4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 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 """Configuration management for Ganeti
24 This module provides the interface to the Ganeti cluster configuration.
26 The configuration data is stored on every node but is updated on the master
27 only. After each update, the master distributes the data to the other nodes.
29 Currently, the data storage format is JSON. YAML was slow and consuming too
34 # pylint: disable=R0904
35 # R0904: Too many public methods
44 from ganeti import errors
45 from ganeti import locking
46 from ganeti import utils
47 from ganeti import constants
48 from ganeti import rpc
49 from ganeti import objects
50 from ganeti import serializer
51 from ganeti import uidpool
52 from ganeti import netutils
53 from ganeti import runtime
54 from ganeti import pathutils
55 from ganeti import network
58 _config_lock = locking.SharedLock("ConfigWriter")
60 # job id used for resource management at config upgrade time
61 _UPGRADE_CONFIG_JID = "jid-cfg-upgrade"
64 def _ValidateConfig(data):
65 """Verifies that a configuration objects looks valid.
67 This only verifies the version of the configuration.
69 @raise errors.ConfigurationError: if the version differs from what
73 if data.version != constants.CONFIG_VERSION:
74 raise errors.ConfigVersionMismatch(constants.CONFIG_VERSION, data.version)
77 class TemporaryReservationManager:
78 """A temporary resource reservation manager.
80 This is used to reserve resources in a job, before using them, making sure
81 other jobs cannot get them in the meantime.
85 self._ec_reserved = {}
87 def Reserved(self, resource):
88 for holder_reserved in self._ec_reserved.values():
89 if resource in holder_reserved:
93 def Reserve(self, ec_id, resource):
94 if self.Reserved(resource):
95 raise errors.ReservationError("Duplicate reservation for resource '%s'"
97 if ec_id not in self._ec_reserved:
98 self._ec_reserved[ec_id] = set([resource])
100 self._ec_reserved[ec_id].add(resource)
102 def DropECReservations(self, ec_id):
103 if ec_id in self._ec_reserved:
104 del self._ec_reserved[ec_id]
106 def GetReserved(self):
108 for holder_reserved in self._ec_reserved.values():
109 all_reserved.update(holder_reserved)
112 def GetECReserved(self, ec_id):
113 """ Used when you want to retrieve all reservations for a specific
114 execution context. E.g when commiting reserved IPs for a specific
119 if ec_id in self._ec_reserved:
120 ec_reserved.update(self._ec_reserved[ec_id])
123 def Generate(self, existing, generate_one_fn, ec_id):
124 """Generate a new resource of this type
127 assert callable(generate_one_fn)
129 all_elems = self.GetReserved()
130 all_elems.update(existing)
133 new_resource = generate_one_fn()
134 if new_resource is not None and new_resource not in all_elems:
137 raise errors.ConfigurationError("Not able generate new resource"
138 " (last tried: %s)" % new_resource)
139 self.Reserve(ec_id, new_resource)
143 def _MatchNameComponentIgnoreCase(short_name, names):
144 """Wrapper around L{utils.text.MatchNameComponent}.
147 return utils.MatchNameComponent(short_name, names, case_sensitive=False)
150 def _CheckInstanceDiskIvNames(disks):
151 """Checks if instance's disks' C{iv_name} attributes are in order.
153 @type disks: list of L{objects.Disk}
154 @param disks: List of disks
155 @rtype: list of tuples; (int, string, string)
156 @return: List of wrongly named disks, each tuple contains disk index,
157 expected and actual name
162 for (idx, disk) in enumerate(disks):
163 exp_iv_name = "disk/%s" % idx
164 if disk.iv_name != exp_iv_name:
165 result.append((idx, exp_iv_name, disk.iv_name))
171 """The interface to the cluster configuration.
173 @ivar _temporary_lvs: reservation manager for temporary LVs
174 @ivar _all_rms: a list of all temporary reservation managers
177 def __init__(self, cfg_file=None, offline=False, _getents=runtime.GetEnts,
178 accept_foreign=False):
180 self._lock = _config_lock
181 self._config_data = None
182 self._offline = offline
184 self._cfg_file = pathutils.CLUSTER_CONF_FILE
186 self._cfg_file = cfg_file
187 self._getents = _getents
188 self._temporary_ids = TemporaryReservationManager()
189 self._temporary_drbds = {}
190 self._temporary_macs = TemporaryReservationManager()
191 self._temporary_secrets = TemporaryReservationManager()
192 self._temporary_lvs = TemporaryReservationManager()
193 self._temporary_ips = TemporaryReservationManager()
194 self._all_rms = [self._temporary_ids, self._temporary_macs,
195 self._temporary_secrets, self._temporary_lvs,
197 # Note: in order to prevent errors when resolving our name in
198 # _DistributeConfig, we compute it here once and reuse it; it's
199 # better to raise an error before starting to modify the config
200 # file than after it was modified
201 self._my_hostname = netutils.Hostname.GetSysName()
202 self._last_cluster_serial = -1
205 self._OpenConfig(accept_foreign)
207 def _GetRpc(self, address_list):
208 """Returns RPC runner for configuration.
211 return rpc.ConfigRunner(self._context, address_list)
213 def SetContext(self, context):
214 """Sets Ganeti context.
217 self._context = context
219 # this method needs to be static, so that we can call it on the class
222 """Check if the cluster is configured.
225 return os.path.exists(pathutils.CLUSTER_CONF_FILE)
227 @locking.ssynchronized(_config_lock, shared=1)
228 def GetNdParams(self, node):
229 """Get the node params populated with cluster defaults.
231 @type node: L{objects.Node}
232 @param node: The node we want to know the params for
233 @return: A dict with the filled in node params
236 nodegroup = self._UnlockedGetNodeGroup(node.group)
237 return self._config_data.cluster.FillND(node, nodegroup)
239 @locking.ssynchronized(_config_lock, shared=1)
240 def GetInstanceDiskParams(self, instance):
241 """Get the disk params populated with inherit chain.
243 @type instance: L{objects.Instance}
244 @param instance: The instance we want to know the params for
245 @return: A dict with the filled in disk params
248 node = self._UnlockedGetNodeInfo(instance.primary_node)
249 nodegroup = self._UnlockedGetNodeGroup(node.group)
250 return self._UnlockedGetGroupDiskParams(nodegroup)
252 @locking.ssynchronized(_config_lock, shared=1)
253 def GetGroupDiskParams(self, group):
254 """Get the disk params populated with inherit chain.
256 @type group: L{objects.NodeGroup}
257 @param group: The group we want to know the params for
258 @return: A dict with the filled in disk params
261 return self._UnlockedGetGroupDiskParams(group)
263 def _UnlockedGetGroupDiskParams(self, group):
264 """Get the disk params populated with inherit chain down to node-group.
266 @type group: L{objects.NodeGroup}
267 @param group: The group we want to know the params for
268 @return: A dict with the filled in disk params
271 return self._config_data.cluster.SimpleFillDP(group.diskparams)
273 def _UnlockedGetNetworkMACPrefix(self, net_uuid):
274 """Return the network mac prefix if it exists or the cluster level default.
279 nobj = self._UnlockedGetNetwork(net_uuid)
281 prefix = nobj.mac_prefix
285 def _GenerateOneMAC(self, prefix=None):
286 """Return a function that randomly generates a MAC suffic
287 and appends it to the given prefix. If prefix is not given get
288 the cluster level default.
292 prefix = self._config_data.cluster.mac_prefix
295 byte1 = random.randrange(0, 256)
296 byte2 = random.randrange(0, 256)
297 byte3 = random.randrange(0, 256)
298 mac = "%s:%02x:%02x:%02x" % (prefix, byte1, byte2, byte3)
303 @locking.ssynchronized(_config_lock, shared=1)
304 def GenerateMAC(self, net_uuid, ec_id):
305 """Generate a MAC for an instance.
307 This should check the current instances for duplicates.
310 existing = self._AllMACs()
311 prefix = self._UnlockedGetNetworkMACPrefix(net_uuid)
312 gen_mac = self._GenerateOneMAC(prefix)
313 return self._temporary_ids.Generate(existing, gen_mac, ec_id)
315 @locking.ssynchronized(_config_lock, shared=1)
316 def ReserveMAC(self, mac, ec_id):
317 """Reserve a MAC for an instance.
319 This only checks instances managed by this cluster, it does not
320 check for potential collisions elsewhere.
323 all_macs = self._AllMACs()
325 raise errors.ReservationError("mac already in use")
327 self._temporary_macs.Reserve(ec_id, mac)
329 def _UnlockedCommitTemporaryIps(self, ec_id):
330 """Commit all reserved IP address to their respective pools
333 for action, address, net_uuid in self._temporary_ips.GetECReserved(ec_id):
334 self._UnlockedCommitIp(action, net_uuid, address)
336 def _UnlockedCommitIp(self, action, net_uuid, address):
337 """Commit a reserved IP address to an IP pool.
339 The IP address is taken from the network's IP pool and marked as reserved.
342 nobj = self._UnlockedGetNetwork(net_uuid)
343 pool = network.AddressPool(nobj)
344 if action == constants.RESERVE_ACTION:
345 pool.Reserve(address)
346 elif action == constants.RELEASE_ACTION:
347 pool.Release(address)
349 def _UnlockedReleaseIp(self, net_uuid, address, ec_id):
350 """Give a specific IP address back to an IP pool.
352 The IP address is returned to the IP pool designated by pool_id and marked
356 self._temporary_ips.Reserve(ec_id,
357 (constants.RELEASE_ACTION, address, net_uuid))
359 @locking.ssynchronized(_config_lock, shared=1)
360 def ReleaseIp(self, net_uuid, address, ec_id):
361 """Give a specified IP address back to an IP pool.
