cluster verify and instance disks on offline nodes
[ganeti-local] / lib / bootstrap.py
index 85fd0bc..d18e68b 100644 (file)
@@ -1,7 +1,7 @@
 #
 #
 
-# Copyright (C) 2006, 2007, 2008 Google Inc.
+# Copyright (C) 2006, 2007, 2008, 2010 Google Inc.
 #
 # This program is free software; you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -27,7 +27,6 @@ import os
 import os.path
 import re
 import logging
-import tempfile
 import time
 
 from ganeti import rpc
@@ -40,6 +39,17 @@ from ganeti import objects
 from ganeti import ssconf
 from ganeti import serializer
 from ganeti import hypervisor
+from ganeti import bdev
+from ganeti import netutils
+from ganeti import backend
+from ganeti import luxi
+
+
+# ec_id for InitConfig's temporary reservation manager
+_INITCONF_ECID = "initconfig-ecid"
+
+#: After how many seconds daemon must be responsive
+_DAEMON_READY_TIMEOUT = 10.0
 
 
 def _InitSSHSetup():
@@ -66,40 +76,6 @@ def _InitSSHSetup():
   utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
 
 
-def GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
-  """Generates a self-signed SSL certificate.
-
-  @type file_name: str
-  @param file_name: Path to output file
-  @type validity: int
-  @param validity: Validity for certificate in days
-
-  """
-  (fd, tmp_file_name) = tempfile.mkstemp(dir=os.path.dirname(file_name))
-  try:
-    try:
-      # Set permissions before writing key
-      os.chmod(tmp_file_name, 0600)
-
-      result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
-                             "-days", str(validity), "-nodes", "-x509",
-                             "-keyout", tmp_file_name, "-out", tmp_file_name,
-                             "-batch"])
-      if result.failed:
-        raise errors.OpExecError("Could not generate SSL certificate, command"
-                                 " %s had exitcode %s and error message %s" %
-                                 (result.cmd, result.exit_code, result.output))
-
-      # Make read-only
-      os.chmod(tmp_file_name, 0400)
-
-      os.rename(tmp_file_name, file_name)
-    finally:
-      utils.RemoveFile(tmp_file_name)
-  finally:
-    os.close(fd)
-
-
 def GenerateHmacKey(file_name):
   """Writes a new HMAC key.
 
@@ -111,62 +87,85 @@ def GenerateHmacKey(file_name):
                   backup=True)
 
 
-def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_hmac_key,
-                          rapi_cert_pem=None):
+def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key,
+                          new_cds, rapi_cert_pem=None, cds=None,
+                          nodecert_file=constants.NODED_CERT_FILE,
+                          rapicert_file=constants.RAPI_CERT_FILE,
+                          hmackey_file=constants.CONFD_HMAC_KEY,
+                          cds_file=constants.CLUSTER_DOMAIN_SECRET_FILE):
   """Updates the cluster certificates, keys and secrets.
 
   @type new_cluster_cert: bool
   @param new_cluster_cert: Whether to generate a new cluster certificate
   @type new_rapi_cert: bool
   @param new_rapi_cert: Whether to generate a new RAPI certificate
-  @type new_hmac_key: bool
-  @param new_hmac_key: Whether to generate a new HMAC key
+  @type new_confd_hmac_key: bool
+  @param new_confd_hmac_key: Whether to generate a new HMAC key
+  @type new_cds: bool
+  @param new_cds: Whether to generate a new cluster domain secret
   @type rapi_cert_pem: string
   @param rapi_cert_pem: New RAPI certificate in PEM format
+  @type cds: string
+  @param cds: New cluster domain secret
+  @type nodecert_file: string
+  @param nodecert_file: optional override of the node cert file path
+  @type rapicert_file: string
+  @param rapicert_file: optional override of the rapi cert file path
+  @type hmackey_file: string
+  @param hmackey_file: optional override of the hmac key file path
 
   """
-  # SSL certificate
-  cluster_cert_exists = os.path.exists(constants.SSL_CERT_FILE)
+  # noded SSL certificate
+  cluster_cert_exists = os.path.exists(nodecert_file)
   if new_cluster_cert or not cluster_cert_exists:
     if cluster_cert_exists:
-      utils.CreateBackup(constants.SSL_CERT_FILE)
+      utils.CreateBackup(nodecert_file)
 
