/lib/cmdlib.py - snf-ganeti - Greek Research and Technology Network's projects

| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ e6b8d02d

History | View | Annotate | Download (349 kB)

+      #
+      #
       # Copyright (C) 2006, 2007, 2008 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.
       """Module implementing the master-side code."""
       # pylint: disable-msg=W0201
       # W0201 since most LU attributes are defined in CheckPrereq or similar
       # functions
       import os
       import os.path
       import time
       import re
       import platform
       import logging
       import copy
       import OpenSSL
       from ganeti import ssh
       from ganeti import utils
       from ganeti import errors
       from ganeti import hypervisor
       from ganeti import locking
       from ganeti import constants
       from ganeti import objects
       from ganeti import serializer
       from ganeti import ssconf
       from ganeti import uidpool
       from ganeti import compat
       from ganeti import masterd
       import ganeti.masterd.instance # pylint: disable-msg=W0611
       class LogicalUnit(object):
         """Logical Unit base class.
         Subclasses must follow these rules:
           - implement ExpandNames
           - implement CheckPrereq (except when tasklets are used)
           - implement Exec (except when tasklets are used)
           - implement BuildHooksEnv
           - redefine HPATH and HTYPE
           - optionally redefine their run requirements:
               REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
         Note that all commands require root permissions.
         @ivar dry_run_result: the value (if any) that will be returned to the caller
             in dry-run mode (signalled by opcode dry_run parameter)
         """
         HPATH = None
         HTYPE = None
         _OP_REQP = []
         REQ_BGL = True
         def __init__(self, processor, op, context, rpc):
           """Constructor for LogicalUnit.
           This needs to be overridden in derived classes in order to check op
           validity.
           """
           self.proc = processor
           self.op = op
           self.cfg = context.cfg
           self.context = context
           self.rpc = rpc
           # Dicts used to declare locking needs to mcpu
           self.needed_locks = None
           self.acquired_locks = {}
           self.share_locks = dict.fromkeys(locking.LEVELS, 0)
           self.add_locks = {}
           self.remove_locks = {}
           # Used to force good behavior when calling helper functions
           self.recalculate_locks = {}
           self.__ssh = None
           # logging
           self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
           self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
           self.LogStep = processor.LogStep # pylint: disable-msg=C0103
           # support for dry-run
           self.dry_run_result = None
           # support for generic debug attribute
           if (not hasattr(self.op, "debug_level") or
               not isinstance(self.op.debug_level, int)):
             self.op.debug_level = 0
           # Tasklets
           self.tasklets = None
           for attr_name in self._OP_REQP:
             attr_val = getattr(op, attr_name, None)
             if attr_val is None:
               raise errors.OpPrereqError("Required parameter '%s' missing" %
                                          attr_name, errors.ECODE_INVAL)
           self.CheckArguments()
         def __GetSSH(self):
           """Returns the SshRunner object
           """
           if not self.__ssh:
             self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
           return self.__ssh
         ssh = property(fget=__GetSSH)
         def CheckArguments(self):
           """Check syntactic validity for the opcode arguments.
           This method is for doing a simple syntactic check and ensure
           validity of opcode parameters, without any cluster-related
           checks. While the same can be accomplished in ExpandNames and/or
           CheckPrereq, doing these separate is better because:
             - ExpandNames is left as as purely a lock-related function
             - CheckPrereq is run after we have acquired locks (and possible
               waited for them)
           The function is allowed to change the self.op attribute so that
           later methods can no longer worry about missing parameters.
           """
           pass
         def ExpandNames(self):
           """Expand names for this LU.
           This method is called before starting to execute the opcode, and it should
           update all the parameters of the opcode to their canonical form (e.g. a
           short node name must be fully expanded after this method has successfully
           completed). This way locking, hooks, logging, ecc. can work correctly.
           LUs which implement this method must also populate the self.needed_locks
           member, as a dict with lock levels as keys, and a list of needed lock names
           as values. Rules:
             - use an empty dict if you don't need any lock
             - if you don't need any lock at a particular level omit that level
             - don't put anything for the BGL level
             - if you want all locks at a level use locking.ALL_SET as a value
           If you need to share locks (rather than acquire them exclusively) at one
           level you can modify self.share_locks, setting a true value (usually 1) for
           that level. By default locks are not shared.
           This function can also define a list of tasklets, which then will be
           executed in order instead of the usual LU-level CheckPrereq and Exec
           functions, if those are not defined by the LU.
           Examples::
             # Acquire all nodes and one instance
             self.needed_locks = {
               locking.LEVEL_NODE: locking.ALL_SET,
               locking.LEVEL_INSTANCE: ['instance1.example.tld'],
+            }
             # Acquire just two nodes
             self.needed_locks = {
               locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
+            }
             # Acquire no locks
             self.needed_locks = {} # No, you can't leave it to the default value None
           """
           # The implementation of this method is mandatory only if the new LU is
           # concurrent, so that old LUs don't need to be changed all at the same
           # time.
           if self.REQ_BGL:
             self.needed_locks = {} # Exclusive LUs don't need locks.
           else:
             raise NotImplementedError
         def DeclareLocks(self, level):
           """Declare LU locking needs for a level
           While most LUs can just declare their locking needs at ExpandNames time,
           sometimes there's the need to calculate some locks after having acquired
           the ones before. This function is called just before acquiring locks at a
           particular level, but after acquiring the ones at lower levels, and permits
           such calculations. It can be used to modify self.needed_locks, and by
           default it does nothing.
           This function is only called if you have something already set in
           self.needed_locks for the level.
           @param level: Locking level which is going to be locked
           @type level: member of ganeti.locking.LEVELS
           """
         def CheckPrereq(self):
           """Check prerequisites for this LU.
           This method should check that the prerequisites for the execution
           of this LU are fulfilled. It can do internode communication, but
           it should be idempotent - no cluster or system changes are
           allowed.
           The method should raise errors.OpPrereqError in case something is
           not fulfilled. Its return value is ignored.
           This method should also update all the parameters of the opcode to
           their canonical form if it hasn't been done by ExpandNames before.
           """
           if self.tasklets is not None:
             for (idx, tl) in enumerate(self.tasklets):
               logging.debug("Checking prerequisites for tasklet %s/%s",
                             idx + 1, len(self.tasklets))
               tl.CheckPrereq()
           else:
             raise NotImplementedError
         def Exec(self, feedback_fn):
           """Execute the LU.
           This method should implement the actual work. It should raise
           errors.OpExecError for failures that are somewhat dealt with in
           code, or expected.
           """
           if self.tasklets is not None:
             for (idx, tl) in enumerate(self.tasklets):
               logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
               tl.Exec(feedback_fn)
           else:
             raise NotImplementedError
         def BuildHooksEnv(self):
           """Build hooks environment for this LU.
           This method should return a three-node tuple consisting of: a dict
           containing the environment that will be used for running the
           specific hook for this LU, a list of node names on which the hook
           should run before the execution, and a list of node names on which
           the hook should run after the execution.
           The keys of the dict must not have 'GANETI_' prefixed as this will
           be handled in the hooks runner. Also note additional keys will be
           added by the hooks runner. If the LU doesn't define any
           environment, an empty dict (and not None) should be returned.
           No nodes should be returned as an empty list (and not None).
           Note that if the HPATH for a LU class is None, this function will
           not be called.
           """
           raise NotImplementedError
         def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
           """Notify the LU about the results of its hooks.
           This method is called every time a hooks phase is executed, and notifies
           the Logical Unit about the hooks' result. The LU can then use it to alter
           its result based on the hooks.  By default the method does nothing and the
           previous result is passed back unchanged but any LU can define it if it
           wants to use the local cluster hook-scripts somehow.
           @param phase: one of L{constants.HOOKS_PHASE_POST} or
               L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
           @param hook_results: the results of the multi-node hooks rpc call
           @param feedback_fn: function used send feedback back to the caller
           @param lu_result: the previous Exec result this LU had, or None
               in the PRE phase
           @return: the new Exec result, based on the previous result
               and hook results
           """
           # API must be kept, thus we ignore the unused argument and could
           # be a function warnings
           # pylint: disable-msg=W0613,R0201
           return lu_result
         def _ExpandAndLockInstance(self):
           """Helper function to expand and lock an instance.
           Many LUs that work on an instance take its name in self.op.instance_name
           and need to expand it and then declare the expanded name for locking. This
           function does it, and then updates self.op.instance_name to the expanded
           name. It also initializes needed_locks as a dict, if this hasn't been done
           before.
           """
           if self.needed_locks is None:
             self.needed_locks = {}
           else:
             assert locking.LEVEL_INSTANCE not in self.needed_locks, \
               "_ExpandAndLockInstance called with instance-level locks set"
           self.op.instance_name = _ExpandInstanceName(self.cfg,
                                                       self.op.instance_name)
           self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
         def _LockInstancesNodes(self, primary_only=False):
           """Helper function to declare instances' nodes for locking.
           This function should be called after locking one or more instances to lock
           their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
           with all primary or secondary nodes for instances already locked and
           present in self.needed_locks[locking.LEVEL_INSTANCE].
           It should be called from DeclareLocks, and for safety only works if
           self.recalculate_locks[locking.LEVEL_NODE] is set.
           In the future it may grow parameters to just lock some instance's nodes, or
           to just lock primaries or secondary nodes, if needed.
           If should be called in DeclareLocks in a way similar to::
             if level == locking.LEVEL_NODE:
               self._LockInstancesNodes()
           @type primary_only: boolean
           @param primary_only: only lock primary nodes of locked instances
           """
           assert locking.LEVEL_NODE in self.recalculate_locks, \
             "_LockInstancesNodes helper function called with no nodes to recalculate"
           # TODO: check if we're really been called with the instance locks held
           # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
           # future we might want to have different behaviors depending on the value
           # of self.recalculate_locks[locking.LEVEL_NODE]
           wanted_nodes = []
           for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
             instance = self.context.cfg.GetInstanceInfo(instance_name)
             wanted_nodes.append(instance.primary_node)
             if not primary_only:
               wanted_nodes.extend(instance.secondary_nodes)
           if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
             self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
           elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
             self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
           del self.recalculate_locks[locking.LEVEL_NODE]
       class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
         """Simple LU which runs no hooks.
         This LU is intended as a parent for other LogicalUnits which will
         run no hooks, in order to reduce duplicate code.
         """
         HPATH = None
         HTYPE = None
         def BuildHooksEnv(self):
           """Empty BuildHooksEnv for NoHooksLu.
           This just raises an error.
           """
           assert False, "BuildHooksEnv called for NoHooksLUs"
       class Tasklet:
         """Tasklet base class.
         Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
         they can mix legacy code with tasklets. Locking needs to be done in the LU,
         tasklets know nothing about locks.
         Subclasses must follow these rules:
           - Implement CheckPrereq
           - Implement Exec
         """
         def __init__(self, lu):
           self.lu = lu
           # Shortcuts
           self.cfg = lu.cfg
           self.rpc = lu.rpc
         def CheckPrereq(self):
           """Check prerequisites for this tasklets.
           This method should check whether the prerequisites for the execution of
           this tasklet are fulfilled. It can do internode communication, but it
           should be idempotent - no cluster or system changes are allowed.
           The method should raise errors.OpPrereqError in case something is not
           fulfilled. Its return value is ignored.
           This method should also update all parameters to their canonical form if it
           hasn't been done before.
           """
           raise NotImplementedError
         def Exec(self, feedback_fn):
           """Execute the tasklet.
           This method should implement the actual work. It should raise
           errors.OpExecError for failures that are somewhat dealt with in code, or
           expected.
           """
           raise NotImplementedError
       def _GetWantedNodes(lu, nodes):
         """Returns list of checked and expanded node names.
         @type lu: L{LogicalUnit}
         @param lu: the logical unit on whose behalf we execute
         @type nodes: list
         @param nodes: list of node names or None for all nodes
         @rtype: list
         @return: the list of nodes, sorted
         @raise errors.ProgrammerError: if the nodes parameter is wrong type
         """
         if not isinstance(nodes, list):
           raise errors.OpPrereqError("Invalid argument type 'nodes'",
                                      errors.ECODE_INVAL)
         if not nodes:
           raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
             " non-empty list of nodes whose name is to be expanded.")
         wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
         return utils.NiceSort(wanted)
       def _GetWantedInstances(lu, instances):
         """Returns list of checked and expanded instance names.
         @type lu: L{LogicalUnit}
         @param lu: the logical unit on whose behalf we execute
         @type instances: list
         @param instances: list of instance names or None for all instances
         @rtype: list
         @return: the list of instances, sorted
         @raise errors.OpPrereqError: if the instances parameter is wrong type
         @raise errors.OpPrereqError: if any of the passed instances is not found
         """
         if not isinstance(instances, list):
           raise errors.OpPrereqError("Invalid argument type 'instances'",
                                      errors.ECODE_INVAL)
         if instances:
           wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
         else:
           wanted = utils.NiceSort(lu.cfg.GetInstanceList())
         return wanted
       def _CheckOutputFields(static, dynamic, selected):
         """Checks whether all selected fields are valid.
         @type static: L{utils.FieldSet}
         @param static: static fields set
         @type dynamic: L{utils.FieldSet}
         @param dynamic: dynamic fields set
         """
         f = utils.FieldSet()
         f.Extend(static)
         f.Extend(dynamic)
         delta = f.NonMatching(selected)
         if delta:
           raise errors.OpPrereqError("Unknown output fields selected: %s"
                                      % ",".join(delta), errors.ECODE_INVAL)
       def _CheckBooleanOpField(op, name):
         """Validates boolean opcode parameters.
         This will ensure that an opcode parameter is either a boolean value,
         or None (but that it always exists).
         """
         val = getattr(op, name, None)
         if not (val is None or isinstance(val, bool)):
           raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
                                      (name, str(val)), errors.ECODE_INVAL)
         setattr(op, name, val)
       def _CheckGlobalHvParams(params):
         """Validates that given hypervisor params are not global ones.
         This will ensure that instances don't get customised versions of
         global params.
         """
         used_globals = constants.HVC_GLOBALS.intersection(params)
         if used_globals:
           msg = ("The following hypervisor parameters are global and cannot"
                  " be customized at instance level, please modify them at"
                  " cluster level: %s" % utils.CommaJoin(used_globals))
           raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
       def _CheckNodeOnline(lu, node):
         """Ensure that a given node is online.
         @param lu: the LU on behalf of which we make the check
         @param node: the node to check
         @raise errors.OpPrereqError: if the node is offline
         """
         if lu.cfg.GetNodeInfo(node).offline:
           raise errors.OpPrereqError("Can't use offline node %s" % node,
                                      errors.ECODE_INVAL)
       def _CheckNodeNotDrained(lu, node):
         """Ensure that a given node is not drained.