363 This is just a wrapper around _UnlockedReleaseIp.
367 self._UnlockedReleaseIp(net_uuid, address, ec_id)
369 @locking.ssynchronized(_config_lock, shared=1)
370 def GenerateIp(self, net_uuid, ec_id):
371 """Find a free IPv4 address for an instance.
374 nobj = self._UnlockedGetNetwork(net_uuid)
375 pool = network.AddressPool(nobj)
379 ip = pool.GenerateFree()
380 except errors.AddressPoolError:
381 raise errors.ReservationError("Cannot generate IP. Network is full")
382 return (constants.RESERVE_ACTION, ip, net_uuid)
384 _, address, _ = self._temporary_ips.Generate([], gen_one, ec_id)
387 def _UnlockedReserveIp(self, net_uuid, address, ec_id):
388 """Reserve a given IPv4 address for use by an instance.
391 nobj = self._UnlockedGetNetwork(net_uuid)
392 pool = network.AddressPool(nobj)
394 isreserved = pool.IsReserved(address)
395 except errors.AddressPoolError:
396 raise errors.ReservationError("IP address not in network")
398 raise errors.ReservationError("IP address already in use")
400 return self._temporary_ips.Reserve(ec_id,
401 (constants.RESERVE_ACTION,
404 @locking.ssynchronized(_config_lock, shared=1)
405 def ReserveIp(self, net_uuid, address, ec_id):
406 """Reserve a given IPv4 address for use by an instance.
410 return self._UnlockedReserveIp(net_uuid, address, ec_id)
412 @locking.ssynchronized(_config_lock, shared=1)
413 def ReserveLV(self, lv_name, ec_id):
414 """Reserve an VG/LV pair for an instance.
416 @type lv_name: string
417 @param lv_name: the logical volume name to reserve
420 all_lvs = self._AllLVs()
421 if lv_name in all_lvs:
422 raise errors.ReservationError("LV already in use")
424 self._temporary_lvs.Reserve(ec_id, lv_name)
426 @locking.ssynchronized(_config_lock, shared=1)
427 def GenerateDRBDSecret(self, ec_id):
428 """Generate a DRBD secret.
430 This checks the current disks for duplicates.
433 return self._temporary_secrets.Generate(self._AllDRBDSecrets(),
434 utils.GenerateSecret,
438 """Compute the list of all LVs.
442 for instance in self._config_data.instances.values():
443 node_data = instance.MapLVsByNode()
444 for lv_list in node_data.values():
445 lvnames.update(lv_list)
449 """Compute the list of all Disks.
453 for instance in self._config_data.instances.values():
454 disks.extend(instance.disks)
458 """Compute the list of all NICs.
462 for instance in self._config_data.instances.values():
463 nics.extend(instance.nics)
466 def _AllIDs(self, include_temporary):
467 """Compute the list of all UUIDs and names we have.
469 @type include_temporary: boolean
470 @param include_temporary: whether to include the _temporary_ids set
472 @return: a set of IDs
476 if include_temporary:
477 existing.update(self._temporary_ids.GetReserved())
478 existing.update(self._AllLVs())
479 existing.update(self._config_data.instances.keys())
480 existing.update(self._config_data.nodes.keys())
481 existing.update([i.uuid for i in self._AllUUIDObjects() if i.uuid])
484 def _GenerateUniqueID(self, ec_id):
485 """Generate an unique UUID.
487 This checks the current node, instances and disk names for
491 @return: the unique id
494 existing = self._AllIDs(include_temporary=False)
495 return self._temporary_ids.Generate(existing, utils.NewUUID, ec_id)
497 @locking.ssynchronized(_config_lock, shared=1)
498 def GenerateUniqueID(self, ec_id):
499 """Generate an unique ID.
501 This is just a wrapper over the unlocked version.
504 @param ec_id: unique id for the job to reserve the id to
507 return self._GenerateUniqueID(ec_id)
510 """Return all MACs present in the config.
513 @return: the list of all MACs
517 for instance in self._config_data.instances.values():
518 for nic in instance.nics:
519 result.append(nic.mac)
523 def _AllDRBDSecrets(self):
524 """Return all DRBD secrets present in the config.
527 @return: the list of all DRBD secrets
530 def helper(disk, result):
531 """Recursively gather secrets from this disk."""
532 if disk.dev_type == constants.DT_DRBD8:
533 result.append(disk.logical_id[5])
535 for child in disk.children:
536 helper(child, result)
539 for instance in self._config_data.instances.values():
540 for disk in instance.disks:
545 def _CheckDiskIDs(self, disk, l_ids, p_ids):
546 """Compute duplicate disk IDs
548 @type disk: L{objects.Disk}
549 @param disk: the disk at which to start searching
551 @param l_ids: list of current logical ids
553 @param p_ids: list of current physical ids
555 @return: a list of error messages
559 if disk.logical_id is not None:
560 if disk.logical_id in l_ids:
561 result.append("duplicate logical id %s" % str(disk.logical_id))
563 l_ids.append(disk.logical_id)
564 if disk.physical_id is not None:
565 if disk.physical_id in p_ids:
566 result.append("duplicate physical id %s" % str(disk.physical_id))
568 p_ids.append(disk.physical_id)
571 for child in disk.children:
572 result.extend(self._CheckDiskIDs(child, l_ids, p_ids))
575 def _UnlockedVerifyConfig(self):
579 @return: a list of error messages; a non-empty list signifies
583 # pylint: disable=R0914
587 data = self._config_data
588 cluster = data.cluster
592 # global cluster checks
593 if not cluster.enabled_hypervisors:
594 result.append("enabled hypervisors list doesn't have any entries")
595 invalid_hvs = set(cluster.enabled_hypervisors) - constants.HYPER_TYPES
597 result.append("enabled hypervisors contains invalid entries: %s" %
598 utils.CommaJoin(invalid_hvs))
599 missing_hvp = (set(cluster.enabled_hypervisors) -
600 set(cluster.hvparams.keys()))
602 result.append("hypervisor parameters missing for the enabled"
603 " hypervisor(s) %s" % utils.CommaJoin(missing_hvp))
605 if not cluster.enabled_disk_templates:
606 result.append("enabled disk templates list doesn't have any entries")
607 invalid_disk_templates = set(cluster.enabled_disk_templates) \
608 - constants.DISK_TEMPLATES
609 if invalid_disk_templates:
610 result.append("enabled disk templates list contains invalid entries:"
611 " %s" % utils.CommaJoin(invalid_disk_templates))
613 if cluster.master_node not in data.nodes:
614 result.append("cluster has invalid primary node '%s'" %
617 def _helper(owner, attr, value, template):
619 utils.ForceDictType(value, template)
620 except errors.GenericError, err:
621 result.append("%s has invalid %s: %s" % (owner, attr, err))
623 def _helper_nic(owner, params):
625 objects.NIC.CheckParameterSyntax(params)
626 except errors.ConfigurationError, err:
627 result.append("%s has invalid nicparams: %s" % (owner, err))
629 def _helper_ipolicy(owner, ipolicy, iscluster):
631 objects.InstancePolicy.CheckParameterSyntax(ipolicy, iscluster)
632 except errors.ConfigurationError, err:
633 result.append("%s has invalid instance policy: %s" % (owner, err))
634 for key, value in ipolicy.items():
635 if key == constants.ISPECS_MINMAX:
636 for k in range(len(value)):
637 _helper_ispecs(owner, "ipolicy/%s[%s]" % (key, k), value[k])
638 elif key == constants.ISPECS_STD:
639 _helper(owner, "ipolicy/" + key, value,
640 constants.ISPECS_PARAMETER_TYPES)
642 # FIXME: assuming list type
643 if key in constants.IPOLICY_PARAMETERS:
647 if not isinstance(value, exp_type):
648 result.append("%s has invalid instance policy: for %s,"
649 " expecting %s, got %s" %
650 (owner, key, exp_type.__name__, type(value)))
652 def _helper_ispecs(owner, parentkey, params):
653 for (key, value) in params.items():
654 fullkey = "/".join([parentkey, key])
655 _helper(owner, fullkey, value, constants.ISPECS_PARAMETER_TYPES)
657 # check cluster parameters
658 _helper("cluster", "beparams", cluster.SimpleFillBE({}),
659 constants.BES_PARAMETER_TYPES)
660 _helper("cluster", "nicparams", cluster.SimpleFillNIC({}),
661 constants.NICS_PARAMETER_TYPES)
662 _helper_nic("cluster", cluster.SimpleFillNIC({}))
663 _helper("cluster", "ndparams", cluster.SimpleFillND({}),
664 constants.NDS_PARAMETER_TYPES)
665 _helper_ipolicy("cluster", cluster.ipolicy, True)
667 # per-instance checks
668 for instance_name in data.instances:
669 instance = data.instances[instance_name]
670 if instance.name != instance_name:
671 result.append("instance '%s' is indexed by wrong name '%s'" %
672 (instance.name, instance_name))
673 if instance.primary_node not in data.nodes:
674 result.append("instance '%s' has invalid primary node '%s'" %
675 (instance_name, instance.primary_node))
676 for snode in instance.secondary_nodes:
677 if snode not in data.nodes:
678 result.append("instance '%s' has invalid secondary node '%s'" %
679 (instance_name, snode))
680 for idx, nic in enumerate(instance.nics):
681 if nic.mac in seen_macs:
682 result.append("instance '%s' has NIC %d mac %s duplicate" %
683 (instance_name, idx, nic.mac))
685 seen_macs.append(nic.mac)
687 filled = cluster.SimpleFillNIC(nic.nicparams)
688 owner = "instance %s nic %d" % (instance.name, idx)
689 _helper(owner, "nicparams",
690 filled, constants.NICS_PARAMETER_TYPES)
691 _helper_nic(owner, filled)
693 # disk template checks
694 if not instance.disk_template in data.cluster.enabled_disk_templates:
695 result.append("instance '%s' uses the disabled disk template '%s'." %
696 (instance_name, instance.disk_template))
699 if instance.beparams:
700 _helper("instance %s" % instance.name, "beparams",
701 cluster.FillBE(instance), constants.BES_PARAMETER_TYPES)
703 # gather the drbd ports for duplicate checks
704 for (idx, dsk) in enumerate(instance.disks):
705 if dsk.dev_type in constants.LDS_DRBD:
706 tcp_port = dsk.logical_id[2]
707 if tcp_port not in ports:
709 ports[tcp_port].append((instance.name, "drbd disk %s" % idx))
