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)
2915 msg = remote_info.RemoteFailMsg()
2917 raise errors.OpPrereqError("Error checking node %s: %s" %
2918 (instance.primary_node, msg))
2919 if not remote_info.payload: # not running already
2920 _CheckNodeFreeMemory(self, instance.primary_node,
2921 "starting instance %s" % instance.name,
2922 bep[constants.BE_MEMORY], instance.hypervisor)
2924 def Exec(self, feedback_fn):
2925 """Start the instance.
2928 instance = self.instance
2929 force = self.op.force
2931 self.cfg.MarkInstanceUp(instance.name)
2933 node_current = instance.primary_node
2935 _StartInstanceDisks(self, instance, force)
2937 result = self.rpc.call_instance_start(node_current, instance,
2938 self.hvparams, self.beparams)
2939 msg = result.RemoteFailMsg()
2941 _ShutdownInstanceDisks(self, instance)
2942 raise errors.OpExecError("Could not start instance: %s" % msg)
2945 class LURebootInstance(LogicalUnit):
2946 """Reboot an instance.
2949 HPATH = "instance-reboot"
2950 HTYPE = constants.HTYPE_INSTANCE
2951 _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2954 def ExpandNames(self):
2955 if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2956 constants.INSTANCE_REBOOT_HARD,
2957 constants.INSTANCE_REBOOT_FULL]:
2958 raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2959 (constants.INSTANCE_REBOOT_SOFT,
2960 constants.INSTANCE_REBOOT_HARD,
2961 constants.INSTANCE_REBOOT_FULL))
2962 self._ExpandAndLockInstance()
2964 def BuildHooksEnv(self):
2967 This runs on master, primary and secondary nodes of the instance.
2971 "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2972 "REBOOT_TYPE": self.op.reboot_type,
2974 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2975 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2978 def CheckPrereq(self):
2979 """Check prerequisites.
2981 This checks that the instance is in the cluster.
2984 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2985 assert self.instance is not None, \
2986 "Cannot retrieve locked instance %s" % self.op.instance_name
2988 _CheckNodeOnline(self, instance.primary_node)
2990 # check bridges existance
2991 _CheckInstanceBridgesExist(self, instance)
2993 def Exec(self, feedback_fn):
2994 """Reboot the instance.
2997 instance = self.instance
2998 ignore_secondaries = self.op.ignore_secondaries
2999 reboot_type = self.op.reboot_type
3001 node_current = instance.primary_node
3003 if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
3004 constants.INSTANCE_REBOOT_HARD]:
3005 for disk in instance.disks:
3006 self.cfg.SetDiskID(disk, node_current)
3007 result = self.rpc.call_instance_reboot(node_current, instance,
3009 msg = result.RemoteFailMsg()
3011 raise errors.OpExecError("Could not reboot instance: %s" % msg)
3013 result = self.rpc.call_instance_shutdown(node_current, instance)
3014 msg = result.RemoteFailMsg()
3016 raise errors.OpExecError("Could not shutdown instance for"
3017 " full reboot: %s" % msg)
3018 _ShutdownInstanceDisks(self, instance)
3019 _StartInstanceDisks(self, instance, ignore_secondaries)
3020 result = self.rpc.call_instance_start(node_current, instance, None, None)
3021 msg = result.RemoteFailMsg()
3023 _ShutdownInstanceDisks(self, instance)
3024 raise errors.OpExecError("Could not start instance for"
3025 " full reboot: %s" % msg)
3027 self.cfg.MarkInstanceUp(instance.name)
3030 class LUShutdownInstance(LogicalUnit):
3031 """Shutdown an instance.
3034 HPATH = "instance-stop"
3035 HTYPE = constants.HTYPE_INSTANCE
3036 _OP_REQP = ["instance_name"]
3039 def ExpandNames(self):
3040 self._ExpandAndLockInstance()
3042 def BuildHooksEnv(self):
3045 This runs on master, primary and secondary nodes of the instance.
3048 env = _BuildInstanceHookEnvByObject(self, self.instance)
3049 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3052 def CheckPrereq(self):
3053 """Check prerequisites.
3055 This checks that the instance is in the cluster.
3058 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3059 assert self.instance is not None, \
3060 "Cannot retrieve locked instance %s" % self.op.instance_name
3061 _CheckNodeOnline(self, self.instance.primary_node)
3063 def Exec(self, feedback_fn):
3064 """Shutdown the instance.
3067 instance = self.instance
3068 node_current = instance.primary_node
3069 self.cfg.MarkInstanceDown(instance.name)
3070 result = self.rpc.call_instance_shutdown(node_current, instance)
3071 msg = result.RemoteFailMsg()
3073 self.proc.LogWarning("Could not shutdown instance: %s" % msg)
3075 _ShutdownInstanceDisks(self, instance)
3078 class LUReinstallInstance(LogicalUnit):
3079 """Reinstall an instance.
3082 HPATH = "instance-reinstall"
3083 HTYPE = constants.HTYPE_INSTANCE
3084 _OP_REQP = ["instance_name"]
3087 def ExpandNames(self):
3088 self._ExpandAndLockInstance()
3090 def BuildHooksEnv(self):
3093 This runs on master, primary and secondary nodes of the instance.
3096 env = _BuildInstanceHookEnvByObject(self, self.instance)
3097 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3100 def CheckPrereq(self):
3101 """Check prerequisites.
3103 This checks that the instance is in the cluster and is not running.
3106 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3107 assert instance is not None, \
3108 "Cannot retrieve locked instance %s" % self.op.instance_name
3109 _CheckNodeOnline(self, instance.primary_node)
3111 if instance.disk_template == constants.DT_DISKLESS:
3112 raise errors.OpPrereqError("Instance '%s' has no disks" %
3113 self.op.instance_name)
3114 if instance.admin_up:
3115 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3116 self.op.instance_name)
3117 remote_info = self.rpc.call_instance_info(instance.primary_node,
3119 instance.hypervisor)
3120 msg = remote_info.RemoteFailMsg()
3122 raise errors.OpPrereqError("Error checking node %s: %s" %
3123 (instance.primary_node, msg))
3124 if remote_info.payload:
3125 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3126 (self.op.instance_name,
3127 instance.primary_node))
3129 self.op.os_type = getattr(self.op, "os_type", None)
3130 if self.op.os_type is not None:
3132 pnode = self.cfg.GetNodeInfo(
3133 self.cfg.ExpandNodeName(instance.primary_node))
3135 raise errors.OpPrereqError("Primary node '%s' is unknown" %
3137 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3139 if not isinstance(result.data, objects.OS):
3140 raise errors.OpPrereqError("OS '%s' not in supported OS list for"
3141 " primary node" % self.op.os_type)
3143 self.instance = instance
3145 def Exec(self, feedback_fn):
3146 """Reinstall the instance.
3149 inst = self.instance
3151 if self.op.os_type is not None:
3152 feedback_fn("Changing OS to '%s'..." % self.op.os_type)
3153 inst.os = self.op.os_type
3154 self.cfg.Update(inst)
3156 _StartInstanceDisks(self, inst, None)
3158 feedback_fn("Running the instance OS create scripts...")
3159 result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
3160 msg = result.RemoteFailMsg()
3162 raise errors.OpExecError("Could not install OS for instance %s"
3164 (inst.name, inst.primary_node, msg))
3166 _ShutdownInstanceDisks(self, inst)
3169 class LURenameInstance(LogicalUnit):
3170 """Rename an instance.
3173 HPATH = "instance-rename"
3174 HTYPE = constants.HTYPE_INSTANCE
3175 _OP_REQP = ["instance_name", "new_name"]
3177 def BuildHooksEnv(self):
3180 This runs on master, primary and secondary nodes of the instance.
3183 env = _BuildInstanceHookEnvByObject(self, self.instance)
3184 env["INSTANCE_NEW_NAME"] = self.op.new_name
3185 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3188 def CheckPrereq(self):
3189 """Check prerequisites.
3191 This checks that the instance is in the cluster and is not running.
3194 instance = self.cfg.GetInstanceInfo(
3195 self.cfg.ExpandInstanceName(self.op.instance_name))
3196 if instance is None:
3197 raise errors.OpPrereqError("Instance '%s' not known" %
3198 self.op.instance_name)
3199 _CheckNodeOnline(self, instance.primary_node)
3201 if instance.admin_up:
3202 raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3203 self.op.instance_name)
3204 remote_info = self.rpc.call_instance_info(instance.primary_node,
3206 instance.hypervisor)
3207 msg = remote_info.RemoteFailMsg()
3209 raise errors.OpPrereqError("Error checking node %s: %s" %
3210 (instance.primary_node, msg))
3211 if remote_info.payload:
3212 raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3213 (self.op.instance_name,
3214 instance.primary_node))
3215 self.instance = instance
3217 # new name verification
3218 name_info = utils.HostInfo(self.op.new_name)
3220 self.op.new_name = new_name = name_info.name
3221 instance_list = self.cfg.GetInstanceList()
3222 if new_name in instance_list:
3223 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
3226 if not getattr(self.op, "ignore_ip", False):
3227 if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
3228 raise errors.OpPrereqError("IP %s of instance %s already in use" %
3229 (name_info.ip, new_name))
3232 def Exec(self, feedback_fn):
3233 """Reinstall the instance.
3236 inst = self.instance
3237 old_name = inst.name
3239 if inst.disk_template == constants.DT_FILE:
3240 old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
3242 self.cfg.RenameInstance(inst.name, self.op.new_name)
3243 # Change the instance lock. This is definitely safe while we hold the BGL
3244 self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
3245 self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
3247 # re-read the instance from the configuration after rename
3248 inst = self.cfg.GetInstanceInfo(self.op.new_name)
3250 if inst.disk_template == constants.DT_FILE:
3251 new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
3252 result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
3253 old_file_storage_dir,
3254 new_file_storage_dir)
3257 raise errors.OpExecError("Could not connect to node '%s' to rename"
3258 " directory '%s' to '%s' (but the instance"
3259 " has been renamed in Ganeti)" % (
3260 inst.primary_node, old_file_storage_dir,
3261 new_file_storage_dir))
3263 if not result.data[0]:
3264 raise errors.OpExecError("Could not rename directory '%s' to '%s'"
3265 " (but the instance has been renamed in"
3266 " Ganeti)" % (old_file_storage_dir,
3267 new_file_storage_dir))
3269 _StartInstanceDisks(self, inst, None)
3271 result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
3273 msg = result.RemoteFailMsg()
3275 msg = ("Could not run OS rename script for instance %s on node %s"
3276 " (but the instance has been renamed in Ganeti): %s" %
3277 (inst.name, inst.primary_node, msg))
3278 self.proc.LogWarning(msg)
3280 _ShutdownInstanceDisks(self, inst)
3283 class LURemoveInstance(LogicalUnit):
3284 """Remove an instance.
3287 HPATH = "instance-remove"
3288 HTYPE = constants.HTYPE_INSTANCE
3289 _OP_REQP = ["instance_name", "ignore_failures"]
3292 def ExpandNames(self):
3293 self._ExpandAndLockInstance()
3294 self.needed_locks[locking.LEVEL_NODE] = []
3295 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3297 def DeclareLocks(self, level):
3298 if level == locking.LEVEL_NODE:
3299 self._LockInstancesNodes()
3301 def BuildHooksEnv(self):
3304 This runs on master, primary and secondary nodes of the instance.
3307 env = _BuildInstanceHookEnvByObject(self, self.instance)
3308 nl = [self.cfg.GetMasterNode()]
3311 def CheckPrereq(self):
3312 """Check prerequisites.
3314 This checks that the instance is in the cluster.
3317 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3318 assert self.instance is not None, \
3319 "Cannot retrieve locked instance %s" % self.op.instance_name
3321 def Exec(self, feedback_fn):
3322 """Remove the instance.
3325 instance = self.instance
3326 logging.info("Shutting down instance %s on node %s",
3327 instance.name, instance.primary_node)
3329 result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3330 msg = result.RemoteFailMsg()
3332 if self.op.ignore_failures:
3333 feedback_fn("Warning: can't shutdown instance: %s" % msg)
3335 raise errors.OpExecError("Could not shutdown instance %s on"
3337 (instance.name, instance.primary_node, msg))
3339 logging.info("Removing block devices for instance %s", instance.name)
3341 if not _RemoveDisks(self, instance):
3342 if self.op.ignore_failures:
3343 feedback_fn("Warning: can't remove instance's disks")
3345 raise errors.OpExecError("Can't remove instance's disks")
3347 logging.info("Removing instance %s out of cluster config", instance.name)
3349 self.cfg.RemoveInstance(instance.name)
3350 self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3353 class LUQueryInstances(NoHooksLU):
3354 """Logical unit for querying instances.
3357 _OP_REQP = ["output_fields", "names", "use_locking"]
3359 _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3361 "disk_template", "ip", "mac", "bridge",
3362 "sda_size", "sdb_size", "vcpus", "tags",
3363 "network_port", "beparams",
3364 r"(disk)\.(size)/([0-9]+)",
3365 r"(disk)\.(sizes)", "disk_usage",
3366 r"(nic)\.(mac|ip|bridge)/([0-9]+)",
3367 r"(nic)\.(macs|ips|bridges)",
3368 r"(disk|nic)\.(count)",
3369 "serial_no", "hypervisor", "hvparams",] +
3371 for name in constants.HVS_PARAMETERS] +
3373 for name in constants.BES_PARAMETERS])
3374 _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3377 def ExpandNames(self):
3378 _CheckOutputFields(static=self._FIELDS_STATIC,
3379 dynamic=self._FIELDS_DYNAMIC,
3380 selected=self.op.output_fields)
3382 self.needed_locks = {}
3383 self.share_locks[locking.LEVEL_INSTANCE] = 1
3384 self.share_locks[locking.LEVEL_NODE] = 1
3387 self.wanted = _GetWantedInstances(self, self.op.names)
3389 self.wanted = locking.ALL_SET
3391 self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3392 self.do_locking = self.do_node_query and self.op.use_locking
3394 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3395 self.needed_locks[locking.LEVEL_NODE] = []
3396 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3398 def DeclareLocks(self, level):
3399 if level == locking.LEVEL_NODE and self.do_locking:
3400 self._LockInstancesNodes()
3402 def CheckPrereq(self):
3403 """Check prerequisites.
3408 def Exec(self, feedback_fn):
3409 """Computes the list of nodes and their attributes.
3412 all_info = self.cfg.GetAllInstancesInfo()
3413 if self.wanted == locking.ALL_SET:
3414 # caller didn't specify instance names, so ordering is not important
3416 instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3418 instance_names = all_info.keys()
3419 instance_names = utils.NiceSort(instance_names)
3421 # caller did specify names, so we must keep the ordering
3423 tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3425 tgt_set = all_info.keys()
3426 missing = set(self.wanted).difference(tgt_set)
3428 raise errors.OpExecError("Some instances were removed before"
3429 " retrieving their data: %s" % missing)
3430 instance_names = self.wanted
3432 instance_list = [all_info[iname] for iname in instance_names]
3434 # begin data gathering
3436 nodes = frozenset([inst.primary_node for inst in instance_list])
3437 hv_list = list(set([inst.hypervisor for inst in instance_list]))
3441 if self.do_node_query:
3443 node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3445 result = node_data[name]
3447 # offline nodes will be in both lists
3448 off_nodes.append(name)
3450 bad_nodes.append(name)
3453 live_data.update(result.data)
3454 # else no instance is alive
3456 live_data = dict([(name, {}) for name in instance_names])
3458 # end data gathering
3463 for instance in instance_list:
3465 i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3466 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3467 for field in self.op.output_fields:
3468 st_match = self._FIELDS_STATIC.Matches(field)
3473 elif field == "pnode":
3474 val = instance.primary_node
3475 elif field == "snodes":
3476 val = list(instance.secondary_nodes)
3477 elif field == "admin_state":
3478 val = instance.admin_up
3479 elif field == "oper_state":
3480 if instance.primary_node in bad_nodes:
3483 val = bool(live_data.get(instance.name))
3484 elif field == "status":
3485 if instance.primary_node in off_nodes:
3486 val = "ERROR_nodeoffline"
3487 elif instance.primary_node in bad_nodes:
3488 val = "ERROR_nodedown"
3490 running = bool(live_data.get(instance.name))
3492 if instance.admin_up:
3497 if instance.admin_up:
3501 elif field == "oper_ram":
3502 if instance.primary_node in bad_nodes:
3504 elif instance.name in live_data:
3505 val = live_data[instance.name].get("memory", "?")