         @param lu: the LU on behalf of which we make the check
         @param node: the node to check
         @raise errors.OpPrereqError: if the node is drained
         """
         if lu.cfg.GetNodeInfo(node).drained:
           raise errors.OpPrereqError("Can't use drained node %s" % node,
                                      errors.ECODE_INVAL)
       def _CheckNodeHasOS(lu, node, os_name, force_variant):
         """Ensure that a node supports a given OS.
         @param lu: the LU on behalf of which we make the check
         @param node: the node to check
         @param os_name: the OS to query about
         @param force_variant: whether to ignore variant errors
         @raise errors.OpPrereqError: if the node is not supporting the OS
         """
         result = lu.rpc.call_os_get(node, os_name)
         result.Raise("OS '%s' not in supported OS list for node %s" %
                      (os_name, node),
                      prereq=True, ecode=errors.ECODE_INVAL)
         if not force_variant:
           _CheckOSVariant(result.payload, os_name)
       def _RequireFileStorage():
         """Checks that file storage is enabled.
         @raise errors.OpPrereqError: when file storage is disabled
         """
         if not constants.ENABLE_FILE_STORAGE:
           raise errors.OpPrereqError("File storage disabled at configure time",
                                      errors.ECODE_INVAL)
       def _CheckDiskTemplate(template):
         """Ensure a given disk template is valid.
         """
         if template not in constants.DISK_TEMPLATES:
           msg = ("Invalid disk template name '%s', valid templates are: %s" %
                  (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
           raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
         if template == constants.DT_FILE:
           _RequireFileStorage()
       def _CheckStorageType(storage_type):
         """Ensure a given storage type is valid.
         """
         if storage_type not in constants.VALID_STORAGE_TYPES:
           raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
                                      errors.ECODE_INVAL)
         if storage_type == constants.ST_FILE:
           _RequireFileStorage()
       def _GetClusterDomainSecret():
         """Reads the cluster domain secret.
         """
         return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
                                      strict=True)
       def _CheckInstanceDown(lu, instance, reason):
         """Ensure that an instance is not running."""
         if instance.admin_up:
           raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
                                      (instance.name, reason), errors.ECODE_STATE)
         pnode = instance.primary_node
         ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
         ins_l.Raise("Can't contact node %s for instance information" % pnode,
                     prereq=True, ecode=errors.ECODE_ENVIRON)
         if instance.name in ins_l.payload:
           raise errors.OpPrereqError("Instance %s is running, %s" %
                                      (instance.name, reason), errors.ECODE_STATE)
       def _ExpandItemName(fn, name, kind):
         """Expand an item name.
         @param fn: the function to use for expansion
         @param name: requested item name
         @param kind: text description ('Node' or 'Instance')
         @return: the resolved (full) name
         @raise errors.OpPrereqError: if the item is not found
         """
         full_name = fn(name)
         if full_name is None:
           raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
                                      errors.ECODE_NOENT)
         return full_name
       def _ExpandNodeName(cfg, name):
         """Wrapper over L{_ExpandItemName} for nodes."""
         return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
       def _ExpandInstanceName(cfg, name):
         """Wrapper over L{_ExpandItemName} for instance."""
         return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
       def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
                                 memory, vcpus, nics, disk_template, disks,
                                 bep, hvp, hypervisor_name):
         """Builds instance related env variables for hooks
         This builds the hook environment from individual variables.
         @type name: string
         @param name: the name of the instance
         @type primary_node: string
         @param primary_node: the name of the instance's primary node
         @type secondary_nodes: list
         @param secondary_nodes: list of secondary nodes as strings
         @type os_type: string
         @param os_type: the name of the instance's OS
         @type status: boolean
         @param status: the should_run status of the instance
         @type memory: string
         @param memory: the memory size of the instance
         @type vcpus: string
         @param vcpus: the count of VCPUs the instance has
         @type nics: list
         @param nics: list of tuples (ip, mac, mode, link) representing
             the NICs the instance has
         @type disk_template: string
         @param disk_template: the disk template of the instance
         @type disks: list
         @param disks: the list of (size, mode) pairs
         @type bep: dict
         @param bep: the backend parameters for the instance
         @type hvp: dict
         @param hvp: the hypervisor parameters for the instance
         @type hypervisor_name: string
         @param hypervisor_name: the hypervisor for the instance
         @rtype: dict
         @return: the hook environment for this instance
         """
         if status:
           str_status = "up"
         else:
           str_status = "down"
         env = {
           "OP_TARGET": name,
           "INSTANCE_NAME": name,
           "INSTANCE_PRIMARY": primary_node,
           "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
           "INSTANCE_OS_TYPE": os_type,
           "INSTANCE_STATUS": str_status,
           "INSTANCE_MEMORY": memory,
           "INSTANCE_VCPUS": vcpus,
           "INSTANCE_DISK_TEMPLATE": disk_template,
           "INSTANCE_HYPERVISOR": hypervisor_name,
+        }
         if nics:
           nic_count = len(nics)
           for idx, (ip, mac, mode, link) in enumerate(nics):
             if ip is None:
               ip = ""
             env["INSTANCE_NIC%d_IP" % idx] = ip
             env["INSTANCE_NIC%d_MAC" % idx] = mac
             env["INSTANCE_NIC%d_MODE" % idx] = mode
             env["INSTANCE_NIC%d_LINK" % idx] = link
             if mode == constants.NIC_MODE_BRIDGED:
               env["INSTANCE_NIC%d_BRIDGE" % idx] = link
         else:
           nic_count = 0
         env["INSTANCE_NIC_COUNT"] = nic_count
         if disks:
           disk_count = len(disks)
           for idx, (size, mode) in enumerate(disks):
             env["INSTANCE_DISK%d_SIZE" % idx] = size
             env["INSTANCE_DISK%d_MODE" % idx] = mode
         else:
           disk_count = 0
         env["INSTANCE_DISK_COUNT"] = disk_count
         for source, kind in [(bep, "BE"), (hvp, "HV")]:
           for key, value in source.items():
             env["INSTANCE_%s_%s" % (kind, key)] = value
         return env
       def _NICListToTuple(lu, nics):
         """Build a list of nic information tuples.
         This list is suitable to be passed to _BuildInstanceHookEnv or as a return
         value in LUQueryInstanceData.
         @type lu:  L{LogicalUnit}
         @param lu: the logical unit on whose behalf we execute
         @type nics: list of L{objects.NIC}
         @param nics: list of nics to convert to hooks tuples
         """
         hooks_nics = []
         c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
         for nic in nics:
           ip = nic.ip
           mac = nic.mac
           filled_params = objects.FillDict(c_nicparams, nic.nicparams)
           mode = filled_params[constants.NIC_MODE]
           link = filled_params[constants.NIC_LINK]
           hooks_nics.append((ip, mac, mode, link))
         return hooks_nics
       def _BuildInstanceHookEnvByObject(lu, instance, override=None):
         """Builds instance related env variables for hooks from an object.
         @type lu: L{LogicalUnit}
         @param lu: the logical unit on whose behalf we execute
         @type instance: L{objects.Instance}
         @param instance: the instance for which we should build the
             environment
         @type override: dict
         @param override: dictionary with key/values that will override
             our values
         @rtype: dict
         @return: the hook environment dictionary
         """
         cluster = lu.cfg.GetClusterInfo()
         bep = cluster.FillBE(instance)
         hvp = cluster.FillHV(instance)
         args = {
           'name': instance.name,
           'primary_node': instance.primary_node,
           'secondary_nodes': instance.secondary_nodes,
           'os_type': instance.os,
           'status': instance.admin_up,
           'memory': bep[constants.BE_MEMORY],
           'vcpus': bep[constants.BE_VCPUS],
           'nics': _NICListToTuple(lu, instance.nics),
           'disk_template': instance.disk_template,
           'disks': [(disk.size, disk.mode) for disk in instance.disks],
           'bep': bep,
           'hvp': hvp,
           'hypervisor_name': instance.hypervisor,
+        }
         if override:
           args.update(override)
         return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
       def _AdjustCandidatePool(lu, exceptions):
         """Adjust the candidate pool after node operations.
         """
         mod_list = lu.cfg.MaintainCandidatePool(exceptions)
         if mod_list:
           lu.LogInfo("Promoted nodes to master candidate role: %s",
                      utils.CommaJoin(node.name for node in mod_list))
           for name in mod_list:
             lu.context.ReaddNode(name)
         mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
         if mc_now > mc_max:
           lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
                      (mc_now, mc_max))
       def _DecideSelfPromotion(lu, exceptions=None):
         """Decide whether I should promote myself as a master candidate.
         """
         cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
         mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
         # the new node will increase mc_max with one, so:
         mc_should = min(mc_should + 1, cp_size)
         return mc_now < mc_should
       def _CheckNicsBridgesExist(lu, target_nics, target_node,
                                      profile=constants.PP_DEFAULT):
         """Check that the brigdes needed by a list of nics exist.
         """
         c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
         paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
                       for nic in target_nics]
         brlist = [params[constants.NIC_LINK] for params in paramslist
                   if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
         if brlist:
           result = lu.rpc.call_bridges_exist(target_node, brlist)
           result.Raise("Error checking bridges on destination node '%s'" %
                        target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
       def _CheckInstanceBridgesExist(lu, instance, node=None):
         """Check that the brigdes needed by an instance exist.
         """
         if node is None:
           node = instance.primary_node
         _CheckNicsBridgesExist(lu, instance.nics, node)
       def _CheckOSVariant(os_obj, name):
         """Check whether an OS name conforms to the os variants specification.
         @type os_obj: L{objects.OS}
         @param os_obj: OS object to check
         @type name: string
         @param name: OS name passed by the user, to check for validity
         """
         if not os_obj.supported_variants:
           return
         try:
           variant = name.split("+", 1)[1]
         except IndexError:
           raise errors.OpPrereqError("OS name must include a variant",
                                      errors.ECODE_INVAL)
         if variant not in os_obj.supported_variants:
           raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
       def _GetNodeInstancesInner(cfg, fn):
         return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
       def _GetNodeInstances(cfg, node_name):
         """Returns a list of all primary and secondary instances on a node.
         """
         return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
       def _GetNodePrimaryInstances(cfg, node_name):
         """Returns primary instances on a node.
         """
         return _GetNodeInstancesInner(cfg,
                                       lambda inst: node_name == inst.primary_node)
       def _GetNodeSecondaryInstances(cfg, node_name):
         """Returns secondary instances on a node.
         """
         return _GetNodeInstancesInner(cfg,
                                       lambda inst: node_name in inst.secondary_nodes)
       def _GetStorageTypeArgs(cfg, storage_type):
         """Returns the arguments for a storage type.
         """
         # Special case for file storage
         if storage_type == constants.ST_FILE:
           # storage.FileStorage wants a list of storage directories
           return [[cfg.GetFileStorageDir()]]
         return []
       def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
         faulty = []
         for dev in instance.disks:
           cfg.SetDiskID(dev, node_name)
         result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
         result.Raise("Failed to get disk status from node %s" % node_name,
                      prereq=prereq, ecode=errors.ECODE_ENVIRON)
         for idx, bdev_status in enumerate(result.payload):
           if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
             faulty.append(idx)
         return faulty
       class LUPostInitCluster(LogicalUnit):
         """Logical unit for running hooks after cluster initialization.
         """
         HPATH = "cluster-init"
         HTYPE = constants.HTYPE_CLUSTER
         _OP_REQP = []
         def BuildHooksEnv(self):
           """Build hooks env.
           """
           env = {"OP_TARGET": self.cfg.GetClusterName()}
           mn = self.cfg.GetMasterNode()
           return env, [], [mn]
         def CheckPrereq(self):
           """No prerequisites to check.
           """
           return True
         def Exec(self, feedback_fn):
           """Nothing to do.
           """
           return True
       class LUDestroyCluster(LogicalUnit):
         """Logical unit for destroying the cluster.
         """
         HPATH = "cluster-destroy"
         HTYPE = constants.HTYPE_CLUSTER
         _OP_REQP = []
         def BuildHooksEnv(self):
           """Build hooks env.
           """
           env = {"OP_TARGET": self.cfg.GetClusterName()}
           return env, [], []
         def CheckPrereq(self):
           """Check prerequisites.
           This checks whether the cluster is empty.
           Any errors are signaled by raising errors.OpPrereqError.
           """
           master = self.cfg.GetMasterNode()
           nodelist = self.cfg.GetNodeList()
           if len(nodelist) != 1 or nodelist[0] != master:
             raise errors.OpPrereqError("There are still %d node(s) in"
                                        " this cluster." % (len(nodelist) - 1),
                                        errors.ECODE_INVAL)
           instancelist = self.cfg.GetInstanceList()
           if instancelist:
             raise errors.OpPrereqError("There are still %d instance(s) in"
                                        " this cluster." % len(instancelist),
                                        errors.ECODE_INVAL)
         def Exec(self, feedback_fn):
           """Destroys the cluster.
           """
           master = self.cfg.GetMasterNode()
           modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
           # Run post hooks on master node before it's removed
           hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
           try:
             hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
           except:
             # pylint: disable-msg=W0702
             self.LogWarning("Errors occurred running hooks on %s" % master)
           result = self.rpc.call_node_stop_master(master, False)
           result.Raise("Could not disable the master role")
           if modify_ssh_setup:
             priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
             utils.CreateBackup(priv_key)
             utils.CreateBackup(pub_key)
           return master
       def _VerifyCertificate(filename):
         """Verifies a certificate for LUVerifyCluster.
         @type filename: string
         @param filename: Path to PEM file
         """
         try:
           cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
                                                  utils.ReadFile(filename))
         except Exception, err: # pylint: disable-msg=W0703
           return (LUVerifyCluster.ETYPE_ERROR,
                   "Failed to load X509 certificate %s: %s" % (filename, err))
         (errcode, msg) = \
           utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
                                       constants.SSL_CERT_EXPIRATION_ERROR)
         if msg:
           fnamemsg = "While verifying %s: %s" % (filename, msg)
         else:
           fnamemsg = None
         if errcode is None:
           return (None, fnamemsg)
         elif errcode == utils.CERT_WARNING:
           return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
         elif errcode == utils.CERT_ERROR:
           return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
         raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
       class LUVerifyCluster(LogicalUnit):
         """Verifies the cluster status.
