Some changes on disk failure tests
[ganeti-local] / lib / cmdlib.py
index b31733d..65c52a2 100644 (file)
@@ -267,11 +267,12 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
 
   if nics:
     nic_count = len(nics)
-    for idx, (ip, bridge) in enumerate(nics):
+    for idx, (ip, bridge, mac) 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
   else:
     nic_count = 0
 
@@ -295,7 +296,7 @@ def _BuildInstanceHookEnvByObject(instance, override=None):
     'status': instance.os,
     'memory': instance.memory,
     'vcpus': instance.vcpus,
-    'nics': [(nic.ip, nic.bridge) for nic in instance.nics],
+    'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
   }
   if override:
     args.update(override)
@@ -499,6 +500,12 @@ class LUInitCluster(LogicalUnit):
     if config.ConfigWriter.IsCluster():
       raise errors.OpPrereqError("Cluster is already initialised")
 
+    if self.op.hypervisor_type == constants.HT_XEN_HVM31:
+      if not os.path.exists(constants.VNC_PASSWORD_FILE):
+        raise errors.OpPrereqError("Please prepare the cluster VNC"
+                                   "password file %s" %
+                                   constants.VNC_PASSWORD_FILE)
+
     self.hostname = hostname = utils.HostInfo()
 
     if hostname.ip.startswith("127."):
@@ -538,7 +545,7 @@ class LUInitCluster(LogicalUnit):
       raise errors.OpPrereqError("Invalid mac prefix given '%s'" %
                                  self.op.mac_prefix)
 
-    if self.op.hypervisor_type not in hypervisor.VALID_HTYPES:
+    if self.op.hypervisor_type not in constants.HYPER_TYPES:
       raise errors.OpPrereqError("Invalid hypervisor type given '%s'" %
                                  self.op.hypervisor_type)
 
@@ -834,12 +841,17 @@ class LUVerifyCluster(NoHooksLU):
       # node_volume
       volumeinfo = all_volumeinfo[node]
 
-      if type(volumeinfo) != dict:
+      if isinstance(volumeinfo, basestring):
+        feedback_fn("  - ERROR: LVM problem on node %s: %s" %
+                    (node, volumeinfo[-400:].encode('string_escape')))
+        bad = True
+        node_volume[node] = {}
+      elif not isinstance(volumeinfo, dict):
         feedback_fn("  - ERROR: connection to %s failed" % (node,))
         bad = True
         continue
-
-      node_volume[node] = volumeinfo
+      else:
+        node_volume[node] = volumeinfo
 
       # node_instance
       nodeinstance = all_instanceinfo[node]
@@ -875,6 +887,78 @@ class LUVerifyCluster(NoHooksLU):
     return int(bad)
 
 
+class LUVerifyDisks(NoHooksLU):
+  """Verifies the cluster disks status.
+
+  """
+  _OP_REQP = []
+
+  def CheckPrereq(self):
+    """Check prerequisites.
+
+    This has no prerequisites.
+
+    """
+    pass
+
+  def Exec(self, feedback_fn):
+    """Verify integrity of cluster disks.
+
+    """
+    result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
+
+    vg_name = self.cfg.GetVGName()
+    nodes = utils.NiceSort(self.cfg.GetNodeList())
+    instances = [self.cfg.GetInstanceInfo(name)
+                 for name in self.cfg.GetInstanceList()]
+
+    nv_dict = {}
+    for inst in instances:
+      inst_lvs = {}
+      if (inst.status != "up" or
+          inst.disk_template not in constants.DTS_NET_MIRROR):
+        continue
+      inst.MapLVsByNode(inst_lvs)
+      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
+      for node, vol_list in inst_lvs.iteritems():
+        for vol in vol_list:
+          nv_dict[(node, vol)] = inst
+
+    if not nv_dict:
+      return result
+
+    node_lvs = rpc.call_volume_list(nodes, vg_name)
+
+    to_act = set()
+    for node in nodes:
+      # node_volume
+      lvs = node_lvs[node]
+
+      if isinstance(lvs, basestring):
+        logger.Info("error enumerating LVs on node %s: %s" % (node, lvs))
+        res_nlvm[node] = lvs
+      elif not isinstance(lvs, dict):
+        logger.Info("connection to node %s failed or invalid data returned" %
+                    (node,))
+        res_nodes.append(node)
+        continue
+
+      for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
+        inst = nv_dict.pop((node, lv_name), None)
+        if (not lv_online and inst is not None
+            and inst.name not in res_instances):
+          res_instances.append(inst.name)
+
+    # any leftover items in nv_dict are missing LVs, let's arrange the
+    # data better
+    for key, inst in nv_dict.iteritems():
+      if inst.name not in res_missing:
+        res_missing[inst.name] = []
+      res_missing[inst.name].append(key)
+
+    return result
+
+
 class LURenameCluster(LogicalUnit):
   """Rename the cluster.
 