710 # gather network port reservation
711 net_port = getattr(instance, "network_port", None)
712 if net_port is not None:
713 if net_port not in ports:
715 ports[net_port].append((instance.name, "network port"))
717 # instance disk verify
718 for idx, disk in enumerate(instance.disks):
719 result.extend(["instance '%s' disk %d error: %s" %
720 (instance.name, idx, msg) for msg in disk.Verify()])
721 result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
723 wrong_names = _CheckInstanceDiskIvNames(instance.disks)
725 tmp = "; ".join(("name of disk %s should be '%s', but is '%s'" %
726 (idx, exp_name, actual_name))
727 for (idx, exp_name, actual_name) in wrong_names)
729 result.append("Instance '%s' has wrongly named disks: %s" %
730 (instance.name, tmp))
732 # cluster-wide pool of free ports
733 for free_port in cluster.tcpudp_port_pool:
734 if free_port not in ports:
735 ports[free_port] = []
736 ports[free_port].append(("cluster", "port marked as free"))
738 # compute tcp/udp duplicate ports
744 txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
745 result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
747 # highest used tcp port check
749 if keys[-1] > cluster.highest_used_port:
750 result.append("Highest used port mismatch, saved %s, computed %s" %
751 (cluster.highest_used_port, keys[-1]))
753 if not data.nodes[cluster.master_node].master_candidate:
754 result.append("Master node is not a master candidate")
756 # master candidate checks
757 mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
759 result.append("Not enough master candidates: actual %d, target %d" %
763 for node_name, node in data.nodes.items():
764 if node.name != node_name:
765 result.append("Node '%s' is indexed by wrong name '%s'" %
766 (node.name, node_name))
767 if [node.master_candidate, node.drained, node.offline].count(True) > 1:
768 result.append("Node %s state is invalid: master_candidate=%s,"
769 " drain=%s, offline=%s" %
770 (node.name, node.master_candidate, node.drained,
772 if node.group not in data.nodegroups:
773 result.append("Node '%s' has invalid group '%s'" %
774 (node.name, node.group))
776 _helper("node %s" % node.name, "ndparams",
777 cluster.FillND(node, data.nodegroups[node.group]),
778 constants.NDS_PARAMETER_TYPES)
779 used_globals = constants.NDC_GLOBALS.intersection(node.ndparams)
781 result.append("Node '%s' has some global parameters set: %s" %
782 (node.name, utils.CommaJoin(used_globals)))
785 nodegroups_names = set()
786 for nodegroup_uuid in data.nodegroups:
787 nodegroup = data.nodegroups[nodegroup_uuid]
788 if nodegroup.uuid != nodegroup_uuid:
789 result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'"
790 % (nodegroup.name, nodegroup.uuid, nodegroup_uuid))
791 if utils.UUID_RE.match(nodegroup.name.lower()):
792 result.append("node group '%s' (uuid: '%s') has uuid-like name" %
793 (nodegroup.name, nodegroup.uuid))
794 if nodegroup.name in nodegroups_names:
795 result.append("duplicate node group name '%s'" % nodegroup.name)
797 nodegroups_names.add(nodegroup.name)
798 group_name = "group %s" % nodegroup.name
799 _helper_ipolicy(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy),
801 if nodegroup.ndparams:
802 _helper(group_name, "ndparams",
803 cluster.SimpleFillND(nodegroup.ndparams),
804 constants.NDS_PARAMETER_TYPES)
807 _, duplicates = self._UnlockedComputeDRBDMap()
808 for node, minor, instance_a, instance_b in duplicates:
809 result.append("DRBD minor %d on node %s is assigned twice to instances"
810 " %s and %s" % (minor, node, instance_a, instance_b))
813 default_nicparams = cluster.nicparams[constants.PP_DEFAULT]
816 def _AddIpAddress(ip, name):
817 ips.setdefault(ip, []).append(name)
819 _AddIpAddress(cluster.master_ip, "cluster_ip")
821 for node in data.nodes.values():
822 _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
823 if node.secondary_ip != node.primary_ip:
824 _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
826 for instance in data.instances.values():
827 for idx, nic in enumerate(instance.nics):
831 nicparams = objects.FillDict(default_nicparams, nic.nicparams)
832 nic_mode = nicparams[constants.NIC_MODE]
833 nic_link = nicparams[constants.NIC_LINK]
835 if nic_mode == constants.NIC_MODE_BRIDGED:
836 link = "bridge:%s" % nic_link
837 elif nic_mode == constants.NIC_MODE_ROUTED:
838 link = "route:%s" % nic_link
840 raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
842 _AddIpAddress("%s/%s/%s" % (link, nic.ip, nic.network),
843 "instance:%s/nic:%d" % (instance.name, idx))
845 for ip, owners in ips.items():
847 result.append("IP address %s is used by multiple owners: %s" %
848 (ip, utils.CommaJoin(owners)))
852 @locking.ssynchronized(_config_lock, shared=1)
853 def VerifyConfig(self):
856 This is just a wrapper over L{_UnlockedVerifyConfig}.
859 @return: a list of error messages; a non-empty list signifies
863 return self._UnlockedVerifyConfig()
865 def _UnlockedSetDiskID(self, disk, node_name):
866 """Convert the unique ID to the ID needed on the target nodes.
868 This is used only for drbd, which needs ip/port configuration.
870 The routine descends down and updates its children also, because
871 this helps when the only the top device is passed to the remote
874 This function is for internal use, when the config lock is already held.
878 for child in disk.children:
879 self._UnlockedSetDiskID(child, node_name)
881 if disk.logical_id is None and disk.physical_id is not None:
883 if disk.dev_type == constants.LD_DRBD8:
884 pnode, snode, port, pminor, sminor, secret = disk.logical_id
885 if node_name not in (pnode, snode):
886 raise errors.ConfigurationError("DRBD device not knowing node %s" %
888 pnode_info = self._UnlockedGetNodeInfo(pnode)
889 snode_info = self._UnlockedGetNodeInfo(snode)
890 if pnode_info is None or snode_info is None:
891 raise errors.ConfigurationError("Can't find primary or secondary node"
892 " for %s" % str(disk))
893 p_data = (pnode_info.secondary_ip, port)
894 s_data = (snode_info.secondary_ip, port)
895 if pnode == node_name:
896 disk.physical_id = p_data + s_data + (pminor, secret)
897 else: # it must be secondary, we tested above
898 disk.physical_id = s_data + p_data + (sminor, secret)
900 disk.physical_id = disk.logical_id
903 @locking.ssynchronized(_config_lock)
904 def SetDiskID(self, disk, node_name):
905 """Convert the unique ID to the ID needed on the target nodes.
907 This is used only for drbd, which needs ip/port configuration.
909 The routine descends down and updates its children also, because
910 this helps when the only the top device is passed to the remote
914 return self._UnlockedSetDiskID(disk, node_name)
916 @locking.ssynchronized(_config_lock)
917 def AddTcpUdpPort(self, port):
918 """Adds a new port to the available port pool.
920 @warning: this method does not "flush" the configuration (via
921 L{_WriteConfig}); callers should do that themselves once the
922 configuration is stable
925 if not isinstance(port, int):
926 raise errors.ProgrammerError("Invalid type passed for port")
928 self._config_data.cluster.tcpudp_port_pool.add(port)
930 @locking.ssynchronized(_config_lock, shared=1)
931 def GetPortList(self):
932 """Returns a copy of the current port list.
935 return self._config_data.cluster.tcpudp_port_pool.copy()
937 @locking.ssynchronized(_config_lock)
938 def AllocatePort(self):
941 The port will be taken from the available port pool or from the
942 default port range (and in this case we increase
946 # If there are TCP/IP ports configured, we use them first.
947 if self._config_data.cluster.tcpudp_port_pool:
948 port = self._config_data.cluster.tcpudp_port_pool.pop()
950 port = self._config_data.cluster.highest_used_port + 1
951 if port >= constants.LAST_DRBD_PORT:
952 raise errors.ConfigurationError("The highest used port is greater"
953 " than %s. Aborting." %
954 constants.LAST_DRBD_PORT)
955 self._config_data.cluster.highest_used_port = port
960 def _UnlockedComputeDRBDMap(self):
961 """Compute the used DRBD minor/nodes.
964 @return: dictionary of node_name: dict of minor: instance_name;
965 the returned dict will have all the nodes in it (even if with
966 an empty list), and a list of duplicates; if the duplicates
967 list is not empty, the configuration is corrupted and its caller
968 should raise an exception
971 def _AppendUsedPorts(instance_name, disk, used):
973 if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:
974 node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
975 for node, port in ((node_a, minor_a), (node_b, minor_b)):
976 assert node in used, ("Node '%s' of instance '%s' not found"
977 " in node list" % (node, instance_name))
978 if port in used[node]:
979 duplicates.append((node, port, instance_name, used[node][port]))
981 used[node][port] = instance_name
983 for child in disk.children:
984 duplicates.extend(_AppendUsedPorts(instance_name, child, used))
988 my_dict = dict((node, {}) for node in self._config_data.nodes)
989 for instance in self._config_data.instances.itervalues():
990 for disk in instance.disks:
991 duplicates.extend(_AppendUsedPorts(instance.name, disk, my_dict))
992 for (node, minor), instance in self._temporary_drbds.iteritems():
993 if minor in my_dict[node] and my_dict[node][minor] != instance:
994 duplicates.append((node, minor, instance, my_dict[node][minor]))
996 my_dict[node][minor] = instance
997 return my_dict, duplicates
999 @locking.ssynchronized(_config_lock)
1000 def ComputeDRBDMap(self):
1001 """Compute the used DRBD minor/nodes.
