 if not instance.disks or disks is not None and not disks:
2965 disks = _ExpandCheckDisks(instance, disks)
2968 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2970 node = instance.primary_node
2973 lu.cfg.SetDiskID(dev, node)
2975 # TODO: Convert to utils.Retry
2978 degr_retries = 10 # in seconds, as we sleep 1 second each time
2982 cumul_degraded = False
2983 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
2984 msg = rstats.fail_msg
2986 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2989 raise errors.RemoteError("Can't contact node %s for mirror data,"
2990 " aborting." % node)
2993 rstats = rstats.payload
2995 for i, mstat in enumerate(rstats):
2997 lu.LogWarning("Can't compute data for node %s/%s",
2998 node, disks[i].iv_name)
3001 cumul_degraded = (cumul_degraded or
3002 (mstat.is_degraded and mstat.sync_percent is None))
3003 if mstat.sync_percent is not None:
3005 if mstat.estimated_time is not None:
3006 rem_time = ("%s remaining (estimated)" %
3007 utils.FormatSeconds(mstat.estimated_time))
3008 max_time = mstat.estimated_time
3010 rem_time = "no time estimate"
3011 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3012 (disks[i].iv_name, mstat.sync_percent, rem_time))
3014 # if we're done but degraded, let's do a few small retries, to
3015 # make sure we see a stable and not transient situation; therefore
3016 # we force restart of the loop
3017 if (done or oneshot) and cumul_degraded and degr_retries > 0:
3018 logging.info("Degraded disks found, %d retries left", degr_retries)
3026 time.sleep(min(60, max_time))
3029 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3030 return not cumul_degraded
3033 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3034 """Check that mirrors are not degraded.
3036 The ldisk parameter, if True, will change the test from the
3037 is_degraded attribute (which represents overall non-ok status for
3038 the device(s)) to the ldisk (representing the local storage status).
3041 lu.cfg.SetDiskID(dev, node)
3045 if on_primary or dev.AssembleOnSecondary():
3046 rstats = lu.rpc.call_blockdev_find(node, dev)
3047 msg = rstats.fail_msg
3049 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3051 elif not rstats.payload:
3052 lu.LogWarning("Can't find disk on node %s", node)
3056 result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3058 result = result and not rstats.payload.is_degraded
3061 for child in dev.children:
3062 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3067 class LUDiagnoseOS(NoHooksLU):
3068 """Logical unit for OS diagnose/query.
3073 ("names", _EmptyList, _TListOf(_TNonEmptyString)),
3076 _FIELDS_STATIC = utils.FieldSet()
3077 _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants",
3078 "parameters", "api_versions")
3080 def CheckArguments(self):
3082 raise errors.OpPrereqError("Selective OS query not supported",
3085 _CheckOutputFields(static=self._FIELDS_STATIC,
3086 dynamic=self._FIELDS_DYNAMIC,
3087 selected=self.op.output_fields)
3089 def ExpandNames(self):
3090 # Lock all nodes, in shared mode
3091 # Temporary removal of locks, should be reverted later
3092 # TODO: reintroduce locks when they are lighter-weight
3093 self.needed_locks = {}
3094 #self.share_locks[locking.LEVEL_NODE] = 1
3095 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3098 def _DiagnoseByOS(rlist):
3099 """Remaps a per-node return list into an a per-os per-node dictionary
3101 @param rlist: a map with node names as keys and OS objects as values
3104 @return: a dictionary with osnames as keys and as value another
3105 map, with nodes as keys and tuples of (path, status, diagnose,
3106 variants, parameters, api_versions) as values, eg::
3108 {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3109 (/srv/..., False, "invalid api")],
3110 "node2": [(/srv/..., True, "", [], [])]}
3115 # we build here the list of nodes that didn't fail the RPC (at RPC
3116 # level), so that nodes with a non-responding node daemon don't
3117 # make all OSes invalid
3118 good_nodes = [node_name for node_name in rlist
3119 if not rlist[node_name].fail_msg]
3120 for node_name, nr in rlist.items():
3121 if nr.fail_msg or not nr.payload:
3123 for (name, path, status, diagnose, variants,
3124 params, api_versions) in nr.payload:
3125 if name not in all_os:
3126 # build a list of nodes for this os containing empty lists
3127 # for each node in node_list
3129 for nname in good_nodes:
3130 all_os[name][nname] = []
3131 # convert params from [name, help] to (name, help)
3132 params = [tuple(v) for v in params]
3133 all_os[name][node_name].append((path, status, diagnose,
3134 variants, params, api_versions))
3137 def Exec(self, feedback_fn):
3138 """Compute the list of OSes.
3141 valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
3142 node_data = self.rpc.call_os_diagnose(valid_nodes)
3143 pol = self._DiagnoseByOS(node_data)
3146 for os_name, os_data in pol.items():
3149 (variants, params, api_versions) = null_state = (set(), set(), set())
3150 for idx, osl in enumerate(os_data.values()):
3151 valid = bool(valid and osl and osl[0][1])
3153 (variants, params, api_versions) = null_state
3155 node_variants, node_params, node_api = osl[0][3:6]
3156 if idx == 0: # first entry
3157 variants = set(node_variants)
3158 params = set(node_params)
3159 api_versions = set(node_api)
3160 else: # keep consistency
3161 variants.intersection_update(node_variants)
3162 params.intersection_update(node_params)
3163 api_versions.intersection_update(node_api)
3165 for field in self.op.output_fields:
3168 elif field == "valid":
3170 elif field == "node_status":
3171 # this is just a copy of the dict
3173 for node_name, nos_list in os_data.items():
3174 val[node_name] = nos_list
3175 elif field == "variants":
3176 val = list(variants)
3177 elif field == "parameters":
3179 elif field == "api_versions":
3180 val = list(api_versions)
3182 raise errors.ParameterError(field)
3189 class LURemoveNode(LogicalUnit):
3190 """Logical unit for removing a node.
3193 HPATH = "node-remove"
3194 HTYPE = constants.HTYPE_NODE
3199 def BuildHooksEnv(self):
3202 This doesn't run on the target node in the pre phase as a failed
3203 node would then be impossible to remove.
3207 "OP_TARGET": self.op.node_name,
3208 "NODE_NAME": self.op.node_name,
3210 all_nodes = self.cfg.GetNodeList()
3212 all_nodes.remove(self.op.node_name)
3214 logging.warning("Node %s which is about to be removed not found"
3215 " in the all nodes list", self.op.node_name)
3216 return env, all_nodes, all_nodes
3218 def CheckPrereq(self):
3219 """Check prerequisites.
3222 - the node exists in the configuration
3223 - it does not have primary or secondary instances
3224 - it's not the master
3226 Any errors are signaled by raising errors.OpPrereqError.
3229 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3230 node = self.cfg.GetNodeInfo(self.op.node_name)
3231 assert node is not None
3233 instance_list = self.cfg.GetInstanceList()
3235 masternode = self.cfg.GetMasterNode()
3236 if node.name == masternode:
3237 raise errors.OpPrereqError("Node is the master node,"
3238 " you need to failover first.",
3241 for instance_name in instance_list:
3242 instance = self.cfg.GetInstanceInfo(instance_name)
3243 if node.name in instance.all_nodes:
3244 raise errors.OpPrereqError("Instance %s is still running on the node,"
3245 " please remove first." % instance_name,
3247 self.op.node_name = node.name
3250 def Exec(self, feedback_fn):
3251 """Removes the node from the cluster.
3255 logging.info("Stopping the node daemon and removing configs from node %s",
3258 modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3260 # Promote nodes to master candidate as needed
3261 _AdjustCandidatePool(self, exceptions=[node.name])
3262 self.context.RemoveNode(node.name)
3264 # Run post hooks on the node before it's removed
3265 hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3267 hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3269 # pylint: disable-msg=W0702
3270 self.LogWarning("Errors occurred running hooks on %s" % node.name)
3272 result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3273 msg = result.fail_msg
3275 self.LogWarning("Errors encountered on the remote node while leaving"
3276 " the cluster: %s", msg)
3278 # Remove node from our /etc/hosts
3279 if self.cfg.GetClusterInfo().modify_etc_hosts:
3280 # FIXME: this should be done via an rpc call to node daemon
3281 utils.RemoveHostFromEtcHosts(node.name)
3282 _RedistributeAncillaryFiles(self)
3285 class LUQueryNodes(NoHooksLU):
3286 """Logical unit for querying nodes.
3289 # pylint: disable-msg=W0142
3292 ("names", _EmptyList, _TListOf(_TNonEmptyString)),
3293 ("use_locking", False, _TBool),
3297 _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
3298 "master_candidate", "offline", "drained"]
3300 _FIELDS_DYNAMIC = utils.FieldSet(
3302 "mtotal", "mnode", "mfree",
3304 "ctotal", "cnodes", "csockets",
3307 _FIELDS_STATIC = utils.FieldSet(*[
3308 "pinst_cnt", "sinst_cnt",
3309 "pinst_list", "sinst_list",
3310 "pip", "sip", "tags",
3312 "role"] + _SIMPLE_FIELDS
3315 def CheckArguments(self):
3316 _CheckOutputFields(static=self._FIELDS_STATIC,
3317 dynamic=self._FIELDS_DYNAMIC,
3318 selected=self.op.output_fields)
3320 def ExpandNames(self):
3321 self.needed_locks = {}
3322 self.share_locks[locking.LEVEL_NODE] = 1
3325 self.wanted = _GetWantedNodes(self, self.op.names)
3327 self.wanted = locking.ALL_SET
3329 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3330 self.do_locking = self.do_node_query and self.op.use_locking
3332 # if we don't request only static fields, we need to lock the nodes
3333 self.needed_locks[locking.LEVEL_NODE] = self.wanted
3335 def Exec(self, feedback_fn):
3336 """Computes the list of nodes and their attributes.
3339 all_info = self.cfg.GetAllNodesInfo()
3341 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3342 elif self.wanted != locking.ALL_SET:
3343 nodenames = self.wanted
3344 missing = set(nodenames).difference(all_info.keys())
3346 raise errors.OpExecError(
3347 "Some nodes were removed before retrieving their data: %s" % missing)
3349 nodenames = all_info.keys()
3351 nodenames = utils.NiceSort(nodenames)
3352 nodelist = [all_info[name] for name in nodenames]
3354 # begin data gathering
3356 if self.do_node_query:
3358 node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3359 self.cfg.GetHypervisorType())
3360 for name in nodenames:
3361 nodeinfo = node_data[name]
3362 if not nodeinfo.fail_msg and nodeinfo.payload:
3363 nodeinfo = nodeinfo.payload
3364 fn = utils.TryConvert
3366 "mtotal": fn(int, nodeinfo.get('memory_total', None)),
3367 "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
3368 "mfree": fn(int, nodeinfo.get('memory_free', None)),
3369 "dtotal": fn(int, nodeinfo.get('vg_size', None)),
3370 "dfree": fn(int, nodeinfo.get('vg_free', None)),
3371 "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
3372 "bootid": nodeinfo.get('bootid', None),
3373 "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
3374 "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
3377 live_data[name] = {}
3379 live_data = dict.fromkeys(nodenames, {})
3381 node_to_primary = dict([(name, set()) for name in nodenames])
3382 node_to_secondary = dict([(name, set()) for name in nodenames])
3384 inst_fields = frozenset(("pinst_cnt", "pinst_list",
3385 "sinst_cnt", "sinst_list"))
3386 if inst_fields & frozenset(self.op.output_fields):
3387 inst_data = self.cfg.GetAllInstancesInfo()
3389 for inst in inst_data.values():
3390 if inst.primary_node in node_to_primary:
3391 node_to_primary[inst.primary_node].add(inst.name)
3392 for secnode in inst.secondary_nodes:
3393 if secnode in node_to_secondary:
3394 node_to_secondary[secnode].add(inst.name)
3396 master_node = self.cfg.GetMasterNode()
3398 # end data gathering
3401 for node in nodelist:
3403 for field in self.op.output_fields:
3404 if field in self._SIMPLE_FIELDS:
3405 val = getattr(node, field)
3406 elif field == "pinst_list":
3407 val = list(node_to_primary[node.name])
3408 elif field == "sinst_list":
3409 val = list(node_to_secondary[node.name])
3410 elif field == "pinst_cnt":
3411 val = len(node_to_primary[node.name])
3412 elif field == "sinst_cnt":
3413 val = len(node_to_secondary[node.name])
3414 elif field == "pip":
3415 val = node.primary_ip
3416 elif field == "sip":
3417 val = node.secondary_ip
3418 elif field == "tags":
3419 val = list(node.GetTags())
3420 elif field == "master":
3421 val = node.name == master_node
3422 elif self._FIELDS_DYNAMIC.Matches(field):
3423 val = live_data[node.name].get(field, None)
3424 elif field == "role":
3425 if node.name == master_node:
3427 elif node.master_candidate:
3436 raise errors.ParameterError(field)
3437 node_output.append(val)
3438 output.append(node_output)
3443 class LUQueryNodeVolumes(NoHooksLU):
3444 """Logical unit for getting volumes on node(s).
3448 ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
3449 ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
3452 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3453 _FIELDS_STATIC = utils.FieldSet("node")
3455 def CheckArguments(self):
3456 _CheckOutputFields(static=self._FIELDS_STATIC,
3457 dynamic=self._FIELDS_DYNAMIC,
3458 selected=self.op.output_fields)
3460 def ExpandNames(self):
3461 self.needed_locks = {}
3462 self.share_locks[locking.LEVEL_NODE] = 1
3463 if not self.op.nodes:
3464 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3466 self.needed_locks[locking.LEVEL_NODE] = \
3467 _GetWantedNodes(self, self.op.nodes)
3469 def Exec(self, feedback_fn):
3470 """Computes the list of nodes and their attributes.
3473 nodenames = self.acquired_locks[locking.LEVEL_NODE]
3474 volumes = self.rpc.call_node_volumes(nodenames)
3476 ilist = [self.cfg.GetInstanceInfo(iname) for iname
3477 in self.cfg.GetInstanceList()]
3479 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3482 for node in nodenames:
3483 nresult = volumes[node]
3486 msg = nresult.fail_msg
3488 self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3491 node_vols = nresult.payload[:]
3492 node_vols.sort(key=lambda vol: vol['dev'])
3494 for vol in node_vols:
3496 for field in self.op.output_fields:
3499 elif field == "phys":
3503 elif field == "name":
3505 elif field == "size":
3506 val = int(float(vol['size']))
3507 elif field == "instance":
3509 if node not in lv_by_node[inst]:
3511 if vol['name'] in lv_by_node[inst][node]:
3517 raise errors.ParameterError(field)
3518 node_output.append(str(val))
3520 output.append(node_output)
3525 class LUQueryNodeStorage(NoHooksLU):
3526 """Logical unit for getting information on storage units on node(s).
3529 _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3531 ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
3532 ("storage_type", _NoDefault, _CheckStorageType),
3533 ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
3534 ("name", None, _TMaybeString),
3538 def CheckArguments(self):
3539 _CheckOutputFields(static=self._FIELDS_STATIC,
3540 dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3541 selected=self.op.output_fields)
3543 def ExpandNames(self):
3544 self.needed_locks = {}
3545 self.share_locks[locking.LEVEL_NODE] = 1
3548 self.needed_locks[locking.LEVEL_NODE] = \
3549 _GetWantedNodes(self, self.op.nodes)
3551 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3553 def Exec(self, feedback_fn):
3554 """Computes the list of nodes and their attributes.
3557 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3559 # Always get name to sort by
3560 if constants.SF_NAME in self.op.output_fields:
3561 fields = self.op.output_fields[:]
3563 fields = [constants.SF_NAME] + self.op.output_fields
3565 # Never ask for node or type as it's only known to the LU
3566 for extra in [constants.SF_NODE, constants.SF_TYPE]:
3567 while extra in fields:
3568 fields.remove(extra)
3570 field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3571 name_idx = field_idx[constants.SF_NAME]
3573 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3574 data = self.rpc.call_storage_list(self.nodes,
3575 self.op.storage_type, st_args,
3576 self.op.name, fields)
3580 for node in utils.NiceSort(self.nodes):
3581 nresult = data[node]
3585 msg = nresult.fail_msg
3587 self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3590 rows = dict([(row[name_idx], row) for row in nresult.payload])
3592 for name in utils.NiceSort(rows.keys()):
3597 for field in self.op.output_fields:
3598 if field == constants.SF_NODE:
3600 elif field == constants.SF_TYPE:
3601 val = self.op.storage_type
3602 elif field in field_idx:
3603 val = row[field_idx[field]]
3605 raise errors.ParameterError(field)
3614 class LUModifyNodeStorage(NoHooksLU):
3615 """Logical unit for modifying a storage volume on a node.
3620 ("storage_type", _NoDefault, _CheckStorageType),
3621 ("name", _NoDefault, _TNonEmptyString),
3622 ("changes", _NoDefault, _TDict),
3626 def CheckArguments(self):
3627 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3629 storage_type = self.op.storage_type
3632 modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3634 raise errors.OpPrereqError("Storage units of type '%s' can not be"
3635 " modified" % storage_type,
3638 diff = set(self.op.changes.keys()) - modifiable
3640 raise errors.OpPrereqError("The following fields can not be modified for"
3641 " storage units of type '%s': %r" %
3642 (storage_type, list(diff)),
3645 def ExpandNames(self):
3646 self.needed_locks = {
3647 locking.LEVEL_NODE: self.op.node_name,
3650 def Exec(self, feedback_fn):
3651 """Computes the list of nodes and their attributes.
3654 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3655 result = self.rpc.call_storage_modify(self.op.node_name,
3656 self.op.storage_type, st_args,
3657 self.op.name, self.op.changes)
3658 result.Raise("Failed to modify storage unit '%s' on %s" %
3659 (self.op.name, self.op.node_name))
3662 class LUAddNode(LogicalUnit):
3663 """Logical unit for adding node to the cluster.
3667 HTYPE = constants.HTYPE_NODE
3670 ("primary_ip", None, _NoType),
3671 ("secondary_ip", None, _TMaybeString),
3672 ("readd", False, _TBool),
3675 def CheckArguments(self):
3676 # validate/normalize the node name
3677 self.op.node_name = netutils.HostInfo.NormalizeName(self.op.node_name)
3679 def BuildHooksEnv(self):
3682 This will run on all nodes before, and on all nodes + the new node after.
3686 "OP_TARGET": self.op.node_name,
3687 "NODE_NAME": self.op.node_name,
3688 "NODE_PIP": self.op.primary_ip,
3689 "NODE_SIP": self.op.secondary_ip,
3691 nodes_0 = self.cfg.GetNodeList()
3692 nodes_1 = nodes_0 + [self.op.node_name, ]
3693 return env, nodes_0, nodes_1
3695 def CheckPrereq(self):
3696 """Check prerequisites.
3699 - the new node is not already in the config
3701 - its parameters (single/dual homed) matches the cluster
3703 Any errors are signaled by raising errors.OpPrereqError.
3706 node_name = self.op.node_name
3709 dns_data = netutils.GetHostInfo(node_name)
3711 node = dns_data.name
3712 primary_ip = self.op.primary_ip = dns_data.ip
3713 if self.op.secondary_ip is None:
3714 self.op.secondary_ip = primary_ip
3715 if not netutils.IP4Address.IsValid(self.op.secondary_ip):
3716 raise errors.OpPrereqError("Invalid secondary IP given",
3718 secondary_ip = self.op.secondary_ip
3720 node_list = cfg.GetNodeList()
3721 if not self.op.readd and node in node_list:
3722 raise errors.OpPrereqError("Node %s is already in the configuration" %
3723 node, errors.ECODE_EXISTS)
3724 elif self.op.readd and node not in node_list:
3725 raise errors.OpPrereqError("Node %s is not in the configuration" % node,
3728 self.changed_primary_ip = False
3730 for existing_node_name in node_list:
3731 existing_node = cfg.GetNodeInfo(existing_node_name)
3733 if self.op.readd and node == existing_node_name:
3734 if existing_node.secondary_ip != secondary_ip:
3735 raise errors.OpPrereqError("Readded node doesn't have the same IP"
3736 " address configuration as before",
3738 if existing_node.primary_ip != primary_ip:
3739 self.changed_primary_ip = True
3743 if (existing_node.primary_ip == primary_ip or
3744 existing_node.secondary_ip == primary_ip or
3745 existing_node.primary_ip == secondary_ip or
3746 existing_node.secondary_ip == secondary_ip):
3747 raise errors.OpPrereqError("New node ip address(es) conflict with"
3748 " existing node %s" % existing_node.name,
3749 errors.ECODE_NOTUNIQUE)
3751 # check that the type of the node (single versus dual homed) is the
3752 # same as for the master
3753 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
3754 master_singlehomed = myself.secondary_ip == myself.primary_ip
3755 newbie_singlehomed = secondary_ip == primary_ip
3756 if master_singlehomed != newbie_singlehomed:
3757 if master_singlehomed:
3758 raise errors.OpPrereqError("The master has no private ip but the"
3759 " new node has one",
3762 raise errors.OpPrereqError("The master has a private ip but the"
3763 " new node doesn't have one",
3766 # checks reachability
3767 if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
3768 raise errors.OpPrereqError("Node not reachable by ping",
3769 errors.ECODE_ENVIRON)
3771 if not newbie_singlehomed:
3772 # check reachability from my secondary ip to newbie's secondary ip
3773 if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3774 source=myself.secondary_ip):
3775 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3776 " based ping to noded port",
3777 errors.ECODE_ENVIRON)
3784 self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3787 self.new_node = self.cfg.GetNodeInfo(node)
3788 assert self.new_node is not None, "Can't retrieve locked node %s" % node
3790 self.new_node = objects.Node(name=node,
3791 primary_ip=primary_ip,
3792 secondary_ip=secondary_ip,
3793 master_candidate=self.master_candidate,
3794 offline=False, drained=False)
3796 def Exec(self, feedback_fn):
3797 """Adds the new node to the cluster.
3800 new_node = self.new_node
3801 node = new_node.name
3803 # for re-adds, reset the offline/drained/master-candidate flags;
3804 # we need to reset here, otherwise offline would prevent RPC calls
3805 # later in the procedure; this also means that if the re-add
3806 # fails, we are left with a non-offlined, broken node
3808 new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3809 self.LogInfo("Readding a node, the offline/drained flags were reset")
3810 # if we demote the node, we do cleanup later in the procedure
3811 new_node.master_candidate = self.master_candidate
3812 if self.changed_primary_ip:
3813 new_node.primary_ip = self.op.primary_ip
3815 # notify the user about any possible mc promotion
3816 if new_node.master_candidate:
3817 self.LogInfo("Node will be a master candidate")
3819 # check connectivity
3820 result = self.rpc.call_version([node])[node]
3821 result.Raise("Can't get version information from node %s" % node)
3822 if constants.PROTOCOL_VERSION == result.payload:
3823 logging.info("Communication to node %s fine, sw version %s match",
3824 node, result.payload)
3826 raise errors.OpExecError("Version mismatch master version %s,"
3827 " node version %s" %
3828 (constants.PROTOCOL_VERSION, result.payload))
3831 if self.cfg.GetClusterInfo().modify_ssh_setup:
3832 logging.info("Copy ssh key to node %s", node)
3833 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
3835 keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
3836 constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
3840 keyarray.append(utils.ReadFile(i))
3842 result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
3843 keyarray[2], keyarray[3], keyarray[4],
3845 result.Raise("Cannot transfer ssh keys to the new node")
3847 # Add node to our /etc/hosts, and add key to known_hosts
3848 if self.cfg.GetClusterInfo().modify_etc_hosts:
3849 # FIXME: this should be done via an rpc call to node daemon
3850 utils.AddHostToEtcHosts(new_node.name)
3852 if new_node.secondary_ip != new_node.primary_ip:
3853 result = self.rpc.call_node_has_ip_address(new_node.name,
3854 new_node.secondary_ip)
3855 result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3856 prereq=True, ecode=errors.ECODE_ENVIRON)
3857 if not result.payload:
3858 raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3859 " you gave (%s). Please fix and re-run this"
3860 " command." % new_node.secondary_ip)
3862 node_verify_list = [self.cfg.GetMasterNode()]
3863 node_verify_param = {
3864 constants.NV_NODELIST: [node],
3865 # TODO: do a node-net-test as well?
3868 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3869 self.cfg.GetClusterName())
3870 for verifier in node_verify_list:
3871 result[verifier].Raise("Cannot communicate with node %s" % verifier)
3872 nl_payload = result[verifier].payload[constants.NV_NODELIST]
3874 for failed in nl_payload:
3875 feedback_fn("ssh/hostname verification failed"
3876 " (checking from %s): %s" %
3877 (verifier, nl_payload[failed]))
3878 raise errors.OpExecError("ssh/hostname verification failed.")
3881 _RedistributeAncillaryFiles(self)
3882 self.context.ReaddNode(new_node)
3883 # make sure we redistribute the config
3884 self.cfg.Update(new_node, feedback_fn)
3885 # and make sure the new node will not have old files around
3886 if not new_node.master_candidate:
3887 result = self.rpc.call_node_demote_from_mc(new_node.name)
3888 msg = result.fail_msg
3890 self.LogWarning("Node failed to demote itself from master"
3891 " candidate status: %s" % msg)
3893 _RedistributeAncillaryFiles(self, additional_nodes=[node])
3894 self.context.AddNode(new_node, self.proc.GetECId())
3897 class LUSetNodeParams(LogicalUnit):
3898 """Modifies the parameters of a node.
3901 HPATH = "node-modify"
3902 HTYPE = constants.HTYPE_NODE
3905 ("master_candidate", None, _TMaybeBool),
3906 ("offline", None, _TMaybeBool),
3907 ("drained", None, _TMaybeBool),
3908 ("auto_promote", False, _TBool),
3913 def CheckArguments(self):
3914 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3915 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3916 if all_mods.count(None) == 3:
3917 raise errors.OpPrereqError("Please pass at least one modification",
3919 if all_mods.count(True) > 1:
3920 raise errors.OpPrereqError("Can't set the node into more than one"
3921 " state at the same time",
3924 # Boolean value that tells us whether we're offlining or draining the node
3925 self.offline_or_drain = (self.op.offline == True or
3926 self.op.drained == True)
3927 self.deoffline_or_drain = (self.op.offline == False or
3928 self.op.drained == False)
3929 self.might_demote = (self.op.master_candidate == False or
3930 self.offline_or_drain)
3932 self.lock_all = self.op.auto_promote and self.might_demote
3935 def ExpandNames(self):
3937 self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
3939 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3941 def BuildHooksEnv(self):
3944 This runs on the master node.
3948 "OP_TARGET": self.op.node_name,
3949 "MASTER_CANDIDATE": str(self.op.master_candidate),
3950 "OFFLINE": str(self.op.offline),
3951 "DRAINED": str(self.op.drained),
3953 nl = [self.cfg.GetMasterNode(),
3957 def CheckPrereq(self):
3958 """Check prerequisites.
