4 # Copyright (C) 2006, 2007, 2008 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=W0613,W0201
36 from ganeti import ssh
37 from ganeti import utils
38 from ganeti import errors
39 from ganeti import hypervisor
40 from ganeti import locking
41 from ganeti import constants
42 from ganeti import objects
43 from ganeti import opcodes
44 from ganeti import serializer
45 from ganeti import ssconf
48 class LogicalUnit(object):
49 """Logical Unit base class.
51 Subclasses must follow these rules:
52 - implement ExpandNames
53 - implement CheckPrereq
55 - implement BuildHooksEnv
56 - redefine HPATH and HTYPE
57 - optionally redefine their run requirements:
58 REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
60 Note that all commands require root permissions.
68 def __init__(self, processor, op, context, rpc):
69 """Constructor for LogicalUnit.
71 This needs to be overriden in derived classes in order to check op
77 self.cfg = context.cfg
78 self.context = context
80 # Dicts used to declare locking needs to mcpu
81 self.needed_locks = None
82 self.acquired_locks = {}
83 self.share_locks = dict(((i, 0) for i in locking.LEVELS))
85 self.remove_locks = {}
86 # Used to force good behavior when calling helper functions
87 self.recalculate_locks = {}
90 self.LogWarning = processor.LogWarning
91 self.LogInfo = processor.LogInfo
93 for attr_name in self._OP_REQP:
94 attr_val = getattr(op, attr_name, None)
96 raise errors.OpPrereqError("Required parameter '%s' missing" %
101 """Returns the SshRunner object
105 self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
108 ssh = property(fget=__GetSSH)
110 def CheckArguments(self):
111 """Check syntactic validity for the opcode arguments.
113 This method is for doing a simple syntactic check and ensure
114 validity of opcode parameters, without any cluster-related
115 checks. While the same can be accomplished in ExpandNames and/or
116 CheckPrereq, doing these separate is better because:
118 - ExpandNames is left as as purely a lock-related function
119 - CheckPrereq is run after we have aquired locks (and possible
122 The function is allowed to change the self.op attribute so that
123 later methods can no longer worry about missing parameters.
128 def ExpandNames(self):
129 """Expand names for this LU.
131 This method is called before starting to execute the opcode, and it should
132 update all the parameters of the opcode to their canonical form (e.g. a
133 short node name must be fully expanded after this method has successfully
134 completed). This way locking, hooks, logging, ecc. can work correctly.
136 LUs which implement this method must also populate the self.needed_locks
137 member, as a dict with lock levels as keys, and a list of needed lock names
140 - use an empty dict if you don't need any lock
141 - if you don't need any lock at a particular level omit that level
142 - don't put anything for the BGL level
143 - if you want all locks at a level use locking.ALL_SET as a value
145 If you need to share locks (rather than acquire them exclusively) at one
146 level you can modify self.share_locks, setting a true value (usually 1) for
147 that level. By default locks are not shared.
151 # Acquire all nodes and one instance
152 self.needed_locks = {
153 locking.LEVEL_NODE: locking.ALL_SET,
154 locking.LEVEL_INSTANCE: ['instance1.example.tld'],
156 # Acquire just two nodes
157 self.needed_locks = {
158 locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
161 self.needed_locks = {} # No, you can't leave it to the default value None
164 # The implementation of this method is mandatory only if the new LU is
165 # concurrent, so that old LUs don't need to be changed all at the same
168 self.needed_locks = {} # Exclusive LUs don't need locks.
170 raise NotImplementedError
172 def DeclareLocks(self, level):
173 """Declare LU locking needs for a level
175 While most LUs can just declare their locking needs at ExpandNames time,
176 sometimes there's the need to calculate some locks after having acquired
177 the ones before. This function is called just before acquiring locks at a
178 particular level, but after acquiring the ones at lower levels, and permits
179 such calculations. It can be used to modify self.needed_locks, and by
180 default it does nothing.
182 This function is only called if you have something already set in
183 self.needed_locks for the level.
185 @param level: Locking level which is going to be locked
186 @type level: member of ganeti.locking.LEVELS
190 def CheckPrereq(self):
191 """Check prerequisites for this LU.
193 This method should check that the prerequisites for the execution
194 of this LU are fulfilled. It can do internode communication, but
195 it should be idempotent - no cluster or system changes are
198 The method should raise errors.OpPrereqError in case something is
199 not fulfilled. Its return value is ignored.
201 This method should also update all the parameters of the opcode to
202 their canonical form if it hasn't been done by ExpandNames before.
205 raise NotImplementedError
207 def Exec(self, feedback_fn):
210 This method should implement the actual work. It should raise
211 errors.OpExecError for failures that are somewhat dealt with in
215 raise NotImplementedError
217 def BuildHooksEnv(self):
218 """Build hooks environment for this LU.
220 This method should return a three-node tuple consisting of: a dict
221 containing the environment that will be used for running the
222 specific hook for this LU, a list of node names on which the hook
223 should run before the execution, and a list of node names on which
224 the hook should run after the execution.
226 The keys of the dict must not have 'GANETI_' prefixed as this will
227 be handled in the hooks runner. Also note additional keys will be
228 added by the hooks runner. If the LU doesn't define any
229 environment, an empty dict (and not None) should be returned.
231 No nodes should be returned as an empty list (and not None).
233 Note that if the HPATH for a LU class is None, this function will
237 raise NotImplementedError
239 def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
240 """Notify the LU about the results of its hooks.
242 This method is called every time a hooks phase is executed, and notifies
243 the Logical Unit about the hooks' result. The LU can then use it to alter
244 its result based on the hooks. By default the method does nothing and the
245 previous result is passed back unchanged but any LU can define it if it
246 wants to use the local cluster hook-scripts somehow.
248 @param phase: one of L{constants.HOOKS_PHASE_POST} or
249 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
250 @param hook_results: the results of the multi-node hooks rpc call
251 @param feedback_fn: function used send feedback back to the caller
252 @param lu_result: the previous Exec result this LU had, or None
254 @return: the new Exec result, based on the previous result
260 def _ExpandAndLockInstance(self):
261 """Helper function to expand and lock an instance.
263 Many LUs that work on an instance take its name in self.op.instance_name
264 and need to expand it and then declare the expanded name for locking. This
265 function does it, and then updates self.op.instance_name to the expanded
266 name. It also initializes needed_locks as a dict, if this hasn't been done
270 if self.needed_locks is None:
271 self.needed_locks = {}
273 assert locking.LEVEL_INSTANCE not in self.needed_locks, \
274 "_ExpandAndLockInstance called with instance-level locks set"
275 expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
276 if expanded_name is None:
277 raise errors.OpPrereqError("Instance '%s' not known" %
278 self.op.instance_name)
279 self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
280 self.op.instance_name = expanded_name
282 def _LockInstancesNodes(self, primary_only=False):
283 """Helper function to declare instances' nodes for locking.
285 This function should be called after locking one or more instances to lock
286 their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
287 with all primary or secondary nodes for instances already locked and
288 present in self.needed_locks[locking.LEVEL_INSTANCE].
290 It should be called from DeclareLocks, and for safety only works if
291 self.recalculate_locks[locking.LEVEL_NODE] is set.
293 In the future it may grow parameters to just lock some instance's nodes, or
294 to just lock primaries or secondary nodes, if needed.
296 If should be called in DeclareLocks in a way similar to::
298 if level == locking.LEVEL_NODE:
299 self._LockInstancesNodes()
301 @type primary_only: boolean
302 @param primary_only: only lock primary nodes of locked instances
305 assert locking.LEVEL_NODE in self.recalculate_locks, \
306 "_LockInstancesNodes helper function called with no nodes to recalculate"
308 # TODO: check if we're really been called with the instance locks held
310 # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
311 # future we might want to have different behaviors depending on the value
312 # of self.recalculate_locks[locking.LEVEL_NODE]
314 for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
315 instance = self.context.cfg.GetInstanceInfo(instance_name)
316 wanted_nodes.append(instance.primary_node)
318 wanted_nodes.extend(instance.secondary_nodes)
320 if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
321 self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
322 elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
323 self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
325 del self.recalculate_locks[locking.LEVEL_NODE]
328 class NoHooksLU(LogicalUnit):
329 """Simple LU which runs no hooks.
331 This LU is intended as a parent for other LogicalUnits which will
332 run no hooks, in order to reduce duplicate code.
339 def _GetWantedNodes(lu, nodes):
340 """Returns list of checked and expanded node names.
342 @type lu: L{LogicalUnit}
343 @param lu: the logical unit on whose behalf we execute
345 @param nodes: list of node names or None for all nodes
347 @return: the list of nodes, sorted
348 @raise errors.OpProgrammerError: if the nodes parameter is wrong type
351 if not isinstance(nodes, list):
352 raise errors.OpPrereqError("Invalid argument type 'nodes'")
355 raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
356 " non-empty list of nodes whose name is to be expanded.")
360 node = lu.cfg.ExpandNodeName(name)
362 raise errors.OpPrereqError("No such node name '%s'" % name)
365 return utils.NiceSort(wanted)
368 def _GetWantedInstances(lu, instances):
369 """Returns list of checked and expanded instance names.
371 @type lu: L{LogicalUnit}
372 @param lu: the logical unit on whose behalf we execute
373 @type instances: list
374 @param instances: list of instance names or None for all instances
376 @return: the list of instances, sorted
377 @raise errors.OpPrereqError: if the instances parameter is wrong type
378 @raise errors.OpPrereqError: if any of the passed instances is not found
381 if not isinstance(instances, list):
382 raise errors.OpPrereqError("Invalid argument type 'instances'")
387 for name in instances:
388 instance = lu.cfg.ExpandInstanceName(name)
390 raise errors.OpPrereqError("No such instance name '%s'" % name)
391 wanted.append(instance)
394 wanted = utils.NiceSort(lu.cfg.GetInstanceList())
398 def _CheckOutputFields(static, dynamic, selected):
399 """Checks whether all selected fields are valid.
401 @type static: L{utils.FieldSet}
402 @param static: static fields set
403 @type dynamic: L{utils.FieldSet}
404 @param dynamic: dynamic fields set
411 delta = f.NonMatching(selected)
413 raise errors.OpPrereqError("Unknown output fields selected: %s"
417 def _CheckBooleanOpField(op, name):
418 """Validates boolean opcode parameters.
420 This will ensure that an opcode parameter is either a boolean value,
421 or None (but that it always exists).
424 val = getattr(op, name, None)
425 if not (val is None or isinstance(val, bool)):
426 raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
428 setattr(op, name, val)
431 def _CheckNodeOnline(lu, node):
432 """Ensure that a given node is online.
434 @param lu: the LU on behalf of which we make the check
435 @param node: the node to check
436 @raise errors.OpPrereqError: if the node is offline
439 if lu.cfg.GetNodeInfo(node).offline:
440 raise errors.OpPrereqError("Can't use offline node %s" % node)
443 def _CheckNodeNotDrained(lu, node):
444 """Ensure that a given node is not drained.
446 @param lu: the LU on behalf of which we make the check
447 @param node: the node to check
448 @raise errors.OpPrereqError: if the node is drained
451 if lu.cfg.GetNodeInfo(node).drained:
452 raise errors.OpPrereqError("Can't use drained node %s" % node)
455 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
456 memory, vcpus, nics, disk_template, disks):
457 """Builds instance related env variables for hooks
459 This builds the hook environment from individual variables.
462 @param name: the name of the instance
463 @type primary_node: string
464 @param primary_node: the name of the instance's primary node
465 @type secondary_nodes: list
466 @param secondary_nodes: list of secondary nodes as strings
467 @type os_type: string
468 @param os_type: the name of the instance's OS
469 @type status: boolean
470 @param status: the should_run status of the instance
472 @param memory: the memory size of the instance
474 @param vcpus: the count of VCPUs the instance has
476 @param nics: list of tuples (ip, bridge, mac) representing
477 the NICs the instance has
478 @type disk_template: string
479 @param disk_template: the distk template of the instance
481 @param disks: the list of (size, mode) pairs
483 @return: the hook environment for this instance
492 "INSTANCE_NAME": name,
493 "INSTANCE_PRIMARY": primary_node,
494 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
495 "INSTANCE_OS_TYPE": os_type,
496 "INSTANCE_STATUS": str_status,
497 "INSTANCE_MEMORY": memory,
498 "INSTANCE_VCPUS": vcpus,
499 "INSTANCE_DISK_TEMPLATE": disk_template,
503 nic_count = len(nics)
504 for idx, (ip, mac, mode, link) in enumerate(nics):
507 env["INSTANCE_NIC%d_IP" % idx] = ip
508 env["INSTANCE_NIC%d_MAC" % idx] = mac
509 env["INSTANCE_NIC%d_MODE" % idx] = mode
510 env["INSTANCE_NIC%d_LINK" % idx] = link
511 if mode == constants.NIC_MODE_BRIDGED:
512 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
516 env["INSTANCE_NIC_COUNT"] = nic_count
519 disk_count = len(disks)
520 for idx, (size, mode) in enumerate(disks):
521 env["INSTANCE_DISK%d_SIZE" % idx] = size
522 env["INSTANCE_DISK%d_MODE" % idx] = mode
526 env["INSTANCE_DISK_COUNT"] = disk_count
530 def _PreBuildNICHooksList(lu, nics):
531 """Build a list of nic information tuples.
533 This list is suitable to be passed to _BuildInstanceHookEnv.
535 @type lu: L{LogicalUnit}
536 @param lu: the logical unit on whose behalf we execute
537 @type nics: list of L{objects.NIC}
538 @param nics: list of nics to convert to hooks tuples
542 c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
546 filled_params = objects.FillDict(c_nicparams, nic.nicparams)
547 mode = filled_params[constants.NIC_MODE]
548 link = filled_params[constants.NIC_LINK]
549 hooks_nics.append((ip, mac, mode, link))
552 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
553 """Builds instance related env variables for hooks from an object.
555 @type lu: L{LogicalUnit}
556 @param lu: the logical unit on whose behalf we execute
557 @type instance: L{objects.Instance}
558 @param instance: the instance for which we should build the
561 @param override: dictionary with key/values that will override
564 @return: the hook environment dictionary
567 bep = lu.cfg.GetClusterInfo().FillBE(instance)
569 'name': instance.name,
570 'primary_node': instance.primary_node,
571 'secondary_nodes': instance.secondary_nodes,
572 'os_type': instance.os,
573 'status': instance.admin_up,
574 'memory': bep[constants.BE_MEMORY],
575 'vcpus': bep[constants.BE_VCPUS],
576 'nics': _PreBuildNICHooksList(lu, instance.nics),
577 'disk_template': instance.disk_template,
578 'disks': [(disk.size, disk.mode) for disk in instance.disks],
581 args.update(override)
582 return _BuildInstanceHookEnv(**args)
585 def _AdjustCandidatePool(lu):
586 """Adjust the candidate pool after node operations.
589 mod_list = lu.cfg.MaintainCandidatePool()
591 lu.LogInfo("Promoted nodes to master candidate role: %s",
592 ", ".join(node.name for node in mod_list))
593 for name in mod_list:
594 lu.context.ReaddNode(name)
595 mc_now, mc_max = lu.cfg.GetMasterCandidateStats()
597 lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
601 def _CheckNicsBridgesExist(lu, target_nics, target_node,
602 profile=constants.PP_DEFAULT):
603 """Check that the brigdes needed by a list of nics exist.
606 c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
607 paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
608 for nic in target_nics]
609 brlist = [params[constants.NIC_LINK] for params in paramslist
610 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
612 result = lu.rpc.call_bridges_exist(target_node, brlist)
613 msg = result.RemoteFailMsg()
615 raise errors.OpPrereqError("Error checking bridges on destination node"
616 " '%s': %s" % (target_node, msg))
619 def _CheckInstanceBridgesExist(lu, instance, node=None):
620 """Check that the brigdes needed by an instance exist.
624 node=instance.primary_node
625 _CheckNicsBridgesExist(lu, instance.nics, node)
628 class LUDestroyCluster(NoHooksLU):
629 """Logical unit for destroying the cluster.
634 def CheckPrereq(self):
635 """Check prerequisites.
637 This checks whether the cluster is empty.
639 Any errors are signalled by raising errors.OpPrereqError.
642 master = self.cfg.GetMasterNode()
644 nodelist = self.cfg.GetNodeList()
645 if len(nodelist) != 1 or nodelist[0] != master:
646 raise errors.OpPrereqError("There are still %d node(s) in"
647 " this cluster." % (len(nodelist) - 1))
648 instancelist = self.cfg.GetInstanceList()
650 raise errors.OpPrereqError("There are still %d instance(s) in"
651 " this cluster." % len(instancelist))
653 def Exec(self, feedback_fn):
654 """Destroys the cluster.
657 master = self.cfg.GetMasterNode()
658 result = self.rpc.call_node_stop_master(master, False)
661 raise errors.OpExecError("Could not disable the master role")
662 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
663 utils.CreateBackup(priv_key)
664 utils.CreateBackup(pub_key)
668 class LUVerifyCluster(LogicalUnit):
669 """Verifies the cluster status.
672 HPATH = "cluster-verify"
673 HTYPE = constants.HTYPE_CLUSTER
674 _OP_REQP = ["skip_checks"]
677 def ExpandNames(self):
678 self.needed_locks = {
679 locking.LEVEL_NODE: locking.ALL_SET,
680 locking.LEVEL_INSTANCE: locking.ALL_SET,
682 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
684 def _VerifyNode(self, nodeinfo, file_list, local_cksum,
685 node_result, feedback_fn, master_files,
687 """Run multiple tests against a node.
691 - compares ganeti version
692 - checks vg existance and size > 20G
693 - checks config file checksum
694 - checks ssh to other nodes
696 @type nodeinfo: L{objects.Node}
697 @param nodeinfo: the node to check
698 @param file_list: required list of files
699 @param local_cksum: dictionary of local files and their checksums
700 @param node_result: the results from the node
701 @param feedback_fn: function used to accumulate results
702 @param master_files: list of files that only masters should have
703 @param drbd_map: the useddrbd minors for this node, in
704 form of minor: (instance, must_exist) which correspond to instances
705 and their running status
706 @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
711 # main result, node_result should be a non-empty dict
712 if not node_result or not isinstance(node_result, dict):
713 feedback_fn(" - ERROR: unable to verify node %s." % (node,))
716 # compares ganeti version
717 local_version = constants.PROTOCOL_VERSION
718 remote_version = node_result.get('version', None)
719 if not (remote_version and isinstance(remote_version, (list, tuple)) and
720 len(remote_version) == 2):
721 feedback_fn(" - ERROR: connection to %s failed" % (node))
724 if local_version != remote_version[0]:
725 feedback_fn(" - ERROR: incompatible protocol versions: master %s,"
726 " node %s %s" % (local_version, node, remote_version[0]))
729 # node seems compatible, we can actually try to look into its results
733 # full package version
734 if constants.RELEASE_VERSION != remote_version[1]:
735 feedback_fn(" - WARNING: software version mismatch: master %s,"
737 (constants.RELEASE_VERSION, node, remote_version[1]))
739 # checks vg existence and size > 20G
740 if vg_name is not None:
741 vglist = node_result.get(constants.NV_VGLIST, None)
743 feedback_fn(" - ERROR: unable to check volume groups on node %s." %
747 vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
748 constants.MIN_VG_SIZE)
750 feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
753 # checks config file checksum
755 remote_cksum = node_result.get(constants.NV_FILELIST, None)
756 if not isinstance(remote_cksum, dict):
758 feedback_fn(" - ERROR: node hasn't returned file checksum data")
760 for file_name in file_list:
761 node_is_mc = nodeinfo.master_candidate
762 must_have_file = file_name not in master_files
763 if file_name not in remote_cksum:
764 if node_is_mc or must_have_file:
766 feedback_fn(" - ERROR: file '%s' missing" % file_name)
767 elif remote_cksum[file_name] != local_cksum[file_name]:
768 if node_is_mc or must_have_file:
770 feedback_fn(" - ERROR: file '%s' has wrong checksum" % file_name)
772 # not candidate and this is not a must-have file
774 feedback_fn(" - ERROR: non master-candidate has old/wrong file"
777 # all good, except non-master/non-must have combination
778 if not node_is_mc and not must_have_file:
779 feedback_fn(" - ERROR: file '%s' should not exist on non master"
780 " candidates" % file_name)
784 if constants.NV_NODELIST not in node_result:
786 feedback_fn(" - ERROR: node hasn't returned node ssh connectivity data")
788 if node_result[constants.NV_NODELIST]:
790 for node in node_result[constants.NV_NODELIST]:
791 feedback_fn(" - ERROR: ssh communication with node '%s': %s" %
792 (node, node_result[constants.NV_NODELIST][node]))
794 if constants.NV_NODENETTEST not in node_result:
796 feedback_fn(" - ERROR: node hasn't returned node tcp connectivity data")
798 if node_result[constants.NV_NODENETTEST]:
800 nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
802 feedback_fn(" - ERROR: tcp communication with node '%s': %s" %
803 (node, node_result[constants.NV_NODENETTEST][node]))
805 hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
806 if isinstance(hyp_result, dict):
807 for hv_name, hv_result in hyp_result.iteritems():
808 if hv_result is not None:
809 feedback_fn(" - ERROR: hypervisor %s verify failure: '%s'" %
810 (hv_name, hv_result))
812 # check used drbd list
813 if vg_name is not None:
814 used_minors = node_result.get(constants.NV_DRBDLIST, [])
815 if not isinstance(used_minors, (tuple, list)):
816 feedback_fn(" - ERROR: cannot parse drbd status file: %s" %
819 for minor, (iname, must_exist) in drbd_map.items():
820 if minor not in used_minors and must_exist:
821 feedback_fn(" - ERROR: drbd minor %d of instance %s is"
822 " not active" % (minor, iname))
824 for minor in used_minors:
825 if minor not in drbd_map:
826 feedback_fn(" - ERROR: unallocated drbd minor %d is in use" %
832 def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
833 node_instance, feedback_fn, n_offline):
834 """Verify an instance.
836 This function checks to see if the required block devices are
837 available on the instance's node.
842 node_current = instanceconfig.primary_node
845 instanceconfig.MapLVsByNode(node_vol_should)
847 for node in node_vol_should:
848 if node in n_offline:
849 # ignore missing volumes on offline nodes
851 for volume in node_vol_should[node]:
852 if node not in node_vol_is or volume not in node_vol_is[node]:
853 feedback_fn(" - ERROR: volume %s missing on node %s" %
857 if instanceconfig.admin_up:
858 if ((node_current not in node_instance or
859 not instance in node_instance[node_current]) and
860 node_current not in n_offline):
861 feedback_fn(" - ERROR: instance %s not running on node %s" %
862 (instance, node_current))
865 for node in node_instance:
866 if (not node == node_current):
867 if instance in node_instance[node]:
868 feedback_fn(" - ERROR: instance %s should not run on node %s" %
874 def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
875 """Verify if there are any unknown volumes in the cluster.
877 The .os, .swap and backup volumes are ignored. All other volumes are
883 for node in node_vol_is:
884 for volume in node_vol_is[node]:
885 if node not in node_vol_should or volume not in node_vol_should[node]:
886 feedback_fn(" - ERROR: volume %s on node %s should not exist" %
891 def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
892 """Verify the list of running instances.
894 This checks what instances are running but unknown to the cluster.
898 for node in node_instance:
899 for runninginstance in node_instance[node]:
900 if runninginstance not in instancelist:
901 feedback_fn(" - ERROR: instance %s on node %s should not exist" %
902 (runninginstance, node))
906 def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
907 """Verify N+1 Memory Resilience.
909 Check that if one single node dies we can still start all the instances it
915 for node, nodeinfo in node_info.iteritems():
916 # This code checks that every node which is now listed as secondary has
917 # enough memory to host all instances it is supposed to should a single
918 # other node in the cluster fail.
919 # FIXME: not ready for failover to an arbitrary node
920 # FIXME: does not support file-backed instances
921 # WARNING: we currently take into account down instances as well as up
922 # ones, considering that even if they're down someone might want to start
923 # them even in the event of a node failure.
924 for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
926 for instance in instances:
927 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
928 if bep[constants.BE_AUTO_BALANCE]:
929 needed_mem += bep[constants.BE_MEMORY]
930 if nodeinfo['mfree'] < needed_mem:
931 feedback_fn(" - ERROR: not enough memory on node %s to accomodate"
932 " failovers should node %s fail" % (node, prinode))
936 def CheckPrereq(self):
937 """Check prerequisites.
939 Transform the list of checks we're going to skip into a set and check that
940 all its members are valid.
943 self.skip_set = frozenset(self.op.skip_checks)
944 if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
945 raise errors.OpPrereqError("Invalid checks to be skipped specified")
947 def BuildHooksEnv(self):
950 Cluster-Verify hooks just rone in the post phase and their failure makes
951 the output be logged in the verify output and the verification to fail.
954 all_nodes = self.cfg.GetNodeList()
956 "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
958 for node in self.cfg.GetAllNodesInfo().values():
959 env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
961 return env, [], all_nodes
963 def Exec(self, feedback_fn):
964 """Verify integrity of cluster, performing various test on nodes.