1003 This is just a wrapper over L{_UnlockedComputeDRBDMap}.
1005 @return: dictionary of node_name: dict of minor: instance_name;
1006 the returned dict will have all the nodes in it (even if with
1010 d_map, duplicates = self._UnlockedComputeDRBDMap()
1012 raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1016 @locking.ssynchronized(_config_lock)
1017 def AllocateDRBDMinor(self, nodes, instance):
1018 """Allocate a drbd minor.
1020 The free minor will be automatically computed from the existing
1021 devices. A node can be given multiple times in order to allocate
1022 multiple minors. The result is the list of minors, in the same
1023 order as the passed nodes.
1025 @type instance: string
1026 @param instance: the instance for which we allocate minors
1029 assert isinstance(instance, basestring), \
1030 "Invalid argument '%s' passed to AllocateDRBDMinor" % instance
1032 d_map, duplicates = self._UnlockedComputeDRBDMap()
1034 raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1038 ndata = d_map[nname]
1040 # no minors used, we can start at 0
1043 self._temporary_drbds[(nname, 0)] = instance
1047 ffree = utils.FirstFree(keys)
1049 # return the next minor
1050 # TODO: implement high-limit check
1051 minor = keys[-1] + 1
1054 # double-check minor against current instances
1055 assert minor not in d_map[nname], \
1056 ("Attempt to reuse allocated DRBD minor %d on node %s,"
1057 " already allocated to instance %s" %
1058 (minor, nname, d_map[nname][minor]))
1059 ndata[minor] = instance
1060 # double-check minor against reservation
1061 r_key = (nname, minor)
1062 assert r_key not in self._temporary_drbds, \
1063 ("Attempt to reuse reserved DRBD minor %d on node %s,"
1064 " reserved for instance %s" %
1065 (minor, nname, self._temporary_drbds[r_key]))
1066 self._temporary_drbds[r_key] = instance
1067 result.append(minor)
1068 logging.debug("Request to allocate drbd minors, input: %s, returning %s",
1072 def _UnlockedReleaseDRBDMinors(self, instance):
1073 """Release temporary drbd minors allocated for a given instance.
1075 @type instance: string
1076 @param instance: the instance for which temporary minors should be
1080 assert isinstance(instance, basestring), \
1081 "Invalid argument passed to ReleaseDRBDMinors"
1082 for key, name in self._temporary_drbds.items():
1083 if name == instance:
1084 del self._temporary_drbds[key]
1086 @locking.ssynchronized(_config_lock)
1087 def ReleaseDRBDMinors(self, instance):
1088 """Release temporary drbd minors allocated for a given instance.
1090 This should be called on the error paths, on the success paths
1091 it's automatically called by the ConfigWriter add and update
1094 This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
1096 @type instance: string
1097 @param instance: the instance for which temporary minors should be
1101 self._UnlockedReleaseDRBDMinors(instance)
1103 @locking.ssynchronized(_config_lock, shared=1)
1104 def GetConfigVersion(self):
1105 """Get the configuration version.
1107 @return: Config version
1110 return self._config_data.version
1112 @locking.ssynchronized(_config_lock, shared=1)
1113 def GetClusterName(self):
1114 """Get cluster name.
1116 @return: Cluster name
1119 return self._config_data.cluster.cluster_name
1121 @locking.ssynchronized(_config_lock, shared=1)
1122 def GetMasterNode(self):
1123 """Get the hostname of the master node for this cluster.
1125 @return: Master hostname
1128 return self._config_data.cluster.master_node
1130 @locking.ssynchronized(_config_lock, shared=1)
1131 def GetMasterIP(self):
1132 """Get the IP of the master node for this cluster.
1137 return self._config_data.cluster.master_ip
1139 @locking.ssynchronized(_config_lock, shared=1)
1140 def GetMasterNetdev(self):
1141 """Get the master network device for this cluster.
1144 return self._config_data.cluster.master_netdev
1146 @locking.ssynchronized(_config_lock, shared=1)
1147 def GetMasterNetmask(self):
1148 """Get the netmask of the master node for this cluster.
1151 return self._config_data.cluster.master_netmask
1153 @locking.ssynchronized(_config_lock, shared=1)
1154 def GetUseExternalMipScript(self):
1155 """Get flag representing whether to use the external master IP setup script.
1158 return self._config_data.cluster.use_external_mip_script
1160 @locking.ssynchronized(_config_lock, shared=1)
1161 def GetFileStorageDir(self):
1162 """Get the file storage dir for this cluster.
1165 return self._config_data.cluster.file_storage_dir
1167 @locking.ssynchronized(_config_lock, shared=1)
1168 def GetSharedFileStorageDir(self):
1169 """Get the shared file storage dir for this cluster.
1172 return self._config_data.cluster.shared_file_storage_dir
1174 @locking.ssynchronized(_config_lock, shared=1)
1175 def GetHypervisorType(self):
1176 """Get the hypervisor type for this cluster.
1179 return self._config_data.cluster.enabled_hypervisors[0]
1181 @locking.ssynchronized(_config_lock, shared=1)
1182 def GetHostKey(self):
1183 """Return the rsa hostkey from the config.
1186 @return: the rsa hostkey
1189 return self._config_data.cluster.rsahostkeypub
1191 @locking.ssynchronized(_config_lock, shared=1)
1192 def GetDefaultIAllocator(self):
1193 """Get the default instance allocator for this cluster.
1196 return self._config_data.cluster.default_iallocator
1198 @locking.ssynchronized(_config_lock, shared=1)
1199 def GetPrimaryIPFamily(self):
1200 """Get cluster primary ip family.
1202 @return: primary ip family
1205 return self._config_data.cluster.primary_ip_family
1207 @locking.ssynchronized(_config_lock, shared=1)
1208 def GetMasterNetworkParameters(self):
1209 """Get network parameters of the master node.
1211 @rtype: L{object.MasterNetworkParameters}
1212 @return: network parameters of the master node
1215 cluster = self._config_data.cluster
1216 result = objects.MasterNetworkParameters(
1217 name=cluster.master_node, ip=cluster.master_ip,
1218 netmask=cluster.master_netmask, netdev=cluster.master_netdev,
1219 ip_family=cluster.primary_ip_family)
1223 @locking.ssynchronized(_config_lock)
1224 def AddNodeGroup(self, group, ec_id, check_uuid=True):
1225 """Add a node group to the configuration.
1227 This method calls group.UpgradeConfig() to fill any missing attributes
1228 according to their default values.
1230 @type group: L{objects.NodeGroup}
1231 @param group: the NodeGroup object to add
1233 @param ec_id: unique id for the job to use when creating a missing UUID
1234 @type check_uuid: bool
1235 @param check_uuid: add an UUID to the group if it doesn't have one or, if
1236 it does, ensure that it does not exist in the
1237 configuration already
1240 self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1243 def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid):
1244 """Add a node group to the configuration.
1247 logging.info("Adding node group %s to configuration", group.name)
1249 # Some code might need to add a node group with a pre-populated UUID
1250 # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass
1251 # the "does this UUID" exist already check.
1253 self._EnsureUUID(group, ec_id)
1256 existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1257 except errors.OpPrereqError:
1260 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1261 " node group (UUID: %s)" %
1262 (group.name, existing_uuid),
1263 errors.ECODE_EXISTS)
1266 group.ctime = group.mtime = time.time()
1267 group.UpgradeConfig()
1269 self._config_data.nodegroups[group.uuid] = group
1270 self._config_data.cluster.serial_no += 1
1272 @locking.ssynchronized(_config_lock)
1273 def RemoveNodeGroup(self, group_uuid):
1274 """Remove a node group from the configuration.
1276 @type group_uuid: string
1277 @param group_uuid: the UUID of the node group to remove
1280 logging.info("Removing node group %s from configuration", group_uuid)
1282 if group_uuid not in self._config_data.nodegroups:
1283 raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1285 assert len(self._config_data.nodegroups) != 1, \
1286 "Group '%s' is the only group, cannot be removed" % group_uuid
1288 del self._config_data.nodegroups[group_uuid]
1289 self._config_data.cluster.serial_no += 1
1292 def _UnlockedLookupNodeGroup(self, target):
1293 """Lookup a node group's UUID.
1295 @type target: string or None
1296 @param target: group name or UUID or None to look for the default
1298 @return: nodegroup UUID
1299 @raises errors.OpPrereqError: when the target group cannot be found
1303 if len(self._config_data.nodegroups) != 1:
1304 raise errors.OpPrereqError("More than one node group exists. Target"
1305 " group must be specified explicitly.")
1307 return self._config_data.nodegroups.keys()[0]
1308 if target in self._config_data.nodegroups:
1310 for nodegroup in self._config_data.nodegroups.values():
1311 if nodegroup.name == target:
1312 return nodegroup.uuid
1313 raise errors.OpPrereqError("Node group '%s' not found" % target,
1316 @locking.ssynchronized(_config_lock, shared=1)
1317 def LookupNodeGroup(self, target):
1318 """Lookup a node group's UUID.
1320 This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1322 @type target: string or None
1323 @param target: group name or UUID or None to look for the default
1325 @return: nodegroup UUID
1328 return self._UnlockedLookupNodeGroup(target)
1330 def _UnlockedGetNodeGroup(self, uuid):
1331 """Lookup a node group.
1334 @param uuid: group UUID
1335 @rtype: L{objects.NodeGroup} or None
1336 @return: nodegroup object, or None if not found
1339 if uuid not in self._config_data.nodegroups:
1342 return self._config_data.nodegroups[uuid]
1344 @locking.ssynchronized(_config_lock, shared=1)
1345 def GetNodeGroup(self, uuid):
1346 """Lookup a node group.
