Convert snapshot_export rpc to new style
[ganeti-local] / lib / cmdlib.py
index 6ebdd8c..4d9502f 100644 (file)
@@ -25,7 +25,6 @@
 
 import os
 import os.path
-import sha
 import time
 import tempfile
 import re
@@ -454,7 +453,7 @@ def _CheckNodeNotDrained(lu, node):
 
 
 def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
-                          memory, vcpus, nics):
+                          memory, vcpus, nics, disk_template, disks):
   """Builds instance related env variables for hooks
 
   This builds the hook environment from individual variables.
@@ -476,6 +475,10 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
   @type nics: list
   @param nics: list of tuples (ip, bridge, mac) representing
       the NICs the instance  has
+  @type disk_template: string
+  @param disk_template: the distk template of the instance
+  @type disks: list
+  @param disks: the list of (size, mode) pairs
   @rtype: dict
   @return: the hook environment for this instance
 
@@ -493,23 +496,58 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
     "INSTANCE_STATUS": str_status,
     "INSTANCE_MEMORY": memory,
     "INSTANCE_VCPUS": vcpus,
+    "INSTANCE_DISK_TEMPLATE": disk_template,
   }
 
   if nics:
     nic_count = len(nics)
-    for idx, (ip, bridge, mac) in enumerate(nics):
+    for idx, (ip, mac, mode, link) in enumerate(nics):
       if ip is None:
         ip = ""
       env["INSTANCE_NIC%d_IP" % idx] = ip
-      env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
-      env["INSTANCE_NIC%d_HWADDR" % idx] = mac
+      env["INSTANCE_NIC%d_MAC" % idx] = mac
+      env["INSTANCE_NIC%d_MODE" % idx] = mode
+      env["INSTANCE_NIC%d_LINK" % idx] = link
+      if mode == constants.NIC_MODE_BRIDGED:
+        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
   else:
     nic_count = 0
 
   env["INSTANCE_NIC_COUNT"] = nic_count
 
+  if disks:
+    disk_count = len(disks)
+    for idx, (size, mode) in enumerate(disks):
+      env["INSTANCE_DISK%d_SIZE" % idx] = size
+      env["INSTANCE_DISK%d_MODE" % idx] = mode
+  else:
+    disk_count = 0
+
+  env["INSTANCE_DISK_COUNT"] = disk_count
+
   return env
 
+def _PreBuildNICHooksList(lu, nics):
+  """Build a list of nic information tuples.
+
+  This list is suitable to be passed to _BuildInstanceHookEnv.
+
+  @type lu:  L{LogicalUnit}
+  @param lu: the logical unit on whose behalf we execute
+  @type nics: list of L{objects.NIC}
+  @param nics: list of nics to convert to hooks tuples
+
+  """
+  hooks_nics = []
+  c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
+  for nic in nics:
+    ip = nic.ip
+    mac = nic.mac
+    filled_params = objects.FillDict(c_nicparams, nic.nicparams)
+    mode = filled_params[constants.NIC_MODE]
+    link = filled_params[constants.NIC_LINK]
+    hooks_nics.append((ip, mac, mode, link))
+  return hooks_nics
 
 def _BuildInstanceHookEnvByObject(lu, instance, override=None):
   """Builds instance related env variables for hooks from an object.
@@ -535,7 +573,9 @@ def _BuildInstanceHookEnvByObject(lu, instance, override=None):
     'status': instance.admin_up,
     'memory': bep[constants.BE_MEMORY],
     'vcpus': bep[constants.BE_VCPUS],
-    'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
+    'nics': _PreBuildNICHooksList(lu, instance.nics),
+    'disk_template': instance.disk_template,
+    'disks': [(disk.size, disk.mode) for disk in instance.disks],
   }
   if override:
     args.update(override)
@@ -558,18 +598,32 @@ def _AdjustCandidatePool(lu):
                (mc_now, mc_max))
 
 
-def _CheckInstanceBridgesExist(lu, instance):
+def _CheckNicsBridgesExist(lu, target_nics, target_node,
+                               profile=constants.PP_DEFAULT):
+  """Check that the brigdes needed by a list of nics exist.
+
+  """
+  c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
+  paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
+                for nic in target_nics]
+  brlist = [params[constants.NIC_LINK] for params in paramslist
+            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
+  if brlist:
+    result = lu.rpc.call_bridges_exist(target_node, brlist)
+    result.Raise()
+    if not result.data:
+      raise errors.OpPrereqError("One or more target bridges %s does not"
+                                 " exist on destination node '%s'" %
+                                 (brlist, target_node))
+
+
+def _CheckInstanceBridgesExist(lu, instance, node=None):
   """Check that the brigdes needed by an instance exist.
 
   """
-  # check bridges existance
-  brlist = [nic.bridge for nic in instance.nics]
-  result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
-  result.Raise()
-  if not result.data:
-    raise errors.OpPrereqError("One or more target bridges %s does not"
-                               " exist on destination node '%s'" %
-                               (brlist, instance.primary_node))
+  if node is None:
+    node=instance.primary_node
+  _CheckNicsBridgesExist(lu, instance.nics, node)
 
 
 class LUDestroyCluster(NoHooksLU):
@@ -630,7 +684,7 @@ class LUVerifyCluster(LogicalUnit):
 
   def _VerifyNode(self, nodeinfo, file_list, local_cksum,
                   node_result, feedback_fn, master_files,
-                  drbd_map):
+                  drbd_map, vg_name):
     """Run multiple tests against a node.
 
     Test list:
@@ -650,6 +704,7 @@ class LUVerifyCluster(LogicalUnit):
     @param drbd_map: the useddrbd minors for this node, in
         form of minor: (instance, must_exist) which correspond to instances
         and their running status
+    @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
 
     """
     node = nodeinfo.name
@@ -683,18 +738,18 @@ class LUVerifyCluster(LogicalUnit):
                   (constants.RELEASE_VERSION, node, remote_version[1]))
 
     # checks vg existence and size > 20G
-
-    vglist = node_result.get(constants.NV_VGLIST, None)
-    if not vglist:
-      feedback_fn("  - ERROR: unable to check volume groups on node %s." %
-                      (node,))
-      bad = True
-    else:
-      vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
-                                            constants.MIN_VG_SIZE)
-      if vgstatus:
-        feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
+    if vg_name is not None:
+      vglist = node_result.get(constants.NV_VGLIST, None)
+      if not vglist:
+        feedback_fn("  - ERROR: unable to check volume groups on node %s." %
+                        (node,))
         bad = True
+      else:
+        vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
+                                              constants.MIN_VG_SIZE)
+        if vgstatus:
+          feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
+          bad = True
 
     # checks config file checksum
 
@@ -756,16 +811,22 @@ class LUVerifyCluster(LogicalUnit):
                       (hv_name, hv_result))
 
