Synnefo » snf-ganeti

#

#

# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc.

#

# This program is free software; you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation; either version 2 of the License, or

# (at your option) any later version.

#

# This program is distributed in the hope that it will be useful, but

# WITHOUT ANY WARRANTY; without even the implied warranty of

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

# General Public License for more details.

#

# You should have received a copy of the GNU General Public License

# along with this program; if not, write to the Free Software

# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA

# 02110-1301, USA.

"""Logical units dealing with node groups."""

import itertools

import logging

from ganeti import constants

from ganeti import errors

from ganeti import locking

from ganeti import objects

from ganeti import utils

from ganeti.masterd import iallocator

from ganeti.cmdlib.base import LogicalUnit, NoHooksLU, ResultWithJobs

from ganeti.cmdlib.common import MergeAndVerifyHvState, \

  MergeAndVerifyDiskState, GetWantedNodes, GetUpdatedParams, \

  CheckNodeGroupInstances, GetUpdatedIPolicy, \

  ComputeNewInstanceViolations, GetDefaultIAllocator, ShareAll, \

  CheckInstancesNodeGroups, LoadNodeEvacResult, MapInstanceLvsToNodes, \

  CheckIpolicyVsDiskTemplates, CheckDiskAccessModeValidity, \

  CheckDiskAccessModeConsistency

import ganeti.masterd.instance

class LUGroupAdd(LogicalUnit):

  """Logical unit for creating node groups.

  """

  HPATH = "group-add"

  HTYPE = constants.HTYPE_GROUP

  REQ_BGL = False

  def ExpandNames(self):

    # We need the new group's UUID here so that we can create and acquire the

    # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup

    # that it should not check whether the UUID exists in the configuration.

    self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())

    self.needed_locks = {}

    self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid

  def _CheckIpolicy(self):

    """Checks the group's ipolicy for consistency and validity.

    """

    if self.op.ipolicy:

      cluster = self.cfg.GetClusterInfo()

      full_ipolicy = cluster.SimpleFillIPolicy(self.op.ipolicy)

      try:

        objects.InstancePolicy.CheckParameterSyntax(full_ipolicy, False)

      except errors.ConfigurationError, err:

        raise errors.OpPrereqError("Invalid instance policy: %s" % err,

                                   errors.ECODE_INVAL)

      CheckIpolicyVsDiskTemplates(full_ipolicy,

                                  cluster.enabled_disk_templates)

  def CheckPrereq(self):

    """Check prerequisites.

    This checks that the given group name is not an existing node group

    already.

    """

    try:

      existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    except errors.OpPrereqError:

      pass

    else:

      raise errors.OpPrereqError("Desired group name '%s' already exists as a"

                                 " node group (UUID: %s)" %

                                 (self.op.group_name, existing_uuid),

                                 errors.ECODE_EXISTS)

    if self.op.ndparams:

      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)

    if self.op.hv_state:

      self.new_hv_state = MergeAndVerifyHvState(self.op.hv_state, None)

    else:

      self.new_hv_state = None

    if self.op.disk_state:

      self.new_disk_state = MergeAndVerifyDiskState(self.op.disk_state, None)

    else:

      self.new_disk_state = None

    if self.op.diskparams:

      for templ in constants.DISK_TEMPLATES:

        if templ in self.op.diskparams:

          utils.ForceDictType(self.op.diskparams[templ],

                              constants.DISK_DT_TYPES)

      self.new_diskparams = self.op.diskparams

      try:

        utils.VerifyDictOptions(self.new_diskparams, constants.DISK_DT_DEFAULTS)

      except errors.OpPrereqError, err:

        raise errors.OpPrereqError("While verify diskparams options: %s" % err,

                                   errors.ECODE_INVAL)

    else:

      self.new_diskparams = {}

    self._CheckIpolicy()

  def BuildHooksEnv(self):

    """Build hooks env.

    """

    return {

      "GROUP_NAME": self.op.group_name,

      }

  def BuildHooksNodes(self):

    """Build hooks nodes.

    """

    mn = self.cfg.GetMasterNode()

    return ([mn], [mn])

  def Exec(self, feedback_fn):

    """Add the node group to the cluster.

