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 _helper_ispecs(owner, "ipolicy/" + key, value)
637 elif key == constants.ISPECS_STD:
638 _helper(owner, "ipolicy/" + key, value,
639 constants.ISPECS_PARAMETER_TYPES)
641 # FIXME: assuming list type
642 if key in constants.IPOLICY_PARAMETERS:
646 if not isinstance(value, exp_type):
647 result.append("%s has invalid instance policy: for %s,"
648 " expecting %s, got %s" %
649 (owner, key, exp_type.__name__, type(value)))
651 def _helper_ispecs(owner, parentkey, params):
652 for (key, value) in params.items():
653 fullkey = "/".join([parentkey, key])
654 _helper(owner, fullkey, value, constants.ISPECS_PARAMETER_TYPES)
656 # check cluster parameters
657 _helper("cluster", "beparams", cluster.SimpleFillBE({}),
658 constants.BES_PARAMETER_TYPES)
659 _helper("cluster", "nicparams", cluster.SimpleFillNIC({}),
660 constants.NICS_PARAMETER_TYPES)
661 _helper_nic("cluster", cluster.SimpleFillNIC({}))
662 _helper("cluster", "ndparams", cluster.SimpleFillND({}),
663 constants.NDS_PARAMETER_TYPES)
664 _helper_ipolicy("cluster", cluster.ipolicy, True)
666 # per-instance checks
667 for instance_name in data.instances:
668 instance = data.instances[instance_name]
669 if instance.name != instance_name:
670 result.append("instance '%s' is indexed by wrong name '%s'" %
671 (instance.name, instance_name))
672 if instance.primary_node not in data.nodes:
673 result.append("instance '%s' has invalid primary node '%s'" %
674 (instance_name, instance.primary_node))
675 for snode in instance.secondary_nodes:
676 if snode not in data.nodes:
677 result.append("instance '%s' has invalid secondary node '%s'" %
678 (instance_name, snode))
679 for idx, nic in enumerate(instance.nics):
680 if nic.mac in seen_macs:
681 result.append("instance '%s' has NIC %d mac %s duplicate" %
682 (instance_name, idx, nic.mac))
684 seen_macs.append(nic.mac)
686 filled = cluster.SimpleFillNIC(nic.nicparams)
687 owner = "instance %s nic %d" % (instance.name, idx)
688 _helper(owner, "nicparams",
689 filled, constants.NICS_PARAMETER_TYPES)
690 _helper_nic(owner, filled)
692 # disk template checks
693 if not instance.disk_template in data.cluster.enabled_disk_templates:
694 result.append("instance '%s' uses the disabled disk template '%s'." %
695 (instance_name, instance.disk_template))
698 if instance.beparams:
699 _helper("instance %s" % instance.name, "beparams",
700 cluster.FillBE(instance), constants.BES_PARAMETER_TYPES)
702 # gather the drbd ports for duplicate checks
703 for (idx, dsk) in enumerate(instance.disks):
704 if dsk.dev_type in constants.LDS_DRBD:
705 tcp_port = dsk.logical_id[2]
706 if tcp_port not in ports:
708 ports[tcp_port].append((instance.name, "drbd disk %s" % idx))
709 # gather network port reservation
710 net_port = getattr(instance, "network_port", None)
711 if net_port is not None:
712 if net_port not in ports:
714 ports[net_port].append((instance.name, "network port"))
716 # instance disk verify
717 for idx, disk in enumerate(instance.disks):
718 result.extend(["instance '%s' disk %d error: %s" %
719 (instance.name, idx, msg) for msg in disk.Verify()])
720 result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
722 wrong_names = _CheckInstanceDiskIvNames(instance.disks)
724 tmp = "; ".join(("name of disk %s should be '%s', but is '%s'" %
725 (idx, exp_name, actual_name))
726 for (idx, exp_name, actual_name) in wrong_names)
728 result.append("Instance '%s' has wrongly named disks: %s" %
729 (instance.name, tmp))
731 # cluster-wide pool of free ports
732 for free_port in cluster.tcpudp_port_pool:
733 if free_port not in ports:
734 ports[free_port] = []
735 ports[free_port].append(("cluster", "port marked as free"))
737 # compute tcp/udp duplicate ports
743 txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
744 result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
746 # highest used tcp port check
748 if keys[-1] > cluster.highest_used_port:
749 result.append("Highest used port mismatch, saved %s, computed %s" %
750 (cluster.highest_used_port, keys[-1]))
752 if not data.nodes[cluster.master_node].master_candidate:
753 result.append("Master node is not a master candidate")
755 # master candidate checks
756 mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
758 result.append("Not enough master candidates: actual %d, target %d" %
762 for node_name, node in data.nodes.items():
763 if node.name != node_name:
764 result.append("Node '%s' is indexed by wrong name '%s'" %
765 (node.name, node_name))
766 if [node.master_candidate, node.drained, node.offline].count(True) > 1:
767 result.append("Node %s state is invalid: master_candidate=%s,"
768 " drain=%s, offline=%s" %
769 (node.name, node.master_candidate, node.drained,
771 if node.group not in data.nodegroups:
772 result.append("Node '%s' has invalid group '%s'" %
773 (node.name, node.group))
775 _helper("node %s" % node.name, "ndparams",
776 cluster.FillND(node, data.nodegroups[node.group]),
777 constants.NDS_PARAMETER_TYPES)
778 used_globals = constants.NDC_GLOBALS.intersection(node.ndparams)
780 result.append("Node '%s' has some global parameters set: %s" %
781 (node.name, utils.CommaJoin(used_globals)))
784 nodegroups_names = set()
785 for nodegroup_uuid in data.nodegroups:
786 nodegroup = data.nodegroups[nodegroup_uuid]
787 if nodegroup.uuid != nodegroup_uuid:
788 result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'"
789 % (nodegroup.name, nodegroup.uuid, nodegroup_uuid))
790 if utils.UUID_RE.match(nodegroup.name.lower()):
791 result.append("node group '%s' (uuid: '%s') has uuid-like name" %
792 (nodegroup.name, nodegroup.uuid))
793 if nodegroup.name in nodegroups_names:
794 result.append("duplicate node group name '%s'" % nodegroup.name)
796 nodegroups_names.add(nodegroup.name)
797 group_name = "group %s" % nodegroup.name
798 _helper_ipolicy(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy),
800 if nodegroup.ndparams:
801 _helper(group_name, "ndparams",
802 cluster.SimpleFillND(nodegroup.ndparams),
803 constants.NDS_PARAMETER_TYPES)
806 _, duplicates = self._UnlockedComputeDRBDMap()
807 for node, minor, instance_a, instance_b in duplicates:
808 result.append("DRBD minor %d on node %s is assigned twice to instances"
809 " %s and %s" % (minor, node, instance_a, instance_b))
812 default_nicparams = cluster.nicparams[constants.PP_DEFAULT]
815 def _AddIpAddress(ip, name):
816 ips.setdefault(ip, []).append(name)
818 _AddIpAddress(cluster.master_ip, "cluster_ip")
820 for node in data.nodes.values():
821 _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
822 if node.secondary_ip != node.primary_ip:
823 _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
825 for instance in data.instances.values():
826 for idx, nic in enumerate(instance.nics):
830 nicparams = objects.FillDict(default_nicparams, nic.nicparams)
831 nic_mode = nicparams[constants.NIC_MODE]
832 nic_link = nicparams[constants.NIC_LINK]
834 if nic_mode == constants.NIC_MODE_BRIDGED:
835 link = "bridge:%s" % nic_link
836 elif nic_mode == constants.NIC_MODE_ROUTED:
837 link = "route:%s" % nic_link
839 raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
841 _AddIpAddress("%s/%s/%s" % (link, nic.ip, nic.network),
842 "instance:%s/nic:%d" % (instance.name, idx))
844 for ip, owners in ips.items():
846 result.append("IP address %s is used by multiple owners: %s" %
847 (ip, utils.CommaJoin(owners)))
851 @locking.ssynchronized(_config_lock, shared=1)
852 def VerifyConfig(self):
855 This is just a wrapper over L{_UnlockedVerifyConfig}.
858 @return: a list of error messages; a non-empty list signifies
862 return self._UnlockedVerifyConfig()
864 def _UnlockedSetDiskID(self, disk, node_name):
865 """Convert the unique ID to the ID needed on the target nodes.