1349 @param uuid: group UUID
1350 @rtype: L{objects.NodeGroup} or None
1351 @return: nodegroup object, or None if not found
1354 return self._UnlockedGetNodeGroup(uuid)
1356 @locking.ssynchronized(_config_lock, shared=1)
1357 def GetAllNodeGroupsInfo(self):
1358 """Get the configuration of all node groups.
1361 return dict(self._config_data.nodegroups)
1363 @locking.ssynchronized(_config_lock, shared=1)
1364 def GetNodeGroupList(self):
1365 """Get a list of node groups.
1368 return self._config_data.nodegroups.keys()
1370 @locking.ssynchronized(_config_lock, shared=1)
1371 def GetNodeGroupMembersByNodes(self, nodes):
1372 """Get nodes which are member in the same nodegroups as the given nodes.
1375 ngfn = lambda node_name: self._UnlockedGetNodeInfo(node_name).group
1376 return frozenset(member_name
1377 for node_name in nodes
1379 self._UnlockedGetNodeGroup(ngfn(node_name)).members)
1381 @locking.ssynchronized(_config_lock, shared=1)
1382 def GetMultiNodeGroupInfo(self, group_uuids):
1383 """Get the configuration of multiple node groups.
1385 @param group_uuids: List of node group UUIDs
1387 @return: List of tuples of (group_uuid, group_info)
1390 return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1392 @locking.ssynchronized(_config_lock)
1393 def AddInstance(self, instance, ec_id):
1394 """Add an instance to the config.
1396 This should be used after creating a new instance.
1398 @type instance: L{objects.Instance}
1399 @param instance: the instance object
1402 if not isinstance(instance, objects.Instance):
1403 raise errors.ProgrammerError("Invalid type passed to AddInstance")
1405 if instance.disk_template != constants.DT_DISKLESS:
1406 all_lvs = instance.MapLVsByNode()
1407 logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
1409 all_macs = self._AllMACs()
1410 for nic in instance.nics:
1411 if nic.mac in all_macs:
1412 raise errors.ConfigurationError("Cannot add instance %s:"
1413 " MAC address '%s' already in use." %
1414 (instance.name, nic.mac))
1416 self._EnsureUUID(instance, ec_id)
1418 instance.serial_no = 1
1419 instance.ctime = instance.mtime = time.time()
1420 self._config_data.instances[instance.name] = instance
1421 self._config_data.cluster.serial_no += 1
1422 self._UnlockedReleaseDRBDMinors(instance.name)
1423 self._UnlockedCommitTemporaryIps(ec_id)
1426 def _EnsureUUID(self, item, ec_id):
1427 """Ensures a given object has a valid UUID.
1429 @param item: the instance or node to be checked
1430 @param ec_id: the execution context id for the uuid reservation
1434 item.uuid = self._GenerateUniqueID(ec_id)
1435 elif item.uuid in self._AllIDs(include_temporary=True):
1436 raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1437 " in use" % (item.name, item.uuid))
1439 def _SetInstanceStatus(self, instance_name, status):
1440 """Set the instance's status to a given value.
1443 assert status in constants.ADMINST_ALL, \
1444 "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1446 if instance_name not in self._config_data.instances:
1447 raise errors.ConfigurationError("Unknown instance '%s'" %
1449 instance = self._config_data.instances[instance_name]
1450 if instance.admin_state != status:
1451 instance.admin_state = status
1452 instance.serial_no += 1
1453 instance.mtime = time.time()
1456 @locking.ssynchronized(_config_lock)
1457 def MarkInstanceUp(self, instance_name):
1458 """Mark the instance status to up in the config.
1461 self._SetInstanceStatus(instance_name, constants.ADMINST_UP)
1463 @locking.ssynchronized(_config_lock)
1464 def MarkInstanceOffline(self, instance_name):
1465 """Mark the instance status to down in the config.
1468 self._SetInstanceStatus(instance_name, constants.ADMINST_OFFLINE)
1470 @locking.ssynchronized(_config_lock)
1471 def RemoveInstance(self, instance_name):
1472 """Remove the instance from the configuration.
1475 if instance_name not in self._config_data.instances:
1476 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1478 # If a network port has been allocated to the instance,
1479 # return it to the pool of free ports.
1480 inst = self._config_data.instances[instance_name]
1481 network_port = getattr(inst, "network_port", None)
1482 if network_port is not None:
1483 self._config_data.cluster.tcpudp_port_pool.add(network_port)
1485 instance = self._UnlockedGetInstanceInfo(instance_name)
1487 for nic in instance.nics:
1488 if nic.network and nic.ip:
1489 # Return all IP addresses to the respective address pools
1490 self._UnlockedCommitIp(constants.RELEASE_ACTION, nic.network, nic.ip)
1492 del self._config_data.instances[instance_name]
1493 self._config_data.cluster.serial_no += 1
1496 @locking.ssynchronized(_config_lock)
1497 def RenameInstance(self, old_name, new_name):
1498 """Rename an instance.
1500 This needs to be done in ConfigWriter and not by RemoveInstance
1501 combined with AddInstance as only we can guarantee an atomic
1505 if old_name not in self._config_data.instances:
1506 raise errors.ConfigurationError("Unknown instance '%s'" % old_name)
1508 # Operate on a copy to not loose instance object in case of a failure
1509 inst = self._config_data.instances[old_name].Copy()
1510 inst.name = new_name
1512 for (idx, disk) in enumerate(inst.disks):
1513 if disk.dev_type == constants.LD_FILE:
1514 # rename the file paths in logical and physical id
1515 file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1516 disk.logical_id = (disk.logical_id[0],
1517 utils.PathJoin(file_storage_dir, inst.name,
1519 disk.physical_id = disk.logical_id
1521 # Actually replace instance object
1522 del self._config_data.instances[old_name]
1523 self._config_data.instances[inst.name] = inst
1525 # Force update of ssconf files
1526 self._config_data.cluster.serial_no += 1
1530 @locking.ssynchronized(_config_lock)
1531 def MarkInstanceDown(self, instance_name):
1532 """Mark the status of an instance to down in the configuration.
1535 self._SetInstanceStatus(instance_name, constants.ADMINST_DOWN)
1537 def _UnlockedGetInstanceList(self):
1538 """Get the list of instances.
1540 This function is for internal use, when the config lock is already held.
1543 return self._config_data.instances.keys()
1545 @locking.ssynchronized(_config_lock, shared=1)
1546 def GetInstanceList(self):
1547 """Get the list of instances.
1549 @return: array of instances, ex. ['instance2.example.com',
1550 'instance1.example.com']
1553 return self._UnlockedGetInstanceList()
1555 def ExpandInstanceName(self, short_name):
1556 """Attempt to expand an incomplete instance name.
1559 # Locking is done in L{ConfigWriter.GetInstanceList}
1560 return _MatchNameComponentIgnoreCase(short_name, self.GetInstanceList())
1562 def _UnlockedGetInstanceInfo(self, instance_name):
1563 """Returns information about an instance.
1565 This function is for internal use, when the config lock is already held.
1568 if instance_name not in self._config_data.instances:
1571 return self._config_data.instances[instance_name]
1573 @locking.ssynchronized(_config_lock, shared=1)
1574 def GetInstanceInfo(self, instance_name):
1575 """Returns information about an instance.
1577 It takes the information from the configuration file. Other information of
1578 an instance are taken from the live systems.
1580 @param instance_name: name of the instance, e.g.
1581 I{instance1.example.com}
1583 @rtype: L{objects.Instance}
1584 @return: the instance object
1587 return self._UnlockedGetInstanceInfo(instance_name)
1589 @locking.ssynchronized(_config_lock, shared=1)
1590 def GetInstanceNodeGroups(self, instance_name, primary_only=False):
1591 """Returns set of node group UUIDs for instance's nodes.
1596 instance = self._UnlockedGetInstanceInfo(instance_name)
1598 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1601 nodes = [instance.primary_node]
1603 nodes = instance.all_nodes
1605 return frozenset(self._UnlockedGetNodeInfo(node_name).group
1606 for node_name in nodes)
1608 @locking.ssynchronized(_config_lock, shared=1)
1609 def GetInstanceNetworks(self, instance_name):
1610 """Returns set of network UUIDs for instance's nics.
1615 instance = self._UnlockedGetInstanceInfo(instance_name)
1617 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1620 for nic in instance.nics:
1622 networks.add(nic.network)
1624 return frozenset(networks)
1626 @locking.ssynchronized(_config_lock, shared=1)
1627 def GetMultiInstanceInfo(self, instances):
1628 """Get the configuration of multiple instances.
1630 @param instances: list of instance names
1632 @return: list of tuples (instance, instance_info), where
1633 instance_info is what would GetInstanceInfo return for the
1634 node, while keeping the original order
1637 return [(name, self._UnlockedGetInstanceInfo(name)) for name in instances]
1639 @locking.ssynchronized(_config_lock, shared=1)
1640 def GetAllInstancesInfo(self):
1641 """Get the configuration of all instances.
1644 @return: dict of (instance, instance_info), where instance_info is what
1645 would GetInstanceInfo return for the node
1648 my_dict = dict([(instance, self._UnlockedGetInstanceInfo(instance))
1649 for instance in self._UnlockedGetInstanceList()])
1652 @locking.ssynchronized(_config_lock, shared=1)
1653 def GetInstancesInfoByFilter(self, filter_fn):
1654 """Get instance configuration with a filter.
1656 @type filter_fn: callable
1657 @param filter_fn: Filter function receiving instance object as parameter,
1658 returning boolean. Important: this function is called while the
1659 configuration locks is held. It must not do any complex work or call
1660 functions potentially leading to a deadlock. Ideally it doesn't call any
1661 other functions and just compares instance attributes.
