Export extractExTags and updateExclTags
[ganeti-local] / lib / bootstrap.py
index 7bef3fa..1b862c5 100644 (file)
@@ -1,7 +1,7 @@
 #
 #
 
 #
 #
 
-# Copyright (C) 2006, 2007, 2008 Google Inc.
+# Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 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
 #
 # 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
 
 import os
 import os.path
 
 import os
 import os.path
-import sha
 import re
 import logging
 import re
 import logging
+import time
 import tempfile
 
 import tempfile
 
+from ganeti.cmdlib import cluster
 from ganeti import rpc
 from ganeti import ssh
 from ganeti import utils
 from ganeti import rpc
 from ganeti import ssh
 from ganeti import utils
@@ -38,6 +39,21 @@ from ganeti import config
 from ganeti import constants
 from ganeti import objects
 from ganeti import ssconf
 from ganeti import constants
 from ganeti import objects
 from ganeti import ssconf
+from ganeti import serializer
+from ganeti import hypervisor
+from ganeti.storage import drbd
+from ganeti.storage import filestorage
+from ganeti import netutils
+from ganeti import luxi
+from ganeti import jstore
+from ganeti import pathutils
+
+
+# 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():
 
 
 def _InitSSHSetup():
@@ -47,7 +63,7 @@ def _InitSSHSetup():
   permitted hosts and adds the hostkey to its own known hosts.
 
   """
   permitted hosts and adds the hostkey to its own known hosts.
 
   """
-  priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
+  priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
 
   for name in priv_key, pub_key:
     if os.path.exists(name):
 
   for name in priv_key, pub_key:
     if os.path.exists(name):
@@ -61,206 +77,703 @@ def _InitSSHSetup():
     raise errors.OpExecError("Could not generate ssh keypair, error %s" %
                              result.output)
 
     raise errors.OpExecError("Could not generate ssh keypair, error %s" %
                              result.output)
 
-  f = open(pub_key, 'r')
-  try:
-    utils.AddAuthorizedKey(auth_keys, f.read(8192))
-  finally:
-    f.close()
+  utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
 
 
 
 
-def _GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
-  """Generates a self-signed SSL certificate.
+def GenerateHmacKey(file_name):
+  """Writes a new HMAC key.
 
   @type file_name: str
   @param file_name: Path to output file
 
   @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:
-    # 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)
-
+  utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400,
+                  backup=True)
+
+
+def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert,
+                          new_confd_hmac_key, new_cds,
+                          rapi_cert_pem=None, spice_cert_pem=None,
+                          spice_cacert_pem=None, cds=None,
+                          nodecert_file=pathutils.NODED_CERT_FILE,
+                          rapicert_file=pathutils.RAPI_CERT_FILE,
+                          spicecert_file=pathutils.SPICE_CERT_FILE,
+                          spicecacert_file=pathutils.SPICE_CACERT_FILE,
+                          hmackey_file=pathutils.CONFD_HMAC_KEY,
+                          cds_file=pathutils.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_spice_cert: bool
+  @param new_spice_cert: Whether to generate a new SPICE certificate
+  @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 spice_cert_pem: string
+  @param spice_cert_pem: New SPICE certificate in PEM format
+  @type spice_cacert_pem: string
+  @param spice_cacert_pem: Certificate of the CA that signed the SPICE
+                           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 spicecert_file: string
+  @param spicecert_file: optional override of the spice cert file path
+  @type spicecacert_file: string
+  @param spicecacert_file: optional override of the spice CA cert file path
+  @type hmackey_file: string
+  @param hmackey_file: optional override of the hmac key file path
 
 
-def _InitGanetiServerSetup():
+  """
+  # 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(nodecert_file)
+
+    logging.debug("Generating new cluster certificate at %s", nodecert_file)
+    utils.GenerateSelfSignedSslCert(nodecert_file)
+
+  # 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(rapicert_file)
+
+  if rapi_cert_pem:
+    # Assume rapi_pem contains a valid PEM-formatted certificate and key
+    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(rapicert_file)
+
+    logging.debug("Generating new RAPI certificate at %s", rapicert_file)
+    utils.GenerateSelfSignedSslCert(rapicert_file)
+
+  # SPICE
+  spice_cert_exists = os.path.exists(spicecert_file)
+  spice_cacert_exists = os.path.exists(spicecacert_file)
+  if spice_cert_pem:
+    # spice_cert_pem implies also spice_cacert_pem
+    logging.debug("Writing SPICE certificate at %s", spicecert_file)
+    utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True)
+    logging.debug("Writing SPICE CA certificate at %s", spicecacert_file)
+    utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True)
+  elif new_spice_cert or not spice_cert_exists:
+    if spice_cert_exists:
+      utils.CreateBackup(spicecert_file)
+    if spice_cacert_exists:
+      utils.CreateBackup(spicecacert_file)
+
+    logging.debug("Generating new self-signed SPICE certificate at %s",
+                  spicecert_file)
+    (_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file)
+
+    # Self-signed certificate -> the public certificate is also the CA public
+    # certificate
+    logging.debug("Writing the public certificate to %s",
+                  spicecert_file)
+    utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem)
+
+  # Cluster domain secret
+  if cds:
+    logging.debug("Writing cluster domain secret to %s", cds_file)
+    utils.WriteFile(cds_file, data=cds, backup=True)
+
+  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
   """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.
 