    """

    group_obj = objects.NodeGroup(name=self.op.group_name, members=[],

                                  uuid=self.group_uuid,

                                  alloc_policy=self.op.alloc_policy,

                                  ndparams=self.op.ndparams,

                                  diskparams=self.new_diskparams,

                                  ipolicy=self.op.ipolicy,

                                  hv_state_static=self.new_hv_state,

                                  disk_state_static=self.new_disk_state)

    self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)

    del self.remove_locks[locking.LEVEL_NODEGROUP]

class LUGroupAssignNodes(NoHooksLU):

  """Logical unit for assigning nodes to groups.

  """

  REQ_BGL = False

  def ExpandNames(self):

    # These raise errors.OpPrereqError on their own:

    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    (self.op.node_uuids, self.op.nodes) = GetWantedNodes(self, self.op.nodes)

    # We want to lock all the affected nodes and groups. We have readily

    # available the list of nodes, and the *destination* group. To gather the

    # list of "source" groups, we need to fetch node information later on.

    self.needed_locks = {

      locking.LEVEL_NODEGROUP: set([self.group_uuid]),

      locking.LEVEL_NODE: self.op.node_uuids,

      }

  def DeclareLocks(self, level):

    if level == locking.LEVEL_NODEGROUP:

      assert len(self.needed_locks[locking.LEVEL_NODEGROUP]) == 1

      # Try to get all affected nodes' groups without having the group or node

      # lock yet. Needs verification later in the code flow.

      groups = self.cfg.GetNodeGroupsFromNodes(self.op.node_uuids)

      self.needed_locks[locking.LEVEL_NODEGROUP].update(groups)

  def CheckPrereq(self):

    """Check prerequisites.

    """

    assert self.needed_locks[locking.LEVEL_NODEGROUP]

    assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) ==

            frozenset(self.op.node_uuids))

    expected_locks = (set([self.group_uuid]) |

                      self.cfg.GetNodeGroupsFromNodes(self.op.node_uuids))

    actual_locks = self.owned_locks(locking.LEVEL_NODEGROUP)

    if actual_locks != expected_locks:

      raise errors.OpExecError("Nodes changed groups since locks were acquired,"

                               " current groups are '%s', used to be '%s'" %

                               (utils.CommaJoin(expected_locks),

                                utils.CommaJoin(actual_locks)))

    self.node_data = self.cfg.GetAllNodesInfo()

    self.group = self.cfg.GetNodeGroup(self.group_uuid)

    instance_data = self.cfg.GetAllInstancesInfo()

    if self.group is None:

      raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %

                               (self.op.group_name, self.group_uuid))

    (new_splits, previous_splits) = \

      self.CheckAssignmentForSplitInstances([(uuid, self.group_uuid)

                                             for uuid in self.op.node_uuids],

                                            self.node_data, instance_data)

    if new_splits:

      fmt_new_splits = utils.CommaJoin(utils.NiceSort(

                         self.cfg.GetInstanceNames(new_splits)))

      if not self.op.force:

        raise errors.OpExecError("The following instances get split by this"

                                 " change and --force was not given: %s" %

                                 fmt_new_splits)

      else:

        self.LogWarning("This operation will split the following instances: %s",

                        fmt_new_splits)

        if previous_splits:

          self.LogWarning("In addition, these already-split instances continue"

                          " to be split across groups: %s",

                          utils.CommaJoin(utils.NiceSort(

                            self.cfg.GetInstanceNames(previous_splits))))

  def Exec(self, feedback_fn):

    """Assign nodes to a new group.

    """

    mods = [(node_uuid, self.group_uuid) for node_uuid in self.op.node_uuids]

    self.cfg.AssignGroupNodes(mods)

  @staticmethod

  def CheckAssignmentForSplitInstances(changes, node_data, instance_data):

    """Check for split instances after a node assignment.

    This method considers a series of node assignments as an atomic operation,

    and returns information about split instances after applying the set of

    changes.

    In particular, it returns information about newly split instances, and

    instances that were already split, and remain so after the change.

    Only instances whose disk template is listed in constants.DTS_INT_MIRROR are

    considered.

    @type changes: list of (node_uuid, new_group_uuid) pairs.

    @param changes: list of node assignments to consider.

    @param node_data: a dict with data for all nodes

    @param instance_data: a dict with all instances to consider

    @rtype: a two-tuple

    @return: a list of instances that were previously okay and result split as a

      consequence of this change, and a list of instances that were previously

      split and this change does not fix.