-    logging.debug("Generating new cluster certificate at %s",
-                  constants.SSL_CERT_FILE)
-    GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
+    logging.debug("Generating new cluster certificate at %s", nodecert_file)
+    utils.GenerateSelfSignedSslCert(nodecert_file)
 
-  # HMAC key
-  if new_hmac_key or not os.path.exists(constants.HMAC_CLUSTER_KEY):
-    logging.debug("Writing new HMAC key to %s", constants.HMAC_CLUSTER_KEY)
-    GenerateHmacKey(constants.HMAC_CLUSTER_KEY)
+  # confd HMAC key
+  if new_confd_hmac_key or not os.path.exists(hmackey_file):
+    logging.debug("Writing new confd HMAC key to %s", hmackey_file)
+    GenerateHmacKey(hmackey_file)
 
   # RAPI
-  rapi_cert_exists = os.path.exists(constants.RAPI_CERT_FILE)
+  rapi_cert_exists = os.path.exists(rapicert_file)
 
   if rapi_cert_pem:
     # Assume rapi_pem contains a valid PEM-formatted certificate and key
-    logging.debug("Writing RAPI certificate at %s",
-                  constants.RAPI_CERT_FILE)
-    utils.WriteFile(constants.RAPI_CERT_FILE, data=rapi_cert_pem, backup=True)
+    logging.debug("Writing RAPI certificate at %s", rapicert_file)
+    utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
 
   elif new_rapi_cert or not rapi_cert_exists:
     if rapi_cert_exists:
-      utils.CreateBackup(constants.RAPI_CERT_FILE)
+      utils.CreateBackup(rapicert_file)
+
+    logging.debug("Generating new RAPI certificate at %s", rapicert_file)
+    utils.GenerateSelfSignedSslCert(rapicert_file)
+
+  # Cluster domain secret
+  if cds:
+    logging.debug("Writing cluster domain secret to %s", cds_file)
+    utils.WriteFile(cds_file, data=cds, backup=True)
 
-    logging.debug("Generating new RAPI certificate at %s",
-                  constants.RAPI_CERT_FILE)
-    GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
+  elif new_cds or not os.path.exists(cds_file):
+    logging.debug("Generating new cluster domain secret at %s", cds_file)
+    GenerateHmacKey(cds_file)
 
 
 def _InitGanetiServerSetup(master_name):
   """Setup the necessary configuration for the initial node daemon.
 
   This creates the nodepass file containing the shared password for
-  the cluster and also generates the SSL certificate.
+  the cluster, generates the SSL certificate and starts the node daemon.
+
+  @type master_name: str
+  @param master_name: Name of the master node
 
   """
   # Generate cluster secrets
-  GenerateClusterCrypto(True, False, False)
+  GenerateClusterCrypto(True, False, False, False)
 
   result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED])
   if result.failed:
@@ -187,17 +186,72 @@ def _WaitForNodeDaemon(node_name):
       raise utils.RetryAgain()
 
   try:
-    utils.Retry(_CheckNodeDaemon, 1.0, 10.0)
+    utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT)
   except utils.RetryTimeout:
     raise errors.OpExecError("Node daemon on %s didn't answer queries within"
-                             " 10 seconds" % node_name)
+                             " %s seconds" % (node_name, _DAEMON_READY_TIMEOUT))
+
+
+def _WaitForMasterDaemon():
+  """Wait for master daemon to become responsive.
 
+  """
+  def _CheckMasterDaemon():
+    try:
+      cl = luxi.Client()
+      (cluster_name, ) = cl.QueryConfigValues(["cluster_name"])
+    except Exception:
+      raise utils.RetryAgain()
+
+    logging.debug("Received cluster name %s from master", cluster_name)
+
+  try:
+    utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT)
+  except utils.RetryTimeout:
+    raise errors.OpExecError("Master daemon didn't answer queries within"
+                             " %s seconds" % _DAEMON_READY_TIMEOUT)
+
+
+def _InitFileStorage(file_storage_dir):
+  """Initialize if needed the file storage.
+
+  @param file_storage_dir: the user-supplied value
+  @return: either empty string (if file storage was disabled at build
+      time) or the normalized path to the storage directory
+
+  """
+  if not constants.ENABLE_FILE_STORAGE:
+    return ""
+
+  file_storage_dir = os.path.normpath(file_storage_dir)
+
+  if not os.path.isabs(file_storage_dir):
+    raise errors.OpPrereqError("The file storage directory you passed is"
+                               " not an absolute path.", errors.ECODE_INVAL)
+
+  if not os.path.exists(file_storage_dir):
+    try:
+      os.makedirs(file_storage_dir, 0750)
+    except OSError, err:
+      raise errors.OpPrereqError("Cannot create file storage directory"
+                                 " '%s': %s" % (file_storage_dir, err),
+                                 errors.ECODE_ENVIRON)
+
+  if not os.path.isdir(file_storage_dir):
+    raise errors.OpPrereqError("The file storage directory '%s' is not"
+                               " a directory." % file_storage_dir,
+                               errors.ECODE_ENVIRON)
+  return file_storage_dir
 