     # check used drbd list
-    used_minors = node_result.get(constants.NV_DRBDLIST, [])
-    for minor, (iname, must_exist) in drbd_map.items():
-      if minor not in used_minors and must_exist:
-        feedback_fn("  - ERROR: drbd minor %d of instance %s is not active" %
-                    (minor, iname))
-        bad = True
-    for minor in used_minors:
-      if minor not in drbd_map:
-        feedback_fn("  - ERROR: unallocated drbd minor %d is in use" % minor)
-        bad = True
+    if vg_name is not None:
+      used_minors = node_result.get(constants.NV_DRBDLIST, [])
+      if not isinstance(used_minors, (tuple, list)):
+        feedback_fn("  - ERROR: cannot parse drbd status file: %s" %
+                    str(used_minors))
+      else:
+        for minor, (iname, must_exist) in drbd_map.items():
+          if minor not in used_minors and must_exist:
+            feedback_fn("  - ERROR: drbd minor %d of instance %s is"
+                        " not active" % (minor, iname))
+            bad = True
+        for minor in used_minors:
+          if minor not in drbd_map:
+            feedback_fn("  - ERROR: unallocated drbd minor %d is in use" %
+                        minor)
+            bad = True
 
     return bad
 
@@ -892,8 +953,12 @@ class LUVerifyCluster(LogicalUnit):
 
     """
     all_nodes = self.cfg.GetNodeList()
-    # TODO: populate the environment with useful information for verify hooks
-    env = {}
+    env = {
+      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
+      }
+    for node in self.cfg.GetAllNodesInfo().values():
+      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
+
     return env, [], all_nodes
 
   def Exec(self, feedback_fn):
@@ -941,13 +1006,14 @@ class LUVerifyCluster(LogicalUnit):
       constants.NV_NODENETTEST: [(node.name, node.primary_ip,
                                   node.secondary_ip) for node in nodeinfo
                                  if not node.offline],
-      constants.NV_LVLIST: vg_name,
       constants.NV_INSTANCELIST: hypervisors,
-      constants.NV_VGLIST: None,
       constants.NV_VERSION: None,
       constants.NV_HVINFO: self.cfg.GetHypervisorType(),
-      constants.NV_DRBDLIST: None,
       }
+    if vg_name is not None:
+      node_verify_param[constants.NV_VGLIST] = None
+      node_verify_param[constants.NV_LVLIST] = vg_name
+      node_verify_param[constants.NV_DRBDLIST] = None
     all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
                                            self.cfg.GetClusterName())
 
@@ -982,15 +1048,25 @@ class LUVerifyCluster(LogicalUnit):
 
       node_drbd = {}
       for minor, instance in all_drbd_map[node].items():
-        instance = instanceinfo[instance]
-        node_drbd[minor] = (instance.name, instance.admin_up)
+        if instance not in instanceinfo:
+          feedback_fn("  - ERROR: ghost instance '%s' in temporary DRBD map" %
+                      instance)
+          # ghost instance should not be running, but otherwise we
+          # don't give double warnings (both ghost instance and
+          # unallocated minor in use)
+          node_drbd[minor] = (instance, False)
+        else:
+          instance = instanceinfo[instance]
+          node_drbd[minor] = (instance.name, instance.admin_up)
       result = self._VerifyNode(node_i, file_names, local_checksums,
                                 nresult, feedback_fn, master_files,
-                                node_drbd)
+                                node_drbd, vg_name)
       bad = bad or result
 
       lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
-      if isinstance(lvdata, basestring):
+      if vg_name is None:
+        node_volume[node] = {}
+      elif isinstance(lvdata, basestring):
         feedback_fn("  - ERROR: LVM problem on node %s: %s" %
                     (node, utils.SafeEncode(lvdata)))
         bad = True
@@ -1022,7 +1098,6 @@ class LUVerifyCluster(LogicalUnit):
       try:
         node_info[node] = {
           "mfree": int(nodeinfo['memory_free']),
-          "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
           "pinst": [],
           "sinst": [],
           # dictionary holding all instances this node is secondary for,
@@ -1033,8 +1108,19 @@ class LUVerifyCluster(LogicalUnit):
           # secondary.
           "sinst-by-pnode": {},
         }
-      except ValueError:
-        feedback_fn("  - ERROR: invalid value returned from node %s" % (node,))
+        # FIXME: devise a free space model for file based instances as well
+        if vg_name is not None:
+          if (constants.NV_VGLIST not in nresult or
+              vg_name not in nresult[constants.NV_VGLIST]):
+            feedback_fn("  - ERROR: node %s didn't return data for the"
+                        " volume group '%s' - it is either missing or broken" %
+                        (node, vg_name))
+            bad = True
+            continue
+          node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
+      except (ValueError, KeyError):
+        feedback_fn("  - ERROR: invalid nodeinfo value returned"
+                    " from node %s" % (node,))
         bad = True
         continue
 
@@ -1241,6 +1327,7 @@ class LUVerifyDisks(NoHooksLU):
       if isinstance(lvs, basestring):
         logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
         res_nlvm[node] = lvs
+        continue
       elif not isinstance(lvs, dict):
         logging.warning("Connection to node %s failed or invalid data"
                         " returned", node)
@@ -1332,9 +1419,11 @@ class LURenameCluster(LogicalUnit):
       result = self.rpc.call_upload_file(node_list,
                                          constants.SSH_KNOWN_HOSTS_FILE)
       for to_node, to_result in result.iteritems():
-        if to_result.failed or not to_result.data:
-          logging.error("Copy of file %s to node %s failed",
-                        constants.SSH_KNOWN_HOSTS_FILE, to_node)
+         msg = to_result.RemoteFailMsg()
+         if msg:
+           msg = ("Copy of file %s to node %s failed: %s" %
+                   (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
+           self.proc.LogWarning(msg)
 
     finally:
       result = self.rpc.call_node_start_master(master, False)
@@ -1368,7 +1457,7 @@ class LUSetClusterParams(LogicalUnit):
   _OP_REQP = []
   REQ_BGL = False
 
-  def CheckParameters(self):
+  def CheckArguments(self):
     """Check parameters
 
     """
@@ -1377,7 +1466,7 @@ class LUSetClusterParams(LogicalUnit):
     if self.op.candidate_pool_size is not None:
       try:
         self.op.candidate_pool_size = int(self.op.candidate_pool_size)
-      except ValueError, err:
+      except (ValueError, TypeError), err:
         raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
                                    str(err))
       if self.op.candidate_pool_size < 1:
@@ -1435,14 +1524,20 @@ class LUSetClusterParams(LogicalUnit):
                                      (node, vgstatus))
 
     self.cluster = cluster = self.cfg.GetClusterInfo()
-    # validate beparams changes
+    # validate params changes
     if self.op.beparams:
       utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
-      self.new_beparams = cluster.FillDict(
-        cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
+      self.new_beparams = objects.FillDict(
+        cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
+
+    if self.op.nicparams:
+      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
+      self.new_nicparams = objects.FillDict(
+        cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
+      objects.NIC.CheckParameterSyntax(self.new_nicparams)
 
     # hypervisor list/parameters
-    self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
+    self.new_hvparams = objects.FillDict(cluster.hvparams, {})
     if self.op.hvparams:
       if not isinstance(self.op.hvparams, dict):
         raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
@@ -1474,8 +1569,11 @@ class LUSetClusterParams(LogicalUnit):
 
     """
     if self.op.vg_name is not None:
-      if self.op.vg_name != self.cfg.GetVGName():
-        self.cfg.SetVGName(self.op.vg_name)
+      new_volume = self.op.vg_name
+      if not new_volume:
+        new_volume = None
+      if new_volume != self.cfg.GetVGName():
+        self.cfg.SetVGName(new_volume)
       else:
         feedback_fn("Cluster LVM configuration already in desired"
                     " state, not changing")
@@ -1484,7 +1582,10 @@ class LUSetClusterParams(LogicalUnit):
     if self.op.enabled_hypervisors is not None:
       self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
     if self.op.beparams:
-      self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
+      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
+    if self.op.nicparams:
+      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
+
     if self.op.candidate_pool_size is not None:
       self.cluster.candidate_pool_size = self.op.candidate_pool_size
 
