Transition into and out of offline instance state

[ganeti-local] / lib / cmdlib.py

diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index 362aa15..57883c1 100644 (file)
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -21,12 +21,12 @@
 
 """Module implementing the master-side code."""
 
 
 """Module implementing the master-side code."""
 

-# pylint: disable-msg=W0201,C0302
+# pylint: disable=W0201,C0302

 
 # W0201 since most LU attributes are defined in CheckPrereq or similar
 # functions
 
 
 # W0201 since most LU attributes are defined in CheckPrereq or similar
 # functions
 

-# C0302: since we have waaaay to many lines in this module
+# C0302: since we have waaaay too many lines in this module

 
 import os
 import os.path
 
 import os
 import os.path

@@ -59,8 +59,20 @@ from ganeti import query
 from ganeti import qlang
 from ganeti import opcodes
 from ganeti import ht
 from ganeti import qlang
 from ganeti import opcodes
 from ganeti import ht

+from ganeti import rpc

 
 

-import ganeti.masterd.instance # pylint: disable-msg=W0611
+import ganeti.masterd.instance # pylint: disable=W0611
+
+
+#: Size of DRBD meta block device
+DRBD_META_SIZE = 128
+
+# States of instance
+INSTANCE_UP = [constants.ADMINST_UP]
+INSTANCE_DOWN = [constants.ADMINST_DOWN]
+INSTANCE_OFFLINE = [constants.ADMINST_OFFLINE]
+INSTANCE_ONLINE = [constants.ADMINST_DOWN, constants.ADMINST_UP]
+INSTANCE_NOT_RUNNING = [constants.ADMINST_DOWN, constants.ADMINST_OFFLINE]

 
 
 class ResultWithJobs:
 
 
 class ResultWithJobs:

@@ -108,7 +120,7 @@ class LogicalUnit(object):
   HTYPE = None
   REQ_BGL = True
 
   HTYPE = None
   REQ_BGL = True
 

-  def __init__(self, processor, op, context, rpc):
+  def __init__(self, processor, op, context, rpc_runner):

     """Constructor for LogicalUnit.
 
     This needs to be overridden in derived classes in order to check op
     """Constructor for LogicalUnit.
 
     This needs to be overridden in derived classes in order to check op

@@ -122,7 +134,7 @@ class LogicalUnit(object):
     # readability alias
     self.owned_locks = context.glm.list_owned
     self.context = context
     # readability alias
     self.owned_locks = context.glm.list_owned
     self.context = context

-    self.rpc = rpc
+    self.rpc = rpc_runner

     # Dicts used to declare locking needs to mcpu
     self.needed_locks = None
     self.share_locks = dict.fromkeys(locking.LEVELS, 0)
     # Dicts used to declare locking needs to mcpu
     self.needed_locks = None
     self.share_locks = dict.fromkeys(locking.LEVELS, 0)

@@ -131,10 +143,10 @@ class LogicalUnit(object):
     # Used to force good behavior when calling helper functions
     self.recalculate_locks = {}
     # logging
     # Used to force good behavior when calling helper functions
     self.recalculate_locks = {}
     # logging

-    self.Log = processor.Log # pylint: disable-msg=C0103
-    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
-    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
-    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
+    self.Log = processor.Log # pylint: disable=C0103
+    self.LogWarning = processor.LogWarning # pylint: disable=C0103
+    self.LogInfo = processor.LogInfo # pylint: disable=C0103
+    self.LogStep = processor.LogStep # pylint: disable=C0103

     # support for dry-run
     self.dry_run_result = None
     # support for generic debug attribute
     # support for dry-run
     self.dry_run_result = None
     # support for generic debug attribute

@@ -322,7 +334,7 @@ class LogicalUnit(object):
     """
     # API must be kept, thus we ignore the unused argument and could
     # be a function warnings
     """
     # API must be kept, thus we ignore the unused argument and could
     # be a function warnings

-    # pylint: disable-msg=W0613,R0201
+    # pylint: disable=W0613,R0201

     return lu_result
 
   def _ExpandAndLockInstance(self):
     return lu_result
 
   def _ExpandAndLockInstance(self):

@@ -344,7 +356,8 @@ class LogicalUnit(object):
                                                 self.op.instance_name)
     self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
 
                                                 self.op.instance_name)
     self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
 

-  def _LockInstancesNodes(self, primary_only=False):
+  def _LockInstancesNodes(self, primary_only=False,
+                          level=locking.LEVEL_NODE):

     """Helper function to declare instances' nodes for locking.
 
     This function should be called after locking one or more instances to lock
     """Helper function to declare instances' nodes for locking.
 
     This function should be called after locking one or more instances to lock

@@ -365,9 +378,10 @@ class LogicalUnit(object):
 
     @type primary_only: boolean
     @param primary_only: only lock primary nodes of locked instances
 
     @type primary_only: boolean
     @param primary_only: only lock primary nodes of locked instances

+    @param level: Which lock level to use for locking nodes

 
     """
 
     """

-    assert locking.LEVEL_NODE in self.recalculate_locks, \
+    assert level 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
       "_LockInstancesNodes helper function called with no nodes to recalculate"
 
     # TODO: check if we're really been called with the instance locks held

@@ -382,15 +396,17 @@ class LogicalUnit(object):
       if not primary_only:
         wanted_nodes.extend(instance.secondary_nodes)
 
       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)
+    if self.recalculate_locks[level] == constants.LOCKS_REPLACE:
+      self.needed_locks[level] = wanted_nodes
+    elif self.recalculate_locks[level] == constants.LOCKS_APPEND:
+      self.needed_locks[level].extend(wanted_nodes)
+    else:
+      raise errors.ProgrammerError("Unknown recalculation mode")

 
 

-    del self.recalculate_locks[locking.LEVEL_NODE]
+    del self.recalculate_locks[level]

 
 
 
 

-class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
+class NoHooksLU(LogicalUnit): # pylint: disable=W0223

   """Simple LU which runs no hooks.
 
   This LU is intended as a parent for other LogicalUnits which will
   """Simple LU which runs no hooks.
 
   This LU is intended as a parent for other LogicalUnits which will

@@ -468,13 +484,13 @@ class _QueryBase:
   #: Attribute holding field definitions
   FIELDS = None
 
   #: Attribute holding field definitions
   FIELDS = None
 

-  def __init__(self, filter_, fields, use_locking):
+  def __init__(self, qfilter, fields, use_locking):

     """Initializes this class.
 
     """
     self.use_locking = use_locking
 
     """Initializes this class.
 
     """
     self.use_locking = use_locking
 

-    self.query = query.Query(self.FIELDS, fields, filter_=filter_,
+    self.query = query.Query(self.FIELDS, fields, qfilter=qfilter,

                              namefield="name")
     self.requested_data = self.query.RequestedData()
     self.names = self.query.RequestedNames()
                              namefield="name")
     self.requested_data = self.query.RequestedData()
     self.names = self.query.RequestedNames()

@@ -712,12 +728,17 @@ def _ReleaseLocks(lu, level, names=None, keep=None):
   else:
     should_release = None
 
   else:
     should_release = None
 

-  if should_release:
+  owned = lu.owned_locks(level)
+  if not owned:
+    # Not owning any lock at this level, do nothing
+    pass
+
+  elif should_release:

     retain = []
     release = []
 
     # Determine which locks to release
     retain = []
     release = []
 
     # Determine which locks to release

-    for name in lu.owned_locks(level):
+    for name in owned:

       if should_release(name):
         release.append(name)
       else:
       if should_release(name):
         release.append(name)
       else:

@@ -753,11 +774,11 @@ def _RunPostHook(lu, node_name):
   """Runs the post-hook for an opcode on a single node.
 
   """
   """Runs the post-hook for an opcode on a single node.
 
   """

-  hm = lu.proc.hmclass(lu.rpc.call_hooks_runner, lu)
+  hm = lu.proc.BuildHooksManager(lu)

   try:
     hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
   except:
   try:
     hm.RunPhase(constants.HOOKS_PHASE_POST, nodes=[node_name])
   except:

-    # pylint: disable-msg=W0702
+    # pylint: disable=W0702

     lu.LogWarning("Errors occurred running hooks on %s" % node_name)
 
 
     lu.LogWarning("Errors occurred running hooks on %s" % node_name)
 
 

@@ -889,20 +910,31 @@ def _GetClusterDomainSecret():
                                strict=True)
 
 
                                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)
+def _CheckInstanceState(lu, instance, req_states, msg=None):
+  """Ensure that an instance is in one of the required states.
+
+  @param lu: the LU on behalf of which we make the check
+  @param instance: the instance to check
+  @param msg: if passed, should be a message to replace the default one
+  @raise errors.OpPrereqError: if the instance is not in the required 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 msg is None:
+    msg = "can't use instance from outside %s states" % ", ".join(req_states)
+  if instance.admin_state not in req_states:
+    raise errors.OpPrereqError("Instance %s is marked to be %s, %s" %
+                               (instance, instance.admin_state, msg),
+                               errors.ECODE_STATE)
+
+  if constants.ADMINST_UP not in req_states:
+    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)
+    if instance.name in ins_l.payload:
+      raise errors.OpPrereqError("Instance %s is running, %s" %
+                                 (instance.name, msg), errors.ECODE_STATE)

 
 
 def _ExpandItemName(fn, name, kind):
 
 
 def _ExpandItemName(fn, name, kind):

@@ -947,8 +979,8 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
   @param secondary_nodes: list of secondary nodes as strings
   @type os_type: string
   @param os_type: the name of the instance's OS
   @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 status: string
+  @param status: the desired status of the instance

   @type memory: string
   @param memory: the memory size of the instance
   @type vcpus: string
   @type memory: string
   @param memory: the memory size of the instance
   @type vcpus: string

@@ -972,17 +1004,13 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
   @return: the hook environment for this instance
 
   """
   @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,
   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_STATUS": status,

     "INSTANCE_MEMORY": memory,
     "INSTANCE_VCPUS": vcpus,
     "INSTANCE_DISK_TEMPLATE": disk_template,
     "INSTANCE_MEMORY": memory,
     "INSTANCE_VCPUS": vcpus,
     "INSTANCE_DISK_TEMPLATE": disk_template,

@@ -1074,7 +1102,7 @@ def _BuildInstanceHookEnvByObject(lu, instance, override=None):
     "primary_node": instance.primary_node,
     "secondary_nodes": instance.secondary_nodes,
     "os_type": instance.os,
     "primary_node": instance.primary_node,
     "secondary_nodes": instance.secondary_nodes,
     "os_type": instance.os,

-    "status": instance.admin_up,
+    "status": instance.admin_state,

     "memory": bep[constants.BE_MEMORY],
     "vcpus": bep[constants.BE_VCPUS],
     "nics": _NICListToTuple(lu, instance.nics),
     "memory": bep[constants.BE_MEMORY],
     "vcpus": bep[constants.BE_VCPUS],
     "nics": _NICListToTuple(lu, instance.nics),

@@ -1087,7 +1115,7 @@ def _BuildInstanceHookEnvByObject(lu, instance, override=None):
   }
   if override:
     args.update(override)
   }
   if override:
     args.update(override)

-  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
+  return _BuildInstanceHookEnv(**args) # pylint: disable=W0142

 
 
 def _AdjustCandidatePool(lu, exceptions):
 
 
 def _AdjustCandidatePool(lu, exceptions):

@@ -1204,13 +1232,13 @@ def _GetStorageTypeArgs(cfg, storage_type):
   return []
 
 
   return []
 
 

-def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
+def _FindFaultyInstanceDisks(cfg, rpc_runner, instance, node_name, prereq):

   faulty = []
 
   for dev in instance.disks:
     cfg.SetDiskID(dev, node_name)
 
   faulty = []
 
   for dev in instance.disks:
     cfg.SetDiskID(dev, node_name)
 

-  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
+  result = rpc_runner.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)
 
   result.Raise("Failed to get disk status from node %s" % node_name,
                prereq=prereq, ecode=errors.ECODE_ENVIRON)
 

@@ -1350,15 +1378,17 @@ class LUClusterDestroy(LogicalUnit):
     """Destroys the cluster.
 
     """
     """Destroys the cluster.
 
     """

-    master = self.cfg.GetMasterNode()
+    master_params = self.cfg.GetMasterNetworkParameters()

 
     # Run post hooks on master node before it's removed
 
     # Run post hooks on master node before it's removed

-    _RunPostHook(self, master)
+    _RunPostHook(self, master_params.name)

 
 

-    result = self.rpc.call_node_stop_master(master, False)
+    ems = self.cfg.GetUseExternalMipScript()
+    result = self.rpc.call_node_deactivate_master_ip(master_params.name,
+                                                     master_params, ems)

     result.Raise("Could not disable the master role")
 
     result.Raise("Could not disable the master role")
 

-    return master
+    return master_params.name

 
 
 def _VerifyCertificate(filename):
 
 
 def _VerifyCertificate(filename):

@@ -1371,7 +1401,7 @@ def _VerifyCertificate(filename):
   try:
     cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
                                            utils.ReadFile(filename))
   try:
     cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
                                            utils.ReadFile(filename))

-  except Exception, err: # pylint: disable-msg=W0703
+  except Exception, err: # pylint: disable=W0703

     return (LUClusterVerifyConfig.ETYPE_ERROR,
             "Failed to load X509 certificate %s: %s" % (filename, err))
 
     return (LUClusterVerifyConfig.ETYPE_ERROR,
             "Failed to load X509 certificate %s: %s" % (filename, err))
 

@@ -1433,39 +1463,6 @@ class _VerifyErrors(object):
   self.op and self._feedback_fn to be available.)
 
   """
   self.op and self._feedback_fn to be available.)
 
   """

-  TCLUSTER = "cluster"
-  TNODE = "node"
-  TINSTANCE = "instance"
-
-  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
-  ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
-  ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK")
-  ECLUSTERDANGLINGNODES = (TNODE, "ECLUSTERDANGLINGNODES")
-  ECLUSTERDANGLINGINST = (TNODE, "ECLUSTERDANGLINGINST")
-  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
-  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
-  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
-  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
-  EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
-  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
-  EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
-  ENODEDRBD = (TNODE, "ENODEDRBD")
-  ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
-  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
-  ENODEHOOKS = (TNODE, "ENODEHOOKS")
-  ENODEHV = (TNODE, "ENODEHV")
-  ENODELVM = (TNODE, "ENODELVM")
-  ENODEN1 = (TNODE, "ENODEN1")
-  ENODENET = (TNODE, "ENODENET")
-  ENODEOS = (TNODE, "ENODEOS")
-  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
-  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
-  ENODERPC = (TNODE, "ENODERPC")
-  ENODESSH = (TNODE, "ENODESSH")
-  ENODEVERSION = (TNODE, "ENODEVERSION")
-  ENODESETUP = (TNODE, "ENODESETUP")
-  ENODETIME = (TNODE, "ENODETIME")
-  ENODEOOBPATH = (TNODE, "ENODEOOBPATH")

 
   ETYPE_FIELD = "code"
   ETYPE_ERROR = "ERROR"
 
   ETYPE_FIELD = "code"
   ETYPE_ERROR = "ERROR"

@@ -1481,12 +1478,12 @@ class _VerifyErrors(object):
 
     """
     ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
 
     """
     ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)

-    itype, etxt = ecode
+    itype, etxt, _ = ecode

     # first complete the msg
     if args:
       msg = msg % args
     # then format the whole message
     # first complete the msg
     if args:
       msg = msg % args
     # then format the whole message

-    if self.op.error_codes: # This is a mix-in. pylint: disable-msg=E1101
+    if self.op.error_codes: # This is a mix-in. pylint: disable=E1101

       msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
     else:
       if item:
       msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
     else:
       if item:

@@ -1495,16 +1492,24 @@ class _VerifyErrors(object):
         item = ""
       msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
     # and finally report it via the feedback_fn
         item = ""
       msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
     # and finally report it via the feedback_fn

-    self._feedback_fn("  - %s" % msg) # Mix-in. pylint: disable-msg=E1101
+    self._feedback_fn("  - %s" % msg) # Mix-in. pylint: disable=E1101

 
 

-  def _ErrorIf(self, cond, *args, **kwargs):
+  def _ErrorIf(self, cond, ecode, *args, **kwargs):

     """Log an error message if the passed condition is True.
 
     """
     cond = (bool(cond)
     """Log an error message if the passed condition is True.
 
     """
     cond = (bool(cond)

-            or self.op.debug_simulate_errors) # pylint: disable-msg=E1101
+            or self.op.debug_simulate_errors) # pylint: disable=E1101
+
+    # If the error code is in the list of ignored errors, demote the error to a
+    # warning
+    (_, etxt, _) = ecode
+    if etxt in self.op.ignore_errors:     # pylint: disable=E1101
+      kwargs[self.ETYPE_FIELD] = self.ETYPE_WARNING
+

     if cond:
     if cond:

-      self._Error(*args, **kwargs)
+      self._Error(ecode, *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
     # 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

@@ -1529,17 +1534,20 @@ class LUClusterVerify(NoHooksLU):
       groups = self.cfg.GetNodeGroupList()
 
       # Verify global configuration
       groups = self.cfg.GetNodeGroupList()
 
       # Verify global configuration

-      jobs.append([opcodes.OpClusterVerifyConfig()])
+      jobs.append([
+        opcodes.OpClusterVerifyConfig(ignore_errors=self.op.ignore_errors)
+        ])

 
       # Always depend on global verification
       depends_fn = lambda: [(-len(jobs), [])]
 
     jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,
 
       # Always depend on global verification
       depends_fn = lambda: [(-len(jobs), [])]
 
     jobs.extend([opcodes.OpClusterVerifyGroup(group_name=group,

-                                              depends=depends_fn())]
+                                            ignore_errors=self.op.ignore_errors,
+                                            depends=depends_fn())]

                 for group in groups)
 
     # Fix up all parameters
                 for group in groups)
 
     # Fix up all parameters

-    for op in itertools.chain(*jobs): # pylint: disable-msg=W0142
+    for op in itertools.chain(*jobs): # pylint: disable=W0142

       op.debug_simulate_errors = self.op.debug_simulate_errors
       op.verbose = self.op.verbose
       op.error_codes = self.op.error_codes
       op.debug_simulate_errors = self.op.debug_simulate_errors
       op.verbose = self.op.verbose
       op.error_codes = self.op.error_codes

@@ -1569,11 +1577,12 @@ class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
         utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
         hv_class.CheckParameterSyntax(hv_params)
       except errors.GenericError, err:
         utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
         hv_class.CheckParameterSyntax(hv_params)
       except errors.GenericError, err:

-        self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
+        self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg % str(err))

 
   def ExpandNames(self):
     # Information can be safely retrieved as the BGL is acquired in exclusive
     # mode
 
   def ExpandNames(self):
     # Information can be safely retrieved as the BGL is acquired in exclusive
     # mode

+    assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER)

     self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
     self.all_node_info = self.cfg.GetAllNodesInfo()
     self.all_inst_info = self.cfg.GetAllInstancesInfo()
     self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
     self.all_node_info = self.cfg.GetAllNodesInfo()
     self.all_inst_info = self.cfg.GetAllInstancesInfo()

@@ -1589,13 +1598,13 @@ class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
     feedback_fn("* Verifying cluster config")
 
     for msg in self.cfg.VerifyConfig():
     feedback_fn("* Verifying cluster config")
 
     for msg in self.cfg.VerifyConfig():

-      self._ErrorIf(True, self.ECLUSTERCFG, None, msg)
+      self._ErrorIf(True, constants.CV_ECLUSTERCFG, None, msg)

 
     feedback_fn("* Verifying cluster certificate files")
 
     for cert_filename in constants.ALL_CERT_FILES:
       (errcode, msg) = _VerifyCertificate(cert_filename)
 
     feedback_fn("* Verifying cluster certificate files")
 
     for cert_filename in constants.ALL_CERT_FILES:
       (errcode, msg) = _VerifyCertificate(cert_filename)

-      self._ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
+      self._ErrorIf(errcode, constants.CV_ECLUSTERCERT, None, msg, code=errcode)

 
     feedback_fn("* Verifying hypervisor parameters")
 
 
     feedback_fn("* Verifying hypervisor parameters")
 

@@ -1627,11 +1636,13 @@ class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
                                                 ["no instances"])))
         for node in dangling_nodes]
 
                                                 ["no instances"])))
         for node in dangling_nodes]
 

-    self._ErrorIf(bool(dangling_nodes), self.ECLUSTERDANGLINGNODES, None,
+    self._ErrorIf(bool(dangling_nodes), constants.CV_ECLUSTERDANGLINGNODES,
+                  None,

                   "the following nodes (and their instances) belong to a non"
                   " existing group: %s", utils.CommaJoin(pretty_dangling))
 
                   "the following nodes (and their instances) belong to a non"
                   " existing group: %s", utils.CommaJoin(pretty_dangling))
 

-    self._ErrorIf(bool(no_node_instances), self.ECLUSTERDANGLINGINST, None,
+    self._ErrorIf(bool(no_node_instances), constants.CV_ECLUSTERDANGLINGINST,
+                  None,

                   "the following instances have a non-existing primary-node:"
                   " %s", utils.CommaJoin(no_node_instances))
 
                   "the following instances have a non-existing primary-node:"
                   " %s", utils.CommaJoin(no_node_instances))
 

@@ -1733,7 +1744,10 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       self.needed_locks[locking.LEVEL_NODE] = nodes
 
   def CheckPrereq(self):
       self.needed_locks[locking.LEVEL_NODE] = nodes
 
   def CheckPrereq(self):

-    group_nodes = set(self.cfg.GetNodeGroup(self.group_uuid).members)
+    assert self.group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
+    self.group_info = self.cfg.GetNodeGroup(self.group_uuid)
+
+    group_nodes = set(self.group_info.members)

     group_instances = self.cfg.GetNodeGroupInstances(self.group_uuid)
 
     unlocked_nodes = \
     group_instances = self.cfg.GetNodeGroupInstances(self.group_uuid)
 
     unlocked_nodes = \

@@ -1797,11 +1811,11 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     # main result, nresult should be a non-empty dict
     test = not nresult or not isinstance(nresult, dict)
 