         """
         HPATH = "cluster-verify"
         HTYPE = constants.HTYPE_CLUSTER
         _OP_REQP = ["skip_checks", "verbose", "error_codes", "debug_simulate_errors"]
         REQ_BGL = False
         TCLUSTER = "cluster"
         TNODE = "node"
         TINSTANCE = "instance"
         ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
         ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
         EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
         EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
         EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
         EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
         EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
         EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
         ENODEDRBD = (TNODE, "ENODEDRBD")
         ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
         ENODEHOOKS = (TNODE, "ENODEHOOKS")
         ENODEHV = (TNODE, "ENODEHV")
         ENODELVM = (TNODE, "ENODELVM")
         ENODEN1 = (TNODE, "ENODEN1")
         ENODENET = (TNODE, "ENODENET")
         ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
         ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
         ENODERPC = (TNODE, "ENODERPC")
         ENODESSH = (TNODE, "ENODESSH")
         ENODEVERSION = (TNODE, "ENODEVERSION")
         ENODESETUP = (TNODE, "ENODESETUP")
         ENODETIME = (TNODE, "ENODETIME")
         ETYPE_FIELD = "code"
         ETYPE_ERROR = "ERROR"
         ETYPE_WARNING = "WARNING"
         class NodeImage(object):
           """A class representing the logical and physical status of a node.
           @ivar volumes: a structure as returned from
               L{ganeti.backend.GetVolumeList} (runtime)
           @ivar instances: a list of running instances (runtime)
           @ivar pinst: list of configured primary instances (config)
           @ivar sinst: list of configured secondary instances (config)
           @ivar sbp: diction of {secondary-node: list of instances} of all peers
               of this node (config)
           @ivar mfree: free memory, as reported by hypervisor (runtime)
           @ivar dfree: free disk, as reported by the node (runtime)
           @ivar offline: the offline status (config)
           @type rpc_fail: boolean
           @ivar rpc_fail: whether the RPC verify call was successfull (overall,
               not whether the individual keys were correct) (runtime)
           @type lvm_fail: boolean
           @ivar lvm_fail: whether the RPC call didn't return valid LVM data
           @type hyp_fail: boolean
           @ivar hyp_fail: whether the RPC call didn't return the instance list
           @type ghost: boolean
           @ivar ghost: whether this is a known node or not (config)
           """
           def __init__(self, offline=False):
             self.volumes = {}
             self.instances = []
             self.pinst = []
             self.sinst = []
             self.sbp = {}
             self.mfree = 0
             self.dfree = 0
             self.offline = offline
             self.rpc_fail = False
             self.lvm_fail = False
             self.hyp_fail = False
             self.ghost = False
         def ExpandNames(self):
           self.needed_locks = {
             locking.LEVEL_NODE: locking.ALL_SET,
             locking.LEVEL_INSTANCE: locking.ALL_SET,
+          }
           self.share_locks = dict.fromkeys(locking.LEVELS, 1)
         def _Error(self, ecode, item, msg, *args, **kwargs):
           """Format an error message.
           Based on the opcode's error_codes parameter, either format a
           parseable error code, or a simpler error string.
           This must be called only from Exec and functions called from Exec.
           """
           ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
           itype, etxt = ecode
           # first complete the msg
           if args:
             msg = msg % args
           # then format the whole message
           if self.op.error_codes:
             msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
           else:
             if item:
               item = " " + item
             else:
               item = ""
             msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
           # and finally report it via the feedback_fn
           self._feedback_fn("  - %s" % msg)
         def _ErrorIf(self, cond, *args, **kwargs):
           """Log an error message if the passed condition is True.
           """
           cond = bool(cond) or self.op.debug_simulate_errors
           if cond:
             self._Error(*args, **kwargs)
           # do not mark the operation as failed for WARN cases only
           if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
             self.bad = self.bad or cond
         def _VerifyNode(self, ninfo, nresult):
           """Run multiple tests against a node.
           Test list:
             - compares ganeti version
             - checks vg existence and size > 20G
             - checks config file checksum
             - checks ssh to other nodes
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the results from the node
           @rtype: boolean
           @return: whether overall this call was successful (and we can expect
                reasonable values in the respose)
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           # main result, nresult should be a non-empty dict
           test = not nresult or not isinstance(nresult, dict)
           _ErrorIf(test, self.ENODERPC, node,
                         "unable to verify node: no data returned")
           if test:
             return False
           # compares ganeti version
           local_version = constants.PROTOCOL_VERSION
           remote_version = nresult.get("version", None)
           test = not (remote_version and
                       isinstance(remote_version, (list, tuple)) and
                       len(remote_version) == 2)
           _ErrorIf(test, self.ENODERPC, node,
                    "connection to node returned invalid data")
           if test:
             return False
           test = local_version != remote_version[0]
           _ErrorIf(test, self.ENODEVERSION, node,
                    "incompatible protocol versions: master %s,"
                    " node %s", local_version, remote_version[0])
           if test:
             return False
           # node seems compatible, we can actually try to look into its results
           # full package version
           self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
                         self.ENODEVERSION, node,
                         "software version mismatch: master %s, node %s",
                         constants.RELEASE_VERSION, remote_version[1],
                         code=self.ETYPE_WARNING)
           hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
           if isinstance(hyp_result, dict):
             for hv_name, hv_result in hyp_result.iteritems():
               test = hv_result is not None
               _ErrorIf(test, self.ENODEHV, node,
                        "hypervisor %s verify failure: '%s'", hv_name, hv_result)
           test = nresult.get(constants.NV_NODESETUP,
                                  ["Missing NODESETUP results"])
           _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
                    "; ".join(test))
           return True
         def _VerifyNodeTime(self, ninfo, nresult,
                             nvinfo_starttime, nvinfo_endtime):
           """Check the node time.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param nvinfo_starttime: the start time of the RPC call
           @param nvinfo_endtime: the end time of the RPC call
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           ntime = nresult.get(constants.NV_TIME, None)
           try:
             ntime_merged = utils.MergeTime(ntime)
           except (ValueError, TypeError):
             _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
             return
           if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
             ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
           elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
             ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
           else:
             ntime_diff = None
           _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
                    "Node time diverges by at least %s from master node time",
                    ntime_diff)
         def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
           """Check the node time.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param vg_name: the configured VG name
           """
           if vg_name is None:
             return
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           # checks vg existence and size > 20G
           vglist = nresult.get(constants.NV_VGLIST, None)
           test = not vglist
           _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
           if not test:
             vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
                                                   constants.MIN_VG_SIZE)
             _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
           # check pv names
           pvlist = nresult.get(constants.NV_PVLIST, None)
           test = pvlist is None
           _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
           if not test:
             # check that ':' is not present in PV names, since it's a
             # special character for lvcreate (denotes the range of PEs to
             # use on the PV)
             for _, pvname, owner_vg in pvlist:
               test = ":" in pvname
               _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
                        " '%s' of VG '%s'", pvname, owner_vg)
         def _VerifyNodeNetwork(self, ninfo, nresult):
           """Check the node time.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           test = constants.NV_NODELIST not in nresult
           _ErrorIf(test, self.ENODESSH, node,
                    "node hasn't returned node ssh connectivity data")
           if not test:
             if nresult[constants.NV_NODELIST]:
               for a_node, a_msg in nresult[constants.NV_NODELIST].items():
                 _ErrorIf(True, self.ENODESSH, node,
                          "ssh communication with node '%s': %s", a_node, a_msg)
           test = constants.NV_NODENETTEST not in nresult
           _ErrorIf(test, self.ENODENET, node,
                    "node hasn't returned node tcp connectivity data")
           if not test:
             if nresult[constants.NV_NODENETTEST]:
               nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
               for anode in nlist:
                 _ErrorIf(True, self.ENODENET, node,
                          "tcp communication with node '%s': %s",
                          anode, nresult[constants.NV_NODENETTEST][anode])
           test = constants.NV_MASTERIP not in nresult
           _ErrorIf(test, self.ENODENET, node,
                    "node hasn't returned node master IP reachability data")
           if not test:
             if not nresult[constants.NV_MASTERIP]:
               if node == self.master_node:
                 msg = "the master node cannot reach the master IP (not configured?)"
               else:
                 msg = "cannot reach the master IP"
               _ErrorIf(True, self.ENODENET, node, msg)
         def _VerifyInstance(self, instance, instanceconfig, node_image):
           """Verify an instance.
           This function checks to see if the required block devices are
           available on the instance's node.
           """
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           node_current = instanceconfig.primary_node
           node_vol_should = {}
           instanceconfig.MapLVsByNode(node_vol_should)
           for node in node_vol_should:
             n_img = node_image[node]
             if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
               # ignore missing volumes on offline or broken nodes
               continue
             for volume in node_vol_should[node]:
               test = volume not in n_img.volumes
               _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
                        "volume %s missing on node %s", volume, node)
           if instanceconfig.admin_up:
             pri_img = node_image[node_current]
             test = instance not in pri_img.instances and not pri_img.offline
             _ErrorIf(test, self.EINSTANCEDOWN, instance,
                      "instance not running on its primary node %s",
                      node_current)
           for node, n_img in node_image.items():
             if (not node == node_current):
               test = instance in n_img.instances
               _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
                        "instance should not run on node %s", node)
         def _VerifyOrphanVolumes(self, node_vol_should, node_image):
           """Verify if there are any unknown volumes in the cluster.
           The .os, .swap and backup volumes are ignored. All other volumes are
           reported as unknown.
           """
           for node, n_img in node_image.items():
             if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
               # skip non-healthy nodes
               continue
             for volume in n_img.volumes:
               test = (node not in node_vol_should or
                       volume not in node_vol_should[node])
               self._ErrorIf(test, self.ENODEORPHANLV, node,
                             "volume %s is unknown", volume)
         def _VerifyOrphanInstances(self, instancelist, node_image):
           """Verify the list of running instances.
           This checks what instances are running but unknown to the cluster.
           """
           for node, n_img in node_image.items():
             for o_inst in n_img.instances:
               test = o_inst not in instancelist
               self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
                             "instance %s on node %s should not exist", o_inst, node)
         def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
           """Verify N+1 Memory Resilience.
           Check that if one single node dies we can still start all the
           instances it was primary for.
           """
           for node, n_img in node_image.items():
             # This code checks that every node which is now listed as
             # secondary has enough memory to host all instances it is
             # supposed to should a single other node in the cluster fail.
             # FIXME: not ready for failover to an arbitrary node
             # FIXME: does not support file-backed instances
             # WARNING: we currently take into account down instances as well
             # as up ones, considering that even if they're down someone
             # might want to start them even in the event of a node failure.
             for prinode, instances in n_img.sbp.items():
               needed_mem = 0
               for instance in instances:
                 bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
                 if bep[constants.BE_AUTO_BALANCE]:
                   needed_mem += bep[constants.BE_MEMORY]
               test = n_img.mfree < needed_mem
               self._ErrorIf(test, self.ENODEN1, node,
                             "not enough memory on to accommodate"
                             " failovers should peer node %s fail", prinode)
         def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
                              master_files):
           """Verifies and computes the node required file checksums.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param file_list: required list of files
           @param local_cksum: dictionary of local files and their checksums
           @param master_files: list of files that only masters should have
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           remote_cksum = nresult.get(constants.NV_FILELIST, None)
           test = not isinstance(remote_cksum, dict)
           _ErrorIf(test, self.ENODEFILECHECK, node,
                    "node hasn't returned file checksum data")
           if test:
             return
           for file_name in file_list:
             node_is_mc = ninfo.master_candidate
             must_have = (file_name not in master_files) or node_is_mc
             # missing
             test1 = file_name not in remote_cksum
             # invalid checksum
             test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
             # existing and good
             test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
             _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
                      "file '%s' missing", file_name)
             _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
                      "file '%s' has wrong checksum", file_name)
             # not candidate and this is not a must-have file
             _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
                      "file '%s' should not exist on non master"
                      " candidates (and the file is outdated)", file_name)
             # all good, except non-master/non-must have combination
             _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
                      "file '%s' should not exist"
                      " on non master candidates", file_name)
         def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_map):
           """Verifies and the node DRBD status.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param instanceinfo: the dict of instances
           @param drbd_map: the DRBD map as returned by
               L{ganeti.config.ConfigWriter.ComputeDRBDMap}
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           # compute the DRBD minors
           node_drbd = {}
           for minor, instance in drbd_map[node].items():
             test = instance not in instanceinfo
             _ErrorIf(test, self.ECLUSTERCFG, None,
                      "ghost instance '%s' in temporary DRBD map", instance)
               # ghost instance should not be running, but otherwise we
               # don't give double warnings (both ghost instance and
               # unallocated minor in use)
             if test:
               node_drbd[minor] = (instance, False)
             else:
               instance = instanceinfo[instance]
               node_drbd[minor] = (instance.name, instance.admin_up)
           # and now check them
           used_minors = nresult.get(constants.NV_DRBDLIST, [])
           test = not isinstance(used_minors, (tuple, list))
           _ErrorIf(test, self.ENODEDRBD, node,
                    "cannot parse drbd status file: %s", str(used_minors))
           if test:
             # we cannot check drbd status
             return
           for minor, (iname, must_exist) in node_drbd.items():
             test = minor not in used_minors and must_exist
             _ErrorIf(test, self.ENODEDRBD, node,
                      "drbd minor %d of instance %s is not active", minor, iname)
           for minor in used_minors:
             test = minor not in node_drbd
             _ErrorIf(test, self.ENODEDRBD, node,
                      "unallocated drbd minor %d is in use", minor)
         def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
           """Verifies and updates the node volume data.
           This function will update a L{NodeImage}'s internal structures
           with data from the remote call.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param nimg: the node image object
           @param vg_name: the configured VG name
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           nimg.lvm_fail = True
           lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
           if vg_name is None:
             pass
           elif isinstance(lvdata, basestring):
             _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
                      utils.SafeEncode(lvdata))
           elif not isinstance(lvdata, dict):
             _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
           else:
             nimg.volumes = lvdata
             nimg.lvm_fail = False
         def _UpdateNodeInstances(self, ninfo, nresult, nimg):
           """Verifies and updates the node instance list.
           If the listing was successful, then updates this node's instance
           list. Otherwise, it marks the RPC call as failed for the instance
           list key.