-def InitCluster(cluster_name, mac_prefix,
+
+def InitCluster(cluster_name, mac_prefix, # pylint: disable-msg=R0913
                 master_netdev, file_storage_dir, candidate_pool_size,
                 secondary_ip=None, vg_name=None, beparams=None,
-                nicparams=None, hvparams=None, enabled_hypervisors=None,
-                modify_etc_hosts=True, modify_ssh_setup=True):
+                nicparams=None, ndparams=None, hvparams=None,
+                enabled_hypervisors=None, modify_etc_hosts=True,
+                modify_ssh_setup=True, maintain_node_health=False,
+                drbd_helper=None, uid_pool=None, default_iallocator=None,
+                primary_ip_version=None, prealloc_wipe_disks=False):
   """Initialise the cluster.
 
   @type candidate_pool_size: int
@@ -218,39 +272,56 @@ def InitCluster(cluster_name, mac_prefix,
                                " entries: %s" % invalid_hvs,
                                errors.ECODE_INVAL)
 
-  hostname = utils.GetHostInfo()
 
-  if hostname.ip.startswith("127."):
-    raise errors.OpPrereqError("This host's IP resolves to the private"
-                               " range (%s). Please fix DNS or %s." %
+  ipcls = None
+  if primary_ip_version == constants.IP4_VERSION:
+    ipcls = netutils.IP4Address
+  elif primary_ip_version == constants.IP6_VERSION:
+    ipcls = netutils.IP6Address
+  else:
+    raise errors.OpPrereqError("Invalid primary ip version: %d." %
+                               primary_ip_version)
+
+  hostname = netutils.GetHostname(family=ipcls.family)
+  if not ipcls.IsValid(hostname.ip):
+    raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d"
+                               " address." % (hostname.ip, primary_ip_version))
+
+  if ipcls.IsLoopback(hostname.ip):
+    raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback"
+                               " address. Please fix DNS or %s." %
                                (hostname.ip, constants.ETC_HOSTS),
                                errors.ECODE_ENVIRON)
 
-  if not utils.OwnIpAddress(hostname.ip):
+  if not ipcls.Own(hostname.ip):
     raise errors.OpPrereqError("Inconsistency: this host's name resolves"
                                " to %s,\nbut this ip address does not"
-                               " belong to this host. Aborting." %
+                               " belong to this host" %
                                hostname.ip, errors.ECODE_ENVIRON)
 
-  clustername = utils.GetHostInfo(utils.HostInfo.NormalizeName(cluster_name))
+  clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family)
 
-  if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
-                   timeout=5):
-    raise errors.OpPrereqError("Cluster IP already active. Aborting.",
+  if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5):
+    raise errors.OpPrereqError("Cluster IP already active",
                                errors.ECODE_NOTUNIQUE)
 
-  if secondary_ip:
-    if not utils.IsValidIP(secondary_ip):
-      raise errors.OpPrereqError("Invalid secondary ip given",
+  if not secondary_ip:
+    if primary_ip_version == constants.IP6_VERSION:
+      raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
+                                 " IPv4 address must be given as secondary",
                                  errors.ECODE_INVAL)
-    if (secondary_ip != hostname.ip and
-        not utils.OwnIpAddress(secondary_ip)):
-      raise errors.OpPrereqError("You gave %s as secondary IP,"
-                                 " but it does not belong to this host." %
-                                 secondary_ip, errors.ECODE_ENVIRON)
-  else:
     secondary_ip = hostname.ip
 