     # main result, nresult should be a non-empty dict
     test = not nresult or not isinstance(nresult, dict)

-    _ErrorIf(test, self.ENODERPC, node,
+    _ErrorIf(test, constants.CV_ENODERPC, node,

                   "unable to verify node: no data returned")
     if test:
       return False
                   "unable to verify node: no data returned")
     if test:
       return False

@@ -1812,13 +1826,13 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     test = not (remote_version and
                 isinstance(remote_version, (list, tuple)) and
                 len(remote_version) == 2)
     test = not (remote_version and
                 isinstance(remote_version, (list, tuple)) and
                 len(remote_version) == 2)

-    _ErrorIf(test, self.ENODERPC, node,
+    _ErrorIf(test, constants.CV_ENODERPC, node,

              "connection to node returned invalid data")
     if test:
       return False
 
     test = local_version != remote_version[0]
              "connection to node returned invalid data")
     if test:
       return False
 
     test = local_version != remote_version[0]

-    _ErrorIf(test, self.ENODEVERSION, node,
+    _ErrorIf(test, constants.CV_ENODEVERSION, node,

              "incompatible protocol versions: master %s,"
              " node %s", local_version, remote_version[0])
     if test:
              "incompatible protocol versions: master %s,"
              " node %s", local_version, remote_version[0])
     if test:

@@ -1828,7 +1842,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     # full package version
     self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
 
     # full package version
     self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],

-                  self.ENODEVERSION, node,
+                  constants.CV_ENODEVERSION, node,

                   "software version mismatch: master %s, node %s",
                   constants.RELEASE_VERSION, remote_version[1],
                   code=self.ETYPE_WARNING)
                   "software version mismatch: master %s, node %s",
                   constants.RELEASE_VERSION, remote_version[1],
                   code=self.ETYPE_WARNING)

@@ -1837,19 +1851,19 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     if ninfo.vm_capable and isinstance(hyp_result, dict):
       for hv_name, hv_result in hyp_result.iteritems():
         test = hv_result is not None
     if ninfo.vm_capable and isinstance(hyp_result, dict):
       for hv_name, hv_result in hyp_result.iteritems():
         test = hv_result is not None

-        _ErrorIf(test, self.ENODEHV, node,
+        _ErrorIf(test, constants.CV_ENODEHV, node,

                  "hypervisor %s verify failure: '%s'", hv_name, hv_result)
 
     hvp_result = nresult.get(constants.NV_HVPARAMS, None)
     if ninfo.vm_capable and isinstance(hvp_result, list):
       for item, hv_name, hv_result in hvp_result:
                  "hypervisor %s verify failure: '%s'", hv_name, hv_result)
 
     hvp_result = nresult.get(constants.NV_HVPARAMS, None)
     if ninfo.vm_capable and isinstance(hvp_result, list):
       for item, hv_name, hv_result in hvp_result:

-        _ErrorIf(True, self.ENODEHV, node,
+        _ErrorIf(True, constants.CV_ENODEHV, node,

                  "hypervisor %s parameter verify failure (source %s): %s",
                  hv_name, item, hv_result)
 
     test = nresult.get(constants.NV_NODESETUP,
                        ["Missing NODESETUP results"])
                  "hypervisor %s parameter verify failure (source %s): %s",
                  hv_name, item, hv_result)
 
     test = nresult.get(constants.NV_NODESETUP,
                        ["Missing NODESETUP results"])

-    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
+    _ErrorIf(test, constants.CV_ENODESETUP, node, "node setup error: %s",

              "; ".join(test))
 
     return True
              "; ".join(test))
 
     return True

@@ -1866,13 +1880,13 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     ntime = nresult.get(constants.NV_TIME, None)
     try:
       ntime_merged = utils.MergeTime(ntime)
     except (ValueError, TypeError):
 
     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")
+      _ErrorIf(True, constants.CV_ENODETIME, node, "Node returned invalid time")

       return
 
     if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
       return
 
     if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):

@@ -1882,7 +1896,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     else:
       ntime_diff = None
 
     else:
       ntime_diff = None
 

-    _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
+    _ErrorIf(ntime_diff is not None, constants.CV_ENODETIME, node,

              "Node time diverges by at least %s from master node time",
              ntime_diff)
 
              "Node time diverges by at least %s from master node time",
              ntime_diff)
 

@@ -1899,29 +1913,30 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       return
 
     node = ninfo.name
       return
 
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     # checks vg existence and size > 20G
     vglist = nresult.get(constants.NV_VGLIST, None)
     test = not vglist
 
     # 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")
+    _ErrorIf(test, constants.CV_ENODELVM, node, "unable to check volume groups")

     if not test:
       vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
                                             constants.MIN_VG_SIZE)
     if not test:
       vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
                                             constants.MIN_VG_SIZE)

-      _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
+      _ErrorIf(vgstatus, constants.CV_ENODELVM, node, vgstatus)

 
     # check pv names
     pvlist = nresult.get(constants.NV_PVLIST, None)
     test = pvlist is None
 
     # 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")
+    _ErrorIf(test, constants.CV_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
     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)
+        _ErrorIf(test, constants.CV_ENODELVM, node,
+                 "Invalid character ':' in PV '%s' of VG '%s'",
+                 pvname, owner_vg)

 
   def _VerifyNodeBridges(self, ninfo, nresult, bridges):
     """Check the node bridges.
 
   def _VerifyNodeBridges(self, ninfo, nresult, bridges):
     """Check the node bridges.

@@ -1936,15 +1951,35 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       return
 
     node = ninfo.name
       return
 
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     missing = nresult.get(constants.NV_BRIDGES, None)
     test = not isinstance(missing, list)
 
     missing = nresult.get(constants.NV_BRIDGES, None)
     test = not isinstance(missing, list)

-    _ErrorIf(test, self.ENODENET, node,
+    _ErrorIf(test, constants.CV_ENODENET, node,

              "did not return valid bridge information")
     if not test:
              "did not return valid bridge information")
     if not test:

-      _ErrorIf(bool(missing), self.ENODENET, node, "missing bridges: %s" %
-               utils.CommaJoin(sorted(missing)))
+      _ErrorIf(bool(missing), constants.CV_ENODENET, node,
+               "missing bridges: %s" % utils.CommaJoin(sorted(missing)))
+
+  def _VerifyNodeUserScripts(self, ninfo, nresult):
+    """Check the results of user scripts presence and executability on the node
+
+    @type ninfo: L{objects.Node}
+    @param ninfo: the node to check
+    @param nresult: the remote results for the node
+
+    """
+    node = ninfo.name
+
+    test = not constants.NV_USERSCRIPTS in nresult
+    self._ErrorIf(test, constants.CV_ENODEUSERSCRIPTS, node,
+                  "did not return user scripts information")
+
+    broken_scripts = nresult.get(constants.NV_USERSCRIPTS, None)
+    if not test:
+      self._ErrorIf(broken_scripts, constants.CV_ENODEUSERSCRIPTS, node,
+                    "user scripts not present or not executable: %s" %
+                    utils.CommaJoin(sorted(broken_scripts)))

 
   def _VerifyNodeNetwork(self, ninfo, nresult):
     """Check the node network connectivity results.
 
   def _VerifyNodeNetwork(self, ninfo, nresult):
     """Check the node network connectivity results.

@@ -1955,30 +1990,30 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     test = constants.NV_NODELIST not in nresult
 
     test = constants.NV_NODELIST not in nresult

-    _ErrorIf(test, self.ENODESSH, node,
+    _ErrorIf(test, constants.CV_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():
              "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,
+          _ErrorIf(True, constants.CV_ENODESSH, node,

                    "ssh communication with node '%s': %s", a_node, a_msg)
 
     test = constants.NV_NODENETTEST not in nresult
                    "ssh communication with node '%s': %s", a_node, a_msg)
 
     test = constants.NV_NODENETTEST not in nresult

-    _ErrorIf(test, self.ENODENET, node,
+    _ErrorIf(test, constants.CV_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:
              "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,
+          _ErrorIf(True, constants.CV_ENODENET, node,

                    "tcp communication with node '%s': %s",
                    anode, nresult[constants.NV_NODENETTEST][anode])
 
     test = constants.NV_MASTERIP not in nresult
                    "tcp communication with node '%s': %s",
                    anode, nresult[constants.NV_NODENETTEST][anode])
 
     test = constants.NV_MASTERIP not in nresult

-    _ErrorIf(test, self.ENODENET, node,
+    _ErrorIf(test, constants.CV_ENODENET, node,

              "node hasn't returned node master IP reachability data")
     if not test:
       if not nresult[constants.NV_MASTERIP]:
              "node hasn't returned node master IP reachability data")
     if not test:
       if not nresult[constants.NV_MASTERIP]:

@@ -1986,7 +2021,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
           msg = "the master node cannot reach the master IP (not configured?)"
         else:
           msg = "cannot reach the master IP"
           msg = "the master node cannot reach the master IP (not configured?)"
         else:
           msg = "cannot reach the master IP"

-        _ErrorIf(True, self.ENODENET, node, msg)
+        _ErrorIf(True, constants.CV_ENODENET, node, msg)

 
   def _VerifyInstance(self, instance, instanceconfig, node_image,
                       diskstatus):
 
   def _VerifyInstance(self, instance, instanceconfig, node_image,
                       diskstatus):

@@ -1996,7 +2031,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     available on the instance's node.
 
     """
     available on the instance's node.
 
     """

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

     node_current = instanceconfig.primary_node
 
     node_vol_should = {}
     node_current = instanceconfig.primary_node
 
     node_vol_should = {}

@@ -2009,13 +2044,13 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         continue
       for volume in node_vol_should[node]:
         test = volume not in n_img.volumes
         continue
       for volume in node_vol_should[node]:
         test = volume not in n_img.volumes

-        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
+        _ErrorIf(test, constants.CV_EINSTANCEMISSINGDISK, instance,

                  "volume %s missing on node %s", volume, node)
 
                  "volume %s missing on node %s", volume, node)
 

-    if instanceconfig.admin_up:
+    if instanceconfig.admin_state == constants.ADMINST_UP:

       pri_img = node_image[node_current]
       test = instance not in pri_img.instances and not pri_img.offline
       pri_img = node_image[node_current]
       test = instance not in pri_img.instances and not pri_img.offline

-      _ErrorIf(test, self.EINSTANCEDOWN, instance,
+      _ErrorIf(test, constants.CV_EINSTANCEDOWN, instance,

                "instance not running on its primary node %s",
                node_current)
 
                "instance not running on its primary node %s",
                node_current)
 

@@ -2028,13 +2063,14 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       # node here
       snode = node_image[nname]
       bad_snode = snode.ghost or snode.offline
       # node here
       snode = node_image[nname]
       bad_snode = snode.ghost or snode.offline

-      _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
-               self.EINSTANCEFAULTYDISK, instance,
+      _ErrorIf(instanceconfig.admin_state == constants.ADMINST_UP and
+               not success and not bad_snode,
+               constants.CV_EINSTANCEFAULTYDISK, instance,

                "couldn't retrieve status for disk/%s on %s: %s",
                idx, nname, bdev_status)
                "couldn't retrieve status for disk/%s on %s: %s",
                idx, nname, bdev_status)

-      _ErrorIf((instanceconfig.admin_up and success and
-                bdev_status.ldisk_status == constants.LDS_FAULTY),
-               self.EINSTANCEFAULTYDISK, instance,
+      _ErrorIf((instanceconfig.admin_state == constants.ADMINST_UP and
+                success and bdev_status.ldisk_status == constants.LDS_FAULTY),
+               constants.CV_EINSTANCEFAULTYDISK, instance,

                "disk/%s on %s is faulty", idx, nname)
 
   def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
                "disk/%s on %s is faulty", idx, nname)
 
   def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):

@@ -2055,7 +2091,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         test = ((node not in node_vol_should or
                 volume not in node_vol_should[node]) and
                 not reserved.Matches(volume))
         test = ((node not in node_vol_should or
                 volume not in node_vol_should[node]) and
                 not reserved.Matches(volume))

-        self._ErrorIf(test, self.ENODEORPHANLV, node,
+        self._ErrorIf(test, constants.CV_ENODEORPHANLV, node,

                       "volume %s is unknown", volume)
 
   def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
                       "volume %s is unknown", volume)
 
   def _VerifyNPlusOneMemory(self, node_image, instance_cfg):

@@ -2088,14 +2124,14 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
           if bep[constants.BE_AUTO_BALANCE]:
             needed_mem += bep[constants.BE_MEMORY]
         test = n_img.mfree < needed_mem
           if bep[constants.BE_AUTO_BALANCE]:
             needed_mem += bep[constants.BE_MEMORY]
         test = n_img.mfree < needed_mem

-        self._ErrorIf(test, self.ENODEN1, node,
+        self._ErrorIf(test, constants.CV_ENODEN1, node,

                       "not enough memory to accomodate instance failovers"
                       " should node %s fail (%dMiB needed, %dMiB available)",
                       prinode, needed_mem, n_img.mfree)
 
   @classmethod
   def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,
                       "not enough memory to accomodate instance failovers"
                       " should node %s fail (%dMiB needed, %dMiB available)",
                       prinode, needed_mem, n_img.mfree)
 
   @classmethod
   def _VerifyFiles(cls, errorif, nodeinfo, master_node, all_nvinfo,

-                   (files_all, files_all_opt, files_mc, files_vm)):
+                   (files_all, files_opt, files_mc, files_vm)):

     """Verifies file checksums collected from all nodes.
 
     @param errorif: Callback for reporting errors
     """Verifies file checksums collected from all nodes.
 
     @param errorif: Callback for reporting errors

@@ -2104,26 +2140,33 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     @param all_nvinfo: RPC results
 
     """
     @param all_nvinfo: RPC results
 
     """

-    node_names = frozenset(node.name for node in nodeinfo if not node.offline)
+    # Define functions determining which nodes to consider for a file
+    files2nodefn = [
+      (files_all, None),
+      (files_mc, lambda node: (node.master_candidate or
+                               node.name == master_node)),
+      (files_vm, lambda node: node.vm_capable),
+      ]

 
 

-    assert master_node in node_names
-    assert (len(files_all | files_all_opt | files_mc | files_vm) ==
-            sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
-           "Found file listed in more than one file list"
+    # Build mapping from filename to list of nodes which should have the file
+    nodefiles = {}
+    for (files, fn) in files2nodefn:
+      if fn is None:
+        filenodes = nodeinfo
+      else:
+        filenodes = filter(fn, nodeinfo)
+      nodefiles.update((filename,
+                        frozenset(map(operator.attrgetter("name"), filenodes)))
+                       for filename in files)

 
 

-    # Define functions determining which nodes to consider for a file
-    file2nodefn = dict([(filename, fn)
-      for (files, fn) in [(files_all, None),
-                          (files_all_opt, None),
-                          (files_mc, lambda node: (node.master_candidate or
-                                                   node.name == master_node)),
-                          (files_vm, lambda node: node.vm_capable)]
-      for filename in files])
+    assert set(nodefiles) == (files_all | files_mc | files_vm)

 
 

-    fileinfo = dict((filename, {}) for filename in file2nodefn.keys())
+    fileinfo = dict((filename, {}) for filename in nodefiles)
+    ignore_nodes = set()

 
     for node in nodeinfo:
       if node.offline:
 
     for node in nodeinfo:
       if node.offline:

+        ignore_nodes.add(node.name)

         continue
 
       nresult = all_nvinfo[node.name]
         continue
 
       nresult = all_nvinfo[node.name]

@@ -2134,16 +2177,16 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         node_files = nresult.payload.get(constants.NV_FILELIST, None)
 
       test = not (node_files and isinstance(node_files, dict))
         node_files = nresult.payload.get(constants.NV_FILELIST, None)
 
       test = not (node_files and isinstance(node_files, dict))

-      errorif(test, cls.ENODEFILECHECK, node.name,
+      errorif(test, constants.CV_ENODEFILECHECK, node.name,

               "Node did not return file checksum data")
       if test:
               "Node did not return file checksum data")
       if test:

+        ignore_nodes.add(node.name)

         continue
 
         continue
 

+      # Build per-checksum mapping from filename to nodes having it

       for (filename, checksum) in node_files.items():
       for (filename, checksum) in node_files.items():

-        # Check if the file should be considered for a node
-        fn = file2nodefn[filename]
-        if fn is None or fn(node):
-          fileinfo[filename].setdefault(checksum, set()).add(node.name)
+        assert filename in nodefiles
+        fileinfo[filename].setdefault(checksum, set()).add(node.name)

 
     for (filename, checksums) in fileinfo.items():
       assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"
 
     for (filename, checksums) in fileinfo.items():
       assert compat.all(len(i) > 10 for i in checksums), "Invalid checksum"

@@ -2151,23 +2194,32 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       # Nodes having the file
       with_file = frozenset(node_name
                             for nodes in fileinfo[filename].values()
       # Nodes having the file
       with_file = frozenset(node_name
                             for nodes in fileinfo[filename].values()

-                            for node_name in nodes)
+                            for node_name in nodes) - ignore_nodes
+
+      expected_nodes = nodefiles[filename] - ignore_nodes

 
       # Nodes missing file
 
       # Nodes missing file

-      missing_file = node_names - with_file
+      missing_file = expected_nodes - with_file

 
 

-      if filename in files_all_opt:
+      if filename in files_opt:

         # All or no nodes
         # All or no nodes

-        errorif(missing_file and missing_file != node_names,
-                cls.ECLUSTERFILECHECK, None,
+        errorif(missing_file and missing_file != expected_nodes,
+                constants.CV_ECLUSTERFILECHECK, None,

                 "File %s is optional, but it must exist on all or no"
                 " nodes (not found on %s)",
                 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
       else:
                 "File %s is optional, but it must exist on all or no"
                 " nodes (not found on %s)",
                 filename, utils.CommaJoin(utils.NiceSort(missing_file)))
       else:

-        errorif(missing_file, cls.ECLUSTERFILECHECK, None,
+        errorif(missing_file, constants.CV_ECLUSTERFILECHECK, None,

                 "File %s is missing from node(s) %s", filename,
                 utils.CommaJoin(utils.NiceSort(missing_file)))
 
                 "File %s is missing from node(s) %s", filename,
                 utils.CommaJoin(utils.NiceSort(missing_file)))
 

+        # Warn if a node has a file it shouldn't
+        unexpected = with_file - expected_nodes
+        errorif(unexpected,
+                constants.CV_ECLUSTERFILECHECK, None,
+                "File %s should not exist on node(s) %s",
+                filename, utils.CommaJoin(utils.NiceSort(unexpected)))
+

       # See if there are multiple versions of the file
       test = len(checksums) > 1
       if test:
       # See if there are multiple versions of the file
       test = len(checksums) > 1
       if test:

@@ -2178,7 +2230,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       else:
         variants = []
 
       else:
         variants = []
 

-      errorif(test, cls.ECLUSTERFILECHECK, None,
+      errorif(test, constants.CV_ECLUSTERFILECHECK, None,

               "File %s found with %s different checksums (%s)",
               filename, len(checksums), "; ".join(variants))
 
               "File %s found with %s different checksums (%s)",
               filename, len(checksums), "; ".join(variants))
 

@@ -2196,27 +2248,27 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     if drbd_helper:
       helper_result = nresult.get(constants.NV_DRBDHELPER, None)
       test = (helper_result == None)
 
     if drbd_helper:
       helper_result = nresult.get(constants.NV_DRBDHELPER, None)
       test = (helper_result == None)

-      _ErrorIf(test, self.ENODEDRBDHELPER, node,
+      _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,

                "no drbd usermode helper returned")
       if helper_result:
         status, payload = helper_result
         test = not status
                "no drbd usermode helper returned")
       if helper_result:
         status, payload = helper_result
         test = not status

-        _ErrorIf(test, self.ENODEDRBDHELPER, node,
+        _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,

                  "drbd usermode helper check unsuccessful: %s", payload)
         test = status and (payload != drbd_helper)
                  "drbd usermode helper check unsuccessful: %s", payload)
         test = status and (payload != drbd_helper)

-        _ErrorIf(test, self.ENODEDRBDHELPER, node,
+        _ErrorIf(test, constants.CV_ENODEDRBDHELPER, node,

                  "wrong drbd usermode helper: %s", payload)
 
     # compute the DRBD minors
     node_drbd = {}
     for minor, instance in drbd_map[node].items():
       test = instance not in instanceinfo
                  "wrong drbd usermode helper: %s", payload)
 
     # compute the DRBD minors
     node_drbd = {}
     for minor, instance in drbd_map[node].items():
       test = instance not in instanceinfo

-      _ErrorIf(test, self.ECLUSTERCFG, None,
+      _ErrorIf(test, constants.CV_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
                "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

@@ -2225,12 +2277,13 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         node_drbd[minor] = (instance, False)
       else:
         instance = instanceinfo[instance]
         node_drbd[minor] = (instance, False)
       else:
         instance = instanceinfo[instance]

-        node_drbd[minor] = (instance.name, instance.admin_up)
+        node_drbd[minor] = (instance.name,
+                            instance.admin_state == constants.ADMINST_UP)

 
     # and now check them
     used_minors = nresult.get(constants.NV_DRBDLIST, [])
     test = not isinstance(used_minors, (tuple, list))
 
     # and now check them
     used_minors = nresult.get(constants.NV_DRBDLIST, [])
     test = not isinstance(used_minors, (tuple, list))

-    _ErrorIf(test, self.ENODEDRBD, node,
+    _ErrorIf(test, constants.CV_ENODEDRBD, node,