 
-  """
-  _GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
+  @type master_name: str
+  @param master_name: Name of the master node
 
 
-  # Don't overwrite existing file
-  if not os.path.exists(constants.RAPI_CERT_FILE):
-    _GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
-
-  result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
+  """
+  # Generate cluster secrets
+  GenerateClusterCrypto(True, False, False, False, False)
 
 
+  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start", constants.NODED])
   if result.failed:
     raise errors.OpExecError("Could not start the node daemon, command %s"
                              " had exitcode %s and error %s" %
                              (result.cmd, result.exit_code, result.output))
 
   if result.failed:
     raise errors.OpExecError("Could not start the node daemon, command %s"
                              " had exitcode %s and error %s" %
                              (result.cmd, result.exit_code, result.output))
 
+  _WaitForNodeDaemon(master_name)
+
+
+def _WaitForNodeDaemon(node_name):
+  """Wait for node daemon to become responsive.
+
+  """
+  def _CheckNodeDaemon():
+    # Pylint bug <http://www.logilab.org/ticket/35642>
+    # pylint: disable=E1101
+    result = rpc.BootstrapRunner().call_version([node_name])[node_name]
+    if result.fail_msg:
+      raise utils.RetryAgain()
+
+  try:
+    utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT)
+  except utils.RetryTimeout:
+    raise errors.OpExecError("Node daemon on %s didn't answer queries within"
+                             " %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 _WaitForSshDaemon(hostname, port, family):
+  """Wait for SSH daemon to become responsive.
+
+  """
+  hostip = netutils.GetHostname(name=hostname, family=family).ip
+
+  def _CheckSshDaemon():
+    if netutils.TcpPing(hostip, port, timeout=1.0, live_port_needed=True):
+      logging.debug("SSH daemon on %s:%s (IP address %s) has become"
+                    " responsive", hostname, port, hostip)
+    else:
+      raise utils.RetryAgain()
+
+  try:
+    utils.Retry(_CheckSshDaemon, 1.0, _DAEMON_READY_TIMEOUT)
+  except utils.RetryTimeout:
+    raise errors.OpExecError("SSH daemon on %s:%s (IP address %s) didn't"
+                             " become responsive within %s seconds" %
+                             (hostname, port, hostip, _DAEMON_READY_TIMEOUT))
+
+
+def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose,
+                    use_cluster_key, ask_key, strict_host_check, data):
+  """Runs a command to configure something on a remote machine.
+
+  @type cluster_name: string
+  @param cluster_name: Cluster name
+  @type node: string
+  @param node: Node name
+  @type basecmd: string
+  @param basecmd: Base command (path on the remote machine)
+  @type debug: bool
+  @param debug: Enable debug output
+  @type verbose: bool
+  @param verbose: Enable verbose output
+  @type use_cluster_key: bool
+  @param use_cluster_key: See L{ssh.SshRunner.BuildCmd}
+  @type ask_key: bool
+  @param ask_key: See L{ssh.SshRunner.BuildCmd}
+  @type strict_host_check: bool
+  @param strict_host_check: See L{ssh.SshRunner.BuildCmd}
+  @param data: JSON-serializable input data for script (passed to stdin)
+
+  """
+  cmd = [basecmd]
+
+  # Pass --debug/--verbose to the external script if set on our invocation
+  if debug:
+    cmd.append("--debug")
+
+  if verbose:
+    cmd.append("--verbose")
+
+  family = ssconf.SimpleStore().GetPrimaryIPFamily()
+  srun = ssh.SshRunner(cluster_name,
+                       ipv6=(family == netutils.IP6Address.family))
+  scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER,
+                       utils.ShellQuoteArgs(cmd),
+                       batch=False, ask_key=ask_key, quiet=False,
+                       strict_host_check=strict_host_check,
+                       use_cluster_key=use_cluster_key)
+
+  tempfh = tempfile.TemporaryFile()
+  try:
+    tempfh.write(serializer.DumpJson(data))
+    tempfh.seek(0)
+
+    result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh)
+  finally:
+    tempfh.close()
+
+  if result.failed:
+    raise errors.OpExecError("Command '%s' failed: %s" %
+                             (result.cmd, result.fail_reason))
+
+  _WaitForSshDaemon(node, netutils.GetDaemonPort(constants.SSH), family)
+
+
+def _InitFileStorageDir(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
+
+  """
+  file_storage_dir = os.path.normpath(file_storage_dir)
+
+  if not os.path.isabs(file_storage_dir):
+    raise errors.OpPrereqError("File storage directory '%s' is not an absolute"
+                               " path" % file_storage_dir, 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 _PrepareFileBasedStorage(
+    enabled_disk_templates, file_storage_dir,
+    default_dir, file_disk_template,
+    init_fn=_InitFileStorageDir, acceptance_fn=None):
+  """Checks if a file-base storage type is enabled and inits the dir.
+
+  @type enabled_disk_templates: list of string
+  @param enabled_disk_templates: list of enabled disk templates
+  @type file_storage_dir: string
+  @param file_storage_dir: the file storage directory
+  @type default_dir: string
+  @param default_dir: default file storage directory when C{file_storage_dir}
+      is 'None'
+  @type file_disk_template: string
+  @param file_disk_template: a disk template whose storage type is 'ST_FILE'
+  @rtype: string
+  @returns: the name of the actual file storage directory
+
+  """
+  assert (file_disk_template in
+          utils.storage.GetDiskTemplatesOfStorageType(constants.ST_FILE))
+  if file_storage_dir is None:
+    file_storage_dir = default_dir
+  if not acceptance_fn:
+    acceptance_fn = \
+        lambda path: filestorage.CheckFileStoragePathAcceptance(
+            path, exact_match_ok=True)
+
+  cluster.CheckFileStoragePathVsEnabledDiskTemplates(
+      logging.warning, file_storage_dir, enabled_disk_templates)
+
+  file_storage_enabled = file_disk_template in enabled_disk_templates
+  if file_storage_enabled:
+    try:
+      acceptance_fn(file_storage_dir)
+    except errors.FileStoragePathError as e:
+      raise errors.OpPrereqError(str(e))
+    result_file_storage_dir = init_fn(file_storage_dir)
+  else:
+    result_file_storage_dir = file_storage_dir
+  return result_file_storage_dir
+
+
+def _PrepareFileStorage(
+    enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir,
+    acceptance_fn=None):
+  """Checks if file storage is enabled and inits the dir.
+
+  @see: C{_PrepareFileBasedStorage}
+
+  """
+  return _PrepareFileBasedStorage(
+      enabled_disk_templates, file_storage_dir,
+      pathutils.DEFAULT_FILE_STORAGE_DIR, constants.DT_FILE,
+      init_fn=init_fn, acceptance_fn=acceptance_fn)
+
+
+def _PrepareSharedFileStorage(
+    enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir,
+    acceptance_fn=None):
+  """Checks if shared file storage is enabled and inits the dir.
+
+  @see: C{_PrepareFileBasedStorage}
 