3960 This only checks the instance list against the existing names.
3963 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3965 if (self.op.master_candidate is not None or
3966 self.op.drained is not None or
3967 self.op.offline is not None):
3968 # we can't change the master's node flags
3969 if self.op.node_name == self.cfg.GetMasterNode():
3970 raise errors.OpPrereqError("The master role can be changed"
3971 " only via master-failover",
3975 if node.master_candidate and self.might_demote and not self.lock_all:
3976 assert not self.op.auto_promote, "auto-promote set but lock_all not"
3977 # check if after removing the current node, we're missing master
3979 (mc_remaining, mc_should, _) = \
3980 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
3981 if mc_remaining < mc_should:
3982 raise errors.OpPrereqError("Not enough master candidates, please"
3983 " pass auto_promote to allow promotion",
3986 if (self.op.master_candidate == True and
3987 ((node.offline and not self.op.offline == False) or
3988 (node.drained and not self.op.drained == False))):
3989 raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3990 " to master_candidate" % node.name,
3993 # If we're being deofflined/drained, we'll MC ourself if needed
3994 if (self.deoffline_or_drain and not self.offline_or_drain and not
3995 self.op.master_candidate == True and not node.master_candidate):
3996 self.op.master_candidate = _DecideSelfPromotion(self)
3997 if self.op.master_candidate:
3998 self.LogInfo("Autopromoting node to master candidate")
4002 def Exec(self, feedback_fn):
4011 if self.op.offline is not None:
4012 node.offline = self.op.offline
4013 result.append(("offline", str(self.op.offline)))
4014 if self.op.offline == True:
4015 if node.master_candidate:
4016 node.master_candidate = False
4018 result.append(("master_candidate", "auto-demotion due to offline"))
4020 node.drained = False
4021 result.append(("drained", "clear drained status due to offline"))
4023 if self.op.master_candidate is not None:
4024 node.master_candidate = self.op.master_candidate
4026 result.append(("master_candidate", str(self.op.master_candidate)))
4027 if self.op.master_candidate == False:
4028 rrc = self.rpc.call_node_demote_from_mc(node.name)
4031 self.LogWarning("Node failed to demote itself: %s" % msg)
4033 if self.op.drained is not None:
4034 node.drained = self.op.drained
4035 result.append(("drained", str(self.op.drained)))
4036 if self.op.drained == True:
4037 if node.master_candidate:
4038 node.master_candidate = False
4040 result.append(("master_candidate", "auto-demotion due to drain"))
4041 rrc = self.rpc.call_node_demote_from_mc(node.name)
4044 self.LogWarning("Node failed to demote itself: %s" % msg)
4046 node.offline = False
4047 result.append(("offline", "clear offline status due to drain"))
4049 # we locked all nodes, we adjust the CP before updating this node
4051 _AdjustCandidatePool(self, [node.name])
4053 # this will trigger configuration file update, if needed
4054 self.cfg.Update(node, feedback_fn)
4056 # this will trigger job queue propagation or cleanup
4058 self.context.ReaddNode(node)
4063 class LUPowercycleNode(NoHooksLU):
4064 """Powercycles a node.
4073 def CheckArguments(self):
4074 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4075 if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4076 raise errors.OpPrereqError("The node is the master and the force"
4077 " parameter was not set",
4080 def ExpandNames(self):
4081 """Locking for PowercycleNode.
4083 This is a last-resort option and shouldn't block on other
4084 jobs. Therefore, we grab no locks.
4087 self.needed_locks = {}
4089 def Exec(self, feedback_fn):
4093 result = self.rpc.call_node_powercycle(self.op.node_name,
4094 self.cfg.GetHypervisorType())
4095 result.Raise("Failed to schedule the reboot")
4096 return result.payload
4099 class LUQueryClusterInfo(NoHooksLU):
4100 """Query cluster configuration.
4105 def ExpandNames(self):
4106 self.needed_locks = {}
4108 def Exec(self, feedback_fn):
4109 """Return cluster config.
4112 cluster = self.cfg.GetClusterInfo()
4115 # Filter just for enabled hypervisors
4116 for os_name, hv_dict in cluster.os_hvp.items():
4117 os_hvp[os_name] = {}
4118 for hv_name, hv_params in hv_dict.items():
4119 if hv_name in cluster.enabled_hypervisors:
4120 os_hvp[os_name][hv_name] = hv_params
4123 "software_version": constants.RELEASE_VERSION,
4124 "protocol_version": constants.PROTOCOL_VERSION,
4125 "config_version": constants.CONFIG_VERSION,
4126 "os_api_version": max(constants.OS_API_VERSIONS),
4127 "export_version": constants.EXPORT_VERSION,
4128 "architecture": (platform.architecture()[0], platform.machine()),
4129 "name": cluster.cluster_name,
4130 "master": cluster.master_node,
4131 "default_hypervisor": cluster.enabled_hypervisors[0],
4132 "enabled_hypervisors": cluster.enabled_hypervisors,
4133 "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4134 for hypervisor_name in cluster.enabled_hypervisors]),
4136 "beparams": cluster.beparams,
4137 "osparams": cluster.osparams,
4138 "nicparams": cluster.nicparams,
4139 "candidate_pool_size": cluster.candidate_pool_size,
4140 "master_netdev": cluster.master_netdev,
4141 "volume_group_name": cluster.volume_group_name,
4142 "drbd_usermode_helper": cluster.drbd_usermode_helper,
4143 "file_storage_dir": cluster.file_storage_dir,
4144 "maintain_node_health": cluster.maintain_node_health,
4145 "ctime": cluster.ctime,
4146 "mtime": cluster.mtime,
4147 "uuid": cluster.uuid,
4148 "tags": list(cluster.GetTags()),
4149 "uid_pool": cluster.uid_pool,
4150 "default_iallocator": cluster.default_iallocator,
4151 "reserved_lvs": cluster.reserved_lvs,
4157 class LUQueryConfigValues(NoHooksLU):
4158 """Return configuration values.
4161 _OP_PARAMS = [_POutputFields]
4163 _FIELDS_DYNAMIC = utils.FieldSet()
4164 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4167 def CheckArguments(self):
4168 _CheckOutputFields(static=self._FIELDS_STATIC,
4169 dynamic=self._FIELDS_DYNAMIC,
4170 selected=self.op.output_fields)
4172 def ExpandNames(self):
4173 self.needed_locks = {}
4175 def Exec(self, feedback_fn):
4176 """Dump a representation of the cluster config to the standard output.
4180 for field in self.op.output_fields:
4181 if field == "cluster_name":
4182 entry = self.cfg.GetClusterName()
4183 elif field == "master_node":
4184 entry = self.cfg.GetMasterNode()
4185 elif field == "drain_flag":
4186 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4187 elif field == "watcher_pause":
4188 entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4190 raise errors.ParameterError(field)
4191 values.append(entry)
4195 class LUActivateInstanceDisks(NoHooksLU):
4196 """Bring up an instance's disks.
4201 ("ignore_size", False, _TBool),
4205 def ExpandNames(self):
4206 self._ExpandAndLockInstance()
4207 self.needed_locks[locking.LEVEL_NODE] = []
4208 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4210 def DeclareLocks(self, level):
4211 if level == locking.LEVEL_NODE:
4212 self._LockInstancesNodes()
4214 def CheckPrereq(self):
4215 """Check prerequisites.
4217 This checks that the instance is in the cluster.
4220 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4221 assert self.instance is not None, \
4222 "Cannot retrieve locked instance %s" % self.op.instance_name
4223 _CheckNodeOnline(self, self.instance.primary_node)
4225 def Exec(self, feedback_fn):
4226 """Activate the disks.
4229 disks_ok, disks_info = \
4230 _AssembleInstanceDisks(self, self.instance,
4231 ignore_size=self.op.ignore_size)
4233 raise errors.OpExecError("Cannot activate block devices")
4238 def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4240 """Prepare the block devices for an instance.
4242 This sets up the block devices on all nodes.
4244 @type lu: L{LogicalUnit}
4245 @param lu: the logical unit on whose behalf we execute
4246 @type instance: L{objects.Instance}
4247 @param instance: the instance for whose disks we assemble
4248 @type disks: list of L{objects.Disk} or None
4249 @param disks: which disks to assemble (or all, if None)
4250 @type ignore_secondaries: boolean
4251 @param ignore_secondaries: if true, errors on secondary nodes
4252 won't result in an error return from the function
4253 @type ignore_size: boolean
4254 @param ignore_size: if true, the current known size of the disk
4255 will not be used during the disk activation, useful for cases
4256 when the size is wrong
4257 @return: False if the operation failed, otherwise a list of
4258 (host, instance_visible_name, node_visible_name)
4259 with the mapping from node devices to instance devices
4264 iname = instance.name
4265 disks = _ExpandCheckDisks(instance, disks)
4267 # With the two passes mechanism we try to reduce the window of
4268 # opportunity for the race condition of switching DRBD to primary
4269 # before handshaking occured, but we do not eliminate it
4271 # The proper fix would be to wait (with some limits) until the
4272 # connection has been made and drbd transitions from WFConnection
4273 # into any other network-connected state (Connected, SyncTarget,
4276 # 1st pass, assemble on all nodes in secondary mode
4277 for inst_disk in disks:
4278 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4280 node_disk = node_disk.Copy()
4281 node_disk.UnsetSize()
4282 lu.cfg.SetDiskID(node_disk, node)
4283 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4284 msg = result.fail_msg
4286 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4287 " (is_primary=False, pass=1): %s",
4288 inst_disk.iv_name, node, msg)
4289 if not ignore_secondaries:
4292 # FIXME: race condition on drbd migration to primary
4294 # 2nd pass, do only the primary node
4295 for inst_disk in disks:
4298 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4299 if node != instance.primary_node:
4302 node_disk = node_disk.Copy()
4303 node_disk.UnsetSize()
4304 lu.cfg.SetDiskID(node_disk, node)
4305 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4306 msg = result.fail_msg
4308 lu.proc.LogWarning("Could not prepare block device %s on node %s"
4309 " (is_primary=True, pass=2): %s",
4310 inst_disk.iv_name, node, msg)
4313 dev_path = result.payload
4315 device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4317 # leave the disks configured for the primary node
4318 # this is a workaround that would be fixed better by
4319 # improving the logical/physical id handling
4321 lu.cfg.SetDiskID(disk, instance.primary_node)
4323 return disks_ok, device_info
4326 def _StartInstanceDisks(lu, instance, force):
4327 """Start the disks of an instance.
4330 disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4331 ignore_secondaries=force)
4333 _ShutdownInstanceDisks(lu, instance)
4334 if force is not None and not force:
4335 lu.proc.LogWarning("", hint="If the message above refers to a"
4337 " you can retry the operation using '--force'.")
4338 raise errors.OpExecError("Disk consistency error")
4341 class LUDeactivateInstanceDisks(NoHooksLU):
4342 """Shutdown an instance's disks.
4350 def ExpandNames(self):
4351 self._ExpandAndLockInstance()
4352 self.needed_locks[locking.LEVEL_NODE] = []
4353 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4355 def DeclareLocks(self, level):
4356 if level == locking.LEVEL_NODE:
4357 self._LockInstancesNodes()
4359 def CheckPrereq(self):
4360 """Check prerequisites.
4362 This checks that the instance is in the cluster.
4365 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4366 assert self.instance is not None, \
4367 "Cannot retrieve locked instance %s" % self.op.instance_name
4369 def Exec(self, feedback_fn):
4370 """Deactivate the disks
4373 instance = self.instance
4374 _SafeShutdownInstanceDisks(self, instance)
4377 def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4378 """Shutdown block devices of an instance.
4380 This function checks if an instance is running, before calling
4381 _ShutdownInstanceDisks.
4384 _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4385 _ShutdownInstanceDisks(lu, instance, disks=disks)
4388 def _ExpandCheckDisks(instance, disks):
4389 """Return the instance disks selected by the disks list
4391 @type disks: list of L{objects.Disk} or None
4392 @param disks: selected disks
4393 @rtype: list of L{objects.Disk}
4394 @return: selected instance disks to act on
4398 return instance.disks
4400 if not set(disks).issubset(instance.disks):
4401 raise errors.ProgrammerError("Can only act on disks belonging to the"
4406 def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4407 """Shutdown block devices of an instance.
4409 This does the shutdown on all nodes of the instance.
4411 If the ignore_primary is false, errors on the primary node are
4416 disks = _ExpandCheckDisks(instance, disks)
4419 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4420 lu.cfg.SetDiskID(top_disk, node)
4421 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4422 msg = result.fail_msg
4424 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4425 disk.iv_name, node, msg)
4426 if not ignore_primary or node != instance.primary_node:
4431 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4432 """Checks if a node has enough free memory.
4434 This function check if a given node has the needed amount of free
4435 memory. In case the node has less memory or we cannot get the
4436 information from the node, this function raise an OpPrereqError
4439 @type lu: C{LogicalUnit}
4440 @param lu: a logical unit from which we get configuration data
4442 @param node: the node to check
4443 @type reason: C{str}
4444 @param reason: string to use in the error message
4445 @type requested: C{int}
4446 @param requested: the amount of memory in MiB to check for
4447 @type hypervisor_name: C{str}
4448 @param hypervisor_name: the hypervisor to ask for memory stats
4449 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4450 we cannot check the node
4453 nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4454 nodeinfo[node].Raise("Can't get data from node %s" % node,
4455 prereq=True, ecode=errors.ECODE_ENVIRON)
4456 free_mem = nodeinfo[node].payload.get('memory_free', None)
4457 if not isinstance(free_mem, int):
4458 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4459 " was '%s'" % (node, free_mem),
4460 errors.ECODE_ENVIRON)
4461 if requested > free_mem:
4462 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4463 " needed %s MiB, available %s MiB" %
4464 (node, reason, requested, free_mem),
4468 def _CheckNodesFreeDisk(lu, nodenames, requested):
4469 """Checks if nodes have enough free disk space in the default VG.
4471 This function check if all given nodes have the needed amount of
4472 free disk. In case any node has less disk or we cannot get the
4473 information from the node, this function raise an OpPrereqError
4476 @type lu: C{LogicalUnit}
4477 @param lu: a logical unit from which we get configuration data
4478 @type nodenames: C{list}
4479 @param nodenames: the list of node names to check
4480 @type requested: C{int}
4481 @param requested: the amount of disk in MiB to check for
4482 @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4483 we cannot check the node
4486 nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4487 lu.cfg.GetHypervisorType())
4488 for node in nodenames:
4489 info = nodeinfo[node]
4490 info.Raise("Cannot get current information from node %s" % node,
4491 prereq=True, ecode=errors.ECODE_ENVIRON)
4492 vg_free = info.payload.get("vg_free", None)
4493 if not isinstance(vg_free, int):
4494 raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4495 " result was '%s'" % (node, vg_free),
4496 errors.ECODE_ENVIRON)
4497 if requested > vg_free:
4498 raise errors.OpPrereqError("Not enough disk space on target node %s:"
4499 " required %d MiB, available %d MiB" %
4500 (node, requested, vg_free),
4504 class LUStartupInstance(LogicalUnit):
4505 """Starts an instance.
4508 HPATH = "instance-start"
4509 HTYPE = constants.HTYPE_INSTANCE
4513 ("hvparams", _EmptyDict, _TDict),
4514 ("beparams", _EmptyDict, _TDict),
4518 def CheckArguments(self):
4520 if self.op.beparams:
4521 # fill the beparams dict
4522 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4524 def ExpandNames(self):
4525 self._ExpandAndLockInstance()
4527 def BuildHooksEnv(self):
4530 This runs on master, primary and secondary nodes of the instance.
4534 "FORCE": self.op.force,
4536 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4537 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4540 def CheckPrereq(self):
4541 """Check prerequisites.
4543 This checks that the instance is in the cluster.
4546 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4547 assert self.instance is not None, \
4548 "Cannot retrieve locked instance %s" % self.op.instance_name
4551 if self.op.hvparams:
4552 # check hypervisor parameter syntax (locally)
4553 cluster = self.cfg.GetClusterInfo()
4554 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4555 filled_hvp = cluster.FillHV(instance)
4556 filled_hvp.update(self.op.hvparams)
4557 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4558 hv_type.CheckParameterSyntax(filled_hvp)
4559 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4561 _CheckNodeOnline(self, instance.primary_node)
4563 bep = self.cfg.GetClusterInfo().FillBE(instance)
4564 # check bridges existence
4565 _CheckInstanceBridgesExist(self, instance)
4567 remote_info = self.rpc.call_instance_info(instance.primary_node,
4569 instance.hypervisor)
4570 remote_info.Raise("Error checking node %s" % instance.primary_node,
4571 prereq=True, ecode=errors.ECODE_ENVIRON)
4572 if not remote_info.payload: # not running already
4573 _CheckNodeFreeMemory(self, instance.primary_node,
4574 "starting instance %s" % instance.name,
4575 bep[constants.BE_MEMORY], instance.hypervisor)
4577 def Exec(self, feedback_fn):
4578 """Start the instance.
4581 instance = self.instance
4582 force = self.op.force
4584 self.cfg.MarkInstanceUp(instance.name)
4586 node_current = instance.primary_node
4588 _StartInstanceDisks(self, instance, force)
4590 result = self.rpc.call_instance_start(node_current, instance,
4591 self.op.hvparams, self.op.beparams)
4592 msg = result.fail_msg
4594 _ShutdownInstanceDisks(self, instance)
4595 raise errors.OpExecError("Could not start instance: %s" % msg)
4598 class LURebootInstance(LogicalUnit):
4599 """Reboot an instance.
4602 HPATH = "instance-reboot"
4603 HTYPE = constants.HTYPE_INSTANCE
4606 ("ignore_secondaries", False, _TBool),
4607 ("reboot_type", _NoDefault, _TElemOf(constants.REBOOT_TYPES)),
4612 def ExpandNames(self):
4613 self._ExpandAndLockInstance()
4615 def BuildHooksEnv(self):
4618 This runs on master, primary and secondary nodes of the instance.
4622 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4623 "REBOOT_TYPE": self.op.reboot_type,
4624 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
4626 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4627 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4630 def CheckPrereq(self):
4631 """Check prerequisites.
4633 This checks that the instance is in the cluster.
4636 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4637 assert self.instance is not None, \
4638 "Cannot retrieve locked instance %s" % self.op.instance_name
4640 _CheckNodeOnline(self, instance.primary_node)
4642 # check bridges existence
4643 _CheckInstanceBridgesExist(self, instance)
4645 def Exec(self, feedback_fn):
4646 """Reboot the instance.
4649 instance = self.instance
4650 ignore_secondaries = self.op.ignore_secondaries
4651 reboot_type = self.op.reboot_type
4653 node_current = instance.primary_node
4655 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4656 constants.INSTANCE_REBOOT_HARD]:
4657 for disk in instance.disks:
4658 self.cfg.SetDiskID(disk, node_current)
4659 result = self.rpc.call_instance_reboot(node_current, instance,
4661 self.op.shutdown_timeout)
4662 result.Raise("Could not reboot instance")
4664 result = self.rpc.call_instance_shutdown(node_current, instance,
4665 self.op.shutdown_timeout)
4666 result.Raise("Could not shutdown instance for full reboot")
4667 _ShutdownInstanceDisks(self, instance)
4668 _StartInstanceDisks(self, instance, ignore_secondaries)
4669 result = self.rpc.call_instance_start(node_current, instance, None, None)
4670 msg = result.fail_msg
4672 _ShutdownInstanceDisks(self, instance)
4673 raise errors.OpExecError("Could not start instance for"
4674 " full reboot: %s" % msg)
4676 self.cfg.MarkInstanceUp(instance.name)
4679 class LUShutdownInstance(LogicalUnit):
4680 """Shutdown an instance.
4683 HPATH = "instance-stop"
4684 HTYPE = constants.HTYPE_INSTANCE
4687 ("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, _TPositiveInt),
4691 def ExpandNames(self):
4692 self._ExpandAndLockInstance()
4694 def BuildHooksEnv(self):
4697 This runs on master, primary and secondary nodes of the instance.
4700 env = _BuildInstanceHookEnvByObject(self, self.instance)
4701 env["TIMEOUT"] = self.op.timeout
4702 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4705 def CheckPrereq(self):
4706 """Check prerequisites.
4708 This checks that the instance is in the cluster.
4711 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4712 assert self.instance is not None, \
4713 "Cannot retrieve locked instance %s" % self.op.instance_name
4714 _CheckNodeOnline(self, self.instance.primary_node)
4716 def Exec(self, feedback_fn):
4717 """Shutdown the instance.
4720 instance = self.instance
4721 node_current = instance.primary_node
4722 timeout = self.op.timeout
4723 self.cfg.MarkInstanceDown(instance.name)
4724 result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4725 msg = result.fail_msg
4727 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4729 _ShutdownInstanceDisks(self, instance)
4732 class LUReinstallInstance(LogicalUnit):
4733 """Reinstall an instance.
4736 HPATH = "instance-reinstall"
4737 HTYPE = constants.HTYPE_INSTANCE
4740 ("os_type", None, _TMaybeString),
4741 ("force_variant", False, _TBool),
4745 def ExpandNames(self):
4746 self._ExpandAndLockInstance()
4748 def BuildHooksEnv(self):
4751 This runs on master, primary and secondary nodes of the instance.
4754 env = _BuildInstanceHookEnvByObject(self, self.instance)
4755 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4758 def CheckPrereq(self):
4759 """Check prerequisites.
4761 This checks that the instance is in the cluster and is not running.
4764 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4765 assert instance is not None, \
4766 "Cannot retrieve locked instance %s" % self.op.instance_name
4767 _CheckNodeOnline(self, instance.primary_node)
4769 if instance.disk_template == constants.DT_DISKLESS:
4770 raise errors.OpPrereqError("Instance '%s' has no disks" %
4771 self.op.instance_name,
4773 _CheckInstanceDown(self, instance, "cannot reinstall")
4775 if self.op.os_type is not None:
4777 pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4778 _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4780 self.instance = instance
4782 def Exec(self, feedback_fn):
4783 """Reinstall the instance.
4786 inst = self.instance
4788 if self.op.os_type is not None:
4789 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4790 inst.os = self.op.os_type
4791 self.cfg.Update(inst, feedback_fn)
4793 _StartInstanceDisks(self, inst, None)
4795 feedback_fn("Running the instance OS create scripts...")
4796 # FIXME: pass debug option from opcode to backend
4797 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4798 self.op.debug_level)
4799 result.Raise("Could not install OS for instance %s on node %s" %
4800 (inst.name, inst.primary_node))
4802 _ShutdownInstanceDisks(self, inst)
4805 class LURecreateInstanceDisks(LogicalUnit):
4806 """Recreate an instance's missing disks.
4809 HPATH = "instance-recreate-disks"
4810 HTYPE = constants.HTYPE_INSTANCE
4813 ("disks", _EmptyList, _TListOf(_TPositiveInt)),
4817 def ExpandNames(self):
4818 self._ExpandAndLockInstance()
4820 def BuildHooksEnv(self):
4823 This runs on master, primary and secondary nodes of the instance.
4826 env = _BuildInstanceHookEnvByObject(self, self.instance)
4827 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4830 def CheckPrereq(self):
4831 """Check prerequisites.
4833 This checks that the instance is in the cluster and is not running.
4836 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4837 assert instance is not None, \
4838 "Cannot retrieve locked instance %s" % self.op.instance_name
4839 _CheckNodeOnline(self, instance.primary_node)
4841 if instance.disk_template == constants.DT_DISKLESS:
4842 raise errors.OpPrereqError("Instance '%s' has no disks" %
4843 self.op.instance_name, errors.ECODE_INVAL)
4844 _CheckInstanceDown(self, instance, "cannot recreate disks")
4846 if not self.op.disks:
4847 self.op.disks = range(len(instance.disks))
4849 for idx in self.op.disks:
4850 if idx >= len(instance.disks):
4851 raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4854 self.instance = instance
4856 def Exec(self, feedback_fn):
4857 """Recreate the disks.
4861 for idx, _ in enumerate(self.instance.disks):
4862 if idx not in self.op.disks: # disk idx has not been passed in
4866 _CreateDisks(self, self.instance, to_skip=to_skip)
4869 class LURenameInstance(LogicalUnit):
4870 """Rename an instance.
4873 HPATH = "instance-rename"
4874 HTYPE = constants.HTYPE_INSTANCE
4877 ("new_name", _NoDefault, _TNonEmptyString),
4878 ("ip_check", False, _TBool),
4879 ("name_check", True, _TBool),
4882 def CheckArguments(self):
4886 if self.op.ip_check and not self.op.name_check:
4887 # TODO: make the ip check more flexible and not depend on the name check
4888 raise errors.OpPrereqError("Cannot do ip check without a name check",
4891 def BuildHooksEnv(self):
4894 This runs on master, primary and secondary nodes of the instance.
4897 env = _BuildInstanceHookEnvByObject(self, self.instance)
4898 env["INSTANCE_NEW_NAME"] = self.op.new_name
4899 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4902 def CheckPrereq(self):
4903 """Check prerequisites.
4905 This checks that the instance is in the cluster and is not running.
4908 self.op.instance_name = _ExpandInstanceName(self.cfg,
4909 self.op.instance_name)
4910 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4911 assert instance is not None
4912 _CheckNodeOnline(self, instance.primary_node)
4913 _CheckInstanceDown(self, instance, "cannot rename")
4914 self.instance = instance
4916 new_name = self.op.new_name
4917 if self.op.name_check:
4918 hostinfo = netutils.HostInfo(netutils.HostInfo.NormalizeName(new_name))
4919 new_name = hostinfo.name
4920 if (self.op.ip_check and
4921 netutils.TcpPing(hostinfo.ip, constants.DEFAULT_NODED_PORT)):
4922 raise errors.OpPrereqError("IP %s of instance %s already in use" %
4923 (hostinfo.ip, new_name),
4924 errors.ECODE_NOTUNIQUE)
4926 instance_list = self.cfg.GetInstanceList()
4927 if new_name in instance_list:
4928 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4929 new_name, errors.ECODE_EXISTS)
4932 def Exec(self, feedback_fn):
4933 """Reinstall the instance.
