Merge branch 'devel-2.5'
[ganeti-local] / lib / cmdlib.py
index 80795d2..3ebc753 100644 (file)
 
 """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
 
-# 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
@@ -60,7 +60,7 @@ from ganeti import qlang
 from ganeti import opcodes
 from ganeti import ht
 
-import ganeti.masterd.instance # pylint: disable-msg=W0611
+import ganeti.masterd.instance # pylint: disable=W0611
 
 
 class ResultWithJobs:
@@ -131,10 +131,10 @@ class LogicalUnit(object):
     # 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
@@ -322,7 +322,7 @@ class LogicalUnit(object):
     """
     # 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):
@@ -390,7 +390,7 @@ class LogicalUnit(object):
     del self.recalculate_locks[locking.LEVEL_NODE]
 
 
-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
@@ -757,7 +757,7 @@ def _RunPostHook(lu, node_name):
   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)
 
 
@@ -1087,7 +1087,7 @@ def _BuildInstanceHookEnvByObject(lu, instance, override=None):
   }
   if override:
     args.update(override)
-  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
+  return _BuildInstanceHookEnv(**args) # pylint: disable=W0142
 
 
 def _AdjustCandidatePool(lu, exceptions):
@@ -1355,7 +1355,7 @@ class LUClusterDestroy(LogicalUnit):
     # Run post hooks on master node before it's removed
     _RunPostHook(self, master)
 
-    result = self.rpc.call_node_stop_master(master, False)
+    result = self.rpc.call_node_deactivate_master_ip(master)
     result.Raise("Could not disable the master role")
 
     return master
@@ -1371,7 +1371,7 @@ def _VerifyCertificate(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))
 
@@ -1486,7 +1486,7 @@ class _VerifyErrors(object):
     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:
@@ -1495,14 +1495,14 @@ class _VerifyErrors(object):
         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):
     """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 cond:
       self._Error(*args, **kwargs)
     # do not mark the operation as failed for WARN cases only
@@ -1539,7 +1539,7 @@ class LUClusterVerify(NoHooksLU):
                 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
@@ -1801,7 +1801,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     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)
@@ -1870,7 +1870,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103
 
     ntime = nresult.get(constants.NV_TIME, None)
     try:
@@ -1903,7 +1903,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       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)
@@ -1940,7 +1940,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       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)
@@ -1959,7 +1959,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     node = ninfo.name
-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103
 
     test = constants.NV_NODELIST not in nresult
     _ErrorIf(test, self.ENODESSH, node,
@@ -2000,7 +2000,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     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 = {}
@@ -2200,7 +2200,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     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)
@@ -2259,7 +2259,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     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
@@ -2300,7 +2300,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     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?"
 
@@ -2370,7 +2370,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     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")
@@ -2418,7 +2418,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
 
     """
     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)
@@ -2460,7 +2460,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         list of tuples (success, payload)
 
     """
-    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
+    _ErrorIf = self._ErrorIf # pylint: disable=C0103
 
     node_disks = {}
     node_disks_devonly = {}
@@ -2568,7 +2568,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
     """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:
@@ -2577,7 +2577,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
       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
 
@@ -2973,10 +2973,8 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
         self._ErrorIf(test, self.ENODEHOOKS, node_name,
                       "Communication failure in hooks execution: %s", msg)
         if res.offline or msg:
-          # No need to investigate payload if node is offline or gave an error.
-          # override manually lu_result here as _ErrorIf only
-          # overrides self.bad
-          lu_result = 1
+          # 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
@@ -2985,7 +2983,7 @@ class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
           if test:
             output = self._HOOKS_INDENT_RE.sub("      ", output)
             feedback_fn("%s" % output)
-            lu_result = 0
+            lu_result = False
 
     return lu_result
 
@@ -3297,7 +3295,7 @@ class LUClusterRename(LogicalUnit):
 
     # shutdown the master IP
     master = self.cfg.GetMasterNode()
-    result = self.rpc.call_node_stop_master(master, False)
+    result = self.rpc.call_node_deactivate_master_ip(master)
     result.Raise("Could not disable the master role")
 
     try:
@@ -3315,7 +3313,7 @@ class LUClusterRename(LogicalUnit):
         pass
       _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
     finally:
-      result = self.rpc.call_node_start_master(master, False, False)
+      result = self.rpc.call_node_activate_master_ip(master)
       msg = result.fail_msg
       if msg:
         self.LogWarning("Could not re-enable the master role on"
@@ -3648,7 +3646,7 @@ class LUClusterSetParams(LogicalUnit):
       master = self.cfg.GetMasterNode()
       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)
       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))
@@ -3659,7 +3657,7 @@ class LUClusterSetParams(LogicalUnit):
     if 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)
+      result = self.rpc.call_node_activate_master_ip(master)
       if result.fail_msg:
         self.LogWarning("Could not re-enable the master ip on"
                         " the master, please restart manually: %s",
@@ -3697,6 +3695,9 @@ def _ComputeAncillaryFiles(cluster, redist):
   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)
@@ -3797,6 +3798,30 @@ class LUClusterRedistConf(NoHooksLU):
     _RedistributeAncillaryFiles(self)
 
 
+class LUClusterActivateMasterIp(NoHooksLU):
+  """Activate the master IP on the master node.
+
+  """
+  def Exec(self, feedback_fn):
+    """Activate the master IP.
+
+    """
+    master = self.cfg.GetMasterNode()
+    self.rpc.call_node_activate_master_ip(master)
+
+
+class LUClusterDeactivateMasterIp(NoHooksLU):
+  """Deactivate the master IP on the master node.
+
+  """
+  def Exec(self, feedback_fn):
+    """Deactivate the master IP.
+
+    """
+    master = self.cfg.GetMasterNode()
+    self.rpc.call_node_deactivate_master_ip(master)
+
+
 def _WaitForSync(lu, instance, disks=None, oneshot=False):
   """Sleep and poll for an instance's disk to sync.
 