@@ -1496,6 +1597,48 @@ class LUSetClusterParams(LogicalUnit):
       _AdjustCandidatePool(self)
 
 
+def _RedistributeAncillaryFiles(lu, additional_nodes=None):
+  """Distribute additional files which are part of the cluster configuration.
+
+  ConfigWriter takes care of distributing the config and ssconf files, but
+  there are more files which should be distributed to all nodes. This function
+  makes sure those are copied.
+
+  @param lu: calling logical unit
+  @param additional_nodes: list of nodes not in the config to distribute to
+
+  """
+  # 1. Gather target nodes
+  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
+  dist_nodes = lu.cfg.GetNodeList()
+  if additional_nodes is not None:
+    dist_nodes.extend(additional_nodes)
+  if myself.name in dist_nodes:
+    dist_nodes.remove(myself.name)
+  # 2. Gather files to distribute
+  dist_files = set([constants.ETC_HOSTS,
+                    constants.SSH_KNOWN_HOSTS_FILE,
+                    constants.RAPI_CERT_FILE,
+                    constants.RAPI_USERS_FILE,
+                   ])
+
+  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
+  for hv_name in enabled_hypervisors:
+    hv_class = hypervisor.GetHypervisor(hv_name)
+    dist_files.update(hv_class.GetAncillaryFiles())
+
+  # 3. Perform the files upload
+  for fname in dist_files:
+    if os.path.exists(fname):
+      result = lu.rpc.call_upload_file(dist_nodes, fname)
+      for to_node, to_result in result.items():
+         msg = to_result.RemoteFailMsg()
+         if msg:
+           msg = ("Copy of file %s to node %s failed: %s" %
+                   (fname, to_node, msg))
+           lu.proc.LogWarning(msg)
+
+
 class LURedistributeConfig(NoHooksLU):
   """Force the redistribution of cluster configuration.
 
@@ -1521,6 +1664,7 @@ class LURedistributeConfig(NoHooksLU):
 
     """
     self.cfg.Update(self.cfg.GetClusterInfo())
+    _RedistributeAncillaryFiles(self)
 
 
 def _WaitForSync(lu, instance, oneshot=False, unlock=False):
@@ -1544,15 +1688,16 @@ def _WaitForSync(lu, instance, oneshot=False, unlock=False):
     done = True
     cumul_degraded = False
     rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
-    if rstats.failed or not rstats.data:
-      lu.LogWarning("Can't get any data from node %s", node)
+    msg = rstats.RemoteFailMsg()
+    if msg:
+      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
       retries += 1
       if retries >= 10:
         raise errors.RemoteError("Can't contact node %s for mirror data,"
                                  " aborting." % node)
       time.sleep(6)
       continue
-    rstats = rstats.data
+    rstats = rstats.payload
     retries = 0
     for i, mstat in enumerate(rstats):
       if mstat is None:
@@ -1632,9 +1777,11 @@ class LUDiagnoseOS(NoHooksLU):
                        selected=self.op.output_fields)
 
     # Lock all nodes, in shared mode
+    # Temporary removal of locks, should be reverted later
+    # TODO: reintroduce locks when they are lighter-weight
     self.needed_locks = {}
-    self.share_locks[locking.LEVEL_NODE] = 1
-    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
+    #self.share_locks[locking.LEVEL_NODE] = 1
+    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
 
   def CheckPrereq(self):
     """Check prerequisites.
@@ -1649,7 +1796,7 @@ class LUDiagnoseOS(NoHooksLU):
     @param rlist: a map with node names as keys and OS objects as values
 
     @rtype: dict
-    @returns: a dictionary with osnames as keys and as value another map, with
+    @return: a dictionary with osnames as keys and as value another map, with
         nodes as keys and list of OS objects as values, eg::
 
           {"debian-etch": {"node1": [<object>,...],
@@ -1658,6 +1805,11 @@ class LUDiagnoseOS(NoHooksLU):
 
     """
     all_os = {}
+    # we build here the list of nodes that didn't fail the RPC (at RPC
+    # level), so that nodes with a non-responding node daemon don't
+    # make all OSes invalid
+    good_nodes = [node_name for node_name in rlist
+                  if not rlist[node_name].failed]
     for node_name, nr in rlist.iteritems():
       if nr.failed or not nr.data:
         continue
@@ -1666,7 +1818,7 @@ class LUDiagnoseOS(NoHooksLU):
           # build a list of nodes for this os containing empty lists
           # for each node in node_list
           all_os[os_obj.name] = {}
-          for nname in node_list:
+          for nname in good_nodes:
             all_os[os_obj.name][nname] = []
         all_os[os_obj.name][node_name].append(os_obj)
     return all_os
@@ -1675,9 +1827,7 @@ class LUDiagnoseOS(NoHooksLU):
     """Compute the list of OSes.
 
     """
-    node_list = self.acquired_locks[locking.LEVEL_NODE]
-    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
-                   if node in node_list]
+    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
     node_data = self.rpc.call_os_diagnose(valid_nodes)
     if node_data == False:
       raise errors.OpExecError("Can't gather the list of OSes")
@@ -2163,7 +2313,8 @@ class LUAddNode(LogicalUnit):
                                " new node: %s" % msg)
 
     # Add node to our /etc/hosts, and add key to known_hosts
-    utils.AddHostToEtcHosts(new_node.name)
+    if self.cfg.GetClusterInfo().modify_etc_hosts:
+      utils.AddHostToEtcHosts(new_node.name)
 
     if new_node.secondary_ip != new_node.primary_ip:
       result = self.rpc.call_node_has_ip_address(new_node.name,
@@ -2191,35 +2342,11 @@ class LUAddNode(LogicalUnit):
                       (verifier, result[verifier].data['nodelist'][failed]))
         raise errors.OpExecError("ssh/hostname verification failed.")
 
-    # Distribute updated /etc/hosts and known_hosts to all nodes,
-    # including the node just added
-    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
-    dist_nodes = self.cfg.GetNodeList()
-    if not self.op.readd:
-      dist_nodes.append(node)
-    if myself.name in dist_nodes:
-      dist_nodes.remove(myself.name)
-
-    logging.debug("Copying hosts and known_hosts to all nodes")
-    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
-      result = self.rpc.call_upload_file(dist_nodes, fname)
-      for to_node, to_result in result.iteritems():
-        if to_result.failed or not to_result.data:
-          logging.error("Copy of file %s to node %s failed", fname, to_node)
-
-    to_copy = []
-    enabled_hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
-    if constants.HTS_COPY_VNC_PASSWORD.intersection(enabled_hypervisors):
-      to_copy.append(constants.VNC_PASSWORD_FILE)
-
-    for fname in to_copy:
-      result = self.rpc.call_upload_file([node], fname)
-      if result[node].failed or not result[node]:
-        logging.error("Could not copy file %s to node %s", fname, node)
-
     if self.op.readd:
+      _RedistributeAncillaryFiles(self)
       self.context.ReaddNode(new_node)
     else:
+      _RedistributeAncillaryFiles(self, additional_nodes=[node])
       self.context.AddNode(new_node)
 