@@ -1408,6 +1492,11 @@ class LUAddNode(LogicalUnit):
                                  primary_ip=primary_ip,
                                  secondary_ip=secondary_ip)
 
+    if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31:
+      if not os.path.exists(constants.VNC_PASSWORD_FILE):
+        raise errors.OpPrereqError("Cluster VNC password file %s missing" %
+                                   constants.VNC_PASSWORD_FILE)
+
   def Exec(self, feedback_fn):
     """Adds the new node to the cluster.
 
@@ -1527,6 +1616,8 @@ class LUAddNode(LogicalUnit):
                        (fname, to_node))
 
     to_copy = ss.GetFileList()
+    if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31:
+      to_copy.append(constants.VNC_PASSWORD_FILE)
     for fname in to_copy:
       if not ssh.CopyFileToNode(node, fname):
         logger.Error("could not copy file %s to node %s" % (fname, node))
@@ -1769,23 +1860,41 @@ def _AssembleInstanceDisks(instance, cfg, ignore_secondaries=False):
   """
   device_info = []
   disks_ok = True
+  iname = instance.name
+  # With the two passes mechanism we try to reduce the window of
+  # opportunity for the race condition of switching DRBD to primary
+  # before handshaking occured, but we do not eliminate it
+
+  # The proper fix would be to wait (with some limits) until the
+  # connection has been made and drbd transitions from WFConnection
+  # into any other network-connected state (Connected, SyncTarget,
+  # SyncSource, etc.)
+
+  # 1st pass, assemble on all nodes in secondary mode
   for inst_disk in instance.disks:
-    master_result = None
     for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
       cfg.SetDiskID(node_disk, node)
-      is_primary = node == instance.primary_node
-      result = rpc.call_blockdev_assemble(node, node_disk,
-                                          instance.name, is_primary)
+      result = rpc.call_blockdev_assemble(node, node_disk, iname, False)
       if not result:
         logger.Error("could not prepare block device %s on node %s"
-                     " (is_primary=%s)" %
-                     (inst_disk.iv_name, node, is_primary))
-        if is_primary or not ignore_secondaries:
+                     " (is_primary=False, pass=1)" % (inst_disk.iv_name, node))
+        if not ignore_secondaries:
           disks_ok = False
-      if is_primary:
-        master_result = result
-    device_info.append((instance.primary_node, inst_disk.iv_name,
-                        master_result))
+
+  # FIXME: race condition on drbd migration to primary
+
+  # 2nd pass, do only the primary node
+  for inst_disk in instance.disks:
+    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
+      if node != instance.primary_node:
+        continue
+      cfg.SetDiskID(node_disk, node)
+      result = rpc.call_blockdev_assemble(node, node_disk, iname, True)
+      if not result:
+        logger.Error("could not prepare block device %s on node %s"
+                     " (is_primary=True, pass=2)" % (inst_disk.iv_name, node))
+        disks_ok = False
+    device_info.append((instance.primary_node, inst_disk.iv_name, result))
 