4936 inst = self.instance
4937 old_name = inst.name
4939 if inst.disk_template == constants.DT_FILE:
4940 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4942 self.cfg.RenameInstance(inst.name, self.op.new_name)
4943 # Change the instance lock. This is definitely safe while we hold the BGL
4944 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4945 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4947 # re-read the instance from the configuration after rename
4948 inst = self.cfg.GetInstanceInfo(self.op.new_name)
4950 if inst.disk_template == constants.DT_FILE:
4951 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4952 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4953 old_file_storage_dir,
4954 new_file_storage_dir)
4955 result.Raise("Could not rename on node %s directory '%s' to '%s'"
4956 " (but the instance has been renamed in Ganeti)" %
4957 (inst.primary_node, old_file_storage_dir,
4958 new_file_storage_dir))
4960 _StartInstanceDisks(self, inst, None)
4962 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4963 old_name, self.op.debug_level)
4964 msg = result.fail_msg
4966 msg = ("Could not run OS rename script for instance %s on node %s"
4967 " (but the instance has been renamed in Ganeti): %s" %
4968 (inst.name, inst.primary_node, msg))
4969 self.proc.LogWarning(msg)
4971 _ShutdownInstanceDisks(self, inst)
4976 class LURemoveInstance(LogicalUnit):
4977 """Remove an instance.
4980 HPATH = "instance-remove"
4981 HTYPE = constants.HTYPE_INSTANCE
4984 ("ignore_failures", False, _TBool),
4989 def ExpandNames(self):
4990 self._ExpandAndLockInstance()
4991 self.needed_locks[locking.LEVEL_NODE] = []
4992 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4994 def DeclareLocks(self, level):
4995 if level == locking.LEVEL_NODE:
4996 self._LockInstancesNodes()
4998 def BuildHooksEnv(self):
5001 This runs on master, primary and secondary nodes of the instance.
5004 env = _BuildInstanceHookEnvByObject(self, self.instance)
5005 env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5006 nl = [self.cfg.GetMasterNode()]
5007 nl_post = list(self.instance.all_nodes) + nl
5008 return env, nl, nl_post
5010 def CheckPrereq(self):
5011 """Check prerequisites.
5013 This checks that the instance is in the cluster.
5016 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5017 assert self.instance is not None, \
5018 "Cannot retrieve locked instance %s" % self.op.instance_name
5020 def Exec(self, feedback_fn):
5021 """Remove the instance.
5024 instance = self.instance
5025 logging.info("Shutting down instance %s on node %s",
5026 instance.name, instance.primary_node)
5028 result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5029 self.op.shutdown_timeout)
5030 msg = result.fail_msg
5032 if self.op.ignore_failures:
5033 feedback_fn("Warning: can't shutdown instance: %s" % msg)
5035 raise errors.OpExecError("Could not shutdown instance %s on"
5037 (instance.name, instance.primary_node, msg))
5039 _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5042 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5043 """Utility function to remove an instance.
5046 logging.info("Removing block devices for instance %s", instance.name)
5048 if not _RemoveDisks(lu, instance):
5049 if not ignore_failures:
5050 raise errors.OpExecError("Can't remove instance's disks")
5051 feedback_fn("Warning: can't remove instance's disks")
5053 logging.info("Removing instance %s out of cluster config", instance.name)
5055 lu.cfg.RemoveInstance(instance.name)
5057 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5058 "Instance lock removal conflict"
5060 # Remove lock for the instance
5061 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5064 class LUQueryInstances(NoHooksLU):
5065 """Logical unit for querying instances.
5068 # pylint: disable-msg=W0142
5070 ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
5071 ("names", _EmptyList, _TListOf(_TNonEmptyString)),
5072 ("use_locking", False, _TBool),
5075 _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
5076 "serial_no", "ctime", "mtime", "uuid"]
5077 _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
5079 "disk_template", "ip", "mac", "bridge",
5080 "nic_mode", "nic_link",
5081 "sda_size", "sdb_size", "vcpus", "tags",
5082 "network_port", "beparams",
5083 r"(disk)\.(size)/([0-9]+)",
5084 r"(disk)\.(sizes)", "disk_usage",
5085 r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
5086 r"(nic)\.(bridge)/([0-9]+)",
5087 r"(nic)\.(macs|ips|modes|links|bridges)",
5088 r"(disk|nic)\.(count)",
5090 ] + _SIMPLE_FIELDS +
5092 for name in constants.HVS_PARAMETERS
5093 if name not in constants.HVC_GLOBALS] +
5095 for name in constants.BES_PARAMETERS])
5096 _FIELDS_DYNAMIC = utils.FieldSet("oper_state",
5102 def CheckArguments(self):
5103 _CheckOutputFields(static=self._FIELDS_STATIC,
5104 dynamic=self._FIELDS_DYNAMIC,
5105 selected=self.op.output_fields)
5107 def ExpandNames(self):
5108 self.needed_locks = {}
5109 self.share_locks[locking.LEVEL_INSTANCE] = 1
5110 self.share_locks[locking.LEVEL_NODE] = 1
5113 self.wanted = _GetWantedInstances(self, self.op.names)
5115 self.wanted = locking.ALL_SET
5117 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
5118 self.do_locking = self.do_node_query and self.op.use_locking
5120 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5121 self.needed_locks[locking.LEVEL_NODE] = []
5122 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5124 def DeclareLocks(self, level):
5125 if level == locking.LEVEL_NODE and self.do_locking:
5126 self._LockInstancesNodes()
5128 def Exec(self, feedback_fn):
5129 """Computes the list of nodes and their attributes.
5132 # pylint: disable-msg=R0912
5133 # way too many branches here
5134 all_info = self.cfg.GetAllInstancesInfo()
5135 if self.wanted == locking.ALL_SET:
5136 # caller didn't specify instance names, so ordering is not important
5138 instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5140 instance_names = all_info.keys()
5141 instance_names = utils.NiceSort(instance_names)
5143 # caller did specify names, so we must keep the ordering
5145 tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
5147 tgt_set = all_info.keys()
5148 missing = set(self.wanted).difference(tgt_set)
5150 raise errors.OpExecError("Some instances were removed before"
5151 " retrieving their data: %s" % missing)
5152 instance_names = self.wanted
5154 instance_list = [all_info[iname] for iname in instance_names]
5156 # begin data gathering
5158 nodes = frozenset([inst.primary_node for inst in instance_list])
5159 hv_list = list(set([inst.hypervisor for inst in instance_list]))
5163 if self.do_node_query:
5165 node_data = self.rpc.call_all_instances_info(nodes, hv_list)
5167 result = node_data[name]
5169 # offline nodes will be in both lists
5170 off_nodes.append(name)
5172 bad_nodes.append(name)
5175 live_data.update(result.payload)
5176 # else no instance is alive
5178 live_data = dict([(name, {}) for name in instance_names])
5180 # end data gathering
5185 cluster = self.cfg.GetClusterInfo()
5186 for instance in instance_list:
5188 i_hv = cluster.FillHV(instance, skip_globals=True)
5189 i_be = cluster.FillBE(instance)
5190 i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics]
5191 for field in self.op.output_fields:
5192 st_match = self._FIELDS_STATIC.Matches(field)
5193 if field in self._SIMPLE_FIELDS:
5194 val = getattr(instance, field)
5195 elif field == "pnode":
5196 val = instance.primary_node
5197 elif field == "snodes":
5198 val = list(instance.secondary_nodes)
5199 elif field == "admin_state":
5200 val = instance.admin_up
5201 elif field == "oper_state":
5202 if instance.primary_node in bad_nodes:
5205 val = bool(live_data.get(instance.name))
5206 elif field == "status":
5207 if instance.primary_node in off_nodes:
5208 val = "ERROR_nodeoffline"
5209 elif instance.primary_node in bad_nodes:
5210 val = "ERROR_nodedown"
5212 running = bool(live_data.get(instance.name))
5214 if instance.admin_up:
5219 if instance.admin_up:
5223 elif field == "oper_ram":
5224 if instance.primary_node in bad_nodes:
5226 elif instance.name in live_data:
5227 val = live_data[instance.name].get("memory", "?")
5230 elif field == "oper_vcpus":
5231 if instance.primary_node in bad_nodes:
5233 elif instance.name in live_data:
5234 val = live_data[instance.name].get("vcpus", "?")
5237 elif field == "vcpus":
5238 val = i_be[constants.BE_VCPUS]
5239 elif field == "disk_template":
5240 val = instance.disk_template
5243 val = instance.nics[0].ip
5246 elif field == "nic_mode":
5248 val = i_nicp[0][constants.NIC_MODE]
5251 elif field == "nic_link":
5253 val = i_nicp[0][constants.NIC_LINK]
5256 elif field == "bridge":
5257 if (instance.nics and
5258 i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
5259 val = i_nicp[0][constants.NIC_LINK]
5262 elif field == "mac":
5264 val = instance.nics[0].mac
5267 elif field == "sda_size" or field == "sdb_size":
5268 idx = ord(field[2]) - ord('a')
5270 val = instance.FindDisk(idx).size
5271 except errors.OpPrereqError:
5273 elif field == "disk_usage": # total disk usage per node
5274 disk_sizes = [{'size': disk.size} for disk in instance.disks]
5275 val = _ComputeDiskSize(instance.disk_template, disk_sizes)
5276 elif field == "tags":
5277 val = list(instance.GetTags())
5278 elif field == "hvparams":
5280 elif (field.startswith(HVPREFIX) and
5281 field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
5282 field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
5283 val = i_hv.get(field[len(HVPREFIX):], None)
5284 elif field == "beparams":
5286 elif (field.startswith(BEPREFIX) and
5287 field[len(BEPREFIX):] in constants.BES_PARAMETERS):
5288 val = i_be.get(field[len(BEPREFIX):], None)
5289 elif st_match and st_match.groups():
5290 # matches a variable list
5291 st_groups = st_match.groups()
5292 if st_groups and st_groups[0] == "disk":
5293 if st_groups[1] == "count":
5294 val = len(instance.disks)
5295 elif st_groups[1] == "sizes":
5296 val = [disk.size for disk in instance.disks]
5297 elif st_groups[1] == "size":
5299 val = instance.FindDisk(st_groups[2]).size
5300 except errors.OpPrereqError:
5303 assert False, "Unhandled disk parameter"
5304 elif st_groups[0] == "nic":
5305 if st_groups[1] == "count":
5306 val = len(instance.nics)
5307 elif st_groups[1] == "macs":
5308 val = [nic.mac for nic in instance.nics]
5309 elif st_groups[1] == "ips":
5310 val = [nic.ip for nic in instance.nics]
5311 elif st_groups[1] == "modes":
5312 val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
5313 elif st_groups[1] == "links":
5314 val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
5315 elif st_groups[1] == "bridges":
5318 if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
5319 val.append(nicp[constants.NIC_LINK])
5324 nic_idx = int(st_groups[2])
5325 if nic_idx >= len(instance.nics):
5328 if st_groups[1] == "mac":
5329 val = instance.nics[nic_idx].mac
5330 elif st_groups[1] == "ip":
5331 val = instance.nics[nic_idx].ip
5332 elif st_groups[1] == "mode":
5333 val = i_nicp[nic_idx][constants.NIC_MODE]
5334 elif st_groups[1] == "link":
5335 val = i_nicp[nic_idx][constants.NIC_LINK]
5336 elif st_groups[1] == "bridge":
5337 nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
5338 if nic_mode == constants.NIC_MODE_BRIDGED:
5339 val = i_nicp[nic_idx][constants.NIC_LINK]
5343 assert False, "Unhandled NIC parameter"
5345 assert False, ("Declared but unhandled variable parameter '%s'" %
5348 assert False, "Declared but unhandled parameter '%s'" % field
5355 class LUFailoverInstance(LogicalUnit):
5356 """Failover an instance.
5359 HPATH = "instance-failover"
5360 HTYPE = constants.HTYPE_INSTANCE
5363 ("ignore_consistency", False, _TBool),
5368 def ExpandNames(self):
5369 self._ExpandAndLockInstance()
5370 self.needed_locks[locking.LEVEL_NODE] = []
5371 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5373 def DeclareLocks(self, level):
5374 if level == locking.LEVEL_NODE:
5375 self._LockInstancesNodes()
5377 def BuildHooksEnv(self):
5380 This runs on master, primary and secondary nodes of the instance.
5383 instance = self.instance
5384 source_node = instance.primary_node
5385 target_node = instance.secondary_nodes[0]
5387 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5388 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5389 "OLD_PRIMARY": source_node,
5390 "OLD_SECONDARY": target_node,
5391 "NEW_PRIMARY": target_node,
5392 "NEW_SECONDARY": source_node,
5394 env.update(_BuildInstanceHookEnvByObject(self, instance))
5395 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5397 nl_post.append(source_node)
5398 return env, nl, nl_post
5400 def CheckPrereq(self):
5401 """Check prerequisites.
5403 This checks that the instance is in the cluster.
5406 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5407 assert self.instance is not None, \
5408 "Cannot retrieve locked instance %s" % self.op.instance_name
5410 bep = self.cfg.GetClusterInfo().FillBE(instance)
5411 if instance.disk_template not in constants.DTS_NET_MIRROR:
5412 raise errors.OpPrereqError("Instance's disk layout is not"
5413 " network mirrored, cannot failover.",
5416 secondary_nodes = instance.secondary_nodes
5417 if not secondary_nodes:
5418 raise errors.ProgrammerError("no secondary node but using "
5419 "a mirrored disk template")
5421 target_node = secondary_nodes[0]
5422 _CheckNodeOnline(self, target_node)
5423 _CheckNodeNotDrained(self, target_node)
5424 if instance.admin_up:
5425 # check memory requirements on the secondary node
5426 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5427 instance.name, bep[constants.BE_MEMORY],
5428 instance.hypervisor)
5430 self.LogInfo("Not checking memory on the secondary node as"
5431 " instance will not be started")
5433 # check bridge existance
5434 _CheckInstanceBridgesExist(self, instance, node=target_node)
5436 def Exec(self, feedback_fn):
5437 """Failover an instance.
5439 The failover is done by shutting it down on its present node and
5440 starting it on the secondary.
5443 instance = self.instance
5445 source_node = instance.primary_node
5446 target_node = instance.secondary_nodes[0]
5448 if instance.admin_up:
5449 feedback_fn("* checking disk consistency between source and target")
5450 for dev in instance.disks:
5451 # for drbd, these are drbd over lvm
5452 if not _CheckDiskConsistency(self, dev, target_node, False):
5453 if not self.op.ignore_consistency:
5454 raise errors.OpExecError("Disk %s is degraded on target node,"
5455 " aborting failover." % dev.iv_name)
5457 feedback_fn("* not checking disk consistency as instance is not running")
5459 feedback_fn("* shutting down instance on source node")
5460 logging.info("Shutting down instance %s on node %s",
5461 instance.name, source_node)
5463 result = self.rpc.call_instance_shutdown(source_node, instance,
5464 self.op.shutdown_timeout)
5465 msg = result.fail_msg
5467 if self.op.ignore_consistency:
5468 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5469 " Proceeding anyway. Please make sure node"
5470 " %s is down. Error details: %s",
5471 instance.name, source_node, source_node, msg)
5473 raise errors.OpExecError("Could not shutdown instance %s on"
5475 (instance.name, source_node, msg))
5477 feedback_fn("* deactivating the instance's disks on source node")
5478 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5479 raise errors.OpExecError("Can't shut down the instance's disks.")
5481 instance.primary_node = target_node
5482 # distribute new instance config to the other nodes
5483 self.cfg.Update(instance, feedback_fn)
5485 # Only start the instance if it's marked as up
5486 if instance.admin_up:
5487 feedback_fn("* activating the instance's disks on target node")
5488 logging.info("Starting instance %s on node %s",
5489 instance.name, target_node)
5491 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5492 ignore_secondaries=True)
5494 _ShutdownInstanceDisks(self, instance)
5495 raise errors.OpExecError("Can't activate the instance's disks")
5497 feedback_fn("* starting the instance on the target node")
5498 result = self.rpc.call_instance_start(target_node, instance, None, None)
5499 msg = result.fail_msg
5501 _ShutdownInstanceDisks(self, instance)
5502 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5503 (instance.name, target_node, msg))
5506 class LUMigrateInstance(LogicalUnit):
5507 """Migrate an instance.
5509 This is migration without shutting down, compared to the failover,
5510 which is done with shutdown.
5513 HPATH = "instance-migrate"
5514 HTYPE = constants.HTYPE_INSTANCE
5519 ("cleanup", False, _TBool),
5524 def ExpandNames(self):
5525 self._ExpandAndLockInstance()
5527 self.needed_locks[locking.LEVEL_NODE] = []
5528 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5530 self._migrater = TLMigrateInstance(self, self.op.instance_name,
5532 self.tasklets = [self._migrater]
5534 def DeclareLocks(self, level):
5535 if level == locking.LEVEL_NODE:
5536 self._LockInstancesNodes()
5538 def BuildHooksEnv(self):
5541 This runs on master, primary and secondary nodes of the instance.
5544 instance = self._migrater.instance
5545 source_node = instance.primary_node
5546 target_node = instance.secondary_nodes[0]
5547 env = _BuildInstanceHookEnvByObject(self, instance)
5548 env["MIGRATE_LIVE"] = self._migrater.live
5549 env["MIGRATE_CLEANUP"] = self.op.cleanup
5551 "OLD_PRIMARY": source_node,
5552 "OLD_SECONDARY": target_node,
5553 "NEW_PRIMARY": target_node,
5554 "NEW_SECONDARY": source_node,
5556 nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5558 nl_post.append(source_node)
5559 return env, nl, nl_post
5562 class LUMoveInstance(LogicalUnit):
5563 """Move an instance by data-copying.
5566 HPATH = "instance-move"
5567 HTYPE = constants.HTYPE_INSTANCE
5570 ("target_node", _NoDefault, _TNonEmptyString),
5575 def ExpandNames(self):
5576 self._ExpandAndLockInstance()
5577 target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5578 self.op.target_node = target_node
5579 self.needed_locks[locking.LEVEL_NODE] = [target_node]
5580 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5582 def DeclareLocks(self, level):
5583 if level == locking.LEVEL_NODE:
5584 self._LockInstancesNodes(primary_only=True)
5586 def BuildHooksEnv(self):
5589 This runs on master, primary and secondary nodes of the instance.
5593 "TARGET_NODE": self.op.target_node,
5594 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5596 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5597 nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5598 self.op.target_node]
5601 def CheckPrereq(self):
5602 """Check prerequisites.
5604 This checks that the instance is in the cluster.
5607 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5608 assert self.instance is not None, \
5609 "Cannot retrieve locked instance %s" % self.op.instance_name
5611 node = self.cfg.GetNodeInfo(self.op.target_node)
5612 assert node is not None, \
5613 "Cannot retrieve locked node %s" % self.op.target_node
5615 self.target_node = target_node = node.name
5617 if target_node == instance.primary_node:
5618 raise errors.OpPrereqError("Instance %s is already on the node %s" %
5619 (instance.name, target_node),
5622 bep = self.cfg.GetClusterInfo().FillBE(instance)
5624 for idx, dsk in enumerate(instance.disks):
5625 if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5626 raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5627 " cannot copy" % idx, errors.ECODE_STATE)
5629 _CheckNodeOnline(self, target_node)
5630 _CheckNodeNotDrained(self, target_node)
5632 if instance.admin_up:
5633 # check memory requirements on the secondary node
5634 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5635 instance.name, bep[constants.BE_MEMORY],
5636 instance.hypervisor)
5638 self.LogInfo("Not checking memory on the secondary node as"
5639 " instance will not be started")
5641 # check bridge existance
5642 _CheckInstanceBridgesExist(self, instance, node=target_node)
5644 def Exec(self, feedback_fn):
5645 """Move an instance.
5647 The move is done by shutting it down on its present node, copying
5648 the data over (slow) and starting it on the new node.
5651 instance = self.instance
5653 source_node = instance.primary_node
5654 target_node = self.target_node
5656 self.LogInfo("Shutting down instance %s on source node %s",
5657 instance.name, source_node)
5659 result = self.rpc.call_instance_shutdown(source_node, instance,
5660 self.op.shutdown_timeout)
5661 msg = result.fail_msg
5663 if self.op.ignore_consistency:
5664 self.proc.LogWarning("Could not shutdown instance %s on node %s."
5665 " Proceeding anyway. Please make sure node"
5666 " %s is down. Error details: %s",
5667 instance.name, source_node, source_node, msg)
5669 raise errors.OpExecError("Could not shutdown instance %s on"
5671 (instance.name, source_node, msg))
5673 # create the target disks
5675 _CreateDisks(self, instance, target_node=target_node)
5676 except errors.OpExecError:
5677 self.LogWarning("Device creation failed, reverting...")
5679 _RemoveDisks(self, instance, target_node=target_node)
5681 self.cfg.ReleaseDRBDMinors(instance.name)
5684 cluster_name = self.cfg.GetClusterInfo().cluster_name
5687 # activate, get path, copy the data over
5688 for idx, disk in enumerate(instance.disks):
5689 self.LogInfo("Copying data for disk %d", idx)
5690 result = self.rpc.call_blockdev_assemble(target_node, disk,
5691 instance.name, True)
5693 self.LogWarning("Can't assemble newly created disk %d: %s",
5694 idx, result.fail_msg)
5695 errs.append(result.fail_msg)
5697 dev_path = result.payload
5698 result = self.rpc.call_blockdev_export(source_node, disk,
5699 target_node, dev_path,
5702 self.LogWarning("Can't copy data over for disk %d: %s",
5703 idx, result.fail_msg)
5704 errs.append(result.fail_msg)
5708 self.LogWarning("Some disks failed to copy, aborting")
5710 _RemoveDisks(self, instance, target_node=target_node)
5712 self.cfg.ReleaseDRBDMinors(instance.name)
5713 raise errors.OpExecError("Errors during disk copy: %s" %
5716 instance.primary_node = target_node
5717 self.cfg.Update(instance, feedback_fn)
5719 self.LogInfo("Removing the disks on the original node")
5720 _RemoveDisks(self, instance, target_node=source_node)
5722 # Only start the instance if it's marked as up
5723 if instance.admin_up:
5724 self.LogInfo("Starting instance %s on node %s",
5725 instance.name, target_node)
5727 disks_ok, _ = _AssembleInstanceDisks(self, instance,
5728 ignore_secondaries=True)
5730 _ShutdownInstanceDisks(self, instance)
5731 raise errors.OpExecError("Can't activate the instance's disks")
5733 result = self.rpc.call_instance_start(target_node, instance, None, None)
5734 msg = result.fail_msg
5736 _ShutdownInstanceDisks(self, instance)
5737 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5738 (instance.name, target_node, msg))
5741 class LUMigrateNode(LogicalUnit):
5742 """Migrate all instances from a node.
5745 HPATH = "node-migrate"
5746 HTYPE = constants.HTYPE_NODE
5754 def ExpandNames(self):
5755 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5757 self.needed_locks = {
5758 locking.LEVEL_NODE: [self.op.node_name],
5761 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5763 # Create tasklets for migrating instances for all instances on this node
5767 for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5768 logging.debug("Migrating instance %s", inst.name)
5769 names.append(inst.name)
5771 tasklets.append(TLMigrateInstance(self, inst.name, False))
5773 self.tasklets = tasklets
5775 # Declare instance locks
5776 self.needed_locks[locking.LEVEL_INSTANCE] = names
5778 def DeclareLocks(self, level):
5779 if level == locking.LEVEL_NODE:
5780 self._LockInstancesNodes()
5782 def BuildHooksEnv(self):
5785 This runs on the master, the primary and all the secondaries.
5789 "NODE_NAME": self.op.node_name,
5792 nl = [self.cfg.GetMasterNode()]
5794 return (env, nl, nl)
5797 class TLMigrateInstance(Tasklet):
5798 """Tasklet class for instance migration.
5801 @ivar live: whether the migration will be done live or non-live;
5802 this variable is initalized only after CheckPrereq has run
5805 def __init__(self, lu, instance_name, cleanup):
5806 """Initializes this class.
5809 Tasklet.__init__(self, lu)
5812 self.instance_name = instance_name
5813 self.cleanup = cleanup
5814 self.live = False # will be overridden later
5816 def CheckPrereq(self):
5817 """Check prerequisites.
5819 This checks that the instance is in the cluster.
5822 instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5823 instance = self.cfg.GetInstanceInfo(instance_name)
5824 assert instance is not None
5826 if instance.disk_template != constants.DT_DRBD8:
5827 raise errors.OpPrereqError("Instance's disk layout is not"
5828 " drbd8, cannot migrate.", errors.ECODE_STATE)
5830 secondary_nodes = instance.secondary_nodes
5831 if not secondary_nodes:
5832 raise errors.ConfigurationError("No secondary node but using"
5833 " drbd8 disk template")
5835 i_be = self.cfg.GetClusterInfo().FillBE(instance)
5837 target_node = secondary_nodes[0]
5838 # check memory requirements on the secondary node
5839 _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
5840 instance.name, i_be[constants.BE_MEMORY],
5841 instance.hypervisor)
5843 # check bridge existance
5844 _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
5846 if not self.cleanup:
5847 _CheckNodeNotDrained(self.lu, target_node)
5848 result = self.rpc.call_instance_migratable(instance.primary_node,
5850 result.Raise("Can't migrate, please use failover",
5851 prereq=True, ecode=errors.ECODE_STATE)
5853 self.instance = instance
5855 if self.lu.op.live is not None and self.lu.op.mode is not None:
5856 raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
5857 " parameters are accepted",
5859 if self.lu.op.live is not None:
5861 self.lu.op.mode = constants.HT_MIGRATION_LIVE
5863 self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
5864 # reset the 'live' parameter to None so that repeated
5865 # invocations of CheckPrereq do not raise an exception
5866 self.lu.op.live = None
5867 elif self.lu.op.mode is None:
5868 # read the default value from the hypervisor
5869 i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
5870 self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
5872 self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
5874 def _WaitUntilSync(self):
5875 """Poll with custom rpc for disk sync.