867 This is used only for drbd, which needs ip/port configuration.
869 The routine descends down and updates its children also, because
870 this helps when the only the top device is passed to the remote
873 This function is for internal use, when the config lock is already held.
877 for child in disk.children:
878 self._UnlockedSetDiskID(child, node_name)
880 if disk.logical_id is None and disk.physical_id is not None:
882 if disk.dev_type == constants.LD_DRBD8:
883 pnode, snode, port, pminor, sminor, secret = disk.logical_id
884 if node_name not in (pnode, snode):
885 raise errors.ConfigurationError("DRBD device not knowing node %s" %
887 pnode_info = self._UnlockedGetNodeInfo(pnode)
888 snode_info = self._UnlockedGetNodeInfo(snode)
889 if pnode_info is None or snode_info is None:
890 raise errors.ConfigurationError("Can't find primary or secondary node"
891 " for %s" % str(disk))
892 p_data = (pnode_info.secondary_ip, port)
893 s_data = (snode_info.secondary_ip, port)
894 if pnode == node_name:
895 disk.physical_id = p_data + s_data + (pminor, secret)
896 else: # it must be secondary, we tested above
897 disk.physical_id = s_data + p_data + (sminor, secret)
899 disk.physical_id = disk.logical_id
902 @locking.ssynchronized(_config_lock)
903 def SetDiskID(self, disk, node_name):
904 """Convert the unique ID to the ID needed on the target nodes.
906 This is used only for drbd, which needs ip/port configuration.
908 The routine descends down and updates its children also, because
909 this helps when the only the top device is passed to the remote
913 return self._UnlockedSetDiskID(disk, node_name)
915 @locking.ssynchronized(_config_lock)
916 def AddTcpUdpPort(self, port):
917 """Adds a new port to the available port pool.
919 @warning: this method does not "flush" the configuration (via
920 L{_WriteConfig}); callers should do that themselves once the
921 configuration is stable
924 if not isinstance(port, int):
925 raise errors.ProgrammerError("Invalid type passed for port")
927 self._config_data.cluster.tcpudp_port_pool.add(port)
929 @locking.ssynchronized(_config_lock, shared=1)
930 def GetPortList(self):
931 """Returns a copy of the current port list.
934 return self._config_data.cluster.tcpudp_port_pool.copy()
936 @locking.ssynchronized(_config_lock)
937 def AllocatePort(self):
940 The port will be taken from the available port pool or from the
941 default port range (and in this case we increase
945 # If there are TCP/IP ports configured, we use them first.
946 if self._config_data.cluster.tcpudp_port_pool:
947 port = self._config_data.cluster.tcpudp_port_pool.pop()
949 port = self._config_data.cluster.highest_used_port + 1
950 if port >= constants.LAST_DRBD_PORT:
951 raise errors.ConfigurationError("The highest used port is greater"
952 " than %s. Aborting." %
953 constants.LAST_DRBD_PORT)
954 self._config_data.cluster.highest_used_port = port
959 def _UnlockedComputeDRBDMap(self):
960 """Compute the used DRBD minor/nodes.
963 @return: dictionary of node_name: dict of minor: instance_name;
964 the returned dict will have all the nodes in it (even if with
965 an empty list), and a list of duplicates; if the duplicates
966 list is not empty, the configuration is corrupted and its caller
967 should raise an exception
970 def _AppendUsedPorts(instance_name, disk, used):
972 if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:
973 node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
974 for node, port in ((node_a, minor_a), (node_b, minor_b)):
975 assert node in used, ("Node '%s' of instance '%s' not found"
976 " in node list" % (node, instance_name))
977 if port in used[node]:
978 duplicates.append((node, port, instance_name, used[node][port]))
980 used[node][port] = instance_name
982 for child in disk.children:
983 duplicates.extend(_AppendUsedPorts(instance_name, child, used))
987 my_dict = dict((node, {}) for node in self._config_data.nodes)
988 for instance in self._config_data.instances.itervalues():
989 for disk in instance.disks:
990 duplicates.extend(_AppendUsedPorts(instance.name, disk, my_dict))
991 for (node, minor), instance in self._temporary_drbds.iteritems():
992 if minor in my_dict[node] and my_dict[node][minor] != instance:
993 duplicates.append((node, minor, instance, my_dict[node][minor]))
995 my_dict[node][minor] = instance
996 return my_dict, duplicates
998 @locking.ssynchronized(_config_lock)
999 def ComputeDRBDMap(self):
1000 """Compute the used DRBD minor/nodes.
1002 This is just a wrapper over L{_UnlockedComputeDRBDMap}.
1004 @return: dictionary of node_name: dict of minor: instance_name;
1005 the returned dict will have all the nodes in it (even if with
1009 d_map, duplicates = self._UnlockedComputeDRBDMap()
1011 raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1015 @locking.ssynchronized(_config_lock)
1016 def AllocateDRBDMinor(self, nodes, instance):
1017 """Allocate a drbd minor.
1019 The free minor will be automatically computed from the existing
1020 devices. A node can be given multiple times in order to allocate
1021 multiple minors. The result is the list of minors, in the same
1022 order as the passed nodes.
1024 @type instance: string
1025 @param instance: the instance for which we allocate minors
1028 assert isinstance(instance, basestring), \
1029 "Invalid argument '%s' passed to AllocateDRBDMinor" % instance
1031 d_map, duplicates = self._UnlockedComputeDRBDMap()
1033 raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1037 ndata = d_map[nname]
1039 # no minors used, we can start at 0
1042 self._temporary_drbds[(nname, 0)] = instance
1046 ffree = utils.FirstFree(keys)
1048 # return the next minor
1049 # TODO: implement high-limit check
1050 minor = keys[-1] + 1
1053 # double-check minor against current instances
1054 assert minor not in d_map[nname], \
1055 ("Attempt to reuse allocated DRBD minor %d on node %s,"
1056 " already allocated to instance %s" %
1057 (minor, nname, d_map[nname][minor]))
1058 ndata[minor] = instance
1059 # double-check minor against reservation
1060 r_key = (nname, minor)
1061 assert r_key not in self._temporary_drbds, \
1062 ("Attempt to reuse reserved DRBD minor %d on node %s,"
1063 " reserved for instance %s" %
1064 (minor, nname, self._temporary_drbds[r_key]))
1065 self._temporary_drbds[r_key] = instance
1066 result.append(minor)
1067 logging.debug("Request to allocate drbd minors, input: %s, returning %s",
1071 def _UnlockedReleaseDRBDMinors(self, instance):
1072 """Release temporary drbd minors allocated for a given instance.
1074 @type instance: string
1075 @param instance: the instance for which temporary minors should be
1079 assert isinstance(instance, basestring), \
1080 "Invalid argument passed to ReleaseDRBDMinors"
1081 for key, name in self._temporary_drbds.items():
1082 if name == instance:
1083 del self._temporary_drbds[key]
1085 @locking.ssynchronized(_config_lock)
1086 def ReleaseDRBDMinors(self, instance):
1087 """Release temporary drbd minors allocated for a given instance.