   # leave the disks configured for the primary node
   # this is a workaround that would be fixed better by
@@ -1868,6 +1977,36 @@ def _ShutdownInstanceDisks(instance, cfg, ignore_primary=False):
   return result
 
 
+def _CheckNodeFreeMemory(cfg, node, reason, requested):
+  """Checks if a node has enough free memory.
+
+  This function check if a given node has the needed amount of free
+  memory. In case the node has less memory or we cannot get the
+  information from the node, this function raise an OpPrereqError
+  exception.
+
+  Args:
+    - cfg: a ConfigWriter instance
+    - node: the node name
+    - reason: string to use in the error message
+    - requested: the amount of memory in MiB
+
+  """
+  nodeinfo = rpc.call_node_info([node], cfg.GetVGName())
+  if not nodeinfo or not isinstance(nodeinfo, dict):
+    raise errors.OpPrereqError("Could not contact node %s for resource"
+                             " information" % (node,))
+
+  free_mem = nodeinfo[node].get('memory_free')
+  if not isinstance(free_mem, int):
+    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
+                             " was '%s'" % (node, free_mem))
+  if requested > free_mem:
+    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
+                             " needed %s MiB, available %s MiB" %
+                             (node, reason, requested, free_mem))
+
+
 class LUStartupInstance(LogicalUnit):
   """Starts an instance.
 
@@ -1905,6 +2044,10 @@ class LUStartupInstance(LogicalUnit):
     # check bridges existance
     _CheckInstanceBridgesExist(instance)
 
+    _CheckNodeFreeMemory(self.cfg, instance.primary_node,
+                         "starting instance %s" % instance.name,
+                         instance.memory)
+
     self.instance = instance
     self.op.instance_name = instance.name
 
@@ -1918,20 +2061,6 @@ class LUStartupInstance(LogicalUnit):
 
     node_current = instance.primary_node
 
-    nodeinfo = rpc.call_node_info([node_current], self.cfg.GetVGName())
-    if not nodeinfo:
-      raise errors.OpExecError("Could not contact node %s for infos" %
-                               (node_current))
-
-    freememory = nodeinfo[node_current]['memory_free']
-    memory = instance.memory
-    if memory > freememory:
-      raise errors.OpExecError("Not enough memory to start instance"
-                               " %s on node %s"
-                               " needed %s MiB, available %s MiB" %
-                               (instance.name, node_current, memory,
-                                freememory))
-
     _StartInstanceDisks(self.cfg, instance, force)
 
     if not rpc.call_instance_start(node_current, instance, extra_args):
@@ -2289,11 +2418,11 @@ class LUQueryInstances(NoHooksLU):
     This checks that the fields required are valid output fields.
 
     """
-    self.dynamic_fields = frozenset(["oper_state", "oper_ram"])
+    self.dynamic_fields = frozenset(["oper_state", "oper_ram", "status"])
     _CheckOutputFields(static=["name", "os", "pnode", "snodes",
                                "admin_state", "admin_ram",
                                "disk_template", "ip", "mac", "bridge",
-                               "sda_size", "sdb_size"],
+                               "sda_size", "sdb_size", "vcpus"],
                        dynamic=self.dynamic_fields,
                        selected=self.op.output_fields)
 