5877 This uses our own step-based rpc call.
5880 self.feedback_fn("* wait until resync is done")
5884 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5886 self.instance.disks)
5888 for node, nres in result.items():
5889 nres.Raise("Cannot resync disks on node %s" % node)
5890 node_done, node_percent = nres.payload
5891 all_done = all_done and node_done
5892 if node_percent is not None:
5893 min_percent = min(min_percent, node_percent)
5895 if min_percent < 100:
5896 self.feedback_fn(" - progress: %.1f%%" % min_percent)
5899 def _EnsureSecondary(self, node):
5900 """Demote a node to secondary.
5903 self.feedback_fn("* switching node %s to secondary mode" % node)
5905 for dev in self.instance.disks:
5906 self.cfg.SetDiskID(dev, node)
5908 result = self.rpc.call_blockdev_close(node, self.instance.name,
5909 self.instance.disks)
5910 result.Raise("Cannot change disk to secondary on node %s" % node)
5912 def _GoStandalone(self):
5913 """Disconnect from the network.
5916 self.feedback_fn("* changing into standalone mode")
5917 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5918 self.instance.disks)
5919 for node, nres in result.items():
5920 nres.Raise("Cannot disconnect disks node %s" % node)
5922 def _GoReconnect(self, multimaster):
5923 """Reconnect to the network.
5929 msg = "single-master"
5930 self.feedback_fn("* changing disks into %s mode" % msg)
5931 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5932 self.instance.disks,
5933 self.instance.name, multimaster)
5934 for node, nres in result.items():
5935 nres.Raise("Cannot change disks config on node %s" % node)
5937 def _ExecCleanup(self):
5938 """Try to cleanup after a failed migration.
5940 The cleanup is done by:
5941 - check that the instance is running only on one node
5942 (and update the config if needed)
5943 - change disks on its secondary node to secondary
5944 - wait until disks are fully synchronized
5945 - disconnect from the network
5946 - change disks into single-master mode
5947 - wait again until disks are fully synchronized
5950 instance = self.instance
5951 target_node = self.target_node
5952 source_node = self.source_node
5954 # check running on only one node
5955 self.feedback_fn("* checking where the instance actually runs"
5956 " (if this hangs, the hypervisor might be in"
5958 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5959 for node, result in ins_l.items():
5960 result.Raise("Can't contact node %s" % node)
5962 runningon_source = instance.name in ins_l[source_node].payload
5963 runningon_target = instance.name in ins_l[target_node].payload
5965 if runningon_source and runningon_target:
5966 raise errors.OpExecError("Instance seems to be running on two nodes,"
5967 " or the hypervisor is confused. You will have"
5968 " to ensure manually that it runs only on one"
5969 " and restart this operation.")
5971 if not (runningon_source or runningon_target):
5972 raise errors.OpExecError("Instance does not seem to be running at all."
5973 " In this case, it's safer to repair by"
5974 " running 'gnt-instance stop' to ensure disk"
5975 " shutdown, and then restarting it.")
5977 if runningon_target:
5978 # the migration has actually succeeded, we need to update the config
5979 self.feedback_fn("* instance running on secondary node (%s),"
5980 " updating config" % target_node)
5981 instance.primary_node = target_node
5982 self.cfg.Update(instance, self.feedback_fn)
5983 demoted_node = source_node
5985 self.feedback_fn("* instance confirmed to be running on its"
5986 " primary node (%s)" % source_node)
5987 demoted_node = target_node
5989 self._EnsureSecondary(demoted_node)
5991 self._WaitUntilSync()
5992 except errors.OpExecError:
5993 # we ignore here errors, since if the device is standalone, it
5994 # won't be able to sync
5996 self._GoStandalone()
5997 self._GoReconnect(False)
5998 self._WaitUntilSync()
6000 self.feedback_fn("* done")
6002 def _RevertDiskStatus(self):
6003 """Try to revert the disk status after a failed migration.
6006 target_node = self.target_node
6008 self._EnsureSecondary(target_node)
6009 self._GoStandalone()
6010 self._GoReconnect(False)
6011 self._WaitUntilSync()
6012 except errors.OpExecError, err:
6013 self.lu.LogWarning("Migration failed and I can't reconnect the"
6014 " drives: error '%s'\n"
6015 "Please look and recover the instance status" %
6018 def _AbortMigration(self):
6019 """Call the hypervisor code to abort a started migration.
6022 instance = self.instance
6023 target_node = self.target_node
6024 migration_info = self.migration_info
6026 abort_result = self.rpc.call_finalize_migration(target_node,
6030 abort_msg = abort_result.fail_msg
6032 logging.error("Aborting migration failed on target node %s: %s",
6033 target_node, abort_msg)
6034 # Don't raise an exception here, as we stil have to try to revert the
6035 # disk status, even if this step failed.
6037 def _ExecMigration(self):
6038 """Migrate an instance.
6040 The migrate is done by:
6041 - change the disks into dual-master mode
6042 - wait until disks are fully synchronized again
6043 - migrate the instance
6044 - change disks on the new secondary node (the old primary) to secondary
6045 - wait until disks are fully synchronized
6046 - change disks into single-master mode
6049 instance = self.instance
6050 target_node = self.target_node
6051 source_node = self.source_node
6053 self.feedback_fn("* checking disk consistency between source and target")
6054 for dev in instance.disks:
6055 if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6056 raise errors.OpExecError("Disk %s is degraded or not fully"
6057 " synchronized on target node,"
6058 " aborting migrate." % dev.iv_name)
6060 # First get the migration information from the remote node
6061 result = self.rpc.call_migration_info(source_node, instance)
6062 msg = result.fail_msg
6064 log_err = ("Failed fetching source migration information from %s: %s" %
6066 logging.error(log_err)
6067 raise errors.OpExecError(log_err)
6069 self.migration_info = migration_info = result.payload
6071 # Then switch the disks to master/master mode
6072 self._EnsureSecondary(target_node)
6073 self._GoStandalone()
6074 self._GoReconnect(True)
6075 self._WaitUntilSync()
6077 self.feedback_fn("* preparing %s to accept the instance" % target_node)
6078 result = self.rpc.call_accept_instance(target_node,
6081 self.nodes_ip[target_node])
6083 msg = result.fail_msg
6085 logging.error("Instance pre-migration failed, trying to revert"
6086 " disk status: %s", msg)
6087 self.feedback_fn("Pre-migration failed, aborting")
6088 self._AbortMigration()
6089 self._RevertDiskStatus()
6090 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6091 (instance.name, msg))
6093 self.feedback_fn("* migrating instance to %s" % target_node)
6095 result = self.rpc.call_instance_migrate(source_node, instance,
6096 self.nodes_ip[target_node],
6098 msg = result.fail_msg
6100 logging.error("Instance migration failed, trying to revert"
6101 " disk status: %s", msg)
6102 self.feedback_fn("Migration failed, aborting")
6103 self._AbortMigration()
6104 self._RevertDiskStatus()
6105 raise errors.OpExecError("Could not migrate instance %s: %s" %
6106 (instance.name, msg))
6109 instance.primary_node = target_node
6110 # distribute new instance config to the other nodes
6111 self.cfg.Update(instance, self.feedback_fn)
6113 result = self.rpc.call_finalize_migration(target_node,
6117 msg = result.fail_msg
6119 logging.error("Instance migration succeeded, but finalization failed:"
6121 raise errors.OpExecError("Could not finalize instance migration: %s" %
6124 self._EnsureSecondary(source_node)
6125 self._WaitUntilSync()
6126 self._GoStandalone()
6127 self._GoReconnect(False)
6128 self._WaitUntilSync()
6130 self.feedback_fn("* done")
6132 def Exec(self, feedback_fn):
6133 """Perform the migration.
6136 feedback_fn("Migrating instance %s" % self.instance.name)
6138 self.feedback_fn = feedback_fn
6140 self.source_node = self.instance.primary_node
6141 self.target_node = self.instance.secondary_nodes[0]
6142 self.all_nodes = [self.source_node, self.target_node]
6144 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6145 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6149 return self._ExecCleanup()
6151 return self._ExecMigration()
6154 def _CreateBlockDev(lu, node, instance, device, force_create,
6156 """Create a tree of block devices on a given node.
6158 If this device type has to be created on secondaries, create it and
6161 If not, just recurse to children keeping the same 'force' value.
6163 @param lu: the lu on whose behalf we execute
6164 @param node: the node on which to create the device
6165 @type instance: L{objects.Instance}
6166 @param instance: the instance which owns the device
6167 @type device: L{objects.Disk}
6168 @param device: the device to create
6169 @type force_create: boolean
6170 @param force_create: whether to force creation of this device; this
6171 will be change to True whenever we find a device which has
6172 CreateOnSecondary() attribute
6173 @param info: the extra 'metadata' we should attach to the device
6174 (this will be represented as a LVM tag)
6175 @type force_open: boolean
6176 @param force_open: this parameter will be passes to the
6177 L{backend.BlockdevCreate} function where it specifies
6178 whether we run on primary or not, and it affects both
6179 the child assembly and the device own Open() execution
6182 if device.CreateOnSecondary():
6186 for child in device.children:
6187 _CreateBlockDev(lu, node, instance, child, force_create,
6190 if not force_create:
6193 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6196 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6197 """Create a single block device on a given node.
6199 This will not recurse over children of the device, so they must be
6202 @param lu: the lu on whose behalf we execute
6203 @param node: the node on which to create the device
6204 @type instance: L{objects.Instance}
6205 @param instance: the instance which owns the device
6206 @type device: L{objects.Disk}
6207 @param device: the device to create
6208 @param info: the extra 'metadata' we should attach to the device
6209 (this will be represented as a LVM tag)
6210 @type force_open: boolean
6211 @param force_open: this parameter will be passes to the
6212 L{backend.BlockdevCreate} function where it specifies
6213 whether we run on primary or not, and it affects both
6214 the child assembly and the device own Open() execution
6217 lu.cfg.SetDiskID(device, node)
6218 result = lu.rpc.call_blockdev_create(node, device, device.size,
6219 instance.name, force_open, info)
6220 result.Raise("Can't create block device %s on"
6221 " node %s for instance %s" % (device, node, instance.name))
6222 if device.physical_id is None:
6223 device.physical_id = result.payload
6226 def _GenerateUniqueNames(lu, exts):
6227 """Generate a suitable LV name.
6229 This will generate a logical volume name for the given instance.
6234 new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6235 results.append("%s%s" % (new_id, val))
6239 def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
6241 """Generate a drbd8 device complete with its children.
6244 port = lu.cfg.AllocatePort()
6245 vgname = lu.cfg.GetVGName()
6246 shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6247 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6248 logical_id=(vgname, names[0]))
6249 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6250 logical_id=(vgname, names[1]))
6251 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6252 logical_id=(primary, secondary, port,
6255 children=[dev_data, dev_meta],
6260 def _GenerateDiskTemplate(lu, template_name,
6261 instance_name, primary_node,
6262 secondary_nodes, disk_info,
6263 file_storage_dir, file_driver,
6265 """Generate the entire disk layout for a given template type.
6268 #TODO: compute space requirements
6270 vgname = lu.cfg.GetVGName()
6271 disk_count = len(disk_info)
6273 if template_name == constants.DT_DISKLESS:
6275 elif template_name == constants.DT_PLAIN:
6276 if len(secondary_nodes) != 0:
6277 raise errors.ProgrammerError("Wrong template configuration")
6279 names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6280 for i in range(disk_count)])
6281 for idx, disk in enumerate(disk_info):
6282 disk_index = idx + base_index
6283 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6284 logical_id=(vgname, names[idx]),
6285 iv_name="disk/%d" % disk_index,
6287 disks.append(disk_dev)
6288 elif template_name == constants.DT_DRBD8:
6289 if len(secondary_nodes) != 1:
6290 raise errors.ProgrammerError("Wrong template configuration")
6291 remote_node = secondary_nodes[0]
6292 minors = lu.cfg.AllocateDRBDMinor(
6293 [primary_node, remote_node] * len(disk_info), instance_name)
6296 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6297 for i in range(disk_count)]):
6298 names.append(lv_prefix + "_data")
6299 names.append(lv_prefix + "_meta")
6300 for idx, disk in enumerate(disk_info):
6301 disk_index = idx + base_index
6302 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6303 disk["size"], names[idx*2:idx*2+2],
6304 "disk/%d" % disk_index,
6305 minors[idx*2], minors[idx*2+1])
6306 disk_dev.mode = disk["mode"]
6307 disks.append(disk_dev)
6308 elif template_name == constants.DT_FILE:
6309 if len(secondary_nodes) != 0:
6310 raise errors.ProgrammerError("Wrong template configuration")
6312 _RequireFileStorage()
6314 for idx, disk in enumerate(disk_info):
6315 disk_index = idx + base_index
6316 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6317 iv_name="disk/%d" % disk_index,
6318 logical_id=(file_driver,
6319 "%s/disk%d" % (file_storage_dir,
6322 disks.append(disk_dev)
6324 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6328 def _GetInstanceInfoText(instance):
6329 """Compute that text that should be added to the disk's metadata.
6332 return "originstname+%s" % instance.name
6335 def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6336 """Create all disks for an instance.
6338 This abstracts away some work from AddInstance.
6340 @type lu: L{LogicalUnit}
6341 @param lu: the logical unit on whose behalf we execute
6342 @type instance: L{objects.Instance}
6343 @param instance: the instance whose disks we should create
6345 @param to_skip: list of indices to skip
6346 @type target_node: string
6347 @param target_node: if passed, overrides the target node for creation
6349 @return: the success of the creation
6352 info = _GetInstanceInfoText(instance)
6353 if target_node is None:
6354 pnode = instance.primary_node
6355 all_nodes = instance.all_nodes
6360 if instance.disk_template == constants.DT_FILE:
6361 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6362 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6364 result.Raise("Failed to create directory '%s' on"
6365 " node %s" % (file_storage_dir, pnode))
6367 # Note: this needs to be kept in sync with adding of disks in
6368 # LUSetInstanceParams
6369 for idx, device in enumerate(instance.disks):
6370 if to_skip and idx in to_skip:
6372 logging.info("Creating volume %s for instance %s",
6373 device.iv_name, instance.name)
6375 for node in all_nodes:
6376 f_create = node == pnode
6377 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6380 def _RemoveDisks(lu, instance, target_node=None):
6381 """Remove all disks for an instance.
6383 This abstracts away some work from `AddInstance()` and
6384 `RemoveInstance()`. Note that in case some of the devices couldn't
6385 be removed, the removal will continue with the other ones (compare
6386 with `_CreateDisks()`).
6388 @type lu: L{LogicalUnit}
6389 @param lu: the logical unit on whose behalf we execute
6390 @type instance: L{objects.Instance}
6391 @param instance: the instance whose disks we should remove
6392 @type target_node: string
6393 @param target_node: used to override the node on which to remove the disks
6395 @return: the success of the removal
6398 logging.info("Removing block devices for instance %s", instance.name)
6401 for device in instance.disks:
6403 edata = [(target_node, device)]
6405 edata = device.ComputeNodeTree(instance.primary_node)
6406 for node, disk in edata:
6407 lu.cfg.SetDiskID(disk, node)
6408 msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6410 lu.LogWarning("Could not remove block device %s on node %s,"
6411 " continuing anyway: %s", device.iv_name, node, msg)
6414 if instance.disk_template == constants.DT_FILE:
6415 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6419 tgt = instance.primary_node
6420 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6422 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6423 file_storage_dir, instance.primary_node, result.fail_msg)
6429 def _ComputeDiskSize(disk_template, disks):
6430 """Compute disk size requirements in the volume group
6433 # Required free disk space as a function of disk and swap space
6435 constants.DT_DISKLESS: None,
6436 constants.DT_PLAIN: sum(d["size"] for d in disks),
6437 # 128 MB are added for drbd metadata for each disk
6438 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6439 constants.DT_FILE: None,
6442 if disk_template not in req_size_dict:
6443 raise errors.ProgrammerError("Disk template '%s' size requirement"
6444 " is unknown" % disk_template)
6446 return req_size_dict[disk_template]
6449 def _CheckHVParams(lu, nodenames, hvname, hvparams):
6450 """Hypervisor parameter validation.
6452 This function abstract the hypervisor parameter validation to be
6453 used in both instance create and instance modify.
6455 @type lu: L{LogicalUnit}
6456 @param lu: the logical unit for which we check
6457 @type nodenames: list
6458 @param nodenames: the list of nodes on which we should check
6459 @type hvname: string
6460 @param hvname: the name of the hypervisor we should use
6461 @type hvparams: dict
6462 @param hvparams: the parameters which we need to check
6463 @raise errors.OpPrereqError: if the parameters are not valid
6466 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6469 for node in nodenames:
6473 info.Raise("Hypervisor parameter validation failed on node %s" % node)
6476 def _CheckOSParams(lu, required, nodenames, osname, osparams):
6477 """OS parameters validation.
6479 @type lu: L{LogicalUnit}
6480 @param lu: the logical unit for which we check
6481 @type required: boolean
6482 @param required: whether the validation should fail if the OS is not
6484 @type nodenames: list
6485 @param nodenames: the list of nodes on which we should check
6486 @type osname: string
6487 @param osname: the name of the hypervisor we should use
6488 @type osparams: dict
6489 @param osparams: the parameters which we need to check
6490 @raise errors.OpPrereqError: if the parameters are not valid
6493 result = lu.rpc.call_os_validate(required, nodenames, osname,
6494 [constants.OS_VALIDATE_PARAMETERS],
6496 for node, nres in result.items():
6497 # we don't check for offline cases since this should be run only
6498 # against the master node and/or an instance's nodes
6499 nres.Raise("OS Parameters validation failed on node %s" % node)
6500 if not nres.payload:
6501 lu.LogInfo("OS %s not found on node %s, validation skipped",
6505 class LUCreateInstance(LogicalUnit):
6506 """Create an instance.
6509 HPATH = "instance-add"
6510 HTYPE = constants.HTYPE_INSTANCE
6513 ("mode", _NoDefault, _TElemOf(constants.INSTANCE_CREATE_MODES)),
6514 ("start", True, _TBool),
6515 ("wait_for_sync", True, _TBool),
6516 ("ip_check", True, _TBool),
6517 ("name_check", True, _TBool),
6518 ("disks", _NoDefault, _TListOf(_TDict)),
6519 ("nics", _NoDefault, _TListOf(_TDict)),
6520 ("hvparams", _EmptyDict, _TDict),
6521 ("beparams", _EmptyDict, _TDict),
6522 ("osparams", _EmptyDict, _TDict),
6523 ("no_install", None, _TMaybeBool),
6524 ("os_type", None, _TMaybeString),
6525 ("force_variant", False, _TBool),
6526 ("source_handshake", None, _TOr(_TList, _TNone)),
6527 ("source_x509_ca", None, _TMaybeString),
6528 ("source_instance_name", None, _TMaybeString),
6529 ("src_node", None, _TMaybeString),
6530 ("src_path", None, _TMaybeString),
6531 ("pnode", None, _TMaybeString),
6532 ("snode", None, _TMaybeString),
6533 ("iallocator", None, _TMaybeString),
6534 ("hypervisor", None, _TMaybeString),
6535 ("disk_template", _NoDefault, _CheckDiskTemplate),
6536 ("identify_defaults", False, _TBool),
6537 ("file_driver", None, _TOr(_TNone, _TElemOf(constants.FILE_DRIVER))),
6538 ("file_storage_dir", None, _TMaybeString),
6539 ("dry_run", False, _TBool),
6543 def CheckArguments(self):
6547 # do not require name_check to ease forward/backward compatibility
6549 if self.op.no_install and self.op.start:
6550 self.LogInfo("No-installation mode selected, disabling startup")
6551 self.op.start = False
6552 # validate/normalize the instance name
6553 self.op.instance_name = \
6554 netutils.HostInfo.NormalizeName(self.op.instance_name)
6556 if self.op.ip_check and not self.op.name_check:
6557 # TODO: make the ip check more flexible and not depend on the name check
6558 raise errors.OpPrereqError("Cannot do ip check without a name check",
6561 # check nics' parameter names
6562 for nic in self.op.nics:
6563 utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6565 # check disks. parameter names and consistent adopt/no-adopt strategy
6566 has_adopt = has_no_adopt = False
6567 for disk in self.op.disks:
6568 utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6573 if has_adopt and has_no_adopt:
6574 raise errors.OpPrereqError("Either all disks are adopted or none is",
6577 if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6578 raise errors.OpPrereqError("Disk adoption is not supported for the"
6579 " '%s' disk template" %
6580 self.op.disk_template,
6582 if self.op.iallocator is not None:
6583 raise errors.OpPrereqError("Disk adoption not allowed with an"
6584 " iallocator script", errors.ECODE_INVAL)
6585 if self.op.mode == constants.INSTANCE_IMPORT:
6586 raise errors.OpPrereqError("Disk adoption not allowed for"
6587 " instance import", errors.ECODE_INVAL)
6589 self.adopt_disks = has_adopt
6591 # instance name verification
6592 if self.op.name_check:
6593 self.hostname1 = netutils.GetHostInfo(self.op.instance_name)
6594 self.op.instance_name = self.hostname1.name
6595 # used in CheckPrereq for ip ping check
6596 self.check_ip = self.hostname1.ip
6597 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6598 raise errors.OpPrereqError("Remote imports require names to be checked" %
6601 self.check_ip = None
6603 # file storage checks
6604 if (self.op.file_driver and
6605 not self.op.file_driver in constants.FILE_DRIVER):
6606 raise errors.OpPrereqError("Invalid file driver name '%s'" %
6607 self.op.file_driver, errors.ECODE_INVAL)
6609 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6610 raise errors.OpPrereqError("File storage directory path not absolute",
6613 ### Node/iallocator related checks
6614 _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6616 self._cds = _GetClusterDomainSecret()
6618 if self.op.mode == constants.INSTANCE_IMPORT:
6619 # On import force_variant must be True, because if we forced it at
6620 # initial install, our only chance when importing it back is that it
6622 self.op.force_variant = True
6624 if self.op.no_install:
6625 self.LogInfo("No-installation mode has no effect during import")
6627 elif self.op.mode == constants.INSTANCE_CREATE:
6628 if self.op.os_type is None:
6629 raise errors.OpPrereqError("No guest OS specified",
6631 if self.op.disk_template is None:
6632 raise errors.OpPrereqError("No disk template specified",
6635 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6636 # Check handshake to ensure both clusters have the same domain secret
6637 src_handshake = self.op.source_handshake
6638 if not src_handshake:
6639 raise errors.OpPrereqError("Missing source handshake",
6642 errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6645 raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6648 # Load and check source CA
6649 self.source_x509_ca_pem = self.op.source_x509_ca
6650 if not self.source_x509_ca_pem:
6651 raise errors.OpPrereqError("Missing source X509 CA",
6655 (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6657 except OpenSSL.crypto.Error, err:
6658 raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6659 (err, ), errors.ECODE_INVAL)
6661 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6662 if errcode is not None:
6663 raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6666 self.source_x509_ca = cert
6668 src_instance_name = self.op.source_instance_name
6669 if not src_instance_name:
6670 raise errors.OpPrereqError("Missing source instance name",
6673 norm_name = netutils.HostInfo.NormalizeName(src_instance_name)
6674 self.source_instance_name = netutils.GetHostInfo(norm_name).name
6677 raise errors.OpPrereqError("Invalid instance creation mode %r" %
6678 self.op.mode, errors.ECODE_INVAL)
6680 def ExpandNames(self):
6681 """ExpandNames for CreateInstance.
6683 Figure out the right locks for instance creation.
6686 self.needed_locks = {}
6688 instance_name = self.op.instance_name
6689 # this is just a preventive check, but someone might still add this
6690 # instance in the meantime, and creation will fail at lock-add time
6691 if instance_name in self.cfg.GetInstanceList():
6692 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6693 instance_name, errors.ECODE_EXISTS)
6695 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6697 if self.op.iallocator:
6698 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6700 self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6701 nodelist = [self.op.pnode]
6702 if self.op.snode is not None:
6703 self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6704 nodelist.append(self.op.snode)
6705 self.needed_locks[locking.LEVEL_NODE] = nodelist
6707 # in case of import lock the source node too
6708 if self.op.mode == constants.INSTANCE_IMPORT:
6709 src_node = self.op.src_node
6710 src_path = self.op.src_path
6712 if src_path is None:
6713 self.op.src_path = src_path = self.op.instance_name
6715 if src_node is None:
6716 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6717 self.op.src_node = None
6718 if os.path.isabs(src_path):
6719 raise errors.OpPrereqError("Importing an instance from an absolute"
6720 " path requires a source node option.",
6723 self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6724 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6725 self.needed_locks[locking.LEVEL_NODE].append(src_node)
6726 if not os.path.isabs(src_path):
6727 self.op.src_path = src_path = \
6728 utils.PathJoin(constants.EXPORT_DIR, src_path)
6730 def _RunAllocator(self):
6731 """Run the allocator based on input opcode.
6734 nics = [n.ToDict() for n in self.nics]
6735 ial = IAllocator(self.cfg, self.rpc,
6736 mode=constants.IALLOCATOR_MODE_ALLOC,
6737 name=self.op.instance_name,
6738 disk_template=self.op.disk_template,
6741 vcpus=self.be_full[constants.BE_VCPUS],
6742 mem_size=self.be_full[constants.BE_MEMORY],
6745 hypervisor=self.op.hypervisor,
6748 ial.Run(self.op.iallocator)
6751 raise errors.OpPrereqError("Can't compute nodes using"
6752 " iallocator '%s': %s" %
6753 (self.op.iallocator, ial.info),
6755 if len(ial.result) != ial.required_nodes:
6756 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6757 " of nodes (%s), required %s" %
6758 (self.op.iallocator, len(ial.result),
6759 ial.required_nodes), errors.ECODE_FAULT)
6760 self.op.pnode = ial.result[0]
6761 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6762 self.op.instance_name, self.op.iallocator,
6763 utils.CommaJoin(ial.result))
6764 if ial.required_nodes == 2:
6765 self.op.snode = ial.result[1]
6767 def BuildHooksEnv(self):
6770 This runs on master, primary and secondary nodes of the instance.
6774 "ADD_MODE": self.op.mode,
6776 if self.op.mode == constants.INSTANCE_IMPORT:
6777 env["SRC_NODE"] = self.op.src_node
6778 env["SRC_PATH"] = self.op.src_path
6779 env["SRC_IMAGES"] = self.src_images
6781 env.update(_BuildInstanceHookEnv(
6782 name=self.op.instance_name,
6783 primary_node=self.op.pnode,
6784 secondary_nodes=self.secondaries,
6785 status=self.op.start,
6786 os_type=self.op.os_type,
6787 memory=self.be_full[constants.BE_MEMORY],
6788 vcpus=self.be_full[constants.BE_VCPUS],
6789 nics=_NICListToTuple(self, self.nics),
6790 disk_template=self.op.disk_template,
6791 disks=[(d["size"], d["mode"]) for d in self.disks],
6794 hypervisor_name=self.op.hypervisor,
6797 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6801 def _ReadExportInfo(self):
6802 """Reads the export information from disk.
