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 """Utility function mainly, but not only used by instance LU's."""
27 from ganeti import constants
28 from ganeti import errors
29 from ganeti import locking
30 from ganeti import network
31 from ganeti import objects
32 from ganeti import pathutils
33 from ganeti import utils
34 from ganeti.cmdlib.common import _AnnotateDiskParams, \
35 _ComputeIPolicyInstanceViolation
38 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
39 minmem, maxmem, vcpus, nics, disk_template, disks,
40 bep, hvp, hypervisor_name, tags):
41 """Builds instance related env variables for hooks
43 This builds the hook environment from individual variables.
46 @param name: the name of the instance
47 @type primary_node: string
48 @param primary_node: the name of the instance's primary node
49 @type secondary_nodes: list
50 @param secondary_nodes: list of secondary nodes as strings
52 @param os_type: the name of the instance's OS
54 @param status: the desired status of the instance
56 @param minmem: the minimum memory size of the instance
58 @param maxmem: the maximum memory size of the instance
60 @param vcpus: the count of VCPUs the instance has
62 @param nics: list of tuples (name, uuid, ip, mac, mode, link, net, netinfo)
63 representing the NICs the instance has
64 @type disk_template: string
65 @param disk_template: the disk template of the instance
67 @param disks: list of tuples (name, uuid, size, mode)
69 @param bep: the backend parameters for the instance
71 @param hvp: the hypervisor parameters for the instance
72 @type hypervisor_name: string
73 @param hypervisor_name: the hypervisor for the instance
75 @param tags: list of instance tags as strings
77 @return: the hook environment for this instance
82 "INSTANCE_NAME": name,
83 "INSTANCE_PRIMARY": primary_node,
84 "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
85 "INSTANCE_OS_TYPE": os_type,
86 "INSTANCE_STATUS": status,
87 "INSTANCE_MINMEM": minmem,
88 "INSTANCE_MAXMEM": maxmem,
89 # TODO(2.9) remove deprecated "memory" value
90 "INSTANCE_MEMORY": maxmem,
91 "INSTANCE_VCPUS": vcpus,
92 "INSTANCE_DISK_TEMPLATE": disk_template,
93 "INSTANCE_HYPERVISOR": hypervisor_name,
97 for idx, (name, _, ip, mac, mode, link, net, netinfo) in enumerate(nics):
100 env["INSTANCE_NIC%d_NAME" % idx] = name
101 env["INSTANCE_NIC%d_IP" % idx] = ip
102 env["INSTANCE_NIC%d_MAC" % idx] = mac
103 env["INSTANCE_NIC%d_MODE" % idx] = mode
104 env["INSTANCE_NIC%d_LINK" % idx] = link
106 nobj = objects.Network.FromDict(netinfo)
107 env.update(nobj.HooksDict("INSTANCE_NIC%d_" % idx))
109 # FIXME: broken network reference: the instance NIC specifies a
110 # network, but the relevant network entry was not in the config. This
111 # should be made impossible.
112 env["INSTANCE_NIC%d_NETWORK_NAME" % idx] = net
113 if mode == constants.NIC_MODE_BRIDGED:
114 env["INSTANCE_NIC%d_BRIDGE" % idx] = link
118 env["INSTANCE_NIC_COUNT"] = nic_count
121 disk_count = len(disks)
122 for idx, (name, size, mode) in enumerate(disks):
123 env["INSTANCE_DISK%d_NAME" % idx] = name
124 env["INSTANCE_DISK%d_SIZE" % idx] = size
125 env["INSTANCE_DISK%d_MODE" % idx] = mode
129 env["INSTANCE_DISK_COUNT"] = disk_count
134 env["INSTANCE_TAGS"] = " ".join(tags)
136 for source, kind in [(bep, "BE"), (hvp, "HV")]:
137 for key, value in source.items():
138 env["INSTANCE_%s_%s" % (kind, key)] = value
143 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
144 """Builds instance related env variables for hooks from an object.