@@ -2346,6 +2475,21 @@ class LUQueryInstances(NoHooksLU):
             val = None
           else:
             val = bool(live_data.get(instance.name))
+        elif field == "status":
+          if instance.primary_node in bad_nodes:
+            val = "ERROR_nodedown"
+          else:
+            running = bool(live_data.get(instance.name))
+            if running:
+              if instance.status != "down":
+                val = "running"
+              else:
+                val = "ERROR_up"
+            else:
+              if instance.status != "down":
+                val = "ERROR_down"
+              else:
+                val = "ADMIN_down"
         elif field == "admin_ram":
           val = instance.memory
         elif field == "oper_ram":
@@ -2369,6 +2513,8 @@ class LUQueryInstances(NoHooksLU):
             val = None
           else:
             val = disk.size
+        elif field == "vcpus":
+          val = instance.vcpus
         else:
           raise errors.ParameterError(field)
         iout.append(val)
@@ -2419,18 +2565,10 @@ class LUFailoverInstance(LogicalUnit):
       raise errors.ProgrammerError("no secondary node but using "
                                    "DT_REMOTE_RAID1 template")
 
-    # check memory requirements on the secondary node
     target_node = secondary_nodes[0]
-    nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName())
-    info = nodeinfo.get(target_node, None)
-    if not info:
-      raise errors.OpPrereqError("Cannot get current information"
-                                 " from node '%s'" % nodeinfo)
-    if instance.memory > info['memory_free']:
-      raise errors.OpPrereqError("Not enough memory on target node %s."
-                                 " %d MB available, %d MB required" %
-                                 (target_node, info['memory_free'],
-                                  instance.memory))
+    # check memory requirements on the secondary node
+    _CheckNodeFreeMemory(self.cfg, target_node, "failing over instance %s" %
+                         instance.name, instance.memory)
 
     # check bridge existance
     brlist = [nic.bridge for nic in instance.nics]
@@ -2461,21 +2599,6 @@ class LUFailoverInstance(LogicalUnit):
           raise errors.OpExecError("Disk %s is degraded on target node,"
                                    " aborting failover." % dev.iv_name)
 
-    feedback_fn("* checking target node resource availability")
-    nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName())
-
-    if not nodeinfo:
-      raise errors.OpExecError("Could not contact target node %s." %
-                               target_node)
-
-    free_memory = int(nodeinfo[target_node]['memory_free'])
-    memory = instance.memory
-    if memory > free_memory:
-      raise errors.OpExecError("Not enough memory to create instance %s on"
-                               " node %s. needed %s MiB, available %s MiB" %
-                               (instance.name, target_node, memory,
-                                free_memory))
-
     feedback_fn("* shutting down instance on source node")
     logger.Info("Shutting down instance %s on node %s" %
                 (instance.name, source_node))
@@ -2618,9 +2741,9 @@ def _GenerateDiskTemplate(cfg, template_name,
   #TODO: compute space requirements
 
   vgname = cfg.GetVGName()
-  if template_name == "diskless":
+  if template_name == constants.DT_DISKLESS:
     disks = []
-  elif template_name == "plain":
+  elif template_name == constants.DT_PLAIN:
     if len(secondary_nodes) != 0:
       raise errors.ProgrammerError("Wrong template configuration")
 
@@ -2632,7 +2755,7 @@ def _GenerateDiskTemplate(cfg, template_name,
                            logical_id=(vgname, names[1]),
                            iv_name = "sdb")
     disks = [sda_dev, sdb_dev]
-  elif template_name == "local_raid1":
+  elif template_name == constants.DT_LOCAL_RAID1:
     if len(secondary_nodes) != 0:
       raise errors.ProgrammerError("Wrong template configuration")
 
@@ -2762,7 +2885,7 @@ class LUCreateInstance(LogicalUnit):
   HTYPE = constants.HTYPE_INSTANCE
   _OP_REQP = ["instance_name", "mem_size", "disk_size", "pnode",
               "disk_template", "swap_size", "mode", "start", "vcpus",
-              "wait_for_sync", "ip_check"]
+              "wait_for_sync", "ip_check", "mac"]
 
   def BuildHooksEnv(self):
     """Build hooks env.
@@ -2788,7 +2911,7 @@ class LUCreateInstance(LogicalUnit):
       os_type=self.op.os_type,
       memory=self.op.mem_size,
       vcpus=self.op.vcpus,
-      nics=[(self.inst_ip, self.op.bridge)],
+      nics=[(self.inst_ip, self.op.bridge, self.op.mac)],
     ))
 
     nl = ([self.sstore.GetMasterNode(), self.op.pnode] +
@@ -2800,6 +2923,10 @@ class LUCreateInstance(LogicalUnit):
     """Check prerequisites.
 