1664 return dict((name, inst)
1665 for (name, inst) in self._config_data.instances.items()
1668 @locking.ssynchronized(_config_lock)
1669 def AddNode(self, node, ec_id):
1670 """Add a node to the configuration.
1672 @type node: L{objects.Node}
1673 @param node: a Node instance
1676 logging.info("Adding node %s to configuration", node.name)
1678 self._EnsureUUID(node, ec_id)
1681 node.ctime = node.mtime = time.time()
1682 self._UnlockedAddNodeToGroup(node.name, node.group)
1683 self._config_data.nodes[node.name] = node
1684 self._config_data.cluster.serial_no += 1
1687 @locking.ssynchronized(_config_lock)
1688 def RemoveNode(self, node_name):
1689 """Remove a node from the configuration.
1692 logging.info("Removing node %s from configuration", node_name)
1694 if node_name not in self._config_data.nodes:
1695 raise errors.ConfigurationError("Unknown node '%s'" % node_name)
1697 self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_name])
1698 del self._config_data.nodes[node_name]
1699 self._config_data.cluster.serial_no += 1
1702 def ExpandNodeName(self, short_name):
1703 """Attempt to expand an incomplete node name.
1706 # Locking is done in L{ConfigWriter.GetNodeList}
1707 return _MatchNameComponentIgnoreCase(short_name, self.GetNodeList())
1709 def _UnlockedGetNodeInfo(self, node_name):
1710 """Get the configuration of a node, as stored in the config.
1712 This function is for internal use, when the config lock is already
1715 @param node_name: the node name, e.g. I{node1.example.com}
1717 @rtype: L{objects.Node}
1718 @return: the node object
1721 if node_name not in self._config_data.nodes:
1724 return self._config_data.nodes[node_name]
1726 @locking.ssynchronized(_config_lock, shared=1)
1727 def GetNodeInfo(self, node_name):
1728 """Get the configuration of a node, as stored in the config.
1730 This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1732 @param node_name: the node name, e.g. I{node1.example.com}
1734 @rtype: L{objects.Node}
1735 @return: the node object
1738 return self._UnlockedGetNodeInfo(node_name)
1740 @locking.ssynchronized(_config_lock, shared=1)
1741 def GetNodeInstances(self, node_name):
1742 """Get the instances of a node, as stored in the config.
1744 @param node_name: the node name, e.g. I{node1.example.com}
1746 @rtype: (list, list)
1747 @return: a tuple with two lists: the primary and the secondary instances
1752 for inst in self._config_data.instances.values():
1753 if inst.primary_node == node_name:
1754 pri.append(inst.name)
1755 if node_name in inst.secondary_nodes:
1756 sec.append(inst.name)
1759 @locking.ssynchronized(_config_lock, shared=1)
1760 def GetNodeGroupInstances(self, uuid, primary_only=False):
1761 """Get the instances of a node group.
1763 @param uuid: Node group UUID
1764 @param primary_only: Whether to only consider primary nodes
1766 @return: List of instance names in node group
1770 nodes_fn = lambda inst: [inst.primary_node]
1772 nodes_fn = lambda inst: inst.all_nodes
1774 return frozenset(inst.name
1775 for inst in self._config_data.instances.values()
1776 for node_name in nodes_fn(inst)
1777 if self._UnlockedGetNodeInfo(node_name).group == uuid)
1779 def _UnlockedGetNodeList(self):
1780 """Return the list of nodes which are in the configuration.
1782 This function is for internal use, when the config lock is already
1788 return self._config_data.nodes.keys()
1790 @locking.ssynchronized(_config_lock, shared=1)
1791 def GetNodeList(self):
1792 """Return the list of nodes which are in the configuration.
1795 return self._UnlockedGetNodeList()
1797 def _UnlockedGetOnlineNodeList(self):
1798 """Return the list of nodes which are online.
1801 all_nodes = [self._UnlockedGetNodeInfo(node)
1802 for node in self._UnlockedGetNodeList()]
1803 return [node.name for node in all_nodes if not node.offline]
1805 @locking.ssynchronized(_config_lock, shared=1)
1806 def GetOnlineNodeList(self):
1807 """Return the list of nodes which are online.
1810 return self._UnlockedGetOnlineNodeList()
1812 @locking.ssynchronized(_config_lock, shared=1)
1813 def GetVmCapableNodeList(self):
1814 """Return the list of nodes which are not vm capable.
1817 all_nodes = [self._UnlockedGetNodeInfo(node)
1818 for node in self._UnlockedGetNodeList()]
1819 return [node.name for node in all_nodes if node.vm_capable]
1821 @locking.ssynchronized(_config_lock, shared=1)
1822 def GetNonVmCapableNodeList(self):
1823 """Return the list of nodes which are not vm capable.
1826 all_nodes = [self._UnlockedGetNodeInfo(node)
1827 for node in self._UnlockedGetNodeList()]
1828 return [node.name for node in all_nodes if not node.vm_capable]
1830 @locking.ssynchronized(_config_lock, shared=1)
1831 def GetMultiNodeInfo(self, nodes):
1832 """Get the configuration of multiple nodes.
1834 @param nodes: list of node names
1836 @return: list of tuples of (node, node_info), where node_info is
1837 what would GetNodeInfo return for the node, in the original
1841 return [(name, self._UnlockedGetNodeInfo(name)) for name in nodes]
1843 @locking.ssynchronized(_config_lock, shared=1)
1844 def GetAllNodesInfo(self):
1845 """Get the configuration of all nodes.
1848 @return: dict of (node, node_info), where node_info is what
1849 would GetNodeInfo return for the node
1852 return self._UnlockedGetAllNodesInfo()
1854 def _UnlockedGetAllNodesInfo(self):
1855 """Gets configuration of all nodes.
1857 @note: See L{GetAllNodesInfo}
1860 return dict([(node, self._UnlockedGetNodeInfo(node))
1861 for node in self._UnlockedGetNodeList()])
1863 @locking.ssynchronized(_config_lock, shared=1)
1864 def GetNodeGroupsFromNodes(self, nodes):
1865 """Returns groups for a list of nodes.
1867 @type nodes: list of string
1868 @param nodes: List of node names
1872 return frozenset(self._UnlockedGetNodeInfo(name).group for name in nodes)
1874 def _UnlockedGetMasterCandidateStats(self, exceptions=None):
1875 """Get the number of current and maximum desired and possible candidates.
1877 @type exceptions: list
1878 @param exceptions: if passed, list of nodes that should be ignored
1880 @return: tuple of (current, desired and possible, possible)
1883 mc_now = mc_should = mc_max = 0
1884 for node in self._config_data.nodes.values():
1885 if exceptions and node.name in exceptions:
1887 if not (node.offline or node.drained) and node.master_capable:
1889 if node.master_candidate:
1891 mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
1892 return (mc_now, mc_should, mc_max)
1894 @locking.ssynchronized(_config_lock, shared=1)
1895 def GetMasterCandidateStats(self, exceptions=None):
1896 """Get the number of current and maximum possible candidates.
1898 This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
1900 @type exceptions: list
1901 @param exceptions: if passed, list of nodes that should be ignored
1903 @return: tuple of (current, max)
1906 return self._UnlockedGetMasterCandidateStats(exceptions)
1908 @locking.ssynchronized(_config_lock)
1909 def MaintainCandidatePool(self, exceptions):
1910 """Try to grow the candidate pool to the desired size.
1912 @type exceptions: list
1913 @param exceptions: if passed, list of nodes that should be ignored
1915 @return: list with the adjusted nodes (L{objects.Node} instances)
1918 mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(exceptions)
1921 node_list = self._config_data.nodes.keys()
1922 random.shuffle(node_list)
1923 for name in node_list:
1924 if mc_now >= mc_max:
1926 node = self._config_data.nodes[name]
1927 if (node.master_candidate or node.offline or node.drained or
1928 node.name in exceptions or not node.master_capable):
1930 mod_list.append(node)
1931 node.master_candidate = True
1934 if mc_now != mc_max:
1935 # this should not happen
1936 logging.warning("Warning: MaintainCandidatePool didn't manage to"
1937 " fill the candidate pool (%d/%d)", mc_now, mc_max)
1939 self._config_data.cluster.serial_no += 1
1944 def _UnlockedAddNodeToGroup(self, node_name, nodegroup_uuid):
1945 """Add a given node to the specified group.
1948 if nodegroup_uuid not in self._config_data.nodegroups:
1949 # This can happen if a node group gets deleted between its lookup and
1950 # when we're adding the first node to it, since we don't keep a lock in
1951 # the meantime. It's ok though, as we'll fail cleanly if the node group
1952 # is not found anymore.
1953 raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
1954 if node_name not in self._config_data.nodegroups[nodegroup_uuid].members:
1955 self._config_data.nodegroups[nodegroup_uuid].members.append(node_name)
1957 def _UnlockedRemoveNodeFromGroup(self, node):
1958 """Remove a given node from its group.
1961 nodegroup = node.group
1962 if nodegroup not in self._config_data.nodegroups:
1963 logging.warning("Warning: node '%s' has unknown node group '%s'"
1964 " (while being removed from it)", node.name, nodegroup)
1965 nodegroup_obj = self._config_data.nodegroups[nodegroup]
1966 if node.name not in nodegroup_obj.members:
1967 logging.warning("Warning: node '%s' not a member of its node group '%s'"
1968 " (while being removed from it)", node.name, nodegroup)
1970 nodegroup_obj.members.remove(node.name)
1972 @locking.ssynchronized(_config_lock)
1973 def AssignGroupNodes(self, mods):
1974 """Changes the group of a number of nodes.