    """

    changed_nodes = dict((uuid, group) for uuid, group in changes

                         if node_data[uuid].group != group)

    all_split_instances = set()

    previously_split_instances = set()

    for inst in instance_data.values():

      if inst.disk_template not in constants.DTS_INT_MIRROR:

        continue

      if len(set(node_data[node_uuid].group

                 for node_uuid in inst.all_nodes)) > 1:

        previously_split_instances.add(inst.uuid)

      if len(set(changed_nodes.get(node_uuid, node_data[node_uuid].group)

                 for node_uuid in inst.all_nodes)) > 1:

        all_split_instances.add(inst.uuid)

    return (list(all_split_instances - previously_split_instances),

            list(previously_split_instances & all_split_instances))

class LUGroupSetParams(LogicalUnit):

  """Modifies the parameters of a node group.

  """

  HPATH = "group-modify"

  HTYPE = constants.HTYPE_GROUP

  REQ_BGL = False

  def CheckArguments(self):

    all_changes = [

      self.op.ndparams,

      self.op.diskparams,

      self.op.alloc_policy,

      self.op.hv_state,

      self.op.disk_state,

      self.op.ipolicy,

      ]

    if all_changes.count(None) == len(all_changes):

      raise errors.OpPrereqError("Please pass at least one modification",

                                 errors.ECODE_INVAL)

    if self.op.diskparams:

      CheckDiskAccessModeValidity(self.op.diskparams)

  def ExpandNames(self):

    # This raises errors.OpPrereqError on its own:

    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    self.needed_locks = {

      locking.LEVEL_INSTANCE: [],

      locking.LEVEL_NODEGROUP: [self.group_uuid],

      }

    self.share_locks[locking.LEVEL_INSTANCE] = 1

  def DeclareLocks(self, level):

    if level == locking.LEVEL_INSTANCE:

      assert not self.needed_locks[locking.LEVEL_INSTANCE]

      # Lock instances optimistically, needs verification once group lock has

      # been acquired

      self.needed_locks[locking.LEVEL_INSTANCE] = \

        self.cfg.GetInstanceNames(

          self.cfg.GetNodeGroupInstances(self.group_uuid))

  @staticmethod

  def _UpdateAndVerifyDiskParams(old, new):

    """Updates and verifies disk parameters.

    """

    new_params = GetUpdatedParams(old, new)

    utils.ForceDictType(new_params, constants.DISK_DT_TYPES)

    return new_params

  def _CheckIpolicy(self, cluster, owned_instance_names):

    """Sanity checks for the ipolicy.

    @type cluster: C{objects.Cluster}

    @param cluster: the cluster's configuration

    @type owned_instance_names: list of string

    @param owned_instance_names: list of instances

    """

    if self.op.ipolicy:

      self.new_ipolicy = GetUpdatedIPolicy(self.group.ipolicy,

                                           self.op.ipolicy,

                                           group_policy=True)

      new_ipolicy = cluster.SimpleFillIPolicy(self.new_ipolicy)

      CheckIpolicyVsDiskTemplates(new_ipolicy,

                                  cluster.enabled_disk_templates)

      instances = \

        dict(self.cfg.GetMultiInstanceInfoByName(owned_instance_names))

      gmi = ganeti.masterd.instance

      violations = \

          ComputeNewInstanceViolations(gmi.CalculateGroupIPolicy(cluster,

                                                                 self.group),

                                       new_ipolicy, instances.values(),

                                       self.cfg)

      if violations:

        self.LogWarning("After the ipolicy change the following instances"

                        " violate them: %s",

                        utils.CommaJoin(violations))

  def CheckPrereq(self):

    """Check prerequisites.