1089 This should be called on the error paths, on the success paths
1090 it's automatically called by the ConfigWriter add and update
1093 This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
1095 @type instance: string
1096 @param instance: the instance for which temporary minors should be
1100 self._UnlockedReleaseDRBDMinors(instance)
1102 @locking.ssynchronized(_config_lock, shared=1)
1103 def GetConfigVersion(self):
1104 """Get the configuration version.
1106 @return: Config version
1109 return self._config_data.version
1111 @locking.ssynchronized(_config_lock, shared=1)
1112 def GetClusterName(self):
1113 """Get cluster name.
1115 @return: Cluster name
1118 return self._config_data.cluster.cluster_name
1120 @locking.ssynchronized(_config_lock, shared=1)
1121 def GetMasterNode(self):
1122 """Get the hostname of the master node for this cluster.
1124 @return: Master hostname
1127 return self._config_data.cluster.master_node
1129 @locking.ssynchronized(_config_lock, shared=1)
1130 def GetMasterIP(self):
1131 """Get the IP of the master node for this cluster.
1136 return self._config_data.cluster.master_ip
1138 @locking.ssynchronized(_config_lock, shared=1)
1139 def GetMasterNetdev(self):
1140 """Get the master network device for this cluster.
1143 return self._config_data.cluster.master_netdev
1145 @locking.ssynchronized(_config_lock, shared=1)
1146 def GetMasterNetmask(self):
1147 """Get the netmask of the master node for this cluster.
1150 return self._config_data.cluster.master_netmask
1152 @locking.ssynchronized(_config_lock, shared=1)
1153 def GetUseExternalMipScript(self):
1154 """Get flag representing whether to use the external master IP setup script.
1157 return self._config_data.cluster.use_external_mip_script
1159 @locking.ssynchronized(_config_lock, shared=1)
1160 def GetFileStorageDir(self):
1161 """Get the file storage dir for this cluster.
1164 return self._config_data.cluster.file_storage_dir
1166 @locking.ssynchronized(_config_lock, shared=1)
1167 def GetSharedFileStorageDir(self):
1168 """Get the shared file storage dir for this cluster.
1171 return self._config_data.cluster.shared_file_storage_dir
1173 @locking.ssynchronized(_config_lock, shared=1)
1174 def GetHypervisorType(self):
1175 """Get the hypervisor type for this cluster.
1178 return self._config_data.cluster.enabled_hypervisors[0]
1180 @locking.ssynchronized(_config_lock, shared=1)
1181 def GetHostKey(self):
1182 """Return the rsa hostkey from the config.
1185 @return: the rsa hostkey
1188 return self._config_data.cluster.rsahostkeypub
1190 @locking.ssynchronized(_config_lock, shared=1)
1191 def GetDefaultIAllocator(self):
1192 """Get the default instance allocator for this cluster.
1195 return self._config_data.cluster.default_iallocator
1197 @locking.ssynchronized(_config_lock, shared=1)
1198 def GetPrimaryIPFamily(self):
1199 """Get cluster primary ip family.
1201 @return: primary ip family
1204 return self._config_data.cluster.primary_ip_family
1206 @locking.ssynchronized(_config_lock, shared=1)
1207 def GetMasterNetworkParameters(self):
1208 """Get network parameters of the master node.
1210 @rtype: L{object.MasterNetworkParameters}
1211 @return: network parameters of the master node
1214 cluster = self._config_data.cluster
1215 result = objects.MasterNetworkParameters(
1216 name=cluster.master_node, ip=cluster.master_ip,
1217 netmask=cluster.master_netmask, netdev=cluster.master_netdev,
1218 ip_family=cluster.primary_ip_family)
1222 @locking.ssynchronized(_config_lock)
1223 def AddNodeGroup(self, group, ec_id, check_uuid=True):
1224 """Add a node group to the configuration.
1226 This method calls group.UpgradeConfig() to fill any missing attributes
1227 according to their default values.
1229 @type group: L{objects.NodeGroup}
1230 @param group: the NodeGroup object to add
1232 @param ec_id: unique id for the job to use when creating a missing UUID
1233 @type check_uuid: bool
1234 @param check_uuid: add an UUID to the group if it doesn't have one or, if
1235 it does, ensure that it does not exist in the
1236 configuration already
1239 self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1242 def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid):
1243 """Add a node group to the configuration.
1246 logging.info("Adding node group %s to configuration", group.name)
1248 # Some code might need to add a node group with a pre-populated UUID
1249 # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass
1250 # the "does this UUID" exist already check.
1252 self._EnsureUUID(group, ec_id)
1255 existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1256 except errors.OpPrereqError:
1259 raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1260 " node group (UUID: %s)" %
1261 (group.name, existing_uuid),
1262 errors.ECODE_EXISTS)
1265 group.ctime = group.mtime = time.time()
1266 group.UpgradeConfig()
1268 self._config_data.nodegroups[group.uuid] = group
1269 self._config_data.cluster.serial_no += 1
1271 @locking.ssynchronized(_config_lock)
1272 def RemoveNodeGroup(self, group_uuid):
1273 """Remove a node group from the configuration.
1275 @type group_uuid: string
1276 @param group_uuid: the UUID of the node group to remove
1279 logging.info("Removing node group %s from configuration", group_uuid)
1281 if group_uuid not in self._config_data.nodegroups:
1282 raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1284 assert len(self._config_data.nodegroups) != 1, \
1285 "Group '%s' is the only group, cannot be removed" % group_uuid
1287 del self._config_data.nodegroups[group_uuid]
1288 self._config_data.cluster.serial_no += 1
1291 def _UnlockedLookupNodeGroup(self, target):
1292 """Lookup a node group's UUID.
1294 @type target: string or None
1295 @param target: group name or UUID or None to look for the default
1297 @return: nodegroup UUID
1298 @raises errors.OpPrereqError: when the target group cannot be found
1302 if len(self._config_data.nodegroups) != 1:
1303 raise errors.OpPrereqError("More than one node group exists. Target"
1304 " group must be specified explicitly.")
1306 return self._config_data.nodegroups.keys()[0]
1307 if target in self._config_data.nodegroups:
1309 for nodegroup in self._config_data.nodegroups.values():
1310 if nodegroup.name == target:
1311 return nodegroup.uuid
1312 raise errors.OpPrereqError("Node group '%s' not found" % target,
1315 @locking.ssynchronized(_config_lock, shared=1)
1316 def LookupNodeGroup(self, target):
1317 """Lookup a node group's UUID.
1319 This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1321 @type target: string or None
1322 @param target: group name or UUID or None to look for the default
1324 @return: nodegroup UUID
1327 return self._UnlockedLookupNodeGroup(target)
1329 def _UnlockedGetNodeGroup(self, uuid):
1330 """Lookup a node group.
1333 @param uuid: group UUID
1334 @rtype: L{objects.NodeGroup} or None
1335 @return: nodegroup object, or None if not found
1338 if uuid not in self._config_data.nodegroups:
1341 return self._config_data.nodegroups[uuid]
1343 @locking.ssynchronized(_config_lock, shared=1)
1344 def GetNodeGroup(self, uuid):
1345 """Lookup a node group.
1348 @param uuid: group UUID
1349 @rtype: L{objects.NodeGroup} or None
1350 @return: nodegroup object, or None if not found
1353 return self._UnlockedGetNodeGroup(uuid)
1355 @locking.ssynchronized(_config_lock, shared=1)
1356 def GetAllNodeGroupsInfo(self):
1357 """Get the configuration of all node groups.
1360 return dict(self._config_data.nodegroups)
1362 @locking.ssynchronized(_config_lock, shared=1)
1363 def GetNodeGroupList(self):
1364 """Get a list of node groups.
1367 return self._config_data.nodegroups.keys()
1369 @locking.ssynchronized(_config_lock, shared=1)
1370 def GetNodeGroupMembersByNodes(self, nodes):
1371 """Get nodes which are member in the same nodegroups as the given nodes.