+  if not netutils.IP4Address.IsValid(secondary_ip):
+    raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid"
+                               " IPv4 address." % secondary_ip,
+                               errors.ECODE_INVAL)
+
+  if not netutils.IP4Address.Own(secondary_ip):
+    raise errors.OpPrereqError("You gave %s as secondary IP,"
+                               " but it does not belong to this host." %
+                               secondary_ip, errors.ECODE_ENVIRON)
+
   if vg_name is not None:
     # Check if volume group is valid
     vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
@@ -260,24 +331,21 @@ def InitCluster(cluster_name, mac_prefix,
                                  " you are not using lvm" % vgstatus,
                                  errors.ECODE_INVAL)
 
-  file_storage_dir = os.path.normpath(file_storage_dir)
-
-  if not os.path.isabs(file_storage_dir):
-    raise errors.OpPrereqError("The file storage directory you passed is"
-                               " not an absolute path.", errors.ECODE_INVAL)
-
-  if not os.path.exists(file_storage_dir):
+  if drbd_helper is not None:
     try:
-      os.makedirs(file_storage_dir, 0750)
-    except OSError, err:
-      raise errors.OpPrereqError("Cannot create file storage directory"
-                                 " '%s': %s" % (file_storage_dir, err),
+      curr_helper = bdev.BaseDRBD.GetUsermodeHelper()
+    except errors.BlockDeviceError, err:
+      raise errors.OpPrereqError("Error while checking drbd helper"
+                                 " (specify --no-drbd-storage if you are not"
+                                 " using drbd): %s" % str(err),
                                  errors.ECODE_ENVIRON)
+    if drbd_helper != curr_helper:
+      raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s"
+                                 " is the current helper" % (drbd_helper,
+                                                             curr_helper),
+                                 errors.ECODE_INVAL)
 
-  if not os.path.isdir(file_storage_dir):
-    raise errors.OpPrereqError("The file storage directory '%s' is not"
-                               " a directory." % file_storage_dir,
-                               errors.ECODE_ENVIRON)
+  file_storage_dir = _InitFileStorage(file_storage_dir)
 
   if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
     raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
@@ -296,25 +364,36 @@ def InitCluster(cluster_name, mac_prefix,
   utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
   objects.NIC.CheckParameterSyntax(nicparams)
 
+  if ndparams is not None:
+    utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES)
+  else:
+    ndparams = dict(constants.NDC_DEFAULTS)
+
   # hvparams is a mapping of hypervisor->hvparams dict
   for hv_name, hv_params in hvparams.iteritems():
     utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
     hv_class = hypervisor.GetHypervisor(hv_name)
     hv_class.CheckParameterSyntax(hv_params)
 
-  # set up the inter-node password and certificate
-  _InitGanetiServerSetup(hostname.name)
-
   # set up ssh config and /etc/hosts
   sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB)
   sshkey = sshline.split(" ")[1]
 
   if modify_etc_hosts:
-    utils.AddHostToEtcHosts(hostname.name)
+    utils.AddHostToEtcHosts(hostname.name, hostname.ip)
 
   if modify_ssh_setup:
     _InitSSHSetup()
 
+  if default_iallocator is not None:
+    alloc_script = utils.FindFile(default_iallocator,
+                                  constants.IALLOCATOR_SEARCH_PATH,
+                                  os.path.isfile)
+    if alloc_script is None:
+      raise errors.OpPrereqError("Invalid default iallocator script '%s'"
+                                 " specified" % default_iallocator,
+                                 errors.ECODE_INVAL)
+
   now = time.time()
 