968 feedback_fn("* Verifying global settings")
969 for msg in self.cfg.VerifyConfig():
970 feedback_fn(" - ERROR: %s" % msg)
972 vg_name = self.cfg.GetVGName()
973 hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
974 nodelist = utils.NiceSort(self.cfg.GetNodeList())
975 nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
976 instancelist = utils.NiceSort(self.cfg.GetInstanceList())
977 instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
978 for iname in instancelist)
979 i_non_redundant = [] # Non redundant instances
980 i_non_a_balanced = [] # Non auto-balanced instances
981 n_offline = [] # List of offline nodes
982 n_drained = [] # List of nodes being drained
988 # FIXME: verify OS list
990 master_files = [constants.CLUSTER_CONF_FILE]
992 file_names = ssconf.SimpleStore().GetFileList()
993 file_names.append(constants.SSL_CERT_FILE)
994 file_names.append(constants.RAPI_CERT_FILE)
995 file_names.extend(master_files)
997 local_checksums = utils.FingerprintFiles(file_names)
999 feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1000 node_verify_param = {
1001 constants.NV_FILELIST: file_names,
1002 constants.NV_NODELIST: [node.name for node in nodeinfo
1003 if not node.offline],
1004 constants.NV_HYPERVISOR: hypervisors,
1005 constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1006 node.secondary_ip) for node in nodeinfo
1007 if not node.offline],
1008 constants.NV_INSTANCELIST: hypervisors,
1009 constants.NV_VERSION: None,
1010 constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1012 if vg_name is not None:
1013 node_verify_param[constants.NV_VGLIST] = None
1014 node_verify_param[constants.NV_LVLIST] = vg_name
1015 node_verify_param[constants.NV_DRBDLIST] = None
1016 all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1017 self.cfg.GetClusterName())
1019 cluster = self.cfg.GetClusterInfo()
1020 master_node = self.cfg.GetMasterNode()
1021 all_drbd_map = self.cfg.ComputeDRBDMap()
1023 for node_i in nodeinfo:
1025 nresult = all_nvinfo[node].data
1028 feedback_fn("* Skipping offline node %s" % (node,))
1029 n_offline.append(node)
1032 if node == master_node:
1034 elif node_i.master_candidate:
1035 ntype = "master candidate"
1036 elif node_i.drained:
1038 n_drained.append(node)
1041 feedback_fn("* Verifying node %s (%s)" % (node, ntype))
1043 if all_nvinfo[node].failed or not isinstance(nresult, dict):
1044 feedback_fn(" - ERROR: connection to %s failed" % (node,))
1049 for minor, instance in all_drbd_map[node].items():
1050 if instance not in instanceinfo:
1051 feedback_fn(" - ERROR: ghost instance '%s' in temporary DRBD map" %
1053 # ghost instance should not be running, but otherwise we
1054 # don't give double warnings (both ghost instance and
1055 # unallocated minor in use)
1056 node_drbd[minor] = (instance, False)
1058 instance = instanceinfo[instance]
1059 node_drbd[minor] = (instance.name, instance.admin_up)
1060 result = self._VerifyNode(node_i, file_names, local_checksums,
1061 nresult, feedback_fn, master_files,
1065 lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1067 node_volume[node] = {}
1068 elif isinstance(lvdata, basestring):
1069 feedback_fn(" - ERROR: LVM problem on node %s: %s" %
1070 (node, utils.SafeEncode(lvdata)))
1072 node_volume[node] = {}
1073 elif not isinstance(lvdata, dict):
1074 feedback_fn(" - ERROR: connection to %s failed (lvlist)" % (node,))
1078 node_volume[node] = lvdata
1081 idata = nresult.get(constants.NV_INSTANCELIST, None)
1082 if not isinstance(idata, list):
1083 feedback_fn(" - ERROR: connection to %s failed (instancelist)" %
1088 node_instance[node] = idata
1091 nodeinfo = nresult.get(constants.NV_HVINFO, None)
1092 if not isinstance(nodeinfo, dict):
1093 feedback_fn(" - ERROR: connection to %s failed (hvinfo)" % (node,))
1099 "mfree": int(nodeinfo['memory_free']),
1102 # dictionary holding all instances this node is secondary for,
1103 # grouped by their primary node. Each key is a cluster node, and each
1104 # value is a list of instances which have the key as primary and the
1105 # current node as secondary. this is handy to calculate N+1 memory
1106 # availability if you can only failover from a primary to its
1108 "sinst-by-pnode": {},
1110 # FIXME: devise a free space model for file based instances as well
1111 if vg_name is not None:
1112 if (constants.NV_VGLIST not in nresult or
1113 vg_name not in nresult[constants.NV_VGLIST]):
1114 feedback_fn(" - ERROR: node %s didn't return data for the"
1115 " volume group '%s' - it is either missing or broken" %
1119 node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
1120 except (ValueError, KeyError):
1121 feedback_fn(" - ERROR: invalid nodeinfo value returned"
1122 " from node %s" % (node,))
1126 node_vol_should = {}
1128 for instance in instancelist:
1129 feedback_fn("* Verifying instance %s" % instance)
1130 inst_config = instanceinfo[instance]
1131 result = self._VerifyInstance(instance, inst_config, node_volume,
1132 node_instance, feedback_fn, n_offline)
1134 inst_nodes_offline = []
1136 inst_config.MapLVsByNode(node_vol_should)
1138 instance_cfg[instance] = inst_config
1140 pnode = inst_config.primary_node
1141 if pnode in node_info:
1142 node_info[pnode]['pinst'].append(instance)
1143 elif pnode not in n_offline:
1144 feedback_fn(" - ERROR: instance %s, connection to primary node"
1145 " %s failed" % (instance, pnode))
1148 if pnode in n_offline:
1149 inst_nodes_offline.append(pnode)
1151 # If the instance is non-redundant we cannot survive losing its primary
1152 # node, so we are not N+1 compliant. On the other hand we have no disk
1153 # templates with more than one secondary so that situation is not well
1155 # FIXME: does not support file-backed instances
1156 if len(inst_config.secondary_nodes) == 0:
1157 i_non_redundant.append(instance)
1158 elif len(inst_config.secondary_nodes) > 1:
1159 feedback_fn(" - WARNING: multiple secondaries for instance %s"
1162 if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1163 i_non_a_balanced.append(instance)
1165 for snode in inst_config.secondary_nodes:
1166 if snode in node_info:
1167 node_info[snode]['sinst'].append(instance)
1168 if pnode not in node_info[snode]['sinst-by-pnode']:
1169 node_info[snode]['sinst-by-pnode'][pnode] = []
1170 node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1171 elif snode not in n_offline:
1172 feedback_fn(" - ERROR: instance %s, connection to secondary node"
1173 " %s failed" % (instance, snode))
1175 if snode in n_offline:
1176 inst_nodes_offline.append(snode)
1178 if inst_nodes_offline:
1179 # warn that the instance lives on offline nodes, and set bad=True
1180 feedback_fn(" - ERROR: instance lives on offline node(s) %s" %
1181 ", ".join(inst_nodes_offline))
1184 feedback_fn("* Verifying orphan volumes")
1185 result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
1189 feedback_fn("* Verifying remaining instances")
1190 result = self._VerifyOrphanInstances(instancelist, node_instance,
1194 if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1195 feedback_fn("* Verifying N+1 Memory redundancy")
1196 result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
1199 feedback_fn("* Other Notes")
1201 feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
1202 % len(i_non_redundant))
1204 if i_non_a_balanced:
1205 feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
1206 % len(i_non_a_balanced))
1209 feedback_fn(" - NOTICE: %d offline node(s) found." % len(n_offline))
1212 feedback_fn(" - NOTICE: %d drained node(s) found." % len(n_drained))
1216 def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1217 """Analize the post-hooks' result
1219 This method analyses the hook result, handles it, and sends some
1220 nicely-formatted feedback back to the user.
1222 @param phase: one of L{constants.HOOKS_PHASE_POST} or
1223 L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1224 @param hooks_results: the results of the multi-node hooks rpc call
1225 @param feedback_fn: function used send feedback back to the caller
1226 @param lu_result: previous Exec result
1227 @return: the new Exec result, based on the previous result
1231 # We only really run POST phase hooks, and are only interested in
1233 if phase == constants.HOOKS_PHASE_POST:
1234 # Used to change hooks' output to proper indentation
1235 indent_re = re.compile('^', re.M)
1236 feedback_fn("* Hooks Results")
1237 if not hooks_results:
1238 feedback_fn(" - ERROR: general communication failure")
1241 for node_name in hooks_results:
1242 show_node_header = True
1243 res = hooks_results[node_name]
1244 if res.failed or res.data is False or not isinstance(res.data, list):
1246 # no need to warn or set fail return value
1248 feedback_fn(" Communication failure in hooks execution")
1251 for script, hkr, output in res.data:
1252 if hkr == constants.HKR_FAIL:
1253 # The node header is only shown once, if there are
1254 # failing hooks on that node
1255 if show_node_header:
1256 feedback_fn(" Node %s:" % node_name)
1257 show_node_header = False
1258 feedback_fn(" ERROR: Script %s failed, output:" % script)
1259 output = indent_re.sub(' ', output)
1260 feedback_fn("%s" % output)
1266 class LUVerifyDisks(NoHooksLU):
1267 """Verifies the cluster disks status.
1273 def ExpandNames(self):
1274 self.needed_locks = {
1275 locking.LEVEL_NODE: locking.ALL_SET,
1276 locking.LEVEL_INSTANCE: locking.ALL_SET,
1278 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1280 def CheckPrereq(self):
1281 """Check prerequisites.
1283 This has no prerequisites.
1288 def Exec(self, feedback_fn):
1289 """Verify integrity of cluster disks.
1291 @rtype: tuple of three items
1292 @return: a tuple of (dict of node-to-node_error, list of instances
1293 which need activate-disks, dict of instance: (node, volume) for
1297 result = res_nodes, res_instances, res_missing = {}, [], {}
1299 vg_name = self.cfg.GetVGName()
1300 nodes = utils.NiceSort(self.cfg.GetNodeList())
1301 instances = [self.cfg.GetInstanceInfo(name)
1302 for name in self.cfg.GetInstanceList()]
1305 for inst in instances:
1307 if (not inst.admin_up or
1308 inst.disk_template not in constants.DTS_NET_MIRROR):
1310 inst.MapLVsByNode(inst_lvs)
1311 # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1312 for node, vol_list in inst_lvs.iteritems():
1313 for vol in vol_list:
1314 nv_dict[(node, vol)] = inst
1319 node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1324 node_res = node_lvs[node]
1325 if node_res.offline:
1327 msg = node_res.RemoteFailMsg()
1329 logging.warning("Error enumerating LVs on node %s: %s", node, msg)
1330 res_nodes[node] = msg
1333 lvs = node_res.payload
1334 for lv_name, (_, lv_inactive, lv_online) in lvs.items():
1335 inst = nv_dict.pop((node, lv_name), None)
1336 if (not lv_online and inst is not None
1337 and inst.name not in res_instances):
1338 res_instances.append(inst.name)
1340 # any leftover items in nv_dict are missing LVs, let's arrange the
1342 for key, inst in nv_dict.iteritems():
1343 if inst.name not in res_missing:
1344 res_missing[inst.name] = []
1345 res_missing[inst.name].append(key)
1350 class LURenameCluster(LogicalUnit):
1351 """Rename the cluster.
1354 HPATH = "cluster-rename"
1355 HTYPE = constants.HTYPE_CLUSTER
1358 def BuildHooksEnv(self):
1363 "OP_TARGET": self.cfg.GetClusterName(),
1364 "NEW_NAME": self.op.name,
1366 mn = self.cfg.GetMasterNode()
1367 return env, [mn], [mn]
1369 def CheckPrereq(self):
1370 """Verify that the passed name is a valid one.
1373 hostname = utils.HostInfo(self.op.name)
1375 new_name = hostname.name
1376 self.ip = new_ip = hostname.ip
1377 old_name = self.cfg.GetClusterName()
1378 old_ip = self.cfg.GetMasterIP()
1379 if new_name == old_name and new_ip == old_ip:
1380 raise errors.OpPrereqError("Neither the name nor the IP address of the"
1381 " cluster has changed")
1382 if new_ip != old_ip:
1383 if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1384 raise errors.OpPrereqError("The given cluster IP address (%s) is"
1385 " reachable on the network. Aborting." %
1388 self.op.name = new_name
1390 def Exec(self, feedback_fn):
1391 """Rename the cluster.
1394 clustername = self.op.name
1397 # shutdown the master IP
1398 master = self.cfg.GetMasterNode()
1399 result = self.rpc.call_node_stop_master(master, False)
1400 if result.failed or not result.data:
1401 raise errors.OpExecError("Could not disable the master role")
1404 cluster = self.cfg.GetClusterInfo()
1405 cluster.cluster_name = clustername
1406 cluster.master_ip = ip
1407 self.cfg.Update(cluster)
1409 # update the known hosts file
1410 ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1411 node_list = self.cfg.GetNodeList()
1413 node_list.remove(master)
1416 result = self.rpc.call_upload_file(node_list,
1417 constants.SSH_KNOWN_HOSTS_FILE)
1418 for to_node, to_result in result.iteritems():
1419 msg = to_result.RemoteFailMsg()
1421 msg = ("Copy of file %s to node %s failed: %s" %
1422 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
1423 self.proc.LogWarning(msg)
1426 result = self.rpc.call_node_start_master(master, False)
1427 if result.failed or not result.data:
1428 self.LogWarning("Could not re-enable the master role on"
1429 " the master, please restart manually.")
1432 def _RecursiveCheckIfLVMBased(disk):
1433 """Check if the given disk or its children are lvm-based.
1435 @type disk: L{objects.Disk}
1436 @param disk: the disk to check
1438 @return: boolean indicating whether a LD_LV dev_type was found or not
1442 for chdisk in disk.children:
1443 if _RecursiveCheckIfLVMBased(chdisk):
1445 return disk.dev_type == constants.LD_LV
1448 class LUSetClusterParams(LogicalUnit):
1449 """Change the parameters of the cluster.
1452 HPATH = "cluster-modify"
1453 HTYPE = constants.HTYPE_CLUSTER
1457 def CheckArguments(self):
1461 if not hasattr(self.op, "candidate_pool_size"):
1462 self.op.candidate_pool_size = None
1463 if self.op.candidate_pool_size is not None:
1465 self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1466 except (ValueError, TypeError), err:
1467 raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1469 if self.op.candidate_pool_size < 1:
1470 raise errors.OpPrereqError("At least one master candidate needed")
1472 def ExpandNames(self):
1473 # FIXME: in the future maybe other cluster params won't require checking on
1474 # all nodes to be modified.
1475 self.needed_locks = {
1476 locking.LEVEL_NODE: locking.ALL_SET,
1478 self.share_locks[locking.LEVEL_NODE] = 1
1480 def BuildHooksEnv(self):
1485 "OP_TARGET": self.cfg.GetClusterName(),
1486 "NEW_VG_NAME": self.op.vg_name,
1488 mn = self.cfg.GetMasterNode()
1489 return env, [mn], [mn]
1491 def CheckPrereq(self):
1492 """Check prerequisites.
1494 This checks whether the given params don't conflict and
1495 if the given volume group is valid.
1498 if self.op.vg_name is not None and not self.op.vg_name:
1499 instances = self.cfg.GetAllInstancesInfo().values()
1500 for inst in instances:
1501 for disk in inst.disks:
1502 if _RecursiveCheckIfLVMBased(disk):
1503 raise errors.OpPrereqError("Cannot disable lvm storage while"
1504 " lvm-based instances exist")
1506 node_list = self.acquired_locks[locking.LEVEL_NODE]
1508 # if vg_name not None, checks given volume group on all nodes
1510 vglist = self.rpc.call_vg_list(node_list)
1511 for node in node_list:
1512 msg = vglist[node].RemoteFailMsg()
1514 # ignoring down node
1515 self.LogWarning("Error while gathering data on node %s"
1516 " (ignoring node): %s", node, msg)
1518 vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
1520 constants.MIN_VG_SIZE)
1522 raise errors.OpPrereqError("Error on node '%s': %s" %
1525 self.cluster = cluster = self.cfg.GetClusterInfo()
1526 # validate params changes
1527 if self.op.beparams:
1528 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
1529 self.new_beparams = objects.FillDict(
1530 cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
1532 if self.op.nicparams:
1533 utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
1534 self.new_nicparams = objects.FillDict(
1535 cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
1536 objects.NIC.CheckParameterSyntax(self.new_nicparams)
1538 # hypervisor list/parameters
1539 self.new_hvparams = objects.FillDict(cluster.hvparams, {})
1540 if self.op.hvparams:
1541 if not isinstance(self.op.hvparams, dict):
1542 raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1543 for hv_name, hv_dict in self.op.hvparams.items():
1544 if hv_name not in self.new_hvparams:
1545 self.new_hvparams[hv_name] = hv_dict
1547 self.new_hvparams[hv_name].update(hv_dict)
1549 if self.op.enabled_hypervisors is not None:
1550 self.hv_list = self.op.enabled_hypervisors
1552 self.hv_list = cluster.enabled_hypervisors
1554 if self.op.hvparams or self.op.enabled_hypervisors is not None:
1555 # either the enabled list has changed, or the parameters have, validate
1556 for hv_name, hv_params in self.new_hvparams.items():
1557 if ((self.op.hvparams and hv_name in self.op.hvparams) or
1558 (self.op.enabled_hypervisors and
1559 hv_name in self.op.enabled_hypervisors)):
1560 # either this is a new hypervisor, or its parameters have changed
1561 hv_class = hypervisor.GetHypervisor(hv_name)
1562 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
1563 hv_class.CheckParameterSyntax(hv_params)
1564 _CheckHVParams(self, node_list, hv_name, hv_params)
1566 def Exec(self, feedback_fn):
1567 """Change the parameters of the cluster.
1570 if self.op.vg_name is not None:
1571 new_volume = self.op.vg_name
1574 if new_volume != self.cfg.GetVGName():
1575 self.cfg.SetVGName(new_volume)
1577 feedback_fn("Cluster LVM configuration already in desired"
1578 " state, not changing")
1579 if self.op.hvparams:
1580 self.cluster.hvparams = self.new_hvparams
1581 if self.op.enabled_hypervisors is not None:
1582 self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1583 if self.op.beparams:
1584 self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
1585 if self.op.nicparams:
1586 self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
1588 if self.op.candidate_pool_size is not None:
1589 self.cluster.candidate_pool_size = self.op.candidate_pool_size
1591 self.cfg.Update(self.cluster)
1593 # we want to update nodes after the cluster so that if any errors
1594 # happen, we have recorded and saved the cluster info
1595 if self.op.candidate_pool_size is not None:
1596 _AdjustCandidatePool(self)
1599 def _RedistributeAncillaryFiles(lu, additional_nodes=None):
1600 """Distribute additional files which are part of the cluster configuration.
1602 ConfigWriter takes care of distributing the config and ssconf files, but
1603 there are more files which should be distributed to all nodes. This function
1604 makes sure those are copied.
1606 @param lu: calling logical unit
1607 @param additional_nodes: list of nodes not in the config to distribute to
1610 # 1. Gather target nodes
1611 myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
1612 dist_nodes = lu.cfg.GetNodeList()
1613 if additional_nodes is not None:
1614 dist_nodes.extend(additional_nodes)
1615 if myself.name in dist_nodes:
1616 dist_nodes.remove(myself.name)
1617 # 2. Gather files to distribute
1618 dist_files = set([constants.ETC_HOSTS,
1619 constants.SSH_KNOWN_HOSTS_FILE,
1620 constants.RAPI_CERT_FILE,
1621 constants.RAPI_USERS_FILE,
1624 enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
1625 for hv_name in enabled_hypervisors:
1626 hv_class = hypervisor.GetHypervisor(hv_name)
1627 dist_files.update(hv_class.GetAncillaryFiles())
1629 # 3. Perform the files upload
1630 for fname in dist_files:
1631 if os.path.exists(fname):
1632 result = lu.rpc.call_upload_file(dist_nodes, fname)
1633 for to_node, to_result in result.items():
1634 msg = to_result.RemoteFailMsg()
1636 msg = ("Copy of file %s to node %s failed: %s" %
1637 (fname, to_node, msg))
1638 lu.proc.LogWarning(msg)
1641 class LURedistributeConfig(NoHooksLU):
1642 """Force the redistribution of cluster configuration.
1644 This is a very simple LU.
1650 def ExpandNames(self):
1651 self.needed_locks = {
1652 locking.LEVEL_NODE: locking.ALL_SET,
1654 self.share_locks[locking.LEVEL_NODE] = 1
1656 def CheckPrereq(self):
1657 """Check prerequisites.
1661 def Exec(self, feedback_fn):
1662 """Redistribute the configuration.
1665 self.cfg.Update(self.cfg.GetClusterInfo())
1666 _RedistributeAncillaryFiles(self)
1669 def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1670 """Sleep and poll for an instance's disk to sync.
1673 if not instance.disks:
1677 lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1679 node = instance.primary_node
1681 for dev in instance.disks:
1682 lu.cfg.SetDiskID(dev, node)
1688 cumul_degraded = False
1689 rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1690 msg = rstats.RemoteFailMsg()
1692 lu.LogWarning("Can't get any data from node %s: %s", node, msg)
1695 raise errors.RemoteError("Can't contact node %s for mirror data,"
1696 " aborting." % node)
1699 rstats = rstats.payload
1701 for i, mstat in enumerate(rstats):
1703 lu.LogWarning("Can't compute data for node %s/%s",
1704 node, instance.disks[i].iv_name)
1706 # we ignore the ldisk parameter
1707 perc_done, est_time, is_degraded, _ = mstat
1708 cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1709 if perc_done is not None:
1711 if est_time is not None:
1712 rem_time = "%d estimated seconds remaining" % est_time
1715 rem_time = "no time estimate"
1716 lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1717 (instance.disks[i].iv_name, perc_done, rem_time))
1721 time.sleep(min(60, max_time))
1724 lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1725 return not cumul_degraded
1728 def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1729 """Check that mirrors are not degraded.
1731 The ldisk parameter, if True, will change the test from the
1732 is_degraded attribute (which represents overall non-ok status for
1733 the device(s)) to the ldisk (representing the local storage status).
1736 lu.cfg.SetDiskID(dev, node)
1743 if on_primary or dev.AssembleOnSecondary():
1744 rstats = lu.rpc.call_blockdev_find(node, dev)
1745 msg = rstats.RemoteFailMsg()
1747 lu.LogWarning("Can't find disk on node %s: %s", node, msg)
1749 elif not rstats.payload:
1750 lu.LogWarning("Can't find disk on node %s", node)
1753 result = result and (not rstats.payload[idx])
1755 for child in dev.children:
1756 result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1761 class LUDiagnoseOS(NoHooksLU):
1762 """Logical unit for OS diagnose/query.
1765 _OP_REQP = ["output_fields", "names"]
1767 _FIELDS_STATIC = utils.FieldSet()
1768 _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1770 def ExpandNames(self):
1772 raise errors.OpPrereqError("Selective OS query not supported")
1774 _CheckOutputFields(static=self._FIELDS_STATIC,
1775 dynamic=self._FIELDS_DYNAMIC,
1776 selected=self.op.output_fields)
1778 # Lock all nodes, in shared mode
1779 # Temporary removal of locks, should be reverted later
1780 # TODO: reintroduce locks when they are lighter-weight
1781 self.needed_locks = {}
1782 #self.share_locks[locking.LEVEL_NODE] = 1
1783 #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1785 def CheckPrereq(self):
1786 """Check prerequisites.
1791 def _DiagnoseByOS(node_list, rlist):
1792 """Remaps a per-node return list into an a per-os per-node dictionary
1794 @param node_list: a list with the names of all nodes
1795 @param rlist: a map with node names as keys and OS objects as values
1798 @return: a dictionary with osnames as keys and as value another map, with
1799 nodes as keys and list of OS objects as values, eg::
1801 {"debian-etch": {"node1": [<object>,...],
1802 "node2": [<object>,]}
1807 # we build here the list of nodes that didn't fail the RPC (at RPC
1808 # level), so that nodes with a non-responding node daemon don't
1809 # make all OSes invalid
1810 good_nodes = [node_name for node_name in rlist
1811 if not rlist[node_name].failed]
1812 for node_name, nr in rlist.iteritems():
1813 if nr.failed or not nr.data:
1815 for os_obj in nr.data:
1816 if os_obj.name not in all_os:
1817 # build a list of nodes for this os containing empty lists
1818 # for each node in node_list
1819 all_os[os_obj.name] = {}
1820 for nname in good_nodes:
1821 all_os[os_obj.name][nname] = []
1822 all_os[os_obj.name][node_name].append(os_obj)
1825 def Exec(self, feedback_fn):
1826 """Compute the list of OSes.
1829 valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
1830 node_data = self.rpc.call_os_diagnose(valid_nodes)
1831 if node_data == False:
1832 raise errors.OpExecError("Can't gather the list of OSes")
1833 pol = self._DiagnoseByOS(valid_nodes, node_data)
1835 for os_name, os_data in pol.iteritems():
1837 for field in self.op.output_fields:
1840 elif field == "valid":
1841 val = utils.all([osl and osl[0] for osl in os_data.values()])
1842 elif field == "node_status":
1844 for node_name, nos_list in os_data.iteritems():
1845 val[node_name] = [(v.status, v.path) for v in nos_list]
1847 raise errors.ParameterError(field)
1854 class LURemoveNode(LogicalUnit):
1855 """Logical unit for removing a node.
1858 HPATH = "node-remove"
1859 HTYPE = constants.HTYPE_NODE
1860 _OP_REQP = ["node_name"]
1862 def BuildHooksEnv(self):
1865 This doesn't run on the target node in the pre phase as a failed
1866 node would then be impossible to remove.
1870 "OP_TARGET": self.op.node_name,
1871 "NODE_NAME": self.op.node_name,
1873 all_nodes = self.cfg.GetNodeList()
1874 all_nodes.remove(self.op.node_name)
1875 return env, all_nodes, all_nodes
1877 def CheckPrereq(self):
1878 """Check prerequisites.