1374 ngfn = lambda node_name: self._UnlockedGetNodeInfo(node_name).group
1375 return frozenset(member_name
1376 for node_name in nodes
1378 self._UnlockedGetNodeGroup(ngfn(node_name)).members)
1380 @locking.ssynchronized(_config_lock, shared=1)
1381 def GetMultiNodeGroupInfo(self, group_uuids):
1382 """Get the configuration of multiple node groups.
1384 @param group_uuids: List of node group UUIDs
1386 @return: List of tuples of (group_uuid, group_info)
1389 return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1391 @locking.ssynchronized(_config_lock)
1392 def AddInstance(self, instance, ec_id):
1393 """Add an instance to the config.
1395 This should be used after creating a new instance.
1397 @type instance: L{objects.Instance}
1398 @param instance: the instance object
1401 if not isinstance(instance, objects.Instance):
1402 raise errors.ProgrammerError("Invalid type passed to AddInstance")
1404 if instance.disk_template != constants.DT_DISKLESS:
1405 all_lvs = instance.MapLVsByNode()
1406 logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
1408 all_macs = self._AllMACs()
1409 for nic in instance.nics:
1410 if nic.mac in all_macs:
1411 raise errors.ConfigurationError("Cannot add instance %s:"
1412 " MAC address '%s' already in use." %
1413 (instance.name, nic.mac))
1415 self._EnsureUUID(instance, ec_id)
1417 instance.serial_no = 1
1418 instance.ctime = instance.mtime = time.time()
1419 self._config_data.instances[instance.name] = instance
1420 self._config_data.cluster.serial_no += 1
1421 self._UnlockedReleaseDRBDMinors(instance.name)
1422 self._UnlockedCommitTemporaryIps(ec_id)
1425 def _EnsureUUID(self, item, ec_id):
1426 """Ensures a given object has a valid UUID.
1428 @param item: the instance or node to be checked
1429 @param ec_id: the execution context id for the uuid reservation
1433 item.uuid = self._GenerateUniqueID(ec_id)
1434 elif item.uuid in self._AllIDs(include_temporary=True):
1435 raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1436 " in use" % (item.name, item.uuid))
1438 def _SetInstanceStatus(self, instance_name, status):
1439 """Set the instance's status to a given value.
1442 assert status in constants.ADMINST_ALL, \
1443 "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1445 if instance_name not in self._config_data.instances:
1446 raise errors.ConfigurationError("Unknown instance '%s'" %
1448 instance = self._config_data.instances[instance_name]
1449 if instance.admin_state != status:
1450 instance.admin_state = status
1451 instance.serial_no += 1
1452 instance.mtime = time.time()
1455 @locking.ssynchronized(_config_lock)
1456 def MarkInstanceUp(self, instance_name):
1457 """Mark the instance status to up in the config.
1460 self._SetInstanceStatus(instance_name, constants.ADMINST_UP)
1462 @locking.ssynchronized(_config_lock)
1463 def MarkInstanceOffline(self, instance_name):
1464 """Mark the instance status to down in the config.
1467 self._SetInstanceStatus(instance_name, constants.ADMINST_OFFLINE)
1469 @locking.ssynchronized(_config_lock)
1470 def RemoveInstance(self, instance_name):
1471 """Remove the instance from the configuration.
1474 if instance_name not in self._config_data.instances:
1475 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1477 # If a network port has been allocated to the instance,
1478 # return it to the pool of free ports.
1479 inst = self._config_data.instances[instance_name]
1480 network_port = getattr(inst, "network_port", None)
1481 if network_port is not None:
1482 self._config_data.cluster.tcpudp_port_pool.add(network_port)
1484 instance = self._UnlockedGetInstanceInfo(instance_name)
1486 for nic in instance.nics:
1487 if nic.network and nic.ip:
1488 # Return all IP addresses to the respective address pools
1489 self._UnlockedCommitIp(constants.RELEASE_ACTION, nic.network, nic.ip)
1491 del self._config_data.instances[instance_name]
1492 self._config_data.cluster.serial_no += 1
1495 @locking.ssynchronized(_config_lock)
1496 def RenameInstance(self, old_name, new_name):
1497 """Rename an instance.
1499 This needs to be done in ConfigWriter and not by RemoveInstance
1500 combined with AddInstance as only we can guarantee an atomic
1504 if old_name not in self._config_data.instances:
1505 raise errors.ConfigurationError("Unknown instance '%s'" % old_name)
1507 # Operate on a copy to not loose instance object in case of a failure
1508 inst = self._config_data.instances[old_name].Copy()
1509 inst.name = new_name
1511 for (idx, disk) in enumerate(inst.disks):
1512 if disk.dev_type == constants.LD_FILE:
1513 # rename the file paths in logical and physical id
1514 file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1515 disk.logical_id = (disk.logical_id[0],
1516 utils.PathJoin(file_storage_dir, inst.name,
1518 disk.physical_id = disk.logical_id
1520 # Actually replace instance object
1521 del self._config_data.instances[old_name]
1522 self._config_data.instances[inst.name] = inst
1524 # Force update of ssconf files
1525 self._config_data.cluster.serial_no += 1
1529 @locking.ssynchronized(_config_lock)
1530 def MarkInstanceDown(self, instance_name):
1531 """Mark the status of an instance to down in the configuration.
1534 self._SetInstanceStatus(instance_name, constants.ADMINST_DOWN)
1536 def _UnlockedGetInstanceList(self):
1537 """Get the list of instances.
1539 This function is for internal use, when the config lock is already held.
1542 return self._config_data.instances.keys()
1544 @locking.ssynchronized(_config_lock, shared=1)
1545 def GetInstanceList(self):
1546 """Get the list of instances.
1548 @return: array of instances, ex. ['instance2.example.com',
1549 'instance1.example.com']
1552 return self._UnlockedGetInstanceList()
1554 def ExpandInstanceName(self, short_name):
1555 """Attempt to expand an incomplete instance name.
1558 # Locking is done in L{ConfigWriter.GetInstanceList}
1559 return _MatchNameComponentIgnoreCase(short_name, self.GetInstanceList())
1561 def _UnlockedGetInstanceInfo(self, instance_name):
1562 """Returns information about an instance.
1564 This function is for internal use, when the config lock is already held.
1567 if instance_name not in self._config_data.instances:
1570 return self._config_data.instances[instance_name]
1572 @locking.ssynchronized(_config_lock, shared=1)
1573 def GetInstanceInfo(self, instance_name):
1574 """Returns information about an instance.
1576 It takes the information from the configuration file. Other information of
1577 an instance are taken from the live systems.
1579 @param instance_name: name of the instance, e.g.
1580 I{instance1.example.com}
1582 @rtype: L{objects.Instance}
1583 @return: the instance object
1586 return self._UnlockedGetInstanceInfo(instance_name)
1588 @locking.ssynchronized(_config_lock, shared=1)
1589 def GetInstanceNodeGroups(self, instance_name, primary_only=False):
1590 """Returns set of node group UUIDs for instance's nodes.
1595 instance = self._UnlockedGetInstanceInfo(instance_name)
1597 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1600 nodes = [instance.primary_node]
1602 nodes = instance.all_nodes
1604 return frozenset(self._UnlockedGetNodeInfo(node_name).group
1605 for node_name in nodes)
1607 @locking.ssynchronized(_config_lock, shared=1)
1608 def GetInstanceNetworks(self, instance_name):
1609 """Returns set of network UUIDs for instance's nics.
1614 instance = self._UnlockedGetInstanceInfo(instance_name)
1616 raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1619 for nic in instance.nics:
1621 networks.add(nic.network)
1623 return frozenset(networks)
1625 @locking.ssynchronized(_config_lock, shared=1)
1626 def GetMultiInstanceInfo(self, instances):
1627 """Get the configuration of multiple instances.