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param nimg: the node image object
           """
           idata = nresult.get(constants.NV_INSTANCELIST, None)
           test = not isinstance(idata, list)
           self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
                         " (instancelist): %s", utils.SafeEncode(str(idata)))
           if test:
             nimg.hyp_fail = True
           else:
             nimg.instances = idata
         def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
           """Verifies and computes a node information map
           @type ninfo: L{objects.Node}
           @param ninfo: the node to check
           @param nresult: the remote results for the node
           @param nimg: the node image object
           @param vg_name: the configured VG name
           """
           node = ninfo.name
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           # try to read free memory (from the hypervisor)
           hv_info = nresult.get(constants.NV_HVINFO, None)
           test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
           _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
           if not test:
             try:
               nimg.mfree = int(hv_info["memory_free"])
             except (ValueError, TypeError):
               _ErrorIf(True, self.ENODERPC, node,
                        "node returned invalid nodeinfo, check hypervisor")
           # FIXME: devise a free space model for file based instances as well
           if vg_name is not None:
             test = (constants.NV_VGLIST not in nresult or
                     vg_name not in nresult[constants.NV_VGLIST])
             _ErrorIf(test, self.ENODELVM, node,
                      "node didn't return data for the volume group '%s'"
                      " - it is either missing or broken", vg_name)
             if not test:
               try:
                 nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
               except (ValueError, TypeError):
                 _ErrorIf(True, self.ENODERPC, node,
                          "node returned invalid LVM info, check LVM status")
         def CheckPrereq(self):
           """Check prerequisites.
           Transform the list of checks we're going to skip into a set and check that
           all its members are valid.
           """
           self.skip_set = frozenset(self.op.skip_checks)
           if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
             raise errors.OpPrereqError("Invalid checks to be skipped specified",
                                        errors.ECODE_INVAL)
         def BuildHooksEnv(self):
           """Build hooks env.
           Cluster-Verify hooks just ran in the post phase and their failure makes
           the output be logged in the verify output and the verification to fail.
           """
           all_nodes = self.cfg.GetNodeList()
           env = {
             "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
+            }
           for node in self.cfg.GetAllNodesInfo().values():
             env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
           return env, [], all_nodes
         def Exec(self, feedback_fn):
           """Verify integrity of cluster, performing various test on nodes.
           """
           self.bad = False
           _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
           verbose = self.op.verbose
           self._feedback_fn = feedback_fn
           feedback_fn("* Verifying global settings")
           for msg in self.cfg.VerifyConfig():
             _ErrorIf(True, self.ECLUSTERCFG, None, msg)
           # Check the cluster certificates
           for cert_filename in constants.ALL_CERT_FILES:
             (errcode, msg) = _VerifyCertificate(cert_filename)
             _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
           vg_name = self.cfg.GetVGName()
           hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
           cluster = self.cfg.GetClusterInfo()
           nodelist = utils.NiceSort(self.cfg.GetNodeList())
           nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
           instancelist = utils.NiceSort(self.cfg.GetInstanceList())
           instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
                               for iname in instancelist)
           i_non_redundant = [] # Non redundant instances
           i_non_a_balanced = [] # Non auto-balanced instances
           n_offline = 0 # Count of offline nodes
           n_drained = 0 # Count of nodes being drained
           node_vol_should = {}
           # FIXME: verify OS list
           # do local checksums
           master_files = [constants.CLUSTER_CONF_FILE]
           master_node = self.master_node = self.cfg.GetMasterNode()
           master_ip = self.cfg.GetMasterIP()
           file_names = ssconf.SimpleStore().GetFileList()
           file_names.extend(constants.ALL_CERT_FILES)
           file_names.extend(master_files)
           if cluster.modify_etc_hosts:
             file_names.append(constants.ETC_HOSTS)
           local_checksums = utils.FingerprintFiles(file_names)
           feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
           node_verify_param = {
             constants.NV_FILELIST: file_names,
             constants.NV_NODELIST: [node.name for node in nodeinfo
                                     if not node.offline],
             constants.NV_HYPERVISOR: hypervisors,
             constants.NV_NODENETTEST: [(node.name, node.primary_ip,
                                         node.secondary_ip) for node in nodeinfo
                                        if not node.offline],
             constants.NV_INSTANCELIST: hypervisors,
             constants.NV_VERSION: None,
             constants.NV_HVINFO: self.cfg.GetHypervisorType(),
             constants.NV_NODESETUP: None,
             constants.NV_TIME: None,
             constants.NV_MASTERIP: (master_node, master_ip),
+            }
           if vg_name is not None:
             node_verify_param[constants.NV_VGLIST] = None
             node_verify_param[constants.NV_LVLIST] = vg_name
             node_verify_param[constants.NV_PVLIST] = [vg_name]
             node_verify_param[constants.NV_DRBDLIST] = None
           # Build our expected cluster state
           node_image = dict((node.name, self.NodeImage(offline=node.offline))
                             for node in nodeinfo)
           for instance in instancelist:
             inst_config = instanceinfo[instance]
             for nname in inst_config.all_nodes:
               if nname not in node_image:
                 # ghost node
                 gnode = self.NodeImage()
                 gnode.ghost = True
                 node_image[nname] = gnode
             inst_config.MapLVsByNode(node_vol_should)
             pnode = inst_config.primary_node
             node_image[pnode].pinst.append(instance)
             for snode in inst_config.secondary_nodes:
               nimg = node_image[snode]
               nimg.sinst.append(instance)
               if pnode not in nimg.sbp:
                 nimg.sbp[pnode] = []
               nimg.sbp[pnode].append(instance)
           # At this point, we have the in-memory data structures complete,
           # except for the runtime information, which we'll gather next
           # Due to the way our RPC system works, exact response times cannot be
           # guaranteed (e.g. a broken node could run into a timeout). By keeping the
           # time before and after executing the request, we can at least have a time
           # window.
           nvinfo_starttime = time.time()
           all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
                                                  self.cfg.GetClusterName())
           nvinfo_endtime = time.time()
           all_drbd_map = self.cfg.ComputeDRBDMap()
           feedback_fn("* Verifying node status")
           for node_i in nodeinfo:
             node = node_i.name
             nimg = node_image[node]
             if node_i.offline:
               if verbose:
                 feedback_fn("* Skipping offline node %s" % (node,))
               n_offline += 1
               continue
             if node == master_node:
               ntype = "master"
             elif node_i.master_candidate:
               ntype = "master candidate"
             elif node_i.drained:
               ntype = "drained"
               n_drained += 1
             else:
               ntype = "regular"
             if verbose:
               feedback_fn("* Verifying node %s (%s)" % (node, ntype))
             msg = all_nvinfo[node].fail_msg
             _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
             if msg:
               nimg.rpc_fail = True
               continue
             nresult = all_nvinfo[node].payload
             nimg.call_ok = self._VerifyNode(node_i, nresult)
             self._VerifyNodeNetwork(node_i, nresult)
             self._VerifyNodeLVM(node_i, nresult, vg_name)
             self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
                                   master_files)
             self._VerifyNodeDrbd(node_i, nresult, instanceinfo, all_drbd_map)
             self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
             self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
             self._UpdateNodeInstances(node_i, nresult, nimg)
             self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
           feedback_fn("* Verifying instance status")
           for instance in instancelist:
             if verbose:
               feedback_fn("* Verifying instance %s" % instance)
             inst_config = instanceinfo[instance]
             self._VerifyInstance(instance, inst_config, node_image)
             inst_nodes_offline = []
             pnode = inst_config.primary_node
             pnode_img = node_image[pnode]
             _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
                      self.ENODERPC, pnode, "instance %s, connection to"
                      " primary node failed", instance)
             if pnode_img.offline:
               inst_nodes_offline.append(pnode)
             # If the instance is non-redundant we cannot survive losing its primary
             # node, so we are not N+1 compliant. On the other hand we have no disk
             # templates with more than one secondary so that situation is not well
             # supported either.
             # FIXME: does not support file-backed instances
             if not inst_config.secondary_nodes:
               i_non_redundant.append(instance)
             _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
                      instance, "instance has multiple secondary nodes: %s",
                      utils.CommaJoin(inst_config.secondary_nodes),
                      code=self.ETYPE_WARNING)
             if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
               i_non_a_balanced.append(instance)
             for snode in inst_config.secondary_nodes:
               s_img = node_image[snode]
               _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
                        "instance %s, connection to secondary node failed", instance)
               if s_img.offline:
                 inst_nodes_offline.append(snode)
             # warn that the instance lives on offline nodes
             _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
                      "instance lives on offline node(s) %s",
                      utils.CommaJoin(inst_nodes_offline))
             # ... or ghost nodes
             for node in inst_config.all_nodes:
               _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
                        "instance lives on ghost node %s", node)
           feedback_fn("* Verifying orphan volumes")
           self._VerifyOrphanVolumes(node_vol_should, node_image)
           feedback_fn("* Verifying orphan instances")
           self._VerifyOrphanInstances(instancelist, node_image)
           if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
             feedback_fn("* Verifying N+1 Memory redundancy")
             self._VerifyNPlusOneMemory(node_image, instanceinfo)
           feedback_fn("* Other Notes")
           if i_non_redundant:
             feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
                         % len(i_non_redundant))
           if i_non_a_balanced:
             feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
                         % len(i_non_a_balanced))
           if n_offline:
             feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
           if n_drained:
             feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
           return not self.bad
         def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
           """Analyze the post-hooks' result
           This method analyses the hook result, handles it, and sends some
           nicely-formatted feedback back to the user.
           @param phase: one of L{constants.HOOKS_PHASE_POST} or
               L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
           @param hooks_results: the results of the multi-node hooks rpc call
           @param feedback_fn: function used send feedback back to the caller
           @param lu_result: previous Exec result
           @return: the new Exec result, based on the previous result
               and hook results
           """
           # We only really run POST phase hooks, and are only interested in
           # their results
           if phase == constants.HOOKS_PHASE_POST:
             # Used to change hooks' output to proper indentation
             indent_re = re.compile('^', re.M)
             feedback_fn("* Hooks Results")
             assert hooks_results, "invalid result from hooks"
             for node_name in hooks_results:
               res = hooks_results[node_name]
               msg = res.fail_msg
               test = msg and not res.offline
               self._ErrorIf(test, self.ENODEHOOKS, node_name,
                             "Communication failure in hooks execution: %s", msg)
               if res.offline or msg:
                 # No need to investigate payload if node is offline or gave an error.
                 # override manually lu_result here as _ErrorIf only
                 # overrides self.bad
                 lu_result = 1
                 continue
               for script, hkr, output in res.payload:
                 test = hkr == constants.HKR_FAIL
                 self._ErrorIf(test, self.ENODEHOOKS, node_name,
                               "Script %s failed, output:", script)
                 if test:
                   output = indent_re.sub('      ', output)
                   feedback_fn("%s" % output)
                   lu_result = 0
             return lu_result
       class LUVerifyDisks(NoHooksLU):
         """Verifies the cluster disks status.
         """
         _OP_REQP = []
         REQ_BGL = False
         def ExpandNames(self):
           self.needed_locks = {
             locking.LEVEL_NODE: locking.ALL_SET,
             locking.LEVEL_INSTANCE: locking.ALL_SET,
+          }
           self.share_locks = dict.fromkeys(locking.LEVELS, 1)
         def CheckPrereq(self):
           """Check prerequisites.
           This has no prerequisites.
           """
           pass
         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
           """
           result = res_nodes, res_instances, res_missing = {}, [], {}
           vg_name = self.cfg.GetVGName()
           nodes = utils.NiceSort(self.cfg.GetNodeList())
           instances = [self.cfg.GetInstanceInfo(name)
                        for name in self.cfg.GetInstanceList()]
           nv_dict = {}
           for inst in instances:
             inst_lvs = {}
             if (not inst.admin_up or
                 inst.disk_template not in constants.DTS_NET_MIRROR):
               continue
             inst.MapLVsByNode(inst_lvs)
             # transform { iname: {node: [vol,],},} to {(node, vol): iname}
             for node, vol_list in inst_lvs.iteritems():
               for vol in vol_list:
                 nv_dict[(node, vol)] = inst
           if not nv_dict:
             return result
           node_lvs = self.rpc.call_lv_list(nodes, vg_name)
           for node in nodes:
             # node_volume
             node_res = node_lvs[node]
             if node_res.offline:
               continue
             msg = node_res.fail_msg
             if msg:
               logging.warning("Error enumerating LVs on node %s: %s", node, msg)
               res_nodes[node] = msg
               continue
             lvs = node_res.payload
             for lv_name, (_, _, lv_online) in lvs.items():
               inst = nv_dict.pop((node, lv_name), None)
               if (not lv_online and inst is not None
                   and inst.name not in res_instances):
                 res_instances.append(inst.name)
           # any leftover items in nv_dict are missing LVs, let's arrange the
           # data better
           for key, inst in nv_dict.iteritems():
             if inst.name not in res_missing:
               res_missing[inst.name] = []
             res_missing[inst.name].append(key)
           return result
       class LURepairDiskSizes(NoHooksLU):
         """Verifies the cluster disks sizes.
         """
         _OP_REQP = ["instances"]
         REQ_BGL = False
         def ExpandNames(self):
           if not isinstance(self.op.instances, list):
             raise errors.OpPrereqError("Invalid argument type 'instances'",
                                        errors.ECODE_INVAL)
           if self.op.instances:
             self.wanted_names = []
             for name in self.op.instances:
               full_name = _ExpandInstanceName(self.cfg, name)
               self.wanted_names.append(full_name)
             self.needed_locks = {
               locking.LEVEL_NODE: [],
               locking.LEVEL_INSTANCE: self.wanted_names,
+              }
             self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
           else:
             self.wanted_names = None
             self.needed_locks = {
               locking.LEVEL_NODE: locking.ALL_SET,
               locking.LEVEL_INSTANCE: locking.ALL_SET,
+              }
           self.share_locks = dict(((i, 1) for i in locking.LEVELS))
         def DeclareLocks(self, level):
           if level == locking.LEVEL_NODE and self.wanted_names is not None:
             self._LockInstancesNodes(primary_only=True)
         def CheckPrereq(self):
           """Check prerequisites.
           This only checks the optional instance list against the existing names.
           """
           if self.wanted_names is None:
             self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
           self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
                                    in self.wanted_names]
         def _EnsureChildSizes(self, disk):
           """Ensure children of the disk have the needed disk size.
           This is valid mainly for DRBD8 and fixes an issue where the
           children have smaller disk size.
           @param disk: an L{ganeti.objects.Disk} object
           """
           if disk.dev_type == constants.LD_DRBD8:
             assert disk.children, "Empty children for DRBD8?"
             fchild = disk.children[0]
             mismatch = fchild.size < disk.size
             if mismatch:
               self.LogInfo("Child disk has size %d, parent %d, fixing",
                            fchild.size, disk.size)
               fchild.size = disk.size
             # and we recurse on this child only, not on the metadev
             return self._EnsureChildSizes(fchild) or mismatch
           else:
             return False
         def Exec(self, feedback_fn):
           """Verify the size of cluster disks.