 
@@ -2294,7 +2421,7 @@ class LUSetNodeParams(LogicalUnit):
         ((node.offline and not self.op.offline == False) or
          (node.drained and not self.op.drained == False))):
       raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
-                                 " to master_candidate")
+                                 " to master_candidate" % node.name)
 
     return
 
@@ -2350,6 +2477,51 @@ class LUSetNodeParams(LogicalUnit):
     return result
 
 
+class LUPowercycleNode(NoHooksLU):
+  """Powercycles a node.
+
+  """
+  _OP_REQP = ["node_name", "force"]
+  REQ_BGL = False
+
+  def CheckArguments(self):
+    node_name = self.cfg.ExpandNodeName(self.op.node_name)
+    if node_name is None:
+      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
+    self.op.node_name = node_name
+    if node_name == self.cfg.GetMasterNode() and not self.op.force:
+      raise errors.OpPrereqError("The node is the master and the force"
+                                 " parameter was not set")
+
+  def ExpandNames(self):
+    """Locking for PowercycleNode.
+
+    This is a last-resource option and shouldn't block on other
+    jobs. Therefore, we grab no locks.
+
+    """
+    self.needed_locks = {}
+
+  def CheckPrereq(self):
+    """Check prerequisites.
+
+    This LU has no prereqs.
+
+    """
+    pass
+
+  def Exec(self, feedback_fn):
+    """Reboots a node.
+
+    """
+    result = self.rpc.call_node_powercycle(self.op.node_name,
+                                           self.cfg.GetHypervisorType())
+    msg = result.RemoteFailMsg()
+    if msg:
+      raise errors.OpExecError("Failed to schedule the reboot: %s" % msg)
+    return result.payload
+
+
 class LUQueryClusterInfo(NoHooksLU):
   """Query cluster configuration.
 
@@ -2385,7 +2557,11 @@ class LUQueryClusterInfo(NoHooksLU):
       "hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
                         for hypervisor in cluster.enabled_hypervisors]),
       "beparams": cluster.beparams,
+      "nicparams": cluster.nicparams,
       "candidate_pool_size": cluster.candidate_pool_size,
+      "master_netdev": cluster.master_netdev,
+      "volume_group_name": cluster.volume_group_name,
+      "file_storage_dir": cluster.file_storage_dir,
       }
 
     return result
@@ -2700,15 +2876,48 @@ class LUStartupInstance(LogicalUnit):
     assert self.instance is not None, \
       "Cannot retrieve locked instance %s" % self.op.instance_name
 
+    # extra beparams
+    self.beparams = getattr(self.op, "beparams", {})
+    if self.beparams:
+      if not isinstance(self.beparams, dict):
+        raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
+                                   " dict" % (type(self.beparams), ))
+      # fill the beparams dict
+      utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
+      self.op.beparams = self.beparams
+
+    # extra hvparams
+    self.hvparams = getattr(self.op, "hvparams", {})
+    if self.hvparams:
+      if not isinstance(self.hvparams, dict):
+        raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
+                                   " dict" % (type(self.hvparams), ))
+
+      # check hypervisor parameter syntax (locally)
+      cluster = self.cfg.GetClusterInfo()
+      utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
+      filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
+                                    instance.hvparams)
+      filled_hvp.update(self.hvparams)
+      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
+      hv_type.CheckParameterSyntax(filled_hvp)
+      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
+      self.op.hvparams = self.hvparams
+
     _CheckNodeOnline(self, instance.primary_node)
 
     bep = self.cfg.GetClusterInfo().FillBE(instance)
     # check bridges existance
     _CheckInstanceBridgesExist(self, instance)
 
-    _CheckNodeFreeMemory(self, instance.primary_node,
-                         "starting instance %s" % instance.name,
-                         bep[constants.BE_MEMORY], instance.hypervisor)
+    remote_info = self.rpc.call_instance_info(instance.primary_node,
+                                              instance.name,
+                                              instance.hypervisor)
+    remote_info.Raise()
+    if not remote_info.data:
+      _CheckNodeFreeMemory(self, instance.primary_node,
+                           "starting instance %s" % instance.name,
+                           bep[constants.BE_MEMORY], instance.hypervisor)
 
   def Exec(self, feedback_fn):
     """Start the instance.
@@ -2716,7 +2925,6 @@ class LUStartupInstance(LogicalUnit):
     """
     instance = self.instance
     force = self.op.force
-    extra_args = getattr(self.op, "extra_args", "")
 
     self.cfg.MarkInstanceUp(instance.name)
 
@@ -2724,7 +2932,8 @@ class LUStartupInstance(LogicalUnit):
 
     _StartInstanceDisks(self, instance, force)
 
-    result = self.rpc.call_instance_start(node_current, instance, extra_args)
+    result = self.rpc.call_instance_start(node_current, instance,
+                                          self.hvparams, self.beparams)
     msg = result.RemoteFailMsg()
     if msg:
       _ShutdownInstanceDisks(self, instance)
@@ -2758,6 +2967,7 @@ class LURebootInstance(LogicalUnit):
     """
     env = {
       "IGNORE_SECONDARIES": self.op.ignore_secondaries,
+      "REBOOT_TYPE": self.op.reboot_type,
       }
     env.update(_BuildInstanceHookEnvByObject(self, self.instance))
     nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
@@ -2785,14 +2995,15 @@ class LURebootInstance(LogicalUnit):
     instance = self.instance
     ignore_secondaries = self.op.ignore_secondaries
     reboot_type = self.op.reboot_type
-    extra_args = getattr(self.op, "extra_args", "")
 
     node_current = instance.primary_node
 
     if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
                        constants.INSTANCE_REBOOT_HARD]:
+      for disk in instance.disks:
+        self.cfg.SetDiskID(disk, node_current)
       result = self.rpc.call_instance_reboot(node_current, instance,
-                                             reboot_type, extra_args)
+                                             reboot_type)
       msg = result.RemoteFailMsg()
       if msg:
         raise errors.OpExecError("Could not reboot instance: %s" % msg)
@@ -2804,7 +3015,7 @@ class LURebootInstance(LogicalUnit):
                                  " full reboot: %s" % msg)
       _ShutdownInstanceDisks(self, instance)
       _StartInstanceDisks(self, instance, ignore_secondaries)
-      result = self.rpc.call_instance_start(node_current, instance, extra_args)
+      result = self.rpc.call_instance_start(node_current, instance, None, None)
       msg = result.RemoteFailMsg()
       if msg:
         _ShutdownInstanceDisks(self, instance)
@@ -2904,7 +3115,8 @@ class LUReinstallInstance(LogicalUnit):
     remote_info = self.rpc.call_instance_info(instance.primary_node,
                                               instance.name,
                                               instance.hypervisor)
-    if remote_info.failed or remote_info.data:
+    remote_info.Raise()
+    if remote_info.data:
       raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
                                  (self.op.instance_name,
                                   instance.primary_node))
@@ -2939,7 +3151,7 @@ class LUReinstallInstance(LogicalUnit):
     _StartInstanceDisks(self, inst, None)
     try:
       feedback_fn("Running the instance OS create scripts...")
-      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
+      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
       msg = result.RemoteFailMsg()
       if msg:
         raise errors.OpExecError("Could not install OS for instance %s"
@@ -3137,7 +3349,7 @@ class LUQueryInstances(NoHooksLU):
   _OP_REQP = ["output_fields", "names", "use_locking"]
   REQ_BGL = False
   _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
-                                    "admin_state", "admin_ram",
+                                    "admin_state",
                                     "disk_template", "ip", "mac", "bridge",
                                     "sda_size", "sdb_size", "vcpus", "tags",
                                     "network_port", "beparams",
@@ -3426,15 +3638,8 @@ class LUFailoverInstance(LogicalUnit):
     _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
                          instance.name, bep[constants.BE_MEMORY],
                          instance.hypervisor)
-
     # check bridge existance
-    brlist = [nic.bridge for nic in instance.nics]
-    result = self.rpc.call_bridges_exist(target_node, brlist)
-    result.Raise()
-    if not result.data:
-      raise errors.OpPrereqError("One or more target bridges %s does not"
-                                 " exist on destination node '%s'" %
-                                 (brlist, target_node))
+    _CheckInstanceBridgesExist(self, instance, node=target_node)
 