1629 @param instances: list of instance names
1631 @return: list of tuples (instance, instance_info), where
1632 instance_info is what would GetInstanceInfo return for the
1633 node, while keeping the original order
1636 return [(name, self._UnlockedGetInstanceInfo(name)) for name in instances]
1638 @locking.ssynchronized(_config_lock, shared=1)
1639 def GetAllInstancesInfo(self):
1640 """Get the configuration of all instances.
1643 @return: dict of (instance, instance_info), where instance_info is what
1644 would GetInstanceInfo return for the node
1647 my_dict = dict([(instance, self._UnlockedGetInstanceInfo(instance))
1648 for instance in self._UnlockedGetInstanceList()])
1651 @locking.ssynchronized(_config_lock, shared=1)
1652 def GetInstancesInfoByFilter(self, filter_fn):
1653 """Get instance configuration with a filter.
1655 @type filter_fn: callable
1656 @param filter_fn: Filter function receiving instance object as parameter,
1657 returning boolean. Important: this function is called while the
1658 configuration locks is held. It must not do any complex work or call
1659 functions potentially leading to a deadlock. Ideally it doesn't call any
1660 other functions and just compares instance attributes.
1663 return dict((name, inst)
1664 for (name, inst) in self._config_data.instances.items()
1667 @locking.ssynchronized(_config_lock)
1668 def AddNode(self, node, ec_id):
1669 """Add a node to the configuration.
1671 @type node: L{objects.Node}
1672 @param node: a Node instance
1675 logging.info("Adding node %s to configuration", node.name)
1677 self._EnsureUUID(node, ec_id)
1680 node.ctime = node.mtime = time.time()
1681 self._UnlockedAddNodeToGroup(node.name, node.group)
1682 self._config_data.nodes[node.name] = node
1683 self._config_data.cluster.serial_no += 1
1686 @locking.ssynchronized(_config_lock)
1687 def RemoveNode(self, node_name):
1688 """Remove a node from the configuration.
1691 logging.info("Removing node %s from configuration", node_name)
1693 if node_name not in self._config_data.nodes:
1694 raise errors.ConfigurationError("Unknown node '%s'" % node_name)
1696 self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_name])
1697 del self._config_data.nodes[node_name]
1698 self._config_data.cluster.serial_no += 1
1701 def ExpandNodeName(self, short_name):
1702 """Attempt to expand an incomplete node name.
1705 # Locking is done in L{ConfigWriter.GetNodeList}
1706 return _MatchNameComponentIgnoreCase(short_name, self.GetNodeList())
1708 def _UnlockedGetNodeInfo(self, node_name):
1709 """Get the configuration of a node, as stored in the config.
1711 This function is for internal use, when the config lock is already
1714 @param node_name: the node name, e.g. I{node1.example.com}
1716 @rtype: L{objects.Node}
1717 @return: the node object
1720 if node_name not in self._config_data.nodes:
1723 return self._config_data.nodes[node_name]
1725 @locking.ssynchronized(_config_lock, shared=1)
1726 def GetNodeInfo(self, node_name):
1727 """Get the configuration of a node, as stored in the config.
1729 This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1731 @param node_name: the node name, e.g. I{node1.example.com}
1733 @rtype: L{objects.Node}
1734 @return: the node object
1737 return self._UnlockedGetNodeInfo(node_name)
1739 @locking.ssynchronized(_config_lock, shared=1)
1740 def GetNodeInstances(self, node_name):
1741 """Get the instances of a node, as stored in the config.
1743 @param node_name: the node name, e.g. I{node1.example.com}
1745 @rtype: (list, list)
1746 @return: a tuple with two lists: the primary and the secondary instances
1751 for inst in self._config_data.instances.values():
1752 if inst.primary_node == node_name:
1753 pri.append(inst.name)
1754 if node_name in inst.secondary_nodes:
1755 sec.append(inst.name)
1758 @locking.ssynchronized(_config_lock, shared=1)
1759 def GetNodeGroupInstances(self, uuid, primary_only=False):
1760 """Get the instances of a node group.
1762 @param uuid: Node group UUID
1763 @param primary_only: Whether to only consider primary nodes
1765 @return: List of instance names in node group
1769 nodes_fn = lambda inst: [inst.primary_node]
1771 nodes_fn = lambda inst: inst.all_nodes
1773 return frozenset(inst.name
1774 for inst in self._config_data.instances.values()
1775 for node_name in nodes_fn(inst)
1776 if self._UnlockedGetNodeInfo(node_name).group == uuid)
1778 def _UnlockedGetNodeList(self):
1779 """Return the list of nodes which are in the configuration.
1781 This function is for internal use, when the config lock is already
1787 return self._config_data.nodes.keys()
1789 @locking.ssynchronized(_config_lock, shared=1)
1790 def GetNodeList(self):
1791 """Return the list of nodes which are in the configuration.
1794 return self._UnlockedGetNodeList()
1796 def _UnlockedGetOnlineNodeList(self):
1797 """Return the list of nodes which are online.
1800 all_nodes = [self._UnlockedGetNodeInfo(node)
1801 for node in self._UnlockedGetNodeList()]
1802 return [node.name for node in all_nodes if not node.offline]
1804 @locking.ssynchronized(_config_lock, shared=1)
1805 def GetOnlineNodeList(self):
1806 """Return the list of nodes which are online.
1809 return self._UnlockedGetOnlineNodeList()
1811 @locking.ssynchronized(_config_lock, shared=1)
1812 def GetVmCapableNodeList(self):
1813 """Return the list of nodes which are not vm capable.
1816 all_nodes = [self._UnlockedGetNodeInfo(node)
1817 for node in self._UnlockedGetNodeList()]
1818 return [node.name for node in all_nodes if node.vm_capable]
1820 @locking.ssynchronized(_config_lock, shared=1)
1821 def GetNonVmCapableNodeList(self):
1822 """Return the list of nodes which are not vm capable.
1825 all_nodes = [self._UnlockedGetNodeInfo(node)
1826 for node in self._UnlockedGetNodeList()]
1827 return [node.name for node in all_nodes if not node.vm_capable]
1829 @locking.ssynchronized(_config_lock, shared=1)
1830 def GetMultiNodeInfo(self, nodes):
1831 """Get the configuration of multiple nodes.
1833 @param nodes: list of node names
1835 @return: list of tuples of (node, node_info), where node_info is
1836 what would GetNodeInfo return for the node, in the original
1840 return [(name, self._UnlockedGetNodeInfo(name)) for name in nodes]
1842 @locking.ssynchronized(_config_lock, shared=1)
1843 def GetAllNodesInfo(self):
1844 """Get the configuration of all nodes.
1847 @return: dict of (node, node_info), where node_info is what
1848 would GetNodeInfo return for the node
1851 return self._UnlockedGetAllNodesInfo()
1853 def _UnlockedGetAllNodesInfo(self):
1854 """Gets configuration of all nodes.
1856 @note: See L{GetAllNodesInfo}
1859 return dict([(node, self._UnlockedGetNodeInfo(node))
1860 for node in self._UnlockedGetNodeList()])
1862 @locking.ssynchronized(_config_lock, shared=1)
1863 def GetNodeGroupsFromNodes(self, nodes):
1864 """Returns groups for a list of nodes.
1866 @type nodes: list of string
1867 @param nodes: List of node names
1871 return frozenset(self._UnlockedGetNodeInfo(name).group for name in nodes)
1873 def _UnlockedGetMasterCandidateStats(self, exceptions=None):
1874 """Get the number of current and maximum desired and possible candidates.