           """
           # TODO: check child disks too
           # TODO: check differences in size between primary/secondary nodes
           per_node_disks = {}
           for instance in self.wanted_instances:
             pnode = instance.primary_node
             if pnode not in per_node_disks:
               per_node_disks[pnode] = []
             for idx, disk in enumerate(instance.disks):
               per_node_disks[pnode].append((instance, idx, disk))
           changed = []
           for node, dskl in per_node_disks.items():
             newl = [v[2].Copy() for v in dskl]
             for dsk in newl:
               self.cfg.SetDiskID(dsk, node)
             result = self.rpc.call_blockdev_getsizes(node, newl)
             if result.fail_msg:
               self.LogWarning("Failure in blockdev_getsizes call to node"
                               " %s, ignoring", node)
               continue
             if len(result.data) != len(dskl):
               self.LogWarning("Invalid result from node %s, ignoring node results",
                               node)
               continue
             for ((instance, idx, disk), size) in zip(dskl, result.data):
               if size is None:
                 self.LogWarning("Disk %d of instance %s did not return size"
                                 " information, ignoring", idx, instance.name)
                 continue
               if not isinstance(size, (int, long)):
                 self.LogWarning("Disk %d of instance %s did not return valid"
                                 " size information, ignoring", idx, instance.name)
                 continue
               size = size >> 20
               if size != disk.size:
                 self.LogInfo("Disk %d of instance %s has mismatched size,"
                              " correcting: recorded %d, actual %d", idx,
                              instance.name, disk.size, size)
                 disk.size = size
                 self.cfg.Update(instance, feedback_fn)
                 changed.append((instance.name, idx, size))
               if self._EnsureChildSizes(disk):
                 self.cfg.Update(instance, feedback_fn)
                 changed.append((instance.name, idx, disk.size))
           return changed
       class LURenameCluster(LogicalUnit):
         """Rename the cluster.
         """
         HPATH = "cluster-rename"
         HTYPE = constants.HTYPE_CLUSTER
         _OP_REQP = ["name"]
         def BuildHooksEnv(self):
           """Build hooks env.
           """
           env = {
             "OP_TARGET": self.cfg.GetClusterName(),
             "NEW_NAME": self.op.name,
+            }
           mn = self.cfg.GetMasterNode()
           all_nodes = self.cfg.GetNodeList()
           return env, [mn], all_nodes
         def CheckPrereq(self):
           """Verify that the passed name is a valid one.
           """
           hostname = utils.GetHostInfo(self.op.name)
           new_name = hostname.name
           self.ip = new_ip = hostname.ip
           old_name = self.cfg.GetClusterName()
           old_ip = self.cfg.GetMasterIP()
           if new_name == old_name and new_ip == old_ip:
             raise errors.OpPrereqError("Neither the name nor the IP address of the"
                                        " cluster has changed",
                                        errors.ECODE_INVAL)
           if new_ip != old_ip:
             if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
               raise errors.OpPrereqError("The given cluster IP address (%s) is"
                                          " reachable on the network. Aborting." %
                                          new_ip, errors.ECODE_NOTUNIQUE)
           self.op.name = new_name
         def Exec(self, feedback_fn):
           """Rename the cluster.
           """
           clustername = self.op.name
           ip = self.ip
           # shutdown the master IP
           master = self.cfg.GetMasterNode()
           result = self.rpc.call_node_stop_master(master, False)
           result.Raise("Could not disable the master role")
           try:
             cluster = self.cfg.GetClusterInfo()
             cluster.cluster_name = clustername
             cluster.master_ip = ip
             self.cfg.Update(cluster, feedback_fn)
             # update the known hosts file
             ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
             node_list = self.cfg.GetNodeList()
             try:
               node_list.remove(master)
             except ValueError:
               pass
             result = self.rpc.call_upload_file(node_list,
                                                constants.SSH_KNOWN_HOSTS_FILE)
             for to_node, to_result in result.iteritems():
               msg = to_result.fail_msg
               if msg:
                 msg = ("Copy of file %s to node %s failed: %s" %
                        (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
                 self.proc.LogWarning(msg)
           finally:
             result = self.rpc.call_node_start_master(master, False, False)
             msg = result.fail_msg
             if msg:
               self.LogWarning("Could not re-enable the master role on"
                               " the master, please restart manually: %s", msg)
       def _RecursiveCheckIfLVMBased(disk):
         """Check if the given disk or its children are lvm-based.
         @type disk: L{objects.Disk}
         @param disk: the disk to check
         @rtype: boolean
         @return: boolean indicating whether a LD_LV dev_type was found or not
         """
         if disk.children:
           for chdisk in disk.children:
             if _RecursiveCheckIfLVMBased(chdisk):
               return True
         return disk.dev_type == constants.LD_LV
       class LUSetClusterParams(LogicalUnit):
         """Change the parameters of the cluster.
         """
         HPATH = "cluster-modify"
         HTYPE = constants.HTYPE_CLUSTER
         _OP_REQP = []
         REQ_BGL = False
         def CheckArguments(self):
           """Check parameters
           """
           for attr in ["candidate_pool_size",
                        "uid_pool", "add_uids", "remove_uids"]:
             if not hasattr(self.op, attr):
               setattr(self.op, attr, None)
           if self.op.candidate_pool_size is not None:
             try:
               self.op.candidate_pool_size = int(self.op.candidate_pool_size)
             except (ValueError, TypeError), err:
               raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
                                          str(err), errors.ECODE_INVAL)
             if self.op.candidate_pool_size < 1:
               raise errors.OpPrereqError("At least one master candidate needed",
                                          errors.ECODE_INVAL)
           _CheckBooleanOpField(self.op, "maintain_node_health")
           if self.op.uid_pool:
             uidpool.CheckUidPool(self.op.uid_pool)
           if self.op.add_uids:
             uidpool.CheckUidPool(self.op.add_uids)
           if self.op.remove_uids:
             uidpool.CheckUidPool(self.op.remove_uids)
         def ExpandNames(self):
           # FIXME: in the future maybe other cluster params won't require checking on
           # all nodes to be modified.
           self.needed_locks = {
             locking.LEVEL_NODE: locking.ALL_SET,
+          }
           self.share_locks[locking.LEVEL_NODE] = 1
         def BuildHooksEnv(self):
           """Build hooks env.
           """
           env = {
             "OP_TARGET": self.cfg.GetClusterName(),
             "NEW_VG_NAME": self.op.vg_name,
+            }
           mn = self.cfg.GetMasterNode()
           return env, [mn], [mn]
         def CheckPrereq(self):
           """Check prerequisites.
           This checks whether the given params don't conflict and
           if the given volume group is valid.
           """
           if self.op.vg_name is not None and not self.op.vg_name:
             instances = self.cfg.GetAllInstancesInfo().values()
             for inst in instances:
               for disk in inst.disks:
                 if _RecursiveCheckIfLVMBased(disk):
                   raise errors.OpPrereqError("Cannot disable lvm storage while"
                                              " lvm-based instances exist",
                                              errors.ECODE_INVAL)
           node_list = self.acquired_locks[locking.LEVEL_NODE]
           # if vg_name not None, checks given volume group on all nodes
           if self.op.vg_name:
             vglist = self.rpc.call_vg_list(node_list)
             for node in node_list:
               msg = vglist[node].fail_msg
               if msg:
                 # ignoring down node
                 self.LogWarning("Error while gathering data on node %s"
                                 " (ignoring node): %s", node, msg)
                 continue
               vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
                                                     self.op.vg_name,
                                                     constants.MIN_VG_SIZE)
               if vgstatus:
                 raise errors.OpPrereqError("Error on node '%s': %s" %
                                            (node, vgstatus), errors.ECODE_ENVIRON)
           self.cluster = cluster = self.cfg.GetClusterInfo()
           # validate params changes
           if self.op.beparams:
             utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
             self.new_beparams = objects.FillDict(
               cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
           if self.op.nicparams:
             utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
             self.new_nicparams = objects.FillDict(
               cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
             objects.NIC.CheckParameterSyntax(self.new_nicparams)
             nic_errors = []
             # check all instances for consistency
             for instance in self.cfg.GetAllInstancesInfo().values():
               for nic_idx, nic in enumerate(instance.nics):
                 params_copy = copy.deepcopy(nic.nicparams)
                 params_filled = objects.FillDict(self.new_nicparams, params_copy)
                 # check parameter syntax
                 try:
                   objects.NIC.CheckParameterSyntax(params_filled)
                 except errors.ConfigurationError, err:
                   nic_errors.append("Instance %s, nic/%d: %s" %
                                     (instance.name, nic_idx, err))
                 # if we're moving instances to routed, check that they have an ip
                 target_mode = params_filled[constants.NIC_MODE]
                 if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
                   nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
                                     (instance.name, nic_idx))
             if nic_errors:
               raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
                                          "\n".join(nic_errors))
           # hypervisor list/parameters
           self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
           if self.op.hvparams:
             if not isinstance(self.op.hvparams, dict):
               raise errors.OpPrereqError("Invalid 'hvparams' parameter on input",
                                          errors.ECODE_INVAL)
             for hv_name, hv_dict in self.op.hvparams.items():
               if hv_name not in self.new_hvparams:
                 self.new_hvparams[hv_name] = hv_dict
               else:
                 self.new_hvparams[hv_name].update(hv_dict)
           # os hypervisor parameters
           self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
           if self.op.os_hvp:
             if not isinstance(self.op.os_hvp, dict):
               raise errors.OpPrereqError("Invalid 'os_hvp' parameter on input",
                                          errors.ECODE_INVAL)
             for os_name, hvs in self.op.os_hvp.items():
               if not isinstance(hvs, dict):
                 raise errors.OpPrereqError(("Invalid 'os_hvp' parameter on"
                                             " input"), errors.ECODE_INVAL)
               if os_name not in self.new_os_hvp:
                 self.new_os_hvp[os_name] = hvs
               else:
                 for hv_name, hv_dict in hvs.items():
                   if hv_name not in self.new_os_hvp[os_name]:
                     self.new_os_hvp[os_name][hv_name] = hv_dict
                   else:
                     self.new_os_hvp[os_name][hv_name].update(hv_dict)
           # changes to the hypervisor list
           if self.op.enabled_hypervisors is not None:
             self.hv_list = self.op.enabled_hypervisors
             if not self.hv_list:
               raise errors.OpPrereqError("Enabled hypervisors list must contain at"
                                          " least one member",
                                          errors.ECODE_INVAL)
             invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
             if invalid_hvs:
               raise errors.OpPrereqError("Enabled hypervisors contains invalid"
                                          " entries: %s" %
                                          utils.CommaJoin(invalid_hvs),
                                          errors.ECODE_INVAL)
             for hv in self.hv_list:
               # if the hypervisor doesn't already exist in the cluster
               # hvparams, we initialize it to empty, and then (in both
               # cases) we make sure to fill the defaults, as we might not
               # have a complete defaults list if the hypervisor wasn't
               # enabled before
               if hv not in new_hvp:
                 new_hvp[hv] = {}
               new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
               utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
           else:
             self.hv_list = cluster.enabled_hypervisors
           if self.op.hvparams or self.op.enabled_hypervisors is not None:
             # either the enabled list has changed, or the parameters have, validate
             for hv_name, hv_params in self.new_hvparams.items():
               if ((self.op.hvparams and hv_name in self.op.hvparams) or
                   (self.op.enabled_hypervisors and
                    hv_name in self.op.enabled_hypervisors)):
                 # either this is a new hypervisor, or its parameters have changed
                 hv_class = hypervisor.GetHypervisor(hv_name)
                 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
                 hv_class.CheckParameterSyntax(hv_params)
                 _CheckHVParams(self, node_list, hv_name, hv_params)
           if self.op.os_hvp:
             # no need to check any newly-enabled hypervisors, since the
             # defaults have already been checked in the above code-block
             for os_name, os_hvp in self.new_os_hvp.items():
               for hv_name, hv_params in os_hvp.items():
                 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
                 # we need to fill in the new os_hvp on top of the actual hv_p
                 cluster_defaults = self.new_hvparams.get(hv_name, {})
                 new_osp = objects.FillDict(cluster_defaults, hv_params)
                 hv_class = hypervisor.GetHypervisor(hv_name)
                 hv_class.CheckParameterSyntax(new_osp)
                 _CheckHVParams(self, node_list, hv_name, new_osp)
         def Exec(self, feedback_fn):
           """Change the parameters of the cluster.
           """
           if self.op.vg_name is not None:
             new_volume = self.op.vg_name
             if not new_volume:
               new_volume = None
             if new_volume != self.cfg.GetVGName():
               self.cfg.SetVGName(new_volume)
             else:
               feedback_fn("Cluster LVM configuration already in desired"
                           " state, not changing")
           if self.op.hvparams:
             self.cluster.hvparams = self.new_hvparams
           if self.op.os_hvp:
             self.cluster.os_hvp = self.new_os_hvp
           if self.op.enabled_hypervisors is not None:
             self.cluster.hvparams = self.new_hvparams
             self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
           if self.op.beparams:
             self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
           if self.op.nicparams:
             self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
           if self.op.candidate_pool_size is not None:
             self.cluster.candidate_pool_size = self.op.candidate_pool_size
             # we need to update the pool size here, otherwise the save will fail
             _AdjustCandidatePool(self, [])
           if self.op.maintain_node_health is not None:
             self.cluster.maintain_node_health = self.op.maintain_node_health
           if self.op.add_uids is not None:
             uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
           if self.op.remove_uids is not None:
             uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
           if self.op.uid_pool is not None:
             self.cluster.uid_pool = self.op.uid_pool
           self.cfg.Update(self.cluster, feedback_fn)
       def _RedistributeAncillaryFiles(lu, additional_nodes=None):
         """Distribute additional files which are part of the cluster configuration.
         ConfigWriter takes care of distributing the config and ssconf files, but
         there are more files which should be distributed to all nodes. This function
         makes sure those are copied.