              "cannot parse drbd status file: %s", str(used_minors))
     if test:
       # we cannot check drbd status
              "cannot parse drbd status file: %s", str(used_minors))
     if test:
       # we cannot check drbd status

@@ -2238,11 +2291,11 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     for minor, (iname, must_exist) in node_drbd.items():
       test = minor not in used_minors and must_exist
 
     for minor, (iname, must_exist) in node_drbd.items():
       test = minor not in used_minors and must_exist

-      _ErrorIf(test, self.ENODEDRBD, node,
+      _ErrorIf(test, constants.CV_ENODEDRBD, node,

                "drbd minor %d of instance %s is not active", minor, iname)
     for minor in used_minors:
       test = minor not in node_drbd
                "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,
+      _ErrorIf(test, constants.CV_ENODEDRBD, node,

                "unallocated drbd minor %d is in use", minor)
 
   def _UpdateNodeOS(self, ninfo, nresult, nimg):
                "unallocated drbd minor %d is in use", minor)
 
   def _UpdateNodeOS(self, ninfo, nresult, nimg):

@@ -2255,14 +2308,14 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     remote_os = nresult.get(constants.NV_OSLIST, None)
     test = (not isinstance(remote_os, list) or
             not compat.all(isinstance(v, list) and len(v) == 7
                            for v in remote_os))
 
 
     remote_os = nresult.get(constants.NV_OSLIST, None)
     test = (not isinstance(remote_os, list) or
             not compat.all(isinstance(v, list) and len(v) == 7
                            for v in remote_os))
 

-    _ErrorIf(test, self.ENODEOS, node,
+    _ErrorIf(test, constants.CV_ENODEOS, node,

              "node hasn't returned valid OS data")
 
     nimg.os_fail = test
              "node hasn't returned valid OS data")
 
     nimg.os_fail = test

@@ -2296,7 +2349,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
 
 
     assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
 

@@ -2304,14 +2357,14 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     for os_name, os_data in nimg.oslist.items():
       assert os_data, "Empty OS status for OS %s?!" % os_name
       f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
     for os_name, os_data in nimg.oslist.items():
       assert os_data, "Empty OS status for OS %s?!" % os_name
       f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]

-      _ErrorIf(not f_status, self.ENODEOS, node,
+      _ErrorIf(not f_status, constants.CV_ENODEOS, node,

                "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
                "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)

-      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
+      _ErrorIf(len(os_data) > 1, constants.CV_ENODEOS, node,

                "OS '%s' has multiple entries (first one shadows the rest): %s",
                os_name, utils.CommaJoin([v[0] for v in os_data]))
       # comparisons with the 'base' image
       test = os_name not in base.oslist
                "OS '%s' has multiple entries (first one shadows the rest): %s",
                os_name, utils.CommaJoin([v[0] for v in os_data]))
       # comparisons with the 'base' image
       test = os_name not in base.oslist

-      _ErrorIf(test, self.ENODEOS, node,
+      _ErrorIf(test, constants.CV_ENODEOS, node,

                "Extra OS %s not present on reference node (%s)",
                os_name, base.name)
       if test:
                "Extra OS %s not present on reference node (%s)",
                os_name, base.name)
       if test:

@@ -2325,14 +2378,14 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
                          ("variants list", f_var, b_var),
                          ("parameters", beautify_params(f_param),
                           beautify_params(b_param))]:
                          ("variants list", f_var, b_var),
                          ("parameters", beautify_params(f_param),
                           beautify_params(b_param))]:

-        _ErrorIf(a != b, self.ENODEOS, node,
+        _ErrorIf(a != b, constants.CV_ENODEOS, node,

                  "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
                  kind, os_name, base.name,
                  utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
 
     # check any missing OSes
     missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
                  "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
                  kind, os_name, base.name,
                  utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
 
     # check any missing OSes
     missing = set(base.oslist.keys()).difference(nimg.oslist.keys())

-    _ErrorIf(missing, self.ENODEOS, node,
+    _ErrorIf(missing, constants.CV_ENODEOS, node,

              "OSes present on reference node %s but missing on this node: %s",
              base.name, utils.CommaJoin(missing))
 
              "OSes present on reference node %s but missing on this node: %s",
              base.name, utils.CommaJoin(missing))
 

@@ -2350,7 +2403,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     if ((ninfo.master_candidate or ninfo.master_capable) and
         constants.NV_OOB_PATHS in nresult):
       for path_result in nresult[constants.NV_OOB_PATHS]:
     if ((ninfo.master_candidate or ninfo.master_capable) and
         constants.NV_OOB_PATHS in nresult):
       for path_result in nresult[constants.NV_OOB_PATHS]:

-        self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
+        self._ErrorIf(path_result, constants.CV_ENODEOOBPATH, node, path_result)

 
   def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
     """Verifies and updates the node volume data.
 
   def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
     """Verifies and updates the node volume data.

@@ -2366,17 +2419,18 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     nimg.lvm_fail = True
     lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
     if vg_name is None:
       pass
     elif isinstance(lvdata, basestring):
 
     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",
+      _ErrorIf(True, constants.CV_ENODELVM, node, "LVM problem on node: %s",

                utils.SafeEncode(lvdata))
     elif not isinstance(lvdata, dict):
                utils.SafeEncode(lvdata))
     elif not isinstance(lvdata, dict):

-      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
+      _ErrorIf(True, constants.CV_ENODELVM, node,
+               "rpc call to node failed (lvlist)")

     else:
       nimg.volumes = lvdata
       nimg.lvm_fail = False
     else:
       nimg.volumes = lvdata
       nimg.lvm_fail = False

@@ -2396,8 +2450,9 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     """
     idata = nresult.get(constants.NV_INSTANCELIST, None)
     test = not isinstance(idata, list)
     """
     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)))
+    self._ErrorIf(test, constants.CV_ENODEHV, ninfo.name,
+                  "rpc call to node failed (instancelist): %s",
+                  utils.SafeEncode(str(idata)))

     if test:
       nimg.hyp_fail = True
     else:
     if test:
       nimg.hyp_fail = True
     else:

@@ -2414,31 +2469,32 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
 
     """
     node = ninfo.name

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=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
 
     # 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)")
+    _ErrorIf(test, constants.CV_ENODEHV, node,
+             "rpc call to node failed (hvinfo)")

     if not test:
       try:
         nimg.mfree = int(hv_info["memory_free"])
       except (ValueError, TypeError):
     if not test:
       try:
         nimg.mfree = int(hv_info["memory_free"])
       except (ValueError, TypeError):

-        _ErrorIf(True, self.ENODERPC, node,
+        _ErrorIf(True, constants.CV_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])
                  "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,
+      _ErrorIf(test, constants.CV_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):
                "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,
+          _ErrorIf(True, constants.CV_ENODERPC, node,

                    "node returned invalid LVM info, check LVM status")
 
   def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
                    "node returned invalid LVM info, check LVM status")
 
   def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):

@@ -2456,7 +2512,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         list of tuples (success, payload)
 
     """
         list of tuples (success, payload)
 
     """

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

 
     node_disks = {}
     node_disks_devonly = {}
 
     node_disks = {}
     node_disks_devonly = {}

@@ -2505,7 +2561,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         data = len(disks) * [(False, "node offline")]
       else:
         msg = nres.fail_msg
         data = len(disks) * [(False, "node offline")]
       else:
         msg = nres.fail_msg

-        _ErrorIf(msg, self.ENODERPC, nname,
+        _ErrorIf(msg, constants.CV_ENODERPC, nname,

                  "while getting disk information: %s", msg)
         if msg:
           # No data from this node
                  "while getting disk information: %s", msg)
         if msg:
           # No data from this node

@@ -2538,6 +2594,40 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     return instdisk
 
 
     return instdisk
 

+  @staticmethod
+  def _SshNodeSelector(group_uuid, all_nodes):
+    """Create endless iterators for all potential SSH check hosts.
+
+    """
+    nodes = [node for node in all_nodes
+             if (node.group != group_uuid and
+                 not node.offline)]
+    keyfunc = operator.attrgetter("group")
+
+    return map(itertools.cycle,
+               [sorted(map(operator.attrgetter("name"), names))
+                for _, names in itertools.groupby(sorted(nodes, key=keyfunc),
+                                                  keyfunc)])
+
+  @classmethod
+  def _SelectSshCheckNodes(cls, group_nodes, group_uuid, all_nodes):
+    """Choose which nodes should talk to which other nodes.
+
+    We will make nodes contact all nodes in their group, and one node from
+    every other group.
+
+    @warning: This algorithm has a known issue if one node group is much
+      smaller than others (e.g. just one node). In such a case all other
+      nodes will talk to the single node.
+
+    """
+    online_nodes = sorted(node.name for node in group_nodes if not node.offline)
+    sel = cls._SshNodeSelector(group_uuid, all_nodes)
+
+    return (online_nodes,
+            dict((name, sorted([i.next() for i in sel]))
+                 for name in online_nodes))
+

   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.
 

@@ -2564,7 +2654,8 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     """Verify integrity of the node group, performing various test on nodes.
 
     """
     """Verify integrity of the node group, performing various test on nodes.
 
     """

-    # This method has too many local variables. pylint: disable-msg=R0914
+    # This method has too many local variables. pylint: disable=R0914
+    feedback_fn("* Verifying group '%s'" % self.group_info.name)

 
     if not self.my_node_names:
       # empty node group
 
     if not self.my_node_names:
       # empty node group

@@ -2572,7 +2663,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       return True
 
     self.bad = False
       return True
 
     self.bad = False

-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103

     verbose = self.op.verbose
     self._feedback_fn = feedback_fn
 
     verbose = self.op.verbose
     self._feedback_fn = feedback_fn
 

@@ -2585,6 +2676,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     i_non_redundant = [] # Non redundant instances
     i_non_a_balanced = [] # Non auto-balanced instances
 
     i_non_redundant = [] # Non redundant instances
     i_non_a_balanced = [] # Non auto-balanced instances

+    i_offline = 0 # Count of offline instances

     n_offline = 0 # Count of offline nodes
     n_drained = 0 # Count of nodes being drained
     node_vol_should = {}
     n_offline = 0 # Count of offline nodes
     n_drained = 0 # Count of nodes being drained
     node_vol_should = {}

@@ -2600,25 +2692,18 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
 
 
     feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
 

-    # We will make nodes contact all nodes in their group, and one node from
-    # every other group.
-    # TODO: should it be a *random* node, different every time?
-    online_nodes = [node.name for node in node_data_list if not node.offline]
-    other_group_nodes = {}
-
-    for name in sorted(self.all_node_info):
-      node = self.all_node_info[name]
-      if (node.group not in other_group_nodes
-          and node.group != self.group_uuid
-          and not node.offline):
-        other_group_nodes[node.group] = node.name
+    user_scripts = []
+    if self.cfg.GetUseExternalMipScript():
+      user_scripts.append(constants.EXTERNAL_MASTER_SETUP_SCRIPT)

 
     node_verify_param = {
       constants.NV_FILELIST:
         utils.UniqueSequence(filename
                              for files in filemap
                              for filename in files),
 
     node_verify_param = {
       constants.NV_FILELIST:
         utils.UniqueSequence(filename
                              for files in filemap
                              for filename in files),

-      constants.NV_NODELIST: online_nodes + other_group_nodes.values(),
+      constants.NV_NODELIST:
+        self._SelectSshCheckNodes(node_data_list, self.group_uuid,
+                                  self.all_node_info.values()),

       constants.NV_HYPERVISOR: hypervisors,
       constants.NV_HVPARAMS:
         _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
       constants.NV_HYPERVISOR: hypervisors,
       constants.NV_HVPARAMS:
         _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),

@@ -2633,6 +2718,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       constants.NV_MASTERIP: (master_node, master_ip),
       constants.NV_OSLIST: None,
       constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
       constants.NV_MASTERIP: (master_node, master_ip),
       constants.NV_OSLIST: None,
       constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),

+      constants.NV_USERSCRIPTS: user_scripts,

       }
 
     if vg_name is not None:
       }
 
     if vg_name is not None:

@@ -2780,7 +2866,8 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         feedback_fn("* Verifying node %s (%s)" % (node, ntype))
 
       msg = all_nvinfo[node].fail_msg
         feedback_fn("* Verifying node %s (%s)" % (node, ntype))
 
       msg = all_nvinfo[node].fail_msg

-      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
+      _ErrorIf(msg, constants.CV_ENODERPC, node, "while contacting node: %s",
+               msg)

       if msg:
         nimg.rpc_fail = True
         continue
       if msg:
         nimg.rpc_fail = True
         continue

@@ -2790,6 +2877,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       nimg.call_ok = self._VerifyNode(node_i, nresult)
       self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
       self._VerifyNodeNetwork(node_i, nresult)
       nimg.call_ok = self._VerifyNode(node_i, nresult)
       self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
       self._VerifyNodeNetwork(node_i, nresult)

+      self._VerifyNodeUserScripts(node_i, nresult)

       self._VerifyOob(node_i, nresult)
 
       if nimg.vm_capable:
       self._VerifyOob(node_i, nresult)
 
       if nimg.vm_capable:

@@ -2814,10 +2902,16 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         non_primary_inst = set(nimg.instances).difference(nimg.pinst)
 
         for inst in non_primary_inst:
         non_primary_inst = set(nimg.instances).difference(nimg.pinst)
 
         for inst in non_primary_inst:

+          # FIXME: investigate best way to handle offline insts
+          if inst.admin_state == constants.ADMINST_OFFLINE:
+            if verbose:
+              feedback_fn("* Skipping offline instance %s" % inst.name)
+            i_offline += 1
+            continue

           test = inst in self.all_inst_info
           test = inst in self.all_inst_info

-          _ErrorIf(test, self.EINSTANCEWRONGNODE, inst,
+          _ErrorIf(test, constants.CV_EINSTANCEWRONGNODE, inst,

                    "instance should not run on node %s", node_i.name)
                    "instance should not run on node %s", node_i.name)

-          _ErrorIf(not test, self.ENODEORPHANINSTANCE, node_i.name,
+          _ErrorIf(not test, constants.CV_ENODEORPHANINSTANCE, node_i.name,

                    "node is running unknown instance %s", inst)
 
     for node, result in extra_lv_nvinfo.items():
                    "node is running unknown instance %s", inst)
 
     for node, result in extra_lv_nvinfo.items():

@@ -2836,11 +2930,12 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       pnode = inst_config.primary_node
       pnode_img = node_image[pnode]
       _ErrorIf(pnode_img.rpc_fail and not pnode_img.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"
+               constants.CV_ENODERPC, pnode, "instance %s, connection to"

                " primary node failed", instance)
 
                " primary node failed", instance)
 

-      _ErrorIf(inst_config.admin_up and pnode_img.offline,
-               self.EINSTANCEBADNODE, instance,
+      _ErrorIf(inst_config.admin_state == constants.ADMINST_UP and
+               pnode_img.offline,
+               constants.CV_EINSTANCEBADNODE, instance,

                "instance is marked as running and lives on offline node %s",
                inst_config.primary_node)
 
                "instance is marked as running and lives on offline node %s",
                inst_config.primary_node)
 

@@ -2852,7 +2947,8 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       if not inst_config.secondary_nodes:
         i_non_redundant.append(instance)
 
       if not inst_config.secondary_nodes:
         i_non_redundant.append(instance)
 

-      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
+      _ErrorIf(len(inst_config.secondary_nodes) > 1,
+               constants.CV_EINSTANCELAYOUT,

                instance, "instance has multiple secondary nodes: %s",
                utils.CommaJoin(inst_config.secondary_nodes),
                code=self.ETYPE_WARNING)
                instance, "instance has multiple secondary nodes: %s",
                utils.CommaJoin(inst_config.secondary_nodes),
                code=self.ETYPE_WARNING)

@@ -2873,7 +2969,8 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
                                      key=lambda (_, nodes): pnode in nodes,
                                      reverse=True)]
 
                                      key=lambda (_, nodes): pnode in nodes,
                                      reverse=True)]
 

-        self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
+        self._ErrorIf(len(instance_groups) > 1,
+                      constants.CV_EINSTANCESPLITGROUPS,

                       instance, "instance has primary and secondary nodes in"
                       " different groups: %s", utils.CommaJoin(pretty_list),
                       code=self.ETYPE_WARNING)
                       instance, "instance has primary and secondary nodes in"
                       " different groups: %s", utils.CommaJoin(pretty_list),
                       code=self.ETYPE_WARNING)

@@ -2883,21 +2980,22 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
       for snode in inst_config.secondary_nodes:
         s_img = node_image[snode]
 
       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)
+        _ErrorIf(s_img.rpc_fail and not s_img.offline, constants.CV_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
 
         if s_img.offline:
           inst_nodes_offline.append(snode)
 
       # warn that the instance lives on offline nodes

-      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
+      _ErrorIf(inst_nodes_offline, constants.CV_EINSTANCEBADNODE, instance,

                "instance has offline secondary node(s) %s",
                utils.CommaJoin(inst_nodes_offline))
       # ... or ghost/non-vm_capable nodes
       for node in inst_config.all_nodes:
                "instance has offline secondary node(s) %s",
                utils.CommaJoin(inst_nodes_offline))
       # ... or ghost/non-vm_capable nodes
       for node in inst_config.all_nodes:

-        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
-                 "instance lives on ghost node %s", node)
-        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
+        _ErrorIf(node_image[node].ghost, constants.CV_EINSTANCEBADNODE,
+                 instance, "instance lives on ghost node %s", node)
+        _ErrorIf(not node_image[node].vm_capable, constants.CV_EINSTANCEBADNODE,

                  instance, "instance lives on non-vm_capable node %s", node)
 
     feedback_fn("* Verifying orphan volumes")
                  instance, "instance lives on non-vm_capable node %s", node)
 
     feedback_fn("* Verifying orphan volumes")

@@ -2928,6 +3026,9 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
                   % len(i_non_a_balanced))
 
       feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
                   % len(i_non_a_balanced))
 

+    if i_offline:
+      feedback_fn("  - NOTICE: %d offline instance(s) found." % i_offline)
+

     if n_offline:
       feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
 
     if n_offline:
       feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
 

@@ -2965,22 +3066,20 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         res = hooks_results[node_name]
         msg = res.fail_msg
         test = msg and not res.offline
         res = hooks_results[node_name]
         msg = res.fail_msg
         test = msg and not res.offline

-        self._ErrorIf(test, self.ENODEHOOKS, node_name,
+        self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,

                       "Communication failure in hooks execution: %s", msg)
         if res.offline or msg:
                       "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
+          # No need to investigate payload if node is offline or gave
+          # an error.

           continue
         for script, hkr, output in res.payload:
           test = hkr == constants.HKR_FAIL
           continue
         for script, hkr, output in res.payload:
           test = hkr == constants.HKR_FAIL

-          self._ErrorIf(test, self.ENODEHOOKS, node_name,
+          self._ErrorIf(test, constants.CV_ENODEHOOKS, node_name,

                         "Script %s failed, output:", script)
           if test:
             output = self._HOOKS_INDENT_RE.sub("      ", output)
             feedback_fn("%s" % output)
                         "Script %s failed, output:", script)
           if test:
             output = self._HOOKS_INDENT_RE.sub("      ", output)
             feedback_fn("%s" % output)

-            lu_result = 0
+            lu_result = False

 
     return lu_result
 
 
     return lu_result
 

@@ -3092,8 +3191,8 @@ class LUGroupVerifyDisks(NoHooksLU):
     res_missing = {}
 
     nv_dict = _MapInstanceDisksToNodes([inst
     res_missing = {}
 
     nv_dict = _MapInstanceDisksToNodes([inst

-                                        for inst in self.instances.values()
-                                        if inst.admin_up])
+            for inst in self.instances.values()
+            if inst.admin_state == constants.ADMINST_UP])

 
     if nv_dict:
       nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &
 
     if nv_dict:
       nodes = utils.NiceSort(set(self.owned_locks(locking.LEVEL_NODE)) &

@@ -3119,7 +3218,7 @@ class LUGroupVerifyDisks(NoHooksLU):
       # any leftover items in nv_dict are missing LVs, let's arrange the data
       # better
       for key, inst in nv_dict.iteritems():
       # any leftover items in nv_dict are missing LVs, let's arrange the data
       # better
       for key, inst in nv_dict.iteritems():

-        res_missing.setdefault(inst, []).append(key)
+        res_missing.setdefault(inst, []).append(list(key))

 
     return (res_nodes, list(res_instances), res_missing)
 
 
     return (res_nodes, list(res_instances), res_missing)
 

@@ -3134,21 +3233,21 @@ class LUClusterRepairDiskSizes(NoHooksLU):
     if self.op.instances:
       self.wanted_names = _GetWantedInstances(self, self.op.instances)
       self.needed_locks = {
     if self.op.instances:
       self.wanted_names = _GetWantedInstances(self, self.op.instances)
       self.needed_locks = {

-        locking.LEVEL_NODE: [],
+        locking.LEVEL_NODE_RES: [],

         locking.LEVEL_INSTANCE: self.wanted_names,
         }
         locking.LEVEL_INSTANCE: self.wanted_names,
         }

-      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
+      self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE

     else:
       self.wanted_names = None
       self.needed_locks = {
     else:
       self.wanted_names = None
       self.needed_locks = {

-        locking.LEVEL_NODE: locking.ALL_SET,
+        locking.LEVEL_NODE_RES: locking.ALL_SET,

         locking.LEVEL_INSTANCE: locking.ALL_SET,
         }
     self.share_locks = _ShareAll()
 
   def DeclareLocks(self, level):
         locking.LEVEL_INSTANCE: locking.ALL_SET,
         }
     self.share_locks = _ShareAll()
 
   def DeclareLocks(self, level):

-    if level == locking.LEVEL_NODE and self.wanted_names is not None:
-      self._LockInstancesNodes(primary_only=True)
+    if level == locking.LEVEL_NODE_RES and self.wanted_names is not None:
+      self._LockInstancesNodes(primary_only=True, level=level)

 
   def CheckPrereq(self):
     """Check prerequisites.
 
   def CheckPrereq(self):
     """Check prerequisites.