1876 @type exceptions: list
1877 @param exceptions: if passed, list of nodes that should be ignored
1879 @return: tuple of (current, desired and possible, possible)
1882 mc_now = mc_should = mc_max = 0
1883 for node in self._config_data.nodes.values():
1884 if exceptions and node.name in exceptions:
1886 if not (node.offline or node.drained) and node.master_capable:
1888 if node.master_candidate:
1890 mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
1891 return (mc_now, mc_should, mc_max)
1893 @locking.ssynchronized(_config_lock, shared=1)
1894 def GetMasterCandidateStats(self, exceptions=None):
1895 """Get the number of current and maximum possible candidates.
1897 This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
1899 @type exceptions: list
1900 @param exceptions: if passed, list of nodes that should be ignored
1902 @return: tuple of (current, max)
1905 return self._UnlockedGetMasterCandidateStats(exceptions)
1907 @locking.ssynchronized(_config_lock)
1908 def MaintainCandidatePool(self, exceptions):
1909 """Try to grow the candidate pool to the desired size.
1911 @type exceptions: list
1912 @param exceptions: if passed, list of nodes that should be ignored
1914 @return: list with the adjusted nodes (L{objects.Node} instances)
1917 mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(exceptions)
1920 node_list = self._config_data.nodes.keys()
1921 random.shuffle(node_list)
1922 for name in node_list:
1923 if mc_now >= mc_max:
1925 node = self._config_data.nodes[name]
1926 if (node.master_candidate or node.offline or node.drained or
1927 node.name in exceptions or not node.master_capable):
1929 mod_list.append(node)
1930 node.master_candidate = True
1933 if mc_now != mc_max:
1934 # this should not happen
1935 logging.warning("Warning: MaintainCandidatePool didn't manage to"
1936 " fill the candidate pool (%d/%d)", mc_now, mc_max)
1938 self._config_data.cluster.serial_no += 1
1943 def _UnlockedAddNodeToGroup(self, node_name, nodegroup_uuid):
1944 """Add a given node to the specified group.
1947 if nodegroup_uuid not in self._config_data.nodegroups:
1948 # This can happen if a node group gets deleted between its lookup and
1949 # when we're adding the first node to it, since we don't keep a lock in
1950 # the meantime. It's ok though, as we'll fail cleanly if the node group
1951 # is not found anymore.
1952 raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
1953 if node_name not in self._config_data.nodegroups[nodegroup_uuid].members:
1954 self._config_data.nodegroups[nodegroup_uuid].members.append(node_name)
1956 def _UnlockedRemoveNodeFromGroup(self, node):
1957 """Remove a given node from its group.
1960 nodegroup = node.group
1961 if nodegroup not in self._config_data.nodegroups:
1962 logging.warning("Warning: node '%s' has unknown node group '%s'"
1963 " (while being removed from it)", node.name, nodegroup)
1964 nodegroup_obj = self._config_data.nodegroups[nodegroup]
1965 if node.name not in nodegroup_obj.members:
1966 logging.warning("Warning: node '%s' not a member of its node group '%s'"
1967 " (while being removed from it)", node.name, nodegroup)
1969 nodegroup_obj.members.remove(node.name)
1971 @locking.ssynchronized(_config_lock)
1972 def AssignGroupNodes(self, mods):
1973 """Changes the group of a number of nodes.
1975 @type mods: list of tuples; (node name, new group UUID)
1976 @param mods: Node membership modifications
1979 groups = self._config_data.nodegroups
1980 nodes = self._config_data.nodes
1984 # Try to resolve names/UUIDs first
1985 for (node_name, new_group_uuid) in mods:
1987 node = nodes[node_name]
1989 raise errors.ConfigurationError("Unable to find node '%s'" % node_name)
1991 if node.group == new_group_uuid:
1992 # Node is being assigned to its current group
1993 logging.debug("Node '%s' was assigned to its current group (%s)",
1994 node_name, node.group)
1997 # Try to find current group of node
1999 old_group = groups[node.group]
2001 raise errors.ConfigurationError("Unable to find old group '%s'" %
2004 # Try to find new group for node
2006 new_group = groups[new_group_uuid]
2008 raise errors.ConfigurationError("Unable to find new group '%s'" %
2011 assert node.name in old_group.members, \
2012 ("Inconsistent configuration: node '%s' not listed in members for its"
2013 " old group '%s'" % (node.name, old_group.uuid))
2014 assert node.name not in new_group.members, \
2015 ("Inconsistent configuration: node '%s' already listed in members for"
2016 " its new group '%s'" % (node.name, new_group.uuid))
2018 resmod.append((node, old_group, new_group))
2021 for (node, old_group, new_group) in resmod:
2022 assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
2023 "Assigning to current group is not possible"
2025 node.group = new_group.uuid
2027 # Update members of involved groups
2028 if node.name in old_group.members:
2029 old_group.members.remove(node.name)
2030 if node.name not in new_group.members:
2031 new_group.members.append(node.name)
2033 # Update timestamps and serials (only once per node/group object)
2035 for obj in frozenset(itertools.chain(*resmod)): # pylint: disable=W0142
2039 # Force ssconf update
2040 self._config_data.cluster.serial_no += 1
2044 def _BumpSerialNo(self):
2045 """Bump up the serial number of the config.
2048 self._config_data.serial_no += 1
2049 self._config_data.mtime = time.time()
2051 def _AllUUIDObjects(self):
2052 """Returns all objects with uuid attributes.
2055 return (self._config_data.instances.values() +
2056 self._config_data.nodes.values() +
2057 self._config_data.nodegroups.values() +
2058 self._config_data.networks.values() +
2061 [self._config_data.cluster])
2063 def _OpenConfig(self, accept_foreign):
2064 """Read the config data from disk.
2067 raw_data = utils.ReadFile(self._cfg_file)
2070 data = objects.ConfigData.FromDict(serializer.Load(raw_data))
2071 except Exception, err:
2072 raise errors.ConfigurationError(err)
2074 # Make sure the configuration has the right version
2075 _ValidateConfig(data)
2077 if (not hasattr(data, "cluster") or
2078 not hasattr(data.cluster, "rsahostkeypub")):
2079 raise errors.ConfigurationError("Incomplete configuration"
2080 " (missing cluster.rsahostkeypub)")
2082 if data.cluster.master_node != self._my_hostname and not accept_foreign:
2083 msg = ("The configuration denotes node %s as master, while my"
2084 " hostname is %s; opening a foreign configuration is only"
2085 " possible in accept_foreign mode" %
2086 (data.cluster.master_node, self._my_hostname))
2087 raise errors.ConfigurationError(msg)
2089 self._config_data = data
2090 # reset the last serial as -1 so that the next write will cause
2092 self._last_cluster_serial = -1
2094 # Upgrade configuration if needed
2095 self._UpgradeConfig()
2097 self._cfg_id = utils.GetFileID(path=self._cfg_file)
2099 def _UpgradeConfig(self):
2100 """Run any upgrade steps.
2102 This method performs both in-object upgrades and also update some data
2103 elements that need uniqueness across the whole configuration or interact
2106 @warning: this function will call L{_WriteConfig()}, but also
2107 L{DropECReservations} so it needs to be called only from a
2108 "safe" place (the constructor). If one wanted to call it with
2109 the lock held, a DropECReservationUnlocked would need to be
2110 created first, to avoid causing deadlock.
2113 # Keep a copy of the persistent part of _config_data to check for changes
2114 # Serialization doesn't guarantee order in dictionaries
2115 oldconf = copy.deepcopy(self._config_data.ToDict())
2117 # In-object upgrades
2118 self._config_data.UpgradeConfig()
2120 for item in self._AllUUIDObjects():
2121 if item.uuid is None:
2122 item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
2123 if not self._config_data.nodegroups:
2124 default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
2125 default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
2127 self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
2128 for node in self._config_data.nodes.values():
2130 node.group = self.LookupNodeGroup(None)
2131 # This is technically *not* an upgrade, but needs to be done both when
2132 # nodegroups are being added, and upon normally loading the config,
2133 # because the members list of a node group is discarded upon
2134 # serializing/deserializing the object.