3508 elif field == "disk_template":
3509 val = instance.disk_template
3511 val = instance.nics[0].ip
3512 elif field == "bridge":
3513 val = instance.nics[0].bridge
3514 elif field == "mac":
3515 val = instance.nics[0].mac
3516 elif field == "sda_size" or field == "sdb_size":
3517 idx = ord(field[2]) - ord('a')
3519 val = instance.FindDisk(idx).size
3520 except errors.OpPrereqError:
3522 elif field == "disk_usage": # total disk usage per node
3523 disk_sizes = [{'size': disk.size} for disk in instance.disks]
3524 val = _ComputeDiskSize(instance.disk_template, disk_sizes)
3525 elif field == "tags":
3526 val = list(instance.GetTags())
3527 elif field == "serial_no":
3528 val = instance.serial_no
3529 elif field == "network_port":
3530 val = instance.network_port
3531 elif field == "hypervisor":
3532 val = instance.hypervisor
3533 elif field == "hvparams":
3535 elif (field.startswith(HVPREFIX) and
3536 field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3537 val = i_hv.get(field[len(HVPREFIX):], None)
3538 elif field == "beparams":
3540 elif (field.startswith(BEPREFIX) and
3541 field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3542 val = i_be.get(field[len(BEPREFIX):], None)
3543 elif st_match and st_match.groups():
3544 # matches a variable list
3545 st_groups = st_match.groups()
3546 if st_groups and st_groups[0] == "disk":
3547 if st_groups[1] == "count":
3548 val = len(instance.disks)
3549 elif st_groups[1] == "sizes":
3550 val = [disk.size for disk in instance.disks]
3551 elif st_groups[1] == "size":
3553 val = instance.FindDisk(st_groups[2]).size
3554 except errors.OpPrereqError:
3557 assert False, "Unhandled disk parameter"
3558 elif st_groups[0] == "nic":
3559 if st_groups[1] == "count":
3560 val = len(instance.nics)
3561 elif st_groups[1] == "macs":
3562 val = [nic.mac for nic in instance.nics]
3563 elif st_groups[1] == "ips":
3564 val = [nic.ip for nic in instance.nics]
3565 elif st_groups[1] == "bridges":
3566 val = [nic.bridge for nic in instance.nics]
3569 nic_idx = int(st_groups[2])
3570 if nic_idx >= len(instance.nics):
3573 if st_groups[1] == "mac":
3574 val = instance.nics[nic_idx].mac
3575 elif st_groups[1] == "ip":
3576 val = instance.nics[nic_idx].ip
3577 elif st_groups[1] == "bridge":
3578 val = instance.nics[nic_idx].bridge
3580 assert False, "Unhandled NIC parameter"
3582 assert False, "Unhandled variable parameter"
3584 raise errors.ParameterError(field)
3591 class LUFailoverInstance(LogicalUnit):
3592 """Failover an instance.
3595 HPATH = "instance-failover"
3596 HTYPE = constants.HTYPE_INSTANCE
3597 _OP_REQP = ["instance_name", "ignore_consistency"]
3600 def ExpandNames(self):
3601 self._ExpandAndLockInstance()
3602 self.needed_locks[locking.LEVEL_NODE] = []
3603 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3605 def DeclareLocks(self, level):
3606 if level == locking.LEVEL_NODE:
3607 self._LockInstancesNodes()
3609 def BuildHooksEnv(self):
3612 This runs on master, primary and secondary nodes of the instance.
3616 "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3618 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3619 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3622 def CheckPrereq(self):
3623 """Check prerequisites.
3625 This checks that the instance is in the cluster.
3628 self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3629 assert self.instance is not None, \
3630 "Cannot retrieve locked instance %s" % self.op.instance_name
3632 bep = self.cfg.GetClusterInfo().FillBE(instance)
3633 if instance.disk_template not in constants.DTS_NET_MIRROR:
3634 raise errors.OpPrereqError("Instance's disk layout is not"
3635 " network mirrored, cannot failover.")
3637 secondary_nodes = instance.secondary_nodes
3638 if not secondary_nodes:
3639 raise errors.ProgrammerError("no secondary node but using "
3640 "a mirrored disk template")
3642 target_node = secondary_nodes[0]
3643 _CheckNodeOnline(self, target_node)
3644 _CheckNodeNotDrained(self, target_node)
3645 # check memory requirements on the secondary node
3646 _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3647 instance.name, bep[constants.BE_MEMORY],
3648 instance.hypervisor)
3649 # check bridge existance
3650 _CheckInstanceBridgesExist(self, instance, node=target_node)
3652 def Exec(self, feedback_fn):
3653 """Failover an instance.
3655 The failover is done by shutting it down on its present node and
3656 starting it on the secondary.
3659 instance = self.instance
3661 source_node = instance.primary_node
3662 target_node = instance.secondary_nodes[0]
3664 feedback_fn("* checking disk consistency between source and target")
3665 for dev in instance.disks:
3666 # for drbd, these are drbd over lvm
3667 if not _CheckDiskConsistency(self, dev, target_node, False):
3668 if instance.admin_up and not self.op.ignore_consistency:
3669 raise errors.OpExecError("Disk %s is degraded on target node,"
3670 " aborting failover." % dev.iv_name)
3672 feedback_fn("* shutting down instance on source node")
3673 logging.info("Shutting down instance %s on node %s",
3674 instance.name, source_node)
3676 result = self.rpc.call_instance_shutdown(source_node, instance)
3677 msg = result.RemoteFailMsg()
3679 if self.op.ignore_consistency:
3680 self.proc.LogWarning("Could not shutdown instance %s on node %s."
3681 " Proceeding anyway. Please make sure node"
3682 " %s is down. Error details: %s",
3683 instance.name, source_node, source_node, msg)
3685 raise errors.OpExecError("Could not shutdown instance %s on"
3687 (instance.name, source_node, msg))
3689 feedback_fn("* deactivating the instance's disks on source node")
3690 if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3691 raise errors.OpExecError("Can't shut down the instance's disks.")
3693 instance.primary_node = target_node
3694 # distribute new instance config to the other nodes
3695 self.cfg.Update(instance)
3697 # Only start the instance if it's marked as up
3698 if instance.admin_up:
3699 feedback_fn("* activating the instance's disks on target node")
3700 logging.info("Starting instance %s on node %s",
3701 instance.name, target_node)
3703 disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3704 ignore_secondaries=True)
3706 _ShutdownInstanceDisks(self, instance)
3707 raise errors.OpExecError("Can't activate the instance's disks")
3709 feedback_fn("* starting the instance on the target node")
3710 result = self.rpc.call_instance_start(target_node, instance, None, None)
3711 msg = result.RemoteFailMsg()
3713 _ShutdownInstanceDisks(self, instance)
3714 raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3715 (instance.name, target_node, msg))
3718 class LUMigrateInstance(LogicalUnit):
3719 """Migrate an instance.
3721 This is migration without shutting down, compared to the failover,
3722 which is done with shutdown.
3725 HPATH = "instance-migrate"
3726 HTYPE = constants.HTYPE_INSTANCE
3727 _OP_REQP = ["instance_name", "live", "cleanup"]
3731 def ExpandNames(self):
3732 self._ExpandAndLockInstance()
3733 self.needed_locks[locking.LEVEL_NODE] = []
3734 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3736 def DeclareLocks(self, level):
3737 if level == locking.LEVEL_NODE:
3738 self._LockInstancesNodes()
3740 def BuildHooksEnv(self):
3743 This runs on master, primary and secondary nodes of the instance.
3746 env = _BuildInstanceHookEnvByObject(self, self.instance)
3747 env["MIGRATE_LIVE"] = self.op.live
3748 env["MIGRATE_CLEANUP"] = self.op.cleanup
3749 nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3752 def CheckPrereq(self):
3753 """Check prerequisites.
3755 This checks that the instance is in the cluster.
3758 instance = self.cfg.GetInstanceInfo(
3759 self.cfg.ExpandInstanceName(self.op.instance_name))
3760 if instance is None:
3761 raise errors.OpPrereqError("Instance '%s' not known" %
3762 self.op.instance_name)
3764 if instance.disk_template != constants.DT_DRBD8:
3765 raise errors.OpPrereqError("Instance's disk layout is not"
3766 " drbd8, cannot migrate.")
3768 secondary_nodes = instance.secondary_nodes
3769 if not secondary_nodes:
3770 raise errors.ConfigurationError("No secondary node but using"
3771 " drbd8 disk template")
3773 i_be = self.cfg.GetClusterInfo().FillBE(instance)
3775 target_node = secondary_nodes[0]
3776 # check memory requirements on the secondary node
3777 _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3778 instance.name, i_be[constants.BE_MEMORY],
3779 instance.hypervisor)
3781 # check bridge existance
3782 _CheckInstanceBridgesExist(self, instance, node=target_node)
3784 if not self.op.cleanup:
3785 _CheckNodeNotDrained(self, target_node)
3786 result = self.rpc.call_instance_migratable(instance.primary_node,
3788 msg = result.RemoteFailMsg()
3790 raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3793 self.instance = instance
3795 def _WaitUntilSync(self):
3796 """Poll with custom rpc for disk sync.
3798 This uses our own step-based rpc call.
3801 self.feedback_fn("* wait until resync is done")
3805 result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3807 self.instance.disks)
3809 for node, nres in result.items():
3810 msg = nres.RemoteFailMsg()
3812 raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3814 node_done, node_percent = nres.payload
3815 all_done = all_done and node_done
3816 if node_percent is not None:
3817 min_percent = min(min_percent, node_percent)
3819 if min_percent < 100:
3820 self.feedback_fn(" - progress: %.1f%%" % min_percent)
3823 def _EnsureSecondary(self, node):
3824 """Demote a node to secondary.
3827 self.feedback_fn("* switching node %s to secondary mode" % node)
3829 for dev in self.instance.disks:
3830 self.cfg.SetDiskID(dev, node)
3832 result = self.rpc.call_blockdev_close(node, self.instance.name,
3833 self.instance.disks)
3834 msg = result.RemoteFailMsg()
3836 raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3837 " error %s" % (node, msg))
3839 def _GoStandalone(self):
3840 """Disconnect from the network.
3843 self.feedback_fn("* changing into standalone mode")
3844 result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3845 self.instance.disks)
3846 for node, nres in result.items():
3847 msg = nres.RemoteFailMsg()
3849 raise errors.OpExecError("Cannot disconnect disks node %s,"
3850 " error %s" % (node, msg))
3852 def _GoReconnect(self, multimaster):
3853 """Reconnect to the network.
3859 msg = "single-master"
3860 self.feedback_fn("* changing disks into %s mode" % msg)
3861 result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3862 self.instance.disks,
3863 self.instance.name, multimaster)
3864 for node, nres in result.items():
3865 msg = nres.RemoteFailMsg()
3867 raise errors.OpExecError("Cannot change disks config on node %s,"
3868 " error: %s" % (node, msg))
3870 def _ExecCleanup(self):
3871 """Try to cleanup after a failed migration.
3873 The cleanup is done by:
3874 - check that the instance is running only on one node
3875 (and update the config if needed)
3876 - change disks on its secondary node to secondary
3877 - wait until disks are fully synchronized
3878 - disconnect from the network
3879 - change disks into single-master mode
3880 - wait again until disks are fully synchronized
3883 instance = self.instance
3884 target_node = self.target_node
3885 source_node = self.source_node
3887 # check running on only one node
3888 self.feedback_fn("* checking where the instance actually runs"
3889 " (if this hangs, the hypervisor might be in"
3891 ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3892 for node, result in ins_l.items():
3894 if not isinstance(result.data, list):
3895 raise errors.OpExecError("Can't contact node '%s'" % node)
3897 runningon_source = instance.name in ins_l[source_node].data
3898 runningon_target = instance.name in ins_l[target_node].data
3900 if runningon_source and runningon_target:
3901 raise errors.OpExecError("Instance seems to be running on two nodes,"
3902 " or the hypervisor is confused. You will have"
3903 " to ensure manually that it runs only on one"
3904 " and restart this operation.")
3906 if not (runningon_source or runningon_target):
3907 raise errors.OpExecError("Instance does not seem to be running at all."
3908 " In this case, it's safer to repair by"
3909 " running 'gnt-instance stop' to ensure disk"
3910 " shutdown, and then restarting it.")
3912 if runningon_target:
3913 # the migration has actually succeeded, we need to update the config
3914 self.feedback_fn("* instance running on secondary node (%s),"
3915 " updating config" % target_node)
3916 instance.primary_node = target_node
3917 self.cfg.Update(instance)
3918 demoted_node = source_node
3920 self.feedback_fn("* instance confirmed to be running on its"
3921 " primary node (%s)" % source_node)
3922 demoted_node = target_node
3924 self._EnsureSecondary(demoted_node)
3926 self._WaitUntilSync()
3927 except errors.OpExecError:
3928 # we ignore here errors, since if the device is standalone, it
3929 # won't be able to sync
3931 self._GoStandalone()
3932 self._GoReconnect(False)
3933 self._WaitUntilSync()
3935 self.feedback_fn("* done")
3937 def _RevertDiskStatus(self):
3938 """Try to revert the disk status after a failed migration.
3941 target_node = self.target_node
3943 self._EnsureSecondary(target_node)
3944 self._GoStandalone()
3945 self._GoReconnect(False)
3946 self._WaitUntilSync()
3947 except errors.OpExecError, err:
3948 self.LogWarning("Migration failed and I can't reconnect the"
3949 " drives: error '%s'\n"
3950 "Please look and recover the instance status" %
3953 def _AbortMigration(self):
3954 """Call the hypervisor code to abort a started migration.
3957 instance = self.instance
3958 target_node = self.target_node
3959 migration_info = self.migration_info
3961 abort_result = self.rpc.call_finalize_migration(target_node,
3965 abort_msg = abort_result.RemoteFailMsg()
3967 logging.error("Aborting migration failed on target node %s: %s" %
3968 (target_node, abort_msg))
3969 # Don't raise an exception here, as we stil have to try to revert the
3970 # disk status, even if this step failed.
3972 def _ExecMigration(self):
3973 """Migrate an instance.
3975 The migrate is done by:
3976 - change the disks into dual-master mode
3977 - wait until disks are fully synchronized again
3978 - migrate the instance
3979 - change disks on the new secondary node (the old primary) to secondary
3980 - wait until disks are fully synchronized
3981 - change disks into single-master mode
3984 instance = self.instance
3985 target_node = self.target_node
3986 source_node = self.source_node
3988 self.feedback_fn("* checking disk consistency between source and target")
3989 for dev in instance.disks:
3990 if not _CheckDiskConsistency(self, dev, target_node, False):
3991 raise errors.OpExecError("Disk %s is degraded or not fully"
3992 " synchronized on target node,"
3993 " aborting migrate." % dev.iv_name)
3995 # First get the migration information from the remote node
3996 result = self.rpc.call_migration_info(source_node, instance)
3997 msg = result.RemoteFailMsg()
3999 log_err = ("Failed fetching source migration information from %s: %s" %
4001 logging.error(log_err)
4002 raise errors.OpExecError(log_err)
4004 self.migration_info = migration_info = result.payload
4006 # Then switch the disks to master/master mode
4007 self._EnsureSecondary(target_node)
4008 self._GoStandalone()
4009 self._GoReconnect(True)
4010 self._WaitUntilSync()
4012 self.feedback_fn("* preparing %s to accept the instance" % target_node)
4013 result = self.rpc.call_accept_instance(target_node,
4016 self.nodes_ip[target_node])
4018 msg = result.RemoteFailMsg()
4020 logging.error("Instance pre-migration failed, trying to revert"
4021 " disk status: %s", msg)
4022 self._AbortMigration()
4023 self._RevertDiskStatus()
4024 raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
4025 (instance.name, msg))
4027 self.feedback_fn("* migrating instance to %s" % target_node)
4029 result = self.rpc.call_instance_migrate(source_node, instance,
4030 self.nodes_ip[target_node],
4032 msg = result.RemoteFailMsg()
4034 logging.error("Instance migration failed, trying to revert"
4035 " disk status: %s", msg)
4036 self._AbortMigration()
4037 self._RevertDiskStatus()
4038 raise errors.OpExecError("Could not migrate instance %s: %s" %
4039 (instance.name, msg))
4042 instance.primary_node = target_node
4043 # distribute new instance config to the other nodes
4044 self.cfg.Update(instance)
4046 result = self.rpc.call_finalize_migration(target_node,
4050 msg = result.RemoteFailMsg()
4052 logging.error("Instance migration succeeded, but finalization failed:"
4054 raise errors.OpExecError("Could not finalize instance migration: %s" %
4057 self._EnsureSecondary(source_node)
4058 self._WaitUntilSync()
4059 self._GoStandalone()
4060 self._GoReconnect(False)
4061 self._WaitUntilSync()
4063 self.feedback_fn("* done")
4065 def Exec(self, feedback_fn):
4066 """Perform the migration.