1881 - the node exists in the configuration
1882 - it does not have primary or secondary instances
1883 - it's not the master
1885 Any errors are signalled by raising errors.OpPrereqError.
1888 node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1890 raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1892 instance_list = self.cfg.GetInstanceList()
1894 masternode = self.cfg.GetMasterNode()
1895 if node.name == masternode:
1896 raise errors.OpPrereqError("Node is the master node,"
1897 " you need to failover first.")
1899 for instance_name in instance_list:
1900 instance = self.cfg.GetInstanceInfo(instance_name)
1901 if node.name in instance.all_nodes:
1902 raise errors.OpPrereqError("Instance %s is still running on the node,"
1903 " please remove first." % instance_name)
1904 self.op.node_name = node.name
1907 def Exec(self, feedback_fn):
1908 """Removes the node from the cluster.
1912 logging.info("Stopping the node daemon and removing configs from node %s",
1915 self.context.RemoveNode(node.name)
1917 self.rpc.call_node_leave_cluster(node.name)
1919 # Promote nodes to master candidate as needed
1920 _AdjustCandidatePool(self)
1923 class LUQueryNodes(NoHooksLU):
1924 """Logical unit for querying nodes.
1927 _OP_REQP = ["output_fields", "names", "use_locking"]
1929 _FIELDS_DYNAMIC = utils.FieldSet(
1931 "mtotal", "mnode", "mfree",
1933 "ctotal", "cnodes", "csockets",
1936 _FIELDS_STATIC = utils.FieldSet(
1937 "name", "pinst_cnt", "sinst_cnt",
1938 "pinst_list", "sinst_list",
1939 "pip", "sip", "tags",
1947 def ExpandNames(self):
1948 _CheckOutputFields(static=self._FIELDS_STATIC,
1949 dynamic=self._FIELDS_DYNAMIC,
1950 selected=self.op.output_fields)
1952 self.needed_locks = {}
1953 self.share_locks[locking.LEVEL_NODE] = 1
1956 self.wanted = _GetWantedNodes(self, self.op.names)
1958 self.wanted = locking.ALL_SET
1960 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1961 self.do_locking = self.do_node_query and self.op.use_locking
1963 # if we don't request only static fields, we need to lock the nodes
1964 self.needed_locks[locking.LEVEL_NODE] = self.wanted
1967 def CheckPrereq(self):
1968 """Check prerequisites.
1971 # The validation of the node list is done in the _GetWantedNodes,
1972 # if non empty, and if empty, there's no validation to do
1975 def Exec(self, feedback_fn):
1976 """Computes the list of nodes and their attributes.
1979 all_info = self.cfg.GetAllNodesInfo()
1981 nodenames = self.acquired_locks[locking.LEVEL_NODE]
1982 elif self.wanted != locking.ALL_SET:
1983 nodenames = self.wanted
1984 missing = set(nodenames).difference(all_info.keys())
1986 raise errors.OpExecError(
1987 "Some nodes were removed before retrieving their data: %s" % missing)
1989 nodenames = all_info.keys()
1991 nodenames = utils.NiceSort(nodenames)
1992 nodelist = [all_info[name] for name in nodenames]
1994 # begin data gathering
1996 if self.do_node_query:
1998 node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1999 self.cfg.GetHypervisorType())
2000 for name in nodenames:
2001 nodeinfo = node_data[name]
2002 if not nodeinfo.failed and nodeinfo.data:
2003 nodeinfo = nodeinfo.data
2004 fn = utils.TryConvert
2006 "mtotal": fn(int, nodeinfo.get('memory_total', None)),
2007 "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
2008 "mfree": fn(int, nodeinfo.get('memory_free', None)),
2009 "dtotal": fn(int, nodeinfo.get('vg_size', None)),
2010 "dfree": fn(int, nodeinfo.get('vg_free', None)),
2011 "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
2012 "bootid": nodeinfo.get('bootid', None),
2013 "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
2014 "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
2017 live_data[name] = {}
2019 live_data = dict.fromkeys(nodenames, {})
2021 node_to_primary = dict([(name, set()) for name in nodenames])
2022 node_to_secondary = dict([(name, set()) for name in nodenames])
2024 inst_fields = frozenset(("pinst_cnt", "pinst_list",
2025 "sinst_cnt", "sinst_list"))
2026 if inst_fields & frozenset(self.op.output_fields):
2027 instancelist = self.cfg.GetInstanceList()
2029 for instance_name in instancelist:
2030 inst = self.cfg.GetInstanceInfo(instance_name)
2031 if inst.primary_node in node_to_primary:
2032 node_to_primary[inst.primary_node].add(inst.name)
2033 for secnode in inst.secondary_nodes:
2034 if secnode in node_to_secondary:
2035 node_to_secondary[secnode].add(inst.name)
2037 master_node = self.cfg.GetMasterNode()
2039 # end data gathering
2042 for node in nodelist:
2044 for field in self.op.output_fields:
2047 elif field == "pinst_list":
2048 val = list(node_to_primary[node.name])
2049 elif field == "sinst_list":
2050 val = list(node_to_secondary[node.name])
2051 elif field == "pinst_cnt":
2052 val = len(node_to_primary[node.name])
2053 elif field == "sinst_cnt":
2054 val = len(node_to_secondary[node.name])
2055 elif field == "pip":
2056 val = node.primary_ip
2057 elif field == "sip":
2058 val = node.secondary_ip
2059 elif field == "tags":
2060 val = list(node.GetTags())
2061 elif field == "serial_no":
2062 val = node.serial_no
2063 elif field == "master_candidate":
2064 val = node.master_candidate
2065 elif field == "master":
2066 val = node.name == master_node
2067 elif field == "offline":
2069 elif field == "drained":
2071 elif self._FIELDS_DYNAMIC.Matches(field):
2072 val = live_data[node.name].get(field, None)
2074 raise errors.ParameterError(field)
2075 node_output.append(val)
2076 output.append(node_output)
2081 class LUQueryNodeVolumes(NoHooksLU):
2082 """Logical unit for getting volumes on node(s).
2085 _OP_REQP = ["nodes", "output_fields"]
2087 _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
2088 _FIELDS_STATIC = utils.FieldSet("node")
2090 def ExpandNames(self):
2091 _CheckOutputFields(static=self._FIELDS_STATIC,
2092 dynamic=self._FIELDS_DYNAMIC,
2093 selected=self.op.output_fields)
2095 self.needed_locks = {}
2096 self.share_locks[locking.LEVEL_NODE] = 1
2097 if not self.op.nodes:
2098 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2100 self.needed_locks[locking.LEVEL_NODE] = \
2101 _GetWantedNodes(self, self.op.nodes)
2103 def CheckPrereq(self):
2104 """Check prerequisites.
2106 This checks that the fields required are valid output fields.
2109 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2111 def Exec(self, feedback_fn):
2112 """Computes the list of nodes and their attributes.
2115 nodenames = self.nodes
2116 volumes = self.rpc.call_node_volumes(nodenames)
2118 ilist = [self.cfg.GetInstanceInfo(iname) for iname
2119 in self.cfg.GetInstanceList()]
2121 lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
2124 for node in nodenames:
2125 if node not in volumes or volumes[node].failed or not volumes[node].data:
2128 node_vols = volumes[node].data[:]
2129 node_vols.sort(key=lambda vol: vol['dev'])
2131 for vol in node_vols:
2133 for field in self.op.output_fields:
2136 elif field == "phys":
2140 elif field == "name":
2142 elif field == "size":
2143 val = int(float(vol['size']))
2144 elif field == "instance":
2146 if node not in lv_by_node[inst]:
2148 if vol['name'] in lv_by_node[inst][node]:
2154 raise errors.ParameterError(field)
2155 node_output.append(str(val))
2157 output.append(node_output)
2162 class LUAddNode(LogicalUnit):
2163 """Logical unit for adding node to the cluster.
2167 HTYPE = constants.HTYPE_NODE
2168 _OP_REQP = ["node_name"]
2170 def BuildHooksEnv(self):
2173 This will run on all nodes before, and on all nodes + the new node after.
2177 "OP_TARGET": self.op.node_name,
2178 "NODE_NAME": self.op.node_name,
2179 "NODE_PIP": self.op.primary_ip,
2180 "NODE_SIP": self.op.secondary_ip,
2182 nodes_0 = self.cfg.GetNodeList()
2183 nodes_1 = nodes_0 + [self.op.node_name, ]
2184 return env, nodes_0, nodes_1
2186 def CheckPrereq(self):
2187 """Check prerequisites.
2190 - the new node is not already in the config
2192 - its parameters (single/dual homed) matches the cluster
2194 Any errors are signalled by raising errors.OpPrereqError.
2197 node_name = self.op.node_name
2200 dns_data = utils.HostInfo(node_name)
2202 node = dns_data.name
2203 primary_ip = self.op.primary_ip = dns_data.ip
2204 secondary_ip = getattr(self.op, "secondary_ip", None)
2205 if secondary_ip is None:
2206 secondary_ip = primary_ip
2207 if not utils.IsValidIP(secondary_ip):
2208 raise errors.OpPrereqError("Invalid secondary IP given")
2209 self.op.secondary_ip = secondary_ip
2211 node_list = cfg.GetNodeList()
2212 if not self.op.readd and node in node_list:
2213 raise errors.OpPrereqError("Node %s is already in the configuration" %
2215 elif self.op.readd and node not in node_list:
2216 raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2218 for existing_node_name in node_list:
2219 existing_node = cfg.GetNodeInfo(existing_node_name)
2221 if self.op.readd and node == existing_node_name:
2222 if (existing_node.primary_ip != primary_ip or
2223 existing_node.secondary_ip != secondary_ip):
2224 raise errors.OpPrereqError("Readded node doesn't have the same IP"
2225 " address configuration as before")
2228 if (existing_node.primary_ip == primary_ip or
2229 existing_node.secondary_ip == primary_ip or
2230 existing_node.primary_ip == secondary_ip or
2231 existing_node.secondary_ip == secondary_ip):
2232 raise errors.OpPrereqError("New node ip address(es) conflict with"
2233 " existing node %s" % existing_node.name)
2235 # check that the type of the node (single versus dual homed) is the
2236 # same as for the master
2237 myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2238 master_singlehomed = myself.secondary_ip == myself.primary_ip
2239 newbie_singlehomed = secondary_ip == primary_ip
2240 if master_singlehomed != newbie_singlehomed:
2241 if master_singlehomed:
2242 raise errors.OpPrereqError("The master has no private ip but the"
2243 " new node has one")
2245 raise errors.OpPrereqError("The master has a private ip but the"
2246 " new node doesn't have one")
2248 # checks reachablity
2249 if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2250 raise errors.OpPrereqError("Node not reachable by ping")
2252 if not newbie_singlehomed:
2253 # check reachability from my secondary ip to newbie's secondary ip
2254 if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2255 source=myself.secondary_ip):
2256 raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2257 " based ping to noded port")
2259 cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2260 mc_now, _ = self.cfg.GetMasterCandidateStats()
2261 master_candidate = mc_now < cp_size
2263 self.new_node = objects.Node(name=node,
2264 primary_ip=primary_ip,
2265 secondary_ip=secondary_ip,
2266 master_candidate=master_candidate,
2267 offline=False, drained=False)
2269 def Exec(self, feedback_fn):
2270 """Adds the new node to the cluster.
2273 new_node = self.new_node
2274 node = new_node.name
2276 # check connectivity
2277 result = self.rpc.call_version([node])[node]
2280 if constants.PROTOCOL_VERSION == result.data:
2281 logging.info("Communication to node %s fine, sw version %s match",
2284 raise errors.OpExecError("Version mismatch master version %s,"
2285 " node version %s" %
2286 (constants.PROTOCOL_VERSION, result.data))
2288 raise errors.OpExecError("Cannot get version from the new node")
2291 logging.info("Copy ssh key to node %s", node)
2292 priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2294 keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2295 constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2301 keyarray.append(f.read())
2305 result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2307 keyarray[3], keyarray[4], keyarray[5])
2309 msg = result.RemoteFailMsg()
2311 raise errors.OpExecError("Cannot transfer ssh keys to the"
2312 " new node: %s" % msg)
2314 # Add node to our /etc/hosts, and add key to known_hosts
2315 if self.cfg.GetClusterInfo().modify_etc_hosts:
2316 utils.AddHostToEtcHosts(new_node.name)
2318 if new_node.secondary_ip != new_node.primary_ip:
2319 result = self.rpc.call_node_has_ip_address(new_node.name,
2320 new_node.secondary_ip)
2321 if result.failed or not result.data:
2322 raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2323 " you gave (%s). Please fix and re-run this"
2324 " command." % new_node.secondary_ip)
2326 node_verify_list = [self.cfg.GetMasterNode()]
2327 node_verify_param = {
2329 # TODO: do a node-net-test as well?
2332 result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2333 self.cfg.GetClusterName())
2334 for verifier in node_verify_list:
2335 if result[verifier].failed or not result[verifier].data:
2336 raise errors.OpExecError("Cannot communicate with %s's node daemon"
2337 " for remote verification" % verifier)
2338 if result[verifier].data['nodelist']:
2339 for failed in result[verifier].data['nodelist']:
2340 feedback_fn("ssh/hostname verification failed %s -> %s" %
2341 (verifier, result[verifier].data['nodelist'][failed]))
2342 raise errors.OpExecError("ssh/hostname verification failed.")
2345 _RedistributeAncillaryFiles(self)
2346 self.context.ReaddNode(new_node)
2348 _RedistributeAncillaryFiles(self, additional_nodes=[node])
2349 self.context.AddNode(new_node)
2352 class LUSetNodeParams(LogicalUnit):
2353 """Modifies the parameters of a node.
2356 HPATH = "node-modify"
2357 HTYPE = constants.HTYPE_NODE
2358 _OP_REQP = ["node_name"]
2361 def CheckArguments(self):
2362 node_name = self.cfg.ExpandNodeName(self.op.node_name)
2363 if node_name is None:
2364 raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2365 self.op.node_name = node_name
2366 _CheckBooleanOpField(self.op, 'master_candidate')
2367 _CheckBooleanOpField(self.op, 'offline')
2368 _CheckBooleanOpField(self.op, 'drained')
2369 all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
2370 if all_mods.count(None) == 3:
2371 raise errors.OpPrereqError("Please pass at least one modification")
2372 if all_mods.count(True) > 1:
2373 raise errors.OpPrereqError("Can't set the node into more than one"
2374 " state at the same time")
2376 def ExpandNames(self):
2377 self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2379 def BuildHooksEnv(self):
2382 This runs on the master node.
2386 "OP_TARGET": self.op.node_name,
2387 "MASTER_CANDIDATE": str(self.op.master_candidate),
2388 "OFFLINE": str(self.op.offline),
2389 "DRAINED": str(self.op.drained),
2391 nl = [self.cfg.GetMasterNode(),
2395 def CheckPrereq(self):
2396 """Check prerequisites.
2398 This only checks the instance list against the existing names.
2401 node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2403 if ((self.op.master_candidate == False or self.op.offline == True or
2404 self.op.drained == True) and node.master_candidate):
2405 # we will demote the node from master_candidate
2406 if self.op.node_name == self.cfg.GetMasterNode():
2407 raise errors.OpPrereqError("The master node has to be a"
2408 " master candidate, online and not drained")
2409 cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2410 num_candidates, _ = self.cfg.GetMasterCandidateStats()
2411 if num_candidates <= cp_size:
2412 msg = ("Not enough master candidates (desired"
2413 " %d, new value will be %d)" % (cp_size, num_candidates-1))
2415 self.LogWarning(msg)
2417 raise errors.OpPrereqError(msg)
2419 if (self.op.master_candidate == True and
2420 ((node.offline and not self.op.offline == False) or
2421 (node.drained and not self.op.drained == False))):
2422 raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
2423 " to master_candidate" % node.name)
2427 def Exec(self, feedback_fn):
2436 if self.op.offline is not None:
2437 node.offline = self.op.offline
2438 result.append(("offline", str(self.op.offline)))
2439 if self.op.offline == True:
2440 if node.master_candidate:
2441 node.master_candidate = False
2443 result.append(("master_candidate", "auto-demotion due to offline"))
2445 node.drained = False
2446 result.append(("drained", "clear drained status due to offline"))
2448 if self.op.master_candidate is not None:
2449 node.master_candidate = self.op.master_candidate
2451 result.append(("master_candidate", str(self.op.master_candidate)))
2452 if self.op.master_candidate == False:
2453 rrc = self.rpc.call_node_demote_from_mc(node.name)
2454 msg = rrc.RemoteFailMsg()
2456 self.LogWarning("Node failed to demote itself: %s" % msg)
2458 if self.op.drained is not None:
2459 node.drained = self.op.drained
2460 result.append(("drained", str(self.op.drained)))
2461 if self.op.drained == True:
2462 if node.master_candidate:
2463 node.master_candidate = False
2465 result.append(("master_candidate", "auto-demotion due to drain"))
2467 node.offline = False
2468 result.append(("offline", "clear offline status due to drain"))
2470 # this will trigger configuration file update, if needed
2471 self.cfg.Update(node)
2472 # this will trigger job queue propagation or cleanup
2474 self.context.ReaddNode(node)
2479 class LUPowercycleNode(NoHooksLU):
2480 """Powercycles a node.
2483 _OP_REQP = ["node_name", "force"]
2486 def CheckArguments(self):
2487 node_name = self.cfg.ExpandNodeName(self.op.node_name)
2488 if node_name is None:
2489 raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2490 self.op.node_name = node_name
2491 if node_name == self.cfg.GetMasterNode() and not self.op.force:
2492 raise errors.OpPrereqError("The node is the master and the force"
2493 " parameter was not set")
2495 def ExpandNames(self):
2496 """Locking for PowercycleNode.
2498 This is a last-resource option and shouldn't block on other
2499 jobs. Therefore, we grab no locks.
2502 self.needed_locks = {}
2504 def CheckPrereq(self):
2505 """Check prerequisites.
2507 This LU has no prereqs.
2512 def Exec(self, feedback_fn):
2516 result = self.rpc.call_node_powercycle(self.op.node_name,
2517 self.cfg.GetHypervisorType())
2518 msg = result.RemoteFailMsg()
2520 raise errors.OpExecError("Failed to schedule the reboot: %s" % msg)
2521 return result.payload
2524 class LUQueryClusterInfo(NoHooksLU):
2525 """Query cluster configuration.
2531 def ExpandNames(self):
2532 self.needed_locks = {}
2534 def CheckPrereq(self):
2535 """No prerequsites needed for this LU.
2540 def Exec(self, feedback_fn):
2541 """Return cluster config.
2544 cluster = self.cfg.GetClusterInfo()
2546 "software_version": constants.RELEASE_VERSION,
2547 "protocol_version": constants.PROTOCOL_VERSION,
2548 "config_version": constants.CONFIG_VERSION,
2549 "os_api_version": constants.OS_API_VERSION,
2550 "export_version": constants.EXPORT_VERSION,
2551 "architecture": (platform.architecture()[0], platform.machine()),
2552 "name": cluster.cluster_name,
2553 "master": cluster.master_node,
2554 "default_hypervisor": cluster.default_hypervisor,
2555 "enabled_hypervisors": cluster.enabled_hypervisors,
2556 "hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
2557 for hypervisor in cluster.enabled_hypervisors]),
2558 "beparams": cluster.beparams,
2559 "nicparams": cluster.nicparams,
2560 "candidate_pool_size": cluster.candidate_pool_size,
2561 "master_netdev": cluster.master_netdev,
2562 "volume_group_name": cluster.volume_group_name,
2563 "file_storage_dir": cluster.file_storage_dir,
2569 class LUQueryConfigValues(NoHooksLU):
2570 """Return configuration values.
2575 _FIELDS_DYNAMIC = utils.FieldSet()
2576 _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2578 def ExpandNames(self):
2579 self.needed_locks = {}
2581 _CheckOutputFields(static=self._FIELDS_STATIC,
2582 dynamic=self._FIELDS_DYNAMIC,
2583 selected=self.op.output_fields)
2585 def CheckPrereq(self):
2586 """No prerequisites.
2591 def Exec(self, feedback_fn):
2592 """Dump a representation of the cluster config to the standard output.
2596 for field in self.op.output_fields:
2597 if field == "cluster_name":
2598 entry = self.cfg.GetClusterName()
2599 elif field == "master_node":
2600 entry = self.cfg.GetMasterNode()
2601 elif field == "drain_flag":
2602 entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2604 raise errors.ParameterError(field)
2605 values.append(entry)
2609 class LUActivateInstanceDisks(NoHooksLU):
2610 """Bring up an instance's disks.
2613 _OP_REQP = ["instance_name"]
2616 def ExpandNames(self):
2617 self._ExpandAndLockInstance()
2618 self.needed_locks[locking.LEVEL_NODE] = []
2619 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2621 def DeclareLocks(self, level):
2622 if level == locking.LEVEL_NODE:
2623 self._LockInstancesNodes()
2625 def CheckPrereq(self):
2626 """Check prerequisites.
2628 This checks that the instance is in the cluster.
2631 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2632 assert self.instance is not None, \
2633 "Cannot retrieve locked instance %s" % self.op.instance_name
2634 _CheckNodeOnline(self, self.instance.primary_node)
2636 def Exec(self, feedback_fn):
2637 """Activate the disks.
2640 disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2642 raise errors.OpExecError("Cannot activate block devices")
2647 def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2648 """Prepare the block devices for an instance.
2650 This sets up the block devices on all nodes.
2652 @type lu: L{LogicalUnit}
2653 @param lu: the logical unit on whose behalf we execute
2654 @type instance: L{objects.Instance}
2655 @param instance: the instance for whose disks we assemble
2656 @type ignore_secondaries: boolean
2657 @param ignore_secondaries: if true, errors on secondary nodes
2658 won't result in an error return from the function
2659 @return: False if the operation failed, otherwise a list of
2660 (host, instance_visible_name, node_visible_name)
2661 with the mapping from node devices to instance devices
2666 iname = instance.name
2667 # With the two passes mechanism we try to reduce the window of
2668 # opportunity for the race condition of switching DRBD to primary
2669 # before handshaking occured, but we do not eliminate it
2671 # The proper fix would be to wait (with some limits) until the
2672 # connection has been made and drbd transitions from WFConnection
2673 # into any other network-connected state (Connected, SyncTarget,
2676 # 1st pass, assemble on all nodes in secondary mode
2677 for inst_disk in instance.disks:
2678 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2679 lu.cfg.SetDiskID(node_disk, node)
2680 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2681 msg = result.RemoteFailMsg()
2683 lu.proc.LogWarning("Could not prepare block device %s on node %s"
2684 " (is_primary=False, pass=1): %s",
2685 inst_disk.iv_name, node, msg)
2686 if not ignore_secondaries:
2689 # FIXME: race condition on drbd migration to primary
2691 # 2nd pass, do only the primary node
2692 for inst_disk in instance.disks:
2693 for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2694 if node != instance.primary_node:
2696 lu.cfg.SetDiskID(node_disk, node)
2697 result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2698 msg = result.RemoteFailMsg()
2700 lu.proc.LogWarning("Could not prepare block device %s on node %s"
2701 " (is_primary=True, pass=2): %s",
2702 inst_disk.iv_name, node, msg)
2704 device_info.append((instance.primary_node, inst_disk.iv_name,
2707 # leave the disks configured for the primary node
2708 # this is a workaround that would be fixed better by
2709 # improving the logical/physical id handling
2710 for disk in instance.disks:
2711 lu.cfg.SetDiskID(disk, instance.primary_node)
2713 return disks_ok, device_info
2716 def _StartInstanceDisks(lu, instance, force):
2717 """Start the disks of an instance.
2720 disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2721 ignore_secondaries=force)
2723 _ShutdownInstanceDisks(lu, instance)
2724 if force is not None and not force:
2725 lu.proc.LogWarning("", hint="If the message above refers to a"
2727 " you can retry the operation using '--force'.")
2728 raise errors.OpExecError("Disk consistency error")
2731 class LUDeactivateInstanceDisks(NoHooksLU):
2732 """Shutdown an instance's disks.
2735 _OP_REQP = ["instance_name"]
2738 def ExpandNames(self):
2739 self._ExpandAndLockInstance()
2740 self.needed_locks[locking.LEVEL_NODE] = []
2741 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2743 def DeclareLocks(self, level):
2744 if level == locking.LEVEL_NODE:
2745 self._LockInstancesNodes()
2747 def CheckPrereq(self):
2748 """Check prerequisites.
2750 This checks that the instance is in the cluster.
2753 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2754 assert self.instance is not None, \
2755 "Cannot retrieve locked instance %s" % self.op.instance_name
2757 def Exec(self, feedback_fn):
2758 """Deactivate the disks
2761 instance = self.instance
2762 _SafeShutdownInstanceDisks(self, instance)
2765 def _SafeShutdownInstanceDisks(lu, instance):
2766 """Shutdown block devices of an instance.
2768 This function checks if an instance is running, before calling
2769 _ShutdownInstanceDisks.
2772 ins_l = lu.rpc.call_instance_list([instance.primary_node],
2773 [instance.hypervisor])
2774 ins_l = ins_l[instance.primary_node]
2775 if ins_l.failed or not isinstance(ins_l.data, list):
2776 raise errors.OpExecError("Can't contact node '%s'" %
2777 instance.primary_node)
2779 if instance.name in ins_l.data:
2780 raise errors.OpExecError("Instance is running, can't shutdown"
2783 _ShutdownInstanceDisks(lu, instance)
2786 def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2787 """Shutdown block devices of an instance.