2135 self._UnlockedAddNodeToGroup(node.name, node.group)
2137 modified = (oldconf != self._config_data.ToDict())
2140 # This is ok even if it acquires the internal lock, as _UpgradeConfig is
2141 # only called at config init time, without the lock held
2142 self.DropECReservations(_UPGRADE_CONFIG_JID)
2144 config_errors = self._UnlockedVerifyConfig()
2146 errmsg = ("Loaded configuration data is not consistent: %s" %
2147 (utils.CommaJoin(config_errors)))
2148 logging.critical(errmsg)
2150 def _DistributeConfig(self, feedback_fn):
2151 """Distribute the configuration to the other nodes.
2153 Currently, this only copies the configuration file. In the future,
2154 it could be used to encapsulate the 2/3-phase update mechanism.
2164 myhostname = self._my_hostname
2165 # we can skip checking whether _UnlockedGetNodeInfo returns None
2166 # since the node list comes from _UnlocketGetNodeList, and we are
2167 # called with the lock held, so no modifications should take place
2169 for node_name in self._UnlockedGetNodeList():
2170 if node_name == myhostname:
2172 node_info = self._UnlockedGetNodeInfo(node_name)
2173 if not node_info.master_candidate:
2175 node_list.append(node_info.name)
2176 addr_list.append(node_info.primary_ip)
2178 # TODO: Use dedicated resolver talking to config writer for name resolution
2180 self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
2181 for to_node, to_result in result.items():
2182 msg = to_result.fail_msg
2184 msg = ("Copy of file %s to node %s failed: %s" %
2185 (self._cfg_file, to_node, msg))
2195 def _WriteConfig(self, destination=None, feedback_fn=None):
2196 """Write the configuration data to persistent storage.
2199 assert feedback_fn is None or callable(feedback_fn)
2201 # Warn on config errors, but don't abort the save - the
2202 # configuration has already been modified, and we can't revert;
2203 # the best we can do is to warn the user and save as is, leaving
2204 # recovery to the user
2205 config_errors = self._UnlockedVerifyConfig()
2207 errmsg = ("Configuration data is not consistent: %s" %
2208 (utils.CommaJoin(config_errors)))
2209 logging.critical(errmsg)
2213 if destination is None:
2214 destination = self._cfg_file
2215 self._BumpSerialNo()
2216 txt = serializer.Dump(self._config_data.ToDict())
2218 getents = self._getents()
2220 fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2221 close=False, gid=getents.confd_gid, mode=0640)
2222 except errors.LockError:
2223 raise errors.ConfigurationError("The configuration file has been"
2224 " modified since the last write, cannot"
2227 self._cfg_id = utils.GetFileID(fd=fd)
2231 self.write_count += 1
2233 # and redistribute the config file to master candidates
2234 self._DistributeConfig(feedback_fn)
2236 # Write ssconf files on all nodes (including locally)
2237 if self._last_cluster_serial < self._config_data.cluster.serial_no:
2238 if not self._offline:
2239 result = self._GetRpc(None).call_write_ssconf_files(
2240 self._UnlockedGetOnlineNodeList(),
2241 self._UnlockedGetSsconfValues())
2243 for nname, nresu in result.items():
2244 msg = nresu.fail_msg
2246 errmsg = ("Error while uploading ssconf files to"
2247 " node %s: %s" % (nname, msg))
2248 logging.warning(errmsg)
2253 self._last_cluster_serial = self._config_data.cluster.serial_no
2255 def _UnlockedGetSsconfValues(self):
2256 """Return the values needed by ssconf.
2259 @return: a dictionary with keys the ssconf names and values their
2264 instance_names = utils.NiceSort(self._UnlockedGetInstanceList())
2265 node_names = utils.NiceSort(self._UnlockedGetNodeList())
2266 node_info = [self._UnlockedGetNodeInfo(name) for name in node_names]
2267 node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2268 for ninfo in node_info]
2269 node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2270 for ninfo in node_info]
2272 instance_data = fn(instance_names)
2273 off_data = fn(node.name for node in node_info if node.offline)
2274 on_data = fn(node.name for node in node_info if not node.offline)
2275 mc_data = fn(node.name for node in node_info if node.master_candidate)
2276 mc_ips_data = fn(node.primary_ip for node in node_info
2277 if node.master_candidate)
2278 node_data = fn(node_names)
2279 node_pri_ips_data = fn(node_pri_ips)
2280 node_snd_ips_data = fn(node_snd_ips)
2282 cluster = self._config_data.cluster
2283 cluster_tags = fn(cluster.GetTags())
2285 hypervisor_list = fn(cluster.enabled_hypervisors)
2287 uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2289 nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2290 self._config_data.nodegroups.values()]
2291 nodegroups_data = fn(utils.NiceSort(nodegroups))
2292 networks = ["%s %s" % (net.uuid, net.name) for net in
2293 self._config_data.networks.values()]
2294 networks_data = fn(utils.NiceSort(networks))
2297 constants.SS_CLUSTER_NAME: cluster.cluster_name,
2298 constants.SS_CLUSTER_TAGS: cluster_tags,
2299 constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2300 constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2301 constants.SS_MASTER_CANDIDATES: mc_data,
2302 constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2303 constants.SS_MASTER_IP: cluster.master_ip,
2304 constants.SS_MASTER_NETDEV: cluster.master_netdev,
2305 constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2306 constants.SS_MASTER_NODE: cluster.master_node,
2307 constants.SS_NODE_LIST: node_data,
2308 constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2309 constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2310 constants.SS_OFFLINE_NODES: off_data,
2311 constants.SS_ONLINE_NODES: on_data,
2312 constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2313 constants.SS_INSTANCE_LIST: instance_data,
2314 constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2315 constants.SS_HYPERVISOR_LIST: hypervisor_list,
2316 constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2317 constants.SS_UID_POOL: uid_pool,
2318 constants.SS_NODEGROUPS: nodegroups_data,
2319 constants.SS_NETWORKS: networks_data,
2321 bad_values = [(k, v) for k, v in ssconf_values.items()
2322 if not isinstance(v, (str, basestring))]
2324 err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2325 raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2326 " values: %s" % err)
2327 return ssconf_values
2329 @locking.ssynchronized(_config_lock, shared=1)
2330 def GetSsconfValues(self):
2331 """Wrapper using lock around _UnlockedGetSsconf().
2334 return self._UnlockedGetSsconfValues()
2336 @locking.ssynchronized(_config_lock, shared=1)
2337 def GetVGName(self):
2338 """Return the volume group name.
2341 return self._config_data.cluster.volume_group_name
2343 @locking.ssynchronized(_config_lock)
2344 def SetVGName(self, vg_name):
2345 """Set the volume group name.
2348 self._config_data.cluster.volume_group_name = vg_name
2349 self._config_data.cluster.serial_no += 1
2352 @locking.ssynchronized(_config_lock, shared=1)
2353 def GetDRBDHelper(self):
2354 """Return DRBD usermode helper.
2357 return self._config_data.cluster.drbd_usermode_helper
2359 @locking.ssynchronized(_config_lock)
2360 def SetDRBDHelper(self, drbd_helper):
2361 """Set DRBD usermode helper.
2364 self._config_data.cluster.drbd_usermode_helper = drbd_helper
2365 self._config_data.cluster.serial_no += 1
2368 @locking.ssynchronized(_config_lock, shared=1)
2369 def GetMACPrefix(self):
2370 """Return the mac prefix.