146 @type lu: L{LogicalUnit}
147 @param lu: the logical unit on whose behalf we execute
148 @type instance: L{objects.Instance}
149 @param instance: the instance for which we should build the
152 @param override: dictionary with key/values that will override
155 @return: the hook environment dictionary
158 cluster = lu.cfg.GetClusterInfo()
159 bep = cluster.FillBE(instance)
160 hvp = cluster.FillHV(instance)
162 "name": instance.name,
163 "primary_node": instance.primary_node,
164 "secondary_nodes": instance.secondary_nodes,
165 "os_type": instance.os,
166 "status": instance.admin_state,
167 "maxmem": bep[constants.BE_MAXMEM],
168 "minmem": bep[constants.BE_MINMEM],
169 "vcpus": bep[constants.BE_VCPUS],
170 "nics": _NICListToTuple(lu, instance.nics),
171 "disk_template": instance.disk_template,
172 "disks": [(disk.name, disk.size, disk.mode)
173 for disk in instance.disks],
176 "hypervisor_name": instance.hypervisor,
177 "tags": instance.tags,
180 args.update(override)
181 return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
184 def _GetClusterDomainSecret():
185 """Reads the cluster domain secret.
188 return utils.ReadOneLineFile(pathutils.CLUSTER_DOMAIN_SECRET_FILE,
192 def _CheckNodeNotDrained(lu, node):
193 """Ensure that a given node is not drained.
195 @param lu: the LU on behalf of which we make the check
196 @param node: the node to check
197 @raise errors.OpPrereqError: if the node is drained
200 if lu.cfg.GetNodeInfo(node).drained:
201 raise errors.OpPrereqError("Can't use drained node %s" % node,
205 def _CheckNodeVmCapable(lu, node):
206 """Ensure that a given node is vm capable.
208 @param lu: the LU on behalf of which we make the check
209 @param node: the node to check
210 @raise errors.OpPrereqError: if the node is not vm capable
213 if not lu.cfg.GetNodeInfo(node).vm_capable:
214 raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
218 def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
219 """Utility function to remove an instance.
222 logging.info("Removing block devices for instance %s", instance.name)
224 if not _RemoveDisks(lu, instance, ignore_failures=ignore_failures):
225 if not ignore_failures:
226 raise errors.OpExecError("Can't remove instance's disks")
227 feedback_fn("Warning: can't remove instance's disks")
229 logging.info("Removing instance %s out of cluster config", instance.name)
231 lu.cfg.RemoveInstance(instance.name)
233 assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
234 "Instance lock removal conflict"
236 # Remove lock for the instance
237 lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
240 def _RemoveDisks(lu, instance, target_node=None, ignore_failures=False):
241 """Remove all disks for an instance.
243 This abstracts away some work from `AddInstance()` and
244 `RemoveInstance()`. Note that in case some of the devices couldn't
245 be removed, the removal will continue with the other ones.
247 @type lu: L{LogicalUnit}
248 @param lu: the logical unit on whose behalf we execute
249 @type instance: L{objects.Instance}
250 @param instance: the instance whose disks we should remove
251 @type target_node: string
252 @param target_node: used to override the node on which to remove the disks
254 @return: the success of the removal
257 logging.info("Removing block devices for instance %s", instance.name)
260 ports_to_release = set()
261 anno_disks = _AnnotateDiskParams(instance, instance.disks, lu.cfg)
262 for (idx, device) in enumerate(anno_disks):
264 edata = [(target_node, device)]
266 edata = device.ComputeNodeTree(instance.primary_node)
267 for node, disk in edata:
268 lu.cfg.SetDiskID(disk, node)
269 result = lu.rpc.call_blockdev_remove(node, disk)
271 lu.LogWarning("Could not remove disk %s on node %s,"
272 " continuing anyway: %s", idx, node, result.fail_msg)
273 if not (result.offline and node != instance.primary_node):
276 # if this is a DRBD disk, return its port to the pool
277 if device.dev_type in constants.LDS_DRBD:
278 ports_to_release.add(device.logical_id[2])
280 if all_result or ignore_failures:
281 for port in ports_to_release:
282 lu.cfg.AddTcpUdpPort(port)
284 if instance.disk_template in constants.DTS_FILEBASED:
285 file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
289 tgt = instance.primary_node
290 result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
292 lu.LogWarning("Could not remove directory '%s' on node %s: %s",
293 file_storage_dir, instance.primary_node, result.fail_msg)
299 def _NICToTuple(lu, nic):
300 """Build a tupple of nic information.