   def Exec(self, feedback_fn):
     """Failover an instance.
@@ -3494,7 +3699,7 @@ class LUFailoverInstance(LogicalUnit):
         raise errors.OpExecError("Can't activate the instance's disks")
 
       feedback_fn("* starting the instance on the target node")
-      result = self.rpc.call_instance_start(target_node, instance, None)
+      result = self.rpc.call_instance_start(target_node, instance, None, None)
       msg = result.RemoteFailMsg()
       if msg:
         _ShutdownInstanceDisks(self, instance)
@@ -3531,6 +3736,8 @@ class LUMigrateInstance(LogicalUnit):
 
     """
     env = _BuildInstanceHookEnvByObject(self, self.instance)
+    env["MIGRATE_LIVE"] = self.op.live
+    env["MIGRATE_CLEANUP"] = self.op.cleanup
     nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
     return env, nl, nl
 
@@ -3564,12 +3771,7 @@ class LUMigrateInstance(LogicalUnit):
                          instance.hypervisor)
 
     # check bridge existance
-    brlist = [nic.bridge for nic in instance.nics]
-    result = self.rpc.call_bridges_exist(target_node, brlist)
-    if result.failed or not result.data:
-      raise errors.OpPrereqError("One or more target bridges %s does not"
-                                 " exist on destination node '%s'" %
-                                 (brlist, target_node))
+    _CheckInstanceBridgesExist(self, instance, node=target_node)
 
     if not self.op.cleanup:
       _CheckNodeNotDrained(self, target_node)
@@ -4176,8 +4378,8 @@ def _CheckHVParams(lu, nodenames, hvname, hvparams):
       continue
     msg = info.RemoteFailMsg()
     if msg:
-      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
-                                 " %s" % msg)
+      raise errors.OpPrereqError("Hypervisor parameter validation"
+                                 " failed on node %s: %s" % (node, msg))
 
 
 class LUCreateInstance(LogicalUnit):
@@ -4238,14 +4440,14 @@ class LUCreateInstance(LogicalUnit):
 
     # check hypervisor parameter syntax (locally)
     utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
-    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
+    filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
                                   self.op.hvparams)
     hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
     hv_type.CheckParameterSyntax(filled_hvp)
 
     # fill and remember the beparams dict
     utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
-    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
+    self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
                                     self.op.beparams)
 
     #### instance parameters check
@@ -4264,10 +4466,21 @@ class LUCreateInstance(LogicalUnit):
 
     # NIC buildup
     self.nics = []
-    for nic in self.op.nics:
+    for idx, nic in enumerate(self.op.nics):
+      nic_mode_req = nic.get("mode", None)
+      nic_mode = nic_mode_req
+      if nic_mode is None:
+        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
+
+      # in routed mode, for the first nic, the default ip is 'auto'
+      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
+        default_ip_mode = constants.VALUE_AUTO
+      else:
+        default_ip_mode = constants.VALUE_NONE
+
       # ip validity checks
-      ip = nic.get("ip", None)
-      if ip is None or ip.lower() == "none":
+      ip = nic.get("ip", default_ip_mode)
+      if ip is None or ip.lower() == constants.VALUE_NONE:
         nic_ip = None
       elif ip.lower() == constants.VALUE_AUTO:
         nic_ip = hostname1.ip
@@ -4277,6 +4490,10 @@ class LUCreateInstance(LogicalUnit):
                                      " like a valid IP" % ip)
         nic_ip = ip
 
+      # TODO: check the ip for uniqueness !!
+      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
+        raise errors.OpPrereqError("Routed nic mode requires an ip address")
+
       # MAC address verification
       mac = nic.get("mac", constants.VALUE_AUTO)
       if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
@@ -4285,9 +4502,24 @@ class LUCreateInstance(LogicalUnit):
                                      mac)
       # bridge verification
       bridge = nic.get("bridge", None)
-      if bridge is None:
-        bridge = self.cfg.GetDefBridge()
-      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
+      link = nic.get("link", None)
+      if bridge and link:
+        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
+      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
+        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic")
+      elif bridge:
+        link = bridge
+
+      nicparams = {}
+      if nic_mode_req:
+        nicparams[constants.NIC_MODE] = nic_mode_req
+      if link:
+        nicparams[constants.NIC_LINK] = link
+
+      check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
+                                      nicparams)
+      objects.NIC.CheckParameterSyntax(check_params)
+      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
 
     # disk checks/pre-build
     self.disks = []
@@ -4401,23 +4633,24 @@ class LUCreateInstance(LogicalUnit):
 
     """
     env = {
-      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
-      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
-      "INSTANCE_ADD_MODE": self.op.mode,
+      "ADD_MODE": self.op.mode,
       }
     if self.op.mode == constants.INSTANCE_IMPORT:
-      env["INSTANCE_SRC_NODE"] = self.op.src_node
-      env["INSTANCE_SRC_PATH"] = self.op.src_path
-      env["INSTANCE_SRC_IMAGES"] = self.src_images
+      env["SRC_NODE"] = self.op.src_node
+      env["SRC_PATH"] = self.op.src_path
+      env["SRC_IMAGES"] = self.src_images
 
-    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
+    env.update(_BuildInstanceHookEnv(
+      name=self.op.instance_name,
       primary_node=self.op.pnode,
       secondary_nodes=self.secondaries,
       status=self.op.start,
       os_type=self.op.os_type,
       memory=self.be_full[constants.BE_MEMORY],
       vcpus=self.be_full[constants.BE_VCPUS],
-      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
+      nics=_PreBuildNICHooksList(self, self.nics),
+      disk_template=self.op.disk_template,
+      disks=[(d["size"], d["mode"]) for d in self.disks],
     ))
 
     nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
@@ -4434,7 +4667,6 @@ class LUCreateInstance(LogicalUnit):
       raise errors.OpPrereqError("Cluster does not support lvm-based"
                                  " instances")
 
-
     if self.op.mode == constants.INSTANCE_IMPORT:
       src_node = self.op.src_node
       src_path = self.op.src_path
@@ -4500,6 +4732,7 @@ class LUCreateInstance(LogicalUnit):
             nic_mac_ini = 'nic%d_mac' % idx
             nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
 
+    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
     # ip ping checks (we use the same ip that was resolved in ExpandNames)
     if self.op.start and not self.op.ip_check:
       raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
@@ -4510,6 +4743,18 @@ class LUCreateInstance(LogicalUnit):
         raise errors.OpPrereqError("IP %s of instance %s already in use" %
                                    (self.check_ip, self.op.instance_name))
 