@@ -4419,7 +4444,7 @@ class LUNodeQuery(NoHooksLU):
   """Logical unit for querying nodes.
 
   """
-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142
   REQ_BGL = False
 
   def CheckArguments(self):
@@ -4620,7 +4645,7 @@ 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:
-        lu._LockInstancesNodes() # pylint: disable-msg=W0212
+        lu._LockInstancesNodes() # pylint: disable=W0212
 
   @staticmethod
   def _CheckGroupLocks(lu):
@@ -4721,7 +4746,7 @@ class LUQuery(NoHooksLU):
   """Query for resources/items of a certain kind.
 
   """
-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142
   REQ_BGL = False
 
   def CheckArguments(self):
@@ -4743,7 +4768,7 @@ class LUQueryFields(NoHooksLU):
   """Query for resources/items of a certain kind.
 
   """
-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142
   REQ_BGL = False
 
   def CheckArguments(self):
@@ -4990,7 +5015,7 @@ 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:
-      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
@@ -5838,6 +5863,40 @@ def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
                                  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.
 
@@ -6550,7 +6609,7 @@ class LUInstanceQuery(NoHooksLU):
   """Logical unit for querying instances.
 
   """
-  # pylint: disable-msg=W0142
+  # pylint: disable=W0142
   REQ_BGL = False
 
   def CheckArguments(self):
@@ -6991,6 +7050,11 @@ class TLMigrateInstance(Tasklet):
   @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,
@@ -7314,12 +7378,13 @@ class TLMigrateInstance(Tasklet):
     """
     instance = self.instance
     target_node = self.target_node
+    source_node = self.source_node
     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",
@@ -7327,6 +7392,13 @@ 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.
 
+    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.
 
@@ -7343,6 +7415,21 @@ class TLMigrateInstance(Tasklet):
     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):
@@ -7398,18 +7485,59 @@ class TLMigrateInstance(Tasklet):
       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
+
     # 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:
-      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)
 
@@ -8403,33 +8531,39 @@ class LUInstanceCreate(LogicalUnit):
       if einfo.has_option(constants.INISECT_INS, "disk_template"):
         self.op.disk_template = einfo.get(constants.INISECT_INS,
                                           "disk_template")
+        if 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:
-      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})
-        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)
 
-    if (not self.op.nics and
-        einfo.has_option(constants.INISECT_INS, "nic_count")):
+    if not self.op.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"):
@@ -8515,7 +8649,7 @@ class LUInstanceCreate(LogicalUnit):
 
       joinargs.append(self.op.instance_name)
 
-      # pylint: disable-msg=W0142
+      # pylint: disable=W0142
       self.instance_file_storage_dir = utils.PathJoin(*joinargs)
 
   def CheckPrereq(self):
@@ -8533,7 +8667,8 @@ class LUInstanceCreate(LogicalUnit):
       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()
@@ -8559,6 +8694,10 @@ class LUInstanceCreate(LogicalUnit):
     _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)
 
@@ -8575,7 +8714,7 @@ 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
-      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'
@@ -8619,9 +8758,11 @@ class LUInstanceCreate(LogicalUnit):
 
       #  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[constants.NIC_MODE] = nic_mode_req
+        nicparams[constants.NIC_MODE] = nic_mode
       if link:
         nicparams[constants.NIC_LINK] = link
 
@@ -8658,18 +8799,8 @@ class LUInstanceCreate(LogicalUnit):
       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 = []
-      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?
@@ -8682,15 +8813,9 @@ class LUInstanceCreate(LogicalUnit):
       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 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)
 
@@ -8967,8 +9092,9 @@ class LUInstanceCreate(LogicalUnit):
 
           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, False,
+                                          self.op.debug_level)
           if pause_sync:
             feedback_fn("* resuming disk sync")
             result = self.rpc.call_blockdev_pause_resume_sync(pnode_name,
@@ -8978,8 +9104,8 @@ class LUInstanceCreate(LogicalUnit):
                 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...")
@@ -9674,7 +9800,7 @@ class TLReplaceDisks(Tasklet):
           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:
@@ -9831,6 +9957,8 @@ class TLReplaceDisks(Tasklet):
     """
     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])
@@ -9905,10 +10033,8 @@ class TLReplaceDisks(Tasklet):
                                  " 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:
@@ -10785,9 +10911,11 @@ class LUInstanceSetParams(LogicalUnit):
       # 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_proposed = cluster.SimpleFillHV(instance.hypervisor, instance.os,
+                                              instance.hvparams)
       self.hv_new = self.hv_inst = {}
 
     # beparams processing
@@ -10796,12 +10924,40 @@ class LUInstanceSetParams(LogicalUnit):
                                    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_proposed = cluster.SimpleFillBE(instance.beparams)
     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)
@@ -12436,7 +12592,7 @@ class LUGroupEvacuate(LogicalUnit):
     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.
@@ -12696,7 +12852,7 @@ class LUTestJqueue(NoHooksLU):
     # 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)
@@ -12793,7 +12949,7 @@ class IAllocator(object):
       easy usage
 
   """
-  # pylint: disable-msg=R0902
+  # pylint: disable=R0902
   # lots of instance attributes
 
   def __init__(self, cfg, rpc, mode, **kwargs):
@@ -13130,7 +13286,7 @@ class IAllocator(object):
 
   _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,