2789 This does the shutdown on all nodes of the instance.
2791 If the ignore_primary is false, errors on the primary node are
2796 for disk in instance.disks:
2797 for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2798 lu.cfg.SetDiskID(top_disk, node)
2799 result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2800 msg = result.RemoteFailMsg()
2802 lu.LogWarning("Could not shutdown block device %s on node %s: %s",
2803 disk.iv_name, node, msg)
2804 if not ignore_primary or node != instance.primary_node:
2809 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2810 """Checks if a node has enough free memory.
2812 This function check if a given node has the needed amount of free
2813 memory. In case the node has less memory or we cannot get the
2814 information from the node, this function raise an OpPrereqError
2817 @type lu: C{LogicalUnit}
2818 @param lu: a logical unit from which we get configuration data
2820 @param node: the node to check
2821 @type reason: C{str}
2822 @param reason: string to use in the error message
2823 @type requested: C{int}
2824 @param requested: the amount of memory in MiB to check for
2825 @type hypervisor_name: C{str}
2826 @param hypervisor_name: the hypervisor to ask for memory stats
2827 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2828 we cannot check the node
2831 nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2832 nodeinfo[node].Raise()
2833 free_mem = nodeinfo[node].data.get('memory_free')
2834 if not isinstance(free_mem, int):
2835 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2836 " was '%s'" % (node, free_mem))
2837 if requested > free_mem:
2838 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2839 " needed %s MiB, available %s MiB" %
2840 (node, reason, requested, free_mem))
2843 class LUStartupInstance(LogicalUnit):
2844 """Starts an instance.
2847 HPATH = "instance-start"
2848 HTYPE = constants.HTYPE_INSTANCE
2849 _OP_REQP = ["instance_name", "force"]
2852 def ExpandNames(self):
2853 self._ExpandAndLockInstance()
2855 def BuildHooksEnv(self):
2858 This runs on master, primary and secondary nodes of the instance.
2862 "FORCE": self.op.force,
2864 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2865 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2868 def CheckPrereq(self):
2869 """Check prerequisites.
2871 This checks that the instance is in the cluster.
2874 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2875 assert self.instance is not None, \
2876 "Cannot retrieve locked instance %s" % self.op.instance_name
2879 self.beparams = getattr(self.op, "beparams", {})
2881 if not isinstance(self.beparams, dict):
2882 raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
2883 " dict" % (type(self.beparams), ))
2884 # fill the beparams dict
2885 utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
2886 self.op.beparams = self.beparams
2889 self.hvparams = getattr(self.op, "hvparams", {})
2891 if not isinstance(self.hvparams, dict):
2892 raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
2893 " dict" % (type(self.hvparams), ))
2895 # check hypervisor parameter syntax (locally)
2896 cluster = self.cfg.GetClusterInfo()
2897 utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
2898 filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
2900 filled_hvp.update(self.hvparams)
2901 hv_type = hypervisor.GetHypervisor(instance.hypervisor)
2902 hv_type.CheckParameterSyntax(filled_hvp)
2903 _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
2904 self.op.hvparams = self.hvparams
2906 _CheckNodeOnline(self, instance.primary_node)
2908 bep = self.cfg.GetClusterInfo().FillBE(instance)
2909 # check bridges existance
2910 _CheckInstanceBridgesExist(self, instance)
2912 remote_info = self.rpc.call_instance_info(instance.primary_node,
2914 instance.hypervisor)
2916 if not remote_info.data:
2917 _CheckNodeFreeMemory(self, instance.primary_node,
2918 "starting instance %s" % instance.name,
2919 bep[constants.BE_MEMORY], instance.hypervisor)
2921 def Exec(self, feedback_fn):
2922 """Start the instance.
2925 instance = self.instance
2926 force = self.op.force
2928 self.cfg.MarkInstanceUp(instance.name)
2930 node_current = instance.primary_node
2932 _StartInstanceDisks(self, instance, force)
2934 result = self.rpc.call_instance_start(node_current, instance,
2935 self.hvparams, self.beparams)
2936 msg = result.RemoteFailMsg()
2938 _ShutdownInstanceDisks(self, instance)
2939 raise errors.OpExecError("Could not start instance: %s" % msg)
2942 class LURebootInstance(LogicalUnit):
2943 """Reboot an instance.
2946 HPATH = "instance-reboot"
2947 HTYPE = constants.HTYPE_INSTANCE
2948 _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2951 def ExpandNames(self):
2952 if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2953 constants.INSTANCE_REBOOT_HARD,
2954 constants.INSTANCE_REBOOT_FULL]:
2955 raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2956 (constants.INSTANCE_REBOOT_SOFT,
2957 constants.INSTANCE_REBOOT_HARD,
2958 constants.INSTANCE_REBOOT_FULL))
2959 self._ExpandAndLockInstance()
2961 def BuildHooksEnv(self):
2964 This runs on master, primary and secondary nodes of the instance.
2968 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2969 "REBOOT_TYPE": self.op.reboot_type,
2971 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2972 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2975 def CheckPrereq(self):
2976 """Check prerequisites.
2978 This checks that the instance is in the cluster.
2981 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2982 assert self.instance is not None, \
2983 "Cannot retrieve locked instance %s" % self.op.instance_name
2985 _CheckNodeOnline(self, instance.primary_node)
2987 # check bridges existance
2988 _CheckInstanceBridgesExist(self, instance)
2990 def Exec(self, feedback_fn):
2991 """Reboot the instance.
2994 instance = self.instance
2995 ignore_secondaries = self.op.ignore_secondaries
2996 reboot_type = self.op.reboot_type
2998 node_current = instance.primary_node
3000 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
3001 constants.INSTANCE_REBOOT_HARD]:
3002 for disk in instance.disks:
3003 self.cfg.SetDiskID(disk, node_current)
3004 result = self.rpc.call_instance_reboot(node_current, instance,
3006 msg = result.RemoteFailMsg()
3008 raise errors.OpExecError("Could not reboot instance: %s" % msg)
3010 result = self.rpc.call_instance_shutdown(node_current, instance)
3011 msg = result.RemoteFailMsg()
3013 raise errors.OpExecError("Could not shutdown instance for"
3014 " full reboot: %s" % msg)
3015 _ShutdownInstanceDisks(self, instance)
3016 _StartInstanceDisks(self, instance, ignore_secondaries)
3017 result = self.rpc.call_instance_start(node_current, instance, None, None)
3018 msg = result.RemoteFailMsg()
3020 _ShutdownInstanceDisks(self, instance)
3021 raise errors.OpExecError("Could not start instance for"
3022 " full reboot: %s" % msg)
3024 self.cfg.MarkInstanceUp(instance.name)
3027 class LUShutdownInstance(LogicalUnit):
3028 """Shutdown an instance.
3031 HPATH = "instance-stop"
3032 HTYPE = constants.HTYPE_INSTANCE
3033 _OP_REQP = ["instance_name"]
3036 def ExpandNames(self):
3037 self._ExpandAndLockInstance()
3039 def BuildHooksEnv(self):
3042 This runs on master, primary and secondary nodes of the instance.
3045 env = _BuildInstanceHookEnvByObject(self, self.instance)
3046 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3049 def CheckPrereq(self):
3050 """Check prerequisites.
3052 This checks that the instance is in the cluster.
3055 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3056 assert self.instance is not None, \
3057 "Cannot retrieve locked instance %s" % self.op.instance_name
3058 _CheckNodeOnline(self, self.instance.primary_node)
3060 def Exec(self, feedback_fn):
3061 """Shutdown the instance.
3064 instance = self.instance
3065 node_current = instance.primary_node
3066 self.cfg.MarkInstanceDown(instance.name)
3067 result = self.rpc.call_instance_shutdown(node_current, instance)
3068 msg = result.RemoteFailMsg()
3070 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
3072 _ShutdownInstanceDisks(self, instance)
3075 class LUReinstallInstance(LogicalUnit):
3076 """Reinstall an instance.
3079 HPATH = "instance-reinstall"
3080 HTYPE = constants.HTYPE_INSTANCE
3081 _OP_REQP = ["instance_name"]
3084 def ExpandNames(self):
3085 self._ExpandAndLockInstance()
3087 def BuildHooksEnv(self):
3090 This runs on master, primary and secondary nodes of the instance.
3093 env = _BuildInstanceHookEnvByObject(self, self.instance)
3094 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3097 def CheckPrereq(self):
3098 """Check prerequisites.
3100 This checks that the instance is in the cluster and is not running.
3103 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3104 assert instance is not None, \
3105 "Cannot retrieve locked instance %s" % self.op.instance_name
3106 _CheckNodeOnline(self, instance.primary_node)
3108 if instance.disk_template == constants.DT_DISKLESS:
3109 raise errors.OpPrereqError("Instance '%s' has no disks" %
3110 self.op.instance_name)
3111 if instance.admin_up:
3112 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3113 self.op.instance_name)
3114 remote_info = self.rpc.call_instance_info(instance.primary_node,
3116 instance.hypervisor)
3118 if remote_info.data:
3119 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3120 (self.op.instance_name,
3121 instance.primary_node))
3123 self.op.os_type = getattr(self.op, "os_type", None)
3124 if self.op.os_type is not None:
3126 pnode = self.cfg.GetNodeInfo(
3127 self.cfg.ExpandNodeName(instance.primary_node))
3129 raise errors.OpPrereqError("Primary node '%s' is unknown" %
3131 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3133 if not isinstance(result.data, objects.OS):
3134 raise errors.OpPrereqError("OS '%s' not in supported OS list for"
3135 " primary node" % self.op.os_type)
3137 self.instance = instance
3139 def Exec(self, feedback_fn):
3140 """Reinstall the instance.
3143 inst = self.instance
3145 if self.op.os_type is not None:
3146 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
3147 inst.os = self.op.os_type
3148 self.cfg.Update(inst)
3150 _StartInstanceDisks(self, inst, None)
3152 feedback_fn("Running the instance OS create scripts...")
3153 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
3154 msg = result.RemoteFailMsg()
3156 raise errors.OpExecError("Could not install OS for instance %s"
3158 (inst.name, inst.primary_node, msg))
3160 _ShutdownInstanceDisks(self, inst)
3163 class LURenameInstance(LogicalUnit):
3164 """Rename an instance.
3167 HPATH = "instance-rename"
3168 HTYPE = constants.HTYPE_INSTANCE
3169 _OP_REQP = ["instance_name", "new_name"]
3171 def BuildHooksEnv(self):
3174 This runs on master, primary and secondary nodes of the instance.
3177 env = _BuildInstanceHookEnvByObject(self, self.instance)
3178 env["INSTANCE_NEW_NAME"] = self.op.new_name
3179 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3182 def CheckPrereq(self):
3183 """Check prerequisites.
3185 This checks that the instance is in the cluster and is not running.
3188 instance = self.cfg.GetInstanceInfo(
3189 self.cfg.ExpandInstanceName(self.op.instance_name))
3190 if instance is None:
3191 raise errors.OpPrereqError("Instance '%s' not known" %
3192 self.op.instance_name)
3193 _CheckNodeOnline(self, instance.primary_node)
3195 if instance.admin_up:
3196 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3197 self.op.instance_name)
3198 remote_info = self.rpc.call_instance_info(instance.primary_node,
3200 instance.hypervisor)
3202 if remote_info.data:
3203 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3204 (self.op.instance_name,
3205 instance.primary_node))
3206 self.instance = instance
3208 # new name verification
3209 name_info = utils.HostInfo(self.op.new_name)
3211 self.op.new_name = new_name = name_info.name
3212 instance_list = self.cfg.GetInstanceList()
3213 if new_name in instance_list:
3214 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
3217 if not getattr(self.op, "ignore_ip", False):
3218 if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
3219 raise errors.OpPrereqError("IP %s of instance %s already in use" %
3220 (name_info.ip, new_name))
3223 def Exec(self, feedback_fn):
3224 """Reinstall the instance.
3227 inst = self.instance
3228 old_name = inst.name
3230 if inst.disk_template == constants.DT_FILE:
3231 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
3233 self.cfg.RenameInstance(inst.name, self.op.new_name)
3234 # Change the instance lock. This is definitely safe while we hold the BGL
3235 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
3236 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
3238 # re-read the instance from the configuration after rename
3239 inst = self.cfg.GetInstanceInfo(self.op.new_name)
3241 if inst.disk_template == constants.DT_FILE:
3242 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
3243 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
3244 old_file_storage_dir,
3245 new_file_storage_dir)
3248 raise errors.OpExecError("Could not connect to node '%s' to rename"
3249 " directory '%s' to '%s' (but the instance"
3250 " has been renamed in Ganeti)" % (
3251 inst.primary_node, old_file_storage_dir,
3252 new_file_storage_dir))
3254 if not result.data[0]:
3255 raise errors.OpExecError("Could not rename directory '%s' to '%s'"
3256 " (but the instance has been renamed in"
3257 " Ganeti)" % (old_file_storage_dir,
3258 new_file_storage_dir))
3260 _StartInstanceDisks(self, inst, None)
3262 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
3264 msg = result.RemoteFailMsg()
3266 msg = ("Could not run OS rename script for instance %s on node %s"
3267 " (but the instance has been renamed in Ganeti): %s" %
3268 (inst.name, inst.primary_node, msg))
3269 self.proc.LogWarning(msg)
3271 _ShutdownInstanceDisks(self, inst)
3274 class LURemoveInstance(LogicalUnit):
3275 """Remove an instance.
3278 HPATH = "instance-remove"
3279 HTYPE = constants.HTYPE_INSTANCE
3280 _OP_REQP = ["instance_name", "ignore_failures"]
3283 def ExpandNames(self):
3284 self._ExpandAndLockInstance()
3285 self.needed_locks[locking.LEVEL_NODE] = []
3286 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3288 def DeclareLocks(self, level):
3289 if level == locking.LEVEL_NODE:
3290 self._LockInstancesNodes()
3292 def BuildHooksEnv(self):
3295 This runs on master, primary and secondary nodes of the instance.
3298 env = _BuildInstanceHookEnvByObject(self, self.instance)
3299 nl = [self.cfg.GetMasterNode()]
3302 def CheckPrereq(self):
3303 """Check prerequisites.
3305 This checks that the instance is in the cluster.
3308 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3309 assert self.instance is not None, \
3310 "Cannot retrieve locked instance %s" % self.op.instance_name
3312 def Exec(self, feedback_fn):
3313 """Remove the instance.
3316 instance = self.instance
3317 logging.info("Shutting down instance %s on node %s",
3318 instance.name, instance.primary_node)
3320 result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3321 msg = result.RemoteFailMsg()
3323 if self.op.ignore_failures:
3324 feedback_fn("Warning: can't shutdown instance: %s" % msg)
3326 raise errors.OpExecError("Could not shutdown instance %s on"
3328 (instance.name, instance.primary_node, msg))
3330 logging.info("Removing block devices for instance %s", instance.name)
3332 if not _RemoveDisks(self, instance):
3333 if self.op.ignore_failures:
3334 feedback_fn("Warning: can't remove instance's disks")
3336 raise errors.OpExecError("Can't remove instance's disks")
3338 logging.info("Removing instance %s out of cluster config", instance.name)
3340 self.cfg.RemoveInstance(instance.name)
3341 self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3344 class LUQueryInstances(NoHooksLU):
3345 """Logical unit for querying instances.
3348 _OP_REQP = ["output_fields", "names", "use_locking"]
3350 _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3352 "disk_template", "ip", "mac", "bridge",
3353 "sda_size", "sdb_size", "vcpus", "tags",
3354 "network_port", "beparams",
3355 r"(disk)\.(size)/([0-9]+)",
3356 r"(disk)\.(sizes)", "disk_usage",
3357 r"(nic)\.(mac|ip|bridge)/([0-9]+)",
3358 r"(nic)\.(macs|ips|bridges)",
3359 r"(disk|nic)\.(count)",
3360 "serial_no", "hypervisor", "hvparams",] +
3362 for name in constants.HVS_PARAMETERS] +
3364 for name in constants.BES_PARAMETERS])
3365 _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3368 def ExpandNames(self):
3369 _CheckOutputFields(static=self._FIELDS_STATIC,
3370 dynamic=self._FIELDS_DYNAMIC,
3371 selected=self.op.output_fields)
3373 self.needed_locks = {}
3374 self.share_locks[locking.LEVEL_INSTANCE] = 1
3375 self.share_locks[locking.LEVEL_NODE] = 1
3378 self.wanted = _GetWantedInstances(self, self.op.names)
3380 self.wanted = locking.ALL_SET
3382 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3383 self.do_locking = self.do_node_query and self.op.use_locking
3385 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3386 self.needed_locks[locking.LEVEL_NODE] = []
3387 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3389 def DeclareLocks(self, level):
3390 if level == locking.LEVEL_NODE and self.do_locking:
3391 self._LockInstancesNodes()
3393 def CheckPrereq(self):
3394 """Check prerequisites.
3399 def Exec(self, feedback_fn):
3400 """Computes the list of nodes and their attributes.
3403 all_info = self.cfg.GetAllInstancesInfo()
3404 if self.wanted == locking.ALL_SET:
3405 # caller didn't specify instance names, so ordering is not important
3407 instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3409 instance_names = all_info.keys()
3410 instance_names = utils.NiceSort(instance_names)
3412 # caller did specify names, so we must keep the ordering
3414 tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3416 tgt_set = all_info.keys()
3417 missing = set(self.wanted).difference(tgt_set)
3419 raise errors.OpExecError("Some instances were removed before"
3420 " retrieving their data: %s" % missing)
3421 instance_names = self.wanted
3423 instance_list = [all_info[iname] for iname in instance_names]
3425 # begin data gathering
3427 nodes = frozenset([inst.primary_node for inst in instance_list])
3428 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3432 if self.do_node_query:
3434 node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3436 result = node_data[name]
3438 # offline nodes will be in both lists
3439 off_nodes.append(name)
3441 bad_nodes.append(name)
3444 live_data.update(result.data)
3445 # else no instance is alive
3447 live_data = dict([(name, {}) for name in instance_names])
3449 # end data gathering
3454 for instance in instance_list:
3456 i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3457 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3458 for field in self.op.output_fields:
3459 st_match = self._FIELDS_STATIC.Matches(field)
3464 elif field == "pnode":
3465 val = instance.primary_node
3466 elif field == "snodes":
3467 val = list(instance.secondary_nodes)
3468 elif field == "admin_state":
3469 val = instance.admin_up
3470 elif field == "oper_state":
3471 if instance.primary_node in bad_nodes:
3474 val = bool(live_data.get(instance.name))
3475 elif field == "status":
3476 if instance.primary_node in off_nodes:
3477 val = "ERROR_nodeoffline"
3478 elif instance.primary_node in bad_nodes:
3479 val = "ERROR_nodedown"
3481 running = bool(live_data.get(instance.name))
3483 if instance.admin_up:
3488 if instance.admin_up:
3492 elif field == "oper_ram":
3493 if instance.primary_node in bad_nodes:
3495 elif instance.name in live_data:
3496 val = live_data[instance.name].get("memory", "?")
3499 elif field == "disk_template":
3500 val = instance.disk_template
3502 val = instance.nics[0].ip
3503 elif field == "bridge":
3504 val = instance.nics[0].bridge
3505 elif field == "mac":
3506 val = instance.nics[0].mac
3507 elif field == "sda_size" or field == "sdb_size":
3508 idx = ord(field[2]) - ord('a')
3510 val = instance.FindDisk(idx).size
3511 except errors.OpPrereqError:
3513 elif field == "disk_usage": # total disk usage per node
3514 disk_sizes = [{'size': disk.size} for disk in instance.disks]
3515 val = _ComputeDiskSize(instance.disk_template, disk_sizes)
3516 elif field == "tags":
3517 val = list(instance.GetTags())
3518 elif field == "serial_no":
3519 val = instance.serial_no
3520 elif field == "network_port":
3521 val = instance.network_port
3522 elif field == "hypervisor":
3523 val = instance.hypervisor
3524 elif field == "hvparams":
3526 elif (field.startswith(HVPREFIX) and
3527 field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3528 val = i_hv.get(field[len(HVPREFIX):], None)
3529 elif field == "beparams":
3531 elif (field.startswith(BEPREFIX) and
3532 field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3533 val = i_be.get(field[len(BEPREFIX):], None)
3534 elif st_match and st_match.groups():
3535 # matches a variable list
3536 st_groups = st_match.groups()
3537 if st_groups and st_groups[0] == "disk":
3538 if st_groups[1] == "count":
3539 val = len(instance.disks)
3540 elif st_groups[1] == "sizes":
3541 val = [disk.size for disk in instance.disks]
3542 elif st_groups[1] == "size":
3544 val = instance.FindDisk(st_groups[2]).size
3545 except errors.OpPrereqError:
3548 assert False, "Unhandled disk parameter"
3549 elif st_groups[0] == "nic":
3550 if st_groups[1] == "count":
3551 val = len(instance.nics)
3552 elif st_groups[1] == "macs":
3553 val = [nic.mac for nic in instance.nics]
3554 elif st_groups[1] == "ips":
3555 val = [nic.ip for nic in instance.nics]
3556 elif st_groups[1] == "bridges":
3557 val = [nic.bridge for nic in instance.nics]
3560 nic_idx = int(st_groups[2])
3561 if nic_idx >= len(instance.nics):
3564 if st_groups[1] == "mac":
3565 val = instance.nics[nic_idx].mac
3566 elif st_groups[1] == "ip":
3567 val = instance.nics[nic_idx].ip
3568 elif st_groups[1] == "bridge":
3569 val = instance.nics[nic_idx].bridge
3571 assert False, "Unhandled NIC parameter"
3573 assert False, "Unhandled variable parameter"
3575 raise errors.ParameterError(field)
3582 class LUFailoverInstance(LogicalUnit):
3583 """Failover an instance.
3586 HPATH = "instance-failover"
3587 HTYPE = constants.HTYPE_INSTANCE
3588 _OP_REQP = ["instance_name", "ignore_consistency"]
3591 def ExpandNames(self):
3592 self._ExpandAndLockInstance()
3593 self.needed_locks[locking.LEVEL_NODE] = []
3594 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3596 def DeclareLocks(self, level):
3597 if level == locking.LEVEL_NODE:
3598 self._LockInstancesNodes()
3600 def BuildHooksEnv(self):
3603 This runs on master, primary and secondary nodes of the instance.
3607 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3609 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3610 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3613 def CheckPrereq(self):
3614 """Check prerequisites.
3616 This checks that the instance is in the cluster.
3619 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3620 assert self.instance is not None, \
3621 "Cannot retrieve locked instance %s" % self.op.instance_name
3623 bep = self.cfg.GetClusterInfo().FillBE(instance)
3624 if instance.disk_template not in constants.DTS_NET_MIRROR:
3625 raise errors.OpPrereqError("Instance's disk layout is not"
3626 " network mirrored, cannot failover.")
3628 secondary_nodes = instance.secondary_nodes
3629 if not secondary_nodes:
3630 raise errors.ProgrammerError("no secondary node but using "
3631 "a mirrored disk template")
3633 target_node = secondary_nodes[0]
3634 _CheckNodeOnline(self, target_node)
3635 _CheckNodeNotDrained(self, target_node)
3636 # check memory requirements on the secondary node
3637 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3638 instance.name, bep[constants.BE_MEMORY],
3639 instance.hypervisor)
3640 # check bridge existance
3641 _CheckInstanceBridgesExist(self, instance, node=target_node)
3643 def Exec(self, feedback_fn):
3644 """Failover an instance.
3646 The failover is done by shutting it down on its present node and
3647 starting it on the secondary.
3650 instance = self.instance
3652 source_node = instance.primary_node
3653 target_node = instance.secondary_nodes[0]
3655 feedback_fn("* checking disk consistency between source and target")
3656 for dev in instance.disks:
3657 # for drbd, these are drbd over lvm
3658 if not _CheckDiskConsistency(self, dev, target_node, False):
3659 if instance.admin_up and not self.op.ignore_consistency:
3660 raise errors.OpExecError("Disk %s is degraded on target node,"
3661 " aborting failover." % dev.iv_name)
3663 feedback_fn("* shutting down instance on source node")
3664 logging.info("Shutting down instance %s on node %s",
3665 instance.name, source_node)
3667 result = self.rpc.call_instance_shutdown(source_node, instance)
3668 msg = result.RemoteFailMsg()
3670 if self.op.ignore_consistency:
3671 self.proc.LogWarning("Could not shutdown instance %s on node %s."
3672 " Proceeding anyway. Please make sure node"
3673 " %s is down. Error details: %s",
3674 instance.name, source_node, source_node, msg)
3676 raise errors.OpExecError("Could not shutdown instance %s on"
3678 (instance.name, source_node, msg))
3680 feedback_fn("* deactivating the instance's disks on source node")
3681 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3682 raise errors.OpExecError("Can't shut down the instance's disks.")
3684 instance.primary_node = target_node
3685 # distribute new instance config to the other nodes
3686 self.cfg.Update(instance)
3688 # Only start the instance if it's marked as up
3689 if instance.admin_up:
3690 feedback_fn("* activating the instance's disks on target node")
3691 logging.info("Starting instance %s on node %s",
3692 instance.name, target_node)
3694 disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3695 ignore_secondaries=True)
3697 _ShutdownInstanceDisks(self, instance)
3698 raise errors.OpExecError("Can't activate the instance's disks")
3700 feedback_fn("* starting the instance on the target node")
3701 result = self.rpc.call_instance_start(target_node, instance, None, None)
3702 msg = result.RemoteFailMsg()
3704 _ShutdownInstanceDisks(self, instance)
3705 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3706 (instance.name, target_node, msg))
3709 class LUMigrateInstance(LogicalUnit):
3710 """Migrate an instance.