@@ -3199,6 +3298,11 @@ class LUClusterRepairDiskSizes(NoHooksLU):
       for idx, disk in enumerate(instance.disks):
         per_node_disks[pnode].append((instance, idx, disk))
 
       for idx, disk in enumerate(instance.disks):
         per_node_disks[pnode].append((instance, idx, disk))
 

+    assert not (frozenset(per_node_disks.keys()) -
+                self.owned_locks(locking.LEVEL_NODE_RES)), \
+      "Not owning correct locks"
+    assert not self.owned_locks(locking.LEVEL_NODE)
+

     changed = []
     for node, dskl in per_node_disks.items():
       newl = [v[2].Copy() for v in dskl]
     changed = []
     for node, dskl in per_node_disks.items():
       newl = [v[2].Copy() for v in dskl]

@@ -3288,29 +3392,33 @@ class LUClusterRename(LogicalUnit):
 
     """
     clustername = self.op.name
 
     """
     clustername = self.op.name

-    ip = self.ip
+    new_ip = self.ip

 
     # shutdown the master IP
 
     # shutdown the master IP

-    master = self.cfg.GetMasterNode()
-    result = self.rpc.call_node_stop_master(master, False)
+    master_params = self.cfg.GetMasterNetworkParameters()
+    ems = self.cfg.GetUseExternalMipScript()
+    result = self.rpc.call_node_deactivate_master_ip(master_params.name,
+                                                     master_params, ems)

     result.Raise("Could not disable the master role")
 
     try:
       cluster = self.cfg.GetClusterInfo()
       cluster.cluster_name = clustername
     result.Raise("Could not disable the master role")
 
     try:
       cluster = self.cfg.GetClusterInfo()
       cluster.cluster_name = clustername

-      cluster.master_ip = ip
+      cluster.master_ip = new_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.GetOnlineNodeList()
       try:
       self.cfg.Update(cluster, feedback_fn)
 
       # update the known hosts file
       ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
       node_list = self.cfg.GetOnlineNodeList()
       try:

-        node_list.remove(master)
+        node_list.remove(master_params.name)

       except ValueError:
         pass
       _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
     finally:
       except ValueError:
         pass
       _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
     finally:

-      result = self.rpc.call_node_start_master(master, False, False)
+      master_params.ip = new_ip
+      result = self.rpc.call_node_activate_master_ip(master_params.name,
+                                                     master_params, ems)

       msg = result.fail_msg
       if msg:
         self.LogWarning("Could not re-enable the master role on"
       msg = result.fail_msg
       if msg:
         self.LogWarning("Could not re-enable the master role on"

@@ -3319,6 +3427,27 @@ class LUClusterRename(LogicalUnit):
     return clustername
 
 
     return clustername
 
 

+def _ValidateNetmask(cfg, netmask):
+  """Checks if a netmask is valid.
+
+  @type cfg: L{config.ConfigWriter}
+  @param cfg: The cluster configuration
+  @type netmask: int
+  @param netmask: the netmask to be verified
+  @raise errors.OpPrereqError: if the validation fails
+
+  """
+  ip_family = cfg.GetPrimaryIPFamily()
+  try:
+    ipcls = netutils.IPAddress.GetClassFromIpFamily(ip_family)
+  except errors.ProgrammerError:
+    raise errors.OpPrereqError("Invalid primary ip family: %s." %
+                               ip_family)
+  if not ipcls.ValidateNetmask(netmask):
+    raise errors.OpPrereqError("CIDR netmask (%s) not valid" %
+                                (netmask))
+
+

 class LUClusterSetParams(LogicalUnit):
   """Change the parameters of the cluster.
 
 class LUClusterSetParams(LogicalUnit):
   """Change the parameters of the cluster.
 

@@ -3340,6 +3469,9 @@ class LUClusterSetParams(LogicalUnit):
     if self.op.remove_uids:
       uidpool.CheckUidPool(self.op.remove_uids)
 
     if self.op.remove_uids:
       uidpool.CheckUidPool(self.op.remove_uids)
 

+    if self.op.master_netmask is not None:
+      _ValidateNetmask(self.cfg, self.op.master_netmask)
+

   def ExpandNames(self):
     # FIXME: in the future maybe other cluster params won't require checking on
     # all nodes to be modified.
   def ExpandNames(self):
     # FIXME: in the future maybe other cluster params won't require checking on
     # all nodes to be modified.

@@ -3616,6 +3748,9 @@ class LUClusterSetParams(LogicalUnit):
     if self.op.reserved_lvs is not None:
       self.cluster.reserved_lvs = self.op.reserved_lvs
 
     if self.op.reserved_lvs is not None:
       self.cluster.reserved_lvs = self.op.reserved_lvs
 

+    if self.op.use_external_mip_script is not None:
+      self.cluster.use_external_mip_script = self.op.use_external_mip_script
+

     def helper_os(aname, mods, desc):
       desc += " OS list"
       lst = getattr(self.cluster, aname)
     def helper_os(aname, mods, desc):
       desc += " OS list"
       lst = getattr(self.cluster, aname)

@@ -3640,21 +3775,40 @@ class LUClusterSetParams(LogicalUnit):
       helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
 
     if self.op.master_netdev:
       helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
 
     if self.op.master_netdev:

-      master = self.cfg.GetMasterNode()
+      master_params = self.cfg.GetMasterNetworkParameters()
+      ems = self.cfg.GetUseExternalMipScript()

       feedback_fn("Shutting down master ip on the current netdev (%s)" %
                   self.cluster.master_netdev)
       feedback_fn("Shutting down master ip on the current netdev (%s)" %
                   self.cluster.master_netdev)

-      result = self.rpc.call_node_stop_master(master, False)
+      result = self.rpc.call_node_deactivate_master_ip(master_params.name,
+                                                       master_params, ems)

       result.Raise("Could not disable the master ip")
       feedback_fn("Changing master_netdev from %s to %s" %
       result.Raise("Could not disable the master ip")
       feedback_fn("Changing master_netdev from %s to %s" %

-                  (self.cluster.master_netdev, self.op.master_netdev))
+                  (master_params.netdev, self.op.master_netdev))

       self.cluster.master_netdev = self.op.master_netdev
 
       self.cluster.master_netdev = self.op.master_netdev
 

+    if self.op.master_netmask:
+      master_params = self.cfg.GetMasterNetworkParameters()
+      feedback_fn("Changing master IP netmask to %s" % self.op.master_netmask)
+      result = self.rpc.call_node_change_master_netmask(master_params.name,
+                                                        master_params.netmask,
+                                                        self.op.master_netmask,
+                                                        master_params.ip,
+                                                        master_params.netdev)
+      if result.fail_msg:
+        msg = "Could not change the master IP netmask: %s" % result.fail_msg
+        feedback_fn(msg)
+
+      self.cluster.master_netmask = self.op.master_netmask
+

     self.cfg.Update(self.cluster, feedback_fn)
 
     if self.op.master_netdev:
     self.cfg.Update(self.cluster, feedback_fn)
 
     if self.op.master_netdev:

+      master_params = self.cfg.GetMasterNetworkParameters()

       feedback_fn("Starting the master ip on the new master netdev (%s)" %
                   self.op.master_netdev)
       feedback_fn("Starting the master ip on the new master netdev (%s)" %
                   self.op.master_netdev)

-      result = self.rpc.call_node_start_master(master, False, False)
+      ems = self.cfg.GetUseExternalMipScript()
+      result = self.rpc.call_node_activate_master_ip(master_params.name,
+                                                     master_params, ems)

       if result.fail_msg:
         self.LogWarning("Could not re-enable the master ip on"
                         " the master, please restart manually: %s",
       if result.fail_msg:
         self.LogWarning("Could not re-enable the master ip on"
                         " the master, please restart manually: %s",

@@ -3687,36 +3841,58 @@ def _ComputeAncillaryFiles(cluster, redist):
     constants.SSH_KNOWN_HOSTS_FILE,
     constants.CONFD_HMAC_KEY,
     constants.CLUSTER_DOMAIN_SECRET_FILE,
     constants.SSH_KNOWN_HOSTS_FILE,
     constants.CONFD_HMAC_KEY,
     constants.CLUSTER_DOMAIN_SECRET_FILE,

+    constants.SPICE_CERT_FILE,
+    constants.SPICE_CACERT_FILE,
+    constants.RAPI_USERS_FILE,

     ])
 
   if not redist:
     files_all.update(constants.ALL_CERT_FILES)
     files_all.update(ssconf.SimpleStore().GetFileList())
     ])
 
   if not redist:
     files_all.update(constants.ALL_CERT_FILES)
     files_all.update(ssconf.SimpleStore().GetFileList())

+  else:
+    # we need to ship at least the RAPI certificate
+    files_all.add(constants.RAPI_CERT_FILE)

 
   if cluster.modify_etc_hosts:
     files_all.add(constants.ETC_HOSTS)
 
 
   if cluster.modify_etc_hosts:
     files_all.add(constants.ETC_HOSTS)
 

-  # Files which must either exist on all nodes or on none
-  files_all_opt = set([
+  # Files which are optional, these must:
+  # - be present in one other category as well
+  # - either exist or not exist on all nodes of that category (mc, vm all)
+  files_opt = set([

     constants.RAPI_USERS_FILE,
     ])
 
   # Files which should only be on master candidates
   files_mc = set()
     constants.RAPI_USERS_FILE,
     ])
 
   # Files which should only be on master candidates
   files_mc = set()

+

   if not redist:
     files_mc.add(constants.CLUSTER_CONF_FILE)
 
   if not redist:
     files_mc.add(constants.CLUSTER_CONF_FILE)
 

+    # FIXME: this should also be replicated but Ganeti doesn't support files_mc
+    # replication
+    files_mc.add(constants.DEFAULT_MASTER_SETUP_SCRIPT)
+

   # Files which should only be on VM-capable nodes
   files_vm = set(filename
     for hv_name in cluster.enabled_hypervisors
   # Files which should only be on VM-capable nodes
   files_vm = set(filename
     for hv_name in cluster.enabled_hypervisors

-    for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles())
+    for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[0])
+
+  files_opt |= set(filename
+    for hv_name in cluster.enabled_hypervisors
+    for filename in hypervisor.GetHypervisor(hv_name).GetAncillaryFiles()[1])

 
 

-  # Filenames must be unique
-  assert (len(files_all | files_all_opt | files_mc | files_vm) ==
-          sum(map(len, [files_all, files_all_opt, files_mc, files_vm]))), \
+  # Filenames in each category must be unique
+  all_files_set = files_all | files_mc | files_vm
+  assert (len(all_files_set) ==
+          sum(map(len, [files_all, files_mc, files_vm]))), \

          "Found file listed in more than one file list"
 
          "Found file listed in more than one file list"
 

-  return (files_all, files_all_opt, files_mc, files_vm)
+  # Optional files must be present in one other category
+  assert all_files_set.issuperset(files_opt), \
+         "Optional file not in a different required list"
+
+  return (files_all, files_opt, files_mc, files_vm)

 
 
 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
 
 
 def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):

@@ -3750,7 +3926,7 @@ def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
       nodelist.remove(master_info.name)
 
   # Gather file lists
       nodelist.remove(master_info.name)
 
   # Gather file lists

-  (files_all, files_all_opt, files_mc, files_vm) = \
+  (files_all, _, files_mc, files_vm) = \

     _ComputeAncillaryFiles(cluster, True)
 
   # Never re-distribute configuration file from here
     _ComputeAncillaryFiles(cluster, True)
 
   # Never re-distribute configuration file from here

@@ -3760,7 +3936,6 @@ def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
 
   filemap = [
     (online_nodes, files_all),
 
   filemap = [
     (online_nodes, files_all),

-    (online_nodes, files_all_opt),

     (vm_nodes, files_vm),
     ]
 
     (vm_nodes, files_vm),
     ]
 

@@ -3792,6 +3967,34 @@ class LUClusterRedistConf(NoHooksLU):
     _RedistributeAncillaryFiles(self)
 
 
     _RedistributeAncillaryFiles(self)
 
 

+class LUClusterActivateMasterIp(NoHooksLU):
+  """Activate the master IP on the master node.
+
+  """
+  def Exec(self, feedback_fn):
+    """Activate the master IP.
+
+    """
+    master_params = self.cfg.GetMasterNetworkParameters()
+    ems = self.cfg.GetUseExternalMipScript()
+    self.rpc.call_node_activate_master_ip(master_params.name,
+                                          master_params, ems)
+
+
+class LUClusterDeactivateMasterIp(NoHooksLU):
+  """Deactivate the master IP on the master node.
+
+  """
+  def Exec(self, feedback_fn):
+    """Deactivate the master IP.
+
+    """
+    master_params = self.cfg.GetMasterNetworkParameters()
+    ems = self.cfg.GetUseExternalMipScript()
+    self.rpc.call_node_deactivate_master_ip(master_params.name, master_params,
+                                            ems)
+
+

 def _WaitForSync(lu, instance, disks=None, oneshot=False):
   """Sleep and poll for an instance's disk to sync.
 
 def _WaitForSync(lu, instance, disks=None, oneshot=False):
   """Sleep and poll for an instance's disk to sync.
 

@@ -4080,6 +4283,7 @@ class LUOobCommand(NoHooksLU):
       raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
                                utils.CommaJoin(errs))
 
       raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
                                utils.CommaJoin(errs))
 

+

 class _OsQuery(_QueryBase):
   FIELDS = query.OS_FIELDS
 
 class _OsQuery(_QueryBase):
   FIELDS = query.OS_FIELDS
 

@@ -4310,6 +4514,9 @@ class LUNodeRemove(LogicalUnit):
 
     modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
 
 
     modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
 

+    assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
+      "Not owning BGL"
+

     # Promote nodes to master candidate as needed
     _AdjustCandidatePool(self, exceptions=[node.name])
     self.context.RemoveNode(node.name)
     # Promote nodes to master candidate as needed
     _AdjustCandidatePool(self, exceptions=[node.name])
     self.context.RemoveNode(node.name)

@@ -4338,7 +4545,7 @@ class _NodeQuery(_QueryBase):
 
   def ExpandNames(self, lu):
     lu.needed_locks = {}
 
   def ExpandNames(self, lu):
     lu.needed_locks = {}

-    lu.share_locks[locking.LEVEL_NODE] = 1
+    lu.share_locks = _ShareAll()

 
     if self.names:
       self.wanted = _GetWantedNodes(lu, self.names)
 
     if self.names:
       self.wanted = _GetWantedNodes(lu, self.names)

@@ -4349,7 +4556,7 @@ class _NodeQuery(_QueryBase):
                        query.NQ_LIVE in self.requested_data)
 
     if self.do_locking:
                        query.NQ_LIVE in self.requested_data)
 
     if self.do_locking:

-      # if we don't request only static fields, we need to lock the nodes
+      # If any non-static field is requested we need to lock the nodes

       lu.needed_locks[locking.LEVEL_NODE] = self.wanted
 
   def DeclareLocks(self, lu, level):
       lu.needed_locks[locking.LEVEL_NODE] = self.wanted
 
   def DeclareLocks(self, lu, level):

@@ -4413,7 +4620,7 @@ class LUNodeQuery(NoHooksLU):
   """Logical unit for querying nodes.
 
   """
   """Logical unit for querying nodes.
 
   """

-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142

   REQ_BGL = False
 
   def CheckArguments(self):
   REQ_BGL = False
 
   def CheckArguments(self):

@@ -4423,6 +4630,9 @@ class LUNodeQuery(NoHooksLU):
   def ExpandNames(self):
     self.nq.ExpandNames(self)
 
   def ExpandNames(self):
     self.nq.ExpandNames(self)
 

+  def DeclareLocks(self, level):
+    self.nq.DeclareLocks(self, level)
+

   def Exec(self, feedback_fn):
     return self.nq.OldStyleQuery(self)
 
   def Exec(self, feedback_fn):
     return self.nq.OldStyleQuery(self)
 

@@ -4441,8 +4651,9 @@ class LUNodeQueryvols(NoHooksLU):
                        selected=self.op.output_fields)
 
   def ExpandNames(self):
                        selected=self.op.output_fields)
 
   def ExpandNames(self):

+    self.share_locks = _ShareAll()

     self.needed_locks = {}
     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:
     if not self.op.nodes:
       self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
     else:

@@ -4509,8 +4720,8 @@ class LUNodeQueryStorage(NoHooksLU):
                        selected=self.op.output_fields)
 
   def ExpandNames(self):
                        selected=self.op.output_fields)
 
   def ExpandNames(self):

+    self.share_locks = _ShareAll()

     self.needed_locks = {}
     self.needed_locks = {}

-    self.share_locks[locking.LEVEL_NODE] = 1

 
     if self.op.nodes:
       self.needed_locks[locking.LEVEL_NODE] = \
 
     if self.op.nodes:
       self.needed_locks[locking.LEVEL_NODE] = \

@@ -4614,7 +4825,7 @@ class _InstanceQuery(_QueryBase):
               for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
               for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
       elif level == locking.LEVEL_NODE:
               for instance_name in lu.owned_locks(locking.LEVEL_INSTANCE)
               for group_uuid in lu.cfg.GetInstanceNodeGroups(instance_name))
       elif level == locking.LEVEL_NODE:

-        lu._LockInstancesNodes() # pylint: disable-msg=W0212
+        lu._LockInstancesNodes() # pylint: disable=W0212

 
   @staticmethod
   def _CheckGroupLocks(lu):
 
   @staticmethod
   def _CheckGroupLocks(lu):

@@ -4715,13 +4926,13 @@ class LUQuery(NoHooksLU):
   """Query for resources/items of a certain kind.
 
   """
   """Query for resources/items of a certain kind.
 
   """

-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142

   REQ_BGL = False
 
   def CheckArguments(self):
     qcls = _GetQueryImplementation(self.op.what)
 
   REQ_BGL = False
 
   def CheckArguments(self):
     qcls = _GetQueryImplementation(self.op.what)
 

-    self.impl = qcls(self.op.filter, self.op.fields, False)
+    self.impl = qcls(self.op.qfilter, self.op.fields, self.op.use_locking)

 
   def ExpandNames(self):
     self.impl.ExpandNames(self)
 
   def ExpandNames(self):
     self.impl.ExpandNames(self)

@@ -4737,7 +4948,7 @@ class LUQueryFields(NoHooksLU):
   """Query for resources/items of a certain kind.
 
   """
   """Query for resources/items of a certain kind.
 
   """

-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142

   REQ_BGL = False
 
   def CheckArguments(self):
   REQ_BGL = False
 
   def CheckArguments(self):

@@ -4976,6 +5187,9 @@ class LUNodeAdd(LogicalUnit):
     new_node = self.new_node
     node = new_node.name
 
     new_node = self.new_node
     node = new_node.name
 

+    assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \
+      "Not owning BGL"
+

     # We adding a new node so we assume it's powered
     new_node.powered = True
 
     # We adding a new node so we assume it's powered
     new_node.powered = True
 

@@ -4984,7 +5198,7 @@ class LUNodeAdd(LogicalUnit):
     # 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:
     # 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
+      new_node.drained = new_node.offline = False # pylint: disable=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
       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

@@ -5030,7 +5244,7 @@ class LUNodeAdd(LogicalUnit):
 
     node_verify_list = [self.cfg.GetMasterNode()]
     node_verify_param = {
 
     node_verify_list = [self.cfg.GetMasterNode()]
     node_verify_param = {

-      constants.NV_NODELIST: [node],
+      constants.NV_NODELIST: ([node], {}),

       # TODO: do a node-net-test as well?
     }
 
       # TODO: do a node-net-test as well?
     }
 

@@ -5114,35 +5328,32 @@ class LUNodeSetParams(LogicalUnit):
     self.lock_all = self.op.auto_promote and self.might_demote
     self.lock_instances = self.op.secondary_ip is not None
 
     self.lock_all = self.op.auto_promote and self.might_demote
     self.lock_instances = self.op.secondary_ip is not None
 

+  def _InstanceFilter(self, instance):
+    """Filter for getting affected instances.
+
+    """
+    return (instance.disk_template in constants.DTS_INT_MIRROR and
+            self.op.node_name in instance.all_nodes)
+

   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 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}
 

-    if self.lock_instances:
-      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
+    # Since modifying a node can have severe effects on currently running
+    # operations the resource lock is at least acquired in shared mode
+    self.needed_locks[locking.LEVEL_NODE_RES] = \
+      self.needed_locks[locking.LEVEL_NODE]

 
 

-  def DeclareLocks(self, level):
-    # If we have locked all instances, before waiting to lock nodes, release
-    # all the ones living on nodes unrelated to the current operation.
-    if level == locking.LEVEL_NODE and self.lock_instances:
-      self.affected_instances = []
-      if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
-        instances_keep = []
-
-        # Build list of instances to release
-        locked_i = self.owned_locks(locking.LEVEL_INSTANCE)
-        for instance_name, instance in self.cfg.GetMultiInstanceInfo(locked_i):
-          if (instance.disk_template in constants.DTS_INT_MIRROR and
-              self.op.node_name in instance.all_nodes):
-            instances_keep.append(instance_name)
-            self.affected_instances.append(instance)
-
-        _ReleaseLocks(self, locking.LEVEL_INSTANCE, keep=instances_keep)
-
-        assert (set(self.owned_locks(locking.LEVEL_INSTANCE)) ==
-                set(instances_keep))
+    # Get node resource and instance locks in shared mode; they are not used
+    # for anything but read-only access
+    self.share_locks[locking.LEVEL_NODE_RES] = 1
+    self.share_locks[locking.LEVEL_INSTANCE] = 1
+
+    if self.lock_instances:
+      self.needed_locks[locking.LEVEL_INSTANCE] = \
+        frozenset(self.cfg.GetInstancesInfoByFilter(self._InstanceFilter))

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -5174,6 +5385,25 @@ class LUNodeSetParams(LogicalUnit):
     """
     node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
 
     """
     node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
 

+    if self.lock_instances:
+      affected_instances = \
+        self.cfg.GetInstancesInfoByFilter(self._InstanceFilter)
+
+      # Verify instance locks
+      owned_instances = self.owned_locks(locking.LEVEL_INSTANCE)
+      wanted_instances = frozenset(affected_instances.keys())
+      if wanted_instances - owned_instances:
+        raise errors.OpPrereqError("Instances affected by changing node %s's"
+                                   " secondary IP address have changed since"
+                                   " locks were acquired, wanted '%s', have"
+                                   " '%s'; retry the operation" %
+                                   (self.op.node_name,
+                                    utils.CommaJoin(wanted_instances),
+                                    utils.CommaJoin(owned_instances)),
+                                   errors.ECODE_STATE)
+    else:
+      affected_instances = None
+

     if (self.op.master_candidate is not None or
         self.op.drained is not None or
         self.op.offline is not None):
     if (self.op.master_candidate is not None or
         self.op.drained is not None or
         self.op.offline is not None):

@@ -5263,7 +5493,9 @@ class LUNodeSetParams(LogicalUnit):
 
     if old_role == self._ROLE_OFFLINE and new_role != old_role:
       # Trying to transition out of offline status
 
     if old_role == self._ROLE_OFFLINE and new_role != old_role:
       # Trying to transition out of offline status

-      result = self.rpc.call_version([node.name])[node.name]
+      # TODO: Use standard RPC runner, but make sure it works when the node is
+      # still marked offline
+      result = rpc.BootstrapRunner().call_version([node.name])[node.name]

       if result.fail_msg:
         raise errors.OpPrereqError("Node %s is being de-offlined but fails"
                                    " to report its version: %s" %
       if result.fail_msg:
         raise errors.OpPrereqError("Node %s is being de-offlined but fails"
                                    " to report its version: %s" %

@@ -5282,16 +5514,21 @@ class LUNodeSetParams(LogicalUnit):
         raise errors.OpPrereqError("Cannot change the secondary ip on a single"
                                    " homed cluster", errors.ECODE_INVAL)
 
         raise errors.OpPrereqError("Cannot change the secondary ip on a single"
                                    " homed cluster", errors.ECODE_INVAL)
 

+      assert not (frozenset(affected_instances) -
+                  self.owned_locks(locking.LEVEL_INSTANCE))
+

       if node.offline:
       if node.offline:

-        if self.affected_instances:
-          raise errors.OpPrereqError("Cannot change secondary ip: offline"
-                                     " node has instances (%s) configured"
-                                     " to use it" % self.affected_instances)
+        if affected_instances:
+          raise errors.OpPrereqError("Cannot change secondary IP address:"
+                                     " offline node has instances (%s)"
+                                     " configured to use it" %
+                                     utils.CommaJoin(affected_instances.keys()))

       else:
         # On online nodes, check that no instances are running, and that
         # the node has the new ip and we can reach it.
       else:
         # On online nodes, check that no instances are running, and that
         # the node has the new ip and we can reach it.