4069 self.feedback_fn = feedback_fn
4071 self.source_node = self.instance.primary_node
4072 self.target_node = self.instance.secondary_nodes[0]
4073 self.all_nodes = [self.source_node, self.target_node]
4075 self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
4076 self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
4079 return self._ExecCleanup()
4081 return self._ExecMigration()
4084 def _CreateBlockDev(lu, node, instance, device, force_create,
4086 """Create a tree of block devices on a given node.
4088 If this device type has to be created on secondaries, create it and
4091 If not, just recurse to children keeping the same 'force' value.
4093 @param lu: the lu on whose behalf we execute
4094 @param node: the node on which to create the device
4095 @type instance: L{objects.Instance}
4096 @param instance: the instance which owns the device
4097 @type device: L{objects.Disk}
4098 @param device: the device to create
4099 @type force_create: boolean
4100 @param force_create: whether to force creation of this device; this
4101 will be change to True whenever we find a device which has
4102 CreateOnSecondary() attribute
4103 @param info: the extra 'metadata' we should attach to the device
4104 (this will be represented as a LVM tag)
4105 @type force_open: boolean
4106 @param force_open: this parameter will be passes to the
4107 L{backend.BlockdevCreate} function where it specifies
4108 whether we run on primary or not, and it affects both
4109 the child assembly and the device own Open() execution
4112 if device.CreateOnSecondary():
4116 for child in device.children:
4117 _CreateBlockDev(lu, node, instance, child, force_create,
4120 if not force_create:
4123 _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
4126 def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
4127 """Create a single block device on a given node.
4129 This will not recurse over children of the device, so they must be
4132 @param lu: the lu on whose behalf we execute
4133 @param node: the node on which to create the device
4134 @type instance: L{objects.Instance}
4135 @param instance: the instance which owns the device
4136 @type device: L{objects.Disk}
4137 @param device: the device to create
4138 @param info: the extra 'metadata' we should attach to the device
4139 (this will be represented as a LVM tag)
4140 @type force_open: boolean
4141 @param force_open: this parameter will be passes to the
4142 L{backend.BlockdevCreate} function where it specifies
4143 whether we run on primary or not, and it affects both
4144 the child assembly and the device own Open() execution
4147 lu.cfg.SetDiskID(device, node)
4148 result = lu.rpc.call_blockdev_create(node, device, device.size,
4149 instance.name, force_open, info)
4150 msg = result.RemoteFailMsg()
4152 raise errors.OpExecError("Can't create block device %s on"
4153 " node %s for instance %s: %s" %
4154 (device, node, instance.name, msg))
4155 if device.physical_id is None:
4156 device.physical_id = result.payload
4159 def _GenerateUniqueNames(lu, exts):
4160 """Generate a suitable LV name.
4162 This will generate a logical volume name for the given instance.
4167 new_id = lu.cfg.GenerateUniqueID()
4168 results.append("%s%s" % (new_id, val))
4172 def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
4174 """Generate a drbd8 device complete with its children.
4177 port = lu.cfg.AllocatePort()
4178 vgname = lu.cfg.GetVGName()
4179 shared_secret = lu.cfg.GenerateDRBDSecret()
4180 dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4181 logical_id=(vgname, names[0]))
4182 dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4183 logical_id=(vgname, names[1]))
4184 drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
4185 logical_id=(primary, secondary, port,
4188 children=[dev_data, dev_meta],
4193 def _GenerateDiskTemplate(lu, template_name,
4194 instance_name, primary_node,
4195 secondary_nodes, disk_info,
4196 file_storage_dir, file_driver,
4198 """Generate the entire disk layout for a given template type.
4201 #TODO: compute space requirements
4203 vgname = lu.cfg.GetVGName()
4204 disk_count = len(disk_info)
4206 if template_name == constants.DT_DISKLESS:
4208 elif template_name == constants.DT_PLAIN:
4209 if len(secondary_nodes) != 0:
4210 raise errors.ProgrammerError("Wrong template configuration")
4212 names = _GenerateUniqueNames(lu, [".disk%d" % i
4213 for i in range(disk_count)])
4214 for idx, disk in enumerate(disk_info):
4215 disk_index = idx + base_index
4216 disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
4217 logical_id=(vgname, names[idx]),
4218 iv_name="disk/%d" % disk_index,
4220 disks.append(disk_dev)
4221 elif template_name == constants.DT_DRBD8:
4222 if len(secondary_nodes) != 1:
4223 raise errors.ProgrammerError("Wrong template configuration")
4224 remote_node = secondary_nodes[0]
4225 minors = lu.cfg.AllocateDRBDMinor(
4226 [primary_node, remote_node] * len(disk_info), instance_name)
4229 for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
4230 for i in range(disk_count)]):
4231 names.append(lv_prefix + "_data")
4232 names.append(lv_prefix + "_meta")
4233 for idx, disk in enumerate(disk_info):
4234 disk_index = idx + base_index
4235 disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
4236 disk["size"], names[idx*2:idx*2+2],
4237 "disk/%d" % disk_index,
4238 minors[idx*2], minors[idx*2+1])
4239 disk_dev.mode = disk["mode"]
4240 disks.append(disk_dev)
4241 elif template_name == constants.DT_FILE:
4242 if len(secondary_nodes) != 0:
4243 raise errors.ProgrammerError("Wrong template configuration")
4245 for idx, disk in enumerate(disk_info):
4246 disk_index = idx + base_index
4247 disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
4248 iv_name="disk/%d" % disk_index,
4249 logical_id=(file_driver,
4250 "%s/disk%d" % (file_storage_dir,
4253 disks.append(disk_dev)
4255 raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
4259 def _GetInstanceInfoText(instance):
4260 """Compute that text that should be added to the disk's metadata.
4263 return "originstname+%s" % instance.name
4266 def _CreateDisks(lu, instance):
4267 """Create all disks for an instance.
4269 This abstracts away some work from AddInstance.
4271 @type lu: L{LogicalUnit}
4272 @param lu: the logical unit on whose behalf we execute
4273 @type instance: L{objects.Instance}
4274 @param instance: the instance whose disks we should create
4276 @return: the success of the creation
4279 info = _GetInstanceInfoText(instance)
4280 pnode = instance.primary_node
4282 if instance.disk_template == constants.DT_FILE:
4283 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4284 result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
4286 if result.failed or not result.data:
4287 raise errors.OpExecError("Could not connect to node '%s'" % pnode)
4289 if not result.data[0]:
4290 raise errors.OpExecError("Failed to create directory '%s'" %
4293 # Note: this needs to be kept in sync with adding of disks in
4294 # LUSetInstanceParams
4295 for device in instance.disks:
4296 logging.info("Creating volume %s for instance %s",
4297 device.iv_name, instance.name)
4299 for node in instance.all_nodes:
4300 f_create = node == pnode
4301 _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4304 def _RemoveDisks(lu, instance):
4305 """Remove all disks for an instance.
4307 This abstracts away some work from `AddInstance()` and
4308 `RemoveInstance()`. Note that in case some of the devices couldn't
4309 be removed, the removal will continue with the other ones (compare
4310 with `_CreateDisks()`).
4312 @type lu: L{LogicalUnit}
4313 @param lu: the logical unit on whose behalf we execute
4314 @type instance: L{objects.Instance}
4315 @param instance: the instance whose disks we should remove
4317 @return: the success of the removal
4320 logging.info("Removing block devices for instance %s", instance.name)
4323 for device in instance.disks:
4324 for node, disk in device.ComputeNodeTree(instance.primary_node):
4325 lu.cfg.SetDiskID(disk, node)
4326 msg = lu.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
4328 lu.LogWarning("Could not remove block device %s on node %s,"
4329 " continuing anyway: %s", device.iv_name, node, msg)
4332 if instance.disk_template == constants.DT_FILE:
4333 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4334 result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4336 if result.failed or not result.data:
4337 logging.error("Could not remove directory '%s'", file_storage_dir)
4343 def _ComputeDiskSize(disk_template, disks):
4344 """Compute disk size requirements in the volume group
4347 # Required free disk space as a function of disk and swap space
4349 constants.DT_DISKLESS: None,
4350 constants.DT_PLAIN: sum(d["size"] for d in disks),
4351 # 128 MB are added for drbd metadata for each disk
4352 constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4353 constants.DT_FILE: None,
4356 if disk_template not in req_size_dict:
4357 raise errors.ProgrammerError("Disk template '%s' size requirement"
4358 " is unknown" % disk_template)
4360 return req_size_dict[disk_template]
4363 def _CheckHVParams(lu, nodenames, hvname, hvparams):
4364 """Hypervisor parameter validation.
4366 This function abstract the hypervisor parameter validation to be
4367 used in both instance create and instance modify.
4369 @type lu: L{LogicalUnit}
4370 @param lu: the logical unit for which we check
4371 @type nodenames: list
4372 @param nodenames: the list of nodes on which we should check
4373 @type hvname: string
4374 @param hvname: the name of the hypervisor we should use
4375 @type hvparams: dict
4376 @param hvparams: the parameters which we need to check
4377 @raise errors.OpPrereqError: if the parameters are not valid
4380 hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4383 for node in nodenames:
4387 msg = info.RemoteFailMsg()
4389 raise errors.OpPrereqError("Hypervisor parameter validation"
4390 " failed on node %s: %s" % (node, msg))
4393 class LUCreateInstance(LogicalUnit):
4394 """Create an instance.
4397 HPATH = "instance-add"
4398 HTYPE = constants.HTYPE_INSTANCE
4399 _OP_REQP = ["instance_name", "disks", "disk_template",
4401 "wait_for_sync", "ip_check", "nics",
4402 "hvparams", "beparams"]
4405 def _ExpandNode(self, node):
4406 """Expands and checks one node name.
4409 node_full = self.cfg.ExpandNodeName(node)
4410 if node_full is None:
4411 raise errors.OpPrereqError("Unknown node %s" % node)
4414 def ExpandNames(self):
4415 """ExpandNames for CreateInstance.
4417 Figure out the right locks for instance creation.
4420 self.needed_locks = {}
4422 # set optional parameters to none if they don't exist
4423 for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4424 if not hasattr(self.op, attr):
4425 setattr(self.op, attr, None)
4427 # cheap checks, mostly valid constants given
4429 # verify creation mode
4430 if self.op.mode not in (constants.INSTANCE_CREATE,
4431 constants.INSTANCE_IMPORT):
4432 raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4435 # disk template and mirror node verification
4436 if self.op.disk_template not in constants.DISK_TEMPLATES:
4437 raise errors.OpPrereqError("Invalid disk template name")
4439 if self.op.hypervisor is None:
4440 self.op.hypervisor = self.cfg.GetHypervisorType()
4442 cluster = self.cfg.GetClusterInfo()
4443 enabled_hvs = cluster.enabled_hypervisors
4444 if self.op.hypervisor not in enabled_hvs:
4445 raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4446 " cluster (%s)" % (self.op.hypervisor,
4447 ",".join(enabled_hvs)))
4449 # check hypervisor parameter syntax (locally)
4450 utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4451 filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
4453 hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4454 hv_type.CheckParameterSyntax(filled_hvp)
4456 # fill and remember the beparams dict
4457 utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4458 self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
4461 #### instance parameters check
4463 # instance name verification
4464 hostname1 = utils.HostInfo(self.op.instance_name)
4465 self.op.instance_name = instance_name = hostname1.name
4467 # this is just a preventive check, but someone might still add this
4468 # instance in the meantime, and creation will fail at lock-add time
4469 if instance_name in self.cfg.GetInstanceList():
4470 raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4473 self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4477 for idx, nic in enumerate(self.op.nics):
4478 nic_mode_req = nic.get("mode", None)
4479 nic_mode = nic_mode_req
4480 if nic_mode is None:
4481 nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
4483 # in routed mode, for the first nic, the default ip is 'auto'
4484 if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
4485 default_ip_mode = constants.VALUE_AUTO
4487 default_ip_mode = constants.VALUE_NONE
4489 # ip validity checks
4490 ip = nic.get("ip", default_ip_mode)
4491 if ip is None or ip.lower() == constants.VALUE_NONE:
4493 elif ip.lower() == constants.VALUE_AUTO:
4494 nic_ip = hostname1.ip
4496 if not utils.IsValidIP(ip):
4497 raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4498 " like a valid IP" % ip)
4501 # TODO: check the ip for uniqueness !!
4502 if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
4503 raise errors.OpPrereqError("Routed nic mode requires an ip address")
4505 # MAC address verification
4506 mac = nic.get("mac", constants.VALUE_AUTO)
4507 if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4508 if not utils.IsValidMac(mac.lower()):
4509 raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4511 # bridge verification
4512 bridge = nic.get("bridge", None)
4513 link = nic.get("link", None)
4515 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
4516 elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
4517 raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic")
4523 nicparams[constants.NIC_MODE] = nic_mode_req
4525 nicparams[constants.NIC_LINK] = link
4527 check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
4529 objects.NIC.CheckParameterSyntax(check_params)
4530 self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
4532 # disk checks/pre-build
4534 for disk in self.op.disks:
4535 mode = disk.get("mode", constants.DISK_RDWR)
4536 if mode not in constants.DISK_ACCESS_SET:
4537 raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4539 size = disk.get("size", None)
4541 raise errors.OpPrereqError("Missing disk size")
4545 raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4546 self.disks.append({"size": size, "mode": mode})
4548 # used in CheckPrereq for ip ping check
4549 self.check_ip = hostname1.ip
4551 # file storage checks
4552 if (self.op.file_driver and
4553 not self.op.file_driver in constants.FILE_DRIVER):
4554 raise errors.OpPrereqError("Invalid file driver name '%s'" %
4555 self.op.file_driver)
4557 if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4558 raise errors.OpPrereqError("File storage directory path not absolute")
4560 ### Node/iallocator related checks
4561 if [self.op.iallocator, self.op.pnode].count(None) != 1:
4562 raise errors.OpPrereqError("One and only one of iallocator and primary"
4563 " node must be given")
4565 if self.op.iallocator:
4566 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4568 self.op.pnode = self._ExpandNode(self.op.pnode)
4569 nodelist = [self.op.pnode]
4570 if self.op.snode is not None:
4571 self.op.snode = self._ExpandNode(self.op.snode)
4572 nodelist.append(self.op.snode)
4573 self.needed_locks[locking.LEVEL_NODE] = nodelist
4575 # in case of import lock the source node too
4576 if self.op.mode == constants.INSTANCE_IMPORT:
4577 src_node = getattr(self.op, "src_node", None)
4578 src_path = getattr(self.op, "src_path", None)
4580 if src_path is None:
4581 self.op.src_path = src_path = self.op.instance_name
4583 if src_node is None:
4584 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4585 self.op.src_node = None
4586 if os.path.isabs(src_path):
4587 raise errors.OpPrereqError("Importing an instance from an absolute"
4588 " path requires a source node option.")