3712 This is migration without shutting down, compared to the failover,
3713 which is done with shutdown.
3716 HPATH = "instance-migrate"
3717 HTYPE = constants.HTYPE_INSTANCE
3718 _OP_REQP = ["instance_name", "live", "cleanup"]
3722 def ExpandNames(self):
3723 self._ExpandAndLockInstance()
3724 self.needed_locks[locking.LEVEL_NODE] = []
3725 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3727 def DeclareLocks(self, level):
3728 if level == locking.LEVEL_NODE:
3729 self._LockInstancesNodes()
3731 def BuildHooksEnv(self):
3734 This runs on master, primary and secondary nodes of the instance.
3737 env = _BuildInstanceHookEnvByObject(self, self.instance)
3738 env["MIGRATE_LIVE"] = self.op.live
3739 env["MIGRATE_CLEANUP"] = self.op.cleanup
3740 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3743 def CheckPrereq(self):
3744 """Check prerequisites.
3746 This checks that the instance is in the cluster.
3749 instance = self.cfg.GetInstanceInfo(
3750 self.cfg.ExpandInstanceName(self.op.instance_name))
3751 if instance is None:
3752 raise errors.OpPrereqError("Instance '%s' not known" %
3753 self.op.instance_name)
3755 if instance.disk_template != constants.DT_DRBD8:
3756 raise errors.OpPrereqError("Instance's disk layout is not"
3757 " drbd8, cannot migrate.")
3759 secondary_nodes = instance.secondary_nodes
3760 if not secondary_nodes:
3761 raise errors.ConfigurationError("No secondary node but using"
3762 " drbd8 disk template")
3764 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3766 target_node = secondary_nodes[0]
3767 # check memory requirements on the secondary node
3768 _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3769 instance.name, i_be[constants.BE_MEMORY],
3770 instance.hypervisor)
3772 # check bridge existance
3773 _CheckInstanceBridgesExist(self, instance, node=target_node)
3775 if not self.op.cleanup:
3776 _CheckNodeNotDrained(self, target_node)
3777 result = self.rpc.call_instance_migratable(instance.primary_node,
3779 msg = result.RemoteFailMsg()
3781 raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3784 self.instance = instance
3786 def _WaitUntilSync(self):
3787 """Poll with custom rpc for disk sync.
3789 This uses our own step-based rpc call.
3792 self.feedback_fn("* wait until resync is done")
3796 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3798 self.instance.disks)
3800 for node, nres in result.items():
3801 msg = nres.RemoteFailMsg()
3803 raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3805 node_done, node_percent = nres.payload
3806 all_done = all_done and node_done
3807 if node_percent is not None:
3808 min_percent = min(min_percent, node_percent)
3810 if min_percent < 100:
3811 self.feedback_fn(" - progress: %.1f%%" % min_percent)
3814 def _EnsureSecondary(self, node):
3815 """Demote a node to secondary.
3818 self.feedback_fn("* switching node %s to secondary mode" % node)
3820 for dev in self.instance.disks:
3821 self.cfg.SetDiskID(dev, node)
3823 result = self.rpc.call_blockdev_close(node, self.instance.name,
3824 self.instance.disks)
3825 msg = result.RemoteFailMsg()
3827 raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3828 " error %s" % (node, msg))
3830 def _GoStandalone(self):
3831 """Disconnect from the network.
3834 self.feedback_fn("* changing into standalone mode")
3835 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3836 self.instance.disks)
3837 for node, nres in result.items():
3838 msg = nres.RemoteFailMsg()
3840 raise errors.OpExecError("Cannot disconnect disks node %s,"
3841 " error %s" % (node, msg))
3843 def _GoReconnect(self, multimaster):
3844 """Reconnect to the network.
3850 msg = "single-master"
3851 self.feedback_fn("* changing disks into %s mode" % msg)
3852 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3853 self.instance.disks,
3854 self.instance.name, multimaster)
3855 for node, nres in result.items():
3856 msg = nres.RemoteFailMsg()
3858 raise errors.OpExecError("Cannot change disks config on node %s,"
3859 " error: %s" % (node, msg))
3861 def _ExecCleanup(self):
3862 """Try to cleanup after a failed migration.
3864 The cleanup is done by:
3865 - check that the instance is running only on one node
3866 (and update the config if needed)
3867 - change disks on its secondary node to secondary
3868 - wait until disks are fully synchronized
3869 - disconnect from the network
3870 - change disks into single-master mode
3871 - wait again until disks are fully synchronized
3874 instance = self.instance
3875 target_node = self.target_node
3876 source_node = self.source_node
3878 # check running on only one node
3879 self.feedback_fn("* checking where the instance actually runs"
3880 " (if this hangs, the hypervisor might be in"
3882 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3883 for node, result in ins_l.items():
3885 if not isinstance(result.data, list):
3886 raise errors.OpExecError("Can't contact node '%s'" % node)
3888 runningon_source = instance.name in ins_l[source_node].data
3889 runningon_target = instance.name in ins_l[target_node].data
3891 if runningon_source and runningon_target:
3892 raise errors.OpExecError("Instance seems to be running on two nodes,"
3893 " or the hypervisor is confused. You will have"
3894 " to ensure manually that it runs only on one"
3895 " and restart this operation.")
3897 if not (runningon_source or runningon_target):
3898 raise errors.OpExecError("Instance does not seem to be running at all."
3899 " In this case, it's safer to repair by"
3900 " running 'gnt-instance stop' to ensure disk"
3901 " shutdown, and then restarting it.")
3903 if runningon_target:
3904 # the migration has actually succeeded, we need to update the config
3905 self.feedback_fn("* instance running on secondary node (%s),"
3906 " updating config" % target_node)
3907 instance.primary_node = target_node
3908 self.cfg.Update(instance)
3909 demoted_node = source_node
3911 self.feedback_fn("* instance confirmed to be running on its"
3912 " primary node (%s)" % source_node)
3913 demoted_node = target_node
3915 self._EnsureSecondary(demoted_node)
3917 self._WaitUntilSync()
3918 except errors.OpExecError:
3919 # we ignore here errors, since if the device is standalone, it
3920 # won't be able to sync
3922 self._GoStandalone()
3923 self._GoReconnect(False)
3924 self._WaitUntilSync()
3926 self.feedback_fn("* done")
3928 def _RevertDiskStatus(self):
3929 """Try to revert the disk status after a failed migration.
3932 target_node = self.target_node
3934 self._EnsureSecondary(target_node)
3935 self._GoStandalone()
3936 self._GoReconnect(False)
3937 self._WaitUntilSync()
3938 except errors.OpExecError, err:
3939 self.LogWarning("Migration failed and I can't reconnect the"
3940 " drives: error '%s'\n"
3941 "Please look and recover the instance status" %
3944 def _AbortMigration(self):
3945 """Call the hypervisor code to abort a started migration.
3948 instance = self.instance
3949 target_node = self.target_node
3950 migration_info = self.migration_info
3952 abort_result = self.rpc.call_finalize_migration(target_node,
3956 abort_msg = abort_result.RemoteFailMsg()
3958 logging.error("Aborting migration failed on target node %s: %s" %
3959 (target_node, abort_msg))
3960 # Don't raise an exception here, as we stil have to try to revert the
3961 # disk status, even if this step failed.
3963 def _ExecMigration(self):
3964 """Migrate an instance.
3966 The migrate is done by:
3967 - change the disks into dual-master mode
3968 - wait until disks are fully synchronized again
3969 - migrate the instance
3970 - change disks on the new secondary node (the old primary) to secondary
3971 - wait until disks are fully synchronized
3972 - change disks into single-master mode
3975 instance = self.instance
3976 target_node = self.target_node
3977 source_node = self.source_node
3979 self.feedback_fn("* checking disk consistency between source and target")
3980 for dev in instance.disks:
3981 if not _CheckDiskConsistency(self, dev, target_node, False):
3982 raise errors.OpExecError("Disk %s is degraded or not fully"
3983 " synchronized on target node,"
3984 " aborting migrate." % dev.iv_name)
3986 # First get the migration information from the remote node
3987 result = self.rpc.call_migration_info(source_node, instance)
3988 msg = result.RemoteFailMsg()
3990 log_err = ("Failed fetching source migration information from %s: %s" %
3992 logging.error(log_err)
3993 raise errors.OpExecError(log_err)
3995 self.migration_info = migration_info = result.payload
3997 # Then switch the disks to master/master mode
3998 self._EnsureSecondary(target_node)
3999 self._GoStandalone()
4000 self._GoReconnect(True)
4001 self._WaitUntilSync()
4003 self.feedback_fn("* preparing %s to accept the instance" % target_node)
4004 result = self.rpc.call_accept_instance(target_node,
4007 self.nodes_ip[target_node])
4009 msg = result.RemoteFailMsg()
4011 logging.error("Instance pre-migration failed, trying to revert"
4012 " disk status: %s", msg)
4013 self._AbortMigration()
4014 self._RevertDiskStatus()
4015 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
4016 (instance.name, msg))
4018 self.feedback_fn("* migrating instance to %s" % target_node)
4020 result = self.rpc.call_instance_migrate(source_node, instance,
4021 self.nodes_ip[target_node],
4023 msg = result.RemoteFailMsg()
4025 logging.error("Instance migration failed, trying to revert"
4026 " disk status: %s", msg)
4027 self._AbortMigration()
4028 self._RevertDiskStatus()
4029 raise errors.OpExecError("Could not migrate instance %s: %s" %
4030 (instance.name, msg))
4033 instance.primary_node = target_node
4034 # distribute new instance config to the other nodes
4035 self.cfg.Update(instance)
4037 result = self.rpc.call_finalize_migration(target_node,
4041 msg = result.RemoteFailMsg()
4043 logging.error("Instance migration succeeded, but finalization failed:"
4045 raise errors.OpExecError("Could not finalize instance migration: %s" %
4048 self._EnsureSecondary(source_node)
4049 self._WaitUntilSync()
4050 self._GoStandalone()
4051 self._GoReconnect(False)
4052 self._WaitUntilSync()
4054 self.feedback_fn("* done")
4056 def Exec(self, feedback_fn):
4057 """Perform the migration.
4060 self.feedback_fn = feedback_fn
4062 self.source_node = self.instance.primary_node
4063 self.target_node = self.instance.secondary_nodes[0]
4064 self.all_nodes = [self.source_node, self.target_node]
4066 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
4067 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
4070 return self._ExecCleanup()
4072 return self._ExecMigration()
4075 def _CreateBlockDev(lu, node, instance, device, force_create,
4077 """Create a tree of block devices on a given node.
4079 If this device type has to be created on secondaries, create it and
4082 If not, just recurse to children keeping the same 'force' value.
4084 @param lu: the lu on whose behalf we execute
4085 @param node: the node on which to create the device
4086 @type instance: L{objects.Instance}
4087 @param instance: the instance which owns the device
4088 @type device: L{objects.Disk}
4089 @param device: the device to create
4090 @type force_create: boolean
4091 @param force_create: whether to force creation of this device; this
4092 will be change to True whenever we find a device which has
4093 CreateOnSecondary() attribute
4094 @param info: the extra 'metadata' we should attach to the device
4095 (this will be represented as a LVM tag)
4096 @type force_open: boolean
4097 @param force_open: this parameter will be passes to the
4098 L{backend.BlockdevCreate} function where it specifies
4099 whether we run on primary or not, and it affects both
4100 the child assembly and the device own Open() execution
4103 if device.CreateOnSecondary():
4107 for child in device.children:
4108 _CreateBlockDev(lu, node, instance, child, force_create,
4111 if not force_create:
4114 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
4117 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
4118 """Create a single block device on a given node.
4120 This will not recurse over children of the device, so they must be
4123 @param lu: the lu on whose behalf we execute
4124 @param node: the node on which to create the device
4125 @type instance: L{objects.Instance}
4126 @param instance: the instance which owns the device
4127 @type device: L{objects.Disk}
4128 @param device: the device to create
4129 @param info: the extra 'metadata' we should attach to the device
4130 (this will be represented as a LVM tag)
4131 @type force_open: boolean
4132 @param force_open: this parameter will be passes to the
4133 L{backend.BlockdevCreate} function where it specifies
4134 whether we run on primary or not, and it affects both
4135 the child assembly and the device own Open() execution
4138 lu.cfg.SetDiskID(device, node)
4139 result = lu.rpc.call_blockdev_create(node, device, device.size,
4140 instance.name, force_open, info)
4141 msg = result.RemoteFailMsg()
4143 raise errors.OpExecError("Can't create block device %s on"
4144 " node %s for instance %s: %s" %
4145 (device, node, instance.name, msg))
4146 if device.physical_id is None:
4147 device.physical_id = result.payload
4150 def _GenerateUniqueNames(lu, exts):
4151 """Generate a suitable LV name.
4153 This will generate a logical volume name for the given instance.
4158 new_id = lu.cfg.GenerateUniqueID()
4159 results.append("%s%s" % (new_id, val))
4163 def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
4165 """Generate a drbd8 device complete with its children.
4168 port = lu.cfg.AllocatePort()
4169 vgname = lu.cfg.GetVGName()
4170 shared_secret = lu.cfg.GenerateDRBDSecret()
4171 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4172 logical_id=(vgname, names[0]))
4173 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4174 logical_id=(vgname, names[1]))
4175 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
4176 logical_id=(primary, secondary, port,
4179 children=[dev_data, dev_meta],
4184 def _GenerateDiskTemplate(lu, template_name,
4185 instance_name, primary_node,
4186 secondary_nodes, disk_info,
4187 file_storage_dir, file_driver,
4189 """Generate the entire disk layout for a given template type.
4192 #TODO: compute space requirements
4194 vgname = lu.cfg.GetVGName()
4195 disk_count = len(disk_info)
4197 if template_name == constants.DT_DISKLESS:
4199 elif template_name == constants.DT_PLAIN:
4200 if len(secondary_nodes) != 0:
4201 raise errors.ProgrammerError("Wrong template configuration")
4203 names = _GenerateUniqueNames(lu, [".disk%d" % i
4204 for i in range(disk_count)])
4205 for idx, disk in enumerate(disk_info):
4206 disk_index = idx + base_index
4207 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
4208 logical_id=(vgname, names[idx]),
4209 iv_name="disk/%d" % disk_index,
4211 disks.append(disk_dev)
4212 elif template_name == constants.DT_DRBD8:
4213 if len(secondary_nodes) != 1:
4214 raise errors.ProgrammerError("Wrong template configuration")
4215 remote_node = secondary_nodes[0]
4216 minors = lu.cfg.AllocateDRBDMinor(
4217 [primary_node, remote_node] * len(disk_info), instance_name)
4220 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
4221 for i in range(disk_count)]):
4222 names.append(lv_prefix + "_data")
4223 names.append(lv_prefix + "_meta")
4224 for idx, disk in enumerate(disk_info):
4225 disk_index = idx + base_index
4226 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
4227 disk["size"], names[idx*2:idx*2+2],
4228 "disk/%d" % disk_index,
4229 minors[idx*2], minors[idx*2+1])
4230 disk_dev.mode = disk["mode"]
4231 disks.append(disk_dev)
4232 elif template_name == constants.DT_FILE:
4233 if len(secondary_nodes) != 0:
4234 raise errors.ProgrammerError("Wrong template configuration")
4236 for idx, disk in enumerate(disk_info):
4237 disk_index = idx + base_index
4238 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
4239 iv_name="disk/%d" % disk_index,
4240 logical_id=(file_driver,
4241 "%s/disk%d" % (file_storage_dir,
4244 disks.append(disk_dev)
4246 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
4250 def _GetInstanceInfoText(instance):
4251 """Compute that text that should be added to the disk's metadata.
4254 return "originstname+%s" % instance.name
4257 def _CreateDisks(lu, instance):
4258 """Create all disks for an instance.
4260 This abstracts away some work from AddInstance.
4262 @type lu: L{LogicalUnit}
4263 @param lu: the logical unit on whose behalf we execute
4264 @type instance: L{objects.Instance}
4265 @param instance: the instance whose disks we should create
4267 @return: the success of the creation
4270 info = _GetInstanceInfoText(instance)
4271 pnode = instance.primary_node
4273 if instance.disk_template == constants.DT_FILE:
4274 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4275 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
4277 if result.failed or not result.data:
4278 raise errors.OpExecError("Could not connect to node '%s'" % pnode)
4280 if not result.data[0]:
4281 raise errors.OpExecError("Failed to create directory '%s'" %
4284 # Note: this needs to be kept in sync with adding of disks in
4285 # LUSetInstanceParams
4286 for device in instance.disks:
4287 logging.info("Creating volume %s for instance %s",
4288 device.iv_name, instance.name)
4290 for node in instance.all_nodes:
4291 f_create = node == pnode
4292 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4295 def _RemoveDisks(lu, instance):
4296 """Remove all disks for an instance.
4298 This abstracts away some work from `AddInstance()` and
4299 `RemoveInstance()`. Note that in case some of the devices couldn't
4300 be removed, the removal will continue with the other ones (compare
4301 with `_CreateDisks()`).
4303 @type lu: L{LogicalUnit}
4304 @param lu: the logical unit on whose behalf we execute
4305 @type instance: L{objects.Instance}
4306 @param instance: the instance whose disks we should remove
4308 @return: the success of the removal
4311 logging.info("Removing block devices for instance %s", instance.name)
4314 for device in instance.disks:
4315 for node, disk in device.ComputeNodeTree(instance.primary_node):
4316 lu.cfg.SetDiskID(disk, node)
4317 msg = lu.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
4319 lu.LogWarning("Could not remove block device %s on node %s,"
4320 " continuing anyway: %s", device.iv_name, node, msg)
4323 if instance.disk_template == constants.DT_FILE:
4324 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4325 result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4327 if result.failed or not result.data:
4328 logging.error("Could not remove directory '%s'", file_storage_dir)
4334 def _ComputeDiskSize(disk_template, disks):
4335 """Compute disk size requirements in the volume group
4338 # Required free disk space as a function of disk and swap space
4340 constants.DT_DISKLESS: None,
4341 constants.DT_PLAIN: sum(d["size"] for d in disks),
4342 # 128 MB are added for drbd metadata for each disk
4343 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4344 constants.DT_FILE: None,
4347 if disk_template not in req_size_dict:
4348 raise errors.ProgrammerError("Disk template '%s' size requirement"
4349 " is unknown" % disk_template)
4351 return req_size_dict[disk_template]
4354 def _CheckHVParams(lu, nodenames, hvname, hvparams):
4355 """Hypervisor parameter validation.
4357 This function abstract the hypervisor parameter validation to be
4358 used in both instance create and instance modify.
4360 @type lu: L{LogicalUnit}
4361 @param lu: the logical unit for which we check
4362 @type nodenames: list
4363 @param nodenames: the list of nodes on which we should check
4364 @type hvname: string
4365 @param hvname: the name of the hypervisor we should use
4366 @type hvparams: dict
4367 @param hvparams: the parameters which we need to check
4368 @raise errors.OpPrereqError: if the parameters are not valid
4371 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4374 for node in nodenames:
4378 msg = info.RemoteFailMsg()
4380 raise errors.OpPrereqError("Hypervisor parameter validation"
4381 " failed on node %s: %s" % (node, msg))
4384 class LUCreateInstance(LogicalUnit):
4385 """Create an instance.
4388 HPATH = "instance-add"
4389 HTYPE = constants.HTYPE_INSTANCE
4390 _OP_REQP = ["instance_name", "disks", "disk_template",
4392 "wait_for_sync", "ip_check", "nics",
4393 "hvparams", "beparams"]
4396 def _ExpandNode(self, node):
4397 """Expands and checks one node name.
4400 node_full = self.cfg.ExpandNodeName(node)
4401 if node_full is None:
4402 raise errors.OpPrereqError("Unknown node %s" % node)
4405 def ExpandNames(self):
4406 """ExpandNames for CreateInstance.
4408 Figure out the right locks for instance creation.
4411 self.needed_locks = {}
4413 # set optional parameters to none if they don't exist
4414 for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4415 if not hasattr(self.op, attr):
4416 setattr(self.op, attr, None)
4418 # cheap checks, mostly valid constants given
4420 # verify creation mode
4421 if self.op.mode not in (constants.INSTANCE_CREATE,
4422 constants.INSTANCE_IMPORT):
4423 raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4426 # disk template and mirror node verification
4427 if self.op.disk_template not in constants.DISK_TEMPLATES:
4428 raise errors.OpPrereqError("Invalid disk template name")
4430 if self.op.hypervisor is None:
4431 self.op.hypervisor = self.cfg.GetHypervisorType()
4433 cluster = self.cfg.GetClusterInfo()
4434 enabled_hvs = cluster.enabled_hypervisors
4435 if self.op.hypervisor not in enabled_hvs:
4436 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4437 " cluster (%s)" % (self.op.hypervisor,
4438 ",".join(enabled_hvs)))
4440 # check hypervisor parameter syntax (locally)
4441 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4442 filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
4444 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4445 hv_type.CheckParameterSyntax(filled_hvp)
4447 # fill and remember the beparams dict
4448 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4449 self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
4452 #### instance parameters check
4454 # instance name verification
4455 hostname1 = utils.HostInfo(self.op.instance_name)
4456 self.op.instance_name = instance_name = hostname1.name
4458 # this is just a preventive check, but someone might still add this
4459 # instance in the meantime, and creation will fail at lock-add time
4460 if instance_name in self.cfg.GetInstanceList():
4461 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4464 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4468 for idx, nic in enumerate(self.op.nics):
4469 nic_mode_req = nic.get("mode", None)
4470 nic_mode = nic_mode_req
4471 if nic_mode is None:
4472 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
4474 # in routed mode, for the first nic, the default ip is 'auto'
4475 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
4476 default_ip_mode = constants.VALUE_AUTO
4478 default_ip_mode = constants.VALUE_NONE
4480 # ip validity checks
4481 ip = nic.get("ip", default_ip_mode)
4482 if ip is None or ip.lower() == constants.VALUE_NONE:
4484 elif ip.lower() == constants.VALUE_AUTO:
4485 nic_ip = hostname1.ip
4487 if not utils.IsValidIP(ip):
4488 raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4489 " like a valid IP" % ip)
4492 # TODO: check the ip for uniqueness !!
4493 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
4494 raise errors.OpPrereqError("Routed nic mode requires an ip address")
4496 # MAC address verification
4497 mac = nic.get("mac", constants.VALUE_AUTO)
4498 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4499 if not utils.IsValidMac(mac.lower()):
4500 raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4502 # bridge verification
4503 bridge = nic.get("bridge", None)
4504 link = nic.get("link", None)
4506 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
4507 elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
4508 raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic")
4514 nicparams[constants.NIC_MODE] = nic_mode_req
4516 nicparams[constants.NIC_LINK] = link
4518 check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
4520 objects.NIC.CheckParameterSyntax(check_params)
4521 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
4523 # disk checks/pre-build
4525 for disk in self.op.disks:
4526 mode = disk.get("mode", constants.DISK_RDWR)
4527 if mode not in constants.DISK_ACCESS_SET:
4528 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4530 size = disk.get("size", None)
4532 raise errors.OpPrereqError("Missing disk size")
4536 raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4537 self.disks.append({"size": size, "mode": mode})
4539 # used in CheckPrereq for ip ping check
4540 self.check_ip = hostname1.ip
4542 # file storage checks
4543 if (self.op.file_driver and
4544 not self.op.file_driver in constants.FILE_DRIVER):
4545 raise errors.OpPrereqError("Invalid file driver name '%s'" %
4546 self.op.file_driver)
4548 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4549 raise errors.OpPrereqError("File storage directory path not absolute")
4551 ### Node/iallocator related checks
4552 if [self.op.iallocator, self.op.pnode].count(None) != 1:
4553 raise errors.OpPrereqError("One and only one of iallocator and primary"
4554 " node must be given")
4556 if self.op.iallocator:
4557 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4559 self.op.pnode = self._ExpandNode(self.op.pnode)
4560 nodelist = [self.op.pnode]
4561 if self.op.snode is not None:
4562 self.op.snode = self._ExpandNode(self.op.snode)
4563 nodelist.append(self.op.snode)
4564 self.needed_locks[locking.LEVEL_NODE] = nodelist
4566 # in case of import lock the source node too
4567 if self.op.mode == constants.INSTANCE_IMPORT:
4568 src_node = getattr(self.op, "src_node", None)
4569 src_path = getattr(self.op, "src_path", None)
4571 if src_path is None:
4572 self.op.src_path = src_path = self.op.instance_name
4574 if src_node is None:
4575 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4576 self.op.src_node = None
4577 if os.path.isabs(src_path):
4578 raise errors.OpPrereqError("Importing an instance from an absolute"
4579 " path requires a source node option.")
4581 self.op.src_node = src_node = self._ExpandNode(src_node)
4582 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4583 self.needed_locks[locking.LEVEL_NODE].append(src_node)
4584 if not os.path.isabs(src_path):
4585 self.op.src_path = src_path = \
4586 os.path.join(constants.EXPORT_DIR, src_path)
4588 else: # INSTANCE_CREATE
4589 if getattr(self.op, "os_type", None) is None:
4590 raise errors.OpPrereqError("No guest OS specified")
4592 def _RunAllocator(self):
4593 """Run the allocator based on input opcode.