+    #### mac address generation
+    # By generating here the mac address both the allocator and the hooks get
+    # the real final mac address rather than the 'auto' or 'generate' value.
+    # There is a race condition between the generation and the instance object
+    # creation, which means that we know the mac is valid now, but we're not
+    # sure it will be when we actually add the instance. If things go bad
+    # adding the instance will abort because of a duplicate mac, and the
+    # creation job will fail.
+    for nic in self.nics:
+      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
+        nic.mac = self.cfg.GenerateMAC()
+
     #### allocator run
 
     if self.op.iallocator is not None:
@@ -4576,14 +4821,7 @@ class LUCreateInstance(LogicalUnit):
       raise errors.OpPrereqError("OS '%s' not in supported os list for"
                                  " primary node"  % self.op.os_type)
 
-    # bridge check on primary node
-    bridges = [n.bridge for n in self.nics]
-    result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
-    result.Raise()
-    if not result.data:
-      raise errors.OpPrereqError("One of the target bridges '%s' does not"
-                                 " exist on destination node '%s'" %
-                                 (",".join(bridges), pnode.name))
+    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
 
     # memory check on primary node
     if self.op.start:
@@ -4599,10 +4837,6 @@ class LUCreateInstance(LogicalUnit):
     instance = self.op.instance_name
     pnode_name = self.pnode.name
 
-    for nic in self.nics:
-      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
-        nic.mac = self.cfg.GenerateMAC()
-
     ht_kind = self.op.hypervisor
     if ht_kind in constants.HTS_REQ_PORT:
       network_port = self.cfg.AllocatePort()
@@ -4696,7 +4930,7 @@ class LUCreateInstance(LogicalUnit):
     if iobj.disk_template != constants.DT_DISKLESS:
       if self.op.mode == constants.INSTANCE_CREATE:
         feedback_fn("* running the instance OS create scripts...")
-        result = self.rpc.call_instance_os_add(pnode_name, iobj)
+        result = self.rpc.call_instance_os_add(pnode_name, iobj, False)
         msg = result.RemoteFailMsg()
         if msg:
           raise errors.OpExecError("Could not add os for instance %s"
@@ -4727,7 +4961,7 @@ class LUCreateInstance(LogicalUnit):
       self.cfg.Update(iobj)
       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)
+      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
       msg = result.RemoteFailMsg()
       if msg:
         raise errors.OpExecError("Could not start instance: %s" % msg)
@@ -5049,10 +5283,11 @@ class LUReplaceDisks(LogicalUnit):
     for dev, old_lvs, new_lvs in iv_names.itervalues():
       info("detaching %s drbd from local storage" % dev.iv_name)
       result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
-      result.Raise()
-      if not result.data:
+      msg = result.RemoteFailMsg()
+      if msg:
         raise errors.OpExecError("Can't detach drbd from local storage on node"
-                                 " %s for device %s" % (tgt_node, dev.iv_name))
+                                 " %s for device %s: %s" %
+                                 (tgt_node, dev.iv_name, msg))
       #dev.children = []
       #cfg.Update(instance)
 
@@ -5076,16 +5311,18 @@ class LUReplaceDisks(LogicalUnit):
 
       info("renaming the old LVs on the target node")
       result = self.rpc.call_blockdev_rename(tgt_node, rlist)
-      result.Raise()
-      if not result.data:
-        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
+      msg = result.RemoteFailMsg()
+      if msg:
+        raise errors.OpExecError("Can't rename old LVs on node %s: %s" %
+                                 (tgt_node, msg))
       # now we rename the new LVs to the old LVs
       info("renaming the new LVs on the target node")
       rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
       result = self.rpc.call_blockdev_rename(tgt_node, rlist)
-      result.Raise()
-      if not result.data:
-        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
+      msg = result.RemoteFailMsg()
+      if msg:
+        raise errors.OpExecError("Can't rename new LVs on node %s: %s" %
+                                 (tgt_node, msg))
 
       for old, new in zip(old_lvs, new_lvs):
         new.logical_id = old.logical_id
@@ -5098,13 +5335,14 @@ class LUReplaceDisks(LogicalUnit):
       # now that the new lvs have the old name, we can add them to the device
       info("adding new mirror component on %s" % tgt_node)
       result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
-      if result.failed or not result.data:
+      msg = result.RemoteFailMsg()
+      if msg:
         for new_lv in new_lvs:
           msg = self.rpc.call_blockdev_remove(tgt_node, new_lv).RemoteFailMsg()
           if msg:
             warning("Can't rollback device %s: %s", dev, msg,
                     hint="cleanup manually the unused logical volumes")
-        raise errors.OpExecError("Can't add local storage to drbd")
+        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
 
       dev.children = new_lvs
       cfg.Update(instance)
@@ -5252,7 +5490,7 @@ class LUReplaceDisks(LogicalUnit):
       try:
         _CreateSingleBlockDev(self, new_node, instance, new_drbd,
                               _GetInstanceInfoText(instance), False)
-      except errors.BlockDeviceError:
+      except errors.GenericError:
         self.cfg.ReleaseDRBDMinors(instance.name)
         raise
 
@@ -5501,11 +5739,14 @@ class LUQueryInstanceData(NoHooksLU):
     if not static:
       self.cfg.SetDiskID(dev, instance.primary_node)
       dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
-      msg = dev_pstatus.RemoteFailMsg()
-      if msg:
-        raise errors.OpExecError("Can't compute disk status for %s: %s" %
-                                 (instance.name, msg))
-      dev_pstatus = dev_pstatus.payload
+      if dev_pstatus.offline:
+        dev_pstatus = None
+      else:
+        msg = dev_pstatus.RemoteFailMsg()
+        if msg:
+          raise errors.OpExecError("Can't compute disk status for %s: %s" %
+                                   (instance.name, msg))
+        dev_pstatus = dev_pstatus.payload
     else:
       dev_pstatus = None
 
@@ -5519,11 +5760,14 @@ class LUQueryInstanceData(NoHooksLU):
     if snode and not static:
       self.cfg.SetDiskID(dev, snode)
       dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
-      msg = dev_sstatus.RemoteFailMsg()
-      if msg:
-        raise errors.OpExecError("Can't compute disk status for %s: %s" %
-                                 (instance.name, msg))
-      dev_sstatus = dev_sstatus.payload
+      if dev_sstatus.offline:
+        dev_sstatus = None
+      else:
+        msg = dev_sstatus.RemoteFailMsg()
+        if msg:
+          raise errors.OpExecError("Can't compute disk status for %s: %s" %
+                                   (instance.name, msg))
+        dev_sstatus = dev_sstatus.payload
     else:
       dev_sstatus = None
 