4590 self.op.src_node = src_node = self._ExpandNode(src_node)
4591 if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4592 self.needed_locks[locking.LEVEL_NODE].append(src_node)
4593 if not os.path.isabs(src_path):
4594 self.op.src_path = src_path = \
4595 os.path.join(constants.EXPORT_DIR, src_path)
4597 else: # INSTANCE_CREATE
4598 if getattr(self.op, "os_type", None) is None:
4599 raise errors.OpPrereqError("No guest OS specified")
4601 def _RunAllocator(self):
4602 """Run the allocator based on input opcode.
4605 nics = [n.ToDict() for n in self.nics]
4606 ial = IAllocator(self,
4607 mode=constants.IALLOCATOR_MODE_ALLOC,
4608 name=self.op.instance_name,
4609 disk_template=self.op.disk_template,
4612 vcpus=self.be_full[constants.BE_VCPUS],
4613 mem_size=self.be_full[constants.BE_MEMORY],
4616 hypervisor=self.op.hypervisor,
4619 ial.Run(self.op.iallocator)
4622 raise errors.OpPrereqError("Can't compute nodes using"
4623 " iallocator '%s': %s" % (self.op.iallocator,
4625 if len(ial.nodes) != ial.required_nodes:
4626 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4627 " of nodes (%s), required %s" %
4628 (self.op.iallocator, len(ial.nodes),
4629 ial.required_nodes))
4630 self.op.pnode = ial.nodes[0]
4631 self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4632 self.op.instance_name, self.op.iallocator,
4633 ", ".join(ial.nodes))
4634 if ial.required_nodes == 2:
4635 self.op.snode = ial.nodes[1]
4637 def BuildHooksEnv(self):
4640 This runs on master, primary and secondary nodes of the instance.
4644 "ADD_MODE": self.op.mode,
4646 if self.op.mode == constants.INSTANCE_IMPORT:
4647 env["SRC_NODE"] = self.op.src_node
4648 env["SRC_PATH"] = self.op.src_path
4649 env["SRC_IMAGES"] = self.src_images
4651 env.update(_BuildInstanceHookEnv(
4652 name=self.op.instance_name,
4653 primary_node=self.op.pnode,
4654 secondary_nodes=self.secondaries,
4655 status=self.op.start,
4656 os_type=self.op.os_type,
4657 memory=self.be_full[constants.BE_MEMORY],
4658 vcpus=self.be_full[constants.BE_VCPUS],
4659 nics=_PreBuildNICHooksList(self, self.nics),
4660 disk_template=self.op.disk_template,
4661 disks=[(d["size"], d["mode"]) for d in self.disks],
4664 nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4669 def CheckPrereq(self):
4670 """Check prerequisites.
4673 if (not self.cfg.GetVGName() and
4674 self.op.disk_template not in constants.DTS_NOT_LVM):
4675 raise errors.OpPrereqError("Cluster does not support lvm-based"
4678 if self.op.mode == constants.INSTANCE_IMPORT:
4679 src_node = self.op.src_node
4680 src_path = self.op.src_path
4682 if src_node is None:
4683 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
4684 exp_list = self.rpc.call_export_list(locked_nodes)
4686 for node in exp_list:
4687 if exp_list[node].RemoteFailMsg():
4689 if src_path in exp_list[node].payload:
4691 self.op.src_node = src_node = node
4692 self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4696 raise errors.OpPrereqError("No export found for relative path %s" %
4699 _CheckNodeOnline(self, src_node)
4700 result = self.rpc.call_export_info(src_node, src_path)
4701 msg = result.RemoteFailMsg()
4703 raise errors.OpPrereqError("No export or invalid export found in"
4704 " dir %s: %s" % (src_path, msg))
4706 export_info = objects.SerializableConfigParser.Loads(str(result.payload))
4707 if not export_info.has_section(constants.INISECT_EXP):
4708 raise errors.ProgrammerError("Corrupted export config")
4710 ei_version = export_info.get(constants.INISECT_EXP, 'version')
4711 if (int(ei_version) != constants.EXPORT_VERSION):
4712 raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4713 (ei_version, constants.EXPORT_VERSION))
4715 # Check that the new instance doesn't have less disks than the export
4716 instance_disks = len(self.disks)
4717 export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4718 if instance_disks < export_disks:
4719 raise errors.OpPrereqError("Not enough disks to import."
4720 " (instance: %d, export: %d)" %
4721 (instance_disks, export_disks))
4723 self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4725 for idx in range(export_disks):
4726 option = 'disk%d_dump' % idx
4727 if export_info.has_option(constants.INISECT_INS, option):
4728 # FIXME: are the old os-es, disk sizes, etc. useful?
4729 export_name = export_info.get(constants.INISECT_INS, option)
4730 image = os.path.join(src_path, export_name)
4731 disk_images.append(image)
4733 disk_images.append(False)
4735 self.src_images = disk_images
4737 old_name = export_info.get(constants.INISECT_INS, 'name')
4738 # FIXME: int() here could throw a ValueError on broken exports
4739 exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4740 if self.op.instance_name == old_name:
4741 for idx, nic in enumerate(self.nics):
4742 if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4743 nic_mac_ini = 'nic%d_mac' % idx
4744 nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4746 # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
4747 # ip ping checks (we use the same ip that was resolved in ExpandNames)
4748 if self.op.start and not self.op.ip_check:
4749 raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4750 " adding an instance in start mode")
4752 if self.op.ip_check:
4753 if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4754 raise errors.OpPrereqError("IP %s of instance %s already in use" %
4755 (self.check_ip, self.op.instance_name))
4757 #### mac address generation
4758 # By generating here the mac address both the allocator and the hooks get
4759 # the real final mac address rather than the 'auto' or 'generate' value.
4760 # There is a race condition between the generation and the instance object
4761 # creation, which means that we know the mac is valid now, but we're not
4762 # sure it will be when we actually add the instance. If things go bad
4763 # adding the instance will abort because of a duplicate mac, and the
4764 # creation job will fail.
4765 for nic in self.nics:
4766 if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4767 nic.mac = self.cfg.GenerateMAC()
4771 if self.op.iallocator is not None:
4772 self._RunAllocator()
4774 #### node related checks
4776 # check primary node
4777 self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4778 assert self.pnode is not None, \
4779 "Cannot retrieve locked node %s" % self.op.pnode
4781 raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4784 raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
4787 self.secondaries = []
4789 # mirror node verification
4790 if self.op.disk_template in constants.DTS_NET_MIRROR:
4791 if self.op.snode is None:
4792 raise errors.OpPrereqError("The networked disk templates need"
4794 if self.op.snode == pnode.name:
4795 raise errors.OpPrereqError("The secondary node cannot be"
4796 " the primary node.")
4797 _CheckNodeOnline(self, self.op.snode)
4798 _CheckNodeNotDrained(self, self.op.snode)
4799 self.secondaries.append(self.op.snode)
4801 nodenames = [pnode.name] + self.secondaries
4803 req_size = _ComputeDiskSize(self.op.disk_template,
4806 # Check lv size requirements
4807 if req_size is not None:
4808 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4810 for node in nodenames:
4811 info = nodeinfo[node]
4815 raise errors.OpPrereqError("Cannot get current information"
4816 " from node '%s'" % node)
4817 vg_free = info.get('vg_free', None)
4818 if not isinstance(vg_free, int):
4819 raise errors.OpPrereqError("Can't compute free disk space on"
4821 if req_size > info['vg_free']:
4822 raise errors.OpPrereqError("Not enough disk space on target node %s."
4823 " %d MB available, %d MB required" %
4824 (node, info['vg_free'], req_size))
4826 _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4829 result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4831 if not isinstance(result.data, objects.OS):
4832 raise errors.OpPrereqError("OS '%s' not in supported os list for"
4833 " primary node" % self.op.os_type)
4835 _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
4837 # memory check on primary node
4839 _CheckNodeFreeMemory(self, self.pnode.name,
4840 "creating instance %s" % self.op.instance_name,
4841 self.be_full[constants.BE_MEMORY],
4844 def Exec(self, feedback_fn):
4845 """Create and add the instance to the cluster.
4848 instance = self.op.instance_name
4849 pnode_name = self.pnode.name
4851 ht_kind = self.op.hypervisor
4852 if ht_kind in constants.HTS_REQ_PORT:
4853 network_port = self.cfg.AllocatePort()
4857 ##if self.op.vnc_bind_address is None:
4858 ## self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4860 # this is needed because os.path.join does not accept None arguments
4861 if self.op.file_storage_dir is None:
4862 string_file_storage_dir = ""
4864 string_file_storage_dir = self.op.file_storage_dir
4866 # build the full file storage dir path
4867 file_storage_dir = os.path.normpath(os.path.join(
4868 self.cfg.GetFileStorageDir(),
4869 string_file_storage_dir, instance))
4872 disks = _GenerateDiskTemplate(self,
4873 self.op.disk_template,
4874 instance, pnode_name,
4878 self.op.file_driver,
4881 iobj = objects.Instance(name=instance, os=self.op.os_type,
4882 primary_node=pnode_name,
4883 nics=self.nics, disks=disks,
4884 disk_template=self.op.disk_template,
4886 network_port=network_port,
4887 beparams=self.op.beparams,
4888 hvparams=self.op.hvparams,
4889 hypervisor=self.op.hypervisor,
4892 feedback_fn("* creating instance disks...")
4894 _CreateDisks(self, iobj)
4895 except errors.OpExecError:
4896 self.LogWarning("Device creation failed, reverting...")
4898 _RemoveDisks(self, iobj)
4900 self.cfg.ReleaseDRBDMinors(instance)
4903 feedback_fn("adding instance %s to cluster config" % instance)
4905 self.cfg.AddInstance(iobj)
4906 # Declare that we don't want to remove the instance lock anymore, as we've
4907 # added the instance to the config
4908 del self.remove_locks[locking.LEVEL_INSTANCE]
4909 # Unlock all the nodes
4910 if self.op.mode == constants.INSTANCE_IMPORT:
4911 nodes_keep = [self.op.src_node]
4912 nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4913 if node != self.op.src_node]
4914 self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4915 self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4917 self.context.glm.release(locking.LEVEL_NODE)
4918 del self.acquired_locks[locking.LEVEL_NODE]
4920 if self.op.wait_for_sync:
4921 disk_abort = not _WaitForSync(self, iobj)
4922 elif iobj.disk_template in constants.DTS_NET_MIRROR:
4923 # make sure the disks are not degraded (still sync-ing is ok)
4925 feedback_fn("* checking mirrors status")
4926 disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4931 _RemoveDisks(self, iobj)
4932 self.cfg.RemoveInstance(iobj.name)
4933 # Make sure the instance lock gets removed
4934 self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4935 raise errors.OpExecError("There are some degraded disks for"
4938 feedback_fn("creating os for instance %s on node %s" %
4939 (instance, pnode_name))
4941 if iobj.disk_template != constants.DT_DISKLESS:
4942 if self.op.mode == constants.INSTANCE_CREATE:
4943 feedback_fn("* running the instance OS create scripts...")
4944 result = self.rpc.call_instance_os_add(pnode_name, iobj, False)
4945 msg = result.RemoteFailMsg()
4947 raise errors.OpExecError("Could not add os for instance %s"
4949 (instance, pnode_name, msg))
4951 elif self.op.mode == constants.INSTANCE_IMPORT:
4952 feedback_fn("* running the instance OS import scripts...")
4953 src_node = self.op.src_node
4954 src_images = self.src_images
4955 cluster_name = self.cfg.GetClusterName()
4956 import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4957 src_node, src_images,
4959 msg = import_result.RemoteFailMsg()
4961 self.LogWarning("Error while importing the disk images for instance"
4962 " %s on node %s: %s" % (instance, pnode_name, msg))
4964 # also checked in the prereq part
4965 raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4969 iobj.admin_up = True
4970 self.cfg.Update(iobj)
4971 logging.info("Starting instance %s on node %s", instance, pnode_name)
4972 feedback_fn("* starting instance...")
4973 result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
4974 msg = result.RemoteFailMsg()
4976 raise errors.OpExecError("Could not start instance: %s" % msg)
4979 class LUConnectConsole(NoHooksLU):
4980 """Connect to an instance's console.
4982 This is somewhat special in that it returns the command line that
4983 you need to run on the master node in order to connect to the
4987 _OP_REQP = ["instance_name"]
4990 def ExpandNames(self):
4991 self._ExpandAndLockInstance()
4993 def CheckPrereq(self):
4994 """Check prerequisites.
4996 This checks that the instance is in the cluster.
4999 self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5000 assert self.instance is not None, \
5001 "Cannot retrieve locked instance %s" % self.op.instance_name
5002 _CheckNodeOnline(self, self.instance.primary_node)
5004 def Exec(self, feedback_fn):
5005 """Connect to the console of an instance
5008 instance = self.instance
5009 node = instance.primary_node
5011 node_insts = self.rpc.call_instance_list([node],
5012 [instance.hypervisor])[node]
5015 if instance.name not in node_insts.data:
5016 raise errors.OpExecError("Instance %s is not running." % instance.name)
5018 logging.debug("Connecting to console of %s on %s", instance.name, node)
5020 hyper = hypervisor.GetHypervisor(instance.hypervisor)
5021 cluster = self.cfg.GetClusterInfo()
5022 # beparams and hvparams are passed separately, to avoid editing the
5023 # instance and then saving the defaults in the instance itself.
5024 hvparams = cluster.FillHV(instance)
5025 beparams = cluster.FillBE(instance)
5026 console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
5029 return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
5032 class LUReplaceDisks(LogicalUnit):
5033 """Replace the disks of an instance.
5036 HPATH = "mirrors-replace"
5037 HTYPE = constants.HTYPE_INSTANCE
5038 _OP_REQP = ["instance_name", "mode", "disks"]
5041 def CheckArguments(self):
5042 if not hasattr(self.op, "remote_node"):
5043 self.op.remote_node = None
5044 if not hasattr(self.op, "iallocator"):
5045 self.op.iallocator = None
5047 # check for valid parameter combination
5048 cnt = [self.op.remote_node, self.op.iallocator].count(None)
5049 if self.op.mode == constants.REPLACE_DISK_CHG:
5051 raise errors.OpPrereqError("When changing the secondary either an"
5052 " iallocator script must be used or the"
5055 raise errors.OpPrereqError("Give either the iallocator or the new"
5056 " secondary, not both")
5057 else: # not replacing the secondary
5059 raise errors.OpPrereqError("The iallocator and new node options can"
5060 " be used only when changing the"
5063 def ExpandNames(self):
5064 self._ExpandAndLockInstance()
5066 if self.op.iallocator is not None:
5067 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5068 elif self.op.remote_node is not None:
5069 remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
5070 if remote_node is None:
5071 raise errors.OpPrereqError("Node '%s' not known" %
5072 self.op.remote_node)
5073 self.op.remote_node = remote_node
5074 # Warning: do not remove the locking of the new secondary here
5075 # unless DRBD8.AddChildren is changed to work in parallel;
5076 # currently it doesn't since parallel invocations of
5077 # FindUnusedMinor will conflict
5078 self.needed_locks[locking.LEVEL_NODE] = [remote_node]
5079 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5081 self.needed_locks[locking.LEVEL_NODE] = []
5082 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5084 def DeclareLocks(self, level):
5085 # If we're not already locking all nodes in the set we have to declare the
5086 # instance's primary/secondary nodes.
5087 if (level == locking.LEVEL_NODE and
5088 self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
5089 self._LockInstancesNodes()
5091 def _RunAllocator(self):
5092 """Compute a new secondary node using an IAllocator.