   # init of cluster config file
@@ -333,13 +412,19 @@ def InitCluster(cluster_name, mac_prefix,
     enabled_hypervisors=enabled_hypervisors,
     beparams={constants.PP_DEFAULT: beparams},
     nicparams={constants.PP_DEFAULT: nicparams},
+    ndparams=ndparams,
     hvparams=hvparams,
     candidate_pool_size=candidate_pool_size,
     modify_etc_hosts=modify_etc_hosts,
     modify_ssh_setup=modify_ssh_setup,
+    uid_pool=uid_pool,
     ctime=now,
     mtime=now,
-    uuid=utils.NewUUID(),
+    maintain_node_health=maintain_node_health,
+    drbd_usermode_helper=drbd_helper,
+    default_iallocator=default_iallocator,
+    primary_ip_family=ipcls.family,
+    prealloc_wipe_disks=prealloc_wipe_disks,
     )
   master_node_config = objects.Node(name=hostname.name,
                                     primary_ip=hostname.ip,
@@ -347,16 +432,25 @@ def InitCluster(cluster_name, mac_prefix,
                                     serial_no=1,
                                     master_candidate=True,
                                     offline=False, drained=False,
+                                    ctime=now, mtime=now,
                                     )
   InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
-  cfg = config.ConfigWriter()
+  cfg = config.ConfigWriter(offline=True)
   ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
   cfg.Update(cfg.GetClusterInfo(), logging.error)
+  backend.WriteSsconfFiles(cfg.GetSsconfValues())
 
-  # start the master ip
-  # TODO: Review rpc call from bootstrap
-  # TODO: Warn on failed start master
-  rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
+  # set up the inter-node password and certificate
+  _InitGanetiServerSetup(hostname.name)
+
+  logging.debug("Starting daemons")
+  result = utils.RunCmd([constants.DAEMON_UTIL, "start-all"])
+  if result.failed:
+    raise errors.OpExecError("Could not start daemons, command %s"
+                             " had exitcode %s and error %s" %
+                             (result.cmd, result.exit_code, result.output))
+
+  _WaitForMasterDaemon()
 
 
 def InitConfig(version, cluster_config, master_node_config,
@@ -376,13 +470,26 @@ def InitConfig(version, cluster_config, master_node_config,
   @param cfg_file: configuration file path
 
   """
+  uuid_generator = config.TemporaryReservationManager()
+  cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID,
+                                                _INITCONF_ECID)
+  master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID,
+                                                    _INITCONF_ECID)
   nodes = {
     master_node_config.name: master_node_config,
     }
-
+  default_nodegroup = objects.NodeGroup(
+    uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID),
+    name=constants.INITIAL_NODE_GROUP_NAME,
+    members=[master_node_config.name],
+    )
+  nodegroups = {
+    default_nodegroup.uuid: default_nodegroup,
+    }
   now = time.time()
   config_data = objects.ConfigData(version=version,
                                    cluster=cluster_config,
+                                   nodegroups=nodegroups,
                                    nodes=nodes,
                                    instances={},
                                    serial_no=1,
@@ -424,51 +531,28 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check):
   @param ssh_key_check: whether to do a strict key check
 
   """
-  sshrunner = ssh.SshRunner(cluster_name)
-
-  noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
-  rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
-  hmac_key = utils.ReadFile(constants.HMAC_CLUSTER_KEY)
-
-  # in the base64 pem encoding, neither '!' nor '.' are valid chars,
-  # so we use this to detect an invalid certificate; as long as the
-  # cert doesn't contain this, the here-document will be correctly
-  # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
-  # so the same restrictions apply.
-  for content in (noded_cert, rapi_cert, hmac_key):
-    if re.search('^!EOF\.', content, re.MULTILINE):
-      raise errors.OpExecError("invalid SSL certificate or HMAC key")
-
-  if not noded_cert.endswith("\n"):
-    noded_cert += "\n"
-  if not rapi_cert.endswith("\n"):
-    rapi_cert += "\n"
-  if not hmac_key.endswith("\n"):
-    hmac_key += "\n"
+  family = ssconf.SimpleStore().GetPrimaryIPFamily()
+  sshrunner = ssh.SshRunner(cluster_name,
+                            ipv6=(family == netutils.IP6Address.family))
+
+  bind_address = constants.IP4_ADDRESS_ANY
+  if family == netutils.IP6Address.family:
+    bind_address = constants.IP6_ADDRESS_ANY
 