6804 It will override the opcode source node and path with the actual
6805 information, if these two were not specified before.
6807 @return: the export information
6810 assert self.op.mode == constants.INSTANCE_IMPORT
6812 src_node = self.op.src_node
6813 src_path = self.op.src_path
6815 if src_node is None:
6816 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6817 exp_list = self.rpc.call_export_list(locked_nodes)
6819 for node in exp_list:
6820 if exp_list[node].fail_msg:
6822 if src_path in exp_list[node].payload:
6824 self.op.src_node = src_node = node
6825 self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6829 raise errors.OpPrereqError("No export found for relative path %s" %
6830 src_path, errors.ECODE_INVAL)
6832 _CheckNodeOnline(self, src_node)
6833 result = self.rpc.call_export_info(src_node, src_path)
6834 result.Raise("No export or invalid export found in dir %s" % src_path)
6836 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6837 if not export_info.has_section(constants.INISECT_EXP):
6838 raise errors.ProgrammerError("Corrupted export config",
6839 errors.ECODE_ENVIRON)
6841 ei_version = export_info.get(constants.INISECT_EXP, "version")
6842 if (int(ei_version) != constants.EXPORT_VERSION):
6843 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6844 (ei_version, constants.EXPORT_VERSION),
6845 errors.ECODE_ENVIRON)
6848 def _ReadExportParams(self, einfo):
6849 """Use export parameters as defaults.
6851 In case the opcode doesn't specify (as in override) some instance
6852 parameters, then try to use them from the export information, if
6856 self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6858 if self.op.disk_template is None:
6859 if einfo.has_option(constants.INISECT_INS, "disk_template"):
6860 self.op.disk_template = einfo.get(constants.INISECT_INS,
6863 raise errors.OpPrereqError("No disk template specified and the export"
6864 " is missing the disk_template information",
6867 if not self.op.disks:
6868 if einfo.has_option(constants.INISECT_INS, "disk_count"):
6870 # TODO: import the disk iv_name too
6871 for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6872 disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6873 disks.append({"size": disk_sz})
6874 self.op.disks = disks
6876 raise errors.OpPrereqError("No disk info specified and the export"
6877 " is missing the disk information",
6880 if (not self.op.nics and
6881 einfo.has_option(constants.INISECT_INS, "nic_count")):
6883 for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6885 for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6886 v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6891 if (self.op.hypervisor is None and
6892 einfo.has_option(constants.INISECT_INS, "hypervisor")):
6893 self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6894 if einfo.has_section(constants.INISECT_HYP):
6895 # use the export parameters but do not override the ones
6896 # specified by the user
6897 for name, value in einfo.items(constants.INISECT_HYP):
6898 if name not in self.op.hvparams:
6899 self.op.hvparams[name] = value
6901 if einfo.has_section(constants.INISECT_BEP):
6902 # use the parameters, without overriding
6903 for name, value in einfo.items(constants.INISECT_BEP):
6904 if name not in self.op.beparams:
6905 self.op.beparams[name] = value
6907 # try to read the parameters old style, from the main section
6908 for name in constants.BES_PARAMETERS:
6909 if (name not in self.op.beparams and
6910 einfo.has_option(constants.INISECT_INS, name)):
6911 self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6913 if einfo.has_section(constants.INISECT_OSP):
6914 # use the parameters, without overriding
6915 for name, value in einfo.items(constants.INISECT_OSP):
6916 if name not in self.op.osparams:
6917 self.op.osparams[name] = value
6919 def _RevertToDefaults(self, cluster):
6920 """Revert the instance parameters to the default values.
6924 hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
6925 for name in self.op.hvparams.keys():
6926 if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
6927 del self.op.hvparams[name]
6929 be_defs = cluster.SimpleFillBE({})
6930 for name in self.op.beparams.keys():
6931 if name in be_defs and be_defs[name] == self.op.beparams[name]:
6932 del self.op.beparams[name]
6934 nic_defs = cluster.SimpleFillNIC({})
6935 for nic in self.op.nics:
6936 for name in constants.NICS_PARAMETERS:
6937 if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
6940 os_defs = cluster.SimpleFillOS(self.op.os_type, {})
6941 for name in self.op.osparams.keys():
6942 if name in os_defs and os_defs[name] == self.op.osparams[name]:
6943 del self.op.osparams[name]
6945 def CheckPrereq(self):
6946 """Check prerequisites.
6949 if self.op.mode == constants.INSTANCE_IMPORT:
6950 export_info = self._ReadExportInfo()
6951 self._ReadExportParams(export_info)
6953 _CheckDiskTemplate(self.op.disk_template)
6955 if (not self.cfg.GetVGName() and
6956 self.op.disk_template not in constants.DTS_NOT_LVM):
6957 raise errors.OpPrereqError("Cluster does not support lvm-based"
6958 " instances", errors.ECODE_STATE)
6960 if self.op.hypervisor is None:
6961 self.op.hypervisor = self.cfg.GetHypervisorType()
6963 cluster = self.cfg.GetClusterInfo()
6964 enabled_hvs = cluster.enabled_hypervisors
6965 if self.op.hypervisor not in enabled_hvs:
6966 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
6967 " cluster (%s)" % (self.op.hypervisor,
6968 ",".join(enabled_hvs)),
6971 # check hypervisor parameter syntax (locally)
6972 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6973 filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
6975 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
6976 hv_type.CheckParameterSyntax(filled_hvp)
6977 self.hv_full = filled_hvp
6978 # check that we don't specify global parameters on an instance
6979 _CheckGlobalHvParams(self.op.hvparams)
6981 # fill and remember the beparams dict
6982 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6983 self.be_full = cluster.SimpleFillBE(self.op.beparams)
6985 # build os parameters
6986 self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
6988 # now that hvp/bep are in final format, let's reset to defaults,
6990 if self.op.identify_defaults:
6991 self._RevertToDefaults(cluster)
6995 for idx, nic in enumerate(self.op.nics):
6996 nic_mode_req = nic.get("mode", None)
6997 nic_mode = nic_mode_req
6998 if nic_mode is None:
6999 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7001 # in routed mode, for the first nic, the default ip is 'auto'
7002 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7003 default_ip_mode = constants.VALUE_AUTO
7005 default_ip_mode = constants.VALUE_NONE
7007 # ip validity checks
7008 ip = nic.get("ip", default_ip_mode)
7009 if ip is None or ip.lower() == constants.VALUE_NONE:
7011 elif ip.lower() == constants.VALUE_AUTO:
7012 if not self.op.name_check:
7013 raise errors.OpPrereqError("IP address set to auto but name checks"
7014 " have been skipped. Aborting.",
7016 nic_ip = self.hostname1.ip
7018 if not netutils.IP4Address.IsValid(ip):
7019 raise errors.OpPrereqError("Given IP address '%s' doesn't look"
7020 " like a valid IP" % ip,
7024 # TODO: check the ip address for uniqueness
7025 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7026 raise errors.OpPrereqError("Routed nic mode requires an ip address",
7029 # MAC address verification
7030 mac = nic.get("mac", constants.VALUE_AUTO)
7031 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7032 mac = utils.NormalizeAndValidateMac(mac)
7035 self.cfg.ReserveMAC(mac, self.proc.GetECId())
7036 except errors.ReservationError:
7037 raise errors.OpPrereqError("MAC address %s already in use"
7038 " in cluster" % mac,
7039 errors.ECODE_NOTUNIQUE)
7041 # bridge verification
7042 bridge = nic.get("bridge", None)
7043 link = nic.get("link", None)
7045 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7046 " at the same time", errors.ECODE_INVAL)
7047 elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7048 raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7055 nicparams[constants.NIC_MODE] = nic_mode_req
7057 nicparams[constants.NIC_LINK] = link
7059 check_params = cluster.SimpleFillNIC(nicparams)
7060 objects.NIC.CheckParameterSyntax(check_params)
7061 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7063 # disk checks/pre-build
7065 for disk in self.op.disks:
7066 mode = disk.get("mode", constants.DISK_RDWR)
7067 if mode not in constants.DISK_ACCESS_SET:
7068 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7069 mode, errors.ECODE_INVAL)
7070 size = disk.get("size", None)
7072 raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7075 except (TypeError, ValueError):
7076 raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7078 new_disk = {"size": size, "mode": mode}
7080 new_disk["adopt"] = disk["adopt"]
7081 self.disks.append(new_disk)
7083 if self.op.mode == constants.INSTANCE_IMPORT:
7085 # Check that the new instance doesn't have less disks than the export
7086 instance_disks = len(self.disks)
7087 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7088 if instance_disks < export_disks:
7089 raise errors.OpPrereqError("Not enough disks to import."
7090 " (instance: %d, export: %d)" %
7091 (instance_disks, export_disks),
7095 for idx in range(export_disks):
7096 option = 'disk%d_dump' % idx
7097 if export_info.has_option(constants.INISECT_INS, option):
7098 # FIXME: are the old os-es, disk sizes, etc. useful?
7099 export_name = export_info.get(constants.INISECT_INS, option)
7100 image = utils.PathJoin(self.op.src_path, export_name)
7101 disk_images.append(image)
7103 disk_images.append(False)
7105 self.src_images = disk_images
7107 old_name = export_info.get(constants.INISECT_INS, 'name')
7109 exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7110 except (TypeError, ValueError), err:
7111 raise errors.OpPrereqError("Invalid export file, nic_count is not"
7112 " an integer: %s" % str(err),
7114 if self.op.instance_name == old_name:
7115 for idx, nic in enumerate(self.nics):
7116 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7117 nic_mac_ini = 'nic%d_mac' % idx
7118 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7120 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7122 # ip ping checks (we use the same ip that was resolved in ExpandNames)
7123 if self.op.ip_check:
7124 if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7125 raise errors.OpPrereqError("IP %s of instance %s already in use" %
7126 (self.check_ip, self.op.instance_name),
7127 errors.ECODE_NOTUNIQUE)
7129 #### mac address generation
7130 # By generating here the mac address both the allocator and the hooks get
7131 # the real final mac address rather than the 'auto' or 'generate' value.
7132 # There is a race condition between the generation and the instance object
7133 # creation, which means that we know the mac is valid now, but we're not
7134 # sure it will be when we actually add the instance. If things go bad
7135 # adding the instance will abort because of a duplicate mac, and the
7136 # creation job will fail.
7137 for nic in self.nics:
7138 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7139 nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7143 if self.op.iallocator is not None:
7144 self._RunAllocator()
7146 #### node related checks
7148 # check primary node
7149 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7150 assert self.pnode is not None, \
7151 "Cannot retrieve locked node %s" % self.op.pnode
7153 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7154 pnode.name, errors.ECODE_STATE)
7156 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7157 pnode.name, errors.ECODE_STATE)
7159 self.secondaries = []
7161 # mirror node verification
7162 if self.op.disk_template in constants.DTS_NET_MIRROR:
7163 if self.op.snode is None:
7164 raise errors.OpPrereqError("The networked disk templates need"
7165 " a mirror node", errors.ECODE_INVAL)
7166 if self.op.snode == pnode.name:
7167 raise errors.OpPrereqError("The secondary node cannot be the"
7168 " primary node.", errors.ECODE_INVAL)
7169 _CheckNodeOnline(self, self.op.snode)
7170 _CheckNodeNotDrained(self, self.op.snode)
7171 self.secondaries.append(self.op.snode)
7173 nodenames = [pnode.name] + self.secondaries
7175 req_size = _ComputeDiskSize(self.op.disk_template,
7178 # Check lv size requirements, if not adopting
7179 if req_size is not None and not self.adopt_disks:
7180 _CheckNodesFreeDisk(self, nodenames, req_size)
7182 if self.adopt_disks: # instead, we must check the adoption data
7183 all_lvs = set([i["adopt"] for i in self.disks])
7184 if len(all_lvs) != len(self.disks):
7185 raise errors.OpPrereqError("Duplicate volume names given for adoption",
7187 for lv_name in all_lvs:
7189 self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7190 except errors.ReservationError:
7191 raise errors.OpPrereqError("LV named %s used by another instance" %
7192 lv_name, errors.ECODE_NOTUNIQUE)
7194 node_lvs = self.rpc.call_lv_list([pnode.name],
7195 self.cfg.GetVGName())[pnode.name]
7196 node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7197 node_lvs = node_lvs.payload
7198 delta = all_lvs.difference(node_lvs.keys())
7200 raise errors.OpPrereqError("Missing logical volume(s): %s" %
7201 utils.CommaJoin(delta),
7203 online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7205 raise errors.OpPrereqError("Online logical volumes found, cannot"
7206 " adopt: %s" % utils.CommaJoin(online_lvs),
7208 # update the size of disk based on what is found
7209 for dsk in self.disks:
7210 dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
7212 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7214 _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7215 # check OS parameters (remotely)
7216 _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7218 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7220 # memory check on primary node
7222 _CheckNodeFreeMemory(self, self.pnode.name,
7223 "creating instance %s" % self.op.instance_name,
7224 self.be_full[constants.BE_MEMORY],
7227 self.dry_run_result = list(nodenames)
7229 def Exec(self, feedback_fn):
7230 """Create and add the instance to the cluster.
7233 instance = self.op.instance_name
7234 pnode_name = self.pnode.name
7236 ht_kind = self.op.hypervisor
7237 if ht_kind in constants.HTS_REQ_PORT:
7238 network_port = self.cfg.AllocatePort()
7242 if constants.ENABLE_FILE_STORAGE:
7243 # this is needed because os.path.join does not accept None arguments
7244 if self.op.file_storage_dir is None:
7245 string_file_storage_dir = ""
7247 string_file_storage_dir = self.op.file_storage_dir
7249 # build the full file storage dir path
7250 file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7251 string_file_storage_dir, instance)
7253 file_storage_dir = ""
7255 disks = _GenerateDiskTemplate(self,
7256 self.op.disk_template,
7257 instance, pnode_name,
7261 self.op.file_driver,
7264 iobj = objects.Instance(name=instance, os=self.op.os_type,
7265 primary_node=pnode_name,
7266 nics=self.nics, disks=disks,
7267 disk_template=self.op.disk_template,
7269 network_port=network_port,
7270 beparams=self.op.beparams,
7271 hvparams=self.op.hvparams,
7272 hypervisor=self.op.hypervisor,
7273 osparams=self.op.osparams,
7276 if self.adopt_disks:
7277 # rename LVs to the newly-generated names; we need to construct
7278 # 'fake' LV disks with the old data, plus the new unique_id
7279 tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7281 for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7282 rename_to.append(t_dsk.logical_id)
7283 t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7284 self.cfg.SetDiskID(t_dsk, pnode_name)
7285 result = self.rpc.call_blockdev_rename(pnode_name,
7286 zip(tmp_disks, rename_to))
7287 result.Raise("Failed to rename adoped LVs")
7289 feedback_fn("* creating instance disks...")
7291 _CreateDisks(self, iobj)
7292 except errors.OpExecError:
7293 self.LogWarning("Device creation failed, reverting...")
7295 _RemoveDisks(self, iobj)
7297 self.cfg.ReleaseDRBDMinors(instance)
7300 feedback_fn("adding instance %s to cluster config" % instance)
7302 self.cfg.AddInstance(iobj, self.proc.GetECId())
7304 # Declare that we don't want to remove the instance lock anymore, as we've
7305 # added the instance to the config
7306 del self.remove_locks[locking.LEVEL_INSTANCE]
7307 # Unlock all the nodes
7308 if self.op.mode == constants.INSTANCE_IMPORT:
7309 nodes_keep = [self.op.src_node]
7310 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7311 if node != self.op.src_node]
7312 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7313 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7315 self.context.glm.release(locking.LEVEL_NODE)
7316 del self.acquired_locks[locking.LEVEL_NODE]
7318 if self.op.wait_for_sync:
7319 disk_abort = not _WaitForSync(self, iobj)
7320 elif iobj.disk_template in constants.DTS_NET_MIRROR:
7321 # make sure the disks are not degraded (still sync-ing is ok)
7323 feedback_fn("* checking mirrors status")
7324 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7329 _RemoveDisks(self, iobj)
7330 self.cfg.RemoveInstance(iobj.name)
7331 # Make sure the instance lock gets removed
7332 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7333 raise errors.OpExecError("There are some degraded disks for"
7336 if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7337 if self.op.mode == constants.INSTANCE_CREATE:
7338 if not self.op.no_install:
7339 feedback_fn("* running the instance OS create scripts...")
7340 # FIXME: pass debug option from opcode to backend
7341 result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7342 self.op.debug_level)
7343 result.Raise("Could not add os for instance %s"
7344 " on node %s" % (instance, pnode_name))
7346 elif self.op.mode == constants.INSTANCE_IMPORT:
7347 feedback_fn("* running the instance OS import scripts...")
7351 for idx, image in enumerate(self.src_images):
7355 # FIXME: pass debug option from opcode to backend
7356 dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7357 constants.IEIO_FILE, (image, ),
7358 constants.IEIO_SCRIPT,
7359 (iobj.disks[idx], idx),
7361 transfers.append(dt)
7364 masterd.instance.TransferInstanceData(self, feedback_fn,
7365 self.op.src_node, pnode_name,
7366 self.pnode.secondary_ip,
7368 if not compat.all(import_result):
7369 self.LogWarning("Some disks for instance %s on node %s were not"
7370 " imported successfully" % (instance, pnode_name))
7372 elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7373 feedback_fn("* preparing remote import...")
7374 connect_timeout = constants.RIE_CONNECT_TIMEOUT
7375 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7377 disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
7378 self.source_x509_ca,
7379 self._cds, timeouts)
7380 if not compat.all(disk_results):
7381 # TODO: Should the instance still be started, even if some disks
7382 # failed to import (valid for local imports, too)?
7383 self.LogWarning("Some disks for instance %s on node %s were not"
7384 " imported successfully" % (instance, pnode_name))
7386 # Run rename script on newly imported instance
7387 assert iobj.name == instance
7388 feedback_fn("Running rename script for %s" % instance)
7389 result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7390 self.source_instance_name,
7391 self.op.debug_level)
7393 self.LogWarning("Failed to run rename script for %s on node"
7394 " %s: %s" % (instance, pnode_name, result.fail_msg))
7397 # also checked in the prereq part
7398 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7402 iobj.admin_up = True
7403 self.cfg.Update(iobj, feedback_fn)
7404 logging.info("Starting instance %s on node %s", instance, pnode_name)
7405 feedback_fn("* starting instance...")
7406 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7407 result.Raise("Could not start instance")
7409 return list(iobj.all_nodes)
7412 class LUConnectConsole(NoHooksLU):
7413 """Connect to an instance's console.
7415 This is somewhat special in that it returns the command line that
7416 you need to run on the master node in order to connect to the
7425 def ExpandNames(self):
7426 self._ExpandAndLockInstance()
7428 def CheckPrereq(self):
7429 """Check prerequisites.
7431 This checks that the instance is in the cluster.
7434 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7435 assert self.instance is not None, \
7436 "Cannot retrieve locked instance %s" % self.op.instance_name
7437 _CheckNodeOnline(self, self.instance.primary_node)
7439 def Exec(self, feedback_fn):
7440 """Connect to the console of an instance
7443 instance = self.instance
7444 node = instance.primary_node
7446 node_insts = self.rpc.call_instance_list([node],
7447 [instance.hypervisor])[node]
7448 node_insts.Raise("Can't get node information from %s" % node)
7450 if instance.name not in node_insts.payload:
7451 raise errors.OpExecError("Instance %s is not running." % instance.name)
7453 logging.debug("Connecting to console of %s on %s", instance.name, node)
7455 hyper = hypervisor.GetHypervisor(instance.hypervisor)
7456 cluster = self.cfg.GetClusterInfo()
7457 # beparams and hvparams are passed separately, to avoid editing the
7458 # instance and then saving the defaults in the instance itself.
7459 hvparams = cluster.FillHV(instance)
7460 beparams = cluster.FillBE(instance)
7461 console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
7464 return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
7467 class LUReplaceDisks(LogicalUnit):
7468 """Replace the disks of an instance.
7471 HPATH = "mirrors-replace"
7472 HTYPE = constants.HTYPE_INSTANCE
7475 ("mode", _NoDefault, _TElemOf(constants.REPLACE_MODES)),
7476 ("disks", _EmptyList, _TListOf(_TPositiveInt)),
7477 ("remote_node", None, _TMaybeString),
7478 ("iallocator", None, _TMaybeString),
7479 ("early_release", False, _TBool),
7483 def CheckArguments(self):
7484 TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7487 def ExpandNames(self):
7488 self._ExpandAndLockInstance()
7490 if self.op.iallocator is not None:
7491 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7493 elif self.op.remote_node is not None:
7494 remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7495 self.op.remote_node = remote_node
7497 # Warning: do not remove the locking of the new secondary here
7498 # unless DRBD8.AddChildren is changed to work in parallel;
7499 # currently it doesn't since parallel invocations of
7500 # FindUnusedMinor will conflict
7501 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7502 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7505 self.needed_locks[locking.LEVEL_NODE] = []
7506 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7508 self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7509 self.op.iallocator, self.op.remote_node,
7510 self.op.disks, False, self.op.early_release)
7512 self.tasklets = [self.replacer]
7514 def DeclareLocks(self, level):
7515 # If we're not already locking all nodes in the set we have to declare the
7516 # instance's primary/secondary nodes.
7517 if (level == locking.LEVEL_NODE and
7518 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7519 self._LockInstancesNodes()
7521 def BuildHooksEnv(self):
7524 This runs on the master, the primary and all the secondaries.
7527 instance = self.replacer.instance
7529 "MODE": self.op.mode,
7530 "NEW_SECONDARY": self.op.remote_node,
7531 "OLD_SECONDARY": instance.secondary_nodes[0],
7533 env.update(_BuildInstanceHookEnvByObject(self, instance))
7535 self.cfg.GetMasterNode(),
7536 instance.primary_node,
7538 if self.op.remote_node is not None:
7539 nl.append(self.op.remote_node)
7543 class TLReplaceDisks(Tasklet):
7544 """Replaces disks for an instance.
7546 Note: Locking is not within the scope of this class.
7549 def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7550 disks, delay_iallocator, early_release):
7551 """Initializes this class.
7554 Tasklet.__init__(self, lu)
7557 self.instance_name = instance_name
7559 self.iallocator_name = iallocator_name
7560 self.remote_node = remote_node
7562 self.delay_iallocator = delay_iallocator
7563 self.early_release = early_release
7566 self.instance = None
7567 self.new_node = None
7568 self.target_node = None
7569 self.other_node = None
7570 self.remote_node_info = None
7571 self.node_secondary_ip = None
7574 def CheckArguments(mode, remote_node, iallocator):
7575 """Helper function for users of this class.
7578 # check for valid parameter combination
7579 if mode == constants.REPLACE_DISK_CHG:
7580 if remote_node is None and iallocator is None:
7581 raise errors.OpPrereqError("When changing the secondary either an"
7582 " iallocator script must be used or the"
7583 " new node given", errors.ECODE_INVAL)
7585 if remote_node is not None and iallocator is not None:
7586 raise errors.OpPrereqError("Give either the iallocator or the new"
7587 " secondary, not both", errors.ECODE_INVAL)
7589 elif remote_node is not None or iallocator is not None:
7590 # Not replacing the secondary
7591 raise errors.OpPrereqError("The iallocator and new node options can"
7592 " only be used when changing the"
7593 " secondary node", errors.ECODE_INVAL)
7596 def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7597 """Compute a new secondary node using an IAllocator.
7600 ial = IAllocator(lu.cfg, lu.rpc,
7601 mode=constants.IALLOCATOR_MODE_RELOC,
7603 relocate_from=relocate_from)
7605 ial.Run(iallocator_name)
7608 raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7609 " %s" % (iallocator_name, ial.info),
7612 if len(ial.result) != ial.required_nodes:
7613 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7614 " of nodes (%s), required %s" %
7616 len(ial.result), ial.required_nodes),
7619 remote_node_name = ial.result[0]
7621 lu.LogInfo("Selected new secondary for instance '%s': %s",
7622 instance_name, remote_node_name)
7624 return remote_node_name
7626 def _FindFaultyDisks(self, node_name):
7627 return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7630 def CheckPrereq(self):
7631 """Check prerequisites.
7633 This checks that the instance is in the cluster.
7636 self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7637 assert instance is not None, \
7638 "Cannot retrieve locked instance %s" % self.instance_name
7640 if instance.disk_template != constants.DT_DRBD8:
7641 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7642 " instances", errors.ECODE_INVAL)
7644 if len(instance.secondary_nodes) != 1:
7645 raise errors.OpPrereqError("The instance has a strange layout,"
7646 " expected one secondary but found %d" %
7647 len(instance.secondary_nodes),
7650 if not self.delay_iallocator:
7651 self._CheckPrereq2()
7653 def _CheckPrereq2(self):
7654 """Check prerequisites, second part.