 
-def InitCluster(cluster_name, mac_prefix, def_bridge,
-                master_netdev, file_storage_dir, candidate_pool_size,
-                secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
-                enabled_hypervisors=None, default_hypervisor=None):
+  """
+  return _PrepareFileBasedStorage(
+      enabled_disk_templates, file_storage_dir,
+      pathutils.DEFAULT_SHARED_FILE_STORAGE_DIR, constants.DT_SHARED_FILE,
+      init_fn=init_fn, acceptance_fn=acceptance_fn)
+
+
+def _InitCheckEnabledDiskTemplates(enabled_disk_templates):
+  """Checks the sanity of the enabled disk templates.
+
+  """
+  if not enabled_disk_templates:
+    raise errors.OpPrereqError("Enabled disk templates list must contain at"
+                               " least one member", errors.ECODE_INVAL)
+  invalid_disk_templates = \
+    set(enabled_disk_templates) - constants.DISK_TEMPLATES
+  if invalid_disk_templates:
+    raise errors.OpPrereqError("Enabled disk templates list contains invalid"
+                               " entries: %s" % invalid_disk_templates,
+                               errors.ECODE_INVAL)
+
+
+def _RestrictIpolicyToEnabledDiskTemplates(ipolicy, enabled_disk_templates):
+  """Restricts the ipolicy's disk templates to the enabled ones.
+
+  This function clears the ipolicy's list of allowed disk templates from the
+  ones that are not enabled by the cluster.
+
+  @type ipolicy: dict
+  @param ipolicy: the instance policy
+  @type enabled_disk_templates: list of string
+  @param enabled_disk_templates: the list of cluster-wide enabled disk
+    templates
+
+  """
+  assert constants.IPOLICY_DTS in ipolicy
+  allowed_disk_templates = ipolicy[constants.IPOLICY_DTS]
+  restricted_disk_templates = list(set(allowed_disk_templates)
+                                   .intersection(set(enabled_disk_templates)))
+  ipolicy[constants.IPOLICY_DTS] = restricted_disk_templates
+
+
+def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914
+                master_netmask, master_netdev, file_storage_dir,
+                shared_file_storage_dir, candidate_pool_size, secondary_ip=None,
+                vg_name=None, beparams=None, nicparams=None, ndparams=None,
+                hvparams=None, diskparams=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, ipolicy=None,
+                prealloc_wipe_disks=False, use_external_mip_script=False,
+                hv_state=None, disk_state=None, enabled_disk_templates=None):
   """Initialise the cluster.
 