5095 ial = IAllocator(self,
5096 mode=constants.IALLOCATOR_MODE_RELOC,
5097 name=self.op.instance_name,
5098 relocate_from=[self.sec_node])
5100 ial.Run(self.op.iallocator)
5103 raise errors.OpPrereqError("Can't compute nodes using"
5104 " iallocator '%s': %s" % (self.op.iallocator,
5106 if len(ial.nodes) != ial.required_nodes:
5107 raise errors.OpPrereqError("iallocator '%s' returned invalid number"
5108 " of nodes (%s), required %s" %
5109 (len(ial.nodes), ial.required_nodes))
5110 self.op.remote_node = ial.nodes[0]
5111 self.LogInfo("Selected new secondary for the instance: %s",
5112 self.op.remote_node)
5114 def BuildHooksEnv(self):
5117 This runs on the master, the primary and all the secondaries.
5121 "MODE": self.op.mode,
5122 "NEW_SECONDARY": self.op.remote_node,
5123 "OLD_SECONDARY": self.instance.secondary_nodes[0],
5125 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5127 self.cfg.GetMasterNode(),
5128 self.instance.primary_node,
5130 if self.op.remote_node is not None:
5131 nl.append(self.op.remote_node)
5134 def CheckPrereq(self):
5135 """Check prerequisites.
5137 This checks that the instance is in the cluster.
5140 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5141 assert instance is not None, \
5142 "Cannot retrieve locked instance %s" % self.op.instance_name
5143 self.instance = instance
5145 if instance.disk_template != constants.DT_DRBD8:
5146 raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
5149 if len(instance.secondary_nodes) != 1:
5150 raise errors.OpPrereqError("The instance has a strange layout,"
5151 " expected one secondary but found %d" %
5152 len(instance.secondary_nodes))
5154 self.sec_node = instance.secondary_nodes[0]
5156 if self.op.iallocator is not None:
5157 self._RunAllocator()
5159 remote_node = self.op.remote_node
5160 if remote_node is not None:
5161 self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
5162 assert self.remote_node_info is not None, \
5163 "Cannot retrieve locked node %s" % remote_node
5165 self.remote_node_info = None
5166 if remote_node == instance.primary_node:
5167 raise errors.OpPrereqError("The specified node is the primary node of"
5169 elif remote_node == self.sec_node:
5170 raise errors.OpPrereqError("The specified node is already the"
5171 " secondary node of the instance.")
5173 if self.op.mode == constants.REPLACE_DISK_PRI:
5174 n1 = self.tgt_node = instance.primary_node
5175 n2 = self.oth_node = self.sec_node
5176 elif self.op.mode == constants.REPLACE_DISK_SEC:
5177 n1 = self.tgt_node = self.sec_node
5178 n2 = self.oth_node = instance.primary_node
5179 elif self.op.mode == constants.REPLACE_DISK_CHG:
5180 n1 = self.new_node = remote_node
5181 n2 = self.oth_node = instance.primary_node
5182 self.tgt_node = self.sec_node
5183 _CheckNodeNotDrained(self, remote_node)
5185 raise errors.ProgrammerError("Unhandled disk replace mode")
5187 _CheckNodeOnline(self, n1)
5188 _CheckNodeOnline(self, n2)
5190 if not self.op.disks:
5191 self.op.disks = range(len(instance.disks))
5193 for disk_idx in self.op.disks:
5194 instance.FindDisk(disk_idx)
5196 def _ExecD8DiskOnly(self, feedback_fn):
5197 """Replace a disk on the primary or secondary for dbrd8.
5199 The algorithm for replace is quite complicated:
5201 1. for each disk to be replaced:
5203 1. create new LVs on the target node with unique names
5204 1. detach old LVs from the drbd device
5205 1. rename old LVs to name_replaced.<time_t>
5206 1. rename new LVs to old LVs
5207 1. attach the new LVs (with the old names now) to the drbd device
5209 1. wait for sync across all devices
5211 1. for each modified disk:
5213 1. remove old LVs (which have the name name_replaces.<time_t>)
5215 Failures are not very well handled.
5219 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5220 instance = self.instance
5222 vgname = self.cfg.GetVGName()
5225 tgt_node = self.tgt_node
5226 oth_node = self.oth_node
5228 # Step: check device activation
5229 self.proc.LogStep(1, steps_total, "check device existence")
5230 info("checking volume groups")
5231 my_vg = cfg.GetVGName()
5232 results = self.rpc.call_vg_list([oth_node, tgt_node])
5234 raise errors.OpExecError("Can't list volume groups on the nodes")
5235 for node in oth_node, tgt_node:
5237 msg = res.RemoteFailMsg()
5239 raise errors.OpExecError("Error checking node %s: %s" % (node, msg))
5240 if my_vg not in res.payload:
5241 raise errors.OpExecError("Volume group '%s' not found on %s" %
5243 for idx, dev in enumerate(instance.disks):
5244 if idx not in self.op.disks:
5246 for node in tgt_node, oth_node:
5247 info("checking disk/%d on %s" % (idx, node))
5248 cfg.SetDiskID(dev, node)
5249 result = self.rpc.call_blockdev_find(node, dev)
5250 msg = result.RemoteFailMsg()
5251 if not msg and not result.payload:
5252 msg = "disk not found"
5254 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5257 # Step: check other node consistency
5258 self.proc.LogStep(2, steps_total, "check peer consistency")
5259 for idx, dev in enumerate(instance.disks):
5260 if idx not in self.op.disks:
5262 info("checking disk/%d consistency on %s" % (idx, oth_node))
5263 if not _CheckDiskConsistency(self, dev, oth_node,
5264 oth_node==instance.primary_node):
5265 raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
5266 " to replace disks on this node (%s)" %
5267 (oth_node, tgt_node))
5269 # Step: create new storage
5270 self.proc.LogStep(3, steps_total, "allocate new storage")
5271 for idx, dev in enumerate(instance.disks):
5272 if idx not in self.op.disks:
5275 cfg.SetDiskID(dev, tgt_node)
5276 lv_names = [".disk%d_%s" % (idx, suf)
5277 for suf in ["data", "meta"]]
5278 names = _GenerateUniqueNames(self, lv_names)
5279 lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5280 logical_id=(vgname, names[0]))
5281 lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5282 logical_id=(vgname, names[1]))
5283 new_lvs = [lv_data, lv_meta]
5284 old_lvs = dev.children
5285 iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
5286 info("creating new local storage on %s for %s" %
5287 (tgt_node, dev.iv_name))
5288 # we pass force_create=True to force the LVM creation
5289 for new_lv in new_lvs:
5290 _CreateBlockDev(self, tgt_node, instance, new_lv, True,
5291 _GetInstanceInfoText(instance), False)
5293 # Step: for each lv, detach+rename*2+attach
5294 self.proc.LogStep(4, steps_total, "change drbd configuration")
5295 for dev, old_lvs, new_lvs in iv_names.itervalues():
5296 info("detaching %s drbd from local storage" % dev.iv_name)
5297 result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
5298 msg = result.RemoteFailMsg()
5300 raise errors.OpExecError("Can't detach drbd from local storage on node"
5301 " %s for device %s: %s" %
5302 (tgt_node, dev.iv_name, msg))
5304 #cfg.Update(instance)
5306 # ok, we created the new LVs, so now we know we have the needed
5307 # storage; as such, we proceed on the target node to rename
5308 # old_lv to _old, and new_lv to old_lv; note that we rename LVs
5309 # using the assumption that logical_id == physical_id (which in
5310 # turn is the unique_id on that node)
5312 # FIXME(iustin): use a better name for the replaced LVs
5313 temp_suffix = int(time.time())
5314 ren_fn = lambda d, suff: (d.physical_id[0],
5315 d.physical_id[1] + "_replaced-%s" % suff)
5316 # build the rename list based on what LVs exist on the node
5318 for to_ren in old_lvs:
5319 result = self.rpc.call_blockdev_find(tgt_node, to_ren)
5320 if not result.RemoteFailMsg() and result.payload:
5322 rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
5324 info("renaming the old LVs on the target node")
5325 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5326 msg = result.RemoteFailMsg()
5328 raise errors.OpExecError("Can't rename old LVs on node %s: %s" %
5330 # now we rename the new LVs to the old LVs
5331 info("renaming the new LVs on the target node")
5332 rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
5333 result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5334 msg = result.RemoteFailMsg()
5336 raise errors.OpExecError("Can't rename new LVs on node %s: %s" %
5339 for old, new in zip(old_lvs, new_lvs):
5340 new.logical_id = old.logical_id
5341 cfg.SetDiskID(new, tgt_node)
5343 for disk in old_lvs:
5344 disk.logical_id = ren_fn(disk, temp_suffix)
5345 cfg.SetDiskID(disk, tgt_node)
5347 # now that the new lvs have the old name, we can add them to the device
5348 info("adding new mirror component on %s" % tgt_node)
5349 result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
5350 msg = result.RemoteFailMsg()
5352 for new_lv in new_lvs:
5353 msg = self.rpc.call_blockdev_remove(tgt_node, new_lv).RemoteFailMsg()
5355 warning("Can't rollback device %s: %s", dev, msg,
5356 hint="cleanup manually the unused logical volumes")
5357 raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
5359 dev.children = new_lvs
5360 cfg.Update(instance)
5362 # Step: wait for sync
5364 # this can fail as the old devices are degraded and _WaitForSync
5365 # does a combined result over all disks, so we don't check its
5367 self.proc.LogStep(5, steps_total, "sync devices")
5368 _WaitForSync(self, instance, unlock=True)
5370 # so check manually all the devices
5371 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5372 cfg.SetDiskID(dev, instance.primary_node)
5373 result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5374 msg = result.RemoteFailMsg()
5375 if not msg and not result.payload:
5376 msg = "disk not found"
5378 raise errors.OpExecError("Can't find DRBD device %s: %s" %
5380 if result.payload[5]:
5381 raise errors.OpExecError("DRBD device %s is degraded!" % name)
5383 # Step: remove old storage
5384 self.proc.LogStep(6, steps_total, "removing old storage")
5385 for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5386 info("remove logical volumes for %s" % name)
5388 cfg.SetDiskID(lv, tgt_node)
5389 msg = self.rpc.call_blockdev_remove(tgt_node, lv).RemoteFailMsg()
5391 warning("Can't remove old LV: %s" % msg,
5392 hint="manually remove unused LVs")
5395 def _ExecD8Secondary(self, feedback_fn):
5396 """Replace the secondary node for drbd8.
5398 The algorithm for replace is quite complicated:
5399 - for all disks of the instance:
5400 - create new LVs on the new node with same names
5401 - shutdown the drbd device on the old secondary
5402 - disconnect the drbd network on the primary
5403 - create the drbd device on the new secondary
5404 - network attach the drbd on the primary, using an artifice:
5405 the drbd code for Attach() will connect to the network if it
5406 finds a device which is connected to the good local disks but
5408 - wait for sync across all devices
5409 - remove all disks from the old secondary
5411 Failures are not very well handled.
5415 warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5416 instance = self.instance
5420 old_node = self.tgt_node
5421 new_node = self.new_node
5422 pri_node = instance.primary_node
5424 old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5425 new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5426 pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5429 # Step: check device activation
5430 self.proc.LogStep(1, steps_total, "check device existence")
5431 info("checking volume groups")
5432 my_vg = cfg.GetVGName()
5433 results = self.rpc.call_vg_list([pri_node, new_node])
5434 for node in pri_node, new_node:
5436 msg = res.RemoteFailMsg()
5438 raise errors.OpExecError("Error checking node %s: %s" % (node, msg))
5439 if my_vg not in res.payload:
5440 raise errors.OpExecError("Volume group '%s' not found on %s" %
5442 for idx, dev in enumerate(instance.disks):
5443 if idx not in self.op.disks:
5445 info("checking disk/%d on %s" % (idx, pri_node))
5446 cfg.SetDiskID(dev, pri_node)
5447 result = self.rpc.call_blockdev_find(pri_node, dev)
5448 msg = result.RemoteFailMsg()
5449 if not msg and not result.payload:
5450 msg = "disk not found"
5452 raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5453 (idx, pri_node, msg))
5455 # Step: check other node consistency
5456 self.proc.LogStep(2, steps_total, "check peer consistency")
5457 for idx, dev in enumerate(instance.disks):
5458 if idx not in self.op.disks:
5460 info("checking disk/%d consistency on %s" % (idx, pri_node))
5461 if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5462 raise errors.OpExecError("Primary node (%s) has degraded storage,"
5463 " unsafe to replace the secondary" %
5466 # Step: create new storage
5467 self.proc.LogStep(3, steps_total, "allocate new storage")
5468 for idx, dev in enumerate(instance.disks):
5469 info("adding new local storage on %s for disk/%d" %
5471 # we pass force_create=True to force LVM creation
5472 for new_lv in dev.children:
5473 _CreateBlockDev(self, new_node, instance, new_lv, True,
5474 _GetInstanceInfoText(instance), False)
5476 # Step 4: dbrd minors and drbd setups changes
5477 # after this, we must manually remove the drbd minors on both the
5478 # error and the success paths
5479 minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5481 logging.debug("Allocated minors %s" % (minors,))
5482 self.proc.LogStep(4, steps_total, "changing drbd configuration")
5483 for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5485 info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5486 # create new devices on new_node; note that we create two IDs:
5487 # one without port, so the drbd will be activated without
5488 # networking information on the new node at this stage, and one
5489 # with network, for the latter activation in step 4
5490 (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5491 if pri_node == o_node1:
5496 new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5497 new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5499 iv_names[idx] = (dev, dev.children, new_net_id)
5500 logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5502 new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5503 logical_id=new_alone_id,
5504 children=dev.children)
5506 _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5507 _GetInstanceInfoText(instance), False)
5508 except errors.GenericError:
5509 self.cfg.ReleaseDRBDMinors(instance.name)
5512 for idx, dev in enumerate(instance.disks):
5513 # we have new devices, shutdown the drbd on the old secondary
5514 info("shutting down drbd for disk/%d on old node" % idx)
5515 cfg.SetDiskID(dev, old_node)
5516 msg = self.rpc.call_blockdev_shutdown(old_node, dev).RemoteFailMsg()
5518 warning("Failed to shutdown drbd for disk/%d on old node: %s" %
5520 hint="Please cleanup this device manually as soon as possible")
5522 info("detaching primary drbds from the network (=> standalone)")
5523 result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5524 instance.disks)[pri_node]
5526 msg = result.RemoteFailMsg()
5528 # detaches didn't succeed (unlikely)
5529 self.cfg.ReleaseDRBDMinors(instance.name)
5530 raise errors.OpExecError("Can't detach the disks from the network on"
5531 " old node: %s" % (msg,))
5533 # if we managed to detach at least one, we update all the disks of
5534 # the instance to point to the new secondary
5535 info("updating instance configuration")
5536 for dev, _, new_logical_id in iv_names.itervalues():
5537 dev.logical_id = new_logical_id
5538 cfg.SetDiskID(dev, pri_node)
5539 cfg.Update(instance)
5541 # and now perform the drbd attach
5542 info("attaching primary drbds to new secondary (standalone => connected)")
5543 result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5544 instance.disks, instance.name,
5546 for to_node, to_result in result.items():
5547 msg = to_result.RemoteFailMsg()
5549 warning("can't attach drbd disks on node %s: %s", to_node, msg,
5550 hint="please do a gnt-instance info to see the"
5553 # this can fail as the old devices are degraded and _WaitForSync
5554 # does a combined result over all disks, so we don't check its
5556 self.proc.LogStep(5, steps_total, "sync devices")
5557 _WaitForSync(self, instance, unlock=True)
5559 # so check manually all the devices
5560 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5561 cfg.SetDiskID(dev, pri_node)
5562 result = self.rpc.call_blockdev_find(pri_node, dev)
5563 msg = result.RemoteFailMsg()
5564 if not msg and not result.payload:
5565 msg = "disk not found"
5567 raise errors.OpExecError("Can't find DRBD device disk/%d: %s" %
5569 if result.payload[5]:
5570 raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5572 self.proc.LogStep(6, steps_total, "removing old storage")
5573 for idx, (dev, old_lvs, _) in iv_names.iteritems():
5574 info("remove logical volumes for disk/%d" % idx)
5576 cfg.SetDiskID(lv, old_node)
5577 msg = self.rpc.call_blockdev_remove(old_node, lv).RemoteFailMsg()
5579 warning("Can't remove LV on old secondary: %s", msg,
5580 hint="Cleanup stale volumes by hand")
5582 def Exec(self, feedback_fn):
5583 """Execute disk replacement.