4596 nics = [n.ToDict() for n in self.nics]
4597 ial = IAllocator(self,
4598 mode=constants.IALLOCATOR_MODE_ALLOC,
4599 name=self.op.instance_name,
4600 disk_template=self.op.disk_template,
4603 vcpus=self.be_full[constants.BE_VCPUS],
4604 mem_size=self.be_full[constants.BE_MEMORY],
4607 hypervisor=self.op.hypervisor,
4610 ial.Run(self.op.iallocator)
4613 raise errors.OpPrereqError("Can't compute nodes using"
4614 " iallocator '%s': %s" % (self.op.iallocator,
4616 if len(ial.nodes) != ial.required_nodes:
4617 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4618 " of nodes (%s), required %s" %
4619 (self.op.iallocator, len(ial.nodes),
4620 ial.required_nodes))
4621 self.op.pnode = ial.nodes[0]
4622 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4623 self.op.instance_name, self.op.iallocator,
4624 ", ".join(ial.nodes))
4625 if ial.required_nodes == 2:
4626 self.op.snode = ial.nodes[1]
4628 def BuildHooksEnv(self):
4631 This runs on master, primary and secondary nodes of the instance.
4635 "ADD_MODE": self.op.mode,
4637 if self.op.mode == constants.INSTANCE_IMPORT:
4638 env["SRC_NODE"] = self.op.src_node
4639 env["SRC_PATH"] = self.op.src_path
4640 env["SRC_IMAGES"] = self.src_images
4642 env.update(_BuildInstanceHookEnv(
4643 name=self.op.instance_name,
4644 primary_node=self.op.pnode,
4645 secondary_nodes=self.secondaries,
4646 status=self.op.start,
4647 os_type=self.op.os_type,
4648 memory=self.be_full[constants.BE_MEMORY],
4649 vcpus=self.be_full[constants.BE_VCPUS],
4650 nics=_PreBuildNICHooksList(self, self.nics),
4651 disk_template=self.op.disk_template,
4652 disks=[(d["size"], d["mode"]) for d in self.disks],
4655 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4660 def CheckPrereq(self):
4661 """Check prerequisites.
4664 if (not self.cfg.GetVGName() and
4665 self.op.disk_template not in constants.DTS_NOT_LVM):
4666 raise errors.OpPrereqError("Cluster does not support lvm-based"
4669 if self.op.mode == constants.INSTANCE_IMPORT:
4670 src_node = self.op.src_node
4671 src_path = self.op.src_path
4673 if src_node is None:
4674 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
4675 exp_list = self.rpc.call_export_list(locked_nodes)
4677 for node in exp_list:
4678 if exp_list[node].RemoteFailMsg():
4680 if src_path in exp_list[node].payload:
4682 self.op.src_node = src_node = node
4683 self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4687 raise errors.OpPrereqError("No export found for relative path %s" %
4690 _CheckNodeOnline(self, src_node)
4691 result = self.rpc.call_export_info(src_node, src_path)
4692 msg = result.RemoteFailMsg()
4694 raise errors.OpPrereqError("No export or invalid export found in"
4695 " dir %s: %s" % (src_path, msg))
4697 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
4698 if not export_info.has_section(constants.INISECT_EXP):
4699 raise errors.ProgrammerError("Corrupted export config")
4701 ei_version = export_info.get(constants.INISECT_EXP, 'version')
4702 if (int(ei_version) != constants.EXPORT_VERSION):
4703 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4704 (ei_version, constants.EXPORT_VERSION))
4706 # Check that the new instance doesn't have less disks than the export
4707 instance_disks = len(self.disks)
4708 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4709 if instance_disks < export_disks:
4710 raise errors.OpPrereqError("Not enough disks to import."
4711 " (instance: %d, export: %d)" %
4712 (instance_disks, export_disks))
4714 self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4716 for idx in range(export_disks):
4717 option = 'disk%d_dump' % idx
4718 if export_info.has_option(constants.INISECT_INS, option):
4719 # FIXME: are the old os-es, disk sizes, etc. useful?
4720 export_name = export_info.get(constants.INISECT_INS, option)
4721 image = os.path.join(src_path, export_name)
4722 disk_images.append(image)
4724 disk_images.append(False)
4726 self.src_images = disk_images
4728 old_name = export_info.get(constants.INISECT_INS, 'name')
4729 # FIXME: int() here could throw a ValueError on broken exports
4730 exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4731 if self.op.instance_name == old_name:
4732 for idx, nic in enumerate(self.nics):
4733 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4734 nic_mac_ini = 'nic%d_mac' % idx
4735 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4737 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
4738 # ip ping checks (we use the same ip that was resolved in ExpandNames)
4739 if self.op.start and not self.op.ip_check:
4740 raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4741 " adding an instance in start mode")
4743 if self.op.ip_check:
4744 if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4745 raise errors.OpPrereqError("IP %s of instance %s already in use" %
4746 (self.check_ip, self.op.instance_name))
4748 #### mac address generation
4749 # By generating here the mac address both the allocator and the hooks get
4750 # the real final mac address rather than the 'auto' or 'generate' value.
4751 # There is a race condition between the generation and the instance object
4752 # creation, which means that we know the mac is valid now, but we're not
4753 # sure it will be when we actually add the instance. If things go bad
4754 # adding the instance will abort because of a duplicate mac, and the
4755 # creation job will fail.
4756 for nic in self.nics:
4757 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4758 nic.mac = self.cfg.GenerateMAC()
4762 if self.op.iallocator is not None:
4763 self._RunAllocator()
4765 #### node related checks
4767 # check primary node
4768 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4769 assert self.pnode is not None, \
4770 "Cannot retrieve locked node %s" % self.op.pnode
4772 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4775 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
4778 self.secondaries = []
4780 # mirror node verification
4781 if self.op.disk_template in constants.DTS_NET_MIRROR:
4782 if self.op.snode is None:
4783 raise errors.OpPrereqError("The networked disk templates need"
4785 if self.op.snode == pnode.name:
4786 raise errors.OpPrereqError("The secondary node cannot be"
4787 " the primary node.")
4788 _CheckNodeOnline(self, self.op.snode)
4789 _CheckNodeNotDrained(self, self.op.snode)
4790 self.secondaries.append(self.op.snode)
4792 nodenames = [pnode.name] + self.secondaries
4794 req_size = _ComputeDiskSize(self.op.disk_template,
4797 # Check lv size requirements
4798 if req_size is not None:
4799 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4801 for node in nodenames:
4802 info = nodeinfo[node]
4806 raise errors.OpPrereqError("Cannot get current information"
4807 " from node '%s'" % node)
4808 vg_free = info.get('vg_free', None)
4809 if not isinstance(vg_free, int):
4810 raise errors.OpPrereqError("Can't compute free disk space on"
4812 if req_size > info['vg_free']:
4813 raise errors.OpPrereqError("Not enough disk space on target node %s."
4814 " %d MB available, %d MB required" %
4815 (node, info['vg_free'], req_size))
4817 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4820 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4822 if not isinstance(result.data, objects.OS):
4823 raise errors.OpPrereqError("OS '%s' not in supported os list for"
4824 " primary node" % self.op.os_type)
4826 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
4828 # memory check on primary node
4830 _CheckNodeFreeMemory(self, self.pnode.name,
4831 "creating instance %s" % self.op.instance_name,
4832 self.be_full[constants.BE_MEMORY],
4835 def Exec(self, feedback_fn):
4836 """Create and add the instance to the cluster.
4839 instance = self.op.instance_name
4840 pnode_name = self.pnode.name
4842 ht_kind = self.op.hypervisor
4843 if ht_kind in constants.HTS_REQ_PORT:
4844 network_port = self.cfg.AllocatePort()
4848 ##if self.op.vnc_bind_address is None:
4849 ## self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4851 # this is needed because os.path.join does not accept None arguments
4852 if self.op.file_storage_dir is None:
4853 string_file_storage_dir = ""
4855 string_file_storage_dir = self.op.file_storage_dir
4857 # build the full file storage dir path
4858 file_storage_dir = os.path.normpath(os.path.join(
4859 self.cfg.GetFileStorageDir(),
4860 string_file_storage_dir, instance))
4863 disks = _GenerateDiskTemplate(self,
4864 self.op.disk_template,
4865 instance, pnode_name,
4869 self.op.file_driver,
4872 iobj = objects.Instance(name=instance, os=self.op.os_type,
4873 primary_node=pnode_name,
4874 nics=self.nics, disks=disks,
4875 disk_template=self.op.disk_template,
4877 network_port=network_port,
4878 beparams=self.op.beparams,
4879 hvparams=self.op.hvparams,
4880 hypervisor=self.op.hypervisor,
4883 feedback_fn("* creating instance disks...")
4885 _CreateDisks(self, iobj)
4886 except errors.OpExecError:
4887 self.LogWarning("Device creation failed, reverting...")
4889 _RemoveDisks(self, iobj)
4891 self.cfg.ReleaseDRBDMinors(instance)
4894 feedback_fn("adding instance %s to cluster config" % instance)
4896 self.cfg.AddInstance(iobj)
4897 # Declare that we don't want to remove the instance lock anymore, as we've
4898 # added the instance to the config
4899 del self.remove_locks[locking.LEVEL_INSTANCE]
4900 # Unlock all the nodes
4901 if self.op.mode == constants.INSTANCE_IMPORT:
4902 nodes_keep = [self.op.src_node]
4903 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4904 if node != self.op.src_node]
4905 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4906 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4908 self.context.glm.release(locking.LEVEL_NODE)
4909 del self.acquired_locks[locking.LEVEL_NODE]
4911 if self.op.wait_for_sync:
4912 disk_abort = not _WaitForSync(self, iobj)
4913 elif iobj.disk_template in constants.DTS_NET_MIRROR:
4914 # make sure the disks are not degraded (still sync-ing is ok)
4916 feedback_fn("* checking mirrors status")
4917 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4922 _RemoveDisks(self, iobj)
4923 self.cfg.RemoveInstance(iobj.name)
4924 # Make sure the instance lock gets removed
4925 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4926 raise errors.OpExecError("There are some degraded disks for"
4929 feedback_fn("creating os for instance %s on node %s" %
4930 (instance, pnode_name))
4932 if iobj.disk_template != constants.DT_DISKLESS:
4933 if self.op.mode == constants.INSTANCE_CREATE:
4934 feedback_fn("* running the instance OS create scripts...")
4935 result = self.rpc.call_instance_os_add(pnode_name, iobj, False)
4936 msg = result.RemoteFailMsg()
4938 raise errors.OpExecError("Could not add os for instance %s"
4940 (instance, pnode_name, msg))
4942 elif self.op.mode == constants.INSTANCE_IMPORT:
4943 feedback_fn("* running the instance OS import scripts...")
4944 src_node = self.op.src_node
4945 src_images = self.src_images
4946 cluster_name = self.cfg.GetClusterName()
4947 import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4948 src_node, src_images,
4950 import_result.Raise()
4951 for idx, result in enumerate(import_result.data):
4953 self.LogWarning("Could not import the image %s for instance"
4954 " %s, disk %d, on node %s" %
4955 (src_images[idx], instance, idx, pnode_name))
4957 # also checked in the prereq part
4958 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4962 iobj.admin_up = True
4963 self.cfg.Update(iobj)
4964 logging.info("Starting instance %s on node %s", instance, pnode_name)
4965 feedback_fn("* starting instance...")
4966 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
4967 msg = result.RemoteFailMsg()
4969 raise errors.OpExecError("Could not start instance: %s" % msg)
4972 class LUConnectConsole(NoHooksLU):
4973 """Connect to an instance's console.
4975 This is somewhat special in that it returns the command line that
4976 you need to run on the master node in order to connect to the
4980 _OP_REQP = ["instance_name"]
4983 def ExpandNames(self):
4984 self._ExpandAndLockInstance()
4986 def CheckPrereq(self):
4987 """Check prerequisites.
4989 This checks that the instance is in the cluster.
4992 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4993 assert self.instance is not None, \
4994 "Cannot retrieve locked instance %s" % self.op.instance_name
4995 _CheckNodeOnline(self, self.instance.primary_node)
4997 def Exec(self, feedback_fn):
4998 """Connect to the console of an instance
5001 instance = self.instance
5002 node = instance.primary_node
5004 node_insts = self.rpc.call_instance_list([node],
5005 [instance.hypervisor])[node]
5008 if instance.name not in node_insts.data:
5009 raise errors.OpExecError("Instance %s is not running." % instance.name)
5011 logging.debug("Connecting to console of %s on %s", instance.name, node)
5013 hyper = hypervisor.GetHypervisor(instance.hypervisor)
5014 cluster = self.cfg.GetClusterInfo()
5015 # beparams and hvparams are passed separately, to avoid editing the
5016 # instance and then saving the defaults in the instance itself.
5017 hvparams = cluster.FillHV(instance)
5018 beparams = cluster.FillBE(instance)
5019 console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
5022 return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
5025 class LUReplaceDisks(LogicalUnit):
5026 """Replace the disks of an instance.
5029 HPATH = "mirrors-replace"
5030 HTYPE = constants.HTYPE_INSTANCE
5031 _OP_REQP = ["instance_name", "mode", "disks"]
5034 def CheckArguments(self):
5035 if not hasattr(self.op, "remote_node"):
5036 self.op.remote_node = None
5037 if not hasattr(self.op, "iallocator"):
5038 self.op.iallocator = None
5040 # check for valid parameter combination
5041 cnt = [self.op.remote_node, self.op.iallocator].count(None)
5042 if self.op.mode == constants.REPLACE_DISK_CHG:
5044 raise errors.OpPrereqError("When changing the secondary either an"
5045 " iallocator script must be used or the"
5048 raise errors.OpPrereqError("Give either the iallocator or the new"
5049 " secondary, not both")
5050 else: # not replacing the secondary
5052 raise errors.OpPrereqError("The iallocator and new node options can"
5053 " be used only when changing the"
5056 def ExpandNames(self):
5057 self._ExpandAndLockInstance()
5059 if self.op.iallocator is not None:
5060 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5061 elif self.op.remote_node is not None:
5062 remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
5063 if remote_node is None:
5064 raise errors.OpPrereqError("Node '%s' not known" %
5065 self.op.remote_node)
5066 self.op.remote_node = remote_node
5067 # Warning: do not remove the locking of the new secondary here
5068 # unless DRBD8.AddChildren is changed to work in parallel;
5069 # currently it doesn't since parallel invocations of
5070 # FindUnusedMinor will conflict
5071 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
5072 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5074 self.needed_locks[locking.LEVEL_NODE] = []
5075 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5077 def DeclareLocks(self, level):
5078 # If we're not already locking all nodes in the set we have to declare the
5079 # instance's primary/secondary nodes.
5080 if (level == locking.LEVEL_NODE and
5081 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
5082 self._LockInstancesNodes()
5084 def _RunAllocator(self):
5085 """Compute a new secondary node using an IAllocator.
5088 ial = IAllocator(self,
5089 mode=constants.IALLOCATOR_MODE_RELOC,
5090 name=self.op.instance_name,
5091 relocate_from=[self.sec_node])
5093 ial.Run(self.op.iallocator)
5096 raise errors.OpPrereqError("Can't compute nodes using"
5097 " iallocator '%s': %s" % (self.op.iallocator,
5099 if len(ial.nodes) != ial.required_nodes:
5100 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
5101 " of nodes (%s), required %s" %
5102 (len(ial.nodes), ial.required_nodes))
5103 self.op.remote_node = ial.nodes[0]
5104 self.LogInfo("Selected new secondary for the instance: %s",
5105 self.op.remote_node)
5107 def BuildHooksEnv(self):
5110 This runs on the master, the primary and all the secondaries.
5114 "MODE": self.op.mode,
5115 "NEW_SECONDARY": self.op.remote_node,
5116 "OLD_SECONDARY": self.instance.secondary_nodes[0],
5118 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5120 self.cfg.GetMasterNode(),
5121 self.instance.primary_node,
5123 if self.op.remote_node is not None:
5124 nl.append(self.op.remote_node)
5127 def CheckPrereq(self):
5128 """Check prerequisites.
5130 This checks that the instance is in the cluster.
5133 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5134 assert instance is not None, \
5135 "Cannot retrieve locked instance %s" % self.op.instance_name
5136 self.instance = instance
5138 if instance.disk_template != constants.DT_DRBD8:
5139 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
5142 if len(instance.secondary_nodes) != 1:
5143 raise errors.OpPrereqError("The instance has a strange layout,"
5144 " expected one secondary but found %d" %
5145 len(instance.secondary_nodes))
5147 self.sec_node = instance.secondary_nodes[0]
5149 if self.op.iallocator is not None:
5150 self._RunAllocator()
5152 remote_node = self.op.remote_node
5153 if remote_node is not None:
5154 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
5155 assert self.remote_node_info is not None, \
5156 "Cannot retrieve locked node %s" % remote_node
5158 self.remote_node_info = None
5159 if remote_node == instance.primary_node:
5160 raise errors.OpPrereqError("The specified node is the primary node of"
5162 elif remote_node == self.sec_node:
5163 raise errors.OpPrereqError("The specified node is already the"
5164 " secondary node of the instance.")
5166 if self.op.mode == constants.REPLACE_DISK_PRI:
5167 n1 = self.tgt_node = instance.primary_node
5168 n2 = self.oth_node = self.sec_node
5169 elif self.op.mode == constants.REPLACE_DISK_SEC:
5170 n1 = self.tgt_node = self.sec_node
5171 n2 = self.oth_node = instance.primary_node
5172 elif self.op.mode == constants.REPLACE_DISK_CHG:
5173 n1 = self.new_node = remote_node
5174 n2 = self.oth_node = instance.primary_node
5175 self.tgt_node = self.sec_node
5176 _CheckNodeNotDrained(self, remote_node)
5178 raise errors.ProgrammerError("Unhandled disk replace mode")
5180 _CheckNodeOnline(self, n1)
5181 _CheckNodeOnline(self, n2)
5183 if not self.op.disks:
5184 self.op.disks = range(len(instance.disks))
5186 for disk_idx in self.op.disks:
5187 instance.FindDisk(disk_idx)
5189 def _ExecD8DiskOnly(self, feedback_fn):
5190 """Replace a disk on the primary or secondary for dbrd8.
5192 The algorithm for replace is quite complicated:
5194 1. for each disk to be replaced:
5196 1. create new LVs on the target node with unique names
5197 1. detach old LVs from the drbd device
5198 1. rename old LVs to name_replaced.<time_t>
5199 1. rename new LVs to old LVs
5200 1. attach the new LVs (with the old names now) to the drbd device
5202 1. wait for sync across all devices
5204 1. for each modified disk:
5206 1. remove old LVs (which have the name name_replaces.<time_t>)
5208 Failures are not very well handled.
5212 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5213 instance = self.instance
5215 vgname = self.cfg.GetVGName()
5218 tgt_node = self.tgt_node
5219 oth_node = self.oth_node
5221 # Step: check device activation
5222 self.proc.LogStep(1, steps_total, "check device existence")
5223 info("checking volume groups")
5224 my_vg = cfg.GetVGName()
5225 results = self.rpc.call_vg_list([oth_node, tgt_node])
5227 raise errors.OpExecError("Can't list volume groups on the nodes")
5228 for node in oth_node, tgt_node:
5230 msg = res.RemoteFailMsg()
5232 raise errors.OpExecError("Error checking node %s: %s" % (node, msg))
5233 if my_vg not in res.payload:
5234 raise errors.OpExecError("Volume group '%s' not found on %s" %
5236 for idx, dev in enumerate(instance.disks):
5237 if idx not in self.op.disks:
5239 for node in tgt_node, oth_node:
5240 info("checking disk/%d on %s" % (idx, node))
5241 cfg.SetDiskID(dev, node)
5242 result = self.rpc.call_blockdev_find(node, dev)
5243 msg = result.RemoteFailMsg()
5244 if not msg and not result.payload:
5245 msg = "disk not found"
5247 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5250 # Step: check other node consistency
5251 self.proc.LogStep(2, steps_total, "check peer consistency")
5252 for idx, dev in enumerate(instance.disks):
5253 if idx not in self.op.disks:
5255 info("checking disk/%d consistency on %s" % (idx, oth_node))
5256 if not _CheckDiskConsistency(self, dev, oth_node,
5257 oth_node==instance.primary_node):
5258 raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
5259 " to replace disks on this node (%s)" %
5260 (oth_node, tgt_node))
5262 # Step: create new storage
5263 self.proc.LogStep(3, steps_total, "allocate new storage")
5264 for idx, dev in enumerate(instance.disks):
5265 if idx not in self.op.disks:
5268 cfg.SetDiskID(dev, tgt_node)
5269 lv_names = [".disk%d_%s" % (idx, suf)
5270 for suf in ["data", "meta"]]
5271 names = _GenerateUniqueNames(self, lv_names)
5272 lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5273 logical_id=(vgname, names[0]))
5274 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5275 logical_id=(vgname, names[1]))
5276 new_lvs = [lv_data, lv_meta]
5277 old_lvs = dev.children
5278 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
5279 info("creating new local storage on %s for %s" %
5280 (tgt_node, dev.iv_name))
5281 # we pass force_create=True to force the LVM creation
5282 for new_lv in new_lvs:
5283 _CreateBlockDev(self, tgt_node, instance, new_lv, True,
5284 _GetInstanceInfoText(instance), False)
5286 # Step: for each lv, detach+rename*2+attach
5287 self.proc.LogStep(4, steps_total, "change drbd configuration")
5288 for dev, old_lvs, new_lvs in iv_names.itervalues():
5289 info("detaching %s drbd from local storage" % dev.iv_name)
5290 result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
5291 msg = result.RemoteFailMsg()
5293 raise errors.OpExecError("Can't detach drbd from local storage on node"
5294 " %s for device %s: %s" %
5295 (tgt_node, dev.iv_name, msg))
5297 #cfg.Update(instance)
5299 # ok, we created the new LVs, so now we know we have the needed
5300 # storage; as such, we proceed on the target node to rename
5301 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
5302 # using the assumption that logical_id == physical_id (which in
5303 # turn is the unique_id on that node)
5305 # FIXME(iustin): use a better name for the replaced LVs
5306 temp_suffix = int(time.time())
5307 ren_fn = lambda d, suff: (d.physical_id[0],
5308 d.physical_id[1] + "_replaced-%s" % suff)
5309 # build the rename list based on what LVs exist on the node
5311 for to_ren in old_lvs:
5312 result = self.rpc.call_blockdev_find(tgt_node, to_ren)
5313 if not result.RemoteFailMsg() and result.payload:
5315 rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
5317 info("renaming the old LVs on the target node")
5318 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5319 msg = result.RemoteFailMsg()
5321 raise errors.OpExecError("Can't rename old LVs on node %s: %s" %
5323 # now we rename the new LVs to the old LVs
5324 info("renaming the new LVs on the target node")
5325 rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
5326 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5327 msg = result.RemoteFailMsg()
5329 raise errors.OpExecError("Can't rename new LVs on node %s: %s" %
5332 for old, new in zip(old_lvs, new_lvs):
5333 new.logical_id = old.logical_id
5334 cfg.SetDiskID(new, tgt_node)
5336 for disk in old_lvs:
5337 disk.logical_id = ren_fn(disk, temp_suffix)
5338 cfg.SetDiskID(disk, tgt_node)
5340 # now that the new lvs have the old name, we can add them to the device
5341 info("adding new mirror component on %s" % tgt_node)
5342 result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
5343 msg = result.RemoteFailMsg()
5345 for new_lv in new_lvs:
5346 msg = self.rpc.call_blockdev_remove(tgt_node, new_lv).RemoteFailMsg()
5348 warning("Can't rollback device %s: %s", dev, msg,
5349 hint="cleanup manually the unused logical volumes")
5350 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
5352 dev.children = new_lvs
5353 cfg.Update(instance)
5355 # Step: wait for sync
5357 # this can fail as the old devices are degraded and _WaitForSync
5358 # does a combined result over all disks, so we don't check its
5360 self.proc.LogStep(5, steps_total, "sync devices")
5361 _WaitForSync(self, instance, unlock=True)
5363 # so check manually all the devices
5364 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5365 cfg.SetDiskID(dev, instance.primary_node)
5366 result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5367 msg = result.RemoteFailMsg()
5368 if not msg and not result.payload:
5369 msg = "disk not found"
5371 raise errors.OpExecError("Can't find DRBD device %s: %s" %
5373 if result.payload[5]:
5374 raise errors.OpExecError("DRBD device %s is degraded!" % name)
5376 # Step: remove old storage
5377 self.proc.LogStep(6, steps_total, "removing old storage")
5378 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5379 info("remove logical volumes for %s" % name)
5381 cfg.SetDiskID(lv, tgt_node)
5382 msg = self.rpc.call_blockdev_remove(tgt_node, lv).RemoteFailMsg()
5384 warning("Can't remove old LV: %s" % msg,
5385 hint="manually remove unused LVs")
5388 def _ExecD8Secondary(self, feedback_fn):
5389 """Replace the secondary node for drbd8.