   @type candidate_pool_size: int
   @param candidate_pool_size: master candidate pool size
   """Initialise the cluster.
 
   @type candidate_pool_size: int
   @param candidate_pool_size: master candidate pool size
+  @type enabled_disk_templates: list of string
+  @param enabled_disk_templates: list of disk_templates to be used in this
+    cluster
 
   """
   # TODO: complete the docstring
   if config.ConfigWriter.IsCluster():
 
   """
   # TODO: complete the docstring
   if config.ConfigWriter.IsCluster():
-    raise errors.OpPrereqError("Cluster is already initialised")
+    raise errors.OpPrereqError("Cluster is already initialised",
+                               errors.ECODE_STATE)
 
 
-  hostname = utils.HostInfo()
+  if not enabled_hypervisors:
+    raise errors.OpPrereqError("Enabled hypervisors list must contain at"
+                               " least one member", errors.ECODE_INVAL)
+  invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
+  if invalid_hvs:
+    raise errors.OpPrereqError("Enabled hypervisors contains invalid"
+                               " entries: %s" % invalid_hvs,
+                               errors.ECODE_INVAL)
 
 
-  if hostname.ip.startswith("127."):
-    raise errors.OpPrereqError("This host's IP resolves to the private"
-                               " range (%s). Please fix DNS or %s." %
-                               (hostname.ip, constants.ETC_HOSTS))
+  _InitCheckEnabledDiskTemplates(enabled_disk_templates)
 
 
-  if not utils.OwnIpAddress(hostname.ip):
+  try:
+    ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version)
+  except errors.ProgrammerError:
+    raise errors.OpPrereqError("Invalid primary ip version: %d." %
+                               primary_ip_version, errors.ECODE_INVAL)
+
+  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),
+                               errors.ECODE_INVAL)
+
+  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, pathutils.ETC_HOSTS),
+                               errors.ECODE_ENVIRON)
+
+  if not ipcls.Own(hostname.ip):
     raise errors.OpPrereqError("Inconsistency: this host's name resolves"
                                " to %s,\nbut this ip address does not"
     raise errors.OpPrereqError("Inconsistency: this host's name resolves"
                                " to %s,\nbut this ip address does not"
-                               " belong to this host."
-                               " Aborting." % hostname.ip)
-
-  clustername = utils.HostInfo(cluster_name)
-
-  if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
-                   timeout=5):
-    raise errors.OpPrereqError("Cluster IP already active. Aborting.")
-
-  if secondary_ip:
-    if not utils.IsValidIP(secondary_ip):
-      raise errors.OpPrereqError("Invalid secondary ip given")
-    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)
-  else:
+                               " belong to this host" %
+                               hostname.ip, errors.ECODE_ENVIRON)
+
+  clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family)
+
+  if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5):
+    raise errors.OpPrereqError("Cluster IP already active",
+                               errors.ECODE_NOTUNIQUE)
+
+  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)
     secondary_ip = hostname.ip
 
     secondary_ip = hostname.ip
 
-  if vg_name is not None:
+  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 master_netmask is not None:
+    if not ipcls.ValidateNetmask(master_netmask):
+      raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " %
+                                  (master_netmask, primary_ip_version),
+                                 errors.ECODE_INVAL)
+  else:
+    master_netmask = ipcls.iplen
+
+  if vg_name:
     # Check if volume group is valid
     vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
                                           constants.MIN_VG_SIZE)
     if vgstatus:
     # Check if volume group is valid
     vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
                                           constants.MIN_VG_SIZE)
     if vgstatus:
-      raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
-                                 " you are not using lvm" % vgstatus)
+      raise errors.OpPrereqError("Error: %s" % 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.")
-
-  if not os.path.exists(file_storage_dir):
+  if drbd_helper is not None:
     try:
     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 = drbd.DRBD8.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)
+
+  logging.debug("Stopping daemons (if any are running)")
+  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"])
+  if result.failed:
+    raise errors.OpExecError("Could not stop daemons, command %s"
+                             " had exitcode %s and error '%s'" %
+                             (result.cmd, result.exit_code, result.output))
 
 
-  if not os.path.isdir(file_storage_dir):
-    raise errors.OpPrereqError("The file storage directory '%s' is not"
-                               " a directory." % file_storage_dir)
+  file_storage_dir = _PrepareFileStorage(enabled_disk_templates,
+                                         file_storage_dir)
+  shared_file_storage_dir = _PrepareSharedFileStorage(enabled_disk_templates,
+                                                      shared_file_storage_dir)
 
   if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
 
   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)
+    raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
+                               errors.ECODE_INVAL)
 
   result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
   if result.failed:
     raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
                                (master_netdev,
 
   result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
   if result.failed:
     raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
                                (master_netdev,
-                                result.output.strip()))
+                                result.output.strip()), errors.ECODE_INVAL)
 
 
-  if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
-          os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
-    raise errors.OpPrereqError("Init.d script '%s' missing or not"
-                               " executable." % constants.NODE_INITD_SCRIPT)
+  dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)]
+  utils.EnsureDirs(dirs)
 