@@ -5667,24 +5911,35 @@ class LUSetInstanceParams(LogicalUnit):
       # nic_dict should be a dict
       nic_ip = nic_dict.get('ip', None)
       if nic_ip is not None:
-        if nic_ip.lower() == "none":
+        if nic_ip.lower() == constants.VALUE_NONE:
           nic_dict['ip'] = None
         else:
           if not utils.IsValidIP(nic_ip):
             raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
-      # we can only check None bridges and assign the default one
+
       nic_bridge = nic_dict.get('bridge', None)
-      if nic_bridge is None:
-        nic_dict['bridge'] = self.cfg.GetDefBridge()
-      # but we can validate MACs
-      nic_mac = nic_dict.get('mac', None)
-      if nic_mac is not None:
-        if self.cfg.IsMacInUse(nic_mac):
-          raise errors.OpPrereqError("MAC address %s already in use"
-                                     " in cluster" % nic_mac)
+      nic_link = nic_dict.get('link', None)
+      if nic_bridge and nic_link:
+        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
+      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
+        nic_dict['bridge'] = None
+      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
+        nic_dict['link'] = None
+
+      if nic_op == constants.DDM_ADD:
+        nic_mac = nic_dict.get('mac', None)
+        if nic_mac is None:
+          nic_dict['mac'] = constants.VALUE_AUTO
+
+      if 'mac' in nic_dict:
+        nic_mac = nic_dict['mac']
         if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
           if not utils.IsValidMac(nic_mac):
             raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
+        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
+          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
+                                     " modifying an existing nic")
+
     if nic_addremove > 1:
       raise errors.OpPrereqError("Only one NIC add or remove operation"
                                  " supported at a time")
@@ -5709,11 +5964,78 @@ class LUSetInstanceParams(LogicalUnit):
       args['memory'] = self.be_new[constants.BE_MEMORY]
     if constants.BE_VCPUS in self.be_new:
       args['vcpus'] = self.be_new[constants.BE_VCPUS]
-    # FIXME: readd disk/nic changes
+    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
+    # information at all.
+    if self.op.nics:
+      args['nics'] = []
+      nic_override = dict(self.op.nics)
+      c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
+      for idx, nic in enumerate(self.instance.nics):
+        if idx in nic_override:
+          this_nic_override = nic_override[idx]
+        else:
+          this_nic_override = {}
+        if 'ip' in this_nic_override:
+          ip = this_nic_override['ip']
+        else:
+          ip = nic.ip
+        if 'mac' in this_nic_override:
+          mac = this_nic_override['mac']
+        else:
+          mac = nic.mac
+        if idx in self.nic_pnew:
+          nicparams = self.nic_pnew[idx]
+        else:
+          nicparams = objects.FillDict(c_nicparams, nic.nicparams)
+        mode = nicparams[constants.NIC_MODE]
+        link = nicparams[constants.NIC_LINK]
+        args['nics'].append((ip, mac, mode, link))
+      if constants.DDM_ADD in nic_override:
+        ip = nic_override[constants.DDM_ADD].get('ip', None)
+        mac = nic_override[constants.DDM_ADD]['mac']
+        nicparams = self.nic_pnew[constants.DDM_ADD]
+        mode = nicparams[constants.NIC_MODE]
+        link = nicparams[constants.NIC_LINK]
+        args['nics'].append((ip, mac, mode, link))
+      elif constants.DDM_REMOVE in nic_override:
+        del args['nics'][-1]
+
     env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
     nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
     return env, nl, nl
 
+  def _GetUpdatedParams(self, old_params, update_dict,
+                        default_values, parameter_types):
+    """Return the new params dict for the given params.
+
+    @type old_params: dict
+    @type old_params: old parameters
+    @type update_dict: dict
+    @type update_dict: dict containing new parameter values,
+                       or constants.VALUE_DEFAULT to reset the
+                       parameter to its default value
+    @type default_values: dict
+    @param default_values: default values for the filled parameters
+    @type parameter_types: dict
+    @param parameter_types: dict mapping target dict keys to types
+                            in constants.ENFORCEABLE_TYPES
+    @rtype: (dict, dict)
+    @return: (new_parameters, filled_parameters)
+
+    """
+    params_copy = copy.deepcopy(old_params)
+    for key, val in update_dict.iteritems():
+      if val == constants.VALUE_DEFAULT:
+        try:
+          del params_copy[key]
+        except KeyError:
+          pass
+      else:
+        params_copy[key] = val
+    utils.ForceDictType(params_copy, parameter_types)
+    params_filled = objects.FillDict(default_values, params_copy)
+    return (params_copy, params_filled)
+
   def CheckPrereq(self):
     """Check prerequisites.
 
@@ -5725,6 +6047,7 @@ class LUSetInstanceParams(LogicalUnit):
     # checking the new params on the primary/secondary nodes
 
     instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
+    cluster = self.cluster = self.cfg.GetClusterInfo()
     assert self.instance is not None, \
       "Cannot retrieve locked instance %s" % self.op.instance_name
     pnode = instance.primary_node
@@ -5732,19 +6055,10 @@ class LUSetInstanceParams(LogicalUnit):
 
     # hvparams processing
     if self.op.hvparams:
-      i_hvdict = copy.deepcopy(instance.hvparams)
-      for key, val in self.op.hvparams.iteritems():
-        if val == constants.VALUE_DEFAULT:
-          try:
-            del i_hvdict[key]
-          except KeyError:
-            pass
-        else:
-          i_hvdict[key] = val
-      cluster = self.cfg.GetClusterInfo()
-      utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
-      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
-                                i_hvdict)
+      i_hvdict, hv_new = self._GetUpdatedParams(
+                             instance.hvparams, self.op.hvparams,
+                             cluster.hvparams[instance.hypervisor],
+                             constants.HVS_PARAMETER_TYPES)
       # local check
       hypervisor.GetHypervisor(
         instance.hypervisor).CheckParameterSyntax(hv_new)
@@ -5756,19 +6070,10 @@ class LUSetInstanceParams(LogicalUnit):
 
     # beparams processing
     if self.op.beparams:
-      i_bedict = copy.deepcopy(instance.beparams)
-      for key, val in self.op.beparams.iteritems():
-        if val == constants.VALUE_DEFAULT:
-          try:
-            del i_bedict[key]
-          except KeyError:
-            pass
-        else:
-          i_bedict[key] = val
-      cluster = self.cfg.GetClusterInfo()
-      utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
-      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
-                                i_bedict)
+      i_bedict, be_new = self._GetUpdatedParams(
+                             instance.beparams, self.op.beparams,
+                             cluster.beparams[constants.PP_DEFAULT],
+                             constants.BES_PARAMETER_TYPES)
       self.be_new = be_new # the new actual values
       self.be_inst = i_bedict # the new dict (without defaults)
     else:
@@ -5790,7 +6095,7 @@ class LUSetInstanceParams(LogicalUnit):
         self.warn.append("Can't get info from primary node %s" % pnode)
       else:
         if not instance_info.failed and instance_info.data:
-          current_mem = instance_info.data['memory']
+          current_mem = int(instance_info.data['memory'])
         else:
           # Assume instance not running
           # (there is a slight race condition here, but it's not very probable,
@@ -5814,6 +6119,8 @@ class LUSetInstanceParams(LogicalUnit):
                              " secondary node %s" % node)
 