5585 This dispatches the disk replacement to the appropriate handler.
5588 instance = self.instance
5590 # Activate the instance disks if we're replacing them on a down instance
5591 if not instance.admin_up:
5592 _StartInstanceDisks(self, instance, True)
5594 if self.op.mode == constants.REPLACE_DISK_CHG:
5595 fn = self._ExecD8Secondary
5597 fn = self._ExecD8DiskOnly
5599 ret = fn(feedback_fn)
5601 # Deactivate the instance disks if we're replacing them on a down instance
5602 if not instance.admin_up:
5603 _SafeShutdownInstanceDisks(self, instance)
5608 class LUGrowDisk(LogicalUnit):
5609 """Grow a disk of an instance.
5613 HTYPE = constants.HTYPE_INSTANCE
5614 _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5617 def ExpandNames(self):
5618 self._ExpandAndLockInstance()
5619 self.needed_locks[locking.LEVEL_NODE] = []
5620 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5622 def DeclareLocks(self, level):
5623 if level == locking.LEVEL_NODE:
5624 self._LockInstancesNodes()
5626 def BuildHooksEnv(self):
5629 This runs on the master, the primary and all the secondaries.
5633 "DISK": self.op.disk,
5634 "AMOUNT": self.op.amount,
5636 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5638 self.cfg.GetMasterNode(),
5639 self.instance.primary_node,
5643 def CheckPrereq(self):
5644 """Check prerequisites.
5646 This checks that the instance is in the cluster.
5649 instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5650 assert instance is not None, \
5651 "Cannot retrieve locked instance %s" % self.op.instance_name
5652 nodenames = list(instance.all_nodes)
5653 for node in nodenames:
5654 _CheckNodeOnline(self, node)
5657 self.instance = instance
5659 if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5660 raise errors.OpPrereqError("Instance's disk layout does not support"
5663 self.disk = instance.FindDisk(self.op.disk)
5665 nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5666 instance.hypervisor)
5667 for node in nodenames:
5668 info = nodeinfo[node]
5669 if info.failed or not info.data:
5670 raise errors.OpPrereqError("Cannot get current information"
5671 " from node '%s'" % node)
5672 vg_free = info.data.get('vg_free', None)
5673 if not isinstance(vg_free, int):
5674 raise errors.OpPrereqError("Can't compute free disk space on"
5676 if self.op.amount > vg_free:
5677 raise errors.OpPrereqError("Not enough disk space on target node %s:"
5678 " %d MiB available, %d MiB required" %
5679 (node, vg_free, self.op.amount))
5681 def Exec(self, feedback_fn):
5682 """Execute disk grow.
5685 instance = self.instance
5687 for node in instance.all_nodes:
5688 self.cfg.SetDiskID(disk, node)
5689 result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5690 msg = result.RemoteFailMsg()
5692 raise errors.OpExecError("Grow request failed to node %s: %s" %
5694 disk.RecordGrow(self.op.amount)
5695 self.cfg.Update(instance)
5696 if self.op.wait_for_sync:
5697 disk_abort = not _WaitForSync(self, instance)
5699 self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5700 " status.\nPlease check the instance.")
5703 class LUQueryInstanceData(NoHooksLU):
5704 """Query runtime instance data.
5707 _OP_REQP = ["instances", "static"]
5710 def ExpandNames(self):
5711 self.needed_locks = {}
5712 self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5714 if not isinstance(self.op.instances, list):
5715 raise errors.OpPrereqError("Invalid argument type 'instances'")
5717 if self.op.instances:
5718 self.wanted_names = []
5719 for name in self.op.instances:
5720 full_name = self.cfg.ExpandInstanceName(name)
5721 if full_name is None:
5722 raise errors.OpPrereqError("Instance '%s' not known" % name)
5723 self.wanted_names.append(full_name)
5724 self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5726 self.wanted_names = None
5727 self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5729 self.needed_locks[locking.LEVEL_NODE] = []
5730 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5732 def DeclareLocks(self, level):
5733 if level == locking.LEVEL_NODE:
5734 self._LockInstancesNodes()
5736 def CheckPrereq(self):
5737 """Check prerequisites.
5739 This only checks the optional instance list against the existing names.
5742 if self.wanted_names is None:
5743 self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5745 self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5746 in self.wanted_names]
5749 def _ComputeDiskStatus(self, instance, snode, dev):
5750 """Compute block device status.
5753 static = self.op.static
5755 self.cfg.SetDiskID(dev, instance.primary_node)
5756 dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5757 if dev_pstatus.offline:
5760 msg = dev_pstatus.RemoteFailMsg()
5762 raise errors.OpExecError("Can't compute disk status for %s: %s" %
5763 (instance.name, msg))
5764 dev_pstatus = dev_pstatus.payload
5768 if dev.dev_type in constants.LDS_DRBD:
5769 # we change the snode then (otherwise we use the one passed in)
5770 if dev.logical_id[0] == instance.primary_node:
5771 snode = dev.logical_id[1]
5773 snode = dev.logical_id[0]
5775 if snode and not static:
5776 self.cfg.SetDiskID(dev, snode)
5777 dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5778 if dev_sstatus.offline:
5781 msg = dev_sstatus.RemoteFailMsg()
5783 raise errors.OpExecError("Can't compute disk status for %s: %s" %
5784 (instance.name, msg))
5785 dev_sstatus = dev_sstatus.payload
5790 dev_children = [self._ComputeDiskStatus(instance, snode, child)
5791 for child in dev.children]
5796 "iv_name": dev.iv_name,
5797 "dev_type": dev.dev_type,
5798 "logical_id": dev.logical_id,
5799 "physical_id": dev.physical_id,
5800 "pstatus": dev_pstatus,
5801 "sstatus": dev_sstatus,
5802 "children": dev_children,
5808 def Exec(self, feedback_fn):
5809 """Gather and return data"""
5812 cluster = self.cfg.GetClusterInfo()
5814 for instance in self.wanted_instances:
5815 if not self.op.static:
5816 remote_info = self.rpc.call_instance_info(instance.primary_node,
5818 instance.hypervisor)
5819 msg = remote_info.RemoteFailMsg()
5821 raise errors.OpExecError("Error checking node %s: %s" %
5822 (instance.primary_node, msg))
5823 remote_info = remote_info.payload
5824 if remote_info and "state" in remote_info:
5827 remote_state = "down"
5830 if instance.admin_up:
5833 config_state = "down"
5835 disks = [self._ComputeDiskStatus(instance, None, device)
5836 for device in instance.disks]
5839 "name": instance.name,
5840 "config_state": config_state,
5841 "run_state": remote_state,
5842 "pnode": instance.primary_node,
5843 "snodes": instance.secondary_nodes,
5845 "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5847 "hypervisor": instance.hypervisor,
5848 "network_port": instance.network_port,
5849 "hv_instance": instance.hvparams,
5850 "hv_actual": cluster.FillHV(instance),
5851 "be_instance": instance.beparams,
5852 "be_actual": cluster.FillBE(instance),
5855 result[instance.name] = idict
5860 class LUSetInstanceParams(LogicalUnit):
5861 """Modifies an instances's parameters.
5864 HPATH = "instance-modify"
5865 HTYPE = constants.HTYPE_INSTANCE
5866 _OP_REQP = ["instance_name"]
5869 def CheckArguments(self):
5870 if not hasattr(self.op, 'nics'):
5872 if not hasattr(self.op, 'disks'):
5874 if not hasattr(self.op, 'beparams'):
5875 self.op.beparams = {}
5876 if not hasattr(self.op, 'hvparams'):
5877 self.op.hvparams = {}
5878 self.op.force = getattr(self.op, "force", False)
5879 if not (self.op.nics or self.op.disks or
5880 self.op.hvparams or self.op.beparams):
5881 raise errors.OpPrereqError("No changes submitted")
5885 for disk_op, disk_dict in self.op.disks:
5886 if disk_op == constants.DDM_REMOVE:
5889 elif disk_op == constants.DDM_ADD:
5892 if not isinstance(disk_op, int):
5893 raise errors.OpPrereqError("Invalid disk index")
5894 if disk_op == constants.DDM_ADD:
5895 mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5896 if mode not in constants.DISK_ACCESS_SET:
5897 raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5898 size = disk_dict.get('size', None)
5900 raise errors.OpPrereqError("Required disk parameter size missing")
5903 except ValueError, err:
5904 raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5906 disk_dict['size'] = size
5908 # modification of disk
5909 if 'size' in disk_dict:
5910 raise errors.OpPrereqError("Disk size change not possible, use"
5913 if disk_addremove > 1:
5914 raise errors.OpPrereqError("Only one disk add or remove operation"
5915 " supported at a time")
5919 for nic_op, nic_dict in self.op.nics:
5920 if nic_op == constants.DDM_REMOVE:
5923 elif nic_op == constants.DDM_ADD:
5926 if not isinstance(nic_op, int):
5927 raise errors.OpPrereqError("Invalid nic index")
5929 # nic_dict should be a dict
5930 nic_ip = nic_dict.get('ip', None)
5931 if nic_ip is not None:
5932 if nic_ip.lower() == constants.VALUE_NONE:
5933 nic_dict['ip'] = None
5935 if not utils.IsValidIP(nic_ip):
5936 raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5938 nic_bridge = nic_dict.get('bridge', None)
5939 nic_link = nic_dict.get('link', None)
5940 if nic_bridge and nic_link:
5941 raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
5942 elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
5943 nic_dict['bridge'] = None
5944 elif nic_link and nic_link.lower() == constants.VALUE_NONE:
5945 nic_dict['link'] = None
5947 if nic_op == constants.DDM_ADD:
5948 nic_mac = nic_dict.get('mac', None)
5950 nic_dict['mac'] = constants.VALUE_AUTO
5952 if 'mac' in nic_dict:
5953 nic_mac = nic_dict['mac']
5954 if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5955 if not utils.IsValidMac(nic_mac):
5956 raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5957 if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
5958 raise errors.OpPrereqError("'auto' is not a valid MAC address when"
5959 " modifying an existing nic")
5961 if nic_addremove > 1:
5962 raise errors.OpPrereqError("Only one NIC add or remove operation"
5963 " supported at a time")
5965 def ExpandNames(self):
5966 self._ExpandAndLockInstance()
5967 self.needed_locks[locking.LEVEL_NODE] = []
5968 self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5970 def DeclareLocks(self, level):
5971 if level == locking.LEVEL_NODE:
5972 self._LockInstancesNodes()
5974 def BuildHooksEnv(self):
5977 This runs on the master, primary and secondaries.
5981 if constants.BE_MEMORY in self.be_new:
5982 args['memory'] = self.be_new[constants.BE_MEMORY]
5983 if constants.BE_VCPUS in self.be_new:
5984 args['vcpus'] = self.be_new[constants.BE_VCPUS]
5985 # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
5986 # information at all.
5989 nic_override = dict(self.op.nics)
5990 c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
5991 for idx, nic in enumerate(self.instance.nics):
5992 if idx in nic_override:
5993 this_nic_override = nic_override[idx]
5995 this_nic_override = {}
5996 if 'ip' in this_nic_override:
5997 ip = this_nic_override['ip']
6000 if 'mac' in this_nic_override:
6001 mac = this_nic_override['mac']
6004 if idx in self.nic_pnew:
6005 nicparams = self.nic_pnew[idx]
6007 nicparams = objects.FillDict(c_nicparams, nic.nicparams)
6008 mode = nicparams[constants.NIC_MODE]
6009 link = nicparams[constants.NIC_LINK]
6010 args['nics'].append((ip, mac, mode, link))
6011 if constants.DDM_ADD in nic_override:
6012 ip = nic_override[constants.DDM_ADD].get('ip', None)
6013 mac = nic_override[constants.DDM_ADD]['mac']
6014 nicparams = self.nic_pnew[constants.DDM_ADD]
6015 mode = nicparams[constants.NIC_MODE]
6016 link = nicparams[constants.NIC_LINK]
6017 args['nics'].append((ip, mac, mode, link))
6018 elif constants.DDM_REMOVE in nic_override:
6019 del args['nics'][-1]
6021 env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
6022 nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
6025 def _GetUpdatedParams(self, old_params, update_dict,
6026 default_values, parameter_types):
6027 """Return the new params dict for the given params.
6029 @type old_params: dict
6030 @type old_params: old parameters
6031 @type update_dict: dict
6032 @type update_dict: dict containing new parameter values,
6033 or constants.VALUE_DEFAULT to reset the
6034 parameter to its default value
6035 @type default_values: dict
6036 @param default_values: default values for the filled parameters
6037 @type parameter_types: dict
6038 @param parameter_types: dict mapping target dict keys to types
6039 in constants.ENFORCEABLE_TYPES
6040 @rtype: (dict, dict)
6041 @return: (new_parameters, filled_parameters)
6044 params_copy = copy.deepcopy(old_params)
6045 for key, val in update_dict.iteritems():
6046 if val == constants.VALUE_DEFAULT:
6048 del params_copy[key]
6052 params_copy[key] = val
6053 utils.ForceDictType(params_copy, parameter_types)
6054 params_filled = objects.FillDict(default_values, params_copy)
6055 return (params_copy, params_filled)
6057 def CheckPrereq(self):
6058 """Check prerequisites.