-        for instance in self.affected_instances:
-          _CheckInstanceDown(self, instance, "cannot change secondary ip")
+        for instance in affected_instances.values():
+          _CheckInstanceState(self, instance, INSTANCE_DOWN,
+                              msg="cannot change secondary ip")

 
         _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
         if master.name != node.name:
 
         _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
         if master.name != node.name:

@@ -5442,6 +5679,8 @@ class LUClusterQuery(NoHooksLU):
       "ndparams": cluster.ndparams,
       "candidate_pool_size": cluster.candidate_pool_size,
       "master_netdev": cluster.master_netdev,
       "ndparams": cluster.ndparams,
       "candidate_pool_size": cluster.candidate_pool_size,
       "master_netdev": cluster.master_netdev,

+      "master_netmask": cluster.master_netmask,
+      "use_external_mip_script": cluster.use_external_mip_script,

       "volume_group_name": cluster.volume_group_name,
       "drbd_usermode_helper": cluster.drbd_usermode_helper,
       "file_storage_dir": cluster.file_storage_dir,
       "volume_group_name": cluster.volume_group_name,
       "drbd_usermode_helper": cluster.drbd_usermode_helper,
       "file_storage_dir": cluster.file_storage_dir,

@@ -5687,7 +5926,7 @@ def _SafeShutdownInstanceDisks(lu, instance, disks=None):
   _ShutdownInstanceDisks.
 
   """
   _ShutdownInstanceDisks.
 
   """

-  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
+  _CheckInstanceState(lu, instance, INSTANCE_DOWN, msg="cannot shutdown disks")

   _ShutdownInstanceDisks(lu, instance, disks=disks)
 
 
   _ShutdownInstanceDisks(lu, instance, disks=disks)
 
 

@@ -5832,6 +6071,40 @@ def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
                                  errors.ECODE_NORES)
 
 
                                  errors.ECODE_NORES)
 
 

+def _CheckNodesPhysicalCPUs(lu, nodenames, requested, hypervisor_name):
+  """Checks if nodes have enough physical CPUs
+
+  This function checks if all given nodes have the needed number of
+  physical CPUs. In case any node has less CPUs or we cannot get the
+  information from the node, this function raises an OpPrereqError
+  exception.
+
+  @type lu: C{LogicalUnit}
+  @param lu: a logical unit from which we get configuration data
+  @type nodenames: C{list}
+  @param nodenames: the list of node names to check
+  @type requested: C{int}
+  @param requested: the minimum acceptable number of physical CPUs
+  @raise errors.OpPrereqError: if the node doesn't have enough CPUs,
+      or we cannot check the node
+
+  """
+  nodeinfo = lu.rpc.call_node_info(nodenames, None, hypervisor_name)
+  for node in nodenames:
+    info = nodeinfo[node]
+    info.Raise("Cannot get current information from node %s" % node,
+               prereq=True, ecode=errors.ECODE_ENVIRON)
+    num_cpus = info.payload.get("cpu_total", None)
+    if not isinstance(num_cpus, int):
+      raise errors.OpPrereqError("Can't compute the number of physical CPUs"
+                                 " on node %s, result was '%s'" %
+                                 (node, num_cpus), errors.ECODE_ENVIRON)
+    if requested > num_cpus:
+      raise errors.OpPrereqError("Node %s has %s physical CPUs, but %s are "
+                                 "required" % (node, num_cpus, requested),
+                                 errors.ECODE_NORES)
+
+

 class LUInstanceStartup(LogicalUnit):
   """Starts an instance.
 
 class LUInstanceStartup(LogicalUnit):
   """Starts an instance.
 

@@ -5891,6 +6164,8 @@ class LUInstanceStartup(LogicalUnit):
       hv_type.CheckParameterSyntax(filled_hvp)
       _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
 
       hv_type.CheckParameterSyntax(filled_hvp)
       _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
 

+    _CheckInstanceState(self, instance, INSTANCE_ONLINE)
+

     self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
 
     if self.primary_offline and self.op.ignore_offline_nodes:
     self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
 
     if self.primary_offline and self.op.ignore_offline_nodes:

@@ -5934,9 +6209,11 @@ class LUInstanceStartup(LogicalUnit):
 
       _StartInstanceDisks(self, instance, force)
 
 
       _StartInstanceDisks(self, instance, force)
 

-      result = self.rpc.call_instance_start(node_current, instance,
-                                            self.op.hvparams, self.op.beparams,
-                                            self.op.startup_paused)
+      result = \
+        self.rpc.call_instance_start(node_current,
+                                     (instance, self.op.hvparams,
+                                      self.op.beparams),
+                                     self.op.startup_paused)

       msg = result.fail_msg
       if msg:
         _ShutdownInstanceDisks(self, instance)
       msg = result.fail_msg
       if msg:
         _ShutdownInstanceDisks(self, instance)

@@ -5986,7 +6263,7 @@ class LUInstanceReboot(LogicalUnit):
     self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
     assert self.instance is not None, \
       "Cannot retrieve locked instance %s" % self.op.instance_name
     self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
     assert self.instance is not None, \
       "Cannot retrieve locked instance %s" % self.op.instance_name

-
+    _CheckInstanceState(self, instance, INSTANCE_ONLINE)

     _CheckNodeOnline(self, instance.primary_node)
 
     # check bridges existence
     _CheckNodeOnline(self, instance.primary_node)
 
     # check bridges existence

@@ -6026,8 +6303,8 @@ class LUInstanceReboot(LogicalUnit):
         self.LogInfo("Instance %s was already stopped, starting now",
                      instance.name)
       _StartInstanceDisks(self, instance, ignore_secondaries)
         self.LogInfo("Instance %s was already stopped, starting now",
                      instance.name)
       _StartInstanceDisks(self, instance, ignore_secondaries)

-      result = self.rpc.call_instance_start(node_current, instance,
-                                            None, None, False)
+      result = self.rpc.call_instance_start(node_current,
+                                            (instance, None, None), False)

       msg = result.fail_msg
       if msg:
         _ShutdownInstanceDisks(self, instance)
       msg = result.fail_msg
       if msg:
         _ShutdownInstanceDisks(self, instance)

@@ -6075,6 +6352,8 @@ class LUInstanceShutdown(LogicalUnit):
     assert self.instance is not None, \
       "Cannot retrieve locked instance %s" % self.op.instance_name
 
     assert self.instance is not None, \
       "Cannot retrieve locked instance %s" % self.op.instance_name
 

+    _CheckInstanceState(self, self.instance, INSTANCE_ONLINE)
+

     self.primary_offline = \
       self.cfg.GetNodeInfo(self.instance.primary_node).offline
 
     self.primary_offline = \
       self.cfg.GetNodeInfo(self.instance.primary_node).offline
 

@@ -6151,7 +6430,7 @@ class LUInstanceReinstall(LogicalUnit):
       raise errors.OpPrereqError("Instance '%s' has no disks" %
                                  self.op.instance_name,
                                  errors.ECODE_INVAL)
       raise errors.OpPrereqError("Instance '%s' has no disks" %
                                  self.op.instance_name,
                                  errors.ECODE_INVAL)

-    _CheckInstanceDown(self, instance, "cannot reinstall")
+    _CheckInstanceState(self, instance, INSTANCE_DOWN, msg="cannot reinstall")

 
     if self.op.os_type is not None:
       # OS verification
 
     if self.op.os_type is not None:
       # OS verification

@@ -6188,9 +6467,9 @@ class LUInstanceReinstall(LogicalUnit):
     try:
       feedback_fn("Running the instance OS create scripts...")
       # FIXME: pass debug option from opcode to backend
     try:
       feedback_fn("Running the instance OS create scripts...")
       # FIXME: pass debug option from opcode to backend

-      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
-                                             self.op.debug_level,
-                                             osparams=self.os_inst)
+      result = self.rpc.call_instance_os_add(inst.primary_node,
+                                             (inst, self.os_inst), True,
+                                             self.op.debug_level)

       result.Raise("Could not install OS for instance %s on node %s" %
                    (inst.name, inst.primary_node))
     finally:
       result.Raise("Could not install OS for instance %s on node %s" %
                    (inst.name, inst.primary_node))
     finally:

@@ -6224,6 +6503,10 @@ class LUInstanceRecreateDisks(LogicalUnit):
       # otherwise we need to lock all nodes for disk re-creation
       primary_only = bool(self.op.nodes)
       self._LockInstancesNodes(primary_only=primary_only)
       # otherwise we need to lock all nodes for disk re-creation
       primary_only = bool(self.op.nodes)
       self._LockInstancesNodes(primary_only=primary_only)

+    elif level == locking.LEVEL_NODE_RES:
+      # Copy node locks
+      self.needed_locks[locking.LEVEL_NODE_RES] = \
+        self.needed_locks[locking.LEVEL_NODE][:]

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -6270,10 +6553,12 @@ class LUInstanceRecreateDisks(LogicalUnit):
                                  self.op.instance_name, errors.ECODE_INVAL)
     # if we replace nodes *and* the old primary is offline, we don't
     # check
                                  self.op.instance_name, errors.ECODE_INVAL)
     # if we replace nodes *and* the old primary is offline, we don't
     # check

-    assert instance.primary_node in self.needed_locks[locking.LEVEL_NODE]
+    assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE)
+    assert instance.primary_node in self.owned_locks(locking.LEVEL_NODE_RES)

     old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
     if not (self.op.nodes and old_pnode.offline):
     old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
     if not (self.op.nodes and old_pnode.offline):

-      _CheckInstanceDown(self, instance, "cannot recreate disks")
+      _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
+                          msg="cannot recreate disks")

 
     if not self.op.disks:
       self.op.disks = range(len(instance.disks))
 
     if not self.op.disks:
       self.op.disks = range(len(instance.disks))

@@ -6294,6 +6579,9 @@ class LUInstanceRecreateDisks(LogicalUnit):
     """
     instance = self.instance
 
     """
     instance = self.instance
 

+    assert (self.owned_locks(locking.LEVEL_NODE) ==
+            self.owned_locks(locking.LEVEL_NODE_RES))
+

     to_skip = []
     mods = [] # keeps track of needed logical_id changes
 
     to_skip = []
     mods = [] # keeps track of needed logical_id changes
 

@@ -6376,7 +6664,8 @@ class LUInstanceRename(LogicalUnit):
     instance = self.cfg.GetInstanceInfo(self.op.instance_name)
     assert instance is not None
     _CheckNodeOnline(self, instance.primary_node)
     instance = self.cfg.GetInstanceInfo(self.op.instance_name)
     assert instance is not None
     _CheckNodeOnline(self, instance.primary_node)

-    _CheckInstanceDown(self, instance, "cannot rename")
+    _CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
+                        msg="cannot rename")

     self.instance = instance
 
     new_name = self.op.new_name
     self.instance = instance
 
     new_name = self.op.new_name

@@ -6462,11 +6751,16 @@ class LUInstanceRemove(LogicalUnit):
   def ExpandNames(self):
     self._ExpandAndLockInstance()
     self.needed_locks[locking.LEVEL_NODE] = []
   def ExpandNames(self):
     self._ExpandAndLockInstance()
     self.needed_locks[locking.LEVEL_NODE] = []

+    self.needed_locks[locking.LEVEL_NODE_RES] = []

     self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
 
   def DeclareLocks(self, level):
     if level == locking.LEVEL_NODE:
       self._LockInstancesNodes()
     self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
 
   def DeclareLocks(self, level):
     if level == locking.LEVEL_NODE:
       self._LockInstancesNodes()

+    elif level == locking.LEVEL_NODE_RES:
+      # Copy node locks
+      self.needed_locks[locking.LEVEL_NODE_RES] = \
+        self.needed_locks[locking.LEVEL_NODE][:]

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -6515,6 +6809,12 @@ class LUInstanceRemove(LogicalUnit):
                                  " node %s: %s" %
                                  (instance.name, instance.primary_node, msg))
 
                                  " node %s: %s" %
                                  (instance.name, instance.primary_node, msg))
 

+    assert (self.owned_locks(locking.LEVEL_NODE) ==
+            self.owned_locks(locking.LEVEL_NODE_RES))
+    assert not (set(instance.all_nodes) -
+                self.owned_locks(locking.LEVEL_NODE)), \
+      "Not owning correct locks"
+

     _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
 
 
     _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
 
 

@@ -6544,7 +6844,7 @@ class LUInstanceQuery(NoHooksLU):
   """Logical unit for querying instances.
 
   """
   """Logical unit for querying instances.
 
   """

-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142

   REQ_BGL = False
 
   def CheckArguments(self):
   REQ_BGL = False
 
   def CheckArguments(self):

@@ -6728,11 +7028,16 @@ class LUInstanceMove(LogicalUnit):
     target_node = _ExpandNodeName(self.cfg, self.op.target_node)
     self.op.target_node = target_node
     self.needed_locks[locking.LEVEL_NODE] = [target_node]
     target_node = _ExpandNodeName(self.cfg, self.op.target_node)
     self.op.target_node = target_node
     self.needed_locks[locking.LEVEL_NODE] = [target_node]

+    self.needed_locks[locking.LEVEL_NODE_RES] = []

     self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
 
   def DeclareLocks(self, level):
     if level == locking.LEVEL_NODE:
       self._LockInstancesNodes(primary_only=True)
     self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
 
   def DeclareLocks(self, level):
     if level == locking.LEVEL_NODE:
       self._LockInstancesNodes(primary_only=True)

+    elif level == locking.LEVEL_NODE_RES:
+      # Copy node locks
+      self.needed_locks[locking.LEVEL_NODE_RES] = \
+        self.needed_locks[locking.LEVEL_NODE][:]

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -6790,7 +7095,7 @@ class LUInstanceMove(LogicalUnit):
     _CheckNodeNotDrained(self, target_node)
     _CheckNodeVmCapable(self, target_node)
 
     _CheckNodeNotDrained(self, target_node)
     _CheckNodeVmCapable(self, target_node)
 

-    if instance.admin_up:
+    if instance.admin_state == constants.ADMINST_UP:

       # check memory requirements on the secondary node
       _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
                            instance.name, bep[constants.BE_MEMORY],
       # check memory requirements on the secondary node
       _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
                            instance.name, bep[constants.BE_MEMORY],

@@ -6817,6 +7122,9 @@ class LUInstanceMove(LogicalUnit):
     self.LogInfo("Shutting down instance %s on source node %s",
                  instance.name, source_node)
 
     self.LogInfo("Shutting down instance %s on source node %s",
                  instance.name, source_node)
 

+    assert (self.owned_locks(locking.LEVEL_NODE) ==
+            self.owned_locks(locking.LEVEL_NODE_RES))
+

     result = self.rpc.call_instance_shutdown(source_node, instance,
                                              self.op.shutdown_timeout)
     msg = result.fail_msg
     result = self.rpc.call_instance_shutdown(source_node, instance,
                                              self.op.shutdown_timeout)
     msg = result.fail_msg

@@ -6881,7 +7189,7 @@ class LUInstanceMove(LogicalUnit):
     _RemoveDisks(self, instance, target_node=source_node)
 
     # Only start the instance if it's marked as up
     _RemoveDisks(self, instance, target_node=source_node)
 
     # Only start the instance if it's marked as up

-    if instance.admin_up:
+    if instance.admin_state == constants.ADMINST_UP:

       self.LogInfo("Starting instance %s on node %s",
                    instance.name, target_node)
 
       self.LogInfo("Starting instance %s on node %s",
                    instance.name, target_node)
 

@@ -6891,8 +7199,8 @@ class LUInstanceMove(LogicalUnit):
         _ShutdownInstanceDisks(self, instance)
         raise errors.OpExecError("Can't activate the instance's disks")
 
         _ShutdownInstanceDisks(self, instance)
         raise errors.OpExecError("Can't activate the instance's disks")
 

-      result = self.rpc.call_instance_start(target_node, instance,
-                                            None, None, False)
+      result = self.rpc.call_instance_start(target_node,
+                                            (instance, None, None), False)

       msg = result.fail_msg
       if msg:
         _ShutdownInstanceDisks(self, instance)
       msg = result.fail_msg
       if msg:
         _ShutdownInstanceDisks(self, instance)

@@ -6985,6 +7293,11 @@ class TLMigrateInstance(Tasklet):
   @ivar shutdown_timeout: In case of failover timeout of the shutdown
 
   """
   @ivar shutdown_timeout: In case of failover timeout of the shutdown
 
   """

+
+  # Constants
+  _MIGRATION_POLL_INTERVAL = 1      # seconds
+  _MIGRATION_FEEDBACK_INTERVAL = 10 # seconds
+

   def __init__(self, lu, instance_name, cleanup=False,
                failover=False, fallback=False,
                ignore_consistency=False,
   def __init__(self, lu, instance_name, cleanup=False,
                failover=False, fallback=False,
                ignore_consistency=False,

@@ -7014,10 +7327,11 @@ class TLMigrateInstance(Tasklet):
     assert instance is not None
     self.instance = instance
 
     assert instance is not None
     self.instance = instance
 

-    if (not self.cleanup and not instance.admin_up and not self.failover and
-        self.fallback):
-      self.lu.LogInfo("Instance is marked down, fallback allowed, switching"
-                      " to failover")
+    if (not self.cleanup and
+        not instance.admin_state == constants.ADMINST_UP and
+        not self.failover and self.fallback):
+      self.lu.LogInfo("Instance is marked down or offline, fallback allowed,"
+                      " switching to failover")

       self.failover = True
 
     if instance.disk_template not in constants.DTS_MIRRORED:
       self.failover = True
 
     if instance.disk_template not in constants.DTS_MIRRORED:

@@ -7076,7 +7390,7 @@ class TLMigrateInstance(Tasklet):
     i_be = self.cfg.GetClusterInfo().FillBE(instance)
 
     # check memory requirements on the secondary node
     i_be = self.cfg.GetClusterInfo().FillBE(instance)
 
     # check memory requirements on the secondary node

-    if not self.failover or instance.admin_up:
+    if not self.failover or instance.admin_state == constants.ADMINST_UP:

       _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
                            instance.name, i_be[constants.BE_MEMORY],
                            instance.hypervisor)
       _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
                            instance.name, i_be[constants.BE_MEMORY],
                            instance.hypervisor)

@@ -7308,12 +7622,13 @@ class TLMigrateInstance(Tasklet):
     """
     instance = self.instance
     target_node = self.target_node
     """
     instance = self.instance
     target_node = self.target_node

+    source_node = self.source_node

     migration_info = self.migration_info
 
     migration_info = self.migration_info
 

-    abort_result = self.rpc.call_finalize_migration(target_node,
-                                                    instance,
-                                                    migration_info,
-                                                    False)
+    abort_result = self.rpc.call_instance_finalize_migration_dst(target_node,
+                                                                 instance,
+                                                                 migration_info,
+                                                                 False)

     abort_msg = abort_result.fail_msg
     if abort_msg:
       logging.error("Aborting migration failed on target node %s: %s",
     abort_msg = abort_result.fail_msg
     if abort_msg:
       logging.error("Aborting migration failed on target node %s: %s",

@@ -7321,6 +7636,13 @@ class TLMigrateInstance(Tasklet):
       # Don't raise an exception here, as we stil have to try to revert the
       # disk status, even if this step failed.
 