7656 This function should always be part of CheckPrereq. It was separated and is
7657 now called from Exec because during node evacuation iallocator was only
7658 called with an unmodified cluster model, not taking planned changes into
7662 instance = self.instance
7663 secondary_node = instance.secondary_nodes[0]
7665 if self.iallocator_name is None:
7666 remote_node = self.remote_node
7668 remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7669 instance.name, instance.secondary_nodes)
7671 if remote_node is not None:
7672 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7673 assert self.remote_node_info is not None, \
7674 "Cannot retrieve locked node %s" % remote_node
7676 self.remote_node_info = None
7678 if remote_node == self.instance.primary_node:
7679 raise errors.OpPrereqError("The specified node is the primary node of"
7680 " the instance.", errors.ECODE_INVAL)
7682 if remote_node == secondary_node:
7683 raise errors.OpPrereqError("The specified node is already the"
7684 " secondary node of the instance.",
7687 if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7688 constants.REPLACE_DISK_CHG):
7689 raise errors.OpPrereqError("Cannot specify disks to be replaced",
7692 if self.mode == constants.REPLACE_DISK_AUTO:
7693 faulty_primary = self._FindFaultyDisks(instance.primary_node)
7694 faulty_secondary = self._FindFaultyDisks(secondary_node)
7696 if faulty_primary and faulty_secondary:
7697 raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7698 " one node and can not be repaired"
7699 " automatically" % self.instance_name,
7703 self.disks = faulty_primary
7704 self.target_node = instance.primary_node
7705 self.other_node = secondary_node
7706 check_nodes = [self.target_node, self.other_node]
7707 elif faulty_secondary:
7708 self.disks = faulty_secondary
7709 self.target_node = secondary_node
7710 self.other_node = instance.primary_node
7711 check_nodes = [self.target_node, self.other_node]
7717 # Non-automatic modes
7718 if self.mode == constants.REPLACE_DISK_PRI:
7719 self.target_node = instance.primary_node
7720 self.other_node = secondary_node
7721 check_nodes = [self.target_node, self.other_node]
7723 elif self.mode == constants.REPLACE_DISK_SEC:
7724 self.target_node = secondary_node
7725 self.other_node = instance.primary_node
7726 check_nodes = [self.target_node, self.other_node]
7728 elif self.mode == constants.REPLACE_DISK_CHG:
7729 self.new_node = remote_node
7730 self.other_node = instance.primary_node
7731 self.target_node = secondary_node
7732 check_nodes = [self.new_node, self.other_node]
7734 _CheckNodeNotDrained(self.lu, remote_node)
7736 old_node_info = self.cfg.GetNodeInfo(secondary_node)
7737 assert old_node_info is not None
7738 if old_node_info.offline and not self.early_release:
7739 # doesn't make sense to delay the release
7740 self.early_release = True
7741 self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7742 " early-release mode", secondary_node)
7745 raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7748 # If not specified all disks should be replaced
7750 self.disks = range(len(self.instance.disks))
7752 for node in check_nodes:
7753 _CheckNodeOnline(self.lu, node)
7755 # Check whether disks are valid
7756 for disk_idx in self.disks:
7757 instance.FindDisk(disk_idx)
7759 # Get secondary node IP addresses
7762 for node_name in [self.target_node, self.other_node, self.new_node]:
7763 if node_name is not None:
7764 node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7766 self.node_secondary_ip = node_2nd_ip
7768 def Exec(self, feedback_fn):
7769 """Execute disk replacement.
7771 This dispatches the disk replacement to the appropriate handler.
7774 if self.delay_iallocator:
7775 self._CheckPrereq2()
7778 feedback_fn("No disks need replacement")
7781 feedback_fn("Replacing disk(s) %s for %s" %
7782 (utils.CommaJoin(self.disks), self.instance.name))
7784 activate_disks = (not self.instance.admin_up)
7786 # Activate the instance disks if we're replacing them on a down instance
7788 _StartInstanceDisks(self.lu, self.instance, True)
7791 # Should we replace the secondary node?
7792 if self.new_node is not None:
7793 fn = self._ExecDrbd8Secondary
7795 fn = self._ExecDrbd8DiskOnly
7797 return fn(feedback_fn)
7800 # Deactivate the instance disks if we're replacing them on a
7803 _SafeShutdownInstanceDisks(self.lu, self.instance)
7805 def _CheckVolumeGroup(self, nodes):
7806 self.lu.LogInfo("Checking volume groups")
7808 vgname = self.cfg.GetVGName()
7810 # Make sure volume group exists on all involved nodes
7811 results = self.rpc.call_vg_list(nodes)
7813 raise errors.OpExecError("Can't list volume groups on the nodes")
7817 res.Raise("Error checking node %s" % node)
7818 if vgname not in res.payload:
7819 raise errors.OpExecError("Volume group '%s' not found on node %s" %
7822 def _CheckDisksExistence(self, nodes):
7823 # Check disk existence
7824 for idx, dev in enumerate(self.instance.disks):
7825 if idx not in self.disks:
7829 self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7830 self.cfg.SetDiskID(dev, node)
7832 result = self.rpc.call_blockdev_find(node, dev)
7834 msg = result.fail_msg
7835 if msg or not result.payload:
7837 msg = "disk not found"
7838 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7841 def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7842 for idx, dev in enumerate(self.instance.disks):
7843 if idx not in self.disks:
7846 self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7849 if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7851 raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7852 " replace disks for instance %s" %
7853 (node_name, self.instance.name))
7855 def _CreateNewStorage(self, node_name):
7856 vgname = self.cfg.GetVGName()
7859 for idx, dev in enumerate(self.instance.disks):
7860 if idx not in self.disks:
7863 self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7865 self.cfg.SetDiskID(dev, node_name)
7867 lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7868 names = _GenerateUniqueNames(self.lu, lv_names)
7870 lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7871 logical_id=(vgname, names[0]))
7872 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7873 logical_id=(vgname, names[1]))
7875 new_lvs = [lv_data, lv_meta]
7876 old_lvs = dev.children
7877 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7879 # we pass force_create=True to force the LVM creation
7880 for new_lv in new_lvs:
7881 _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7882 _GetInstanceInfoText(self.instance), False)
7886 def _CheckDevices(self, node_name, iv_names):
7887 for name, (dev, _, _) in iv_names.iteritems():
7888 self.cfg.SetDiskID(dev, node_name)
7890 result = self.rpc.call_blockdev_find(node_name, dev)
7892 msg = result.fail_msg
7893 if msg or not result.payload:
7895 msg = "disk not found"
7896 raise errors.OpExecError("Can't find DRBD device %s: %s" %
7899 if result.payload.is_degraded:
7900 raise errors.OpExecError("DRBD device %s is degraded!" % name)
7902 def _RemoveOldStorage(self, node_name, iv_names):
7903 for name, (_, old_lvs, _) in iv_names.iteritems():
7904 self.lu.LogInfo("Remove logical volumes for %s" % name)
7907 self.cfg.SetDiskID(lv, node_name)
7909 msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7911 self.lu.LogWarning("Can't remove old LV: %s" % msg,
7912 hint="remove unused LVs manually")
7914 def _ReleaseNodeLock(self, node_name):
7915 """Releases the lock for a given node."""
7916 self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7918 def _ExecDrbd8DiskOnly(self, feedback_fn):
7919 """Replace a disk on the primary or secondary for DRBD 8.
7921 The algorithm for replace is quite complicated:
7923 1. for each disk to be replaced:
7925 1. create new LVs on the target node with unique names
7926 1. detach old LVs from the drbd device
7927 1. rename old LVs to name_replaced.<time_t>
7928 1. rename new LVs to old LVs
7929 1. attach the new LVs (with the old names now) to the drbd device
7931 1. wait for sync across all devices
7933 1. for each modified disk:
7935 1. remove old LVs (which have the name name_replaces.<time_t>)
7937 Failures are not very well handled.
7942 # Step: check device activation
7943 self.lu.LogStep(1, steps_total, "Check device existence")
7944 self._CheckDisksExistence([self.other_node, self.target_node])
7945 self._CheckVolumeGroup([self.target_node, self.other_node])
7947 # Step: check other node consistency
7948 self.lu.LogStep(2, steps_total, "Check peer consistency")
7949 self._CheckDisksConsistency(self.other_node,
7950 self.other_node == self.instance.primary_node,
7953 # Step: create new storage
7954 self.lu.LogStep(3, steps_total, "Allocate new storage")
7955 iv_names = self._CreateNewStorage(self.target_node)
7957 # Step: for each lv, detach+rename*2+attach
7958 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
7959 for dev, old_lvs, new_lvs in iv_names.itervalues():
7960 self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
7962 result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
7964 result.Raise("Can't detach drbd from local storage on node"
7965 " %s for device %s" % (self.target_node, dev.iv_name))
7967 #cfg.Update(instance)
7969 # ok, we created the new LVs, so now we know we have the needed
7970 # storage; as such, we proceed on the target node to rename
7971 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
7972 # using the assumption that logical_id == physical_id (which in
7973 # turn is the unique_id on that node)
7975 # FIXME(iustin): use a better name for the replaced LVs
7976 temp_suffix = int(time.time())
7977 ren_fn = lambda d, suff: (d.physical_id[0],
7978 d.physical_id[1] + "_replaced-%s" % suff)
7980 # Build the rename list based on what LVs exist on the node
7981 rename_old_to_new = []
7982 for to_ren in old_lvs:
7983 result = self.rpc.call_blockdev_find(self.target_node, to_ren)
7984 if not result.fail_msg and result.payload:
7986 rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
7988 self.lu.LogInfo("Renaming the old LVs on the target node")
7989 result = self.rpc.call_blockdev_rename(self.target_node,
7991 result.Raise("Can't rename old LVs on node %s" % self.target_node)
7993 # Now we rename the new LVs to the old LVs
7994 self.lu.LogInfo("Renaming the new LVs on the target node")
7995 rename_new_to_old = [(new, old.physical_id)
7996 for old, new in zip(old_lvs, new_lvs)]
7997 result = self.rpc.call_blockdev_rename(self.target_node,
7999 result.Raise("Can't rename new LVs on node %s" % self.target_node)
8001 for old, new in zip(old_lvs, new_lvs):
8002 new.logical_id = old.logical_id
8003 self.cfg.SetDiskID(new, self.target_node)
8005 for disk in old_lvs:
8006 disk.logical_id = ren_fn(disk, temp_suffix)
8007 self.cfg.SetDiskID(disk, self.target_node)
8009 # Now that the new lvs have the old name, we can add them to the device
8010 self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8011 result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8013 msg = result.fail_msg
8015 for new_lv in new_lvs:
8016 msg2 = self.rpc.call_blockdev_remove(self.target_node,
8019 self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8020 hint=("cleanup manually the unused logical"
8022 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8024 dev.children = new_lvs
8026 self.cfg.Update(self.instance, feedback_fn)
8029 if self.early_release:
8030 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8032 self._RemoveOldStorage(self.target_node, iv_names)
8033 # WARNING: we release both node locks here, do not do other RPCs
8034 # than WaitForSync to the primary node
8035 self._ReleaseNodeLock([self.target_node, self.other_node])
8038 # This can fail as the old devices are degraded and _WaitForSync
8039 # does a combined result over all disks, so we don't check its return value
8040 self.lu.LogStep(cstep, steps_total, "Sync devices")
8042 _WaitForSync(self.lu, self.instance)
8044 # Check all devices manually
8045 self._CheckDevices(self.instance.primary_node, iv_names)
8047 # Step: remove old storage
8048 if not self.early_release:
8049 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8051 self._RemoveOldStorage(self.target_node, iv_names)
8053 def _ExecDrbd8Secondary(self, feedback_fn):
8054 """Replace the secondary node for DRBD 8.
8056 The algorithm for replace is quite complicated:
8057 - for all disks of the instance:
8058 - create new LVs on the new node with same names
8059 - shutdown the drbd device on the old secondary
8060 - disconnect the drbd network on the primary
8061 - create the drbd device on the new secondary
8062 - network attach the drbd on the primary, using an artifice:
8063 the drbd code for Attach() will connect to the network if it
8064 finds a device which is connected to the good local disks but
8066 - wait for sync across all devices
8067 - remove all disks from the old secondary
8069 Failures are not very well handled.
8074 # Step: check device activation
8075 self.lu.LogStep(1, steps_total, "Check device existence")
8076 self._CheckDisksExistence([self.instance.primary_node])
8077 self._CheckVolumeGroup([self.instance.primary_node])
8079 # Step: check other node consistency
8080 self.lu.LogStep(2, steps_total, "Check peer consistency")
8081 self._CheckDisksConsistency(self.instance.primary_node, True, True)
8083 # Step: create new storage
8084 self.lu.LogStep(3, steps_total, "Allocate new storage")
8085 for idx, dev in enumerate(self.instance.disks):
8086 self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8087 (self.new_node, idx))
8088 # we pass force_create=True to force LVM creation
8089 for new_lv in dev.children:
8090 _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8091 _GetInstanceInfoText(self.instance), False)
8093 # Step 4: dbrd minors and drbd setups changes
8094 # after this, we must manually remove the drbd minors on both the
8095 # error and the success paths
8096 self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8097 minors = self.cfg.AllocateDRBDMinor([self.new_node
8098 for dev in self.instance.disks],
8100 logging.debug("Allocated minors %r", minors)
8103 for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8104 self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8105 (self.new_node, idx))
8106 # create new devices on new_node; note that we create two IDs:
8107 # one without port, so the drbd will be activated without
8108 # networking information on the new node at this stage, and one
8109 # with network, for the latter activation in step 4
8110 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8111 if self.instance.primary_node == o_node1:
8114 assert self.instance.primary_node == o_node2, "Three-node instance?"
8117 new_alone_id = (self.instance.primary_node, self.new_node, None,
8118 p_minor, new_minor, o_secret)
8119 new_net_id = (self.instance.primary_node, self.new_node, o_port,
8120 p_minor, new_minor, o_secret)
8122 iv_names[idx] = (dev, dev.children, new_net_id)
8123 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8125 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8126 logical_id=new_alone_id,
8127 children=dev.children,
8130 _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8131 _GetInstanceInfoText(self.instance), False)
8132 except errors.GenericError:
8133 self.cfg.ReleaseDRBDMinors(self.instance.name)
8136 # We have new devices, shutdown the drbd on the old secondary
8137 for idx, dev in enumerate(self.instance.disks):
8138 self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8139 self.cfg.SetDiskID(dev, self.target_node)
8140 msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8142 self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8143 "node: %s" % (idx, msg),
8144 hint=("Please cleanup this device manually as"
8145 " soon as possible"))
8147 self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8148 result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8149 self.node_secondary_ip,
8150 self.instance.disks)\
8151 [self.instance.primary_node]
8153 msg = result.fail_msg
8155 # detaches didn't succeed (unlikely)
8156 self.cfg.ReleaseDRBDMinors(self.instance.name)
8157 raise errors.OpExecError("Can't detach the disks from the network on"
8158 " old node: %s" % (msg,))
8160 # if we managed to detach at least one, we update all the disks of
8161 # the instance to point to the new secondary
8162 self.lu.LogInfo("Updating instance configuration")
8163 for dev, _, new_logical_id in iv_names.itervalues():
8164 dev.logical_id = new_logical_id
8165 self.cfg.SetDiskID(dev, self.instance.primary_node)
8167 self.cfg.Update(self.instance, feedback_fn)
8169 # and now perform the drbd attach
8170 self.lu.LogInfo("Attaching primary drbds to new secondary"
8171 " (standalone => connected)")
8172 result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8174 self.node_secondary_ip,
8175 self.instance.disks,
8178 for to_node, to_result in result.items():
8179 msg = to_result.fail_msg
8181 self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8183 hint=("please do a gnt-instance info to see the"
8184 " status of disks"))
8186 if self.early_release:
8187 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8189 self._RemoveOldStorage(self.target_node, iv_names)
8190 # WARNING: we release all node locks here, do not do other RPCs
8191 # than WaitForSync to the primary node
8192 self._ReleaseNodeLock([self.instance.primary_node,
8197 # This can fail as the old devices are degraded and _WaitForSync
8198 # does a combined result over all disks, so we don't check its return value
8199 self.lu.LogStep(cstep, steps_total, "Sync devices")
8201 _WaitForSync(self.lu, self.instance)
8203 # Check all devices manually
8204 self._CheckDevices(self.instance.primary_node, iv_names)
8206 # Step: remove old storage
8207 if not self.early_release:
8208 self.lu.LogStep(cstep, steps_total, "Removing old storage")
8209 self._RemoveOldStorage(self.target_node, iv_names)
8212 class LURepairNodeStorage(NoHooksLU):
8213 """Repairs the volume group on a node.
8218 ("storage_type", _NoDefault, _CheckStorageType),
8219 ("name", _NoDefault, _TNonEmptyString),
8220 ("ignore_consistency", False, _TBool),
8224 def CheckArguments(self):
8225 self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8227 storage_type = self.op.storage_type
8229 if (constants.SO_FIX_CONSISTENCY not in
8230 constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8231 raise errors.OpPrereqError("Storage units of type '%s' can not be"
8232 " repaired" % storage_type,
8235 def ExpandNames(self):
8236 self.needed_locks = {
8237 locking.LEVEL_NODE: [self.op.node_name],
8240 def _CheckFaultyDisks(self, instance, node_name):
8241 """Ensure faulty disks abort the opcode or at least warn."""
8243 if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8245 raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8246 " node '%s'" % (instance.name, node_name),
8248 except errors.OpPrereqError, err:
8249 if self.op.ignore_consistency:
8250 self.proc.LogWarning(str(err.args[0]))
8254 def CheckPrereq(self):
8255 """Check prerequisites.
8258 # Check whether any instance on this node has faulty disks
8259 for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8260 if not inst.admin_up:
8262 check_nodes = set(inst.all_nodes)
8263 check_nodes.discard(self.op.node_name)
8264 for inst_node_name in check_nodes:
8265 self._CheckFaultyDisks(inst, inst_node_name)
8267 def Exec(self, feedback_fn):
8268 feedback_fn("Repairing storage unit '%s' on %s ..." %
8269 (self.op.name, self.op.node_name))
8271 st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8272 result = self.rpc.call_storage_execute(self.op.node_name,
8273 self.op.storage_type, st_args,
8275 constants.SO_FIX_CONSISTENCY)
8276 result.Raise("Failed to repair storage unit '%s' on %s" %
8277 (self.op.name, self.op.node_name))
8280 class LUNodeEvacuationStrategy(NoHooksLU):
8281 """Computes the node evacuation strategy.
8285 ("nodes", _NoDefault, _TListOf(_TNonEmptyString)),
8286 ("remote_node", None, _TMaybeString),
8287 ("iallocator", None, _TMaybeString),
8291 def CheckArguments(self):
8292 _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8294 def ExpandNames(self):
8295 self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8296 self.needed_locks = locks = {}
8297 if self.op.remote_node is None:
8298 locks[locking.LEVEL_NODE] = locking.ALL_SET
8300 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8301 locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8303 def Exec(self, feedback_fn):
8304 if self.op.remote_node is not None:
8306 for node in self.op.nodes:
8307 instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8310 if i.primary_node == self.op.remote_node:
8311 raise errors.OpPrereqError("Node %s is the primary node of"
8312 " instance %s, cannot use it as"
8314 (self.op.remote_node, i.name),
8316 result.append([i.name, self.op.remote_node])
8318 ial = IAllocator(self.cfg, self.rpc,
8319 mode=constants.IALLOCATOR_MODE_MEVAC,
8320 evac_nodes=self.op.nodes)
8321 ial.Run(self.op.iallocator, validate=True)
8323 raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8329 class LUGrowDisk(LogicalUnit):
8330 """Grow a disk of an instance.
8334 HTYPE = constants.HTYPE_INSTANCE
8337 ("disk", _NoDefault, _TInt),
8338 ("amount", _NoDefault, _TInt),
8339 ("wait_for_sync", True, _TBool),
8343 def ExpandNames(self):
8344 self._ExpandAndLockInstance()
8345 self.needed_locks[locking.LEVEL_NODE] = []
8346 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8348 def DeclareLocks(self, level):
8349 if level == locking.LEVEL_NODE:
8350 self._LockInstancesNodes()
8352 def BuildHooksEnv(self):
8355 This runs on the master, the primary and all the secondaries.
8359 "DISK": self.op.disk,
8360 "AMOUNT": self.op.amount,
8362 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8363 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8366 def CheckPrereq(self):
8367 """Check prerequisites.
8369 This checks that the instance is in the cluster.
8372 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8373 assert instance is not None, \
8374 "Cannot retrieve locked instance %s" % self.op.instance_name
8375 nodenames = list(instance.all_nodes)
8376 for node in nodenames:
8377 _CheckNodeOnline(self, node)
8379 self.instance = instance
8381 if instance.disk_template not in constants.DTS_GROWABLE:
8382 raise errors.OpPrereqError("Instance's disk layout does not support"
8383 " growing.", errors.ECODE_INVAL)
8385 self.disk = instance.FindDisk(self.op.disk)
8387 if instance.disk_template != constants.DT_FILE:
8388 # TODO: check the free disk space for file, when that feature will be
8390 _CheckNodesFreeDisk(self, nodenames, self.op.amount)
8392 def Exec(self, feedback_fn):
8393 """Execute disk grow.
8396 instance = self.instance
8399 disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8401 raise errors.OpExecError("Cannot activate block device to grow")
8403 for node in instance.all_nodes:
8404 self.cfg.SetDiskID(disk, node)
8405 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8406 result.Raise("Grow request failed to node %s" % node)
8408 # TODO: Rewrite code to work properly
8409 # DRBD goes into sync mode for a short amount of time after executing the
8410 # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8411 # calling "resize" in sync mode fails. Sleeping for a short amount of
8412 # time is a work-around.
8415 disk.RecordGrow(self.op.amount)
8416 self.cfg.Update(instance, feedback_fn)
8417 if self.op.wait_for_sync:
8418 disk_abort = not _WaitForSync(self, instance, disks=[disk])
8420 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8421 " status.\nPlease check the instance.")
8422 if not instance.admin_up:
8423 _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8424 elif not instance.admin_up:
8425 self.proc.LogWarning("Not shutting down the disk even if the instance is"
8426 " not supposed to be running because no wait for"
8427 " sync mode was requested.")
8430 class LUQueryInstanceData(NoHooksLU):
8431 """Query runtime instance data.
8435 ("instances", _EmptyList, _TListOf(_TNonEmptyString)),
8436 ("static", False, _TBool),
8440 def ExpandNames(self):
8441 self.needed_locks = {}
8442 self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8444 if self.op.instances:
8445 self.wanted_names = []
8446 for name in self.op.instances:
8447 full_name = _ExpandInstanceName(self.cfg, name)
8448 self.wanted_names.append(full_name)
8449 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8451 self.wanted_names = None
8452 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8454 self.needed_locks[locking.LEVEL_NODE] = []
8455 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8457 def DeclareLocks(self, level):
8458 if level == locking.LEVEL_NODE:
8459 self._LockInstancesNodes()
8461 def CheckPrereq(self):
8462 """Check prerequisites.
8464 This only checks the optional instance list against the existing names.
8467 if self.wanted_names is None:
8468 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8470 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8471 in self.wanted_names]
8473 def _ComputeBlockdevStatus(self, node, instance_name, dev):
8474 """Returns the status of a block device
8477 if self.op.static or not node:
8480 self.cfg.SetDiskID(dev, node)
8482 result = self.rpc.call_blockdev_find(node, dev)
8486 result.Raise("Can't compute disk status for %s" % instance_name)
8488 status = result.payload
8492 return (status.dev_path, status.major, status.minor,
8493 status.sync_percent, status.estimated_time,
8494 status.is_degraded, status.ldisk_status)
8496 def _ComputeDiskStatus(self, instance, snode, dev):
8497 """Compute block device status.
8500 if dev.dev_type in constants.LDS_DRBD:
8501 # we change the snode then (otherwise we use the one passed in)
8502 if dev.logical_id[0] == instance.primary_node:
8503 snode = dev.logical_id[1]
8505 snode = dev.logical_id[0]
8507 dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8509 dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8512 dev_children = [self._ComputeDiskStatus(instance, snode, child)
8513 for child in dev.children]
8518 "iv_name": dev.iv_name,
8519 "dev_type": dev.dev_type,
8520 "logical_id": dev.logical_id,
8521 "physical_id": dev.physical_id,
8522 "pstatus": dev_pstatus,
8523 "sstatus": dev_sstatus,
8524 "children": dev_children,
8531 def Exec(self, feedback_fn):
8532 """Gather and return data"""
8535 cluster = self.cfg.GetClusterInfo()
8537 for instance in self.wanted_instances:
8538 if not self.op.static:
8539 remote_info = self.rpc.call_instance_info(instance.primary_node,
8541 instance.hypervisor)
8542 remote_info.Raise("Error checking node %s" % instance.primary_node)
8543 remote_info = remote_info.payload
8544 if remote_info and "state" in remote_info:
8547 remote_state = "down"
8550 if instance.admin_up:
8553 config_state = "down"
8555 disks = [self._ComputeDiskStatus(instance, None, device)
8556 for device in instance.disks]
8559 "name": instance.name,
8560 "config_state": config_state,
8561 "run_state": remote_state,
8562 "pnode": instance.primary_node,
8563 "snodes": instance.secondary_nodes,
8565 # this happens to be the same format used for hooks
8566 "nics": _NICListToTuple(self, instance.nics),
8567 "disk_template": instance.disk_template,
8569 "hypervisor": instance.hypervisor,
8570 "network_port": instance.network_port,
8571 "hv_instance": instance.hvparams,
8572 "hv_actual": cluster.FillHV(instance, skip_globals=True),
8573 "be_instance": instance.beparams,
8574 "be_actual": cluster.FillBE(instance),
8575 "os_instance": instance.osparams,
8576 "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8577 "serial_no": instance.serial_no,
8578 "mtime": instance.mtime,
8579 "ctime": instance.ctime,
8580 "uuid": instance.uuid,
8583 result[instance.name] = idict
8588 class LUSetInstanceParams(LogicalUnit):
8589 """Modifies an instances's parameters.