302 @type lu: L{LogicalUnit}
303 @param lu: the logical unit on whose behalf we execute
304 @type nic: L{objects.NIC}
305 @param nic: nic to convert to hooks tuple
308 cluster = lu.cfg.GetClusterInfo()
309 filled_params = cluster.SimpleFillNIC(nic.nicparams)
310 mode = filled_params[constants.NIC_MODE]
311 link = filled_params[constants.NIC_LINK]
314 nobj = lu.cfg.GetNetwork(nic.network)
315 netinfo = objects.Network.ToDict(nobj)
316 return (nic.name, nic.uuid, nic.ip, nic.mac, mode, link, nic.network, netinfo)
319 def _NICListToTuple(lu, nics):
320 """Build a list of nic information tuples.
322 This list is suitable to be passed to _BuildInstanceHookEnv or as a return
323 value in LUInstanceQueryData.
325 @type lu: L{LogicalUnit}
326 @param lu: the logical unit on whose behalf we execute
327 @type nics: list of L{objects.NIC}
328 @param nics: list of nics to convert to hooks tuples
333 hooks_nics.append(_NICToTuple(lu, nic))
337 def _CopyLockList(names):
338 """Makes a copy of a list of lock names.
340 Handles L{locking.ALL_SET} correctly.
343 if names == locking.ALL_SET:
344 return locking.ALL_SET
349 def _ReleaseLocks(lu, level, names=None, keep=None):
350 """Releases locks owned by an LU.
352 @type lu: L{LogicalUnit}
353 @param level: Lock level
354 @type names: list or None
355 @param names: Names of locks to release
356 @type keep: list or None
357 @param keep: Names of locks to retain
360 assert not (keep is not None and names is not None), \
361 "Only one of the 'names' and the 'keep' parameters can be given"
363 if names is not None:
364 should_release = names.__contains__
366 should_release = lambda name: name not in keep
368 should_release = None
370 owned = lu.owned_locks(level)
372 # Not owning any lock at this level, do nothing
379 # Determine which locks to release
381 if should_release(name):
386 assert len(lu.owned_locks(level)) == (len(retain) + len(release))
388 # Release just some locks
389 lu.glm.release(level, names=release)
391 assert frozenset(lu.owned_locks(level)) == frozenset(retain)
394 lu.glm.release(level)
396 assert not lu.glm.is_owned(level), "No locks should be owned"
399 def _ComputeIPolicyNodeViolation(ipolicy, instance, current_group,
401 _compute_fn=_ComputeIPolicyInstanceViolation):
402 """Compute if instance meets the specs of the new target group.
404 @param ipolicy: The ipolicy to verify
405 @param instance: The instance object to verify
406 @param current_group: The current group of the instance
407 @param target_group: The new group of the instance
408 @type cfg: L{config.ConfigWriter}
409 @param cfg: Cluster configuration
410 @param _compute_fn: The function to verify ipolicy (unittest only)
411 @see: L{ganeti.cmdlib.common._ComputeIPolicySpecViolation}
414 if current_group == target_group:
417 return _compute_fn(ipolicy, instance, cfg)
420 def _CheckTargetNodeIPolicy(lu, ipolicy, instance, node, cfg, ignore=False,
421 _compute_fn=_ComputeIPolicyNodeViolation):
422 """Checks that the target node is correct in terms of instance policy.