     """
+    for attr in ["kernel_path", "initrd_path", "hvm_boot_order"]:
+      if not hasattr(self.op, attr):
+        setattr(self.op, attr, None)
+
     if self.op.mode not in (constants.INSTANCE_CREATE,
                             constants.INSTANCE_IMPORT):
       raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
@@ -2871,13 +2998,9 @@ class LUCreateInstance(LogicalUnit):
                                    " the primary node.")
       self.secondaries.append(snode_name)
 
-    # Check lv size requirements
-    nodenames = [pnode.name] + self.secondaries
-    nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName())
-
     # Required free disk space as a function of disk and swap space
     req_size_dict = {
-      constants.DT_DISKLESS: 0,
+      constants.DT_DISKLESS: None,
       constants.DT_PLAIN: self.op.disk_size + self.op.swap_size,
       constants.DT_LOCAL_RAID1: (self.op.disk_size + self.op.swap_size) * 2,
       # 256 MB are added for drbd metadata, 128MB for each drbd device
@@ -2891,15 +3014,23 @@ class LUCreateInstance(LogicalUnit):
 
     req_size = req_size_dict[self.op.disk_template]
 
-    for node in nodenames:
-      info = nodeinfo.get(node, None)
-      if not info:
-        raise errors.OpPrereqError("Cannot get current information"
-                                   " from node '%s'" % nodeinfo)
-      if req_size > info['vg_free']:
-        raise errors.OpPrereqError("Not enough disk space on target node %s."
-                                   " %d MB available, %d MB required" %
-                                   (node, info['vg_free'], req_size))
+    # Check lv size requirements
+    if req_size is not None:
+      nodenames = [pnode.name] + self.secondaries
+      nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName())
+      for node in nodenames:
+        info = nodeinfo.get(node, None)
+        if not info:
+          raise errors.OpPrereqError("Cannot get current information"
+                                     " from node '%s'" % nodeinfo)
+        vg_free = info.get('vg_free', None)
+        if not isinstance(vg_free, int):
+          raise errors.OpPrereqError("Can't compute free disk space on"
+                                     " node %s" % node)
+        if req_size > info['vg_free']:
+          raise errors.OpPrereqError("Not enough disk space on target node %s."
+                                     " %d MB available, %d MB required" %
+                                     (node, info['vg_free'], req_size))
 
     # os verification
     os_obj = rpc.call_os_get(pnode.name, self.op.os_type)
@@ -2907,6 +3038,9 @@ class LUCreateInstance(LogicalUnit):
       raise errors.OpPrereqError("OS '%s' not in supported os list for"
                                  " primary node"  % self.op.os_type)
 
+    if self.op.kernel_path == constants.VALUE_NONE:
+      raise errors.OpPrereqError("Can't set instance kernel to none")
+
     # instance verification
     hostname1 = utils.HostInfo(self.op.instance_name)
 
@@ -2938,6 +3072,12 @@ class LUCreateInstance(LogicalUnit):
         raise errors.OpPrereqError("IP %s of instance %s already in use" %
                                    (hostname1.ip, instance_name))
 
+    # MAC address verification
+    if self.op.mac != "auto":
+      if not utils.IsValidMac(self.op.mac.lower()):
+        raise errors.OpPrereqError("invalid MAC address specified: %s" %
+                                   self.op.mac)
+
     # bridge verification
     bridge = getattr(self.op, "bridge", None)
     if bridge is None:
@@ -2950,6 +3090,12 @@ class LUCreateInstance(LogicalUnit):
                                  " destination node '%s'" %
                                  (self.op.bridge, pnode.name))
 