8592 HPATH = "instance-modify"
8593 HTYPE = constants.HTYPE_INSTANCE
8596 ("nics", _EmptyList, _TList),
8597 ("disks", _EmptyList, _TList),
8598 ("beparams", _EmptyDict, _TDict),
8599 ("hvparams", _EmptyDict, _TDict),
8600 ("disk_template", None, _TMaybeString),
8601 ("remote_node", None, _TMaybeString),
8602 ("os_name", None, _TMaybeString),
8603 ("force_variant", False, _TBool),
8604 ("osparams", None, _TOr(_TDict, _TNone)),
8609 def CheckArguments(self):
8610 if not (self.op.nics or self.op.disks or self.op.disk_template or
8611 self.op.hvparams or self.op.beparams or self.op.os_name):
8612 raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8614 if self.op.hvparams:
8615 _CheckGlobalHvParams(self.op.hvparams)
8619 for disk_op, disk_dict in self.op.disks:
8620 utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8621 if disk_op == constants.DDM_REMOVE:
8624 elif disk_op == constants.DDM_ADD:
8627 if not isinstance(disk_op, int):
8628 raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8629 if not isinstance(disk_dict, dict):
8630 msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8631 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8633 if disk_op == constants.DDM_ADD:
8634 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8635 if mode not in constants.DISK_ACCESS_SET:
8636 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8638 size = disk_dict.get('size', None)
8640 raise errors.OpPrereqError("Required disk parameter size missing",
8644 except (TypeError, ValueError), err:
8645 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8646 str(err), errors.ECODE_INVAL)
8647 disk_dict['size'] = size
8649 # modification of disk
8650 if 'size' in disk_dict:
8651 raise errors.OpPrereqError("Disk size change not possible, use"
8652 " grow-disk", errors.ECODE_INVAL)
8654 if disk_addremove > 1:
8655 raise errors.OpPrereqError("Only one disk add or remove operation"
8656 " supported at a time", errors.ECODE_INVAL)
8658 if self.op.disks and self.op.disk_template is not None:
8659 raise errors.OpPrereqError("Disk template conversion and other disk"
8660 " changes not supported at the same time",
8663 if self.op.disk_template:
8664 _CheckDiskTemplate(self.op.disk_template)
8665 if (self.op.disk_template in constants.DTS_NET_MIRROR and
8666 self.op.remote_node is None):
8667 raise errors.OpPrereqError("Changing the disk template to a mirrored"
8668 " one requires specifying a secondary node",
8673 for nic_op, nic_dict in self.op.nics:
8674 utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8675 if nic_op == constants.DDM_REMOVE:
8678 elif nic_op == constants.DDM_ADD:
8681 if not isinstance(nic_op, int):
8682 raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8683 if not isinstance(nic_dict, dict):
8684 msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8685 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8687 # nic_dict should be a dict
8688 nic_ip = nic_dict.get('ip', None)
8689 if nic_ip is not None:
8690 if nic_ip.lower() == constants.VALUE_NONE:
8691 nic_dict['ip'] = None
8693 if not netutils.IP4Address.IsValid(nic_ip):
8694 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8697 nic_bridge = nic_dict.get('bridge', None)
8698 nic_link = nic_dict.get('link', None)
8699 if nic_bridge and nic_link:
8700 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8701 " at the same time", errors.ECODE_INVAL)
8702 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8703 nic_dict['bridge'] = None
8704 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8705 nic_dict['link'] = None
8707 if nic_op == constants.DDM_ADD:
8708 nic_mac = nic_dict.get('mac', None)
8710 nic_dict['mac'] = constants.VALUE_AUTO
8712 if 'mac' in nic_dict:
8713 nic_mac = nic_dict['mac']
8714 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8715 nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8717 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8718 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8719 " modifying an existing nic",
8722 if nic_addremove > 1:
8723 raise errors.OpPrereqError("Only one NIC add or remove operation"
8724 " supported at a time", errors.ECODE_INVAL)
8726 def ExpandNames(self):
8727 self._ExpandAndLockInstance()
8728 self.needed_locks[locking.LEVEL_NODE] = []
8729 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8731 def DeclareLocks(self, level):
8732 if level == locking.LEVEL_NODE:
8733 self._LockInstancesNodes()
8734 if self.op.disk_template and self.op.remote_node:
8735 self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8736 self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8738 def BuildHooksEnv(self):
8741 This runs on the master, primary and secondaries.
8745 if constants.BE_MEMORY in self.be_new:
8746 args['memory'] = self.be_new[constants.BE_MEMORY]
8747 if constants.BE_VCPUS in self.be_new:
8748 args['vcpus'] = self.be_new[constants.BE_VCPUS]
8749 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8750 # information at all.
8753 nic_override = dict(self.op.nics)
8754 for idx, nic in enumerate(self.instance.nics):
8755 if idx in nic_override:
8756 this_nic_override = nic_override[idx]
8758 this_nic_override = {}
8759 if 'ip' in this_nic_override:
8760 ip = this_nic_override['ip']
8763 if 'mac' in this_nic_override:
8764 mac = this_nic_override['mac']
8767 if idx in self.nic_pnew:
8768 nicparams = self.nic_pnew[idx]
8770 nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
8771 mode = nicparams[constants.NIC_MODE]
8772 link = nicparams[constants.NIC_LINK]
8773 args['nics'].append((ip, mac, mode, link))
8774 if constants.DDM_ADD in nic_override:
8775 ip = nic_override[constants.DDM_ADD].get('ip', None)
8776 mac = nic_override[constants.DDM_ADD]['mac']
8777 nicparams = self.nic_pnew[constants.DDM_ADD]
8778 mode = nicparams[constants.NIC_MODE]
8779 link = nicparams[constants.NIC_LINK]
8780 args['nics'].append((ip, mac, mode, link))
8781 elif constants.DDM_REMOVE in nic_override:
8782 del args['nics'][-1]
8784 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8785 if self.op.disk_template:
8786 env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8787 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8790 def CheckPrereq(self):
8791 """Check prerequisites.
8793 This only checks the instance list against the existing names.
8796 # checking the new params on the primary/secondary nodes
8798 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8799 cluster = self.cluster = self.cfg.GetClusterInfo()
8800 assert self.instance is not None, \
8801 "Cannot retrieve locked instance %s" % self.op.instance_name
8802 pnode = instance.primary_node
8803 nodelist = list(instance.all_nodes)
8806 if self.op.os_name and not self.op.force:
8807 _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8808 self.op.force_variant)
8809 instance_os = self.op.os_name
8811 instance_os = instance.os
8813 if self.op.disk_template:
8814 if instance.disk_template == self.op.disk_template:
8815 raise errors.OpPrereqError("Instance already has disk template %s" %
8816 instance.disk_template, errors.ECODE_INVAL)
8818 if (instance.disk_template,
8819 self.op.disk_template) not in self._DISK_CONVERSIONS:
8820 raise errors.OpPrereqError("Unsupported disk template conversion from"
8821 " %s to %s" % (instance.disk_template,
8822 self.op.disk_template),
8824 _CheckInstanceDown(self, instance, "cannot change disk template")
8825 if self.op.disk_template in constants.DTS_NET_MIRROR:
8826 if self.op.remote_node == pnode:
8827 raise errors.OpPrereqError("Given new secondary node %s is the same"
8828 " as the primary node of the instance" %
8829 self.op.remote_node, errors.ECODE_STATE)
8830 _CheckNodeOnline(self, self.op.remote_node)
8831 _CheckNodeNotDrained(self, self.op.remote_node)
8832 disks = [{"size": d.size} for d in instance.disks]
8833 required = _ComputeDiskSize(self.op.disk_template, disks)
8834 _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8836 # hvparams processing
8837 if self.op.hvparams:
8838 hv_type = instance.hypervisor
8839 i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
8840 utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
8841 hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
8844 hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
8845 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8846 self.hv_new = hv_new # the new actual values
8847 self.hv_inst = i_hvdict # the new dict (without defaults)
8849 self.hv_new = self.hv_inst = {}
8851 # beparams processing
8852 if self.op.beparams:
8853 i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
8855 utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
8856 be_new = cluster.SimpleFillBE(i_bedict)
8857 self.be_new = be_new # the new actual values
8858 self.be_inst = i_bedict # the new dict (without defaults)
8860 self.be_new = self.be_inst = {}
8862 # osparams processing
8863 if self.op.osparams:
8864 i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
8865 _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
8866 self.os_new = cluster.SimpleFillOS(instance_os, i_osdict)
8867 self.os_inst = i_osdict # the new dict (without defaults)
8869 self.os_new = self.os_inst = {}
8873 if constants.BE_MEMORY in self.op.beparams and not self.op.force:
8874 mem_check_list = [pnode]
8875 if be_new[constants.BE_AUTO_BALANCE]:
8876 # either we changed auto_balance to yes or it was from before
8877 mem_check_list.extend(instance.secondary_nodes)
8878 instance_info = self.rpc.call_instance_info(pnode, instance.name,
8879 instance.hypervisor)
8880 nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8881 instance.hypervisor)
8882 pninfo = nodeinfo[pnode]
8883 msg = pninfo.fail_msg
8885 # Assume the primary node is unreachable and go ahead
8886 self.warn.append("Can't get info from primary node %s: %s" %
8888 elif not isinstance(pninfo.payload.get('memory_free', None), int):
8889 self.warn.append("Node data from primary node %s doesn't contain"
8890 " free memory information" % pnode)
8891 elif instance_info.fail_msg:
8892 self.warn.append("Can't get instance runtime information: %s" %
8893 instance_info.fail_msg)
8895 if instance_info.payload:
8896 current_mem = int(instance_info.payload['memory'])
8898 # Assume instance not running
8899 # (there is a slight race condition here, but it's not very probable,
8900 # and we have no other way to check)
8902 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8903 pninfo.payload['memory_free'])
8905 raise errors.OpPrereqError("This change will prevent the instance"
8906 " from starting, due to %d MB of memory"
8907 " missing on its primary node" % miss_mem,
8910 if be_new[constants.BE_AUTO_BALANCE]:
8911 for node, nres in nodeinfo.items():
8912 if node not in instance.secondary_nodes:
8916 self.warn.append("Can't get info from secondary node %s: %s" %
8918 elif not isinstance(nres.payload.get('memory_free', None), int):
8919 self.warn.append("Secondary node %s didn't return free"
8920 " memory information" % node)
8921 elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8922 self.warn.append("Not enough memory to failover instance to"
8923 " secondary node %s" % node)
8928 for nic_op, nic_dict in self.op.nics:
8929 if nic_op == constants.DDM_REMOVE:
8930 if not instance.nics:
8931 raise errors.OpPrereqError("Instance has no NICs, cannot remove",
8934 if nic_op != constants.DDM_ADD:
8936 if not instance.nics:
8937 raise errors.OpPrereqError("Invalid NIC index %s, instance has"
8938 " no NICs" % nic_op,
8940 if nic_op < 0 or nic_op >= len(instance.nics):
8941 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
8943 (nic_op, len(instance.nics) - 1),
8945 old_nic_params = instance.nics[nic_op].nicparams
8946 old_nic_ip = instance.nics[nic_op].ip
8951 update_params_dict = dict([(key, nic_dict[key])
8952 for key in constants.NICS_PARAMETERS
8953 if key in nic_dict])
8955 if 'bridge' in nic_dict:
8956 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
8958 new_nic_params = _GetUpdatedParams(old_nic_params,
8960 utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
8961 new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
8962 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
8963 self.nic_pinst[nic_op] = new_nic_params
8964 self.nic_pnew[nic_op] = new_filled_nic_params
8965 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
8967 if new_nic_mode == constants.NIC_MODE_BRIDGED:
8968 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
8969 msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
8971 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
8973 self.warn.append(msg)
8975 raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
8976 if new_nic_mode == constants.NIC_MODE_ROUTED:
8977 if 'ip' in nic_dict:
8978 nic_ip = nic_dict['ip']
8982 raise errors.OpPrereqError('Cannot set the nic ip to None'
8983 ' on a routed nic', errors.ECODE_INVAL)
8984 if 'mac' in nic_dict:
8985 nic_mac = nic_dict['mac']
8987 raise errors.OpPrereqError('Cannot set the nic mac to None',
8989 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8990 # otherwise generate the mac
8991 nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
8993 # or validate/reserve the current one
8995 self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
8996 except errors.ReservationError:
8997 raise errors.OpPrereqError("MAC address %s already in use"
8998 " in cluster" % nic_mac,
8999 errors.ECODE_NOTUNIQUE)
9002 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9003 raise errors.OpPrereqError("Disk operations not supported for"
9004 " diskless instances",
9006 for disk_op, _ in self.op.disks:
9007 if disk_op == constants.DDM_REMOVE:
9008 if len(instance.disks) == 1:
9009 raise errors.OpPrereqError("Cannot remove the last disk of"
9010 " an instance", errors.ECODE_INVAL)
9011 _CheckInstanceDown(self, instance, "cannot remove disks")
9013 if (disk_op == constants.DDM_ADD and
9014 len(instance.nics) >= constants.MAX_DISKS):
9015 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9016 " add more" % constants.MAX_DISKS,
9018 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9020 if disk_op < 0 or disk_op >= len(instance.disks):
9021 raise errors.OpPrereqError("Invalid disk index %s, valid values"
9023 (disk_op, len(instance.disks)),
9028 def _ConvertPlainToDrbd(self, feedback_fn):
9029 """Converts an instance from plain to drbd.
9032 feedback_fn("Converting template to drbd")
9033 instance = self.instance
9034 pnode = instance.primary_node
9035 snode = self.op.remote_node
9037 # create a fake disk info for _GenerateDiskTemplate
9038 disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9039 new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9040 instance.name, pnode, [snode],
9041 disk_info, None, None, 0)
9042 info = _GetInstanceInfoText(instance)
9043 feedback_fn("Creating aditional volumes...")
9044 # first, create the missing data and meta devices
9045 for disk in new_disks:
9046 # unfortunately this is... not too nice
9047 _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9049 for child in disk.children:
9050 _CreateSingleBlockDev(self, snode, instance, child, info, True)
9051 # at this stage, all new LVs have been created, we can rename the
9053 feedback_fn("Renaming original volumes...")
9054 rename_list = [(o, n.children[0].logical_id)
9055 for (o, n) in zip(instance.disks, new_disks)]
9056 result = self.rpc.call_blockdev_rename(pnode, rename_list)
9057 result.Raise("Failed to rename original LVs")
9059 feedback_fn("Initializing DRBD devices...")
9060 # all child devices are in place, we can now create the DRBD devices
9061 for disk in new_disks:
9062 for node in [pnode, snode]:
9063 f_create = node == pnode
9064 _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9066 # at this point, the instance has been modified
9067 instance.disk_template = constants.DT_DRBD8
9068 instance.disks = new_disks
9069 self.cfg.Update(instance, feedback_fn)
9071 # disks are created, waiting for sync
9072 disk_abort = not _WaitForSync(self, instance)
9074 raise errors.OpExecError("There are some degraded disks for"
9075 " this instance, please cleanup manually")
9077 def _ConvertDrbdToPlain(self, feedback_fn):
9078 """Converts an instance from drbd to plain.
9081 instance = self.instance
9082 assert len(instance.secondary_nodes) == 1
9083 pnode = instance.primary_node
9084 snode = instance.secondary_nodes[0]
9085 feedback_fn("Converting template to plain")
9087 old_disks = instance.disks
9088 new_disks = [d.children[0] for d in old_disks]
9090 # copy over size and mode
9091 for parent, child in zip(old_disks, new_disks):
9092 child.size = parent.size
9093 child.mode = parent.mode
9095 # update instance structure
9096 instance.disks = new_disks
9097 instance.disk_template = constants.DT_PLAIN
9098 self.cfg.Update(instance, feedback_fn)
9100 feedback_fn("Removing volumes on the secondary node...")
9101 for disk in old_disks:
9102 self.cfg.SetDiskID(disk, snode)
9103 msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9105 self.LogWarning("Could not remove block device %s on node %s,"
9106 " continuing anyway: %s", disk.iv_name, snode, msg)
9108 feedback_fn("Removing unneeded volumes on the primary node...")
9109 for idx, disk in enumerate(old_disks):
9110 meta = disk.children[1]
9111 self.cfg.SetDiskID(meta, pnode)
9112 msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9114 self.LogWarning("Could not remove metadata for disk %d on node %s,"
9115 " continuing anyway: %s", idx, pnode, msg)
9118 def Exec(self, feedback_fn):
9119 """Modifies an instance.
9121 All parameters take effect only at the next restart of the instance.
9124 # Process here the warnings from CheckPrereq, as we don't have a
9125 # feedback_fn there.
9126 for warn in self.warn:
9127 feedback_fn("WARNING: %s" % warn)
9130 instance = self.instance
9132 for disk_op, disk_dict in self.op.disks:
9133 if disk_op == constants.DDM_REMOVE:
9134 # remove the last disk
9135 device = instance.disks.pop()
9136 device_idx = len(instance.disks)
9137 for node, disk in device.ComputeNodeTree(instance.primary_node):
9138 self.cfg.SetDiskID(disk, node)
9139 msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9141 self.LogWarning("Could not remove disk/%d on node %s: %s,"
9142 " continuing anyway", device_idx, node, msg)
9143 result.append(("disk/%d" % device_idx, "remove"))
9144 elif disk_op == constants.DDM_ADD:
9146 if instance.disk_template == constants.DT_FILE:
9147 file_driver, file_path = instance.disks[0].logical_id
9148 file_path = os.path.dirname(file_path)
9150 file_driver = file_path = None
9151 disk_idx_base = len(instance.disks)
9152 new_disk = _GenerateDiskTemplate(self,
9153 instance.disk_template,
9154 instance.name, instance.primary_node,
9155 instance.secondary_nodes,
9160 instance.disks.append(new_disk)
9161 info = _GetInstanceInfoText(instance)
9163 logging.info("Creating volume %s for instance %s",
9164 new_disk.iv_name, instance.name)
9165 # Note: this needs to be kept in sync with _CreateDisks
9167 for node in instance.all_nodes:
9168 f_create = node == instance.primary_node
9170 _CreateBlockDev(self, node, instance, new_disk,
9171 f_create, info, f_create)
9172 except errors.OpExecError, err:
9173 self.LogWarning("Failed to create volume %s (%s) on"
9175 new_disk.iv_name, new_disk, node, err)
9176 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9177 (new_disk.size, new_disk.mode)))
9179 # change a given disk
9180 instance.disks[disk_op].mode = disk_dict['mode']
9181 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9183 if self.op.disk_template:
9184 r_shut = _ShutdownInstanceDisks(self, instance)
9186 raise errors.OpExecError("Cannot shutdow instance disks, unable to"
9187 " proceed with disk template conversion")
9188 mode = (instance.disk_template, self.op.disk_template)
9190 self._DISK_CONVERSIONS[mode](self, feedback_fn)
9192 self.cfg.ReleaseDRBDMinors(instance.name)
9194 result.append(("disk_template", self.op.disk_template))
9197 for nic_op, nic_dict in self.op.nics:
9198 if nic_op == constants.DDM_REMOVE:
9199 # remove the last nic
9200 del instance.nics[-1]
9201 result.append(("nic.%d" % len(instance.nics), "remove"))
9202 elif nic_op == constants.DDM_ADD:
9203 # mac and bridge should be set, by now
9204 mac = nic_dict['mac']
9205 ip = nic_dict.get('ip', None)
9206 nicparams = self.nic_pinst[constants.DDM_ADD]
9207 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9208 instance.nics.append(new_nic)
9209 result.append(("nic.%d" % (len(instance.nics) - 1),
9210 "add:mac=%s,ip=%s,mode=%s,link=%s" %
9211 (new_nic.mac, new_nic.ip,
9212 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9213 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9216 for key in 'mac', 'ip':
9218 setattr(instance.nics[nic_op], key, nic_dict[key])
9219 if nic_op in self.nic_pinst:
9220 instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9221 for key, val in nic_dict.iteritems():
9222 result.append(("nic.%s/%d" % (key, nic_op), val))
9225 if self.op.hvparams:
9226 instance.hvparams = self.hv_inst
9227 for key, val in self.op.hvparams.iteritems():
9228 result.append(("hv/%s" % key, val))
9231 if self.op.beparams:
9232 instance.beparams = self.be_inst
9233 for key, val in self.op.beparams.iteritems():
9234 result.append(("be/%s" % key, val))
9238 instance.os = self.op.os_name
9241 if self.op.osparams:
9242 instance.osparams = self.os_inst
9243 for key, val in self.op.osparams.iteritems():
9244 result.append(("os/%s" % key, val))
9246 self.cfg.Update(instance, feedback_fn)
9250 _DISK_CONVERSIONS = {
9251 (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9252 (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9256 class LUQueryExports(NoHooksLU):
9257 """Query the exports list
9261 ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9262 ("use_locking", False, _TBool),
9266 def ExpandNames(self):
9267 self.needed_locks = {}
9268 self.share_locks[locking.LEVEL_NODE] = 1
9269 if not self.op.nodes:
9270 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9272 self.needed_locks[locking.LEVEL_NODE] = \
9273 _GetWantedNodes(self, self.op.nodes)
9275 def Exec(self, feedback_fn):
9276 """Compute the list of all the exported system images.
9279 @return: a dictionary with the structure node->(export-list)
9280 where export-list is a list of the instances exported on
9284 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9285 rpcresult = self.rpc.call_export_list(self.nodes)
9287 for node in rpcresult:
9288 if rpcresult[node].fail_msg:
9289 result[node] = False
9291 result[node] = rpcresult[node].payload
9296 class LUPrepareExport(NoHooksLU):
9297 """Prepares an instance for an export and returns useful information.
9302 ("mode", _NoDefault, _TElemOf(constants.EXPORT_MODES)),
9306 def ExpandNames(self):
9307 self._ExpandAndLockInstance()
9309 def CheckPrereq(self):
9310 """Check prerequisites.
9313 instance_name = self.op.instance_name
9315 self.instance = self.cfg.GetInstanceInfo(instance_name)
9316 assert self.instance is not None, \
9317 "Cannot retrieve locked instance %s" % self.op.instance_name
9318 _CheckNodeOnline(self, self.instance.primary_node)
9320 self._cds = _GetClusterDomainSecret()
9322 def Exec(self, feedback_fn):
9323 """Prepares an instance for an export.
9326 instance = self.instance
9328 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9329 salt = utils.GenerateSecret(8)
9331 feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9332 result = self.rpc.call_x509_cert_create(instance.primary_node,
9333 constants.RIE_CERT_VALIDITY)
9334 result.Raise("Can't create X509 key and certificate on %s" % result.node)
9336 (name, cert_pem) = result.payload
9338 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9342 "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9343 "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9345 "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9351 class LUExportInstance(LogicalUnit):
9352 """Export an instance to an image in the cluster.
9355 HPATH = "instance-export"
9356 HTYPE = constants.HTYPE_INSTANCE
9359 ("target_node", _NoDefault, _TOr(_TNonEmptyString, _TList)),
9360 ("shutdown", True, _TBool),
9362 ("remove_instance", False, _TBool),
9363 ("ignore_remove_failures", False, _TBool),
9364 ("mode", constants.EXPORT_MODE_LOCAL, _TElemOf(constants.EXPORT_MODES)),
9365 ("x509_key_name", None, _TOr(_TList, _TNone)),
9366 ("destination_x509_ca", None, _TMaybeString),
9370 def CheckArguments(self):
9371 """Check the arguments.
9374 self.x509_key_name = self.op.x509_key_name
9375 self.dest_x509_ca_pem = self.op.destination_x509_ca
9377 if self.op.remove_instance and not self.op.shutdown:
9378 raise errors.OpPrereqError("Can not remove instance without shutting it"
9381 if self.op.mode == constants.EXPORT_MODE_REMOTE:
9382 if not self.x509_key_name:
9383 raise errors.OpPrereqError("Missing X509 key name for encryption",
9386 if not self.dest_x509_ca_pem:
9387 raise errors.OpPrereqError("Missing destination X509 CA",
9390 def ExpandNames(self):
9391 self._ExpandAndLockInstance()
9393 # Lock all nodes for local exports
9394 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9395 # FIXME: lock only instance primary and destination node
9397 # Sad but true, for now we have do lock all nodes, as we don't know where
9398 # the previous export might be, and in this LU we search for it and
9399 # remove it from its current node. In the future we could fix this by:
9400 # - making a tasklet to search (share-lock all), then create the
9401 # new one, then one to remove, after
9402 # - removing the removal operation altogether
9403 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9405 def DeclareLocks(self, level):
9406 """Last minute lock declaration."""
9407 # All nodes are locked anyway, so nothing to do here.
9409 def BuildHooksEnv(self):
9412 This will run on the master, primary node and target node.
9416 "EXPORT_MODE": self.op.mode,
9417 "EXPORT_NODE": self.op.target_node,
9418 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9419 "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9420 # TODO: Generic function for boolean env variables
9421 "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9424 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9426 nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9428 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9429 nl.append(self.op.target_node)
9433 def CheckPrereq(self):
9434 """Check prerequisites.
9436 This checks that the instance and node names are valid.
9439 instance_name = self.op.instance_name
9441 self.instance = self.cfg.GetInstanceInfo(instance_name)
9442 assert self.instance is not None, \
9443 "Cannot retrieve locked instance %s" % self.op.instance_name
9444 _CheckNodeOnline(self, self.instance.primary_node)
9446 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9447 self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9448 self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9449 assert self.dst_node is not None
9451 _CheckNodeOnline(self, self.dst_node.name)
9452 _CheckNodeNotDrained(self, self.dst_node.name)
9455 self.dest_disk_info = None
9456 self.dest_x509_ca = None
9458 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9459 self.dst_node = None
9461 if len(self.op.target_node) != len(self.instance.disks):
9462 raise errors.OpPrereqError(("Received destination information for %s"
9463 " disks, but instance %s has %s disks") %
9464 (len(self.op.target_node), instance_name,
9465 len(self.instance.disks)),
9468 cds = _GetClusterDomainSecret()
9470 # Check X509 key name
9472 (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9473 except (TypeError, ValueError), err:
9474 raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9476 if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9477 raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9480 # Load and verify CA
9482 (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9483 except OpenSSL.crypto.Error, err:
9484 raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9485 (err, ), errors.ECODE_INVAL)
9487 (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9488 if errcode is not None:
9489 raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9490 (msg, ), errors.ECODE_INVAL)
9492 self.dest_x509_ca = cert
9494 # Verify target information
9496 for idx, disk_data in enumerate(self.op.target_node):
9498 (host, port, magic) = \
9499 masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9500 except errors.GenericError, err:
9501 raise errors.OpPrereqError("Target info for disk %s: %s" %
9502 (idx, err), errors.ECODE_INVAL)
9504 disk_info.append((host, port, magic))
9506 assert len(disk_info) == len(self.op.target_node)
9507 self.dest_disk_info = disk_info
9510 raise errors.ProgrammerError("Unhandled export mode %r" %
9513 # instance disk type verification
9514 # TODO: Implement export support for file-based disks
9515 for disk in self.instance.disks:
9516 if disk.dev_type == constants.LD_FILE:
9517 raise errors.OpPrereqError("Export not supported for instances with"
9518 " file-based disks", errors.ECODE_INVAL)
9520 def _CleanupExports(self, feedback_fn):
9521 """Removes exports of current instance from all other nodes.