 
-  utils.CheckBEParams(beparams)
+  objects.UpgradeBeParams(beparams)
+  utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
+  utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
 
 
-  # set up the inter-node password and certificate
-  _InitGanetiServerSetup()
+  objects.NIC.CheckParameterSyntax(nicparams)
+
+  full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy)
+  _RestrictIpolicyToEnabledDiskTemplates(full_ipolicy, enabled_disk_templates)
+
+  if ndparams is not None:
+    utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES)
+  else:
+    ndparams = dict(constants.NDC_DEFAULTS)
+
+  # This is ugly, as we modify the dict itself
+  # FIXME: Make utils.ForceDictType pure functional or write a wrapper
+  # around it
+  if hv_state:
+    for hvname, hvs_data in hv_state.items():
+      utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES)
+      hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data)
+  else:
+    hv_state = dict((hvname, constants.HVST_DEFAULTS)
+                    for hvname in enabled_hypervisors)
+
+  # FIXME: disk_state has no default values yet
+  if disk_state:
+    for storage, ds_data in disk_state.items():
+      if storage not in constants.DS_VALID_TYPES:
+        raise errors.OpPrereqError("Invalid storage type in disk state: %s" %
+                                   storage, errors.ECODE_INVAL)
+      for ds_name, state in ds_data.items():
+        utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES)
+        ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state)
+
+  # 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)
+
+  # diskparams is a mapping of disk-template->diskparams dict
+  for template, dt_params in diskparams.items():
+    param_keys = set(dt_params.keys())
+    default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys())
+    if not (param_keys <= default_param_keys):
+      unknown_params = param_keys - default_param_keys
+      raise errors.OpPrereqError("Invalid parameters for disk template %s:"
+                                 " %s" % (template,
+                                          utils.CommaJoin(unknown_params)),
+                                 errors.ECODE_INVAL)
+    utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
+    if template == constants.DT_DRBD8 and vg_name is not None:
+      # The default METAVG value is equal to the VG name set at init time,
+      # if provided
+      dt_params[constants.DRBD_DEFAULT_METAVG] = vg_name
 
 
-  # set up ssh config and /etc/hosts
-  f = open(constants.SSH_HOST_RSA_PUB, 'r')
   try:
   try:
-    sshline = f.read()
-  finally:
-    f.close()
-  sshkey = sshline.split(" ")[1]
+    utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS)
+  except errors.OpPrereqError, err:
+    raise errors.OpPrereqError("While verify diskparam options: %s" % err,
+                               errors.ECODE_INVAL)
 
 
-  utils.AddHostToEtcHosts(hostname.name)
-  _InitSSHSetup()
+  # set up ssh config and /etc/hosts
+  rsa_sshkey = ""
+  dsa_sshkey = ""
+  if os.path.isfile(pathutils.SSH_HOST_RSA_PUB):
+    sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB)
+    rsa_sshkey = sshline.split(" ")[1]
+  if os.path.isfile(pathutils.SSH_HOST_DSA_PUB):
+    sshline = utils.ReadFile(pathutils.SSH_HOST_DSA_PUB)
+    dsa_sshkey = sshline.split(" ")[1]
+  if not rsa_sshkey and not dsa_sshkey:
+    raise errors.OpPrereqError("Failed to find SSH public keys",
+                               errors.ECODE_ENVIRON)
+
+  if modify_etc_hosts:
+    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)
+  elif constants.HTOOLS:
+    # htools was enabled at build-time, we default to it
+    if utils.FindFile(constants.IALLOC_HAIL,
+                      constants.IALLOCATOR_SEARCH_PATH,
+                      os.path.isfile):
+      default_iallocator = constants.IALLOC_HAIL
+
+  now = time.time()
 