+    # boot order verification
+    if self.op.hvm_boot_order is not None:
+      if len(self.op.hvm_boot_order.strip("acdn")) != 0:
+        raise errors.OpPrereqError("invalid boot order specified,"
+                                   " must be one or more of [acdn]")
+
     if self.op.start:
       self.instance_status = 'up'
     else:
@@ -2962,10 +3108,21 @@ class LUCreateInstance(LogicalUnit):
     instance = self.op.instance_name
     pnode_name = self.pnode.name
 
-    nic = objects.NIC(bridge=self.op.bridge, mac=self.cfg.GenerateMAC())
+    if self.op.mac == "auto":
+      mac_address = self.cfg.GenerateMAC()
+    else:
+      mac_address = self.op.mac
+
+    nic = objects.NIC(bridge=self.op.bridge, mac=mac_address)
     if self.inst_ip is not None:
       nic.ip = self.inst_ip
 
+    ht_kind = self.sstore.GetHypervisorType()
+    if ht_kind in constants.HTS_REQ_PORT:
+      network_port = self.cfg.AllocatePort()
+    else:
+      network_port = None
+
     disks = _GenerateDiskTemplate(self.cfg,
                                   self.op.disk_template,
                                   instance, pnode_name,
@@ -2979,6 +3136,10 @@ class LUCreateInstance(LogicalUnit):
                             nics=[nic], disks=disks,
                             disk_template=self.op.disk_template,
                             status=self.instance_status,
+                            network_port=network_port,
+                            kernel_path=self.op.kernel_path,
+                            initrd_path=self.op.initrd_path,
+                            hvm_boot_order=self.op.hvm_boot_order,
                             )
 
     feedback_fn("* creating instance disks...")
@@ -3078,7 +3239,7 @@ class LUConnectConsole(NoHooksLU):
     logger.Debug("connecting to console of %s on %s" % (instance.name, node))
 
     hyper = hypervisor.GetHypervisor()
-    console_cmd = hyper.GetShellCommandForConsole(instance.name)
+    console_cmd = hyper.GetShellCommandForConsole(instance)
     # build ssh cmdline
     argv = ["ssh", "-q", "-t"]
     argv.extend(ssh.KNOWN_HOSTS_OPTS)
@@ -3595,7 +3756,7 @@ class LUReplaceDisks(LogicalUnit):
       # ok, we created the new LVs, so now we know we have the needed
       # storage; as such, we proceed on the target node to rename
       # old_lv to _old, and new_lv to old_lv; note that we rename LVs
-      # using the assumption than logical_id == physical_id (which in
+      # using the assumption that logical_id == physical_id (which in
       # turn is the unique_id on that node)
 
       # FIXME(iustin): use a better name for the replaced LVs
@@ -3631,7 +3792,7 @@ class LUReplaceDisks(LogicalUnit):
       if not rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs):
         for new_lv in new_lvs:
           if not rpc.call_blockdev_remove(tgt_node, new_lv):
-            warning("Can't rollback device %s", "manually cleanup unused"
+            warning("Can't rollback device %s", hint="manually cleanup unused"
                     " logical volumes")
         raise errors.OpExecError("Can't add local storage to drbd")
 
@@ -3660,7 +3821,7 @@ class LUReplaceDisks(LogicalUnit):
       for lv in old_lvs:
         cfg.SetDiskID(lv, tgt_node)
         if not rpc.call_blockdev_remove(tgt_node, lv):
-          warning("Can't remove old LV", "manually remove unused LVs")
+          warning("Can't remove old LV", hint="manually remove unused LVs")
           continue
 