         @param lu: calling logical unit
         @param additional_nodes: list of nodes not in the config to distribute to
         """
         # 1. Gather target nodes
         myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
         dist_nodes = lu.cfg.GetOnlineNodeList()
         if additional_nodes is not None:
           dist_nodes.extend(additional_nodes)
         if myself.name in dist_nodes:
           dist_nodes.remove(myself.name)
         # 2. Gather files to distribute
         dist_files = set([constants.ETC_HOSTS,
                           constants.SSH_KNOWN_HOSTS_FILE,
                           constants.RAPI_CERT_FILE,
                           constants.RAPI_USERS_FILE,
                           constants.CONFD_HMAC_KEY,
                           constants.CLUSTER_DOMAIN_SECRET_FILE,
                          ])
         enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
         for hv_name in enabled_hypervisors:
           hv_class = hypervisor.GetHypervisor(hv_name)
           dist_files.update(hv_class.GetAncillaryFiles())
         # 3. Perform the files upload
         for fname in dist_files:
           if os.path.exists(fname):
             result = lu.rpc.call_upload_file(dist_nodes, fname)
             for to_node, to_result in result.items():
               msg = to_result.fail_msg
               if msg:
                 msg = ("Copy of file %s to node %s failed: %s" %
                        (fname, to_node, msg))
                 lu.proc.LogWarning(msg)
       class LURedistributeConfig(NoHooksLU):
         """Force the redistribution of cluster configuration.
         This is a very simple LU.
         """
         _OP_REQP = []
         REQ_BGL = False
         def ExpandNames(self):
           self.needed_locks = {
             locking.LEVEL_NODE: locking.ALL_SET,
+          }
           self.share_locks[locking.LEVEL_NODE] = 1
         def CheckPrereq(self):
           """Check prerequisites.
           """
         def Exec(self, feedback_fn):
           """Redistribute the configuration.
           """
           self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
           _RedistributeAncillaryFiles(self)
       def _WaitForSync(lu, instance, disks=None, oneshot=False):
         """Sleep and poll for an instance's disk to sync.
         """
         if not instance.disks or disks is not None and not disks:
           return True
         disks = _ExpandCheckDisks(instance, disks)
         if not oneshot:
           lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
         node = instance.primary_node
         for dev in disks:
           lu.cfg.SetDiskID(dev, node)
         # TODO: Convert to utils.Retry
         retries = 0
         degr_retries = 10 # in seconds, as we sleep 1 second each time
         while True:
           max_time = 0
           done = True
           cumul_degraded = False
           rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
           msg = rstats.fail_msg
           if msg:
             lu.LogWarning("Can't get any data from node %s: %s", node, msg)
             retries += 1
             if retries >= 10:
               raise errors.RemoteError("Can't contact node %s for mirror data,"
                                        " aborting." % node)
             time.sleep(6)
             continue
           rstats = rstats.payload
           retries = 0
           for i, mstat in enumerate(rstats):
             if mstat is None:
               lu.LogWarning("Can't compute data for node %s/%s",
                                  node, disks[i].iv_name)
               continue
             cumul_degraded = (cumul_degraded or
                               (mstat.is_degraded and mstat.sync_percent is None))
             if mstat.sync_percent is not None:
               done = False
               if mstat.estimated_time is not None:
                 rem_time = ("%s remaining (estimated)" %
                             utils.FormatSeconds(mstat.estimated_time))
                 max_time = mstat.estimated_time
               else:
                 rem_time = "no time estimate"
               lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
                               (disks[i].iv_name, mstat.sync_percent, rem_time))
           # if we're done but degraded, let's do a few small retries, to
           # make sure we see a stable and not transient situation; therefore
           # we force restart of the loop
           if (done or oneshot) and cumul_degraded and degr_retries > 0:
             logging.info("Degraded disks found, %d retries left", degr_retries)
             degr_retries -= 1
             time.sleep(1)
             continue
           if done or oneshot:
             break
           time.sleep(min(60, max_time))
         if done:
           lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
         return not cumul_degraded
       def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
         """Check that mirrors are not degraded.
         The ldisk parameter, if True, will change the test from the
         is_degraded attribute (which represents overall non-ok status for
         the device(s)) to the ldisk (representing the local storage status).
         """
         lu.cfg.SetDiskID(dev, node)
         result = True
         if on_primary or dev.AssembleOnSecondary():
           rstats = lu.rpc.call_blockdev_find(node, dev)
           msg = rstats.fail_msg
           if msg:
             lu.LogWarning("Can't find disk on node %s: %s", node, msg)
             result = False
           elif not rstats.payload:
             lu.LogWarning("Can't find disk on node %s", node)
             result = False
           else:
             if ldisk:
               result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
             else:
               result = result and not rstats.payload.is_degraded
         if dev.children:
           for child in dev.children:
             result = result and _CheckDiskConsistency(lu, child, node, on_primary)
         return result
       class LUDiagnoseOS(NoHooksLU):
         """Logical unit for OS diagnose/query.
         """
         _OP_REQP = ["output_fields", "names"]
         REQ_BGL = False
         _FIELDS_STATIC = utils.FieldSet()
         _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants")
         # Fields that need calculation of global os validity
         _FIELDS_NEEDVALID = frozenset(["valid", "variants"])
         def ExpandNames(self):
           if self.op.names:
             raise errors.OpPrereqError("Selective OS query not supported",
                                        errors.ECODE_INVAL)
           _CheckOutputFields(static=self._FIELDS_STATIC,
                              dynamic=self._FIELDS_DYNAMIC,
                              selected=self.op.output_fields)
           # Lock all nodes, in shared mode
           # Temporary removal of locks, should be reverted later
           # TODO: reintroduce locks when they are lighter-weight
           self.needed_locks = {}
           #self.share_locks[locking.LEVEL_NODE] = 1
           #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
         def CheckPrereq(self):
           """Check prerequisites.
           """
         @staticmethod
         def _DiagnoseByOS(rlist):
           """Remaps a per-node return list into an a per-os per-node dictionary
           @param rlist: a map with node names as keys and OS objects as values
           @rtype: dict
           @return: a dictionary with osnames as keys and as value another map, with
               nodes as keys and tuples of (path, status, diagnose) as values, eg::
                 {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
                                            (/srv/..., False, "invalid api")],
                                  "node2": [(/srv/..., True, "")]}
+                }
           """
           all_os = {}
           # we build here the list of nodes that didn't fail the RPC (at RPC
           # level), so that nodes with a non-responding node daemon don't
           # make all OSes invalid
           good_nodes = [node_name for node_name in rlist
                         if not rlist[node_name].fail_msg]
           for node_name, nr in rlist.items():
             if nr.fail_msg or not nr.payload:
               continue
             for name, path, status, diagnose, variants in nr.payload:
               if name not in all_os:
                 # build a list of nodes for this os containing empty lists
                 # for each node in node_list
                 all_os[name] = {}
                 for nname in good_nodes:
                   all_os[name][nname] = []
               all_os[name][node_name].append((path, status, diagnose, variants))
           return all_os
         def Exec(self, feedback_fn):
           """Compute the list of OSes.
           """
           valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
           node_data = self.rpc.call_os_diagnose(valid_nodes)
           pol = self._DiagnoseByOS(node_data)
           output = []
           calc_valid = self._FIELDS_NEEDVALID.intersection(self.op.output_fields)
           calc_variants = "variants" in self.op.output_fields
           for os_name, os_data in pol.items():
             row = []
             if calc_valid:
               valid = True
               variants = None
               for osl in os_data.values():
                 valid = valid and osl and osl[0][1]
                 if not valid:
                   variants = None
                   break
                 if calc_variants:
                   node_variants = osl[0][3]
                   if variants is None:
                     variants = node_variants
                   else:
                     variants = [v for v in variants if v in node_variants]
             for field in self.op.output_fields:
               if field == "name":
                 val = os_name
               elif field == "valid":
                 val = valid
               elif field == "node_status":
                 # this is just a copy of the dict
                 val = {}
                 for node_name, nos_list in os_data.items():
                   val[node_name] = nos_list
               elif field == "variants":
                 val =  variants
               else:
                 raise errors.ParameterError(field)
               row.append(val)
             output.append(row)
           return output
       class LURemoveNode(LogicalUnit):
         """Logical unit for removing a node.
         """
         HPATH = "node-remove"
         HTYPE = constants.HTYPE_NODE
         _OP_REQP = ["node_name"]
         def BuildHooksEnv(self):
           """Build hooks env.
           This doesn't run on the target node in the pre phase as a failed
           node would then be impossible to remove.
           """
           env = {
             "OP_TARGET": self.op.node_name,
             "NODE_NAME": self.op.node_name,
+            }
           all_nodes = self.cfg.GetNodeList()
           try:
             all_nodes.remove(self.op.node_name)
           except ValueError:
             logging.warning("Node %s which is about to be removed not found"
                             " in the all nodes list", self.op.node_name)
           return env, all_nodes, all_nodes
         def CheckPrereq(self):
           """Check prerequisites.
           This checks:
            - the node exists in the configuration
            - it does not have primary or secondary instances
            - it's not the master
           Any errors are signaled by raising errors.OpPrereqError.
           """
           self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
           node = self.cfg.GetNodeInfo(self.op.node_name)
           assert node is not None
           instance_list = self.cfg.GetInstanceList()
           masternode = self.cfg.GetMasterNode()
           if node.name == masternode:
             raise errors.OpPrereqError("Node is the master node,"
                                        " you need to failover first.",
                                        errors.ECODE_INVAL)
           for instance_name in instance_list:
             instance = self.cfg.GetInstanceInfo(instance_name)
             if node.name in instance.all_nodes:
               raise errors.OpPrereqError("Instance %s is still running on the node,"
                                          " please remove first." % instance_name,
                                          errors.ECODE_INVAL)
           self.op.node_name = node.name
           self.node = node
         def Exec(self, feedback_fn):
           """Removes the node from the cluster.
           """
           node = self.node
           logging.info("Stopping the node daemon and removing configs from node %s",
                        node.name)
           modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
           # Promote nodes to master candidate as needed
           _AdjustCandidatePool(self, exceptions=[node.name])
           self.context.RemoveNode(node.name)
           # Run post hooks on the node before it's removed
           hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
           try:
             hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
           except:
             # pylint: disable-msg=W0702
             self.LogWarning("Errors occurred running hooks on %s" % node.name)
           result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
           msg = result.fail_msg
           if msg:
             self.LogWarning("Errors encountered on the remote node while leaving"
                             " the cluster: %s", msg)
           # Remove node from our /etc/hosts
           if self.cfg.GetClusterInfo().modify_etc_hosts:
             # FIXME: this should be done via an rpc call to node daemon
             utils.RemoveHostFromEtcHosts(node.name)
             _RedistributeAncillaryFiles(self)
       class LUQueryNodes(NoHooksLU):
         """Logical unit for querying nodes.
         """
         # pylint: disable-msg=W0142
         _OP_REQP = ["output_fields", "names", "use_locking"]
         REQ_BGL = False
         _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
                           "master_candidate", "offline", "drained"]
         _FIELDS_DYNAMIC = utils.FieldSet(
           "dtotal", "dfree",
           "mtotal", "mnode", "mfree",
           "bootid",
           "ctotal", "cnodes", "csockets",
+          )
         _FIELDS_STATIC = utils.FieldSet(*[
           "pinst_cnt", "sinst_cnt",
           "pinst_list", "sinst_list",
           "pip", "sip", "tags",
           "master",
           "role"] + _SIMPLE_FIELDS
+          )
         def ExpandNames(self):
           _CheckOutputFields(static=self._FIELDS_STATIC,
                              dynamic=self._FIELDS_DYNAMIC,
                              selected=self.op.output_fields)
           self.needed_locks = {}
           self.share_locks[locking.LEVEL_NODE] = 1
           if self.op.names:
             self.wanted = _GetWantedNodes(self, self.op.names)
           else:
             self.wanted = locking.ALL_SET
           self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
           self.do_locking = self.do_node_query and self.op.use_locking
           if self.do_locking:
             # if we don't request only static fields, we need to lock the nodes
             self.needed_locks[locking.LEVEL_NODE] = self.wanted
         def CheckPrereq(self):
           """Check prerequisites.
           """
           # The validation of the node list is done in the _GetWantedNodes,
           # if non empty, and if empty, there's no validation to do
           pass
         def Exec(self, feedback_fn):
           """Computes the list of nodes and their attributes.
           """
           all_info = self.cfg.GetAllNodesInfo()
           if self.do_locking:
             nodenames = self.acquired_locks[locking.LEVEL_NODE]
           elif self.wanted != locking.ALL_SET:
             nodenames = self.wanted
             missing = set(nodenames).difference(all_info.keys())
             if missing:
               raise errors.OpExecError(
                 "Some nodes were removed before retrieving their data: %s" % missing)
           else:
             nodenames = all_info.keys()
           nodenames = utils.NiceSort(nodenames)
           nodelist = [all_info[name] for name in nodenames]
           # begin data gathering
           if self.do_node_query:
             live_data = {}
             node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
                                                 self.cfg.GetHypervisorType())
             for name in nodenames:
               nodeinfo = node_data[name]
               if not nodeinfo.fail_msg and nodeinfo.payload:
                 nodeinfo = nodeinfo.payload
                 fn = utils.TryConvert
                 live_data[name] = {
                   "mtotal": fn(int, nodeinfo.get('memory_total', None)),
                   "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
                   "mfree": fn(int, nodeinfo.get('memory_free', None)),
                   "dtotal": fn(int, nodeinfo.get('vg_size', None)),
                   "dfree": fn(int, nodeinfo.get('vg_free', None)),
                   "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
                   "bootid": nodeinfo.get('bootid', None),
                   "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
                   "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
+                  }
               else:
                 live_data[name] = {}
           else:
             live_data = dict.fromkeys(nodenames, {})
           node_to_primary = dict([(name, set()) for name in nodenames])
           node_to_secondary = dict([(name, set()) for name in nodenames])
           inst_fields = frozenset(("pinst_cnt", "pinst_list",
                                    "sinst_cnt", "sinst_list"))
           if inst_fields & frozenset(self.op.output_fields):
             inst_data = self.cfg.GetAllInstancesInfo()
             for inst in inst_data.values():
               if inst.primary_node in node_to_primary:
                 node_to_primary[inst.primary_node].add(inst.name)
               for secnode in inst.secondary_nodes:
                 if secnode in node_to_secondary:
                   node_to_secondary[secnode].add(inst.name)
           master_node = self.cfg.GetMasterNode()
           # end data gathering
           output = []
           for node in nodelist:
             node_output = []
             for field in self.op.output_fields:
               if field in self._SIMPLE_FIELDS:
                 val = getattr(node, field)
               elif field == "pinst_list":
                 val = list(node_to_primary[node.name])
               elif field == "sinst_list":
                 val = list(node_to_secondary[node.name])
               elif field == "pinst_cnt":
                 val = len(node_to_primary[node.name])
               elif field == "sinst_cnt":
                 val = len(node_to_secondary[node.name])
               elif field == "pip":
                 val = node.primary_ip
               elif field == "sip":
                 val = node.secondary_ip
               elif field == "tags":
                 val = list(node.GetTags())
               elif field == "master":
                 val = node.name == master_node
               elif self._FIELDS_DYNAMIC.Matches(field):
                 val = live_data[node.name].get(field, None)
               elif field == "role":
                 if node.name == master_node:
                   val = "M"
                 elif node.master_candidate:
                   val = "C"
                 elif node.drained:
                   val = "D"
                 elif node.offline:
                   val = "O"
                 else:
                   val = "R"
               else:
                 raise errors.ParameterError(field)
               node_output.append(val)
             output.append(node_output)
           return output
       class LUQueryNodeVolumes(NoHooksLU):
         """Logical unit for getting volumes on node(s).