   # init of cluster config file
   cluster_config = objects.Cluster(
     serial_no=1,
 
   # init of cluster config file
   cluster_config = objects.Cluster(
     serial_no=1,
-    rsahostkeypub=sshkey,
+    rsahostkeypub=rsa_sshkey,
+    dsahostkeypub=dsa_sshkey,
     highest_used_port=(constants.FIRST_DRBD_PORT - 1),
     mac_prefix=mac_prefix,
     volume_group_name=vg_name,
     highest_used_port=(constants.FIRST_DRBD_PORT - 1),
     mac_prefix=mac_prefix,
     volume_group_name=vg_name,
-    default_bridge=def_bridge,
     tcpudp_port_pool=set(),
     tcpudp_port_pool=set(),
-    master_node=hostname.name,
     master_ip=clustername.ip,
     master_ip=clustername.ip,
+    master_netmask=master_netmask,
     master_netdev=master_netdev,
     cluster_name=clustername.name,
     file_storage_dir=file_storage_dir,
     master_netdev=master_netdev,
     cluster_name=clustername.name,
     file_storage_dir=file_storage_dir,
+    shared_file_storage_dir=shared_file_storage_dir,
     enabled_hypervisors=enabled_hypervisors,
     enabled_hypervisors=enabled_hypervisors,
-    default_hypervisor=default_hypervisor,
-    beparams={constants.BEGR_DEFAULT: beparams},
+    beparams={constants.PP_DEFAULT: beparams},
+    nicparams={constants.PP_DEFAULT: nicparams},
+    ndparams=ndparams,
     hvparams=hvparams,
     hvparams=hvparams,
+    diskparams=diskparams,
     candidate_pool_size=candidate_pool_size,
     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,
+    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,
+    use_external_mip_script=use_external_mip_script,
+    ipolicy=full_ipolicy,
+    hv_state_static=hv_state,
+    disk_state_static=disk_state,
+    enabled_disk_templates=enabled_disk_templates,
     )
   master_node_config = objects.Node(name=hostname.name,
                                     primary_ip=hostname.ip,
                                     secondary_ip=secondary_ip,
                                     serial_no=1,
                                     master_candidate=True,
     )
   master_node_config = objects.Node(name=hostname.name,
                                     primary_ip=hostname.ip,
                                     secondary_ip=secondary_ip,
                                     serial_no=1,
                                     master_candidate=True,
-                                    offline=False,
+                                    offline=False, drained=False,
+                                    ctime=now, mtime=now,
                                     )
                                     )
+  InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
+  cfg = config.ConfigWriter(offline=True)
+  ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE)
+  cfg.Update(cfg.GetClusterInfo(), logging.error)
+  ssconf.WriteSsconfFiles(cfg.GetSsconfValues())
 
 
-  sscfg = InitConfig(constants.CONFIG_VERSION,
-                     cluster_config, master_node_config)
-  ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
-  cfg = config.ConfigWriter()
-  cfg.Update(cfg.GetClusterInfo())
+  # set up the inter-node password and certificate
+  _InitGanetiServerSetup(hostname.name)
 
 
-  # start the master ip
-  # TODO: Review rpc call from bootstrap
-  rpc.RpcRunner.call_node_start_master(hostname.name, True)
+  logging.debug("Starting daemons")
+  result = utils.RunCmd([pathutils.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,
 
 
 def InitConfig(version, cluster_config, master_node_config,
-               cfg_file=constants.CLUSTER_CONF_FILE):
+               cfg_file=pathutils.CLUSTER_CONF_FILE):
   """Create the initial cluster configuration.
 
   It will contain the current node, which will also be the master
   """Create the initial cluster configuration.
 
   It will contain the current node, which will also be the master
@@ -275,41 +788,75 @@ def InitConfig(version, cluster_config, master_node_config,
   @type cfg_file: string
   @param cfg_file: configuration file path
 
   @type cfg_file: string
   @param cfg_file: configuration file path
 
-  @rtype: L{ssconf.SimpleConfigWriter}
-  @returns: initialized config instance
-
   """
   """
+  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)
+  cluster_config.master_node = master_node_config.uuid
   nodes = {
   nodes = {
-    master_node_config.name: master_node_config,
+    master_node_config.uuid: 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.uuid],
+    diskparams={},
+    )
+  nodegroups = {
+    default_nodegroup.uuid: default_nodegroup,
+    }
+  now = time.time()
   config_data = objects.ConfigData(version=version,
                                    cluster=cluster_config,
   config_data = objects.ConfigData(version=version,
                                    cluster=cluster_config,
+                                   nodegroups=nodegroups,
                                    nodes=nodes,
                                    instances={},
                                    nodes=nodes,
                                    instances={},
-                                   serial_no=1)
-  cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
-  cfg.Save()
-
-  return cfg
+                                   networks={},
+                                   serial_no=1,
+                                   ctime=now, mtime=now)
+  utils.WriteFile(cfg_file,
+                  data=serializer.Dump(config_data.ToDict()),
+                  mode=0600)
 
 
 
 
-def FinalizeClusterDestroy(master):
+def FinalizeClusterDestroy(master_uuid):
   """Execute the last steps of cluster destroy
 
   This function shuts down all the daemons, completing the destroy
   begun in cmdlib.LUDestroyOpcode.
 
   """
   """Execute the last steps of cluster destroy
 
   This function shuts down all the daemons, completing the destroy
   begun in cmdlib.LUDestroyOpcode.
 