2373 return self._config_data.cluster.mac_prefix
2375 @locking.ssynchronized(_config_lock, shared=1)
2376 def GetClusterInfo(self):
2377 """Returns information about the cluster
2379 @rtype: L{objects.Cluster}
2380 @return: the cluster object
2383 return self._config_data.cluster
2385 @locking.ssynchronized(_config_lock, shared=1)
2386 def HasAnyDiskOfType(self, dev_type):
2387 """Check if in there is at disk of the given type in the configuration.
2390 return self._config_data.HasAnyDiskOfType(dev_type)
2392 @locking.ssynchronized(_config_lock)
2393 def Update(self, target, feedback_fn, ec_id=None):
2394 """Notify function to be called after updates.
2396 This function must be called when an object (as returned by
2397 GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2398 caller wants the modifications saved to the backing store. Note
2399 that all modified objects will be saved, but the target argument
2400 is the one the caller wants to ensure that it's saved.
2402 @param target: an instance of either L{objects.Cluster},
2403 L{objects.Node} or L{objects.Instance} which is existing in
2405 @param feedback_fn: Callable feedback function
2408 if self._config_data is None:
2409 raise errors.ProgrammerError("Configuration file not read,"
2411 update_serial = False
2412 if isinstance(target, objects.Cluster):
2413 test = target == self._config_data.cluster
2414 elif isinstance(target, objects.Node):
2415 test = target in self._config_data.nodes.values()
2416 update_serial = True
2417 elif isinstance(target, objects.Instance):
2418 test = target in self._config_data.instances.values()
2419 elif isinstance(target, objects.NodeGroup):
2420 test = target in self._config_data.nodegroups.values()
2421 elif isinstance(target, objects.Network):
2422 test = target in self._config_data.networks.values()
2424 raise errors.ProgrammerError("Invalid object type (%s) passed to"
2425 " ConfigWriter.Update" % type(target))
2427 raise errors.ConfigurationError("Configuration updated since object"
2428 " has been read or unknown object")
2429 target.serial_no += 1
2430 target.mtime = now = time.time()
2433 # for node updates, we need to increase the cluster serial too
2434 self._config_data.cluster.serial_no += 1
2435 self._config_data.cluster.mtime = now
2437 if isinstance(target, objects.Instance):
2438 self._UnlockedReleaseDRBDMinors(target.name)
2440 if ec_id is not None:
2441 # Commit all ips reserved by OpInstanceSetParams and OpGroupSetParams
2442 self._UnlockedCommitTemporaryIps(ec_id)
2444 self._WriteConfig(feedback_fn=feedback_fn)
2446 @locking.ssynchronized(_config_lock)
2447 def DropECReservations(self, ec_id):
2448 """Drop per-execution-context reservations
2451 for rm in self._all_rms:
2452 rm.DropECReservations(ec_id)
2454 @locking.ssynchronized(_config_lock, shared=1)
2455 def GetAllNetworksInfo(self):
2456 """Get configuration info of all the networks.
2459 return dict(self._config_data.networks)
2461 def _UnlockedGetNetworkList(self):
2462 """Get the list of networks.
2464 This function is for internal use, when the config lock is already held.
2467 return self._config_data.networks.keys()
2469 @locking.ssynchronized(_config_lock, shared=1)
2470 def GetNetworkList(self):
2471 """Get the list of networks.
2473 @return: array of networks, ex. ["main", "vlan100", "200]
2476 return self._UnlockedGetNetworkList()
2478 @locking.ssynchronized(_config_lock, shared=1)
2479 def GetNetworkNames(self):
2480 """Get a list of network names
2484 for net in self._config_data.networks.values()]
2487 def _UnlockedGetNetwork(self, uuid):
2488 """Returns information about a network.
2490 This function is for internal use, when the config lock is already held.
2493 if uuid not in self._config_data.networks:
2496 return self._config_data.networks[uuid]
2498 @locking.ssynchronized(_config_lock, shared=1)
2499 def GetNetwork(self, uuid):
2500 """Returns information about a network.
2502 It takes the information from the configuration file.
2504 @param uuid: UUID of the network
2506 @rtype: L{objects.Network}
2507 @return: the network object
2510 return self._UnlockedGetNetwork(uuid)
2512 @locking.ssynchronized(_config_lock)
2513 def AddNetwork(self, net, ec_id, check_uuid=True):
2514 """Add a network to the configuration.
2516 @type net: L{objects.Network}
2517 @param net: the Network object to add
2519 @param ec_id: unique id for the job to use when creating a missing UUID
2522 self._UnlockedAddNetwork(net, ec_id, check_uuid)
2525 def _UnlockedAddNetwork(self, net, ec_id, check_uuid):
2526 """Add a network to the configuration.
2529 logging.info("Adding network %s to configuration", net.name)
2532 self._EnsureUUID(net, ec_id)
2535 self._config_data.networks[net.uuid] = net
2536 self._config_data.cluster.serial_no += 1
2538 def _UnlockedLookupNetwork(self, target):
2539 """Lookup a network's UUID.
2541 @type target: string
2542 @param target: network name or UUID
2544 @return: network UUID
2545 @raises errors.OpPrereqError: when the target network cannot be found
2550 if target in self._config_data.networks:
2552 for net in self._config_data.networks.values():
2553 if net.name == target:
2555 raise errors.OpPrereqError("Network '%s' not found" % target,
2558 @locking.ssynchronized(_config_lock, shared=1)
2559 def LookupNetwork(self, target):
2560 """Lookup a network's UUID.
2562 This function is just a wrapper over L{_UnlockedLookupNetwork}.
2564 @type target: string
2565 @param target: network name or UUID
2567 @return: network UUID
2570 return self._UnlockedLookupNetwork(target)
2572 @locking.ssynchronized(_config_lock)
2573 def RemoveNetwork(self, network_uuid):
2574 """Remove a network from the configuration.
2576 @type network_uuid: string
2577 @param network_uuid: the UUID of the network to remove
2580 logging.info("Removing network %s from configuration", network_uuid)
2582 if network_uuid not in self._config_data.networks:
2583 raise errors.ConfigurationError("Unknown network '%s'" % network_uuid)
2585 del self._config_data.networks[network_uuid]
2586 self._config_data.cluster.serial_no += 1
2589 def _UnlockedGetGroupNetParams(self, net_uuid, node):
2590 """Get the netparams (mode, link) of a network.
2592 Get a network's netparams for a given node.
2594 @type net_uuid: string
2595 @param net_uuid: network uuid
2597 @param node: node name
2598 @rtype: dict or None
2602 node_info = self._UnlockedGetNodeInfo(node)
2603 nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2604 netparams = nodegroup_info.networks.get(net_uuid, None)
2608 @locking.ssynchronized(_config_lock, shared=1)
2609 def GetGroupNetParams(self, net_uuid, node):
2610 """Locking wrapper of _UnlockedGetGroupNetParams()
2613 return self._UnlockedGetGroupNetParams(net_uuid, node)
2615 @locking.ssynchronized(_config_lock, shared=1)
2616 def CheckIPInNodeGroup(self, ip, node):
2617 """Check IP uniqueness in nodegroup.
2619 Check networks that are connected in the node's node group
2620 if ip is contained in any of them. Used when creating/adding
2621 a NIC to ensure uniqueness among nodegroups.
2624 @param ip: ip address
2626 @param node: node name
2627 @rtype: (string, dict) or (None, None)
2628 @return: (network name, netparams)
2633 node_info = self._UnlockedGetNodeInfo(node)
2634 nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2635 for net_uuid in nodegroup_info.networks.keys():
2636 net_info = self._UnlockedGetNetwork(net_uuid)
2637 pool = network.AddressPool(net_info)
2638 if pool.Contains(ip):
2639 return (net_info.name, nodegroup_info.networks[net_uuid])