5391 The algorithm for replace is quite complicated:
5392 - for all disks of the instance:
5393 - create new LVs on the new node with same names
5394 - shutdown the drbd device on the old secondary
5395 - disconnect the drbd network on the primary
5396 - create the drbd device on the new secondary
5397 - network attach the drbd on the primary, using an artifice:
5398 the drbd code for Attach() will connect to the network if it
5399 finds a device which is connected to the good local disks but
5401 - wait for sync across all devices
5402 - remove all disks from the old secondary
5404 Failures are not very well handled.
5408 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5409 instance = self.instance
5413 old_node = self.tgt_node
5414 new_node = self.new_node
5415 pri_node = instance.primary_node
5417 old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5418 new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5419 pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5422 # Step: check device activation
5423 self.proc.LogStep(1, steps_total, "check device existence")
5424 info("checking volume groups")
5425 my_vg = cfg.GetVGName()
5426 results = self.rpc.call_vg_list([pri_node, new_node])
5427 for node in pri_node, new_node:
5429 msg = res.RemoteFailMsg()
5431 raise errors.OpExecError("Error checking node %s: %s" % (node, msg))
5432 if my_vg not in res.payload:
5433 raise errors.OpExecError("Volume group '%s' not found on %s" %
5435 for idx, dev in enumerate(instance.disks):
5436 if idx not in self.op.disks:
5438 info("checking disk/%d on %s" % (idx, pri_node))
5439 cfg.SetDiskID(dev, pri_node)
5440 result = self.rpc.call_blockdev_find(pri_node, dev)
5441 msg = result.RemoteFailMsg()
5442 if not msg and not result.payload:
5443 msg = "disk not found"
5445 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5446 (idx, pri_node, msg))
5448 # Step: check other node consistency
5449 self.proc.LogStep(2, steps_total, "check peer consistency")
5450 for idx, dev in enumerate(instance.disks):
5451 if idx not in self.op.disks:
5453 info("checking disk/%d consistency on %s" % (idx, pri_node))
5454 if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5455 raise errors.OpExecError("Primary node (%s) has degraded storage,"
5456 " unsafe to replace the secondary" %
5459 # Step: create new storage
5460 self.proc.LogStep(3, steps_total, "allocate new storage")
5461 for idx, dev in enumerate(instance.disks):
5462 info("adding new local storage on %s for disk/%d" %
5464 # we pass force_create=True to force LVM creation
5465 for new_lv in dev.children:
5466 _CreateBlockDev(self, new_node, instance, new_lv, True,
5467 _GetInstanceInfoText(instance), False)
5469 # Step 4: dbrd minors and drbd setups changes
5470 # after this, we must manually remove the drbd minors on both the
5471 # error and the success paths
5472 minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5474 logging.debug("Allocated minors %s" % (minors,))
5475 self.proc.LogStep(4, steps_total, "changing drbd configuration")
5476 for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5478 info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5479 # create new devices on new_node; note that we create two IDs:
5480 # one without port, so the drbd will be activated without
5481 # networking information on the new node at this stage, and one
5482 # with network, for the latter activation in step 4
5483 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5484 if pri_node == o_node1:
5489 new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5490 new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5492 iv_names[idx] = (dev, dev.children, new_net_id)
5493 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5495 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5496 logical_id=new_alone_id,
5497 children=dev.children)
5499 _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5500 _GetInstanceInfoText(instance), False)
5501 except errors.GenericError:
5502 self.cfg.ReleaseDRBDMinors(instance.name)
5505 for idx, dev in enumerate(instance.disks):
5506 # we have new devices, shutdown the drbd on the old secondary
5507 info("shutting down drbd for disk/%d on old node" % idx)
5508 cfg.SetDiskID(dev, old_node)
5509 msg = self.rpc.call_blockdev_shutdown(old_node, dev).RemoteFailMsg()
5511 warning("Failed to shutdown drbd for disk/%d on old node: %s" %
5513 hint="Please cleanup this device manually as soon as possible")
5515 info("detaching primary drbds from the network (=> standalone)")
5516 result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5517 instance.disks)[pri_node]
5519 msg = result.RemoteFailMsg()
5521 # detaches didn't succeed (unlikely)
5522 self.cfg.ReleaseDRBDMinors(instance.name)
5523 raise errors.OpExecError("Can't detach the disks from the network on"
5524 " old node: %s" % (msg,))
5526 # if we managed to detach at least one, we update all the disks of
5527 # the instance to point to the new secondary
5528 info("updating instance configuration")
5529 for dev, _, new_logical_id in iv_names.itervalues():
5530 dev.logical_id = new_logical_id
5531 cfg.SetDiskID(dev, pri_node)
5532 cfg.Update(instance)
5534 # and now perform the drbd attach
5535 info("attaching primary drbds to new secondary (standalone => connected)")
5536 result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5537 instance.disks, instance.name,
5539 for to_node, to_result in result.items():
5540 msg = to_result.RemoteFailMsg()
5542 warning("can't attach drbd disks on node %s: %s", to_node, msg,
5543 hint="please do a gnt-instance info to see the"
5546 # this can fail as the old devices are degraded and _WaitForSync
5547 # does a combined result over all disks, so we don't check its
5549 self.proc.LogStep(5, steps_total, "sync devices")
5550 _WaitForSync(self, instance, unlock=True)
5552 # so check manually all the devices
5553 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5554 cfg.SetDiskID(dev, pri_node)
5555 result = self.rpc.call_blockdev_find(pri_node, dev)
5556 msg = result.RemoteFailMsg()
5557 if not msg and not result.payload:
5558 msg = "disk not found"
5560 raise errors.OpExecError("Can't find DRBD device disk/%d: %s" %
5562 if result.payload[5]:
5563 raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5565 self.proc.LogStep(6, steps_total, "removing old storage")
5566 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5567 info("remove logical volumes for disk/%d" % idx)
5569 cfg.SetDiskID(lv, old_node)
5570 msg = self.rpc.call_blockdev_remove(old_node, lv).RemoteFailMsg()
5572 warning("Can't remove LV on old secondary: %s", msg,
5573 hint="Cleanup stale volumes by hand")
5575 def Exec(self, feedback_fn):
5576 """Execute disk replacement.
5578 This dispatches the disk replacement to the appropriate handler.
5581 instance = self.instance
5583 # Activate the instance disks if we're replacing them on a down instance
5584 if not instance.admin_up:
5585 _StartInstanceDisks(self, instance, True)
5587 if self.op.mode == constants.REPLACE_DISK_CHG:
5588 fn = self._ExecD8Secondary
5590 fn = self._ExecD8DiskOnly
5592 ret = fn(feedback_fn)
5594 # Deactivate the instance disks if we're replacing them on a down instance
5595 if not instance.admin_up:
5596 _SafeShutdownInstanceDisks(self, instance)
5601 class LUGrowDisk(LogicalUnit):
5602 """Grow a disk of an instance.
5606 HTYPE = constants.HTYPE_INSTANCE
5607 _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5610 def ExpandNames(self):
5611 self._ExpandAndLockInstance()
5612 self.needed_locks[locking.LEVEL_NODE] = []
5613 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5615 def DeclareLocks(self, level):
5616 if level == locking.LEVEL_NODE:
5617 self._LockInstancesNodes()
5619 def BuildHooksEnv(self):
5622 This runs on the master, the primary and all the secondaries.
5626 "DISK": self.op.disk,
5627 "AMOUNT": self.op.amount,
5629 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5631 self.cfg.GetMasterNode(),
5632 self.instance.primary_node,
5636 def CheckPrereq(self):
5637 """Check prerequisites.
5639 This checks that the instance is in the cluster.
5642 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5643 assert instance is not None, \
5644 "Cannot retrieve locked instance %s" % self.op.instance_name
5645 nodenames = list(instance.all_nodes)
5646 for node in nodenames:
5647 _CheckNodeOnline(self, node)
5650 self.instance = instance
5652 if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5653 raise errors.OpPrereqError("Instance's disk layout does not support"
5656 self.disk = instance.FindDisk(self.op.disk)
5658 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5659 instance.hypervisor)
5660 for node in nodenames:
5661 info = nodeinfo[node]
5662 if info.failed or not info.data:
5663 raise errors.OpPrereqError("Cannot get current information"
5664 " from node '%s'" % node)
5665 vg_free = info.data.get('vg_free', None)
5666 if not isinstance(vg_free, int):
5667 raise errors.OpPrereqError("Can't compute free disk space on"
5669 if self.op.amount > vg_free:
5670 raise errors.OpPrereqError("Not enough disk space on target node %s:"
5671 " %d MiB available, %d MiB required" %
5672 (node, vg_free, self.op.amount))
5674 def Exec(self, feedback_fn):
5675 """Execute disk grow.
5678 instance = self.instance
5680 for node in instance.all_nodes:
5681 self.cfg.SetDiskID(disk, node)
5682 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5683 msg = result.RemoteFailMsg()
5685 raise errors.OpExecError("Grow request failed to node %s: %s" %
5687 disk.RecordGrow(self.op.amount)
5688 self.cfg.Update(instance)
5689 if self.op.wait_for_sync:
5690 disk_abort = not _WaitForSync(self, instance)
5692 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5693 " status.\nPlease check the instance.")
5696 class LUQueryInstanceData(NoHooksLU):
5697 """Query runtime instance data.
5700 _OP_REQP = ["instances", "static"]
5703 def ExpandNames(self):
5704 self.needed_locks = {}
5705 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5707 if not isinstance(self.op.instances, list):
5708 raise errors.OpPrereqError("Invalid argument type 'instances'")
5710 if self.op.instances:
5711 self.wanted_names = []
5712 for name in self.op.instances:
5713 full_name = self.cfg.ExpandInstanceName(name)
5714 if full_name is None:
5715 raise errors.OpPrereqError("Instance '%s' not known" % name)
5716 self.wanted_names.append(full_name)
5717 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5719 self.wanted_names = None
5720 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5722 self.needed_locks[locking.LEVEL_NODE] = []
5723 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5725 def DeclareLocks(self, level):
5726 if level == locking.LEVEL_NODE:
5727 self._LockInstancesNodes()
5729 def CheckPrereq(self):
5730 """Check prerequisites.
5732 This only checks the optional instance list against the existing names.
5735 if self.wanted_names is None:
5736 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5738 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5739 in self.wanted_names]
5742 def _ComputeDiskStatus(self, instance, snode, dev):
5743 """Compute block device status.
5746 static = self.op.static
5748 self.cfg.SetDiskID(dev, instance.primary_node)
5749 dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5750 if dev_pstatus.offline:
5753 msg = dev_pstatus.RemoteFailMsg()
5755 raise errors.OpExecError("Can't compute disk status for %s: %s" %
5756 (instance.name, msg))
5757 dev_pstatus = dev_pstatus.payload
5761 if dev.dev_type in constants.LDS_DRBD:
5762 # we change the snode then (otherwise we use the one passed in)
5763 if dev.logical_id[0] == instance.primary_node:
5764 snode = dev.logical_id[1]
5766 snode = dev.logical_id[0]
5768 if snode and not static:
5769 self.cfg.SetDiskID(dev, snode)
5770 dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5771 if dev_sstatus.offline:
5774 msg = dev_sstatus.RemoteFailMsg()
5776 raise errors.OpExecError("Can't compute disk status for %s: %s" %
5777 (instance.name, msg))
5778 dev_sstatus = dev_sstatus.payload
5783 dev_children = [self._ComputeDiskStatus(instance, snode, child)
5784 for child in dev.children]
5789 "iv_name": dev.iv_name,
5790 "dev_type": dev.dev_type,
5791 "logical_id": dev.logical_id,
5792 "physical_id": dev.physical_id,
5793 "pstatus": dev_pstatus,
5794 "sstatus": dev_sstatus,
5795 "children": dev_children,
5801 def Exec(self, feedback_fn):
5802 """Gather and return data"""
5805 cluster = self.cfg.GetClusterInfo()
5807 for instance in self.wanted_instances:
5808 if not self.op.static:
5809 remote_info = self.rpc.call_instance_info(instance.primary_node,
5811 instance.hypervisor)
5813 remote_info = remote_info.data
5814 if remote_info and "state" in remote_info:
5817 remote_state = "down"
5820 if instance.admin_up:
5823 config_state = "down"
5825 disks = [self._ComputeDiskStatus(instance, None, device)
5826 for device in instance.disks]
5829 "name": instance.name,
5830 "config_state": config_state,
5831 "run_state": remote_state,
5832 "pnode": instance.primary_node,
5833 "snodes": instance.secondary_nodes,
5835 "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5837 "hypervisor": instance.hypervisor,
5838 "network_port": instance.network_port,
5839 "hv_instance": instance.hvparams,
5840 "hv_actual": cluster.FillHV(instance),
5841 "be_instance": instance.beparams,
5842 "be_actual": cluster.FillBE(instance),
5845 result[instance.name] = idict
5850 class LUSetInstanceParams(LogicalUnit):
5851 """Modifies an instances's parameters.
5854 HPATH = "instance-modify"
5855 HTYPE = constants.HTYPE_INSTANCE
5856 _OP_REQP = ["instance_name"]
5859 def CheckArguments(self):
5860 if not hasattr(self.op, 'nics'):
5862 if not hasattr(self.op, 'disks'):
5864 if not hasattr(self.op, 'beparams'):
5865 self.op.beparams = {}
5866 if not hasattr(self.op, 'hvparams'):
5867 self.op.hvparams = {}
5868 self.op.force = getattr(self.op, "force", False)
5869 if not (self.op.nics or self.op.disks or
5870 self.op.hvparams or self.op.beparams):
5871 raise errors.OpPrereqError("No changes submitted")
5875 for disk_op, disk_dict in self.op.disks:
5876 if disk_op == constants.DDM_REMOVE:
5879 elif disk_op == constants.DDM_ADD:
5882 if not isinstance(disk_op, int):
5883 raise errors.OpPrereqError("Invalid disk index")
5884 if disk_op == constants.DDM_ADD:
5885 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5886 if mode not in constants.DISK_ACCESS_SET:
5887 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5888 size = disk_dict.get('size', None)
5890 raise errors.OpPrereqError("Required disk parameter size missing")
5893 except ValueError, err:
5894 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5896 disk_dict['size'] = size
5898 # modification of disk
5899 if 'size' in disk_dict:
5900 raise errors.OpPrereqError("Disk size change not possible, use"
5903 if disk_addremove > 1:
5904 raise errors.OpPrereqError("Only one disk add or remove operation"
5905 " supported at a time")
5909 for nic_op, nic_dict in self.op.nics:
5910 if nic_op == constants.DDM_REMOVE:
5913 elif nic_op == constants.DDM_ADD:
5916 if not isinstance(nic_op, int):
5917 raise errors.OpPrereqError("Invalid nic index")
5919 # nic_dict should be a dict
5920 nic_ip = nic_dict.get('ip', None)
5921 if nic_ip is not None:
5922 if nic_ip.lower() == constants.VALUE_NONE:
5923 nic_dict['ip'] = None
5925 if not utils.IsValidIP(nic_ip):
5926 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5928 nic_bridge = nic_dict.get('bridge', None)
5929 nic_link = nic_dict.get('link', None)
5930 if nic_bridge and nic_link:
5931 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
5932 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
5933 nic_dict['bridge'] = None
5934 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
5935 nic_dict['link'] = None
5937 if nic_op == constants.DDM_ADD:
5938 nic_mac = nic_dict.get('mac', None)
5940 nic_dict['mac'] = constants.VALUE_AUTO
5942 if 'mac' in nic_dict:
5943 nic_mac = nic_dict['mac']
5944 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5945 if not utils.IsValidMac(nic_mac):
5946 raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5947 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
5948 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
5949 " modifying an existing nic")
5951 if nic_addremove > 1:
5952 raise errors.OpPrereqError("Only one NIC add or remove operation"
5953 " supported at a time")
5955 def ExpandNames(self):
5956 self._ExpandAndLockInstance()
5957 self.needed_locks[locking.LEVEL_NODE] = []
5958 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5960 def DeclareLocks(self, level):
5961 if level == locking.LEVEL_NODE:
5962 self._LockInstancesNodes()
5964 def BuildHooksEnv(self):
5967 This runs on the master, primary and secondaries.
5971 if constants.BE_MEMORY in self.be_new:
5972 args['memory'] = self.be_new[constants.BE_MEMORY]
5973 if constants.BE_VCPUS in self.be_new:
5974 args['vcpus'] = self.be_new[constants.BE_VCPUS]
5975 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
5976 # information at all.
5979 nic_override = dict(self.op.nics)
5980 c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
5981 for idx, nic in enumerate(self.instance.nics):
5982 if idx in nic_override:
5983 this_nic_override = nic_override[idx]
5985 this_nic_override = {}
5986 if 'ip' in this_nic_override:
5987 ip = this_nic_override['ip']
5990 if 'mac' in this_nic_override:
5991 mac = this_nic_override['mac']
5994 if idx in self.nic_pnew:
5995 nicparams = self.nic_pnew[idx]
5997 nicparams = objects.FillDict(c_nicparams, nic.nicparams)
5998 mode = nicparams[constants.NIC_MODE]
5999 link = nicparams[constants.NIC_LINK]
6000 args['nics'].append((ip, mac, mode, link))
6001 if constants.DDM_ADD in nic_override:
6002 ip = nic_override[constants.DDM_ADD].get('ip', None)
6003 mac = nic_override[constants.DDM_ADD]['mac']
6004 nicparams = self.nic_pnew[constants.DDM_ADD]
6005 mode = nicparams[constants.NIC_MODE]
6006 link = nicparams[constants.NIC_LINK]
6007 args['nics'].append((ip, mac, mode, link))
6008 elif constants.DDM_REMOVE in nic_override:
6009 del args['nics'][-1]
6011 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
6012 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6015 def _GetUpdatedParams(self, old_params, update_dict,
6016 default_values, parameter_types):
6017 """Return the new params dict for the given params.
6019 @type old_params: dict
6020 @type old_params: old parameters
6021 @type update_dict: dict
6022 @type update_dict: dict containing new parameter values,
6023 or constants.VALUE_DEFAULT to reset the
6024 parameter to its default value
6025 @type default_values: dict
6026 @param default_values: default values for the filled parameters
6027 @type parameter_types: dict
6028 @param parameter_types: dict mapping target dict keys to types
6029 in constants.ENFORCEABLE_TYPES
6030 @rtype: (dict, dict)
6031 @return: (new_parameters, filled_parameters)
6034 params_copy = copy.deepcopy(old_params)
6035 for key, val in update_dict.iteritems():
6036 if val == constants.VALUE_DEFAULT:
6038 del params_copy[key]
6042 params_copy[key] = val
6043 utils.ForceDictType(params_copy, parameter_types)
6044 params_filled = objects.FillDict(default_values, params_copy)
6045 return (params_copy, params_filled)
6047 def CheckPrereq(self):
6048 """Check prerequisites.
6050 This only checks the instance list against the existing names.
6053 force = self.force = self.op.force
6055 # checking the new params on the primary/secondary nodes
6057 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6058 cluster = self.cluster = self.cfg.GetClusterInfo()
6059 assert self.instance is not None, \
6060 "Cannot retrieve locked instance %s" % self.op.instance_name
6061 pnode = instance.primary_node
6062 nodelist = list(instance.all_nodes)
6064 # hvparams processing
6065 if self.op.hvparams:
6066 i_hvdict, hv_new = self._GetUpdatedParams(
6067 instance.hvparams, self.op.hvparams,
6068 cluster.hvparams[instance.hypervisor],
6069 constants.HVS_PARAMETER_TYPES)
6071 hypervisor.GetHypervisor(
6072 instance.hypervisor).CheckParameterSyntax(hv_new)
6073 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
6074 self.hv_new = hv_new # the new actual values
6075 self.hv_inst = i_hvdict # the new dict (without defaults)
6077 self.hv_new = self.hv_inst = {}
6079 # beparams processing
6080 if self.op.beparams:
6081 i_bedict, be_new = self._GetUpdatedParams(
6082 instance.beparams, self.op.beparams,
6083 cluster.beparams[constants.PP_DEFAULT],
6084 constants.BES_PARAMETER_TYPES)
6085 self.be_new = be_new # the new actual values
6086 self.be_inst = i_bedict # the new dict (without defaults)
6088 self.be_new = self.be_inst = {}
6092 if constants.BE_MEMORY in self.op.beparams and not self.force:
6093 mem_check_list = [pnode]
6094 if be_new[constants.BE_AUTO_BALANCE]:
6095 # either we changed auto_balance to yes or it was from before
6096 mem_check_list.extend(instance.secondary_nodes)
6097 instance_info = self.rpc.call_instance_info(pnode, instance.name,
6098 instance.hypervisor)
6099 nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
6100 instance.hypervisor)
6101 if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
6102 # Assume the primary node is unreachable and go ahead
6103 self.warn.append("Can't get info from primary node %s" % pnode)
6105 if not instance_info.failed and instance_info.data:
6106 current_mem = int(instance_info.data['memory'])
6108 # Assume instance not running
6109 # (there is a slight race condition here, but it's not very probable,
6110 # and we have no other way to check)
6112 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
6113 nodeinfo[pnode].data['memory_free'])
6115 raise errors.OpPrereqError("This change will prevent the instance"
6116 " from starting, due to %d MB of memory"
6117 " missing on its primary node" % miss_mem)
6119 if be_new[constants.BE_AUTO_BALANCE]:
6120 for node, nres in nodeinfo.iteritems():
6121 if node not in instance.secondary_nodes:
6123 if nres.failed or not isinstance(nres.data, dict):
6124 self.warn.append("Can't get info from secondary node %s" % node)
6125 elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
6126 self.warn.append("Not enough memory to failover instance to"
6127 " secondary node %s" % node)
6132 for nic_op, nic_dict in self.op.nics:
6133 if nic_op == constants.DDM_REMOVE:
6134 if not instance.nics:
6135 raise errors.OpPrereqError("Instance has no NICs, cannot remove")
6137 if nic_op != constants.DDM_ADD:
6139 if nic_op < 0 or nic_op >= len(instance.nics):
6140 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
6142 (nic_op, len(instance.nics)))
6143 old_nic_params = instance.nics[nic_op].nicparams
6144 old_nic_ip = instance.nics[nic_op].ip
6149 update_params_dict = dict([(key, nic_dict[key])
6150 for key in constants.NICS_PARAMETERS
6151 if key in nic_dict])
6153 if 'bridge' in nic_dict:
6154 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
6156 new_nic_params, new_filled_nic_params = \
6157 self._GetUpdatedParams(old_nic_params, update_params_dict,
6158 cluster.nicparams[constants.PP_DEFAULT],
6159 constants.NICS_PARAMETER_TYPES)
6160 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
6161 self.nic_pinst[nic_op] = new_nic_params
6162 self.nic_pnew[nic_op] = new_filled_nic_params
6163 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
6165 if new_nic_mode == constants.NIC_MODE_BRIDGED:
6166 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
6167 result = self.rpc.call_bridges_exist(pnode, [nic_bridge])
6168 msg = result.RemoteFailMsg()
6170 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
6172 self.warn.append(msg)
6174 raise errors.OpPrereqError(msg)
6175 if new_nic_mode == constants.NIC_MODE_ROUTED:
6176 if 'ip' in nic_dict:
6177 nic_ip = nic_dict['ip']
6181 raise errors.OpPrereqError('Cannot set the nic ip to None'
6183 if 'mac' in nic_dict:
6184 nic_mac = nic_dict['mac']
6186 raise errors.OpPrereqError('Cannot set the nic mac to None')
6187 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6188 # otherwise generate the mac
6189 nic_dict['mac'] = self.cfg.GenerateMAC()
6191 # or validate/reserve the current one
6192 if self.cfg.IsMacInUse(nic_mac):
6193 raise errors.OpPrereqError("MAC address %s already in use"
6194 " in cluster" % nic_mac)
6197 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
6198 raise errors.OpPrereqError("Disk operations not supported for"
6199 " diskless instances")
6200 for disk_op, disk_dict in self.op.disks:
6201 if disk_op == constants.DDM_REMOVE:
6202 if len(instance.disks) == 1:
6203 raise errors.OpPrereqError("Cannot remove the last disk of"
6205 ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
6206 ins_l = ins_l[pnode]
6207 if ins_l.failed or not isinstance(ins_l.data, list):
6208 raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
6209 if instance.name in ins_l.data:
6210 raise errors.OpPrereqError("Instance is running, can't remove"
6213 if (disk_op == constants.DDM_ADD and
6214 len(instance.nics) >= constants.MAX_DISKS):
6215 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
6216 " add more" % constants.MAX_DISKS)
6217 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
6219 if disk_op < 0 or disk_op >= len(instance.disks):
6220 raise errors.OpPrereqError("Invalid disk index %s, valid values"
6222 (disk_op, len(instance.disks)))
6226 def Exec(self, feedback_fn):
6227 """Modifies an instance.
6229 All parameters take effect only at the next restart of the instance.
6232 # Process here the warnings from CheckPrereq, as we don't have a
6233 # feedback_fn there.