1976 @type mods: list of tuples; (node name, new group UUID)
1977 @param mods: Node membership modifications
1980 groups = self._config_data.nodegroups
1981 nodes = self._config_data.nodes
1985 # Try to resolve names/UUIDs first
1986 for (node_name, new_group_uuid) in mods:
1988 node = nodes[node_name]
1990 raise errors.ConfigurationError("Unable to find node '%s'" % node_name)
1992 if node.group == new_group_uuid:
1993 # Node is being assigned to its current group
1994 logging.debug("Node '%s' was assigned to its current group (%s)",
1995 node_name, node.group)
1998 # Try to find current group of node
2000 old_group = groups[node.group]
2002 raise errors.ConfigurationError("Unable to find old group '%s'" %
2005 # Try to find new group for node
2007 new_group = groups[new_group_uuid]
2009 raise errors.ConfigurationError("Unable to find new group '%s'" %
2012 assert node.name in old_group.members, \
2013 ("Inconsistent configuration: node '%s' not listed in members for its"
2014 " old group '%s'" % (node.name, old_group.uuid))
2015 assert node.name not in new_group.members, \
2016 ("Inconsistent configuration: node '%s' already listed in members for"
2017 " its new group '%s'" % (node.name, new_group.uuid))
2019 resmod.append((node, old_group, new_group))
2022 for (node, old_group, new_group) in resmod:
2023 assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
2024 "Assigning to current group is not possible"
2026 node.group = new_group.uuid
2028 # Update members of involved groups
2029 if node.name in old_group.members:
2030 old_group.members.remove(node.name)
2031 if node.name not in new_group.members:
2032 new_group.members.append(node.name)
2034 # Update timestamps and serials (only once per node/group object)
2036 for obj in frozenset(itertools.chain(*resmod)): # pylint: disable=W0142
2040 # Force ssconf update
2041 self._config_data.cluster.serial_no += 1
2045 def _BumpSerialNo(self):
2046 """Bump up the serial number of the config.
2049 self._config_data.serial_no += 1
2050 self._config_data.mtime = time.time()
2052 def _AllUUIDObjects(self):
2053 """Returns all objects with uuid attributes.
2056 return (self._config_data.instances.values() +
2057 self._config_data.nodes.values() +
2058 self._config_data.nodegroups.values() +
2059 self._config_data.networks.values() +
2062 [self._config_data.cluster])
2064 def _OpenConfig(self, accept_foreign):
2065 """Read the config data from disk.
2068 raw_data = utils.ReadFile(self._cfg_file)
2071 data = objects.ConfigData.FromDict(serializer.Load(raw_data))
2072 except Exception, err:
2073 raise errors.ConfigurationError(err)
2075 # Make sure the configuration has the right version
2076 _ValidateConfig(data)
2078 if (not hasattr(data, "cluster") or
2079 not hasattr(data.cluster, "rsahostkeypub")):
2080 raise errors.ConfigurationError("Incomplete configuration"
2081 " (missing cluster.rsahostkeypub)")
2083 if data.cluster.master_node != self._my_hostname and not accept_foreign:
2084 msg = ("The configuration denotes node %s as master, while my"
2085 " hostname is %s; opening a foreign configuration is only"
2086 " possible in accept_foreign mode" %
2087 (data.cluster.master_node, self._my_hostname))
2088 raise errors.ConfigurationError(msg)
2090 self._config_data = data
2091 # reset the last serial as -1 so that the next write will cause
2093 self._last_cluster_serial = -1
2095 # Upgrade configuration if needed
2096 self._UpgradeConfig()
2098 self._cfg_id = utils.GetFileID(path=self._cfg_file)
2100 def _UpgradeConfig(self):
2101 """Run any upgrade steps.
2103 This method performs both in-object upgrades and also update some data
2104 elements that need uniqueness across the whole configuration or interact
2107 @warning: this function will call L{_WriteConfig()}, but also
2108 L{DropECReservations} so it needs to be called only from a
2109 "safe" place (the constructor). If one wanted to call it with
2110 the lock held, a DropECReservationUnlocked would need to be
2111 created first, to avoid causing deadlock.
2114 # Keep a copy of the persistent part of _config_data to check for changes
2115 # Serialization doesn't guarantee order in dictionaries
2116 oldconf = copy.deepcopy(self._config_data.ToDict())
2118 # In-object upgrades
2119 self._config_data.UpgradeConfig()
2121 for item in self._AllUUIDObjects():
2122 if item.uuid is None:
2123 item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
2124 if not self._config_data.nodegroups:
2125 default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
2126 default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
2128 self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
2129 for node in self._config_data.nodes.values():
2131 node.group = self.LookupNodeGroup(None)
2132 # This is technically *not* an upgrade, but needs to be done both when
2133 # nodegroups are being added, and upon normally loading the config,
2134 # because the members list of a node group is discarded upon
2135 # serializing/deserializing the object.
2136 self._UnlockedAddNodeToGroup(node.name, node.group)
2138 modified = (oldconf != self._config_data.ToDict())
2141 # This is ok even if it acquires the internal lock, as _UpgradeConfig is
2142 # only called at config init time, without the lock held
2143 self.DropECReservations(_UPGRADE_CONFIG_JID)
2145 config_errors = self._UnlockedVerifyConfig()
2147 errmsg = ("Loaded configuration data is not consistent: %s" %
2148 (utils.CommaJoin(config_errors)))
2149 logging.critical(errmsg)
2151 def _DistributeConfig(self, feedback_fn):
2152 """Distribute the configuration to the other nodes.
2154 Currently, this only copies the configuration file. In the future,
2155 it could be used to encapsulate the 2/3-phase update mechanism.
2165 myhostname = self._my_hostname
2166 # we can skip checking whether _UnlockedGetNodeInfo returns None
2167 # since the node list comes from _UnlocketGetNodeList, and we are
2168 # called with the lock held, so no modifications should take place
2170 for node_name in self._UnlockedGetNodeList():
2171 if node_name == myhostname:
2173 node_info = self._UnlockedGetNodeInfo(node_name)
2174 if not node_info.master_candidate:
2176 node_list.append(node_info.name)
2177 addr_list.append(node_info.primary_ip)
2179 # TODO: Use dedicated resolver talking to config writer for name resolution
2181 self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
2182 for to_node, to_result in result.items():
2183 msg = to_result.fail_msg
2185 msg = ("Copy of file %s to node %s failed: %s" %
2186 (self._cfg_file, to_node, msg))
2196 def _WriteConfig(self, destination=None, feedback_fn=None):
2197 """Write the configuration data to persistent storage.
2200 assert feedback_fn is None or callable(feedback_fn)
2202 # Warn on config errors, but don't abort the save - the
2203 # configuration has already been modified, and we can't revert;
2204 # the best we can do is to warn the user and save as is, leaving
2205 # recovery to the user
2206 config_errors = self._UnlockedVerifyConfig()
2208 errmsg = ("Configuration data is not consistent: %s" %
2209 (utils.CommaJoin(config_errors)))
2210 logging.critical(errmsg)
2214 if destination is None:
2215 destination = self._cfg_file
2216 self._BumpSerialNo()
2217 txt = serializer.Dump(self._config_data.ToDict())
2219 getents = self._getents()
2221 fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2222 close=False, gid=getents.confd_gid, mode=0640)
2223 except errors.LockError:
2224 raise errors.ConfigurationError("The configuration file has been"
2225 " modified since the last write, cannot"
2228 self._cfg_id = utils.GetFileID(fd=fd)
2232 self.write_count += 1
2234 # and redistribute the config file to master candidates
2235 self._DistributeConfig(feedback_fn)
2237 # Write ssconf files on all nodes (including locally)
2238 if self._last_cluster_serial < self._config_data.cluster.serial_no:
2239 if not self._offline:
2240 result = self._GetRpc(None).call_write_ssconf_files(
2241 self._UnlockedGetOnlineNodeList(),
2242 self._UnlockedGetSsconfValues())
2244 for nname, nresu in result.items():
2245 msg = nresu.fail_msg
2247 errmsg = ("Error while uploading ssconf files to"
2248 " node %s: %s" % (nname, msg))
2249 logging.warning(errmsg)
2254 self._last_cluster_serial = self._config_data.cluster.serial_no
2256 def _UnlockedGetSsconfValues(self):
2257 """Return the values needed by ssconf.
2260 @return: a dictionary with keys the ssconf names and values their
2265 instance_names = utils.NiceSort(self._UnlockedGetInstanceList())
2266 node_names = utils.NiceSort(self._UnlockedGetNodeList())
2267 node_info = [self._UnlockedGetNodeInfo(name) for name in node_names]
2268 node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2269 for ninfo in node_info]
2270 node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2271 for ninfo in node_info]
2273 instance_data = fn(instance_names)
2274 off_data = fn(node.name for node in node_info if node.offline)
2275 on_data = fn(node.name for node in node_info if not node.offline)
2276 mc_data = fn(node.name for node in node_info if node.master_candidate)
2277 mc_ips_data = fn(node.primary_ip for node in node_info
2278 if node.master_candidate)
2279 node_data = fn(node_names)
2280 node_pri_ips_data = fn(node_pri_ips)
2281 node_snd_ips_data = fn(node_snd_ips)
2283 cluster = self._config_data.cluster
2284 cluster_tags = fn(cluster.GetTags())
2286 hypervisor_list = fn(cluster.enabled_hypervisors)
2288 uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2290 nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2291 self._config_data.nodegroups.values()]
2292 nodegroups_data = fn(utils.NiceSort(nodegroups))
2293 networks = ["%s %s" % (net.uuid, net.name) for net in
2294 self._config_data.networks.values()]
2295 networks_data = fn(utils.NiceSort(networks))
2298 constants.SS_CLUSTER_NAME: cluster.cluster_name,
2299 constants.SS_CLUSTER_TAGS: cluster_tags,
2300 constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2301 constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2302 constants.SS_MASTER_CANDIDATES: mc_data,
2303 constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2304 constants.SS_MASTER_IP: cluster.master_ip,
2305 constants.SS_MASTER_NETDEV: cluster.master_netdev,
2306 constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2307 constants.SS_MASTER_NODE: cluster.master_node,
2308 constants.SS_NODE_LIST: node_data,
2309 constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2310 constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2311 constants.SS_OFFLINE_NODES: off_data,
2312 constants.SS_ONLINE_NODES: on_data,
2313 constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2314 constants.SS_INSTANCE_LIST: instance_data,
2315 constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2316 constants.SS_HYPERVISOR_LIST: hypervisor_list,
2317 constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2318 constants.SS_UID_POOL: uid_pool,
2319 constants.SS_NODEGROUPS: nodegroups_data,
2320 constants.SS_NETWORKS: networks_data,
2322 bad_values = [(k, v) for k, v in ssconf_values.items()
2323 if not isinstance(v, (str, basestring))]
2325 err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2326 raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2327 " values: %s" % err)
2328 return ssconf_values
2330 @locking.ssynchronized(_config_lock, shared=1)
2331 def GetSsconfValues(self):
2332 """Wrapper using lock around _UnlockedGetSsconf().