     # NIC processing
+    self.nic_pnew = {}
+    self.nic_pinst = {}
     for nic_op, nic_dict in self.op.nics:
       if nic_op == constants.DDM_REMOVE:
         if not instance.nics:
@@ -5825,15 +6132,59 @@ class LUSetInstanceParams(LogicalUnit):
           raise errors.OpPrereqError("Invalid NIC index %s, valid values"
                                      " are 0 to %d" %
                                      (nic_op, len(instance.nics)))
-      nic_bridge = nic_dict.get('bridge', None)
-      if nic_bridge is not None:
-        if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
+        old_nic_params = instance.nics[nic_op].nicparams
+        old_nic_ip = instance.nics[nic_op].ip
+      else:
+        old_nic_params = {}
+        old_nic_ip = None
+
+      update_params_dict = dict([(key, nic_dict[key])
+                                 for key in constants.NICS_PARAMETERS
+                                 if key in nic_dict])
+
+      if 'bridge' in nic_dict:
+        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
+
+      new_nic_params, new_filled_nic_params = \
+          self._GetUpdatedParams(old_nic_params, update_params_dict,
+                                 cluster.nicparams[constants.PP_DEFAULT],
+                                 constants.NICS_PARAMETER_TYPES)
+      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
+      self.nic_pinst[nic_op] = new_nic_params
+      self.nic_pnew[nic_op] = new_filled_nic_params
+      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
+
+      if new_nic_mode == constants.NIC_MODE_BRIDGED:
+        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
+        result = self.rpc.call_bridges_exist(pnode, [nic_bridge])
+        result.Raise()
+        if not result.data:
           msg = ("Bridge '%s' doesn't exist on one of"
                  " the instance nodes" % nic_bridge)
           if self.force:
             self.warn.append(msg)
           else:
             raise errors.OpPrereqError(msg)
+      if new_nic_mode == constants.NIC_MODE_ROUTED:
+        if 'ip' in nic_dict:
+          nic_ip = nic_dict['ip']
+        else:
+          nic_ip = old_nic_ip
+        if nic_ip is None:
+          raise errors.OpPrereqError('Cannot set the nic ip to None'
+                                     ' on a routed nic')
+      if 'mac' in nic_dict:
+        nic_mac = nic_dict['mac']
+        if nic_mac is None:
+          raise errors.OpPrereqError('Cannot set the nic mac to None')
+        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
+          # otherwise generate the mac
+          nic_dict['mac'] = self.cfg.GenerateMAC()
+        else:
+          # or validate/reserve the current one
+          if self.cfg.IsMacInUse(nic_mac):
+            raise errors.OpPrereqError("MAC address %s already in use"
+                                       " in cluster" % nic_mac)
 
     # DISK processing
     if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
@@ -5878,6 +6229,7 @@ class LUSetInstanceParams(LogicalUnit):
 
     result = []
     instance = self.instance
+    cluster = self.cluster
     # disk changes
     for disk_op, disk_dict in self.op.disks:
       if disk_op == constants.DDM_REMOVE:
@@ -5936,25 +6288,26 @@ class LUSetInstanceParams(LogicalUnit):
         del instance.nics[-1]
         result.append(("nic.%d" % len(instance.nics), "remove"))
       elif nic_op == constants.DDM_ADD:
-        # add a new nic
-        if 'mac' not in nic_dict:
-          mac = constants.VALUE_GENERATE
-        else:
-          mac = nic_dict['mac']
-        if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
-          mac = self.cfg.GenerateMAC()
-        new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
-                              bridge=nic_dict.get('bridge', None))
+        # mac and bridge should be set, by now
+        mac = nic_dict['mac']
+        ip = nic_dict.get('ip', None)
+        nicparams = self.nic_pinst[constants.DDM_ADD]
+        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
         instance.nics.append(new_nic)
         result.append(("nic.%d" % (len(instance.nics) - 1),
-                       "add:mac=%s,ip=%s,bridge=%s" %
-                       (new_nic.mac, new_nic.ip, new_nic.bridge)))
+                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
+                       (new_nic.mac, new_nic.ip,
+                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
+                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
+                       )))
       else:
-        # change a given nic
-        for key in 'mac', 'ip', 'bridge':
+        for key in 'mac', 'ip':
           if key in nic_dict:
             setattr(instance.nics[nic_op], key, nic_dict[key])
-            result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
+        if nic_op in self.nic_pnew:
+          instance.nics[nic_op].nicparams = self.nic_pnew[nic_op]
+        for key, val in nic_dict.iteritems():
+          result.append(("nic.%s/%d" % (key, nic_op), val))
 
     # hvparams changes
     if self.op.hvparams:
@@ -6109,22 +6462,23 @@ class LUExportInstance(LogicalUnit):
 
     try:
       for disk in instance.disks:
-        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
-        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
-        if new_dev_name.failed or not new_dev_name.data:
-          self.LogWarning("Could not snapshot block device %s on node %s",
-                          disk.logical_id[1], src_node)
+        # result.payload will be a snapshot of an lvm leaf of the one we passed
+        result = self.rpc.call_blockdev_snapshot(src_node, disk)
+        msg = result.RemoteFailMsg()
+        if msg:
+          self.LogWarning("Could not snapshot block device %s on node %s: %s",
+                          disk.logical_id[1], src_node, msg)
           snap_disks.append(False)
         else:
+          disk_id = (vgname, result.payload)
           new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
-                                 logical_id=(vgname, new_dev_name.data),
-                                 physical_id=(vgname, new_dev_name.data),
+                                 logical_id=disk_id, physical_id=disk_id,
                                  iv_name=disk.iv_name)
           snap_disks.append(new_dev)
 
     finally:
       if self.op.shutdown and instance.admin_up:
-        result = self.rpc.call_instance_start(src_node, instance, None)
+        result = self.rpc.call_instance_start(src_node, instance, None, None)
         msg = result.RemoteFailMsg()
         if msg:
           _ShutdownInstanceDisks(self, instance)
@@ -6137,10 +6491,11 @@ class LUExportInstance(LogicalUnit):
       if dev:
         result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
                                                instance, cluster_name, idx)
-        if result.failed or not result.data:
+        msg = result.RemoteFailMsg()
+        if msg:
           self.LogWarning("Could not export block device %s from node %s to"
-                          " node %s", dev.logical_id[1], src_node,
-                          dst_node.name)
+                          " node %s: %s", dev.logical_id[1], src_node,
+                          dst_node.name, msg)
         msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
         if msg:
           self.LogWarning("Could not remove snapshot block device %s from node"
@@ -6496,7 +6851,7 @@ class IAllocator(object):
     cluster_info = cfg.GetClusterInfo()
     # cluster data
     data = {
-      "version": 1,
+      "version": constants.IALLOCATOR_VERSION,
       "cluster_name": cfg.GetClusterName(),
       "cluster_tags": list(cluster_info.GetTags()),
       "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
@@ -6579,8 +6934,19 @@ class IAllocator(object):
     # instance data
     instance_data = {}
     for iinfo, beinfo in i_list:
-      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
-                  for n in iinfo.nics]
+      nic_data = []
+      for nic in iinfo.nics:
+        filled_params = objects.FillDict(
+            cluster_info.nicparams[constants.PP_DEFAULT],
+            nic.nicparams)
+        nic_dict = {"mac": nic.mac,
+                    "ip": nic.ip,
+                    "mode": filled_params[constants.NIC_MODE],
+                    "link": filled_params[constants.NIC_LINK],
+                   }
+        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
+          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
+        nic_data.append(nic_dict)
       pir = {
         "tags": list(iinfo.GetTags()),
         "admin_up": iinfo.admin_up,
@@ -6593,6 +6959,8 @@ class IAllocator(object):
         "disk_template": iinfo.disk_template,
         "hypervisor": iinfo.hypervisor,
         }
+      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
+                                                 pir["disks"])
       instance_data[iinfo.name] = pir
 
     data["instances"] = instance_data