       # Don't raise an exception here, as we stil have to try to revert the
       # disk status, even if this step failed.
 

+    abort_result = self.rpc.call_instance_finalize_migration_src(source_node,
+        instance, False, self.live)
+    abort_msg = abort_result.fail_msg
+    if abort_msg:
+      logging.error("Aborting migration failed on source node %s: %s",
+                    source_node, abort_msg)
+

   def _ExecMigration(self):
     """Migrate an instance.
 
   def _ExecMigration(self):
     """Migrate an instance.
 

@@ -7337,6 +7659,21 @@ class TLMigrateInstance(Tasklet):
     target_node = self.target_node
     source_node = self.source_node
 
     target_node = self.target_node
     source_node = self.source_node
 

+    # Check for hypervisor version mismatch and warn the user.
+    nodeinfo = self.rpc.call_node_info([source_node, target_node],
+                                       None, self.instance.hypervisor)
+    src_info = nodeinfo[source_node]
+    dst_info = nodeinfo[target_node]
+
+    if ((constants.HV_NODEINFO_KEY_VERSION in src_info.payload) and
+        (constants.HV_NODEINFO_KEY_VERSION in dst_info.payload)):
+      src_version = src_info.payload[constants.HV_NODEINFO_KEY_VERSION]
+      dst_version = dst_info.payload[constants.HV_NODEINFO_KEY_VERSION]
+      if src_version != dst_version:
+        self.feedback_fn("* warning: hypervisor version mismatch between"
+                         " source (%s) and target (%s) node" %
+                         (src_version, dst_version))
+

     self.feedback_fn("* checking disk consistency between source and target")
     for dev in instance.disks:
       if not _CheckDiskConsistency(self.lu, dev, target_node, False):
     self.feedback_fn("* checking disk consistency between source and target")
     for dev in instance.disks:
       if not _CheckDiskConsistency(self.lu, dev, target_node, False):

@@ -7392,18 +7729,59 @@ class TLMigrateInstance(Tasklet):
       raise errors.OpExecError("Could not migrate instance %s: %s" %
                                (instance.name, msg))
 
       raise errors.OpExecError("Could not migrate instance %s: %s" %
                                (instance.name, msg))
 

+    self.feedback_fn("* starting memory transfer")
+    last_feedback = time.time()
+    while True:
+      result = self.rpc.call_instance_get_migration_status(source_node,
+                                                           instance)
+      msg = result.fail_msg
+      ms = result.payload   # MigrationStatus instance
+      if msg or (ms.status in constants.HV_MIGRATION_FAILED_STATUSES):
+        logging.error("Instance migration failed, trying to revert"
+                      " disk status: %s", msg)
+        self.feedback_fn("Migration failed, aborting")
+        self._AbortMigration()
+        self._RevertDiskStatus()
+        raise errors.OpExecError("Could not migrate instance %s: %s" %
+                                 (instance.name, msg))
+
+      if result.payload.status != constants.HV_MIGRATION_ACTIVE:
+        self.feedback_fn("* memory transfer complete")
+        break
+
+      if (utils.TimeoutExpired(last_feedback,
+                               self._MIGRATION_FEEDBACK_INTERVAL) and
+          ms.transferred_ram is not None):
+        mem_progress = 100 * float(ms.transferred_ram) / float(ms.total_ram)
+        self.feedback_fn("* memory transfer progress: %.2f %%" % mem_progress)
+        last_feedback = time.time()
+
+      time.sleep(self._MIGRATION_POLL_INTERVAL)
+
+    result = self.rpc.call_instance_finalize_migration_src(source_node,
+                                                           instance,
+                                                           True,
+                                                           self.live)
+    msg = result.fail_msg
+    if msg:
+      logging.error("Instance migration succeeded, but finalization failed"
+                    " on the source node: %s", msg)
+      raise errors.OpExecError("Could not finalize instance migration: %s" %
+                               msg)
+

     instance.primary_node = target_node
     instance.primary_node = target_node

+

     # distribute new instance config to the other nodes
     self.cfg.Update(instance, self.feedback_fn)
 
     # distribute new instance config to the other nodes
     self.cfg.Update(instance, self.feedback_fn)
 

-    result = self.rpc.call_finalize_migration(target_node,
-                                              instance,
-                                              migration_info,
-                                              True)
+    result = self.rpc.call_instance_finalize_migration_dst(target_node,
+                                                           instance,
+                                                           migration_info,
+                                                           True)

     msg = result.fail_msg
     if msg:
     msg = result.fail_msg
     if msg:

-      logging.error("Instance migration succeeded, but finalization failed:"
-                    " %s", msg)
+      logging.error("Instance migration succeeded, but finalization failed"
+                    " on the target node: %s", msg)

       raise errors.OpExecError("Could not finalize instance migration: %s" %
                                msg)
 
       raise errors.OpExecError("Could not finalize instance migration: %s" %
                                msg)
 

@@ -7429,7 +7807,7 @@ class TLMigrateInstance(Tasklet):
     source_node = instance.primary_node
     target_node = self.target_node
 
     source_node = instance.primary_node
     target_node = self.target_node
 

-    if instance.admin_up:
+    if instance.admin_state == constants.ADMINST_UP:

       self.feedback_fn("* checking disk consistency between source and target")
       for dev in instance.disks:
         # for drbd, these are drbd over lvm
       self.feedback_fn("* checking disk consistency between source and target")
       for dev in instance.disks:
         # for drbd, these are drbd over lvm

@@ -7472,7 +7850,7 @@ class TLMigrateInstance(Tasklet):
     self.cfg.Update(instance, self.feedback_fn)
 
     # Only start the instance if it's marked as up
     self.cfg.Update(instance, self.feedback_fn)
 
     # Only start the instance if it's marked as up

-    if instance.admin_up:
+    if instance.admin_state == constants.ADMINST_UP:

       self.feedback_fn("* activating the instance's disks on target node %s" %
                        target_node)
       logging.info("Starting instance %s on node %s",
       self.feedback_fn("* activating the instance's disks on target node %s" %
                        target_node)
       logging.info("Starting instance %s on node %s",

@@ -7486,7 +7864,7 @@ class TLMigrateInstance(Tasklet):
 
       self.feedback_fn("* starting the instance on the target node %s" %
                        target_node)
 
       self.feedback_fn("* starting the instance on the target node %s" %
                        target_node)

-      result = self.rpc.call_instance_start(target_node, instance, None, None,
+      result = self.rpc.call_instance_start(target_node, (instance, None, None),

                                             False)
       msg = result.fail_msg
       if msg:
                                             False)
       msg = result.fail_msg
       if msg:

@@ -7618,7 +7996,7 @@ def _GenerateDRBD8Branch(lu, primary, secondary, size, vgnames, names,
   shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
   dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
                           logical_id=(vgnames[0], names[0]))
   shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
   dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
                           logical_id=(vgnames[0], names[0]))

-  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
+  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,

                           logical_id=(vgnames[1], names[1]))
   drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
                           logical_id=(primary, secondary, port,
                           logical_id=(vgnames[1], names[1]))
   drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
                           logical_id=(primary, secondary, port,

@@ -7938,14 +8316,14 @@ def _ComputeDiskSizePerVG(disk_template, disks):
     constants.DT_DISKLESS: {},
     constants.DT_PLAIN: _compute(disks, 0),
     # 128 MB are added for drbd metadata for each disk
     constants.DT_DISKLESS: {},
     constants.DT_PLAIN: _compute(disks, 0),
     # 128 MB are added for drbd metadata for each disk

-    constants.DT_DRBD8: _compute(disks, 128),
+    constants.DT_DRBD8: _compute(disks, DRBD_META_SIZE),

     constants.DT_FILE: {},
     constants.DT_SHARED_FILE: {},
   }
 
   if disk_template not in req_size_dict:
     raise errors.ProgrammerError("Disk template '%s' size requirement"
     constants.DT_FILE: {},
     constants.DT_SHARED_FILE: {},
   }
 
   if disk_template not in req_size_dict:
     raise errors.ProgrammerError("Disk template '%s' size requirement"

-                                 " is unknown" %  disk_template)
+                                 " is unknown" % disk_template)

 
   return req_size_dict[disk_template]
 
 
   return req_size_dict[disk_template]
 

@@ -7959,7 +8337,8 @@ def _ComputeDiskSize(disk_template, disks):
     constants.DT_DISKLESS: None,
     constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
     # 128 MB are added for drbd metadata for each disk
     constants.DT_DISKLESS: None,
     constants.DT_PLAIN: sum(d[constants.IDISK_SIZE] for d in disks),
     # 128 MB are added for drbd metadata for each disk

-    constants.DT_DRBD8: sum(d[constants.IDISK_SIZE] + 128 for d in disks),
+    constants.DT_DRBD8:
+      sum(d[constants.IDISK_SIZE] + DRBD_META_SIZE for d in disks),

     constants.DT_FILE: None,
     constants.DT_SHARED_FILE: 0,
     constants.DT_BLOCK: 0,
     constants.DT_FILE: None,
     constants.DT_SHARED_FILE: 0,
     constants.DT_BLOCK: 0,

@@ -7967,7 +8346,7 @@ def _ComputeDiskSize(disk_template, disks):
 
   if disk_template not in req_size_dict:
     raise errors.ProgrammerError("Disk template '%s' size requirement"
 
   if disk_template not in req_size_dict:
     raise errors.ProgrammerError("Disk template '%s' size requirement"

-                                 " is unknown" %  disk_template)
+                                 " is unknown" % disk_template)

 
   return req_size_dict[disk_template]
 
 
   return req_size_dict[disk_template]
 

@@ -8005,9 +8384,11 @@ def _CheckHVParams(lu, nodenames, hvname, hvparams):
 
   """
   nodenames = _FilterVmNodes(lu, nodenames)
 
   """
   nodenames = _FilterVmNodes(lu, nodenames)

-  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
-                                                  hvname,
-                                                  hvparams)
+
+  cluster = lu.cfg.GetClusterInfo()
+  hvfull = objects.FillDict(cluster.hvparams.get(hvname, {}), hvparams)
+
+  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames, hvname, hvfull)

   for node in nodenames:
     info = hvinfo[node]
     if info.offline:
   for node in nodenames:
     info = hvinfo[node]
     if info.offline:

@@ -8033,7 +8414,7 @@ def _CheckOSParams(lu, required, nodenames, osname, osparams):
 
   """
   nodenames = _FilterVmNodes(lu, nodenames)
 
   """
   nodenames = _FilterVmNodes(lu, nodenames)

-  result = lu.rpc.call_os_validate(required, nodenames, osname,
+  result = lu.rpc.call_os_validate(nodenames, required, osname,

                                    [constants.OS_VALIDATE_PARAMETERS],
                                    osparams)
   for node, nres in result.items():
                                    [constants.OS_VALIDATE_PARAMETERS],
                                    osparams)
   for node, nres in result.items():

@@ -8226,7 +8607,11 @@ class LUInstanceCreate(LogicalUnit):
     self.add_locks[locking.LEVEL_INSTANCE] = instance_name
 
     if self.op.iallocator:
     self.add_locks[locking.LEVEL_INSTANCE] = instance_name
 
     if self.op.iallocator:

+      # TODO: Find a solution to not lock all nodes in the cluster, e.g. by
+      # specifying a group on instance creation and then selecting nodes from
+      # that group

       self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
       self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET

+      self.needed_locks[locking.LEVEL_NODE_RES] = locking.ALL_SET

     else:
       self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
       nodelist = [self.op.pnode]
     else:
       self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
       nodelist = [self.op.pnode]

@@ -8234,6 +8619,9 @@ class LUInstanceCreate(LogicalUnit):
         self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
         nodelist.append(self.op.snode)
       self.needed_locks[locking.LEVEL_NODE] = nodelist
         self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
         nodelist.append(self.op.snode)
       self.needed_locks[locking.LEVEL_NODE] = nodelist

+      # Lock resources of instance's primary and secondary nodes (copy to
+      # prevent accidential modification)
+      self.needed_locks[locking.LEVEL_NODE_RES] = list(nodelist)

 
     # in case of import lock the source node too
     if self.op.mode == constants.INSTANCE_IMPORT:
 
     # in case of import lock the source node too
     if self.op.mode == constants.INSTANCE_IMPORT:

@@ -8247,8 +8635,8 @@ class LUInstanceCreate(LogicalUnit):
         self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
         self.op.src_node = None
         if os.path.isabs(src_path):
         self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
         self.op.src_node = None
         if os.path.isabs(src_path):

-          raise errors.OpPrereqError("Importing an instance from an absolute"
-                                     " path requires a source node option",
+          raise errors.OpPrereqError("Importing an instance from a path"
+                                     " requires a source node option",

                                      errors.ECODE_INVAL)
       else:
         self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
                                      errors.ECODE_INVAL)
       else:
         self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)

@@ -8397,33 +8785,39 @@ class LUInstanceCreate(LogicalUnit):
       if einfo.has_option(constants.INISECT_INS, "disk_template"):
         self.op.disk_template = einfo.get(constants.INISECT_INS,
                                           "disk_template")
       if einfo.has_option(constants.INISECT_INS, "disk_template"):
         self.op.disk_template = einfo.get(constants.INISECT_INS,
                                           "disk_template")

+        if self.op.disk_template not in constants.DISK_TEMPLATES:
+          raise errors.OpPrereqError("Disk template specified in configuration"
+                                     " file is not one of the allowed values:"
+                                     " %s" % " ".join(constants.DISK_TEMPLATES))

       else:
         raise errors.OpPrereqError("No disk template specified and the export"
                                    " is missing the disk_template information",
                                    errors.ECODE_INVAL)
 
     if not self.op.disks:
       else:
         raise errors.OpPrereqError("No disk template specified and the export"
                                    " is missing the disk_template information",
                                    errors.ECODE_INVAL)
 
     if not self.op.disks:

-      if einfo.has_option(constants.INISECT_INS, "disk_count"):
-        disks = []
-        # TODO: import the disk iv_name too
-        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
+      disks = []
+      # TODO: import the disk iv_name too
+      for idx in range(constants.MAX_DISKS):
+        if einfo.has_option(constants.INISECT_INS, "disk%d_size" % idx):

           disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
           disks.append({constants.IDISK_SIZE: disk_sz})
           disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
           disks.append({constants.IDISK_SIZE: disk_sz})

-        self.op.disks = disks
-      else:
+      self.op.disks = disks
+      if not disks and self.op.disk_template != constants.DT_DISKLESS:

         raise errors.OpPrereqError("No disk info specified and the export"
                                    " is missing the disk information",
                                    errors.ECODE_INVAL)
 
         raise errors.OpPrereqError("No disk info specified and the export"
                                    " is missing the disk information",
                                    errors.ECODE_INVAL)
 

-    if (not self.op.nics and
-        einfo.has_option(constants.INISECT_INS, "nic_count")):
+    if not self.op.nics:

       nics = []
       nics = []

-      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
-        ndict = {}
-        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
-          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
-          ndict[name] = v
-        nics.append(ndict)
+      for idx in range(constants.MAX_NICS):
+        if einfo.has_option(constants.INISECT_INS, "nic%d_mac" % idx):
+          ndict = {}
+          for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
+            v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
+            ndict[name] = v
+          nics.append(ndict)
+        else:
+          break

       self.op.nics = nics
 
     if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):
       self.op.nics = nics
 
     if not self.op.tags and einfo.has_option(constants.INISECT_INS, "tags"):

@@ -8509,7 +8903,7 @@ class LUInstanceCreate(LogicalUnit):
 
       joinargs.append(self.op.instance_name)
 
 
       joinargs.append(self.op.instance_name)
 

-      # pylint: disable-msg=W0142
+      # pylint: disable=W0142

       self.instance_file_storage_dir = utils.PathJoin(*joinargs)
 
   def CheckPrereq(self):
       self.instance_file_storage_dir = utils.PathJoin(*joinargs)
 
   def CheckPrereq(self):

@@ -8527,7 +8921,8 @@ class LUInstanceCreate(LogicalUnit):
       raise errors.OpPrereqError("Cluster does not support lvm-based"
                                  " instances", errors.ECODE_STATE)
 
       raise errors.OpPrereqError("Cluster does not support lvm-based"
                                  " instances", errors.ECODE_STATE)
 

-    if self.op.hypervisor is None:
+    if (self.op.hypervisor is None or
+        self.op.hypervisor == constants.VALUE_AUTO):

       self.op.hypervisor = self.cfg.GetHypervisorType()
 
     cluster = self.cfg.GetClusterInfo()
       self.op.hypervisor = self.cfg.GetHypervisorType()
 
     cluster = self.cfg.GetClusterInfo()

@@ -8553,6 +8948,10 @@ class LUInstanceCreate(LogicalUnit):
     _CheckGlobalHvParams(self.op.hvparams)
 
     # fill and remember the beparams dict
     _CheckGlobalHvParams(self.op.hvparams)
 
     # fill and remember the beparams dict

+    default_beparams = cluster.beparams[constants.PP_DEFAULT]
+    for param, value in self.op.beparams.iteritems():
+      if value == constants.VALUE_AUTO:
+        self.op.beparams[param] = default_beparams[param]

     utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
     self.be_full = cluster.SimpleFillBE(self.op.beparams)
 
     utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
     self.be_full = cluster.SimpleFillBE(self.op.beparams)
 

@@ -8569,7 +8968,7 @@ class LUInstanceCreate(LogicalUnit):
     for idx, nic in enumerate(self.op.nics):
       nic_mode_req = nic.get(constants.INIC_MODE, None)
       nic_mode = nic_mode_req
     for idx, nic in enumerate(self.op.nics):
       nic_mode_req = nic.get(constants.INIC_MODE, None)
       nic_mode = nic_mode_req

-      if nic_mode is None:
+      if nic_mode is None or nic_mode == constants.VALUE_AUTO:

         nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
 
       # in routed mode, for the first nic, the default ip is 'auto'
         nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
 
       # in routed mode, for the first nic, the default ip is 'auto'

@@ -8613,9 +9012,11 @@ class LUInstanceCreate(LogicalUnit):
 
       #  Build nic parameters
       link = nic.get(constants.INIC_LINK, None)
 
       #  Build nic parameters
       link = nic.get(constants.INIC_LINK, None)

+      if link == constants.VALUE_AUTO:
+        link = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_LINK]

       nicparams = {}
       if nic_mode_req:
       nicparams = {}
       if nic_mode_req:

-        nicparams[constants.NIC_MODE] = nic_mode_req
+        nicparams[constants.NIC_MODE] = nic_mode

       if link:
         nicparams[constants.NIC_LINK] = link
 
       if link:
         nicparams[constants.NIC_LINK] = link
 

@@ -8652,18 +9053,8 @@ class LUInstanceCreate(LogicalUnit):
       self.disks.append(new_disk)
 
     if self.op.mode == constants.INSTANCE_IMPORT:
       self.disks.append(new_disk)
 
     if self.op.mode == constants.INSTANCE_IMPORT:

-
-      # Check that the new instance doesn't have less disks than the export
-      instance_disks = len(self.disks)
-      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
-      if instance_disks < export_disks:
-        raise errors.OpPrereqError("Not enough disks to import."
-                                   " (instance: %d, export: %d)" %
-                                   (instance_disks, export_disks),
-                                   errors.ECODE_INVAL)
-

       disk_images = []
       disk_images = []

-      for idx in range(export_disks):
+      for idx in range(len(self.disks)):

         option = "disk%d_dump" % idx
         if export_info.has_option(constants.INISECT_INS, option):
           # FIXME: are the old os-es, disk sizes, etc. useful?
         option = "disk%d_dump" % idx
         if export_info.has_option(constants.INISECT_INS, option):
           # FIXME: are the old os-es, disk sizes, etc. useful?