2335 return self._UnlockedGetSsconfValues()
2337 @locking.ssynchronized(_config_lock, shared=1)
2338 def GetVGName(self):
2339 """Return the volume group name.
2342 return self._config_data.cluster.volume_group_name
2344 @locking.ssynchronized(_config_lock)
2345 def SetVGName(self, vg_name):
2346 """Set the volume group name.
2349 self._config_data.cluster.volume_group_name = vg_name
2350 self._config_data.cluster.serial_no += 1
2353 @locking.ssynchronized(_config_lock, shared=1)
2354 def GetDRBDHelper(self):
2355 """Return DRBD usermode helper.
2358 return self._config_data.cluster.drbd_usermode_helper
2360 @locking.ssynchronized(_config_lock)
2361 def SetDRBDHelper(self, drbd_helper):
2362 """Set DRBD usermode helper.
2365 self._config_data.cluster.drbd_usermode_helper = drbd_helper
2366 self._config_data.cluster.serial_no += 1
2369 @locking.ssynchronized(_config_lock, shared=1)
2370 def GetMACPrefix(self):
2371 """Return the mac prefix.
2374 return self._config_data.cluster.mac_prefix
2376 @locking.ssynchronized(_config_lock, shared=1)
2377 def GetClusterInfo(self):
2378 """Returns information about the cluster
2380 @rtype: L{objects.Cluster}
2381 @return: the cluster object
2384 return self._config_data.cluster
2386 @locking.ssynchronized(_config_lock, shared=1)
2387 def HasAnyDiskOfType(self, dev_type):
2388 """Check if in there is at disk of the given type in the configuration.
2391 return self._config_data.HasAnyDiskOfType(dev_type)
2393 @locking.ssynchronized(_config_lock)
2394 def Update(self, target, feedback_fn, ec_id=None):
2395 """Notify function to be called after updates.
2397 This function must be called when an object (as returned by
2398 GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2399 caller wants the modifications saved to the backing store. Note
2400 that all modified objects will be saved, but the target argument
2401 is the one the caller wants to ensure that it's saved.
2403 @param target: an instance of either L{objects.Cluster},
2404 L{objects.Node} or L{objects.Instance} which is existing in
2406 @param feedback_fn: Callable feedback function
2409 if self._config_data is None:
2410 raise errors.ProgrammerError("Configuration file not read,"
2412 update_serial = False
2413 if isinstance(target, objects.Cluster):
2414 test = target == self._config_data.cluster
2415 elif isinstance(target, objects.Node):
2416 test = target in self._config_data.nodes.values()
2417 update_serial = True
2418 elif isinstance(target, objects.Instance):
2419 test = target in self._config_data.instances.values()
2420 elif isinstance(target, objects.NodeGroup):
2421 test = target in self._config_data.nodegroups.values()
2422 elif isinstance(target, objects.Network):
2423 test = target in self._config_data.networks.values()
2425 raise errors.ProgrammerError("Invalid object type (%s) passed to"
2426 " ConfigWriter.Update" % type(target))
2428 raise errors.ConfigurationError("Configuration updated since object"
2429 " has been read or unknown object")
2430 target.serial_no += 1
2431 target.mtime = now = time.time()
2434 # for node updates, we need to increase the cluster serial too
2435 self._config_data.cluster.serial_no += 1
2436 self._config_data.cluster.mtime = now
2438 if isinstance(target, objects.Instance):
2439 self._UnlockedReleaseDRBDMinors(target.name)
2441 if ec_id is not None:
2442 # Commit all ips reserved by OpInstanceSetParams and OpGroupSetParams
2443 self._UnlockedCommitTemporaryIps(ec_id)
2445 self._WriteConfig(feedback_fn=feedback_fn)
2447 @locking.ssynchronized(_config_lock)
2448 def DropECReservations(self, ec_id):
2449 """Drop per-execution-context reservations
2452 for rm in self._all_rms:
2453 rm.DropECReservations(ec_id)
2455 @locking.ssynchronized(_config_lock, shared=1)
2456 def GetAllNetworksInfo(self):
2457 """Get configuration info of all the networks.
2460 return dict(self._config_data.networks)
2462 def _UnlockedGetNetworkList(self):
2463 """Get the list of networks.
2465 This function is for internal use, when the config lock is already held.
2468 return self._config_data.networks.keys()
2470 @locking.ssynchronized(_config_lock, shared=1)
2471 def GetNetworkList(self):
2472 """Get the list of networks.
2474 @return: array of networks, ex. ["main", "vlan100", "200]
2477 return self._UnlockedGetNetworkList()
2479 @locking.ssynchronized(_config_lock, shared=1)
2480 def GetNetworkNames(self):
2481 """Get a list of network names
2485 for net in self._config_data.networks.values()]
2488 def _UnlockedGetNetwork(self, uuid):
2489 """Returns information about a network.
2491 This function is for internal use, when the config lock is already held.
2494 if uuid not in self._config_data.networks:
2497 return self._config_data.networks[uuid]
2499 @locking.ssynchronized(_config_lock, shared=1)
2500 def GetNetwork(self, uuid):
2501 """Returns information about a network.
2503 It takes the information from the configuration file.
2505 @param uuid: UUID of the network
2507 @rtype: L{objects.Network}
2508 @return: the network object
2511 return self._UnlockedGetNetwork(uuid)
2513 @locking.ssynchronized(_config_lock)
2514 def AddNetwork(self, net, ec_id, check_uuid=True):
2515 """Add a network to the configuration.
2517 @type net: L{objects.Network}
2518 @param net: the Network object to add
2520 @param ec_id: unique id for the job to use when creating a missing UUID
2523 self._UnlockedAddNetwork(net, ec_id, check_uuid)
2526 def _UnlockedAddNetwork(self, net, ec_id, check_uuid):
2527 """Add a network to the configuration.
2530 logging.info("Adding network %s to configuration", net.name)
2533 self._EnsureUUID(net, ec_id)
2536 self._config_data.networks[net.uuid] = net
2537 self._config_data.cluster.serial_no += 1
2539 def _UnlockedLookupNetwork(self, target):
2540 """Lookup a network's UUID.
2542 @type target: string
2543 @param target: network name or UUID
2545 @return: network UUID
2546 @raises errors.OpPrereqError: when the target network cannot be found
2551 if target in self._config_data.networks:
2553 for net in self._config_data.networks.values():
2554 if net.name == target:
2556 raise errors.OpPrereqError("Network '%s' not found" % target,
2559 @locking.ssynchronized(_config_lock, shared=1)
2560 def LookupNetwork(self, target):
2561 """Lookup a network's UUID.
2563 This function is just a wrapper over L{_UnlockedLookupNetwork}.
2565 @type target: string
2566 @param target: network name or UUID
2568 @return: network UUID
2571 return self._UnlockedLookupNetwork(target)
2573 @locking.ssynchronized(_config_lock)
2574 def RemoveNetwork(self, network_uuid):
2575 """Remove a network from the configuration.
2577 @type network_uuid: string
2578 @param network_uuid: the UUID of the network to remove
2581 logging.info("Removing network %s from configuration", network_uuid)
2583 if network_uuid not in self._config_data.networks:
2584 raise errors.ConfigurationError("Unknown network '%s'" % network_uuid)
2586 del self._config_data.networks[network_uuid]
2587 self._config_data.cluster.serial_no += 1
2590 def _UnlockedGetGroupNetParams(self, net_uuid, node):
2591 """Get the netparams (mode, link) of a network.
2593 Get a network's netparams for a given node.
2595 @type net_uuid: string
2596 @param net_uuid: network uuid
2598 @param node: node name
2599 @rtype: dict or None
2603 node_info = self._UnlockedGetNodeInfo(node)
2604 nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2605 netparams = nodegroup_info.networks.get(net_uuid, None)
2609 @locking.ssynchronized(_config_lock, shared=1)
2610 def GetGroupNetParams(self, net_uuid, node):
2611 """Locking wrapper of _UnlockedGetGroupNetParams()
2614 return self._UnlockedGetGroupNetParams(net_uuid, node)
2616 @locking.ssynchronized(_config_lock, shared=1)
2617 def CheckIPInNodeGroup(self, ip, node):
2618 """Check IP uniqueness in nodegroup.
2620 Check networks that are connected in the node's node group
2621 if ip is contained in any of them. Used when creating/adding
2622 a NIC to ensure uniqueness among nodegroups.
2625 @param ip: ip address
2627 @param node: node name
2628 @rtype: (string, dict) or (None, None)
2629 @return: (network name, netparams)
2634 node_info = self._UnlockedGetNodeInfo(node)
2635 nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2636 for net_uuid in nodegroup_info.networks.keys():
2637 net_info = self._UnlockedGetNetwork(net_uuid)
2638 pool = network.AddressPool(net_info)
2639 if pool.Contains(ip):
2640 return (net_info.name, nodegroup_info.networks[net_uuid])