9523 If an instance in a cluster with nodes A..D was exported to node C, its
9524 exports will be removed from the nodes A, B and D.
9527 assert self.op.mode != constants.EXPORT_MODE_REMOTE
9529 nodelist = self.cfg.GetNodeList()
9530 nodelist.remove(self.dst_node.name)
9532 # on one-node clusters nodelist will be empty after the removal
9533 # if we proceed the backup would be removed because OpQueryExports
9534 # substitutes an empty list with the full cluster node list.
9535 iname = self.instance.name
9537 feedback_fn("Removing old exports for instance %s" % iname)
9538 exportlist = self.rpc.call_export_list(nodelist)
9539 for node in exportlist:
9540 if exportlist[node].fail_msg:
9542 if iname in exportlist[node].payload:
9543 msg = self.rpc.call_export_remove(node, iname).fail_msg
9545 self.LogWarning("Could not remove older export for instance %s"
9546 " on node %s: %s", iname, node, msg)
9548 def Exec(self, feedback_fn):
9549 """Export an instance to an image in the cluster.
9552 assert self.op.mode in constants.EXPORT_MODES
9554 instance = self.instance
9555 src_node = instance.primary_node
9557 if self.op.shutdown:
9558 # shutdown the instance, but not the disks
9559 feedback_fn("Shutting down instance %s" % instance.name)
9560 result = self.rpc.call_instance_shutdown(src_node, instance,
9561 self.op.shutdown_timeout)
9562 # TODO: Maybe ignore failures if ignore_remove_failures is set
9563 result.Raise("Could not shutdown instance %s on"
9564 " node %s" % (instance.name, src_node))
9566 # set the disks ID correctly since call_instance_start needs the
9567 # correct drbd minor to create the symlinks
9568 for disk in instance.disks:
9569 self.cfg.SetDiskID(disk, src_node)
9571 activate_disks = (not instance.admin_up)
9574 # Activate the instance disks if we'exporting a stopped instance
9575 feedback_fn("Activating disks for %s" % instance.name)
9576 _StartInstanceDisks(self, instance, None)
9579 helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9582 helper.CreateSnapshots()
9584 if (self.op.shutdown and instance.admin_up and
9585 not self.op.remove_instance):
9586 assert not activate_disks
9587 feedback_fn("Starting instance %s" % instance.name)
9588 result = self.rpc.call_instance_start(src_node, instance, None, None)
9589 msg = result.fail_msg
9591 feedback_fn("Failed to start instance: %s" % msg)
9592 _ShutdownInstanceDisks(self, instance)
9593 raise errors.OpExecError("Could not start instance: %s" % msg)
9595 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9596 (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9597 elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9598 connect_timeout = constants.RIE_CONNECT_TIMEOUT
9599 timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9601 (key_name, _, _) = self.x509_key_name
9604 OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9607 (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9608 key_name, dest_ca_pem,
9613 # Check for backwards compatibility
9614 assert len(dresults) == len(instance.disks)
9615 assert compat.all(isinstance(i, bool) for i in dresults), \
9616 "Not all results are boolean: %r" % dresults
9620 feedback_fn("Deactivating disks for %s" % instance.name)
9621 _ShutdownInstanceDisks(self, instance)
9623 if not (compat.all(dresults) and fin_resu):
9626 failures.append("export finalization")
9627 if not compat.all(dresults):
9628 fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9630 failures.append("disk export: disk(s) %s" % fdsk)
9632 raise errors.OpExecError("Export failed, errors in %s" %
9633 utils.CommaJoin(failures))
9635 # At this point, the export was successful, we can cleanup/finish
9637 # Remove instance if requested
9638 if self.op.remove_instance:
9639 feedback_fn("Removing instance %s" % instance.name)
9640 _RemoveInstance(self, feedback_fn, instance,
9641 self.op.ignore_remove_failures)
9643 if self.op.mode == constants.EXPORT_MODE_LOCAL:
9644 self._CleanupExports(feedback_fn)
9646 return fin_resu, dresults
9649 class LURemoveExport(NoHooksLU):
9650 """Remove exports related to the named instance.
9658 def ExpandNames(self):
9659 self.needed_locks = {}
9660 # We need all nodes to be locked in order for RemoveExport to work, but we
9661 # don't need to lock the instance itself, as nothing will happen to it (and
9662 # we can remove exports also for a removed instance)
9663 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9665 def Exec(self, feedback_fn):
9666 """Remove any export.
9669 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9670 # If the instance was not found we'll try with the name that was passed in.
9671 # This will only work if it was an FQDN, though.
9673 if not instance_name:
9675 instance_name = self.op.instance_name
9677 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9678 exportlist = self.rpc.call_export_list(locked_nodes)
9680 for node in exportlist:
9681 msg = exportlist[node].fail_msg
9683 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9685 if instance_name in exportlist[node].payload:
9687 result = self.rpc.call_export_remove(node, instance_name)
9688 msg = result.fail_msg
9690 logging.error("Could not remove export for instance %s"
9691 " on node %s: %s", instance_name, node, msg)
9693 if fqdn_warn and not found:
9694 feedback_fn("Export not found. If trying to remove an export belonging"
9695 " to a deleted instance please use its Fully Qualified"
9699 class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9702 This is an abstract class which is the parent of all the other tags LUs.
9706 def ExpandNames(self):
9707 self.needed_locks = {}
9708 if self.op.kind == constants.TAG_NODE:
9709 self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9710 self.needed_locks[locking.LEVEL_NODE] = self.op.name
9711 elif self.op.kind == constants.TAG_INSTANCE:
9712 self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9713 self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9715 def CheckPrereq(self):
9716 """Check prerequisites.
9719 if self.op.kind == constants.TAG_CLUSTER:
9720 self.target = self.cfg.GetClusterInfo()
9721 elif self.op.kind == constants.TAG_NODE:
9722 self.target = self.cfg.GetNodeInfo(self.op.name)
9723 elif self.op.kind == constants.TAG_INSTANCE:
9724 self.target = self.cfg.GetInstanceInfo(self.op.name)
9726 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9727 str(self.op.kind), errors.ECODE_INVAL)
9730 class LUGetTags(TagsLU):
9731 """Returns the tags of a given object.
9735 ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9736 ("name", _NoDefault, _TNonEmptyString),
9740 def Exec(self, feedback_fn):
9741 """Returns the tag list.
9744 return list(self.target.GetTags())
9747 class LUSearchTags(NoHooksLU):
9748 """Searches the tags for a given pattern.
9752 ("pattern", _NoDefault, _TNonEmptyString),
9756 def ExpandNames(self):
9757 self.needed_locks = {}
9759 def CheckPrereq(self):
9760 """Check prerequisites.
9762 This checks the pattern passed for validity by compiling it.
9766 self.re = re.compile(self.op.pattern)
9767 except re.error, err:
9768 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9769 (self.op.pattern, err), errors.ECODE_INVAL)
9771 def Exec(self, feedback_fn):
9772 """Returns the tag list.
9776 tgts = [("/cluster", cfg.GetClusterInfo())]
9777 ilist = cfg.GetAllInstancesInfo().values()
9778 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9779 nlist = cfg.GetAllNodesInfo().values()
9780 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9782 for path, target in tgts:
9783 for tag in target.GetTags():
9784 if self.re.search(tag):
9785 results.append((path, tag))
9789 class LUAddTags(TagsLU):
9790 """Sets a tag on a given object.
9794 ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9795 ("name", _NoDefault, _TNonEmptyString),
9796 ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9800 def CheckPrereq(self):
9801 """Check prerequisites.
9803 This checks the type and length of the tag name and value.
9806 TagsLU.CheckPrereq(self)
9807 for tag in self.op.tags:
9808 objects.TaggableObject.ValidateTag(tag)
9810 def Exec(self, feedback_fn):
9815 for tag in self.op.tags:
9816 self.target.AddTag(tag)
9817 except errors.TagError, err:
9818 raise errors.OpExecError("Error while setting tag: %s" % str(err))
9819 self.cfg.Update(self.target, feedback_fn)
9822 class LUDelTags(TagsLU):
9823 """Delete a list of tags from a given object.
9827 ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9828 ("name", _NoDefault, _TNonEmptyString),
9829 ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9833 def CheckPrereq(self):
9834 """Check prerequisites.
9836 This checks that we have the given tag.
9839 TagsLU.CheckPrereq(self)
9840 for tag in self.op.tags:
9841 objects.TaggableObject.ValidateTag(tag)
9842 del_tags = frozenset(self.op.tags)
9843 cur_tags = self.target.GetTags()
9844 if not del_tags <= cur_tags:
9845 diff_tags = del_tags - cur_tags
9846 diff_names = ["'%s'" % tag for tag in diff_tags]
9848 raise errors.OpPrereqError("Tag(s) %s not found" %
9849 (",".join(diff_names)), errors.ECODE_NOENT)
9851 def Exec(self, feedback_fn):
9852 """Remove the tag from the object.
9855 for tag in self.op.tags:
9856 self.target.RemoveTag(tag)
9857 self.cfg.Update(self.target, feedback_fn)
9860 class LUTestDelay(NoHooksLU):
9861 """Sleep for a specified amount of time.
9863 This LU sleeps on the master and/or nodes for a specified amount of
9868 ("duration", _NoDefault, _TFloat),
9869 ("on_master", True, _TBool),
9870 ("on_nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9871 ("repeat", 0, _TPositiveInt)
9875 def ExpandNames(self):
9876 """Expand names and set required locks.
9878 This expands the node list, if any.
9881 self.needed_locks = {}
9882 if self.op.on_nodes:
9883 # _GetWantedNodes can be used here, but is not always appropriate to use
9884 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9886 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9887 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9889 def _TestDelay(self):
9890 """Do the actual sleep.
9893 if self.op.on_master:
9894 if not utils.TestDelay(self.op.duration):
9895 raise errors.OpExecError("Error during master delay test")
9896 if self.op.on_nodes:
9897 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9898 for node, node_result in result.items():
9899 node_result.Raise("Failure during rpc call to node %s" % node)
9901 def Exec(self, feedback_fn):
9902 """Execute the test delay opcode, with the wanted repetitions.
9905 if self.op.repeat == 0:
9908 top_value = self.op.repeat - 1
9909 for i in range(self.op.repeat):
9910 self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
9914 class LUTestJobqueue(NoHooksLU):
9915 """Utility LU to test some aspects of the job queue.
9919 ("notify_waitlock", False, _TBool),
9920 ("notify_exec", False, _TBool),
9921 ("log_messages", _EmptyList, _TListOf(_TString)),
9922 ("fail", False, _TBool),
9926 # Must be lower than default timeout for WaitForJobChange to see whether it
9927 # notices changed jobs
9928 _CLIENT_CONNECT_TIMEOUT = 20.0
9929 _CLIENT_CONFIRM_TIMEOUT = 60.0
9932 def _NotifyUsingSocket(cls, cb, errcls):
9933 """Opens a Unix socket and waits for another program to connect.
9936 @param cb: Callback to send socket name to client
9938 @param errcls: Exception class to use for errors
9941 # Using a temporary directory as there's no easy way to create temporary
9942 # sockets without writing a custom loop around tempfile.mktemp and
9944 tmpdir = tempfile.mkdtemp()
9946 tmpsock = utils.PathJoin(tmpdir, "sock")
9948 logging.debug("Creating temporary socket at %s", tmpsock)
9949 sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
9954 # Send details to client
9957 # Wait for client to connect before continuing
9958 sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
9960 (conn, _) = sock.accept()
9961 except socket.error, err:
9962 raise errcls("Client didn't connect in time (%s)" % err)
9966 # Remove as soon as client is connected
9967 shutil.rmtree(tmpdir)
9969 # Wait for client to close
9972 # pylint: disable-msg=E1101
9973 # Instance of '_socketobject' has no ... member
9974 conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
9976 except socket.error, err:
9977 raise errcls("Client failed to confirm notification (%s)" % err)
9981 def _SendNotification(self, test, arg, sockname):
9982 """Sends a notification to the client.
9985 @param test: Test name
9986 @param arg: Test argument (depends on test)
9987 @type sockname: string
9988 @param sockname: Socket path
9991 self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
9993 def _Notify(self, prereq, test, arg):
9994 """Notifies the client of a test.
9997 @param prereq: Whether this is a prereq-phase test
9999 @param test: Test name
10000 @param arg: Test argument (depends on test)
10004 errcls = errors.OpPrereqError
10006 errcls = errors.OpExecError
10008 return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10012 def CheckArguments(self):
10013 self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10014 self.expandnames_calls = 0
10016 def ExpandNames(self):
10017 checkargs_calls = getattr(self, "checkargs_calls", 0)
10018 if checkargs_calls < 1:
10019 raise errors.ProgrammerError("CheckArguments was not called")
10021 self.expandnames_calls += 1
10023 if self.op.notify_waitlock:
10024 self._Notify(True, constants.JQT_EXPANDNAMES, None)
10026 self.LogInfo("Expanding names")
10028 # Get lock on master node (just to get a lock, not for a particular reason)
10029 self.needed_locks = {
10030 locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10033 def Exec(self, feedback_fn):
10034 if self.expandnames_calls < 1:
10035 raise errors.ProgrammerError("ExpandNames was not called")
10037 if self.op.notify_exec:
10038 self._Notify(False, constants.JQT_EXEC, None)
10040 self.LogInfo("Executing")
10042 if self.op.log_messages:
10043 for idx, msg in enumerate(self.op.log_messages):
10044 self.LogInfo("Sending log message %s", idx + 1)
10045 feedback_fn(constants.JQT_MSGPREFIX + msg)
10046 # Report how many test messages have been sent
10047 self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10050 raise errors.OpExecError("Opcode failure was requested")
10055 class IAllocator(object):
10056 """IAllocator framework.
10058 An IAllocator instance has three sets of attributes:
10059 - cfg that is needed to query the cluster
10060 - input data (all members of the _KEYS class attribute are required)
10061 - four buffer attributes (in|out_data|text), that represent the
10062 input (to the external script) in text and data structure format,
10063 and the output from it, again in two formats
10064 - the result variables from the script (success, info, nodes) for
10068 # pylint: disable-msg=R0902
10069 # lots of instance attributes
10071 "name", "mem_size", "disks", "disk_template",
10072 "os", "tags", "nics", "vcpus", "hypervisor",
10075 "name", "relocate_from",
10081 def __init__(self, cfg, rpc, mode, **kwargs):
10084 # init buffer variables
10085 self.in_text = self.out_text = self.in_data = self.out_data = None
10086 # init all input fields so that pylint is happy
10088 self.mem_size = self.disks = self.disk_template = None
10089 self.os = self.tags = self.nics = self.vcpus = None
10090 self.hypervisor = None
10091 self.relocate_from = None
10093 self.evac_nodes = None
10095 self.required_nodes = None
10096 # init result fields
10097 self.success = self.info = self.result = None
10098 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10099 keyset = self._ALLO_KEYS
10100 fn = self._AddNewInstance
10101 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10102 keyset = self._RELO_KEYS
10103 fn = self._AddRelocateInstance
10104 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10105 keyset = self._EVAC_KEYS
10106 fn = self._AddEvacuateNodes
10108 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10109 " IAllocator" % self.mode)
10111 if key not in keyset:
10112 raise errors.ProgrammerError("Invalid input parameter '%s' to"
10113 " IAllocator" % key)
10114 setattr(self, key, kwargs[key])
10117 if key not in kwargs:
10118 raise errors.ProgrammerError("Missing input parameter '%s' to"
10119 " IAllocator" % key)
10120 self._BuildInputData(fn)
10122 def _ComputeClusterData(self):
10123 """Compute the generic allocator input data.
10125 This is the data that is independent of the actual operation.
10129 cluster_info = cfg.GetClusterInfo()
10132 "version": constants.IALLOCATOR_VERSION,
10133 "cluster_name": cfg.GetClusterName(),
10134 "cluster_tags": list(cluster_info.GetTags()),
10135 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10136 # we don't have job IDs
10138 iinfo = cfg.GetAllInstancesInfo().values()
10139 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10143 node_list = cfg.GetNodeList()
10145 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10146 hypervisor_name = self.hypervisor
10147 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10148 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10149 elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10150 hypervisor_name = cluster_info.enabled_hypervisors[0]
10152 node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10155 self.rpc.call_all_instances_info(node_list,
10156 cluster_info.enabled_hypervisors)
10157 for nname, nresult in node_data.items():
10158 # first fill in static (config-based) values
10159 ninfo = cfg.GetNodeInfo(nname)
10161 "tags": list(ninfo.GetTags()),
10162 "primary_ip": ninfo.primary_ip,
10163 "secondary_ip": ninfo.secondary_ip,
10164 "offline": ninfo.offline,
10165 "drained": ninfo.drained,
10166 "master_candidate": ninfo.master_candidate,
10169 if not (ninfo.offline or ninfo.drained):
10170 nresult.Raise("Can't get data for node %s" % nname)
10171 node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10173 remote_info = nresult.payload
10175 for attr in ['memory_total', 'memory_free', 'memory_dom0',
10176 'vg_size', 'vg_free', 'cpu_total']:
10177 if attr not in remote_info:
10178 raise errors.OpExecError("Node '%s' didn't return attribute"
10179 " '%s'" % (nname, attr))
10180 if not isinstance(remote_info[attr], int):
10181 raise errors.OpExecError("Node '%s' returned invalid value"
10183 (nname, attr, remote_info[attr]))
10184 # compute memory used by primary instances
10185 i_p_mem = i_p_up_mem = 0
10186 for iinfo, beinfo in i_list:
10187 if iinfo.primary_node == nname:
10188 i_p_mem += beinfo[constants.BE_MEMORY]
10189 if iinfo.name not in node_iinfo[nname].payload:
10192 i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10193 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10194 remote_info['memory_free'] -= max(0, i_mem_diff)
10197 i_p_up_mem += beinfo[constants.BE_MEMORY]
10199 # compute memory used by instances
10201 "total_memory": remote_info['memory_total'],
10202 "reserved_memory": remote_info['memory_dom0'],
10203 "free_memory": remote_info['memory_free'],
10204 "total_disk": remote_info['vg_size'],
10205 "free_disk": remote_info['vg_free'],
10206 "total_cpus": remote_info['cpu_total'],
10207 "i_pri_memory": i_p_mem,
10208 "i_pri_up_memory": i_p_up_mem,
10210 pnr.update(pnr_dyn)
10212 node_results[nname] = pnr
10213 data["nodes"] = node_results
10217 for iinfo, beinfo in i_list:
10219 for nic in iinfo.nics:
10220 filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10221 nic_dict = {"mac": nic.mac,
10223 "mode": filled_params[constants.NIC_MODE],
10224 "link": filled_params[constants.NIC_LINK],
10226 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10227 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10228 nic_data.append(nic_dict)
10230 "tags": list(iinfo.GetTags()),
10231 "admin_up": iinfo.admin_up,
10232 "vcpus": beinfo[constants.BE_VCPUS],
10233 "memory": beinfo[constants.BE_MEMORY],
10235 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10237 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10238 "disk_template": iinfo.disk_template,
10239 "hypervisor": iinfo.hypervisor,
10241 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10243 instance_data[iinfo.name] = pir
10245 data["instances"] = instance_data
10247 self.in_data = data
10249 def _AddNewInstance(self):
10250 """Add new instance data to allocator structure.
10252 This in combination with _AllocatorGetClusterData will create the
10253 correct structure needed as input for the allocator.
10255 The checks for the completeness of the opcode must have already been
10259 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10261 if self.disk_template in constants.DTS_NET_MIRROR:
10262 self.required_nodes = 2
10264 self.required_nodes = 1
10267 "disk_template": self.disk_template,
10270 "vcpus": self.vcpus,
10271 "memory": self.mem_size,
10272 "disks": self.disks,
10273 "disk_space_total": disk_space,
10275 "required_nodes": self.required_nodes,
10279 def _AddRelocateInstance(self):
10280 """Add relocate instance data to allocator structure.
10282 This in combination with _IAllocatorGetClusterData will create the
10283 correct structure needed as input for the allocator.
10285 The checks for the completeness of the opcode must have already been
10289 instance = self.cfg.GetInstanceInfo(self.name)
10290 if instance is None:
10291 raise errors.ProgrammerError("Unknown instance '%s' passed to"
10292 " IAllocator" % self.name)
10294 if instance.disk_template not in constants.DTS_NET_MIRROR:
10295 raise errors.OpPrereqError("Can't relocate non-mirrored instances",
10296 errors.ECODE_INVAL)
10298 if len(instance.secondary_nodes) != 1:
10299 raise errors.OpPrereqError("Instance has not exactly one secondary node",
10300 errors.ECODE_STATE)
10302 self.required_nodes = 1
10303 disk_sizes = [{'size': disk.size} for disk in instance.disks]
10304 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
10308 "disk_space_total": disk_space,
10309 "required_nodes": self.required_nodes,
10310 "relocate_from": self.relocate_from,
10314 def _AddEvacuateNodes(self):
10315 """Add evacuate nodes data to allocator structure.
10319 "evac_nodes": self.evac_nodes
10323 def _BuildInputData(self, fn):
10324 """Build input data structures.
10327 self._ComputeClusterData()
10330 request["type"] = self.mode
10331 self.in_data["request"] = request
10333 self.in_text = serializer.Dump(self.in_data)
10335 def Run(self, name, validate=True, call_fn=None):
10336 """Run an instance allocator and return the results.
10339 if call_fn is None:
10340 call_fn = self.rpc.call_iallocator_runner
10342 result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
10343 result.Raise("Failure while running the iallocator script")
10345 self.out_text = result.payload
10347 self._ValidateResult()
10349 def _ValidateResult(self):
10350 """Process the allocator results.
10352 This will process and if successful save the result in
10353 self.out_data and the other parameters.
10357 rdict = serializer.Load(self.out_text)
10358 except Exception, err:
10359 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
10361 if not isinstance(rdict, dict):
10362 raise errors.OpExecError("Can't parse iallocator results: not a dict")
10364 # TODO: remove backwards compatiblity in later versions
10365 if "nodes" in rdict and "result" not in rdict:
10366 rdict["result"] = rdict["nodes"]
10369 for key in "success", "info", "result":
10370 if key not in rdict:
10371 raise errors.OpExecError("Can't parse iallocator results:"
10372 " missing key '%s'" % key)
10373 setattr(self, key, rdict[key])
10375 if not isinstance(rdict["result"], list):
10376 raise errors.OpExecError("Can't parse iallocator results: 'result' key"
10378 self.out_data = rdict
10381 class LUTestAllocator(NoHooksLU):
10382 """Run allocator tests.
10384 This LU runs the allocator tests
10388 ("direction", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
10389 ("mode", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_MODES)),
10390 ("name", _NoDefault, _TNonEmptyString),
10391 ("nics", _NoDefault, _TOr(_TNone, _TListOf(
10392 _TDictOf(_TElemOf(["mac", "ip", "bridge"]),
10393 _TOr(_TNone, _TNonEmptyString))))),
10394 ("disks", _NoDefault, _TOr(_TNone, _TList)),
10395 ("hypervisor", None, _TMaybeString),
10396 ("allocator", None, _TMaybeString),
10397 ("tags", _EmptyList, _TListOf(_TNonEmptyString)),
10398 ("mem_size", None, _TOr(_TNone, _TPositiveInt)),
10399 ("vcpus", None, _TOr(_TNone, _TPositiveInt)),
10400 ("os", None, _TMaybeString),
10401 ("disk_template", None, _TMaybeString),
10402 ("evac_nodes", None, _TOr(_TNone, _TListOf(_TNonEmptyString))),
10405 def CheckPrereq(self):
10406 """Check prerequisites.
10408 This checks the opcode parameters depending on the director and mode test.
10411 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10412 for attr in ["mem_size", "disks", "disk_template",
10413 "os", "tags", "nics", "vcpus"]:
10414 if not hasattr(self.op, attr):
10415 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
10416 attr, errors.ECODE_INVAL)
10417 iname = self.cfg.ExpandInstanceName(self.op.name)
10418 if iname is not None:
10419 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
10420 iname, errors.ECODE_EXISTS)
10421 if not isinstance(self.op.nics, list):
10422 raise errors.OpPrereqError("Invalid parameter 'nics'",
10423 errors.ECODE_INVAL)
10424 if not isinstance(self.op.disks, list):
10425 raise errors.OpPrereqError("Invalid parameter 'disks'",
10426 errors.ECODE_INVAL)
10427 for row in self.op.disks:
10428 if (not isinstance(row, dict) or
10429 "size" not in row or
10430 not isinstance(row["size"], int) or
10431 "mode" not in row or
10432 row["mode"] not in ['r', 'w']):
10433 raise errors.OpPrereqError("Invalid contents of the 'disks'"
10434 " parameter", errors.ECODE_INVAL)
10435 if self.op.hypervisor is None:
10436 self.op.hypervisor = self.cfg.GetHypervisorType()
10437 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10438 fname = _ExpandInstanceName(self.cfg, self.op.name)
10439 self.op.name = fname
10440 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
10441 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10442 if not hasattr(self.op, "evac_nodes"):
10443 raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
10444 " opcode input", errors.ECODE_INVAL)
10446 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
10447 self.op.mode, errors.ECODE_INVAL)
10449 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
10450 if self.op.allocator is None:
10451 raise errors.OpPrereqError("Missing allocator name",
10452 errors.ECODE_INVAL)
10453 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
10454 raise errors.OpPrereqError("Wrong allocator test '%s'" %
10455 self.op.direction, errors.ECODE_INVAL)
10457 def Exec(self, feedback_fn):
10458 """Run the allocator test.
10461 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10462 ial = IAllocator(self.cfg, self.rpc,
10465 mem_size=self.op.mem_size,
10466 disks=self.op.disks,
10467 disk_template=self.op.disk_template,
10471 vcpus=self.op.vcpus,
10472 hypervisor=self.op.hypervisor,
10474 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10475 ial = IAllocator(self.cfg, self.rpc,
10478 relocate_from=list(self.relocate_from),
10480 elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10481 ial = IAllocator(self.cfg, self.rpc,
10483 evac_nodes=self.op.evac_nodes)
10485 raise errors.ProgrammerError("Uncatched mode %s in"
10486 " LUTestAllocator.Exec", self.op.mode)
10488 if self.op.direction == constants.IALLOCATOR_DIR_IN:
10489 result = ial.in_text
10491 ial.Run(self.op.allocator, validate=False)
10492 result = ial.out_text