   """
-  result = rpc.RpcRunner.call_node_stop_master(master, True)
-  if result.failed or not result.data:
-    logging.warning("Could not disable the master role")
-  result = rpc.RpcRunner.call_node_leave_cluster(master)
-  if result.failed or not result.data:
-    logging.warning("Could not shutdown the node daemon and cleanup the node")
+  cfg = config.ConfigWriter()
+  modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
+  runner = rpc.BootstrapRunner()
+
+  master_name = cfg.GetNodeName(master_uuid)
+
+  master_params = cfg.GetMasterNetworkParameters()
+  master_params.uuid = master_uuid
+  ems = cfg.GetUseExternalMipScript()
+  result = runner.call_node_deactivate_master_ip(master_name, master_params,
+                                                 ems)
+
+  msg = result.fail_msg
+  if msg:
+    logging.warning("Could not disable the master IP: %s", msg)
+
+  result = runner.call_node_stop_master(master_name)
+  msg = result.fail_msg
+  if msg:
+    logging.warning("Could not disable the master role: %s", msg)
+
+  result = runner.call_node_leave_cluster(master_name, modify_ssh_setup)
+  msg = result.fail_msg
+  if msg:
+    logging.warning("Could not shutdown the node daemon and cleanup"
+                    " the node: %s", msg)
 
 
 
 
-def SetupNodeDaemon(cluster_name, node, ssh_key_check):
+def SetupNodeDaemon(opts, cluster_name, node):
   """Add a node to the cluster.
 
   This function must be called before the actual opcode, and will ssh
   """Add a node to the cluster.
 
   This function must be called before the actual opcode, and will ssh
@@ -318,59 +865,46 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check):
 
   @param cluster_name: the cluster name
   @param node: the name of the new node
 
   @param cluster_name: the cluster name
   @param node: the name of the new node
-  @param ssh_key_check: whether to do a strict key check
 
   """
 
   """
-  sshrunner = ssh.SshRunner(cluster_name)
-  gntpem = utils.ReadFile(constants.SSL_CERT_FILE)
-  # 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
-  if re.search('^!EOF\.', gntpem, re.MULTILINE):
-    raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
-  if not gntpem.endswith("\n"):
-    raise errors.OpExecError("PEM must end with newline")
-
-  # 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%s restart" %
-               (constants.SSL_CERT_FILE, gntpem,
-                constants.NODE_INITD_SCRIPT))
-
-  result = sshrunner.Run(node, 'root', mycommand, batch=False,
-                         ask_key=ssh_key_check,
-                         use_cluster_key=False,
-                         strict_host_check=ssh_key_check)
-  if result.failed:
-    raise errors.OpExecError("Remote command on node %s, error: %s,"
-                             " output: %s" %
-                             (node, result.fail_reason, result.output))
+  data = {
+    constants.NDS_CLUSTER_NAME: cluster_name,
+    constants.NDS_NODE_DAEMON_CERTIFICATE:
+      utils.ReadFile(pathutils.NODED_CERT_FILE),
+    constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(),
+    constants.NDS_START_NODE_DAEMON: True,
+    }
+
+  RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP,
+                  opts.debug, opts.verbose,
+                  True, opts.ssh_key_check, opts.ssh_key_check, data)
 
 
+  _WaitForNodeDaemon(node)
 
 
-def MasterFailover():
+
+def MasterFailover(no_voting=False):
   """Failover the master node.
 
   This checks that we are not already the master, and will cause the
   current master to cease being master, and the non-master to become
   new master.
 
   """Failover the master node.
 
   This checks that we are not already the master, and will cause the
   current master to cease being master, and the non-master to become
   new master.
 
+  @type no_voting: boolean
+  @param no_voting: force the operation without remote nodes agreement
+                      (dangerous)
+
   """
   sstore = ssconf.SimpleStore()
 
   old_master, new_master = ssconf.GetMasterAndMyself(sstore)
   """
   sstore = ssconf.SimpleStore()
 
   old_master, new_master = ssconf.GetMasterAndMyself(sstore)
-  node_list = sstore.GetNodeList()
+  node_names = sstore.GetNodeList()
   mc_list = sstore.GetMasterCandidates()
 
   if old_master == new_master:
     raise errors.OpPrereqError("This commands must be run on the node"
                                " where you want the new master to be."
                                " %s is already the master" %
   mc_list = sstore.GetMasterCandidates()
 
   if old_master == new_master:
     raise errors.OpPrereqError("This commands must be run on the node"
                                " where you want the new master to be."
                                " %s is already the master" %
-                               old_master)
+                               old_master, errors.ECODE_INVAL)
 
   if new_master not in mc_list:
     mc_no_master = [name for name in mc_list if name != old_master]
 
   if new_master not in mc_list:
     mc_no_master = [name for name in mc_list if name != old_master]
@@ -378,53 +912,133 @@ def MasterFailover():
                                " as master candidates. Only these nodes"
                                " can become masters. Current list of"
                                " master candidates is:\n"
                                " as master candidates. Only these nodes"
                                " can become masters. Current list of"
                                " master candidates is:\n"
-                               "%s" % ('\n'.join(mc_no_master)))
-
-  vote_list = GatherMasterVotes(node_list)
-
-  if vote_list:
-    voted_master = vote_list[0][0]
-    if voted_master is None:
-      raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
-                                 " respond.")
-    elif voted_master != old_master:
-      raise errors.OpPrereqError("I have wrong configuration, I believe the"
-                                 " master is %s but the other nodes voted for"
-                                 " %s. Please resync the configuration of"
-                                 " this node." % (old_master, voted_master))
+                               "%s" % ("\n".join(mc_no_master)),
+                               errors.ECODE_STATE)
+
+  if not no_voting:
+    vote_list = GatherMasterVotes(node_names)
+
+    if vote_list:
+      voted_master = vote_list[0][0]
+      if voted_master is None:
+        raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
+                                   " not respond.", errors.ECODE_ENVIRON)
+      elif voted_master != old_master:
+        raise errors.OpPrereqError("I have a wrong configuration, I believe"
+                                   " the master is %s but the other nodes"
+                                   " voted %s. Please resync the configuration"
+                                   " of this node." %
+                                   (old_master, voted_master),
+                                   errors.ECODE_STATE)
   # end checks
 
   rcode = 0
 
   logging.info("Setting master to %s, old master: %s", new_master, old_master)
 