6060 This only checks the instance list against the existing names.
6063 force = self.force = self.op.force
6065 # checking the new params on the primary/secondary nodes
6067 instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6068 cluster = self.cluster = self.cfg.GetClusterInfo()
6069 assert self.instance is not None, \
6070 "Cannot retrieve locked instance %s" % self.op.instance_name
6071 pnode = instance.primary_node
6072 nodelist = list(instance.all_nodes)
6074 # hvparams processing
6075 if self.op.hvparams:
6076 i_hvdict, hv_new = self._GetUpdatedParams(
6077 instance.hvparams, self.op.hvparams,
6078 cluster.hvparams[instance.hypervisor],
6079 constants.HVS_PARAMETER_TYPES)
6081 hypervisor.GetHypervisor(
6082 instance.hypervisor).CheckParameterSyntax(hv_new)
6083 _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
6084 self.hv_new = hv_new # the new actual values
6085 self.hv_inst = i_hvdict # the new dict (without defaults)
6087 self.hv_new = self.hv_inst = {}
6089 # beparams processing
6090 if self.op.beparams:
6091 i_bedict, be_new = self._GetUpdatedParams(
6092 instance.beparams, self.op.beparams,
6093 cluster.beparams[constants.PP_DEFAULT],
6094 constants.BES_PARAMETER_TYPES)
6095 self.be_new = be_new # the new actual values
6096 self.be_inst = i_bedict # the new dict (without defaults)
6098 self.be_new = self.be_inst = {}
6102 if constants.BE_MEMORY in self.op.beparams and not self.force:
6103 mem_check_list = [pnode]
6104 if be_new[constants.BE_AUTO_BALANCE]:
6105 # either we changed auto_balance to yes or it was from before
6106 mem_check_list.extend(instance.secondary_nodes)
6107 instance_info = self.rpc.call_instance_info(pnode, instance.name,
6108 instance.hypervisor)
6109 nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
6110 instance.hypervisor)
6111 if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
6112 # Assume the primary node is unreachable and go ahead
6113 self.warn.append("Can't get info from primary node %s" % pnode)
6114 elif instance_info.RemoteFailMsg():
6115 self.warn.append("Can't get instance runtime information: %s" %
6116 instance_info.RemoteFailMsg())
6118 if instance_info.payload:
6119 current_mem = int(instance_info.payload['memory'])
6121 # Assume instance not running
6122 # (there is a slight race condition here, but it's not very probable,
6123 # and we have no other way to check)
6125 miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
6126 nodeinfo[pnode].data['memory_free'])
6128 raise errors.OpPrereqError("This change will prevent the instance"
6129 " from starting, due to %d MB of memory"
6130 " missing on its primary node" % miss_mem)
6132 if be_new[constants.BE_AUTO_BALANCE]:
6133 for node, nres in nodeinfo.iteritems():
6134 if node not in instance.secondary_nodes:
6136 if nres.failed or not isinstance(nres.data, dict):
6137 self.warn.append("Can't get info from secondary node %s" % node)
6138 elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
6139 self.warn.append("Not enough memory to failover instance to"
6140 " secondary node %s" % node)
6145 for nic_op, nic_dict in self.op.nics:
6146 if nic_op == constants.DDM_REMOVE:
6147 if not instance.nics:
6148 raise errors.OpPrereqError("Instance has no NICs, cannot remove")
6150 if nic_op != constants.DDM_ADD:
6152 if nic_op < 0 or nic_op >= len(instance.nics):
6153 raise errors.OpPrereqError("Invalid NIC index %s, valid values"
6155 (nic_op, len(instance.nics)))
6156 old_nic_params = instance.nics[nic_op].nicparams
6157 old_nic_ip = instance.nics[nic_op].ip
6162 update_params_dict = dict([(key, nic_dict[key])
6163 for key in constants.NICS_PARAMETERS
6164 if key in nic_dict])
6166 if 'bridge' in nic_dict:
6167 update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
6169 new_nic_params, new_filled_nic_params = \
6170 self._GetUpdatedParams(old_nic_params, update_params_dict,
6171 cluster.nicparams[constants.PP_DEFAULT],
6172 constants.NICS_PARAMETER_TYPES)
6173 objects.NIC.CheckParameterSyntax(new_filled_nic_params)
6174 self.nic_pinst[nic_op] = new_nic_params
6175 self.nic_pnew[nic_op] = new_filled_nic_params
6176 new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
6178 if new_nic_mode == constants.NIC_MODE_BRIDGED:
6179 nic_bridge = new_filled_nic_params[constants.NIC_LINK]
6180 result = self.rpc.call_bridges_exist(pnode, [nic_bridge])
6181 msg = result.RemoteFailMsg()
6183 msg = "Error checking bridges on node %s: %s" % (pnode, msg)
6185 self.warn.append(msg)
6187 raise errors.OpPrereqError(msg)
6188 if new_nic_mode == constants.NIC_MODE_ROUTED:
6189 if 'ip' in nic_dict:
6190 nic_ip = nic_dict['ip']
6194 raise errors.OpPrereqError('Cannot set the nic ip to None'
6196 if 'mac' in nic_dict:
6197 nic_mac = nic_dict['mac']
6199 raise errors.OpPrereqError('Cannot set the nic mac to None')
6200 elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6201 # otherwise generate the mac
6202 nic_dict['mac'] = self.cfg.GenerateMAC()
6204 # or validate/reserve the current one
6205 if self.cfg.IsMacInUse(nic_mac):
6206 raise errors.OpPrereqError("MAC address %s already in use"
6207 " in cluster" % nic_mac)
6210 if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
6211 raise errors.OpPrereqError("Disk operations not supported for"
6212 " diskless instances")
6213 for disk_op, disk_dict in self.op.disks:
6214 if disk_op == constants.DDM_REMOVE:
6215 if len(instance.disks) == 1:
6216 raise errors.OpPrereqError("Cannot remove the last disk of"
6218 ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
6219 ins_l = ins_l[pnode]
6220 if ins_l.failed or not isinstance(ins_l.data, list):
6221 raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
6222 if instance.name in ins_l.data:
6223 raise errors.OpPrereqError("Instance is running, can't remove"
6226 if (disk_op == constants.DDM_ADD and
6227 len(instance.nics) >= constants.MAX_DISKS):
6228 raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
6229 " add more" % constants.MAX_DISKS)
6230 if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
6232 if disk_op < 0 or disk_op >= len(instance.disks):
6233 raise errors.OpPrereqError("Invalid disk index %s, valid values"
6235 (disk_op, len(instance.disks)))
6239 def Exec(self, feedback_fn):
6240 """Modifies an instance.
6242 All parameters take effect only at the next restart of the instance.
6245 # Process here the warnings from CheckPrereq, as we don't have a
6246 # feedback_fn there.
6247 for warn in self.warn:
6248 feedback_fn("WARNING: %s" % warn)
6251 instance = self.instance
6252 cluster = self.cluster
6254 for disk_op, disk_dict in self.op.disks:
6255 if disk_op == constants.DDM_REMOVE:
6256 # remove the last disk
6257 device = instance.disks.pop()
6258 device_idx = len(instance.disks)
6259 for node, disk in device.ComputeNodeTree(instance.primary_node):
6260 self.cfg.SetDiskID(disk, node)
6261 msg = self.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
6263 self.LogWarning("Could not remove disk/%d on node %s: %s,"
6264 " continuing anyway", device_idx, node, msg)
6265 result.append(("disk/%d" % device_idx, "remove"))
6266 elif disk_op == constants.DDM_ADD:
6268 if instance.disk_template == constants.DT_FILE:
6269 file_driver, file_path = instance.disks[0].logical_id
6270 file_path = os.path.dirname(file_path)
6272 file_driver = file_path = None
6273 disk_idx_base = len(instance.disks)
6274 new_disk = _GenerateDiskTemplate(self,
6275 instance.disk_template,
6276 instance.name, instance.primary_node,
6277 instance.secondary_nodes,
6282 instance.disks.append(new_disk)
6283 info = _GetInstanceInfoText(instance)
6285 logging.info("Creating volume %s for instance %s",
6286 new_disk.iv_name, instance.name)
6287 # Note: this needs to be kept in sync with _CreateDisks
6289 for node in instance.all_nodes:
6290 f_create = node == instance.primary_node
6292 _CreateBlockDev(self, node, instance, new_disk,
6293 f_create, info, f_create)
6294 except errors.OpExecError, err:
6295 self.LogWarning("Failed to create volume %s (%s) on"
6297 new_disk.iv_name, new_disk, node, err)
6298 result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
6299 (new_disk.size, new_disk.mode)))
6301 # change a given disk
6302 instance.disks[disk_op].mode = disk_dict['mode']
6303 result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
6305 for nic_op, nic_dict in self.op.nics:
6306 if nic_op == constants.DDM_REMOVE:
6307 # remove the last nic
6308 del instance.nics[-1]
6309 result.append(("nic.%d" % len(instance.nics), "remove"))
6310 elif nic_op == constants.DDM_ADD:
6311 # mac and bridge should be set, by now
6312 mac = nic_dict['mac']
6313 ip = nic_dict.get('ip', None)
6314 nicparams = self.nic_pinst[constants.DDM_ADD]
6315 new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
6316 instance.nics.append(new_nic)
6317 result.append(("nic.%d" % (len(instance.nics) - 1),
6318 "add:mac=%s,ip=%s,mode=%s,link=%s" %
6319 (new_nic.mac, new_nic.ip,
6320 self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
6321 self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
6324 for key in 'mac', 'ip':
6326 setattr(instance.nics[nic_op], key, nic_dict[key])
6327 if nic_op in self.nic_pnew:
6328 instance.nics[nic_op].nicparams = self.nic_pnew[nic_op]
6329 for key, val in nic_dict.iteritems():
6330 result.append(("nic.%s/%d" % (key, nic_op), val))
6333 if self.op.hvparams:
6334 instance.hvparams = self.hv_inst
6335 for key, val in self.op.hvparams.iteritems():
6336 result.append(("hv/%s" % key, val))
6339 if self.op.beparams:
6340 instance.beparams = self.be_inst
6341 for key, val in self.op.beparams.iteritems():
6342 result.append(("be/%s" % key, val))
6344 self.cfg.Update(instance)
6349 class LUQueryExports(NoHooksLU):
6350 """Query the exports list
6353 _OP_REQP = ['nodes']
6356 def ExpandNames(self):
6357 self.needed_locks = {}
6358 self.share_locks[locking.LEVEL_NODE] = 1
6359 if not self.op.nodes:
6360 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6362 self.needed_locks[locking.LEVEL_NODE] = \
6363 _GetWantedNodes(self, self.op.nodes)
6365 def CheckPrereq(self):
6366 """Check prerequisites.
6369 self.nodes = self.acquired_locks[locking.LEVEL_NODE]
6371 def Exec(self, feedback_fn):
6372 """Compute the list of all the exported system images.
6375 @return: a dictionary with the structure node->(export-list)
6376 where export-list is a list of the instances exported on
6380 rpcresult = self.rpc.call_export_list(self.nodes)
6382 for node in rpcresult:
6383 if rpcresult[node].RemoteFailMsg():
6384 result[node] = False
6386 result[node] = rpcresult[node].payload
6391 class LUExportInstance(LogicalUnit):
6392 """Export an instance to an image in the cluster.
6395 HPATH = "instance-export"
6396 HTYPE = constants.HTYPE_INSTANCE
6397 _OP_REQP = ["instance_name", "target_node", "shutdown"]
6400 def ExpandNames(self):
6401 self._ExpandAndLockInstance()
6402 # FIXME: lock only instance primary and destination node
6404 # Sad but true, for now we have do lock all nodes, as we don't know where
6405 # the previous export might be, and and in this LU we search for it and
6406 # remove it from its current node. In the future we could fix this by:
6407 # - making a tasklet to search (share-lock all), then create the new one,
6408 # then one to remove, after
6409 # - removing the removal operation altoghether
6410 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6412 def DeclareLocks(self, level):
6413 """Last minute lock declaration."""
6414 # All nodes are locked anyway, so nothing to do here.
6416 def BuildHooksEnv(self):
6419 This will run on the master, primary node and target node.
6423 "EXPORT_NODE": self.op.target_node,
6424 "EXPORT_DO_SHUTDOWN": self.op.shutdown,
6426 env.update(_BuildInstanceHookEnvByObject(self, self.instance))
6427 nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
6428 self.op.target_node]
6431 def CheckPrereq(self):
6432 """Check prerequisites.
6434 This checks that the instance and node names are valid.
6437 instance_name = self.op.instance_name
6438 self.instance = self.cfg.GetInstanceInfo(instance_name)
6439 assert self.instance is not None, \
6440 "Cannot retrieve locked instance %s" % self.op.instance_name
6441 _CheckNodeOnline(self, self.instance.primary_node)
6443 self.dst_node = self.cfg.GetNodeInfo(
6444 self.cfg.ExpandNodeName(self.op.target_node))
6446 if self.dst_node is None:
6447 # This is wrong node name, not a non-locked node
6448 raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
6449 _CheckNodeOnline(self, self.dst_node.name)
6450 _CheckNodeNotDrained(self, self.dst_node.name)
6452 # instance disk type verification
6453 for disk in self.instance.disks:
6454 if disk.dev_type == constants.LD_FILE:
6455 raise errors.OpPrereqError("Export not supported for instances with"
6456 " file-based disks")
6458 def Exec(self, feedback_fn):
6459 """Export an instance to an image in the cluster.
6462 instance = self.instance
6463 dst_node = self.dst_node
6464 src_node = instance.primary_node
6465 if self.op.shutdown:
6466 # shutdown the instance, but not the disks
6467 result = self.rpc.call_instance_shutdown(src_node, instance)
6468 msg = result.RemoteFailMsg()
6470 raise errors.OpExecError("Could not shutdown instance %s on"
6472 (instance.name, src_node, msg))
6474 vgname = self.cfg.GetVGName()
6478 # set the disks ID correctly since call_instance_start needs the
6479 # correct drbd minor to create the symlinks
6480 for disk in instance.disks:
6481 self.cfg.SetDiskID(disk, src_node)
6484 for disk in instance.disks:
6485 # result.payload will be a snapshot of an lvm leaf of the one we passed
6486 result = self.rpc.call_blockdev_snapshot(src_node, disk)
6487 msg = result.RemoteFailMsg()
6489 self.LogWarning("Could not snapshot block device %s on node %s: %s",
6490 disk.logical_id[1], src_node, msg)
6491 snap_disks.append(False)
6493 disk_id = (vgname, result.payload)
6494 new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
6495 logical_id=disk_id, physical_id=disk_id,
6496 iv_name=disk.iv_name)
6497 snap_disks.append(new_dev)
6500 if self.op.shutdown and instance.admin_up:
6501 result = self.rpc.call_instance_start(src_node, instance, None, None)
6502 msg = result.RemoteFailMsg()
6504 _ShutdownInstanceDisks(self, instance)
6505 raise errors.OpExecError("Could not start instance: %s" % msg)
6507 # TODO: check for size
6509 cluster_name = self.cfg.GetClusterName()
6510 for idx, dev in enumerate(snap_disks):
6512 result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6513 instance, cluster_name, idx)
6514 msg = result.RemoteFailMsg()
6516 self.LogWarning("Could not export block device %s from node %s to"
6517 " node %s: %s", dev.logical_id[1], src_node,
6519 msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
6521 self.LogWarning("Could not remove snapshot block device %s from node"
6522 " %s: %s", dev.logical_id[1], src_node, msg)
6524 result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6525 msg = result.RemoteFailMsg()
6527 self.LogWarning("Could not finalize export for instance %s"
6528 " on node %s: %s", instance.name, dst_node.name, msg)
6530 nodelist = self.cfg.GetNodeList()
6531 nodelist.remove(dst_node.name)
6533 # on one-node clusters nodelist will be empty after the removal
6534 # if we proceed the backup would be removed because OpQueryExports
6535 # substitutes an empty list with the full cluster node list.
6536 iname = instance.name
6538 exportlist = self.rpc.call_export_list(nodelist)
6539 for node in exportlist:
6540 if exportlist[node].RemoteFailMsg():
6542 if iname in exportlist[node].payload:
6543 msg = self.rpc.call_export_remove(node, iname).RemoteFailMsg()
6545 self.LogWarning("Could not remove older export for instance %s"
6546 " on node %s: %s", iname, node, msg)
6549 class LURemoveExport(NoHooksLU):
6550 """Remove exports related to the named instance.
6553 _OP_REQP = ["instance_name"]
6556 def ExpandNames(self):
6557 self.needed_locks = {}
6558 # We need all nodes to be locked in order for RemoveExport to work, but we
6559 # don't need to lock the instance itself, as nothing will happen to it (and
6560 # we can remove exports also for a removed instance)
6561 self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6563 def CheckPrereq(self):
6564 """Check prerequisites.
6568 def Exec(self, feedback_fn):
6569 """Remove any export.
6572 instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6573 # If the instance was not found we'll try with the name that was passed in.
6574 # This will only work if it was an FQDN, though.
6576 if not instance_name:
6578 instance_name = self.op.instance_name
6580 locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6581 exportlist = self.rpc.call_export_list(locked_nodes)
6583 for node in exportlist:
6584 msg = exportlist[node].RemoteFailMsg()
6586 self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
6588 if instance_name in exportlist[node].payload:
6590 result = self.rpc.call_export_remove(node, instance_name)
6591 msg = result.RemoteFailMsg()
6593 logging.error("Could not remove export for instance %s"
6594 " on node %s: %s", instance_name, node, msg)
6596 if fqdn_warn and not found:
6597 feedback_fn("Export not found. If trying to remove an export belonging"
6598 " to a deleted instance please use its Fully Qualified"
6602 class TagsLU(NoHooksLU):
6605 This is an abstract class which is the parent of all the other tags LUs.
6609 def ExpandNames(self):
6610 self.needed_locks = {}
6611 if self.op.kind == constants.TAG_NODE:
6612 name = self.cfg.ExpandNodeName(self.op.name)
6614 raise errors.OpPrereqError("Invalid node name (%s)" %
6617 self.needed_locks[locking.LEVEL_NODE] = name
6618 elif self.op.kind == constants.TAG_INSTANCE:
6619 name = self.cfg.ExpandInstanceName(self.op.name)
6621 raise errors.OpPrereqError("Invalid instance name (%s)" %
6624 self.needed_locks[locking.LEVEL_INSTANCE] = name
6626 def CheckPrereq(self):
6627 """Check prerequisites.