6234 for warn in self.warn:
6235 feedback_fn("WARNING: %s" % warn)
6238 instance = self.instance
6239 cluster = self.cluster
6241 for disk_op, disk_dict in self.op.disks:
6242 if disk_op == constants.DDM_REMOVE:
6243 # remove the last disk
6244 device = instance.disks.pop()
6245 device_idx = len(instance.disks)
6246 for node, disk in device.ComputeNodeTree(instance.primary_node):
6247 self.cfg.SetDiskID(disk, node)
6248 msg = self.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
6250 self.LogWarning("Could not remove disk/%d on node %s: %s,"
6251 " continuing anyway", device_idx, node, msg)
6252 result.append(("disk/%d" % device_idx, "remove"))
6253 elif disk_op == constants.DDM_ADD:
6255 if instance.disk_template == constants.DT_FILE:
6256 file_driver, file_path = instance.disks[0].logical_id
6257 file_path = os.path.dirname(file_path)
6259 file_driver = file_path = None
6260 disk_idx_base = len(instance.disks)
6261 new_disk = _GenerateDiskTemplate(self,
6262 instance.disk_template,
6263 instance.name, instance.primary_node,
6264 instance.secondary_nodes,
6269 instance.disks.append(new_disk)
6270 info = _GetInstanceInfoText(instance)
6272 logging.info("Creating volume %s for instance %s",
6273 new_disk.iv_name, instance.name)
6274 # Note: this needs to be kept in sync with _CreateDisks
6276 for node in instance.all_nodes:
6277 f_create = node == instance.primary_node
6279 _CreateBlockDev(self, node, instance, new_disk,
6280 f_create, info, f_create)
6281 except errors.OpExecError, err:
6282 self.LogWarning("Failed to create volume %s (%s) on"
6284 new_disk.iv_name, new_disk, node, err)
6285 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
6286 (new_disk.size, new_disk.mode)))
6288 # change a given disk
6289 instance.disks[disk_op].mode = disk_dict['mode']
6290 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
6292 for nic_op, nic_dict in self.op.nics:
6293 if nic_op == constants.DDM_REMOVE:
6294 # remove the last nic
6295 del instance.nics[-1]
6296 result.append(("nic.%d" % len(instance.nics), "remove"))
6297 elif nic_op == constants.DDM_ADD:
6298 # mac and bridge should be set, by now
6299 mac = nic_dict['mac']
6300 ip = nic_dict.get('ip', None)
6301 nicparams = self.nic_pinst[constants.DDM_ADD]
6302 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
6303 instance.nics.append(new_nic)
6304 result.append(("nic.%d" % (len(instance.nics) - 1),
6305 "add:mac=%s,ip=%s,mode=%s,link=%s" %
6306 (new_nic.mac, new_nic.ip,
6307 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
6308 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
6311 for key in 'mac', 'ip':
6313 setattr(instance.nics[nic_op], key, nic_dict[key])
6314 if nic_op in self.nic_pnew:
6315 instance.nics[nic_op].nicparams = self.nic_pnew[nic_op]
6316 for key, val in nic_dict.iteritems():
6317 result.append(("nic.%s/%d" % (key, nic_op), val))
6320 if self.op.hvparams:
6321 instance.hvparams = self.hv_inst
6322 for key, val in self.op.hvparams.iteritems():
6323 result.append(("hv/%s" % key, val))
6326 if self.op.beparams:
6327 instance.beparams = self.be_inst
6328 for key, val in self.op.beparams.iteritems():
6329 result.append(("be/%s" % key, val))
6331 self.cfg.Update(instance)
6336 class LUQueryExports(NoHooksLU):
6337 """Query the exports list
6340 _OP_REQP = ['nodes']
6343 def ExpandNames(self):
6344 self.needed_locks = {}
6345 self.share_locks[locking.LEVEL_NODE] = 1
6346 if not self.op.nodes:
6347 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6349 self.needed_locks[locking.LEVEL_NODE] = \
6350 _GetWantedNodes(self, self.op.nodes)
6352 def CheckPrereq(self):
6353 """Check prerequisites.
6356 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
6358 def Exec(self, feedback_fn):
6359 """Compute the list of all the exported system images.
6362 @return: a dictionary with the structure node->(export-list)
6363 where export-list is a list of the instances exported on
6367 rpcresult = self.rpc.call_export_list(self.nodes)
6369 for node in rpcresult:
6370 if rpcresult[node].RemoteFailMsg():
6371 result[node] = False
6373 result[node] = rpcresult[node].payload
6378 class LUExportInstance(LogicalUnit):
6379 """Export an instance to an image in the cluster.
6382 HPATH = "instance-export"
6383 HTYPE = constants.HTYPE_INSTANCE
6384 _OP_REQP = ["instance_name", "target_node", "shutdown"]
6387 def ExpandNames(self):
6388 self._ExpandAndLockInstance()
6389 # FIXME: lock only instance primary and destination node
6391 # Sad but true, for now we have do lock all nodes, as we don't know where
6392 # the previous export might be, and and in this LU we search for it and
6393 # remove it from its current node. In the future we could fix this by:
6394 # - making a tasklet to search (share-lock all), then create the new one,
6395 # then one to remove, after
6396 # - removing the removal operation altoghether
6397 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6399 def DeclareLocks(self, level):
6400 """Last minute lock declaration."""
6401 # All nodes are locked anyway, so nothing to do here.
6403 def BuildHooksEnv(self):
6406 This will run on the master, primary node and target node.
6410 "EXPORT_NODE": self.op.target_node,
6411 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
6413 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6414 nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
6415 self.op.target_node]
6418 def CheckPrereq(self):
6419 """Check prerequisites.
6421 This checks that the instance and node names are valid.
6424 instance_name = self.op.instance_name
6425 self.instance = self.cfg.GetInstanceInfo(instance_name)
6426 assert self.instance is not None, \
6427 "Cannot retrieve locked instance %s" % self.op.instance_name
6428 _CheckNodeOnline(self, self.instance.primary_node)
6430 self.dst_node = self.cfg.GetNodeInfo(
6431 self.cfg.ExpandNodeName(self.op.target_node))
6433 if self.dst_node is None:
6434 # This is wrong node name, not a non-locked node
6435 raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
6436 _CheckNodeOnline(self, self.dst_node.name)
6437 _CheckNodeNotDrained(self, self.dst_node.name)
6439 # instance disk type verification
6440 for disk in self.instance.disks:
6441 if disk.dev_type == constants.LD_FILE:
6442 raise errors.OpPrereqError("Export not supported for instances with"
6443 " file-based disks")
6445 def Exec(self, feedback_fn):
6446 """Export an instance to an image in the cluster.
6449 instance = self.instance
6450 dst_node = self.dst_node
6451 src_node = instance.primary_node
6452 if self.op.shutdown:
6453 # shutdown the instance, but not the disks
6454 result = self.rpc.call_instance_shutdown(src_node, instance)
6455 msg = result.RemoteFailMsg()
6457 raise errors.OpExecError("Could not shutdown instance %s on"
6459 (instance.name, src_node, msg))
6461 vgname = self.cfg.GetVGName()
6465 # set the disks ID correctly since call_instance_start needs the
6466 # correct drbd minor to create the symlinks
6467 for disk in instance.disks:
6468 self.cfg.SetDiskID(disk, src_node)
6471 for disk in instance.disks:
6472 # result.payload will be a snapshot of an lvm leaf of the one we passed
6473 result = self.rpc.call_blockdev_snapshot(src_node, disk)
6474 msg = result.RemoteFailMsg()
6476 self.LogWarning("Could not snapshot block device %s on node %s: %s",
6477 disk.logical_id[1], src_node, msg)
6478 snap_disks.append(False)
6480 disk_id = (vgname, result.payload)
6481 new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
6482 logical_id=disk_id, physical_id=disk_id,
6483 iv_name=disk.iv_name)
6484 snap_disks.append(new_dev)
6487 if self.op.shutdown and instance.admin_up:
6488 result = self.rpc.call_instance_start(src_node, instance, None, None)
6489 msg = result.RemoteFailMsg()
6491 _ShutdownInstanceDisks(self, instance)
6492 raise errors.OpExecError("Could not start instance: %s" % msg)
6494 # TODO: check for size
6496 cluster_name = self.cfg.GetClusterName()
6497 for idx, dev in enumerate(snap_disks):
6499 result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6500 instance, cluster_name, idx)
6501 msg = result.RemoteFailMsg()
6503 self.LogWarning("Could not export block device %s from node %s to"
6504 " node %s: %s", dev.logical_id[1], src_node,
6506 msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
6508 self.LogWarning("Could not remove snapshot block device %s from node"
6509 " %s: %s", dev.logical_id[1], src_node, msg)
6511 result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6512 msg = result.RemoteFailMsg()
6514 self.LogWarning("Could not finalize export for instance %s"
6515 " on node %s: %s", instance.name, dst_node.name, msg)
6517 nodelist = self.cfg.GetNodeList()
6518 nodelist.remove(dst_node.name)
6520 # on one-node clusters nodelist will be empty after the removal
6521 # if we proceed the backup would be removed because OpQueryExports
6522 # substitutes an empty list with the full cluster node list.
6523 iname = instance.name
6525 exportlist = self.rpc.call_export_list(nodelist)
6526 for node in exportlist:
6527 if exportlist[node].RemoteFailMsg():
6529 if iname in exportlist[node].payload:
6530 msg = self.rpc.call_export_remove(node, iname).RemoteFailMsg()
6532 self.LogWarning("Could not remove older export for instance %s"
6533 " on node %s: %s", iname, node, msg)
6536 class LURemoveExport(NoHooksLU):
6537 """Remove exports related to the named instance.
6540 _OP_REQP = ["instance_name"]
6543 def ExpandNames(self):
6544 self.needed_locks = {}
6545 # We need all nodes to be locked in order for RemoveExport to work, but we
6546 # don't need to lock the instance itself, as nothing will happen to it (and
6547 # we can remove exports also for a removed instance)
6548 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6550 def CheckPrereq(self):
6551 """Check prerequisites.
6555 def Exec(self, feedback_fn):
6556 """Remove any export.
6559 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6560 # If the instance was not found we'll try with the name that was passed in.
6561 # This will only work if it was an FQDN, though.
6563 if not instance_name:
6565 instance_name = self.op.instance_name
6567 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6568 exportlist = self.rpc.call_export_list(locked_nodes)
6570 for node in exportlist:
6571 msg = exportlist[node].RemoteFailMsg()
6573 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
6575 if instance_name in exportlist[node].payload:
6577 result = self.rpc.call_export_remove(node, instance_name)
6578 msg = result.RemoteFailMsg()
6580 logging.error("Could not remove export for instance %s"
6581 " on node %s: %s", instance_name, node, msg)
6583 if fqdn_warn and not found:
6584 feedback_fn("Export not found. If trying to remove an export belonging"
6585 " to a deleted instance please use its Fully Qualified"
6589 class TagsLU(NoHooksLU):
6592 This is an abstract class which is the parent of all the other tags LUs.
6596 def ExpandNames(self):
6597 self.needed_locks = {}
6598 if self.op.kind == constants.TAG_NODE:
6599 name = self.cfg.ExpandNodeName(self.op.name)
6601 raise errors.OpPrereqError("Invalid node name (%s)" %
6604 self.needed_locks[locking.LEVEL_NODE] = name
6605 elif self.op.kind == constants.TAG_INSTANCE:
6606 name = self.cfg.ExpandInstanceName(self.op.name)
6608 raise errors.OpPrereqError("Invalid instance name (%s)" %
6611 self.needed_locks[locking.LEVEL_INSTANCE] = name
6613 def CheckPrereq(self):
6614 """Check prerequisites.
6617 if self.op.kind == constants.TAG_CLUSTER:
6618 self.target = self.cfg.GetClusterInfo()
6619 elif self.op.kind == constants.TAG_NODE:
6620 self.target = self.cfg.GetNodeInfo(self.op.name)
6621 elif self.op.kind == constants.TAG_INSTANCE:
6622 self.target = self.cfg.GetInstanceInfo(self.op.name)
6624 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6628 class LUGetTags(TagsLU):
6629 """Returns the tags of a given object.
6632 _OP_REQP = ["kind", "name"]
6635 def Exec(self, feedback_fn):
6636 """Returns the tag list.
6639 return list(self.target.GetTags())
6642 class LUSearchTags(NoHooksLU):
6643 """Searches the tags for a given pattern.
6646 _OP_REQP = ["pattern"]
6649 def ExpandNames(self):
6650 self.needed_locks = {}
6652 def CheckPrereq(self):
6653 """Check prerequisites.
6655 This checks the pattern passed for validity by compiling it.
6659 self.re = re.compile(self.op.pattern)
6660 except re.error, err:
6661 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6662 (self.op.pattern, err))
6664 def Exec(self, feedback_fn):
6665 """Returns the tag list.
6669 tgts = [("/cluster", cfg.GetClusterInfo())]
6670 ilist = cfg.GetAllInstancesInfo().values()
6671 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6672 nlist = cfg.GetAllNodesInfo().values()
6673 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6675 for path, target in tgts:
6676 for tag in target.GetTags():
6677 if self.re.search(tag):
6678 results.append((path, tag))
6682 class LUAddTags(TagsLU):
6683 """Sets a tag on a given object.
6686 _OP_REQP = ["kind", "name", "tags"]
6689 def CheckPrereq(self):
6690 """Check prerequisites.
6692 This checks the type and length of the tag name and value.
6695 TagsLU.CheckPrereq(self)
6696 for tag in self.op.tags:
6697 objects.TaggableObject.ValidateTag(tag)
6699 def Exec(self, feedback_fn):
6704 for tag in self.op.tags:
6705 self.target.AddTag(tag)
6706 except errors.TagError, err:
6707 raise errors.OpExecError("Error while setting tag: %s" % str(err))
6709 self.cfg.Update(self.target)
6710 except errors.ConfigurationError:
6711 raise errors.OpRetryError("There has been a modification to the"
6712 " config file and the operation has been"
6713 " aborted. Please retry.")
6716 class LUDelTags(TagsLU):
6717 """Delete a list of tags from a given object.
6720 _OP_REQP = ["kind", "name", "tags"]
6723 def CheckPrereq(self):
6724 """Check prerequisites.
6726 This checks that we have the given tag.
6729 TagsLU.CheckPrereq(self)
6730 for tag in self.op.tags:
6731 objects.TaggableObject.ValidateTag(tag)
6732 del_tags = frozenset(self.op.tags)
6733 cur_tags = self.target.GetTags()
6734 if not del_tags <= cur_tags:
6735 diff_tags = del_tags - cur_tags
6736 diff_names = ["'%s'" % tag for tag in diff_tags]
6738 raise errors.OpPrereqError("Tag(s) %s not found" %
6739 (",".join(diff_names)))
6741 def Exec(self, feedback_fn):
6742 """Remove the tag from the object.
6745 for tag in self.op.tags:
6746 self.target.RemoveTag(tag)
6748 self.cfg.Update(self.target)
6749 except errors.ConfigurationError:
6750 raise errors.OpRetryError("There has been a modification to the"
6751 " config file and the operation has been"
6752 " aborted. Please retry.")
6755 class LUTestDelay(NoHooksLU):
6756 """Sleep for a specified amount of time.
6758 This LU sleeps on the master and/or nodes for a specified amount of
6762 _OP_REQP = ["duration", "on_master", "on_nodes"]
6765 def ExpandNames(self):
6766 """Expand names and set required locks.
6768 This expands the node list, if any.
6771 self.needed_locks = {}
6772 if self.op.on_nodes:
6773 # _GetWantedNodes can be used here, but is not always appropriate to use
6774 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6776 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6777 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6779 def CheckPrereq(self):
6780 """Check prerequisites.
6784 def Exec(self, feedback_fn):
6785 """Do the actual sleep.
6788 if self.op.on_master:
6789 if not utils.TestDelay(self.op.duration):
6790 raise errors.OpExecError("Error during master delay test")
6791 if self.op.on_nodes:
6792 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6794 raise errors.OpExecError("Complete failure from rpc call")
6795 for node, node_result in result.items():
6797 if not node_result.data:
6798 raise errors.OpExecError("Failure during rpc call to node %s,"
6799 " result: %s" % (node, node_result.data))
6802 class IAllocator(object):
6803 """IAllocator framework.
6805 An IAllocator instance has three sets of attributes:
6806 - cfg that is needed to query the cluster
6807 - input data (all members of the _KEYS class attribute are required)
6808 - four buffer attributes (in|out_data|text), that represent the
6809 input (to the external script) in text and data structure format,
6810 and the output from it, again in two formats
6811 - the result variables from the script (success, info, nodes) for
6816 "mem_size", "disks", "disk_template",
6817 "os", "tags", "nics", "vcpus", "hypervisor",
6823 def __init__(self, lu, mode, name, **kwargs):
6825 # init buffer variables
6826 self.in_text = self.out_text = self.in_data = self.out_data = None
6827 # init all input fields so that pylint is happy
6830 self.mem_size = self.disks = self.disk_template = None
6831 self.os = self.tags = self.nics = self.vcpus = None
6832 self.hypervisor = None
6833 self.relocate_from = None
6835 self.required_nodes = None
6836 # init result fields
6837 self.success = self.info = self.nodes = None
6838 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6839 keyset = self._ALLO_KEYS
6840 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6841 keyset = self._RELO_KEYS
6843 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6844 " IAllocator" % self.mode)
6846 if key not in keyset:
6847 raise errors.ProgrammerError("Invalid input parameter '%s' to"
6848 " IAllocator" % key)
6849 setattr(self, key, kwargs[key])
6851 if key not in kwargs:
6852 raise errors.ProgrammerError("Missing input parameter '%s' to"
6853 " IAllocator" % key)
6854 self._BuildInputData()
6856 def _ComputeClusterData(self):
6857 """Compute the generic allocator input data.
6859 This is the data that is independent of the actual operation.
6863 cluster_info = cfg.GetClusterInfo()
6866 "version": constants.IALLOCATOR_VERSION,
6867 "cluster_name": cfg.GetClusterName(),
6868 "cluster_tags": list(cluster_info.GetTags()),
6869 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
6870 # we don't have job IDs
6872 iinfo = cfg.GetAllInstancesInfo().values()
6873 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6877 node_list = cfg.GetNodeList()
6879 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6880 hypervisor_name = self.hypervisor
6881 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6882 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6884 node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6886 node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6887 cluster_info.enabled_hypervisors)
6888 for nname, nresult in node_data.items():
6889 # first fill in static (config-based) values
6890 ninfo = cfg.GetNodeInfo(nname)
6892 "tags": list(ninfo.GetTags()),
6893 "primary_ip": ninfo.primary_ip,
6894 "secondary_ip": ninfo.secondary_ip,
6895 "offline": ninfo.offline,
6896 "drained": ninfo.drained,
6897 "master_candidate": ninfo.master_candidate,
6900 if not ninfo.offline:
6902 if not isinstance(nresult.data, dict):
6903 raise errors.OpExecError("Can't get data for node %s" % nname)
6904 remote_info = nresult.data
6905 for attr in ['memory_total', 'memory_free', 'memory_dom0',
6906 'vg_size', 'vg_free', 'cpu_total']:
6907 if attr not in remote_info:
6908 raise errors.OpExecError("Node '%s' didn't return attribute"
6909 " '%s'" % (nname, attr))
6911 remote_info[attr] = int(remote_info[attr])
6912 except ValueError, err:
6913 raise errors.OpExecError("Node '%s' returned invalid value"
6914 " for '%s': %s" % (nname, attr, err))
6915 # compute memory used by primary instances
6916 i_p_mem = i_p_up_mem = 0
6917 for iinfo, beinfo in i_list:
6918 if iinfo.primary_node == nname:
6919 i_p_mem += beinfo[constants.BE_MEMORY]
6920 if iinfo.name not in node_iinfo[nname].data:
6923 i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
6924 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6925 remote_info['memory_free'] -= max(0, i_mem_diff)
6928 i_p_up_mem += beinfo[constants.BE_MEMORY]
6930 # compute memory used by instances
6932 "total_memory": remote_info['memory_total'],
6933 "reserved_memory": remote_info['memory_dom0'],
6934 "free_memory": remote_info['memory_free'],
6935 "total_disk": remote_info['vg_size'],
6936 "free_disk": remote_info['vg_free'],
6937 "total_cpus": remote_info['cpu_total'],
6938 "i_pri_memory": i_p_mem,
6939 "i_pri_up_memory": i_p_up_mem,
6943 node_results[nname] = pnr
6944 data["nodes"] = node_results
6948 for iinfo, beinfo in i_list:
6950 for nic in iinfo.nics:
6951 filled_params = objects.FillDict(
6952 cluster_info.nicparams[constants.PP_DEFAULT],
6954 nic_dict = {"mac": nic.mac,
6956 "mode": filled_params[constants.NIC_MODE],
6957 "link": filled_params[constants.NIC_LINK],
6959 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
6960 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
6961 nic_data.append(nic_dict)
6963 "tags": list(iinfo.GetTags()),
6964 "admin_up": iinfo.admin_up,
6965 "vcpus": beinfo[constants.BE_VCPUS],
6966 "memory": beinfo[constants.BE_MEMORY],
6968 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
6970 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
6971 "disk_template": iinfo.disk_template,
6972 "hypervisor": iinfo.hypervisor,
6974 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
6976 instance_data[iinfo.name] = pir
6978 data["instances"] = instance_data
6982 def _AddNewInstance(self):
6983 """Add new instance data to allocator structure.
6985 This in combination with _AllocatorGetClusterData will create the
6986 correct structure needed as input for the allocator.
6988 The checks for the completeness of the opcode must have already been
6994 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6996 if self.disk_template in constants.DTS_NET_MIRROR:
6997 self.required_nodes = 2
6999 self.required_nodes = 1
7003 "disk_template": self.disk_template,
7006 "vcpus": self.vcpus,
7007 "memory": self.mem_size,
7008 "disks": self.disks,
7009 "disk_space_total": disk_space,
7011 "required_nodes": self.required_nodes,
7013 data["request"] = request
7015 def _AddRelocateInstance(self):
7016 """Add relocate instance data to allocator structure.
7018 This in combination with _IAllocatorGetClusterData will create the
7019 correct structure needed as input for the allocator.
7021 The checks for the completeness of the opcode must have already been
7025 instance = self.lu.cfg.GetInstanceInfo(self.name)
7026 if instance is None:
7027 raise errors.ProgrammerError("Unknown instance '%s' passed to"
7028 " IAllocator" % self.name)
7030 if instance.disk_template not in constants.DTS_NET_MIRROR:
7031 raise errors.OpPrereqError("Can't relocate non-mirrored instances")
7033 if len(instance.secondary_nodes) != 1:
7034 raise errors.OpPrereqError("Instance has not exactly one secondary node")
7036 self.required_nodes = 1
7037 disk_sizes = [{'size': disk.size} for disk in instance.disks]
7038 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
7043 "disk_space_total": disk_space,
7044 "required_nodes": self.required_nodes,
7045 "relocate_from": self.relocate_from,
7047 self.in_data["request"] = request
7049 def _BuildInputData(self):
7050 """Build input data structures.
7053 self._ComputeClusterData()
7055 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
7056 self._AddNewInstance()
7058 self._AddRelocateInstance()
7060 self.in_text = serializer.Dump(self.in_data)
7062 def Run(self, name, validate=True, call_fn=None):
7063 """Run an instance allocator and return the results.
7067 call_fn = self.lu.rpc.call_iallocator_runner
7070 result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
7073 if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
7074 raise errors.OpExecError("Invalid result from master iallocator runner")
7076 rcode, stdout, stderr, fail = result.data
7078 if rcode == constants.IARUN_NOTFOUND:
7079 raise errors.OpExecError("Can't find allocator '%s'" % name)
7080 elif rcode == constants.IARUN_FAILURE:
7081 raise errors.OpExecError("Instance allocator call failed: %s,"
7082 " output: %s" % (fail, stdout+stderr))
7083 self.out_text = stdout
7085 self._ValidateResult()
7087 def _ValidateResult(self):
7088 """Process the allocator results.
7090 This will process and if successful save the result in
7091 self.out_data and the other parameters.
7095 rdict = serializer.Load(self.out_text)
7096 except Exception, err:
7097 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
7099 if not isinstance(rdict, dict):
7100 raise errors.OpExecError("Can't parse iallocator results: not a dict")
7102 for key in "success", "info", "nodes":
7103 if key not in rdict:
7104 raise errors.OpExecError("Can't parse iallocator results:"
7105 " missing key '%s'" % key)
7106 setattr(self, key, rdict[key])
7108 if not isinstance(rdict["nodes"], list):
7109 raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
7111 self.out_data = rdict
7114 class LUTestAllocator(NoHooksLU):
7115 """Run allocator tests.
7117 This LU runs the allocator tests
7120 _OP_REQP = ["direction", "mode", "name"]
7122 def CheckPrereq(self):
7123 """Check prerequisites.
7125 This checks the opcode parameters depending on the director and mode test.
7128 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
7129 for attr in ["name", "mem_size", "disks", "disk_template",
7130 "os", "tags", "nics", "vcpus"]:
7131 if not hasattr(self.op, attr):
7132 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
7134 iname = self.cfg.ExpandInstanceName(self.op.name)
7135 if iname is not None:
7136 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
7138 if not isinstance(self.op.nics, list):
7139 raise errors.OpPrereqError("Invalid parameter 'nics'")
7140 for row in self.op.nics:
7141 if (not isinstance(row, dict) or
7144 "bridge" not in row):
7145 raise errors.OpPrereqError("Invalid contents of the"
7146 " 'nics' parameter")
7147 if not isinstance(self.op.disks, list):
7148 raise errors.OpPrereqError("Invalid parameter 'disks'")
7149 for row in self.op.disks:
7150 if (not isinstance(row, dict) or
7151 "size" not in row or
7152 not isinstance(row["size"], int) or
7153 "mode" not in row or
7154 row["mode"] not in ['r', 'w']):
7155 raise errors.OpPrereqError("Invalid contents of the"
7156 " 'disks' parameter")
7157 if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
7158 self.op.hypervisor = self.cfg.GetHypervisorType()
7159 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
7160 if not hasattr(self.op, "name"):
7161 raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
7162 fname = self.cfg.ExpandInstanceName(self.op.name)
7164 raise errors.OpPrereqError("Instance '%s' not found for relocation" %
7166 self.op.name = fname
7167 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
7169 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
7172 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
7173 if not hasattr(self.op, "allocator") or self.op.allocator is None:
7174 raise errors.OpPrereqError("Missing allocator name")
7175 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
7176 raise errors.OpPrereqError("Wrong allocator test '%s'" %
7179 def Exec(self, feedback_fn):
7180 """Run the allocator test.
7183 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
7184 ial = IAllocator(self,
7187 mem_size=self.op.mem_size,
7188 disks=self.op.disks,
7189 disk_template=self.op.disk_template,
7193 vcpus=self.op.vcpus,
7194 hypervisor=self.op.hypervisor,
7197 ial = IAllocator(self,
7200 relocate_from=list(self.relocate_from),
7203 if self.op.direction == constants.IALLOCATOR_DIR_IN:
7204 result = ial.in_text
7206 ial.Run(self.op.allocator, validate=False)
7207 result = ial.out_text