   def _ExecD8Secondary(self, feedback_fn):
@@ -3763,7 +3924,7 @@ class LUReplaceDisks(LogicalUnit):
       cfg.SetDiskID(dev, old_node)
       if not rpc.call_blockdev_shutdown(old_node, dev):
         warning("Failed to shutdown drbd for %s on old node" % dev.iv_name,
-                "Please cleanup this device manually as soon as possible")
+                hint="Please cleanup this device manually as soon as possible")
 
     info("detaching primary drbds from the network (=> standalone)")
     done = 0
@@ -3825,7 +3986,7 @@ class LUReplaceDisks(LogicalUnit):
         cfg.SetDiskID(lv, old_node)
         if not rpc.call_blockdev_remove(old_node, lv):
           warning("Can't remove LV on old secondary",
-                  "Cleanup stale volumes by hand")
+                  hint="Cleanup stale volumes by hand")
 
   def Exec(self, feedback_fn):
     """Execute disk replacement.
@@ -3867,7 +4028,7 @@ class LUQueryInstanceData(NoHooksLU):
         instance = self.cfg.GetInstanceInfo(self.cfg.ExpandInstanceName(name))
         if instance is None:
           raise errors.OpPrereqError("No such instance name '%s'" % name)
-      self.wanted_instances.append(instance)
+        self.wanted_instances.append(instance)
     else:
       self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
                                in self.cfg.GetInstanceList()]
@@ -3939,7 +4100,11 @@ class LUQueryInstanceData(NoHooksLU):
         "memory": instance.memory,
         "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
         "disks": disks,
+        "network_port": instance.network_port,
         "vcpus": instance.vcpus,
+        "kernel_path": instance.kernel_path,
+        "initrd_path": instance.initrd_path,
+        "hvm_boot_order": instance.hvm_boot_order,
         }
 
       result[instance.name] = idict
@@ -3966,7 +4131,7 @@ class LUSetInstanceParms(LogicalUnit):
       args['memory'] = self.mem
     if self.vcpus:
       args['vcpus'] = self.vcpus
-    if self.do_ip or self.do_bridge:
+    if self.do_ip or self.do_bridge or self.mac:
       if self.do_ip:
         ip = self.ip
       else:
@@ -3975,7 +4140,11 @@ class LUSetInstanceParms(LogicalUnit):
         bridge = self.bridge
       else:
         bridge = self.instance.nics[0].bridge
-      args['nics'] = [(ip, bridge)]
+      if self.mac:
+        mac = self.mac
+      else:
+        mac = self.instance.nics[0].mac
+      args['nics'] = [(ip, bridge, mac)]
     env = _BuildInstanceHookEnvByObject(self.instance, override=args)
     nl = [self.sstore.GetMasterNode(),
           self.instance.primary_node] + list(self.instance.secondary_nodes)
@@ -3990,8 +4159,14 @@ class LUSetInstanceParms(LogicalUnit):
     self.mem = getattr(self.op, "mem", None)
     self.vcpus = getattr(self.op, "vcpus", None)
     self.ip = getattr(self.op, "ip", None)
+    self.mac = getattr(self.op, "mac", None)
     self.bridge = getattr(self.op, "bridge", None)
-    if [self.mem, self.vcpus, self.ip, self.bridge].count(None) == 4:
+    self.kernel_path = getattr(self.op, "kernel_path", None)
+    self.initrd_path = getattr(self.op, "initrd_path", None)
+    self.hvm_boot_order = getattr(self.op, "hvm_boot_order", None)
+    all_parms = [self.mem, self.vcpus, self.ip, self.bridge, self.mac,
+                 self.kernel_path, self.initrd_path, self.hvm_boot_order]
+    if all_parms.count(None) == len(all_parms):
       raise errors.OpPrereqError("No changes submitted")
     if self.mem is not None:
       try:
@@ -4013,6 +4188,42 @@ class LUSetInstanceParms(LogicalUnit):
     else:
       self.do_ip = False
     self.do_bridge = (self.bridge is not None)
+    if self.mac is not None:
+      if self.cfg.IsMacInUse(self.mac):
+        raise errors.OpPrereqError('MAC address %s already in use in cluster' %
+                                   self.mac)
+      if not utils.IsValidMac(self.mac):
+        raise errors.OpPrereqError('Invalid MAC address %s' % self.mac)
+
+    if self.kernel_path is not None:
+      self.do_kernel_path = True
+      if self.kernel_path == constants.VALUE_NONE:
+        raise errors.OpPrereqError("Can't set instance to no kernel")
+
+      if self.kernel_path != constants.VALUE_DEFAULT:
+        if not os.path.isabs(self.kernel_path):
+          raise errors.OpPrereqError("The kernel path must be an absolute"
+                                    " filename")
+    else:
+      self.do_kernel_path = False
+
+    if self.initrd_path is not None:
+      self.do_initrd_path = True
+      if self.initrd_path not in (constants.VALUE_NONE,
+                                  constants.VALUE_DEFAULT):
+        if not os.path.isabs(self.initrd_path):
+          raise errors.OpPrereqError("The initrd path must be an absolute"
+                                    " filename")
+    else:
+      self.do_initrd_path = False
+
+    # boot order verification
+    if self.hvm_boot_order is not None:
+      if self.hvm_boot_order != constants.VALUE_DEFAULT:
+        if len(self.hvm_boot_order.strip("acdn")) != 0:
+          raise errors.OpPrereqError("invalid boot order specified,"
+                                     " must be one or more of [acdn]"
+                                     " or 'default'")
 