         """
         _OP_REQP = ["nodes", "output_fields"]
         REQ_BGL = False
         _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
         _FIELDS_STATIC = utils.FieldSet("node")
         def ExpandNames(self):
           _CheckOutputFields(static=self._FIELDS_STATIC,
                              dynamic=self._FIELDS_DYNAMIC,
                              selected=self.op.output_fields)
           self.needed_locks = {}
           self.share_locks[locking.LEVEL_NODE] = 1
           if not self.op.nodes:
             self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
           else:
             self.needed_locks[locking.LEVEL_NODE] = \
               _GetWantedNodes(self, self.op.nodes)
         def CheckPrereq(self):
           """Check prerequisites.
           This checks that the fields required are valid output fields.
           """
           self.nodes = self.acquired_locks[locking.LEVEL_NODE]
         def Exec(self, feedback_fn):
           """Computes the list of nodes and their attributes.
           """
           nodenames = self.nodes
           volumes = self.rpc.call_node_volumes(nodenames)
           ilist = [self.cfg.GetInstanceInfo(iname) for iname
                    in self.cfg.GetInstanceList()]
           lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
           output = []
           for node in nodenames:
             nresult = volumes[node]
             if nresult.offline:
               continue
             msg = nresult.fail_msg
             if msg:
               self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
               continue
             node_vols = nresult.payload[:]
             node_vols.sort(key=lambda vol: vol['dev'])
             for vol in node_vols:
               node_output = []
               for field in self.op.output_fields:
                 if field == "node":
                   val = node
                 elif field == "phys":
                   val = vol['dev']
                 elif field == "vg":
                   val = vol['vg']
                 elif field == "name":
                   val = vol['name']
                 elif field == "size":
                   val = int(float(vol['size']))
                 elif field == "instance":
                   for inst in ilist:
                     if node not in lv_by_node[inst]:
                       continue
                     if vol['name'] in lv_by_node[inst][node]:
                       val = inst.name
                       break
                   else:
                     val = '-'
                 else:
                   raise errors.ParameterError(field)
                 node_output.append(str(val))
               output.append(node_output)
           return output
       class LUQueryNodeStorage(NoHooksLU):
         """Logical unit for getting information on storage units on node(s).
         """
         _OP_REQP = ["nodes", "storage_type", "output_fields"]
         REQ_BGL = False
         _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
         def CheckArguments(self):
           _CheckStorageType(self.op.storage_type)
           _CheckOutputFields(static=self._FIELDS_STATIC,
                              dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
                              selected=self.op.output_fields)
         def ExpandNames(self):
           self.needed_locks = {}
           self.share_locks[locking.LEVEL_NODE] = 1
           if self.op.nodes:
             self.needed_locks[locking.LEVEL_NODE] = \
               _GetWantedNodes(self, self.op.nodes)
           else:
             self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
         def CheckPrereq(self):
           """Check prerequisites.
           This checks that the fields required are valid output fields.
           """
           self.op.name = getattr(self.op, "name", None)
           self.nodes = self.acquired_locks[locking.LEVEL_NODE]
         def Exec(self, feedback_fn):
           """Computes the list of nodes and their attributes.
           """
           # Always get name to sort by
           if constants.SF_NAME in self.op.output_fields:
             fields = self.op.output_fields[:]
           else:
             fields = [constants.SF_NAME] + self.op.output_fields
           # Never ask for node or type as it's only known to the LU
           for extra in [constants.SF_NODE, constants.SF_TYPE]:
             while extra in fields:
               fields.remove(extra)
           field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
           name_idx = field_idx[constants.SF_NAME]
           st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
           data = self.rpc.call_storage_list(self.nodes,
                                             self.op.storage_type, st_args,
                                             self.op.name, fields)
           result = []
           for node in utils.NiceSort(self.nodes):
             nresult = data[node]
             if nresult.offline:
               continue
             msg = nresult.fail_msg
             if msg:
               self.LogWarning("Can't get storage data from node %s: %s", node, msg)
               continue
             rows = dict([(row[name_idx], row) for row in nresult.payload])
             for name in utils.NiceSort(rows.keys()):
               row = rows[name]
               out = []
               for field in self.op.output_fields:
                 if field == constants.SF_NODE:
                   val = node
                 elif field == constants.SF_TYPE:
                   val = self.op.storage_type
                 elif field in field_idx:
                   val = row[field_idx[field]]
                 else:
                   raise errors.ParameterError(field)
                 out.append(val)
               result.append(out)
           return result
       class LUModifyNodeStorage(NoHooksLU):
         """Logical unit for modifying a storage volume on a node.
         """
         _OP_REQP = ["node_name", "storage_type", "name", "changes"]
         REQ_BGL = False
         def CheckArguments(self):
           self.opnode_name = _ExpandNodeName(self.cfg, self.op.node_name)
           _CheckStorageType(self.op.storage_type)
         def ExpandNames(self):
           self.needed_locks = {
             locking.LEVEL_NODE: self.op.node_name,
+            }
         def CheckPrereq(self):
           """Check prerequisites.
           """
           storage_type = self.op.storage_type
           try:
             modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
           except KeyError:
             raise errors.OpPrereqError("Storage units of type '%s' can not be"
                                        " modified" % storage_type,
                                        errors.ECODE_INVAL)
           diff = set(self.op.changes.keys()) - modifiable
           if diff:
             raise errors.OpPrereqError("The following fields can not be modified for"
                                        " storage units of type '%s': %r" %
                                        (storage_type, list(diff)),
                                        errors.ECODE_INVAL)
         def Exec(self, feedback_fn):
           """Computes the list of nodes and their attributes.
           """
           st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
           result = self.rpc.call_storage_modify(self.op.node_name,
                                                 self.op.storage_type, st_args,
                                                 self.op.name, self.op.changes)
           result.Raise("Failed to modify storage unit '%s' on %s" %
                        (self.op.name, self.op.node_name))
       class LUAddNode(LogicalUnit):
         """Logical unit for adding node to the cluster.
         """
         HPATH = "node-add"
         HTYPE = constants.HTYPE_NODE
         _OP_REQP = ["node_name"]
         def CheckArguments(self):
           # validate/normalize the node name
           self.op.node_name = utils.HostInfo.NormalizeName(self.op.node_name)
         def BuildHooksEnv(self):
           """Build hooks env.
           This will run on all nodes before, and on all nodes + the new node after.
           """
           env = {
             "OP_TARGET": self.op.node_name,
             "NODE_NAME": self.op.node_name,
             "NODE_PIP": self.op.primary_ip,
             "NODE_SIP": self.op.secondary_ip,
+            }
           nodes_0 = self.cfg.GetNodeList()
           nodes_1 = nodes_0 + [self.op.node_name, ]
           return env, nodes_0, nodes_1
         def CheckPrereq(self):
           """Check prerequisites.
           This checks:
            - the new node is not already in the config
            - it is resolvable
            - its parameters (single/dual homed) matches the cluster
           Any errors are signaled by raising errors.OpPrereqError.
           """
           node_name = self.op.node_name
           cfg = self.cfg
           dns_data = utils.GetHostInfo(node_name)
           node = dns_data.name
           primary_ip = self.op.primary_ip = dns_data.ip
           secondary_ip = getattr(self.op, "secondary_ip", None)
           if secondary_ip is None:
             secondary_ip = primary_ip
           if not utils.IsValidIP(secondary_ip):
             raise errors.OpPrereqError("Invalid secondary IP given",
                                        errors.ECODE_INVAL)
           self.op.secondary_ip = secondary_ip
           node_list = cfg.GetNodeList()
           if not self.op.readd and node in node_list:
             raise errors.OpPrereqError("Node %s is already in the configuration" %
                                        node, errors.ECODE_EXISTS)
           elif self.op.readd and node not in node_list:
             raise errors.OpPrereqError("Node %s is not in the configuration" % node,
                                        errors.ECODE_NOENT)
           self.changed_primary_ip = False
           for existing_node_name in node_list:
             existing_node = cfg.GetNodeInfo(existing_node_name)
             if self.op.readd and node == existing_node_name:
               if existing_node.secondary_ip != secondary_ip:
                 raise errors.OpPrereqError("Readded node doesn't have the same IP"
                                            " address configuration as before",
                                            errors.ECODE_INVAL)
               if existing_node.primary_ip != primary_ip:
                 self.changed_primary_ip = True
               continue
             if (existing_node.primary_ip == primary_ip or
                 existing_node.secondary_ip == primary_ip or
                 existing_node.primary_ip == secondary_ip or
                 existing_node.secondary_ip == secondary_ip):
               raise errors.OpPrereqError("New node ip address(es) conflict with"
                                          " existing node %s" % existing_node.name,
                                          errors.ECODE_NOTUNIQUE)
           # check that the type of the node (single versus dual homed) is the
           # same as for the master
           myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
           master_singlehomed = myself.secondary_ip == myself.primary_ip
           newbie_singlehomed = secondary_ip == primary_ip
           if master_singlehomed != newbie_singlehomed:
             if master_singlehomed:
               raise errors.OpPrereqError("The master has no private ip but the"
                                          " new node has one",
                                          errors.ECODE_INVAL)
             else:
               raise errors.OpPrereqError("The master has a private ip but the"
                                          " new node doesn't have one",
                                          errors.ECODE_INVAL)
           # checks reachability
           if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
             raise errors.OpPrereqError("Node not reachable by ping",
                                        errors.ECODE_ENVIRON)
           if not newbie_singlehomed:
             # check reachability from my secondary ip to newbie's secondary ip
             if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
                                  source=myself.secondary_ip):
               raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
                                          " based ping to noded port",
                                          errors.ECODE_ENVIRON)
           if self.op.readd:
             exceptions = [node]
           else:
             exceptions = []
           self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
           if self.op.readd:
             self.new_node = self.cfg.GetNodeInfo(node)
             assert self.new_node is not None, "Can't retrieve locked node %s" % node
           else:
             self.new_node = objects.Node(name=node,
                                          primary_ip=primary_ip,
                                          secondary_ip=secondary_ip,
                                          master_candidate=self.master_candidate,
                                          offline=False, drained=False)
         def Exec(self, feedback_fn):
           """Adds the new node to the cluster.
           """
           new_node = self.new_node
           node = new_node.name
           # for re-adds, reset the offline/drained/master-candidate flags;
           # we need to reset here, otherwise offline would prevent RPC calls
           # later in the procedure; this also means that if the re-add
           # fails, we are left with a non-offlined, broken node
           if self.op.readd:
             new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
             self.LogInfo("Readding a node, the offline/drained flags were reset")
             # if we demote the node, we do cleanup later in the procedure
             new_node.master_candidate = self.master_candidate
             if self.changed_primary_ip:
               new_node.primary_ip = self.op.primary_ip
           # notify the user about any possible mc promotion
           if new_node.master_candidate:
             self.LogInfo("Node will be a master candidate")
           # check connectivity
           result = self.rpc.call_version([node])[node]
           result.Raise("Can't get version information from node %s" % node)
           if constants.PROTOCOL_VERSION == result.payload:
             logging.info("Communication to node %s fine, sw version %s match",
                          node, result.payload)
           else:
             raise errors.OpExecError("Version mismatch master version %s,"
                                      " node version %s" %
                                      (constants.PROTOCOL_VERSION, result.payload))
           # setup ssh on node
           if self.cfg.GetClusterInfo().modify_ssh_setup:
             logging.info("Copy ssh key to node %s", node)
             priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
             keyarray = []
             keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
                         constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
                         priv_key, pub_key]
             for i in keyfiles:
               keyarray.append(utils.ReadFile(i))
             result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
                                             keyarray[2], keyarray[3], keyarray[4],
                                             keyarray[5])
             result.Raise("Cannot transfer ssh keys to the new node")
           # Add node to our /etc/hosts, and add key to known_hosts
           if self.cfg.GetClusterInfo().modify_etc_hosts:
             # FIXME: this should be done via an rpc call to node daemon
             utils.AddHostToEtcHosts(new_node.name)
           if new_node.secondary_ip != new_node.primary_ip:
             result = self.rpc.call_node_has_ip_address(new_node.name,
                                                        new_node.secondary_ip)
             result.Raise("Failure checking secondary ip on node %s" % new_node.name,
                          prereq=True, ecode=errors.ECODE_ENVIRON)
             if not result.payload:
               raise errors.OpExecError("Node claims it doesn't have the secondary ip"
                                        " you gave (%s). Please fix and re-run this"
                                        " command." % new_node.secondary_ip)
           node_verify_list = [self.cfg.GetMasterNode()]
           node_verify_param = {
             constants.NV_NODELIST: [node],
             # TODO: do a node-net-test as well?
+          }
           result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
                                              self.cfg.GetClusterName())
           for verifier in node_verify_list:
             result[verifier].Raise("Cannot communicate with node %s" % verifier)
             nl_payload = result[verifier].payload[constants.NV_NODELIST]
             if nl_payload:
               for failed in nl_payload:
                 feedback_fn("ssh/hostname verification failed"
                             " (checking from %s): %s" %
                             (verifier, nl_payload[failed]))
               raise errors.OpExecError("ssh/hostname verification failed.")
           if self.op.readd:
             _RedistributeAncillaryFiles(self)
             self.context.ReaddNode(new_node)
             # make sure we redistribute the config
             self.cfg.Update(new_node, feedback_fn)
             # and make sure the new node will not have old files around
             if not new_node.master_candidate:
               result = self.rpc.call_node_demote_from_mc(new_node.name)
               msg = result.fail_msg
               if msg:
                 self.LogWarning("Node failed to demote itself from master"
                                 " candidate status: %s" % msg)
           else:
             _RedistributeAncillaryFiles(self, additional_nodes=[node])
             self.context.AddNode(new_node, self.proc.GetECId())
       class LUSetNodeParams(LogicalUnit):
         """Modifies the parameters of a node.