6630 if self.op.kind == constants.TAG_CLUSTER:
6631 self.target = self.cfg.GetClusterInfo()
6632 elif self.op.kind == constants.TAG_NODE:
6633 self.target = self.cfg.GetNodeInfo(self.op.name)
6634 elif self.op.kind == constants.TAG_INSTANCE:
6635 self.target = self.cfg.GetInstanceInfo(self.op.name)
6637 raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6641 class LUGetTags(TagsLU):
6642 """Returns the tags of a given object.
6645 _OP_REQP = ["kind", "name"]
6648 def Exec(self, feedback_fn):
6649 """Returns the tag list.
6652 return list(self.target.GetTags())
6655 class LUSearchTags(NoHooksLU):
6656 """Searches the tags for a given pattern.
6659 _OP_REQP = ["pattern"]
6662 def ExpandNames(self):
6663 self.needed_locks = {}
6665 def CheckPrereq(self):
6666 """Check prerequisites.
6668 This checks the pattern passed for validity by compiling it.
6672 self.re = re.compile(self.op.pattern)
6673 except re.error, err:
6674 raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6675 (self.op.pattern, err))
6677 def Exec(self, feedback_fn):
6678 """Returns the tag list.
6682 tgts = [("/cluster", cfg.GetClusterInfo())]
6683 ilist = cfg.GetAllInstancesInfo().values()
6684 tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6685 nlist = cfg.GetAllNodesInfo().values()
6686 tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6688 for path, target in tgts:
6689 for tag in target.GetTags():
6690 if self.re.search(tag):
6691 results.append((path, tag))
6695 class LUAddTags(TagsLU):
6696 """Sets a tag on a given object.
6699 _OP_REQP = ["kind", "name", "tags"]
6702 def CheckPrereq(self):
6703 """Check prerequisites.
6705 This checks the type and length of the tag name and value.
6708 TagsLU.CheckPrereq(self)
6709 for tag in self.op.tags:
6710 objects.TaggableObject.ValidateTag(tag)
6712 def Exec(self, feedback_fn):
6717 for tag in self.op.tags:
6718 self.target.AddTag(tag)
6719 except errors.TagError, err:
6720 raise errors.OpExecError("Error while setting tag: %s" % str(err))
6722 self.cfg.Update(self.target)
6723 except errors.ConfigurationError:
6724 raise errors.OpRetryError("There has been a modification to the"
6725 " config file and the operation has been"
6726 " aborted. Please retry.")
6729 class LUDelTags(TagsLU):
6730 """Delete a list of tags from a given object.
6733 _OP_REQP = ["kind", "name", "tags"]
6736 def CheckPrereq(self):
6737 """Check prerequisites.
6739 This checks that we have the given tag.
6742 TagsLU.CheckPrereq(self)
6743 for tag in self.op.tags:
6744 objects.TaggableObject.ValidateTag(tag)
6745 del_tags = frozenset(self.op.tags)
6746 cur_tags = self.target.GetTags()
6747 if not del_tags <= cur_tags:
6748 diff_tags = del_tags - cur_tags
6749 diff_names = ["'%s'" % tag for tag in diff_tags]
6751 raise errors.OpPrereqError("Tag(s) %s not found" %
6752 (",".join(diff_names)))
6754 def Exec(self, feedback_fn):
6755 """Remove the tag from the object.
6758 for tag in self.op.tags:
6759 self.target.RemoveTag(tag)
6761 self.cfg.Update(self.target)
6762 except errors.ConfigurationError:
6763 raise errors.OpRetryError("There has been a modification to the"
6764 " config file and the operation has been"
6765 " aborted. Please retry.")
6768 class LUTestDelay(NoHooksLU):
6769 """Sleep for a specified amount of time.
6771 This LU sleeps on the master and/or nodes for a specified amount of
6775 _OP_REQP = ["duration", "on_master", "on_nodes"]
6778 def ExpandNames(self):
6779 """Expand names and set required locks.
6781 This expands the node list, if any.
6784 self.needed_locks = {}
6785 if self.op.on_nodes:
6786 # _GetWantedNodes can be used here, but is not always appropriate to use
6787 # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6789 self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6790 self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6792 def CheckPrereq(self):
6793 """Check prerequisites.
6797 def Exec(self, feedback_fn):
6798 """Do the actual sleep.
6801 if self.op.on_master:
6802 if not utils.TestDelay(self.op.duration):
6803 raise errors.OpExecError("Error during master delay test")
6804 if self.op.on_nodes:
6805 result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6807 raise errors.OpExecError("Complete failure from rpc call")
6808 for node, node_result in result.items():
6810 if not node_result.data:
6811 raise errors.OpExecError("Failure during rpc call to node %s,"
6812 " result: %s" % (node, node_result.data))
6815 class IAllocator(object):
6816 """IAllocator framework.
6818 An IAllocator instance has three sets of attributes:
6819 - cfg that is needed to query the cluster
6820 - input data (all members of the _KEYS class attribute are required)
6821 - four buffer attributes (in|out_data|text), that represent the
6822 input (to the external script) in text and data structure format,
6823 and the output from it, again in two formats
6824 - the result variables from the script (success, info, nodes) for
6829 "mem_size", "disks", "disk_template",
6830 "os", "tags", "nics", "vcpus", "hypervisor",
6836 def __init__(self, lu, mode, name, **kwargs):
6838 # init buffer variables
6839 self.in_text = self.out_text = self.in_data = self.out_data = None
6840 # init all input fields so that pylint is happy
6843 self.mem_size = self.disks = self.disk_template = None
6844 self.os = self.tags = self.nics = self.vcpus = None
6845 self.hypervisor = None
6846 self.relocate_from = None
6848 self.required_nodes = None
6849 # init result fields
6850 self.success = self.info = self.nodes = None
6851 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6852 keyset = self._ALLO_KEYS
6853 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6854 keyset = self._RELO_KEYS
6856 raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6857 " IAllocator" % self.mode)
6859 if key not in keyset:
6860 raise errors.ProgrammerError("Invalid input parameter '%s' to"
6861 " IAllocator" % key)
6862 setattr(self, key, kwargs[key])
6864 if key not in kwargs:
6865 raise errors.ProgrammerError("Missing input parameter '%s' to"
6866 " IAllocator" % key)
6867 self._BuildInputData()
6869 def _ComputeClusterData(self):
6870 """Compute the generic allocator input data.
6872 This is the data that is independent of the actual operation.
6876 cluster_info = cfg.GetClusterInfo()
6879 "version": constants.IALLOCATOR_VERSION,
6880 "cluster_name": cfg.GetClusterName(),
6881 "cluster_tags": list(cluster_info.GetTags()),
6882 "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
6883 # we don't have job IDs
6885 iinfo = cfg.GetAllInstancesInfo().values()
6886 i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6890 node_list = cfg.GetNodeList()
6892 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6893 hypervisor_name = self.hypervisor
6894 elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6895 hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6897 node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6899 node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6900 cluster_info.enabled_hypervisors)
6901 for nname, nresult in node_data.items():
6902 # first fill in static (config-based) values
6903 ninfo = cfg.GetNodeInfo(nname)
6905 "tags": list(ninfo.GetTags()),
6906 "primary_ip": ninfo.primary_ip,
6907 "secondary_ip": ninfo.secondary_ip,
6908 "offline": ninfo.offline,
6909 "drained": ninfo.drained,
6910 "master_candidate": ninfo.master_candidate,
6913 if not ninfo.offline:
6915 if not isinstance(nresult.data, dict):
6916 raise errors.OpExecError("Can't get data for node %s" % nname)
6917 remote_info = nresult.data
6918 for attr in ['memory_total', 'memory_free', 'memory_dom0',
6919 'vg_size', 'vg_free', 'cpu_total']:
6920 if attr not in remote_info:
6921 raise errors.OpExecError("Node '%s' didn't return attribute"
6922 " '%s'" % (nname, attr))
6924 remote_info[attr] = int(remote_info[attr])
6925 except ValueError, err:
6926 raise errors.OpExecError("Node '%s' returned invalid value"
6927 " for '%s': %s" % (nname, attr, err))
6928 # compute memory used by primary instances
6929 i_p_mem = i_p_up_mem = 0
6930 for iinfo, beinfo in i_list:
6931 if iinfo.primary_node == nname:
6932 i_p_mem += beinfo[constants.BE_MEMORY]
6933 if iinfo.name not in node_iinfo[nname].data:
6936 i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
6937 i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6938 remote_info['memory_free'] -= max(0, i_mem_diff)
6941 i_p_up_mem += beinfo[constants.BE_MEMORY]
6943 # compute memory used by instances
6945 "total_memory": remote_info['memory_total'],
6946 "reserved_memory": remote_info['memory_dom0'],
6947 "free_memory": remote_info['memory_free'],
6948 "total_disk": remote_info['vg_size'],
6949 "free_disk": remote_info['vg_free'],
6950 "total_cpus": remote_info['cpu_total'],
6951 "i_pri_memory": i_p_mem,
6952 "i_pri_up_memory": i_p_up_mem,
6956 node_results[nname] = pnr
6957 data["nodes"] = node_results
6961 for iinfo, beinfo in i_list:
6963 for nic in iinfo.nics:
6964 filled_params = objects.FillDict(
6965 cluster_info.nicparams[constants.PP_DEFAULT],
6967 nic_dict = {"mac": nic.mac,
6969 "mode": filled_params[constants.NIC_MODE],
6970 "link": filled_params[constants.NIC_LINK],
6972 if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
6973 nic_dict["bridge"] = filled_params[constants.NIC_LINK]
6974 nic_data.append(nic_dict)
6976 "tags": list(iinfo.GetTags()),
6977 "admin_up": iinfo.admin_up,
6978 "vcpus": beinfo[constants.BE_VCPUS],
6979 "memory": beinfo[constants.BE_MEMORY],
6981 "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
6983 "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
6984 "disk_template": iinfo.disk_template,
6985 "hypervisor": iinfo.hypervisor,
6987 pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
6989 instance_data[iinfo.name] = pir
6991 data["instances"] = instance_data
6995 def _AddNewInstance(self):
6996 """Add new instance data to allocator structure.
6998 This in combination with _AllocatorGetClusterData will create the
6999 correct structure needed as input for the allocator.
7001 The checks for the completeness of the opcode must have already been
7007 disk_space = _ComputeDiskSize(self.disk_template, self.disks)
7009 if self.disk_template in constants.DTS_NET_MIRROR:
7010 self.required_nodes = 2
7012 self.required_nodes = 1
7016 "disk_template": self.disk_template,
7019 "vcpus": self.vcpus,
7020 "memory": self.mem_size,
7021 "disks": self.disks,
7022 "disk_space_total": disk_space,
7024 "required_nodes": self.required_nodes,
7026 data["request"] = request
7028 def _AddRelocateInstance(self):
7029 """Add relocate instance data to allocator structure.
7031 This in combination with _IAllocatorGetClusterData will create the
7032 correct structure needed as input for the allocator.
7034 The checks for the completeness of the opcode must have already been
7038 instance = self.lu.cfg.GetInstanceInfo(self.name)
7039 if instance is None:
7040 raise errors.ProgrammerError("Unknown instance '%s' passed to"
7041 " IAllocator" % self.name)
7043 if instance.disk_template not in constants.DTS_NET_MIRROR:
7044 raise errors.OpPrereqError("Can't relocate non-mirrored instances")
7046 if len(instance.secondary_nodes) != 1:
7047 raise errors.OpPrereqError("Instance has not exactly one secondary node")
7049 self.required_nodes = 1
7050 disk_sizes = [{'size': disk.size} for disk in instance.disks]
7051 disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
7056 "disk_space_total": disk_space,
7057 "required_nodes": self.required_nodes,
7058 "relocate_from": self.relocate_from,
7060 self.in_data["request"] = request
7062 def _BuildInputData(self):
7063 """Build input data structures.
7066 self._ComputeClusterData()
7068 if self.mode == constants.IALLOCATOR_MODE_ALLOC:
7069 self._AddNewInstance()
7071 self._AddRelocateInstance()
7073 self.in_text = serializer.Dump(self.in_data)
7075 def Run(self, name, validate=True, call_fn=None):
7076 """Run an instance allocator and return the results.
7080 call_fn = self.lu.rpc.call_iallocator_runner
7083 result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
7086 if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
7087 raise errors.OpExecError("Invalid result from master iallocator runner")
7089 rcode, stdout, stderr, fail = result.data
7091 if rcode == constants.IARUN_NOTFOUND:
7092 raise errors.OpExecError("Can't find allocator '%s'" % name)
7093 elif rcode == constants.IARUN_FAILURE:
7094 raise errors.OpExecError("Instance allocator call failed: %s,"
7095 " output: %s" % (fail, stdout+stderr))
7096 self.out_text = stdout
7098 self._ValidateResult()
7100 def _ValidateResult(self):
7101 """Process the allocator results.
7103 This will process and if successful save the result in
7104 self.out_data and the other parameters.
7108 rdict = serializer.Load(self.out_text)
7109 except Exception, err:
7110 raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
7112 if not isinstance(rdict, dict):
7113 raise errors.OpExecError("Can't parse iallocator results: not a dict")
7115 for key in "success", "info", "nodes":
7116 if key not in rdict:
7117 raise errors.OpExecError("Can't parse iallocator results:"
7118 " missing key '%s'" % key)
7119 setattr(self, key, rdict[key])
7121 if not isinstance(rdict["nodes"], list):
7122 raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
7124 self.out_data = rdict
7127 class LUTestAllocator(NoHooksLU):
7128 """Run allocator tests.
7130 This LU runs the allocator tests
7133 _OP_REQP = ["direction", "mode", "name"]
7135 def CheckPrereq(self):
7136 """Check prerequisites.
7138 This checks the opcode parameters depending on the director and mode test.
7141 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
7142 for attr in ["name", "mem_size", "disks", "disk_template",
7143 "os", "tags", "nics", "vcpus"]:
7144 if not hasattr(self.op, attr):
7145 raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
7147 iname = self.cfg.ExpandInstanceName(self.op.name)
7148 if iname is not None:
7149 raise errors.OpPrereqError("Instance '%s' already in the cluster" %
7151 if not isinstance(self.op.nics, list):
7152 raise errors.OpPrereqError("Invalid parameter 'nics'")
7153 for row in self.op.nics:
7154 if (not isinstance(row, dict) or
7157 "bridge" not in row):
7158 raise errors.OpPrereqError("Invalid contents of the"
7159 " 'nics' parameter")
7160 if not isinstance(self.op.disks, list):
7161 raise errors.OpPrereqError("Invalid parameter 'disks'")
7162 for row in self.op.disks:
7163 if (not isinstance(row, dict) or
7164 "size" not in row or
7165 not isinstance(row["size"], int) or
7166 "mode" not in row or
7167 row["mode"] not in ['r', 'w']):
7168 raise errors.OpPrereqError("Invalid contents of the"
7169 " 'disks' parameter")
7170 if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
7171 self.op.hypervisor = self.cfg.GetHypervisorType()
7172 elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
7173 if not hasattr(self.op, "name"):
7174 raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
7175 fname = self.cfg.ExpandInstanceName(self.op.name)
7177 raise errors.OpPrereqError("Instance '%s' not found for relocation" %
7179 self.op.name = fname
7180 self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
7182 raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
7185 if self.op.direction == constants.IALLOCATOR_DIR_OUT:
7186 if not hasattr(self.op, "allocator") or self.op.allocator is None:
7187 raise errors.OpPrereqError("Missing allocator name")
7188 elif self.op.direction != constants.IALLOCATOR_DIR_IN:
7189 raise errors.OpPrereqError("Wrong allocator test '%s'" %
7192 def Exec(self, feedback_fn):
7193 """Run the allocator test.
7196 if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
7197 ial = IAllocator(self,
7200 mem_size=self.op.mem_size,
7201 disks=self.op.disks,
7202 disk_template=self.op.disk_template,
7206 vcpus=self.op.vcpus,
7207 hypervisor=self.op.hypervisor,
7210 ial = IAllocator(self,
7213 relocate_from=list(self.relocate_from),
7216 if self.op.direction == constants.IALLOCATOR_DIR_IN:
7217 result = ial.in_text
7219 ial.Run(self.op.allocator, validate=False)
7220 result = ial.out_text