424 @param ipolicy: The ipolicy to verify
425 @param instance: The instance object to verify
426 @param node: The new node to relocate
427 @type cfg: L{config.ConfigWriter}
428 @param cfg: Cluster configuration
429 @param ignore: Ignore violations of the ipolicy
430 @param _compute_fn: The function to verify ipolicy (unittest only)
431 @see: L{ganeti.cmdlib.common._ComputeIPolicySpecViolation}
434 primary_node = lu.cfg.GetNodeInfo(instance.primary_node)
435 res = _compute_fn(ipolicy, instance, primary_node.group, node.group, cfg)
438 msg = ("Instance does not meet target node group's (%s) instance"
439 " policy: %s") % (node.group, utils.CommaJoin(res))
443 raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
446 def _GetInstanceInfoText(instance):
447 """Compute that text that should be added to the disk's metadata.
450 return "originstname+%s" % instance.name
453 def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
454 """Checks if a node has enough free memory.
456 This function checks if a given node has the needed amount of free
457 memory. In case the node has less memory or we cannot get the
458 information from the node, this function raises an OpPrereqError
461 @type lu: C{LogicalUnit}
462 @param lu: a logical unit from which we get configuration data
464 @param node: the node to check
466 @param reason: string to use in the error message
467 @type requested: C{int}
468 @param requested: the amount of memory in MiB to check for
469 @type hypervisor_name: C{str}
470 @param hypervisor_name: the hypervisor to ask for memory stats
472 @return: node current free memory
473 @raise errors.OpPrereqError: if the node doesn't have enough memory, or
474 we cannot check the node
477 nodeinfo = lu.rpc.call_node_info([node], None, [hypervisor_name], False)
478 nodeinfo[node].Raise("Can't get data from node %s" % node,
479 prereq=True, ecode=errors.ECODE_ENVIRON)
480 (_, _, (hv_info, )) = nodeinfo[node].payload
482 free_mem = hv_info.get("memory_free", None)
483 if not isinstance(free_mem, int):
484 raise errors.OpPrereqError("Can't compute free memory on node %s, result"
485 " was '%s'" % (node, free_mem),
486 errors.ECODE_ENVIRON)
487 if requested > free_mem:
488 raise errors.OpPrereqError("Not enough memory on node %s for %s:"
489 " needed %s MiB, available %s MiB" %
490 (node, reason, requested, free_mem),
495 def _CheckInstanceBridgesExist(lu, instance, node=None):
496 """Check that the brigdes needed by an instance exist.
500 node = instance.primary_node
501 _CheckNicsBridgesExist(lu, instance.nics, node)
504 def _CheckNicsBridgesExist(lu, target_nics, target_node):
505 """Check that the brigdes needed by a list of nics exist.
508 cluster = lu.cfg.GetClusterInfo()
509 paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
510 brlist = [params[constants.NIC_LINK] for params in paramslist
511 if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
513 result = lu.rpc.call_bridges_exist(target_node, brlist)
514 result.Raise("Error checking bridges on destination node '%s'" %
515 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
518 def _CheckNodeHasOS(lu, node, os_name, force_variant):
519 """Ensure that a node supports a given OS.
521 @param lu: the LU on behalf of which we make the check
522 @param node: the node to check
523 @param os_name: the OS to query about
524 @param force_variant: whether to ignore variant errors
525 @raise errors.OpPrereqError: if the node is not supporting the OS
528 result = lu.rpc.call_os_get(node, os_name)
529 result.Raise("OS '%s' not in supported OS list for node %s" %
531 prereq=True, ecode=errors.ECODE_INVAL)
532 if not force_variant:
533 _CheckOSVariant(result.payload, os_name)
536 def _CheckOSVariant(os_obj, name):
537 """Check whether an OS name conforms to the os variants specification.
539 @type os_obj: L{objects.OS}
540 @param os_obj: OS object to check
542 @param name: OS name passed by the user, to check for validity
545 variant = objects.OS.GetVariant(name)
546 if not os_obj.supported_variants:
548 raise errors.OpPrereqError("OS '%s' doesn't support variants ('%s'"
549 " passed)" % (os_obj.name, variant),
553 raise errors.OpPrereqError("OS name must include a variant",
556 if variant not in os_obj.supported_variants:
557 raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)