    """

    owned_instance_names = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))

    # Check if locked instances are still correct

    CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instance_names)

    self.group = self.cfg.GetNodeGroup(self.group_uuid)

    cluster = self.cfg.GetClusterInfo()

    if self.group is None:

      raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %

                               (self.op.group_name, self.group_uuid))

    if self.op.ndparams:

      new_ndparams = GetUpdatedParams(self.group.ndparams, self.op.ndparams)

      utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)

      self.new_ndparams = new_ndparams

    if self.op.diskparams:

      diskparams = self.group.diskparams

      uavdp = self._UpdateAndVerifyDiskParams

      # For each disktemplate subdict update and verify the values

      new_diskparams = dict((dt,

                             uavdp(diskparams.get(dt, {}),

                                   self.op.diskparams[dt]))

                            for dt in constants.DISK_TEMPLATES

                            if dt in self.op.diskparams)

      # As we've all subdicts of diskparams ready, lets merge the actual

      # dict with all updated subdicts

      self.new_diskparams = objects.FillDict(diskparams, new_diskparams)

      try:

        utils.VerifyDictOptions(self.new_diskparams, constants.DISK_DT_DEFAULTS)

        CheckDiskAccessModeConsistency(self.new_diskparams, self.cfg,

                                       group=self.group)

      except errors.OpPrereqError, err:

        raise errors.OpPrereqError("While verify diskparams options: %s" % err,

                                   errors.ECODE_INVAL)

    if self.op.hv_state:

      self.new_hv_state = MergeAndVerifyHvState(self.op.hv_state,

                                                self.group.hv_state_static)

    if self.op.disk_state:

      self.new_disk_state = \

        MergeAndVerifyDiskState(self.op.disk_state,

                                self.group.disk_state_static)

    self._CheckIpolicy(cluster, owned_instance_names)

  def BuildHooksEnv(self):

    """Build hooks env.

    """

    return {

      "GROUP_NAME": self.op.group_name,

      "NEW_ALLOC_POLICY": self.op.alloc_policy,

      }

  def BuildHooksNodes(self):

    """Build hooks nodes.

    """

    mn = self.cfg.GetMasterNode()

    return ([mn], [mn])

  def Exec(self, feedback_fn):

    """Modifies the node group.

    """

    result = []

    if self.op.ndparams:

      self.group.ndparams = self.new_ndparams

      result.append(("ndparams", str(self.group.ndparams)))

    if self.op.diskparams:

      self.group.diskparams = self.new_diskparams

      result.append(("diskparams", str(self.group.diskparams)))

    if self.op.alloc_policy:

      self.group.alloc_policy = self.op.alloc_policy

    if self.op.hv_state:

      self.group.hv_state_static = self.new_hv_state

    if self.op.disk_state:

      self.group.disk_state_static = self.new_disk_state

    if self.op.ipolicy:

      self.group.ipolicy = self.new_ipolicy

    self.cfg.Update(self.group, feedback_fn)

    return result

class LUGroupRemove(LogicalUnit):

  HPATH = "group-remove"

  HTYPE = constants.HTYPE_GROUP

  REQ_BGL = False

  def ExpandNames(self):

    # This will raises errors.OpPrereqError on its own:

    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    self.needed_locks = {

      locking.LEVEL_NODEGROUP: [self.group_uuid],

      }

  def CheckPrereq(self):

    """Check prerequisites.

    This checks that the given group name exists as a node group, that is

    empty (i.e., contains no nodes), and that is not the last group of the

    cluster.

    """

    # Verify that the group is empty.

    group_nodes = [node.uuid

                   for node in self.cfg.GetAllNodesInfo().values()

                   if node.group == self.group_uuid]

    if group_nodes:

      raise errors.OpPrereqError("Group '%s' not empty, has the following"

                                 " nodes: %s" %

                                 (self.op.group_name,

                                  utils.CommaJoin(utils.NiceSort(group_nodes))),

                                 errors.ECODE_STATE)

    # Verify the cluster would not be left group-less.

    if len(self.cfg.GetNodeGroupList()) == 1:

      raise errors.OpPrereqError("Group '%s' is the only group, cannot be"

                                 " removed" % self.op.group_name,

                                 errors.ECODE_STATE)

  def BuildHooksEnv(self):

    """Build hooks env.

    """

    return {

      "GROUP_NAME": self.op.group_name,

      }

  def BuildHooksNodes(self):

    """Build hooks nodes.

    """

    mn = self.cfg.GetMasterNode()

    return ([mn], [mn])

  def Exec(self, feedback_fn):

    """Remove the node group.