@@ -8676,15 +9067,9 @@ class LUInstanceCreate(LogicalUnit):
       self.src_images = disk_images
 
       old_name = export_info.get(constants.INISECT_INS, "name")
       self.src_images = disk_images
 
       old_name = export_info.get(constants.INISECT_INS, "name")

-      try:
-        exp_nic_count = export_info.getint(constants.INISECT_INS, "nic_count")
-      except (TypeError, ValueError), err:
-        raise errors.OpPrereqError("Invalid export file, nic_count is not"
-                                   " an integer: %s" % str(err),
-                                   errors.ECODE_STATE)

       if self.op.instance_name == old_name:
         for idx, nic in enumerate(self.nics):
       if self.op.instance_name == old_name:
         for idx, nic in enumerate(self.nics):

-          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
+          if nic.mac == constants.VALUE_AUTO:

             nic_mac_ini = "nic%d_mac" % idx
             nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
 
             nic_mac_ini = "nic%d_mac" % idx
             nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
 

@@ -8843,6 +9228,10 @@ class LUInstanceCreate(LogicalUnit):
     instance = self.op.instance_name
     pnode_name = self.pnode.name
 
     instance = self.op.instance_name
     pnode_name = self.pnode.name
 

+    assert not (self.owned_locks(locking.LEVEL_NODE_RES) -
+                self.owned_locks(locking.LEVEL_NODE)), \
+      "Node locks differ from node resource locks"
+

     ht_kind = self.op.hypervisor
     if ht_kind in constants.HTS_REQ_PORT:
       network_port = self.cfg.AllocatePort()
     ht_kind = self.op.hypervisor
     if ht_kind in constants.HTS_REQ_PORT:
       network_port = self.cfg.AllocatePort()

@@ -8863,7 +9252,7 @@ class LUInstanceCreate(LogicalUnit):
                             primary_node=pnode_name,
                             nics=self.nics, disks=disks,
                             disk_template=self.op.disk_template,
                             primary_node=pnode_name,
                             nics=self.nics, disks=disks,
                             disk_template=self.op.disk_template,

-                            admin_up=False,
+                            admin_state=constants.ADMINST_DOWN,

                             network_port=network_port,
                             beparams=self.op.beparams,
                             hvparams=self.op.hvparams,
                             network_port=network_port,
                             beparams=self.op.beparams,
                             hvparams=self.op.hvparams,

@@ -8881,7 +9270,7 @@ class LUInstanceCreate(LogicalUnit):
         # 'fake' LV disks with the old data, plus the new unique_id
         tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
         rename_to = []
         # 'fake' LV disks with the old data, plus the new unique_id
         tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
         rename_to = []

-        for t_dsk, a_dsk in zip (tmp_disks, self.disks):
+        for t_dsk, a_dsk in zip(tmp_disks, self.disks):

           rename_to.append(t_dsk.logical_id)
           t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
           self.cfg.SetDiskID(t_dsk, pnode_name)
           rename_to.append(t_dsk.logical_id)
           t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk[constants.IDISK_ADOPT])
           self.cfg.SetDiskID(t_dsk, pnode_name)

@@ -8945,6 +9334,9 @@ class LUInstanceCreate(LogicalUnit):
       raise errors.OpExecError("There are some degraded disks for"
                                " this instance")
 
       raise errors.OpExecError("There are some degraded disks for"
                                " this instance")
 

+    # Release all node resource locks
+    _ReleaseLocks(self, locking.LEVEL_NODE_RES)
+

     if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
       if self.op.mode == constants.INSTANCE_CREATE:
         if not self.op.no_install:
     if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
       if self.op.mode == constants.INSTANCE_CREATE:
         if not self.op.no_install:

@@ -8961,8 +9353,9 @@ class LUInstanceCreate(LogicalUnit):
 
           feedback_fn("* running the instance OS create scripts...")
           # FIXME: pass debug option from opcode to backend
 
           feedback_fn("* running the instance OS create scripts...")
           # FIXME: pass debug option from opcode to backend

-          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
-                                                 self.op.debug_level)
+          os_add_result = \
+            self.rpc.call_instance_os_add(pnode_name, (iobj, None), False,
+                                          self.op.debug_level)

           if pause_sync:
             feedback_fn("* resuming disk sync")
             result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
           if pause_sync:
             feedback_fn("* resuming disk sync")
             result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,

@@ -8972,8 +9365,8 @@ class LUInstanceCreate(LogicalUnit):
                 logging.warn("resume-sync of instance %s for disk %d failed",
                              instance, idx)
 
                 logging.warn("resume-sync of instance %s for disk %d failed",
                              instance, idx)
 

-          result.Raise("Could not add os for instance %s"
-                       " on node %s" % (instance, pnode_name))
+          os_add_result.Raise("Could not add os for instance %s"
+                              " on node %s" % (instance, pnode_name))

 
       elif self.op.mode == constants.INSTANCE_IMPORT:
         feedback_fn("* running the instance OS import scripts...")
 
       elif self.op.mode == constants.INSTANCE_IMPORT:
         feedback_fn("* running the instance OS import scripts...")

@@ -9036,13 +9429,15 @@ class LUInstanceCreate(LogicalUnit):
         raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
                                      % self.op.mode)
 
         raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
                                      % self.op.mode)
 

+    assert not self.owned_locks(locking.LEVEL_NODE_RES)
+

     if self.op.start:
     if self.op.start:

-      iobj.admin_up = True
+      iobj.admin_state = constants.ADMINST_UP

       self.cfg.Update(iobj, feedback_fn)
       logging.info("Starting instance %s on node %s", instance, pnode_name)
       feedback_fn("* starting instance...")
       self.cfg.Update(iobj, feedback_fn)
       logging.info("Starting instance %s on node %s", instance, pnode_name)
       feedback_fn("* starting instance...")

-      result = self.rpc.call_instance_start(pnode_name, iobj,
-                                            None, None, False)
+      result = self.rpc.call_instance_start(pnode_name, (iobj, None, None),
+                                            False)

       result.Raise("Could not start instance")
 
     return list(iobj.all_nodes)
       result.Raise("Could not start instance")
 
     return list(iobj.all_nodes)

@@ -9059,6 +9454,7 @@ class LUInstanceConsole(NoHooksLU):
   REQ_BGL = False
 
   def ExpandNames(self):
   REQ_BGL = False
 
   def ExpandNames(self):

+    self.share_locks = _ShareAll()

     self._ExpandAndLockInstance()
 
   def CheckPrereq(self):
     self._ExpandAndLockInstance()
 
   def CheckPrereq(self):

@@ -9084,10 +9480,12 @@ class LUInstanceConsole(NoHooksLU):
     node_insts.Raise("Can't get node information from %s" % node)
 
     if instance.name not in node_insts.payload:
     node_insts.Raise("Can't get node information from %s" % node)
 
     if instance.name not in node_insts.payload:

-      if instance.admin_up:
+      if instance.admin_state == constants.ADMINST_UP:

         state = constants.INSTST_ERRORDOWN
         state = constants.INSTST_ERRORDOWN

-      else:
+      elif instance.admin_state == constants.ADMINST_DOWN:

         state = constants.INSTST_ADMINDOWN
         state = constants.INSTST_ADMINDOWN

+      else:
+        state = constants.INSTST_ADMINOFFLINE

       raise errors.OpExecError("Instance %s is not running (state %s)" %
                                (instance.name, state))
 
       raise errors.OpExecError("Instance %s is not running (state %s)" %
                                (instance.name, state))
 

@@ -9133,6 +9531,7 @@ class LUInstanceReplaceDisks(LogicalUnit):
     self._ExpandAndLockInstance()
 
     assert locking.LEVEL_NODE not in self.needed_locks
     self._ExpandAndLockInstance()
 
     assert locking.LEVEL_NODE not in self.needed_locks

+    assert locking.LEVEL_NODE_RES not in self.needed_locks

     assert locking.LEVEL_NODEGROUP not in self.needed_locks
 
     assert self.op.iallocator is None or self.op.remote_node is None, \
     assert locking.LEVEL_NODEGROUP not in self.needed_locks
 
     assert self.op.iallocator is None or self.op.remote_node is None, \

@@ -9155,6 +9554,8 @@ class LUInstanceReplaceDisks(LogicalUnit):
         # iallocator will select a new node in the same group
         self.needed_locks[locking.LEVEL_NODEGROUP] = []
 
         # iallocator will select a new node in the same group
         self.needed_locks[locking.LEVEL_NODEGROUP] = []
 

+    self.needed_locks[locking.LEVEL_NODE_RES] = []
+

     self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
                                    self.op.iallocator, self.op.remote_node,
                                    self.op.disks, False, self.op.early_release)
     self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
                                    self.op.iallocator, self.op.remote_node,
                                    self.op.disks, False, self.op.early_release)

@@ -9168,6 +9569,8 @@ class LUInstanceReplaceDisks(LogicalUnit):
       assert not self.needed_locks[locking.LEVEL_NODEGROUP]
 
       self.share_locks[locking.LEVEL_NODEGROUP] = 1
       assert not self.needed_locks[locking.LEVEL_NODEGROUP]
 
       self.share_locks[locking.LEVEL_NODEGROUP] = 1

+      # Lock all groups used by instance optimistically; this requires going
+      # via the node before it's locked, requiring verification later on

       self.needed_locks[locking.LEVEL_NODEGROUP] = \
         self.cfg.GetInstanceNodeGroups(self.op.instance_name)
 
       self.needed_locks[locking.LEVEL_NODEGROUP] = \
         self.cfg.GetInstanceNodeGroups(self.op.instance_name)
 

@@ -9182,6 +9585,10 @@ class LUInstanceReplaceDisks(LogicalUnit):
           for node_name in self.cfg.GetNodeGroup(group_uuid).members]
       else:
         self._LockInstancesNodes()
           for node_name in self.cfg.GetNodeGroup(group_uuid).members]
       else:
         self._LockInstancesNodes()

+    elif level == locking.LEVEL_NODE_RES:
+      # Reuse node locks
+      self.needed_locks[locking.LEVEL_NODE_RES] = \
+        self.needed_locks[locking.LEVEL_NODE]

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -9218,6 +9625,7 @@ class LUInstanceReplaceDisks(LogicalUnit):
     assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
             self.op.iallocator is None)
 
     assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
             self.op.iallocator is None)
 

+    # Verify if node group locks are still correct

     owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
     if owned_groups:
       _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)
     owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
     if owned_groups:
       _CheckInstanceNodeGroups(self.cfg, self.op.instance_name, owned_groups)

@@ -9476,8 +9884,9 @@ class TLReplaceDisks(Tasklet):
                                                           self.target_node]
                               if node_name is not None)
 
                                                           self.target_node]
                               if node_name is not None)
 

-    # Release unneeded node locks
+    # Release unneeded node and node resource locks

     _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
     _ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)

+    _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES, keep=touched_nodes)

 
     # Release any owned node group
     if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):
 
     # Release any owned node group
     if self.lu.glm.is_owned(locking.LEVEL_NODEGROUP):

@@ -9506,6 +9915,8 @@ class TLReplaceDisks(Tasklet):
       assert set(owned_nodes) == set(self.node_secondary_ip), \
           ("Incorrect node locks, owning %s, expected %s" %
            (owned_nodes, self.node_secondary_ip.keys()))
       assert set(owned_nodes) == set(self.node_secondary_ip), \
           ("Incorrect node locks, owning %s, expected %s" %
            (owned_nodes, self.node_secondary_ip.keys()))

+      assert (self.lu.owned_locks(locking.LEVEL_NODE) ==
+              self.lu.owned_locks(locking.LEVEL_NODE_RES))

 
       owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
       assert list(owned_instances) == [self.instance_name], \
 
       owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
       assert list(owned_instances) == [self.instance_name], \

@@ -9521,7 +9932,7 @@ class TLReplaceDisks(Tasklet):
     feedback_fn("Replacing disk(s) %s for %s" %
                 (utils.CommaJoin(self.disks), self.instance.name))
 
     feedback_fn("Replacing disk(s) %s for %s" %
                 (utils.CommaJoin(self.disks), self.instance.name))
 

-    activate_disks = (not self.instance.admin_up)
+    activate_disks = (self.instance.admin_state != constants.ADMINST_UP)

 
     # Activate the instance disks if we're replacing them on a down instance
     if activate_disks:
 
     # Activate the instance disks if we're replacing them on a down instance
     if activate_disks:

@@ -9541,9 +9952,11 @@ class TLReplaceDisks(Tasklet):
       if activate_disks:
         _SafeShutdownInstanceDisks(self.lu, self.instance)
 
       if activate_disks:
         _SafeShutdownInstanceDisks(self.lu, self.instance)
 

+    assert not self.lu.owned_locks(locking.LEVEL_NODE)
+

     if __debug__:
       # Verify owned locks
     if __debug__:
       # Verify owned locks

-      owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
+      owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE_RES)

       nodes = frozenset(self.node_secondary_ip)
       assert ((self.early_release and not owned_nodes) or
               (not self.early_release and not (set(owned_nodes) - nodes))), \
       nodes = frozenset(self.node_secondary_ip)
       assert ((self.early_release and not owned_nodes) or
               (not self.early_release and not (set(owned_nodes) - nodes))), \

@@ -9626,7 +10039,7 @@ class TLReplaceDisks(Tasklet):
       lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
                              logical_id=(vg_data, names[0]))
       vg_meta = dev.children[1].logical_id[0]
       lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
                              logical_id=(vg_data, names[0]))
       vg_meta = dev.children[1].logical_id[0]

-      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
+      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=DRBD_META_SIZE,

                              logical_id=(vg_meta, names[1]))
 
       new_lvs = [lv_data, lv_meta]
                              logical_id=(vg_meta, names[1]))
 
       new_lvs = [lv_data, lv_meta]

@@ -9668,7 +10081,7 @@ class TLReplaceDisks(Tasklet):
           self.lu.LogWarning("Can't remove old LV: %s" % msg,
                              hint="remove unused LVs manually")
 
           self.lu.LogWarning("Can't remove old LV: %s" % msg,
                              hint="remove unused LVs manually")
 

-  def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable-msg=W0613
+  def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613

     """Replace a disk on the primary or secondary for DRBD 8.
 
     The algorithm for replace is quite complicated:
     """Replace a disk on the primary or secondary for DRBD 8.
 
     The algorithm for replace is quite complicated:

@@ -9783,10 +10196,18 @@ class TLReplaceDisks(Tasklet):
       self.lu.LogStep(cstep, steps_total, "Removing old storage")
       cstep += 1
       self._RemoveOldStorage(self.target_node, iv_names)
       self.lu.LogStep(cstep, steps_total, "Removing old storage")
       cstep += 1
       self._RemoveOldStorage(self.target_node, iv_names)

-      # WARNING: we release both node locks here, do not do other RPCs
-      # than WaitForSync to the primary node
-      _ReleaseLocks(self.lu, locking.LEVEL_NODE,
-                    names=[self.target_node, self.other_node])
+      # TODO: Check if releasing locks early still makes sense
+      _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
+    else:
+      # Release all resource locks except those used by the instance
+      _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
+                    keep=self.node_secondary_ip.keys())
+
+    # Release all node locks while waiting for sync
+    _ReleaseLocks(self.lu, locking.LEVEL_NODE)
+
+    # TODO: Can the instance lock be downgraded here? Take the optional disk
+    # shutdown in the caller into consideration.

 
     # Wait for sync
     # This can fail as the old devices are degraded and _WaitForSync
 
     # Wait for sync
     # This can fail as the old devices are degraded and _WaitForSync

@@ -9825,6 +10246,8 @@ class TLReplaceDisks(Tasklet):
     """
     steps_total = 6
 
     """
     steps_total = 6
 

+    pnode = self.instance.primary_node
+

     # Step: check device activation
     self.lu.LogStep(1, steps_total, "Check device existence")
     self._CheckDisksExistence([self.instance.primary_node])
     # Step: check device activation
     self.lu.LogStep(1, steps_total, "Check device existence")
     self._CheckDisksExistence([self.instance.primary_node])

@@ -9899,10 +10322,8 @@ class TLReplaceDisks(Tasklet):
                                  " soon as possible"))
 
     self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
                                  " soon as possible"))
 
     self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")

-    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
-                                               self.node_secondary_ip,
-                                               self.instance.disks)\
-                                              [self.instance.primary_node]
+    result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
+                                               self.instance.disks)[pnode]

 
     msg = result.fail_msg
     if msg:
 
     msg = result.fail_msg
     if msg:

@@ -9920,6 +10341,9 @@ class TLReplaceDisks(Tasklet):
 
     self.cfg.Update(self.instance, feedback_fn)
 
 
     self.cfg.Update(self.instance, feedback_fn)
 

+    # Release all node locks (the configuration has been updated)
+    _ReleaseLocks(self.lu, locking.LEVEL_NODE)
+

     # and now perform the drbd attach
     self.lu.LogInfo("Attaching primary drbds to new secondary"
                     " (standalone => connected)")
     # and now perform the drbd attach
     self.lu.LogInfo("Attaching primary drbds to new secondary"
                     " (standalone => connected)")

@@ -9941,12 +10365,15 @@ class TLReplaceDisks(Tasklet):
       self.lu.LogStep(cstep, steps_total, "Removing old storage")
       cstep += 1
       self._RemoveOldStorage(self.target_node, iv_names)
       self.lu.LogStep(cstep, steps_total, "Removing old storage")
       cstep += 1
       self._RemoveOldStorage(self.target_node, iv_names)

-      # WARNING: we release all node locks here, do not do other RPCs
-      # than WaitForSync to the primary node
-      _ReleaseLocks(self.lu, locking.LEVEL_NODE,
-                    names=[self.instance.primary_node,
-                           self.target_node,
-                           self.new_node])
+      # TODO: Check if releasing locks early still makes sense
+      _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
+    else:
+      # Release all resource locks except those used by the instance
+      _ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
+                    keep=self.node_secondary_ip.keys())
+
+    # TODO: Can the instance lock be downgraded here? Take the optional disk
+    # shutdown in the caller into consideration.

 
     # Wait for sync
     # This can fail as the old devices are degraded and _WaitForSync
 
     # Wait for sync
     # This can fail as the old devices are degraded and _WaitForSync

@@ -10006,7 +10433,7 @@ class LURepairNodeStorage(NoHooksLU):
     """
     # Check whether any instance on this node has faulty disks
     for inst in _GetNodeInstances(self.cfg, self.op.node_name):
     """
     # Check whether any instance on this node has faulty disks
     for inst in _GetNodeInstances(self.cfg, self.op.node_name):

-      if not inst.admin_up:
+      if inst.admin_state != constants.ADMINST_UP:

         continue
       check_nodes = set(inst.all_nodes)
       check_nodes.discard(self.op.node_name)
         continue
       check_nodes = set(inst.all_nodes)
       check_nodes.discard(self.op.node_name)

@@ -10256,11 +10683,16 @@ class LUInstanceGrowDisk(LogicalUnit):
   def ExpandNames(self):
     self._ExpandAndLockInstance()
     self.needed_locks[locking.LEVEL_NODE] = []
   def ExpandNames(self):
     self._ExpandAndLockInstance()
     self.needed_locks[locking.LEVEL_NODE] = []

-    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
+    self.needed_locks[locking.LEVEL_NODE_RES] = []
+    self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE

 
   def DeclareLocks(self, level):
     if level == locking.LEVEL_NODE:
       self._LockInstancesNodes()
 
   def DeclareLocks(self, level):
     if level == locking.LEVEL_NODE:
       self._LockInstancesNodes()

+    elif level == locking.LEVEL_NODE_RES:
+      # Copy node locks
+      self.needed_locks[locking.LEVEL_NODE_RES] = \
+        self.needed_locks[locking.LEVEL_NODE][:]

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -10317,10 +10749,18 @@ class LUInstanceGrowDisk(LogicalUnit):
     instance = self.instance
     disk = self.disk
 
     instance = self.instance
     disk = self.disk
 

+    assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
+    assert (self.owned_locks(locking.LEVEL_NODE) ==
+            self.owned_locks(locking.LEVEL_NODE_RES))
+

     disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
     if not disks_ok:
       raise errors.OpExecError("Cannot activate block device to grow")
 
     disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
     if not disks_ok:
       raise errors.OpExecError("Cannot activate block device to grow")
 

+    feedback_fn("Growing disk %s of instance '%s' by %s" %
+                (self.op.disk, instance.name,
+                 utils.FormatUnit(self.op.amount, "h")))
+

     # First run all grow ops in dry-run mode
     for node in instance.all_nodes:
       self.cfg.SetDiskID(disk, node)
     # First run all grow ops in dry-run mode
     for node in instance.all_nodes:
       self.cfg.SetDiskID(disk, node)

@@ -10343,18 +10783,28 @@ class LUInstanceGrowDisk(LogicalUnit):
 
     disk.RecordGrow(self.op.amount)
     self.cfg.Update(instance, feedback_fn)
 
     disk.RecordGrow(self.op.amount)
     self.cfg.Update(instance, feedback_fn)

+
+    # Changes have been recorded, release node lock
+    _ReleaseLocks(self, locking.LEVEL_NODE)
+
+    # Downgrade lock while waiting for sync
+    self.glm.downgrade(locking.LEVEL_INSTANCE)
+

     if self.op.wait_for_sync:
       disk_abort = not _WaitForSync(self, instance, disks=[disk])
       if disk_abort:
         self.proc.LogWarning("Disk sync-ing has not returned a good"
                              " status; please check the instance")
     if self.op.wait_for_sync:
       disk_abort = not _WaitForSync(self, instance, disks=[disk])
       if disk_abort:
         self.proc.LogWarning("Disk sync-ing has not returned a good"
                              " status; please check the instance")

-      if not instance.admin_up:
+      if instance.admin_state != constants.ADMINST_UP:

         _SafeShutdownInstanceDisks(self, instance, disks=[disk])
         _SafeShutdownInstanceDisks(self, instance, disks=[disk])

-    elif not instance.admin_up:
+    elif instance.admin_state != constants.ADMINST_UP:

       self.proc.LogWarning("Not shutting down the disk even if the instance is"
                            " not supposed to be running because no wait for"
                            " sync mode was requested")
 
       self.proc.LogWarning("Not shutting down the disk even if the instance is"
                            " not supposed to be running because no wait for"
                            " sync mode was requested")
 

+    assert self.owned_locks(locking.LEVEL_NODE_RES)
+    assert set([instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
+

 
 class LUInstanceQueryData(NoHooksLU):
   """Query runtime instance data.
 
 class LUInstanceQueryData(NoHooksLU):
   """Query runtime instance data.