     instance = self.cfg.GetInstanceInfo(
       self.cfg.ExpandInstanceName(self.op.instance_name))
@@ -4042,6 +4253,21 @@ class LUSetInstanceParms(LogicalUnit):
     if self.bridge:
       instance.nics[0].bridge = self.bridge
       result.append(("bridge", self.bridge))
+    if self.mac:
+      instance.nics[0].mac = self.mac
+      result.append(("mac", self.mac))
+    if self.do_kernel_path:
+      instance.kernel_path = self.kernel_path
+      result.append(("kernel_path", self.kernel_path))
+    if self.do_initrd_path:
+      instance.initrd_path = self.initrd_path
+      result.append(("initrd_path", self.initrd_path))
+    if self.hvm_boot_order:
+      if self.hvm_boot_order == constants.VALUE_DEFAULT:
+        instance.hvm_boot_order = None
+      else:
+        instance.hvm_boot_order = self.hvm_boot_order
+      result.append(("hvm_boot_order", self.hvm_boot_order))
 
     self.cfg.AddInstance(instance)
 
@@ -4335,3 +4561,39 @@ class LUDelTags(TagsLU):
       raise errors.OpRetryError("There has been a modification to the"
                                 " config file and the operation has been"
                                 " aborted. Please retry.")
+
+class LUTestDelay(NoHooksLU):
+  """Sleep for a specified amount of time.
+
+  This LU sleeps on the master and/or nodes for a specified amoutn of
+  time.
+
+  """
+  _OP_REQP = ["duration", "on_master", "on_nodes"]
+
+  def CheckPrereq(self):
+    """Check prerequisites.
+
+    This checks that we have a good list of nodes and/or the duration
+    is valid.
+
+    """
+
+    if self.op.on_nodes:
+      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
+
+  def Exec(self, feedback_fn):
+    """Do the actual sleep.
+
+    """
+    if self.op.on_master:
+      if not utils.TestDelay(self.op.duration):
+        raise errors.OpExecError("Error during master delay test")
+    if self.op.on_nodes:
+      result = rpc.call_test_delay(self.op.on_nodes, self.op.duration)
+      if not result:
+        raise errors.OpExecError("Complete failure from rpc call")
+      for node, node_result in result.items():
+        if not node_result:
+          raise errors.OpExecError("Failure during rpc call to node %s,"
+                                   " result: %s" % (node, node_result))