         """
         HPATH = "node-modify"
         HTYPE = constants.HTYPE_NODE
         _OP_REQP = ["node_name"]
         REQ_BGL = False
         def CheckArguments(self):
           self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
           _CheckBooleanOpField(self.op, 'master_candidate')
           _CheckBooleanOpField(self.op, 'offline')
           _CheckBooleanOpField(self.op, 'drained')
           _CheckBooleanOpField(self.op, 'auto_promote')
           all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
           if all_mods.count(None) == 3:
             raise errors.OpPrereqError("Please pass at least one modification",
                                        errors.ECODE_INVAL)
           if all_mods.count(True) > 1:
             raise errors.OpPrereqError("Can't set the node into more than one"
                                        " state at the same time",
                                        errors.ECODE_INVAL)
           # Boolean value that tells us whether we're offlining or draining the node
           self.offline_or_drain = (self.op.offline == True or
                                    self.op.drained == True)
           self.deoffline_or_drain = (self.op.offline == False or
                                      self.op.drained == False)
           self.might_demote = (self.op.master_candidate == False or
                                self.offline_or_drain)
           self.lock_all = self.op.auto_promote and self.might_demote
         def ExpandNames(self):
           if self.lock_all:
             self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
           else:
             self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
         def BuildHooksEnv(self):
           """Build hooks env.
           This runs on the master node.
           """
           env = {
             "OP_TARGET": self.op.node_name,
             "MASTER_CANDIDATE": str(self.op.master_candidate),
             "OFFLINE": str(self.op.offline),
             "DRAINED": str(self.op.drained),
+            }
           nl = [self.cfg.GetMasterNode(),
                 self.op.node_name]
           return env, nl, nl
         def CheckPrereq(self):
           """Check prerequisites.
           This only checks the instance list against the existing names.
           """
           node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
           if (self.op.master_candidate is not None or
               self.op.drained is not None or
               self.op.offline is not None):
             # we can't change the master's node flags
             if self.op.node_name == self.cfg.GetMasterNode():
               raise errors.OpPrereqError("The master role can be changed"
                                          " only via masterfailover",
                                          errors.ECODE_INVAL)
           if node.master_candidate and self.might_demote and not self.lock_all:
             assert not self.op.auto_promote, "auto-promote set but lock_all not"
             # check if after removing the current node, we're missing master
             # candidates
             (mc_remaining, mc_should, _) = \
                 self.cfg.GetMasterCandidateStats(exceptions=[node.name])
             if mc_remaining < mc_should:
               raise errors.OpPrereqError("Not enough master candidates, please"
                                          " pass auto_promote to allow promotion",
                                          errors.ECODE_INVAL)
           if (self.op.master_candidate == True and
               ((node.offline and not self.op.offline == False) or
                (node.drained and not self.op.drained == False))):
             raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
                                        " to master_candidate" % node.name,
                                        errors.ECODE_INVAL)
           # If we're being deofflined/drained, we'll MC ourself if needed
           if (self.deoffline_or_drain and not self.offline_or_drain and not
               self.op.master_candidate == True and not node.master_candidate):
             self.op.master_candidate = _DecideSelfPromotion(self)
             if self.op.master_candidate:
               self.LogInfo("Autopromoting node to master candidate")
           return
         def Exec(self, feedback_fn):
           """Modifies a node.
           """
           node = self.node
           result = []
           changed_mc = False
           if self.op.offline is not None:
             node.offline = self.op.offline
             result.append(("offline", str(self.op.offline)))
             if self.op.offline == True:
               if node.master_candidate:
                 node.master_candidate = False
                 changed_mc = True
                 result.append(("master_candidate", "auto-demotion due to offline"))
               if node.drained:
                 node.drained = False
                 result.append(("drained", "clear drained status due to offline"))
           if self.op.master_candidate is not None:
             node.master_candidate = self.op.master_candidate
             changed_mc = True
             result.append(("master_candidate", str(self.op.master_candidate)))
             if self.op.master_candidate == False:
               rrc = self.rpc.call_node_demote_from_mc(node.name)
               msg = rrc.fail_msg
               if msg:
                 self.LogWarning("Node failed to demote itself: %s" % msg)
           if self.op.drained is not None:
             node.drained = self.op.drained
             result.append(("drained", str(self.op.drained)))
             if self.op.drained == True:
               if node.master_candidate:
                 node.master_candidate = False
                 changed_mc = True
                 result.append(("master_candidate", "auto-demotion due to drain"))
                 rrc = self.rpc.call_node_demote_from_mc(node.name)
                 msg = rrc.fail_msg
                 if msg:
                   self.LogWarning("Node failed to demote itself: %s" % msg)
               if node.offline:
                 node.offline = False
                 result.append(("offline", "clear offline status due to drain"))
           # we locked all nodes, we adjust the CP before updating this node
           if self.lock_all:
             _AdjustCandidatePool(self, [node.name])
           # this will trigger configuration file update, if needed
           self.cfg.Update(node, feedback_fn)
           # this will trigger job queue propagation or cleanup
           if changed_mc:
             self.context.ReaddNode(node)
           return result
       class LUPowercycleNode(NoHooksLU):
         """Powercycles a node.
         """
         _OP_REQP = ["node_name", "force"]
         REQ_BGL = False
         def CheckArguments(self):
           self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
           if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
             raise errors.OpPrereqError("The node is the master and the force"
                                        " parameter was not set",
                                        errors.ECODE_INVAL)
         def ExpandNames(self):
           """Locking for PowercycleNode.
           This is a last-resort option and shouldn't block on other
           jobs. Therefore, we grab no locks.
           """
           self.needed_locks = {}
         def CheckPrereq(self):
           """Check prerequisites.
           This LU has no prereqs.
           """
           pass
         def Exec(self, feedback_fn):
           """Reboots a node.
           """
           result = self.rpc.call_node_powercycle(self.op.node_name,
                                                  self.cfg.GetHypervisorType())
           result.Raise("Failed to schedule the reboot")
           return result.payload
       class LUQueryClusterInfo(NoHooksLU):
         """Query cluster configuration.
         """
         _OP_REQP = []
         REQ_BGL = False
         def ExpandNames(self):
           self.needed_locks = {}
         def CheckPrereq(self):
           """No prerequsites needed for this LU.
           """
           pass
         def Exec(self, feedback_fn):
           """Return cluster config.
           """
           cluster = self.cfg.GetClusterInfo()
           os_hvp = {}
           # Filter just for enabled hypervisors
           for os_name, hv_dict in cluster.os_hvp.items():
             os_hvp[os_name] = {}
             for hv_name, hv_params in hv_dict.items():
               if hv_name in cluster.enabled_hypervisors:
                 os_hvp[os_name][hv_name] = hv_params
           result = {
             "software_version": constants.RELEASE_VERSION,
             "protocol_version": constants.PROTOCOL_VERSION,
             "config_version": constants.CONFIG_VERSION,
             "os_api_version": max(constants.OS_API_VERSIONS),
             "export_version": constants.EXPORT_VERSION,
             "architecture": (platform.architecture()[0], platform.machine()),
             "name": cluster.cluster_name,
             "master": cluster.master_node,
             "default_hypervisor": cluster.enabled_hypervisors[0],
             "enabled_hypervisors": cluster.enabled_hypervisors,
             "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
                               for hypervisor_name in cluster.enabled_hypervisors]),
             "os_hvp": os_hvp,
             "beparams": cluster.beparams,
             "nicparams": cluster.nicparams,
             "candidate_pool_size": cluster.candidate_pool_size,
             "master_netdev": cluster.master_netdev,
             "volume_group_name": cluster.volume_group_name,
             "file_storage_dir": cluster.file_storage_dir,
             "maintain_node_health": cluster.maintain_node_health,
             "ctime": cluster.ctime,
             "mtime": cluster.mtime,
             "uuid": cluster.uuid,
             "tags": list(cluster.GetTags()),
             "uid_pool": cluster.uid_pool,
+            }
           return result
       class LUQueryConfigValues(NoHooksLU):
         """Return configuration values.
         """
         _OP_REQP = []
         REQ_BGL = False
         _FIELDS_DYNAMIC = utils.FieldSet()
         _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
                                         "watcher_pause")
         def ExpandNames(self):
           self.needed_locks = {}
           _CheckOutputFields(static=self._FIELDS_STATIC,
                              dynamic=self._FIELDS_DYNAMIC,
                              selected=self.op.output_fields)
         def CheckPrereq(self):
           """No prerequisites.
           """
           pass
         def Exec(self, feedback_fn):
           """Dump a representation of the cluster config to the standard output.
           """
           values = []
           for field in self.op.output_fields:
             if field == "cluster_name":
               entry = self.cfg.GetClusterName()
             elif field == "master_node":
               entry = self.cfg.GetMasterNode()
             elif field == "drain_flag":
               entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
             elif field == "watcher_pause":
               entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
             else:
               raise errors.ParameterError(field)
             values.append(entry)
           return values
       class LUActivateInstanceDisks(NoHooksLU):
         """Bring up an instance's disks.
         """
         _OP_REQP = ["instance_name"]
         REQ_BGL = False
         def ExpandNames(self):
           self._ExpandAndLockInstance()
           self.needed_locks[locking.LEVEL_NODE] = []
           self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
         def DeclareLocks(self, level):
           if level == locking.LEVEL_NODE:
             self._LockInstancesNodes()
         def CheckPrereq(self):
           """Check prerequisites.
           This checks that the instance is in the cluster.
           """
           self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
           assert self.instance is not None, \
             "Cannot retrieve locked instance %s" % self.op.instance_name
           _CheckNodeOnline(self, self.instance.primary_node)
           if not hasattr(self.op, "ignore_size"):
             self.op.ignore_size = False
         def Exec(self, feedback_fn):
           """Activate the disks.
           """
           disks_ok, disks_info = \
                     _AssembleInstanceDisks(self, self.instance,
                                            ignore_size=self.op.ignore_size)
           if not disks_ok:
             raise errors.OpExecError("Cannot activate block devices")
           return disks_info
       def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
                                  ignore_size=False):
         """Prepare the block devices for an instance.
         This sets up the block devices on all nodes.
         @type lu: L{LogicalUnit}
         @param lu: the logical unit on whose behalf we execute
         @type instance: L{objects.Instance}
         @param instance: the instance for whose disks we assemble
         @type disks: list of L{objects.Disk} or None
         @param disks: which disks to assemble (or all, if None)
         @type ignore_secondaries: boolean
         @param ignore_secondaries: if true, errors on secondary nodes
             won't result in an error return from the function
         @type ignore_size: boolean
         @param ignore_size: if true, the current known size of the disk
             will not be used during the disk activation, useful for cases
             when the size is wrong
         @return: False if the operation failed, otherwise a list of
             (host, instance_visible_name, node_visible_name)
             with the mapping from node devices to instance devices
         """
         device_info = []
         disks_ok = True
         iname = instance.name
         disks = _ExpandCheckDisks(instance, disks)
         # With the two passes mechanism we try to reduce the window of
         # opportunity for the race condition of switching DRBD to primary
         # before handshaking occured, but we do not eliminate it
         # The proper fix would be to wait (with some limits) until the
         # connection has been made and drbd transitions from WFConnection
         # into any other network-connected state (Connected, SyncTarget,
         # SyncSource, etc.)
         # 1st pass, assemble on all nodes in secondary mode
         for inst_disk in disks:
           for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
             if ignore_size:
               node_disk = node_disk.Copy()
               node_disk.UnsetSize()
             lu.cfg.SetDiskID(node_disk, node)
             result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
             msg = result.fail_msg
             if msg:
               lu.proc.LogWarning("Could not prepare block device %s on node %s"
                                  " (is_primary=False, pass=1): %s",
                                  inst_disk.iv_name, node, msg)
               if not ignore_secondaries:
                 disks_ok = False
         # FIXME: race condition on drbd migration to primary
         # 2nd pass, do only the primary node
         for inst_disk in disks:
           dev_path = None
           for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
             if node != instance.primary_node:
               continue
             if ignore_size:
               node_disk = node_disk.Copy()
               node_disk.UnsetSize()
             lu.cfg.SetDiskID(node_disk, node)
             result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
             msg = result.fail_msg
             if msg:
               lu.proc.LogWarning("Could not prepare block device %s on node %s"
                                  " (is_primary=True, pass=2): %s",
                                  inst_disk.iv_name, node, msg)
               disks_ok = False
             else:
               dev_path = result.payload
           device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
         # leave the disks configured for the primary node
         # this is a workaround that would be fixed better by
         # improving the logical/physical id handling
         for disk in disks:
           lu.cfg.SetDiskID(disk, instance.primary_node)
         return disks_ok, device_info
       def _StartInstanceDisks(lu, instance, force):
         """Start the disks of an instance.
         """
         disks_ok, _ = _AssembleInstanceDisks(lu, instance,
                                                  ignore_secondaries=force)
         if not disks_ok:
           _ShutdownInstanceDisks(lu, instance)
           if force is not None and not force:
             lu.proc.LogWarning("", hint="If the message above refers to a"
                                " secondary node,"
                                " you can retry the operation using '--force'.")
           raise errors.OpExecError("Disk consistency error")
       class LUDeactivateInstanceDisks(NoHooksLU):
         """Shutdown an instance's disks.
         """
         _OP_REQP = ["instance_name"]
         REQ_BGL = False
         def ExpandNames(self):
           self._ExpandAndLockInstance()
           self.needed_locks[locking.LEVEL_NODE] = []
           self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
         def DeclareLocks(self, level):
           if level == locking.LEVEL_NODE:
             self._LockInstancesNodes()
         def CheckPrereq(self):
           """Check prerequisites.
           This checks that the instance is in the cluster.
           """
           self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
           assert self.instance is not None, \
             "Cannot retrieve locked instance %s" % self.op.instance_name
         def Exec(self, feedback_fn):
           """Deactivate the disks
           """
           instance = self.instance
           _SafeShutdownInstanceDisks(self, instance)
       def _SafeShutdownInstanceDisks(lu, instance, disks=None):
         """Shutdown block devices of an instance.
         This function checks if an instance is running, before calling
         _ShutdownInstanceDisks.
         """
         _CheckInstanceDown(lu, instance, "cannot shutdown disks")
         _ShutdownInstanceDisks(lu, instance, disks=disks)
       def _ExpandCheckDisks(instance, disks):
         """Return the instance disks selected by the disks list
         @type disks: list of L{objects.Disk} or None
         @param disks: selected disks
         @rtype: list of L{objects.Disk}
         @return: selected instance disks to act on
         """
         if disks is None:
           return instance.disks
         else:
           if not set(disks).issubset(instance.disks):
             raise errors.ProgrammerError("Can only act on disks belonging to the"

Synnefo » snf-ganeti

root / lib / cmdlib.py @ e6b8d02d