@@ -10486,19 +10936,17 @@ class LUInstanceQueryData(NoHooksLU):
         if remote_info and "state" in remote_info:
           remote_state = "up"
         else:
         if remote_info and "state" in remote_info:
           remote_state = "up"
         else:

-          remote_state = "down"
-
-      if instance.admin_up:
-        config_state = "up"
-      else:
-        config_state = "down"
+          if instance.admin_state == constants.ADMINST_UP:
+            remote_state = "down"
+          else:
+            remote_state = instance.admin_state

 
       disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
                   instance.disks)
 
       result[instance.name] = {
         "name": instance.name,
 
       disks = map(compat.partial(self._ComputeDiskStatus, instance, None),
                   instance.disks)
 
       result[instance.name] = {
         "name": instance.name,

-        "config_state": config_state,
+        "config_state": instance.admin_state,

         "run_state": remote_state,
         "pnode": instance.primary_node,
         "snodes": instance.secondary_nodes,
         "run_state": remote_state,
         "pnode": instance.primary_node,
         "snodes": instance.secondary_nodes,

@@ -10534,7 +10982,8 @@ class LUInstanceSetParams(LogicalUnit):
 
   def CheckArguments(self):
     if not (self.op.nics or self.op.disks or self.op.disk_template or
 
   def CheckArguments(self):
     if not (self.op.nics or self.op.disks or self.op.disk_template or

-            self.op.hvparams or self.op.beparams or self.op.os_name):
+            self.op.hvparams or self.op.beparams or self.op.os_name or
+            self.op.online_inst or self.op.offline_inst):

       raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
 
     if self.op.hvparams:
       raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
 
     if self.op.hvparams:

@@ -10650,7 +11099,10 @@ class LUInstanceSetParams(LogicalUnit):
 
   def ExpandNames(self):
     self._ExpandAndLockInstance()
 
   def ExpandNames(self):
     self._ExpandAndLockInstance()

+    # Can't even acquire node locks in shared mode as upcoming changes in
+    # Ganeti 2.6 will start to modify the node object on disk conversion

     self.needed_locks[locking.LEVEL_NODE] = []
     self.needed_locks[locking.LEVEL_NODE] = []

+    self.needed_locks[locking.LEVEL_NODE_RES] = []

     self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
 
   def DeclareLocks(self, level):
     self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
 
   def DeclareLocks(self, level):

@@ -10659,6 +11111,10 @@ class LUInstanceSetParams(LogicalUnit):
       if self.op.disk_template and self.op.remote_node:
         self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
         self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
       if self.op.disk_template and self.op.remote_node:
         self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
         self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)

+    elif level == locking.LEVEL_NODE_RES and self.op.disk_template:
+      # Copy node locks
+      self.needed_locks[locking.LEVEL_NODE_RES] = \
+        self.needed_locks[locking.LEVEL_NODE][:]

 
   def BuildHooksEnv(self):
     """Build hooks env.
 
   def BuildHooksEnv(self):
     """Build hooks env.

@@ -10753,7 +11209,8 @@ class LUInstanceSetParams(LogicalUnit):
                                    " %s to %s" % (instance.disk_template,
                                                   self.op.disk_template),
                                    errors.ECODE_INVAL)
                                    " %s to %s" % (instance.disk_template,
                                                   self.op.disk_template),
                                    errors.ECODE_INVAL)

-      _CheckInstanceDown(self, instance, "cannot change disk template")
+      _CheckInstanceState(self, instance, INSTANCE_DOWN,
+                          msg="cannot change disk template")

       if self.op.disk_template in constants.DTS_INT_MIRROR:
         if self.op.remote_node == pnode:
           raise errors.OpPrereqError("Given new secondary node %s is the same"
       if self.op.disk_template in constants.DTS_INT_MIRROR:
         if self.op.remote_node == pnode:
           raise errors.OpPrereqError("Given new secondary node %s is the same"

@@ -10779,9 +11236,11 @@ class LUInstanceSetParams(LogicalUnit):
       # local check
       hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
       _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
       # local check
       hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
       _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)

-      self.hv_new = hv_new # the new actual values
+      self.hv_proposed = self.hv_new = hv_new # the new actual values

       self.hv_inst = i_hvdict # the new dict (without defaults)
     else:
       self.hv_inst = i_hvdict # the new dict (without defaults)
     else:

+      self.hv_proposed = cluster.SimpleFillHV(instance.hypervisor, instance.os,
+                                              instance.hvparams)

       self.hv_new = self.hv_inst = {}
 
     # beparams processing
       self.hv_new = self.hv_inst = {}
 
     # beparams processing

@@ -10790,12 +11249,40 @@ class LUInstanceSetParams(LogicalUnit):
                                    use_none=True)
       utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
       be_new = cluster.SimpleFillBE(i_bedict)
                                    use_none=True)
       utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
       be_new = cluster.SimpleFillBE(i_bedict)

-      self.be_new = be_new # the new actual values
+      self.be_proposed = self.be_new = be_new # the new actual values

       self.be_inst = i_bedict # the new dict (without defaults)
     else:
       self.be_new = self.be_inst = {}
       self.be_inst = i_bedict # the new dict (without defaults)
     else:
       self.be_new = self.be_inst = {}

+      self.be_proposed = cluster.SimpleFillBE(instance.beparams)

     be_old = cluster.FillBE(instance)
 
     be_old = cluster.FillBE(instance)
 

+    # CPU param validation -- checking every time a paramtere is
+    # changed to cover all cases where either CPU mask or vcpus have
+    # changed
+    if (constants.BE_VCPUS in self.be_proposed and
+        constants.HV_CPU_MASK in self.hv_proposed):
+      cpu_list = \
+        utils.ParseMultiCpuMask(self.hv_proposed[constants.HV_CPU_MASK])
+      # Verify mask is consistent with number of vCPUs. Can skip this
+      # test if only 1 entry in the CPU mask, which means same mask
+      # is applied to all vCPUs.
+      if (len(cpu_list) > 1 and
+          len(cpu_list) != self.be_proposed[constants.BE_VCPUS]):
+        raise errors.OpPrereqError("Number of vCPUs [%d] does not match the"
+                                   " CPU mask [%s]" %
+                                   (self.be_proposed[constants.BE_VCPUS],
+                                    self.hv_proposed[constants.HV_CPU_MASK]),
+                                   errors.ECODE_INVAL)
+
+      # Only perform this test if a new CPU mask is given
+      if constants.HV_CPU_MASK in self.hv_new:
+        # Calculate the largest CPU number requested
+        max_requested_cpu = max(map(max, cpu_list))
+        # Check that all of the instance's nodes have enough physical CPUs to
+        # satisfy the requested CPU mask
+        _CheckNodesPhysicalCPUs(self, instance.all_nodes,
+                                max_requested_cpu + 1, instance.hypervisor)
+

     # osparams processing
     if self.op.osparams:
       i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
     # osparams processing
     if self.op.osparams:
       i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)

@@ -10821,7 +11308,7 @@ class LUInstanceSetParams(LogicalUnit):
       if msg:
         # Assume the primary node is unreachable and go ahead
         self.warn.append("Can't get info from primary node %s: %s" %
       if msg:
         # Assume the primary node is unreachable and go ahead
         self.warn.append("Can't get info from primary node %s: %s" %

-                         (pnode,  msg))
+                         (pnode, msg))

       elif not isinstance(pninfo.payload.get("memory_free", None), int):
         self.warn.append("Node data from primary node %s doesn't contain"
                          " free memory information" % pnode)
       elif not isinstance(pninfo.payload.get("memory_free", None), int):
         self.warn.append("Node data from primary node %s doesn't contain"
                          " free memory information" % pnode)

@@ -10947,7 +11434,8 @@ class LUInstanceSetParams(LogicalUnit):
         if len(instance.disks) == 1:
           raise errors.OpPrereqError("Cannot remove the last disk of"
                                      " an instance", errors.ECODE_INVAL)
         if len(instance.disks) == 1:
           raise errors.OpPrereqError("Cannot remove the last disk of"
                                      " an instance", errors.ECODE_INVAL)

-        _CheckInstanceDown(self, instance, "cannot remove disks")
+        _CheckInstanceState(self, instance, INSTANCE_DOWN,
+                            msg="cannot remove disks")

 
       if (disk_op == constants.DDM_ADD and
           len(instance.disks) >= constants.MAX_DISKS):
 
       if (disk_op == constants.DDM_ADD and
           len(instance.disks) >= constants.MAX_DISKS):

@@ -10962,7 +11450,15 @@ class LUInstanceSetParams(LogicalUnit):
                                      (disk_op, len(instance.disks)),
                                      errors.ECODE_INVAL)
 
                                      (disk_op, len(instance.disks)),
                                      errors.ECODE_INVAL)
 

-    return
+    # disabling the instance
+    if self.op.offline_inst:
+      _CheckInstanceState(self, instance, INSTANCE_DOWN,
+                          msg="cannot change instance state to offline")
+
+    # enabling the instance
+    if self.op.online_inst:
+      _CheckInstanceState(self, instance, INSTANCE_OFFLINE,
+                          msg="cannot make instance go online")

 
   def _ConvertPlainToDrbd(self, feedback_fn):
     """Converts an instance from plain to drbd.
 
   def _ConvertPlainToDrbd(self, feedback_fn):
     """Converts an instance from plain to drbd.

@@ -10973,6 +11469,8 @@ class LUInstanceSetParams(LogicalUnit):
     pnode = instance.primary_node
     snode = self.op.remote_node
 
     pnode = instance.primary_node
     snode = self.op.remote_node
 

+    assert instance.disk_template == constants.DT_PLAIN
+

     # create a fake disk info for _GenerateDiskTemplate
     disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
                   constants.IDISK_VG: d.logical_id[0]}
     # create a fake disk info for _GenerateDiskTemplate
     disk_info = [{constants.IDISK_SIZE: d.size, constants.IDISK_MODE: d.mode,
                   constants.IDISK_VG: d.logical_id[0]}

@@ -11009,6 +11507,9 @@ class LUInstanceSetParams(LogicalUnit):
     instance.disks = new_disks
     self.cfg.Update(instance, feedback_fn)
 
     instance.disks = new_disks
     self.cfg.Update(instance, feedback_fn)
 

+    # Release node locks while waiting for sync
+    _ReleaseLocks(self, locking.LEVEL_NODE)
+

     # disks are created, waiting for sync
     disk_abort = not _WaitForSync(self, instance,
                                   oneshot=not self.op.wait_for_sync)
     # disks are created, waiting for sync
     disk_abort = not _WaitForSync(self, instance,
                                   oneshot=not self.op.wait_for_sync)

@@ -11016,12 +11517,17 @@ class LUInstanceSetParams(LogicalUnit):
       raise errors.OpExecError("There are some degraded disks for"
                                " this instance, please cleanup manually")
 
       raise errors.OpExecError("There are some degraded disks for"
                                " this instance, please cleanup manually")
 

+    # Node resource locks will be released by caller
+

   def _ConvertDrbdToPlain(self, feedback_fn):
     """Converts an instance from drbd to plain.
 
     """
     instance = self.instance
   def _ConvertDrbdToPlain(self, feedback_fn):
     """Converts an instance from drbd to plain.
 
     """
     instance = self.instance

+

     assert len(instance.secondary_nodes) == 1
     assert len(instance.secondary_nodes) == 1

+    assert instance.disk_template == constants.DT_DRBD8
+

     pnode = instance.primary_node
     snode = instance.secondary_nodes[0]
     feedback_fn("Converting template to plain")
     pnode = instance.primary_node
     snode = instance.secondary_nodes[0]
     feedback_fn("Converting template to plain")

@@ -11039,6 +11545,9 @@ class LUInstanceSetParams(LogicalUnit):
     instance.disk_template = constants.DT_PLAIN
     self.cfg.Update(instance, feedback_fn)
 
     instance.disk_template = constants.DT_PLAIN
     self.cfg.Update(instance, feedback_fn)
 

+    # Release locks in case removing disks takes a while
+    _ReleaseLocks(self, locking.LEVEL_NODE)
+

     feedback_fn("Removing volumes on the secondary node...")
     for disk in old_disks:
       self.cfg.SetDiskID(disk, snode)
     feedback_fn("Removing volumes on the secondary node...")
     for disk in old_disks:
       self.cfg.SetDiskID(disk, snode)

@@ -11056,6 +11565,8 @@ class LUInstanceSetParams(LogicalUnit):
         self.LogWarning("Could not remove metadata for disk %d on node %s,"
                         " continuing anyway: %s", idx, pnode, msg)
 
         self.LogWarning("Could not remove metadata for disk %d on node %s,"
                         " continuing anyway: %s", idx, pnode, msg)
 

+    # Node resource locks will be released by caller
+

   def Exec(self, feedback_fn):
     """Modifies an instance.
 
   def Exec(self, feedback_fn):
     """Modifies an instance.
 

@@ -11067,6 +11578,10 @@ class LUInstanceSetParams(LogicalUnit):
     for warn in self.warn:
       feedback_fn("WARNING: %s" % warn)
 
     for warn in self.warn:
       feedback_fn("WARNING: %s" % warn)
 

+    assert ((self.op.disk_template is None) ^
+            bool(self.owned_locks(locking.LEVEL_NODE_RES))), \
+      "Not owning any node resource locks"
+

     result = []
     instance = self.instance
     # disk changes
     result = []
     instance = self.instance
     # disk changes

@@ -11124,6 +11639,16 @@ class LUInstanceSetParams(LogicalUnit):
                        disk_dict[constants.IDISK_MODE]))
 
     if self.op.disk_template:
                        disk_dict[constants.IDISK_MODE]))
 
     if self.op.disk_template:

+      if __debug__:
+        check_nodes = set(instance.all_nodes)
+        if self.op.remote_node:
+          check_nodes.add(self.op.remote_node)
+        for level in [locking.LEVEL_NODE, locking.LEVEL_NODE_RES]:
+          owned = self.owned_locks(level)
+          assert not (check_nodes - owned), \
+            ("Not owning the correct locks, owning %r, expected at least %r" %
+             (owned, check_nodes))
+

       r_shut = _ShutdownInstanceDisks(self, instance)
       if not r_shut:
         raise errors.OpExecError("Cannot shutdown instance disks, unable to"
       r_shut = _ShutdownInstanceDisks(self, instance)
       if not r_shut:
         raise errors.OpExecError("Cannot shutdown instance disks, unable to"

@@ -11136,6 +11661,15 @@ class LUInstanceSetParams(LogicalUnit):
         raise
       result.append(("disk_template", self.op.disk_template))
 
         raise
       result.append(("disk_template", self.op.disk_template))
 

+      assert instance.disk_template == self.op.disk_template, \
+        ("Expected disk template '%s', found '%s'" %
+         (self.op.disk_template, instance.disk_template))
+
+    # Release node and resource locks if there are any (they might already have
+    # been released during disk conversion)
+    _ReleaseLocks(self, locking.LEVEL_NODE)
+    _ReleaseLocks(self, locking.LEVEL_NODE_RES)
+

     # NIC changes
     for nic_op, nic_dict in self.op.nics:
       if nic_op == constants.DDM_REMOVE:
     # NIC changes
     for nic_op, nic_dict in self.op.nics:
       if nic_op == constants.DDM_REMOVE:

@@ -11186,8 +11720,20 @@ class LUInstanceSetParams(LogicalUnit):
       for key, val in self.op.osparams.iteritems():
         result.append(("os/%s" % key, val))
 
       for key, val in self.op.osparams.iteritems():
         result.append(("os/%s" % key, val))
 

+    # online/offline instance
+    if self.op.online_inst:
+      self.cfg.MarkInstanceDown(instance.name)
+      result.append(("admin_state", constants.ADMINST_DOWN))
+    if self.op.offline_inst:
+      self.cfg.MarkInstanceOffline(instance.name)
+      result.append(("admin_state", constants.ADMINST_OFFLINE))
+

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

+    assert not (self.owned_locks(locking.LEVEL_NODE_RES) or
+                self.owned_locks(locking.LEVEL_NODE)), \
+      "All node locks should have been released by now"
+

     return result
 
   _DISK_CONVERSIONS = {
     return result
 
   _DISK_CONVERSIONS = {

@@ -11510,7 +12056,8 @@ class LUBackupExport(LogicalUnit):
           "Cannot retrieve locked instance %s" % self.op.instance_name
     _CheckNodeOnline(self, self.instance.primary_node)
 
           "Cannot retrieve locked instance %s" % self.op.instance_name
     _CheckNodeOnline(self, self.instance.primary_node)
 

-    if (self.op.remove_instance and self.instance.admin_up and
+    if (self.op.remove_instance and
+        self.instance.admin_state == constants.ADMINST_UP and

         not self.op.shutdown):
       raise errors.OpPrereqError("Can not remove instance without shutting it"
                                  " down before")
         not self.op.shutdown):
       raise errors.OpPrereqError("Can not remove instance without shutting it"
                                  " down before")

@@ -11640,7 +12187,7 @@ class LUBackupExport(LogicalUnit):
     for disk in instance.disks:
       self.cfg.SetDiskID(disk, src_node)
 
     for disk in instance.disks:
       self.cfg.SetDiskID(disk, src_node)
 

-    activate_disks = (not instance.admin_up)
+    activate_disks = (instance.admin_state != constants.ADMINST_UP)

 
     if activate_disks:
       # Activate the instance disks if we'exporting a stopped instance
 
     if activate_disks:
       # Activate the instance disks if we'exporting a stopped instance

@@ -11653,12 +12200,13 @@ class LUBackupExport(LogicalUnit):
 
       helper.CreateSnapshots()
       try:
 
       helper.CreateSnapshots()
       try:

-        if (self.op.shutdown and instance.admin_up and
+        if (self.op.shutdown and
+            instance.admin_state == constants.ADMINST_UP and

             not self.op.remove_instance):
           assert not activate_disks
           feedback_fn("Starting instance %s" % instance.name)
             not self.op.remove_instance):
           assert not activate_disks
           feedback_fn("Starting instance %s" % instance.name)

-          result = self.rpc.call_instance_start(src_node, instance,
-                                                None, None, False)
+          result = self.rpc.call_instance_start(src_node,
+                                                (instance, None, None), False)

           msg = result.fail_msg
           if msg:
             feedback_fn("Failed to start instance: %s" % msg)
           msg = result.fail_msg
           if msg:
             feedback_fn("Failed to start instance: %s" % msg)

@@ -12057,6 +12605,9 @@ class LUGroupQuery(NoHooksLU):
   def ExpandNames(self):
     self.gq.ExpandNames(self)
 
   def ExpandNames(self):
     self.gq.ExpandNames(self)
 

+  def DeclareLocks(self, level):
+    self.gq.DeclareLocks(self, level)
+

   def Exec(self, feedback_fn):
     return self.gq.OldStyleQuery(self)
 
   def Exec(self, feedback_fn):
     return self.gq.OldStyleQuery(self)
 

@@ -12135,7 +12686,6 @@ class LUGroupSetParams(LogicalUnit):
     return result
 
 
     return result
 
 

-

 class LUGroupRemove(LogicalUnit):
   HPATH = "group-remove"
   HTYPE = constants.HTYPE_GROUP
 class LUGroupRemove(LogicalUnit):
   HPATH = "group-remove"
   HTYPE = constants.HTYPE_GROUP

@@ -12428,7 +12978,7 @@ class LUGroupEvacuate(LogicalUnit):
     return ResultWithJobs(jobs)
 
 
     return ResultWithJobs(jobs)
 
 

-class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
+class TagsLU(NoHooksLU): # pylint: disable=W0223

   """Generic tags LU.
 
   This is an abstract class which is the parent of all the other tags LUs.
   """Generic tags LU.
 
   This is an abstract class which is the parent of all the other tags LUs.

@@ -12688,7 +13238,7 @@ class LUTestJqueue(NoHooksLU):
     # Wait for client to close
     try:
       try:
     # Wait for client to close
     try:
       try:

-        # pylint: disable-msg=E1101
+        # pylint: disable=E1101

         # Instance of '_socketobject' has no ... member
         conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
         conn.recv(1)
         # Instance of '_socketobject' has no ... member
         conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
         conn.recv(1)

@@ -12785,12 +13335,12 @@ class IAllocator(object):
       easy usage
 
   """
       easy usage
 
   """

-  # pylint: disable-msg=R0902
+  # pylint: disable=R0902

   # lots of instance attributes
 
   # lots of instance attributes
 

-  def __init__(self, cfg, rpc, mode, **kwargs):
+  def __init__(self, cfg, rpc_runner, mode, **kwargs):

     self.cfg = cfg
     self.cfg = cfg

-    self.rpc = rpc
+    self.rpc = rpc_runner

     # init buffer variables
     self.in_text = self.out_text = self.in_data = self.out_data = None
     # init all input fields so that pylint is happy
     # init buffer variables
     self.in_text = self.out_text = self.in_data = self.out_data = None
     # init all input fields so that pylint is happy

@@ -12954,7 +13504,7 @@ class IAllocator(object):
             i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
             remote_info["memory_free"] -= max(0, i_mem_diff)
 
             i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
             remote_info["memory_free"] -= max(0, i_mem_diff)
 

-            if iinfo.admin_up:
+            if iinfo.admin_state == constants.ADMINST_UP:

               i_p_up_mem += beinfo[constants.BE_MEMORY]
 
         # compute memory used by instances
               i_p_up_mem += beinfo[constants.BE_MEMORY]
 
         # compute memory used by instances

@@ -12994,7 +13544,7 @@ class IAllocator(object):
         nic_data.append(nic_dict)
       pir = {
         "tags": list(iinfo.GetTags()),
         nic_data.append(nic_dict)
       pir = {
         "tags": list(iinfo.GetTags()),

-        "admin_up": iinfo.admin_up,
+        "admin_state": iinfo.admin_state,

         "vcpus": beinfo[constants.BE_VCPUS],
         "memory": beinfo[constants.BE_MEMORY],
         "os": iinfo.os,
         "vcpus": beinfo[constants.BE_VCPUS],
         "memory": beinfo[constants.BE_MEMORY],
         "os": iinfo.os,

@@ -13122,7 +13672,7 @@ class IAllocator(object):
 
   _STRING_LIST = ht.TListOf(ht.TString)
   _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
 
   _STRING_LIST = ht.TListOf(ht.TString)
   _JOB_LIST = ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {

-     # pylint: disable-msg=E1101
+     # pylint: disable=E1101

      # Class '...' has no 'OP_ID' member
      "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
                           opcodes.OpInstanceMigrate.OP_ID,
      # Class '...' has no 'OP_ID' member
      "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
                           opcodes.OpInstanceMigrate.OP_ID,