    """

    try:

      self.cfg.RemoveNodeGroup(self.group_uuid)

    except errors.ConfigurationError:

      raise errors.OpExecError("Group '%s' with UUID %s disappeared" %

                               (self.op.group_name, self.group_uuid))

    self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid

class LUGroupRename(LogicalUnit):

  HPATH = "group-rename"

  HTYPE = constants.HTYPE_GROUP

  REQ_BGL = False

  def ExpandNames(self):

    # This raises errors.OpPrereqError on its own:

    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    self.needed_locks = {

      locking.LEVEL_NODEGROUP: [self.group_uuid],

      }

  def CheckPrereq(self):

    """Check prerequisites.

    Ensures requested new name is not yet used.

    """

    try:

      new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)

    except errors.OpPrereqError:

      pass

    else:

      raise errors.OpPrereqError("Desired new name '%s' clashes with existing"

                                 " node group (UUID: %s)" %

                                 (self.op.new_name, new_name_uuid),

                                 errors.ECODE_EXISTS)

  def BuildHooksEnv(self):

    """Build hooks env.

    """

    return {

      "OLD_NAME": self.op.group_name,

      "NEW_NAME": self.op.new_name,

      }

  def BuildHooksNodes(self):

    """Build hooks nodes.

    """

    mn = self.cfg.GetMasterNode()

    all_nodes = self.cfg.GetAllNodesInfo()

    all_nodes.pop(mn, None)

    run_nodes = [mn]

    run_nodes.extend(node.uuid for node in all_nodes.values()

                     if node.group == self.group_uuid)

    return (run_nodes, run_nodes)

  def Exec(self, feedback_fn):

    """Rename the node group.

    """

    group = self.cfg.GetNodeGroup(self.group_uuid)

    if group is None:

      raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %

                               (self.op.group_name, self.group_uuid))

    group.name = self.op.new_name

    self.cfg.Update(group, feedback_fn)

    return self.op.new_name

class LUGroupEvacuate(LogicalUnit):

  HPATH = "group-evacuate"

  HTYPE = constants.HTYPE_GROUP

  REQ_BGL = False

  def ExpandNames(self):

    # This raises errors.OpPrereqError on its own:

    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    if self.op.target_groups:

      self.req_target_uuids = map(self.cfg.LookupNodeGroup,

                                  self.op.target_groups)

    else:

      self.req_target_uuids = []

    if self.group_uuid in self.req_target_uuids:

      raise errors.OpPrereqError("Group to be evacuated (%s) can not be used"

                                 " as a target group (targets are %s)" %

                                 (self.group_uuid,

                                  utils.CommaJoin(self.req_target_uuids)),

                                 errors.ECODE_INVAL)

    self.op.iallocator = GetDefaultIAllocator(self.cfg, self.op.iallocator)

    self.share_locks = ShareAll()

    self.needed_locks = {

      locking.LEVEL_INSTANCE: [],

      locking.LEVEL_NODEGROUP: [],

      locking.LEVEL_NODE: [],

      }

  def DeclareLocks(self, level):

    if level == locking.LEVEL_INSTANCE:

      assert not self.needed_locks[locking.LEVEL_INSTANCE]

      # Lock instances optimistically, needs verification once node and group

      # locks have been acquired

      self.needed_locks[locking.LEVEL_INSTANCE] = \

        self.cfg.GetInstanceNames(

          self.cfg.GetNodeGroupInstances(self.group_uuid))

    elif level == locking.LEVEL_NODEGROUP:

      assert not self.needed_locks[locking.LEVEL_NODEGROUP]

      if self.req_target_uuids:

        lock_groups = set([self.group_uuid] + self.req_target_uuids)

        # Lock all groups used by instances optimistically; this requires going

        # via the node before it's locked, requiring verification later on

        lock_groups.update(group_uuid

                           for instance_name in

                             self.owned_locks(locking.LEVEL_INSTANCE)

                           for group_uuid in

                             self.cfg.GetInstanceNodeGroups(

                               self.cfg.GetInstanceInfoByName(instance_name)

                                 .uuid))

      else:

        # No target groups, need to lock all of them

        lock_groups = locking.ALL_SET

      self.needed_locks[locking.LEVEL_NODEGROUP] = lock_groups

    elif level == locking.LEVEL_NODE:

      # This will only lock the nodes in the group to be evacuated which

      # contain actual instances

      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND

      self._LockInstancesNodes()