   # set up inter-node password and certificate and restarts the node daemon
   # and then connect with ssh to set password and start ganeti-noded
   # note that all the below variables are sanitized at this point,
   # either by being constants or by the checks above
-  mycommand = ("umask 077 && "
-               "cat > '%s' << '!EOF.' && \n"
-               "%s!EOF.\n"
-               "cat > '%s' << '!EOF.' && \n"
-               "%s!EOF.\n"
-               "cat > '%s' << '!EOF.' && \n"
-               "%s!EOF.\n"
-               "chmod 0400 %s %s %s && "
-               "%s start %s" %
-               (constants.SSL_CERT_FILE, noded_cert,
-                constants.RAPI_CERT_FILE, rapi_cert,
-                constants.HMAC_CLUSTER_KEY, hmac_key,
-                constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
-                constants.HMAC_CLUSTER_KEY,
-                constants.DAEMON_UTIL, constants.NODED))
+  sshrunner.CopyFileToNode(node, constants.NODED_CERT_FILE)
+  sshrunner.CopyFileToNode(node, constants.RAPI_CERT_FILE)
+  sshrunner.CopyFileToNode(node, constants.CONFD_HMAC_KEY)
+  mycommand = ("%s stop-all; %s start %s -b %s" %
+               (constants.DAEMON_UTIL, constants.DAEMON_UTIL, constants.NODED,
+                utils.ShellQuote(bind_address)))
 
   result = sshrunner.Run(node, 'root', mycommand, batch=False,
                          ask_key=ssh_key_check,
-                         use_cluster_key=False,
+                         use_cluster_key=True,
                          strict_host_check=ssh_key_check)
   if result.failed:
     raise errors.OpExecError("Remote command on node %s, error: %s,"
@@ -532,23 +616,51 @@ def MasterFailover(no_voting=False):
 
   logging.info("Setting master to %s, old master: %s", new_master, old_master)
 
+  try:
+    # instantiate a real config writer, as we now know we have the
+    # configuration data
+    cfg = config.ConfigWriter(accept_foreign=True)
+
+    cluster_info = cfg.GetClusterInfo()
+    cluster_info.master_node = new_master
+    # this will also regenerate the ssconf files, since we updated the
+    # cluster info
+    cfg.Update(cluster_info, logging.error)
+  except errors.ConfigurationError, err:
+    logging.error("Error while trying to set the new master: %s",
+                  str(err))
+    return 1
+
+  # if cfg.Update worked, then it means the old master daemon won't be
+  # able now to write its own config file (we rely on locking in both
+  # backend.UploadFile() and ConfigWriter._Write(); hence the next
+  # step is to kill the old master
+
+  logging.info("Stopping the master daemon on node %s", old_master)
+
   result = rpc.RpcRunner.call_node_stop_master(old_master, True)
   msg = result.fail_msg
   if msg:
     logging.error("Could not disable the master role on the old master"
                  " %s, please disable manually: %s", old_master, msg)
 
+  logging.info("Checking master IP non-reachability...")
+
+  master_ip = sstore.GetMasterIP()
+  total_timeout = 30
   # Here we have a phase where no master should be running
+  def _check_ip():
+    if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
+      raise utils.RetryAgain()
 
-  # instantiate a real config writer, as we now know we have the
-  # configuration data
-  cfg = config.ConfigWriter()
+  try:
+    utils.Retry(_check_ip, (1, 1.5, 5), total_timeout)
+  except utils.RetryTimeout:
+    logging.warning("The master IP is still reachable after %s seconds,"
+                    " continuing but activating the master on the current"
+                    " node will probably fail", total_timeout)
 
-  cluster_info = cfg.GetClusterInfo()
-  cluster_info.master_node = new_master
-  # this will also regenerate the ssconf files, since we updated the
-  # cluster info
-  cfg.Update(cluster_info, logging.error)
+  logging.info("Starting the master daemons on the new master")
 
   result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
   msg = result.fail_msg
@@ -557,6 +669,7 @@ def MasterFailover(no_voting=False):
                   " %s, please check: %s", new_master, msg)
     rcode = 1
 
+  logging.info("Master failed over from %s to %s", old_master, new_master)
   return rcode
 
 
@@ -597,7 +710,7 @@ def GatherMasterVotes(node_list):
   @return: list of (node, votes)
 
   """
-  myself = utils.HostInfo().name
+  myself = netutils.Hostname.GetSysName()
   try:
     node_list.remove(myself)
   except ValueError:
@@ -619,6 +732,7 @@ def GatherMasterVotes(node_list):
     if msg:
       logging.warning("Error contacting node %s: %s", node, msg)
       fail = True
+    # for now we accept both length 3 and 4 (data[3] is primary ip version)
     elif not isinstance(data, (tuple, list)) or len(data) < 3:
       logging.warning("Invalid data received from node %s: %s", node, data)
       fail = True