      # Lock all nodes in group to be evacuated and target groups

      owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))

      assert self.group_uuid in owned_groups

      member_node_uuids = [node_uuid

                           for group in owned_groups

                           for node_uuid in

                             self.cfg.GetNodeGroup(group).members]

      self.needed_locks[locking.LEVEL_NODE].extend(member_node_uuids)

  def CheckPrereq(self):

    owned_instance_names = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))

    owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))

    owned_node_uuids = frozenset(self.owned_locks(locking.LEVEL_NODE))

    assert owned_groups.issuperset(self.req_target_uuids)

    assert self.group_uuid in owned_groups

    # Check if locked instances are still correct

    CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_instance_names)

    # Get instance information

    self.instances = \

      dict(self.cfg.GetMultiInstanceInfoByName(owned_instance_names))

    # Check if node groups for locked instances are still correct

    CheckInstancesNodeGroups(self.cfg, self.instances,

                             owned_groups, owned_node_uuids, self.group_uuid)

    if self.req_target_uuids:

      # User requested specific target groups

      self.target_uuids = self.req_target_uuids

    else:

      # All groups except the one to be evacuated are potential targets

      self.target_uuids = [group_uuid for group_uuid in owned_groups

                           if group_uuid != self.group_uuid]

      if not self.target_uuids:

        raise errors.OpPrereqError("There are no possible target groups",

                                   errors.ECODE_INVAL)

  def BuildHooksEnv(self):

    """Build hooks env.

    """

    return {

      "GROUP_NAME": self.op.group_name,

      "TARGET_GROUPS": " ".join(self.target_uuids),

      }

  def BuildHooksNodes(self):

    """Build hooks nodes.

    """

    mn = self.cfg.GetMasterNode()

    assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)

    run_nodes = [mn] + self.cfg.GetNodeGroup(self.group_uuid).members

    return (run_nodes, run_nodes)

  def Exec(self, feedback_fn):

    inst_names = list(self.owned_locks(locking.LEVEL_INSTANCE))

    assert self.group_uuid not in self.target_uuids

    req = iallocator.IAReqGroupChange(instances=inst_names,

                                      target_groups=self.target_uuids)

    ial = iallocator.IAllocator(self.cfg, self.rpc, req)

    ial.Run(self.op.iallocator)

    if not ial.success:

      raise errors.OpPrereqError("Can't compute group evacuation using"

                                 " iallocator '%s': %s" %

                                 (self.op.iallocator, ial.info),

                                 errors.ECODE_NORES)

    jobs = LoadNodeEvacResult(self, ial.result, self.op.early_release, False)

    self.LogInfo("Iallocator returned %s job(s) for evacuating node group %s",

                 len(jobs), self.op.group_name)

    return ResultWithJobs(jobs)

class LUGroupVerifyDisks(NoHooksLU):

  """Verifies the status of all disks in a node group.

  """

  REQ_BGL = False

  def ExpandNames(self):

    # Raises errors.OpPrereqError on its own if group can't be found

    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)

    self.share_locks = ShareAll()

    self.needed_locks = {

      locking.LEVEL_INSTANCE: [],

      locking.LEVEL_NODEGROUP: [],

      locking.LEVEL_NODE: [],

      # This opcode is acquires all node locks in a group. LUClusterVerifyDisks

      # starts one instance of this opcode for every group, which means all

      # nodes will be locked for a short amount of time, so it's better to

      # acquire the node allocation lock as well.

      locking.LEVEL_NODE_ALLOC: locking.ALL_SET,

      }

  def DeclareLocks(self, level):

    if level == locking.LEVEL_INSTANCE:

      assert not self.needed_locks[locking.LEVEL_INSTANCE]

      # Lock instances optimistically, needs verification once node and group

      # locks have been acquired

      self.needed_locks[locking.LEVEL_INSTANCE] = \

        self.cfg.GetInstanceNames(

          self.cfg.GetNodeGroupInstances(self.group_uuid))

    elif level == locking.LEVEL_NODEGROUP:

      assert not self.needed_locks[locking.LEVEL_NODEGROUP]

      self.needed_locks[locking.LEVEL_NODEGROUP] = \

        set([self.group_uuid] +

            # Lock all groups used by instances optimistically; this requires

            # going via the node before it's locked, requiring verification

            # later on

            [group_uuid

             for instance_name in self.owned_locks(locking.LEVEL_INSTANCE)

             for group_uuid in

               self.cfg.GetInstanceNodeGroups(

                 self.cfg.GetInstanceInfoByName(instance_name).uuid)])

    elif level == locking.LEVEL_NODE:

      # This will only lock the nodes in the group to be verified which contain

      # actual instances

      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND

      self._LockInstancesNodes()

      # Lock all nodes in group to be verified

      assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)

      member_node_uuids = self.cfg.GetNodeGroup(self.group_uuid).members

      self.needed_locks[locking.LEVEL_NODE].extend(member_node_uuids)

  def CheckPrereq(self):

    owned_inst_names = frozenset(self.owned_locks(locking.LEVEL_INSTANCE))

    owned_groups = frozenset(self.owned_locks(locking.LEVEL_NODEGROUP))

    owned_node_uuids = frozenset(self.owned_locks(locking.LEVEL_NODE))

    assert self.group_uuid in owned_groups

    # Check if locked instances are still correct

    CheckNodeGroupInstances(self.cfg, self.group_uuid, owned_inst_names)

    # Get instance information

    self.instances = dict(self.cfg.GetMultiInstanceInfoByName(owned_inst_names))

    # Check if node groups for locked instances are still correct

    CheckInstancesNodeGroups(self.cfg, self.instances,

                             owned_groups, owned_node_uuids, self.group_uuid)

  def _VerifyInstanceLvs(self, node_errors, offline_disk_instance_names,

                         missing_disks):

    node_lv_to_inst = MapInstanceLvsToNodes(

      [inst for inst in self.instances.values() if inst.disks_active])

    if node_lv_to_inst:

      node_uuids = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &

                                  set(self.cfg.GetVmCapableNodeList()))

      node_lvs = self.rpc.call_lv_list(node_uuids, [])

      for (node_uuid, node_res) in node_lvs.items():

        if node_res.offline:

          continue

        msg = node_res.fail_msg

        if msg:

          logging.warning("Error enumerating LVs on node %s: %s",

                          self.cfg.GetNodeName(node_uuid), msg)

          node_errors[node_uuid] = msg

          continue

        for lv_name, (_, _, lv_online) in node_res.payload.items():

          inst = node_lv_to_inst.pop((node_uuid, lv_name), None)

          if not lv_online and inst is not None:

            offline_disk_instance_names.add(inst.name)

      # any leftover items in nv_dict are missing LVs, let's arrange the data

      # better

      for key, inst in node_lv_to_inst.iteritems():

        missing_disks.setdefault(inst.name, []).append(list(key))

  def _VerifyDrbdStates(self, node_errors, offline_disk_instance_names):

    node_to_inst = {}

    for inst in self.instances.values():

      if not inst.disks_active or inst.disk_template != constants.DT_DRBD8:

        continue

      for node_uuid in itertools.chain([inst.primary_node],

                                       inst.secondary_nodes):

        node_to_inst.setdefault(node_uuid, []).append(inst)

    for (node_uuid, insts) in node_to_inst.items():

      node_disks = [(inst.disks, inst) for inst in insts]

      node_res = self.rpc.call_drbd_needs_activation(node_uuid, node_disks)

      msg = node_res.fail_msg

      if msg:

        logging.warning("Error getting DRBD status on node %s: %s",

                        self.cfg.GetNodeName(node_uuid), msg)

        node_errors[node_uuid] = msg

        continue

      faulty_disk_uuids = set(node_res.payload)

      for inst in self.instances.values():

        inst_disk_uuids = set([disk.uuid for disk in inst.disks])

        if inst_disk_uuids.intersection(faulty_disk_uuids):

          offline_disk_instance_names.add(inst.name)

  def Exec(self, feedback_fn):

    """Verify integrity of cluster disks.

    @rtype: tuple of three items

    @return: a tuple of (dict of node-to-node_error, list of instances

        which need activate-disks, dict of instance: (node, volume) for

        missing volumes

    """

    node_errors = {}

    offline_disk_instance_names = set()

    missing_disks = {}

    self._VerifyInstanceLvs(node_errors, offline_disk_instance_names,

                            missing_disks)

    self._VerifyDrbdStates(node_errors, offline_disk_instance_names)

    return (node_errors, list(offline_disk_instance_names), missing_disks)