   # end checks
 
   rcode = 0
 
   logging.info("Setting master to %s, old master: %s", new_master, old_master)
 
-  result = rpc.RpcRunner.call_node_stop_master(old_master, True)
-  if result.failed or not result.data:
+  try:
+    # instantiate a real config writer, as we now know we have the
+    # configuration data
+    cfg = config.ConfigWriter(accept_foreign=True)
+
+    old_master_node = cfg.GetNodeInfoByName(old_master)
+    if old_master_node is None:
+      raise errors.OpPrereqError("Could not find old master node '%s' in"
+                                 " cluster configuration." % old_master,
+                                 errors.ECODE_NOENT)
+
+    cluster_info = cfg.GetClusterInfo()
+    new_master_node = cfg.GetNodeInfoByName(new_master)
+    if new_master_node is None:
+      raise errors.OpPrereqError("Could not find new master node '%s' in"
+                                 " cluster configuration." % new_master,
+                                 errors.ECODE_NOENT)
+
+    cluster_info.master_node = new_master_node.uuid
+    # 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)
+
+  runner = rpc.BootstrapRunner()
+  master_params = cfg.GetMasterNetworkParameters()
+  master_params.uuid = old_master_node.uuid
+  ems = cfg.GetUseExternalMipScript()
+  result = runner.call_node_deactivate_master_ip(old_master,
+                                                 master_params, ems)
+
+  msg = result.fail_msg
+  if msg:
+    logging.warning("Could not disable the master IP: %s", msg)
+
+  result = runner.call_node_stop_master(old_master)
+  msg = result.fail_msg
+  if msg:
     logging.error("Could not disable the master role on the old master"
     logging.error("Could not disable the master role on the old master"
-                 " %s, please disable manually", old_master)
+                  " %s, please disable manually: %s", old_master, msg)
 
 
-  # Here we have a phase where no master should be running
+  logging.info("Checking master IP non-reachability...")
 
 
-  # instantiate a real config writer, as we now know we have the
-  # configuration data
-  cfg = config.ConfigWriter()
+  master_ip = sstore.GetMasterIP()
+  total_timeout = 30
 
 
-  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)
+  # 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()
 
 
-  result = rpc.RpcRunner.call_node_start_master(new_master, True)
-  if result.failed or not result.data:
+  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)
+
+  if jstore.CheckDrainFlag():
+    logging.info("Undraining job queue")
+    jstore.SetDrainFlag(False)
+
+  logging.info("Starting the master daemons on the new master")
+
+  result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master,
+                                                                no_voting)
+  msg = result.fail_msg
+  if msg:
     logging.error("Could not start the master role on the new master"
     logging.error("Could not start the master role on the new master"
-                  " %s, please check", new_master)
+                  " %s, please check: %s", new_master, msg)
     rcode = 1
 
     rcode = 1
 
+  logging.info("Master failed over from %s to %s", old_master, new_master)
   return rcode
 
 
   return rcode
 
 
-def GatherMasterVotes(node_list):
+def GetMaster():
+  """Returns the current master node.
+
+  This is a separate function in bootstrap since it's needed by
+  gnt-cluster, and instead of importing directly ssconf, it's better
+  to abstract it in bootstrap, where we do use ssconf in other
+  functions too.
+
+  """
+  sstore = ssconf.SimpleStore()
+
+  old_master, _ = ssconf.GetMasterAndMyself(sstore)
+
+  return old_master
+
+
+def GatherMasterVotes(node_names):
   """Check the agreement on who is the master.
 
   This function will return a list of (node, number of votes), ordered
   """Check the agreement on who is the master.
 
   This function will return a list of (node, number of votes), ordered
@@ -438,32 +1052,42 @@ def GatherMasterVotes(node_list):
   since we use the same source for configuration information for both
   backend and boostrap, we'll always vote for ourselves.
 
   since we use the same source for configuration information for both
   backend and boostrap, we'll always vote for ourselves.
 
-  @type node_list: list
-  @param node_list: the list of nodes to query for master info; the current
-      node wil be removed if it is in the list
+  @type node_names: list
+  @param node_names: the list of nodes to query for master info; the current
+      node will be removed if it is in the list
   @rtype: list
   @return: list of (node, votes)
 
   """
   @rtype: list
   @return: list of (node, votes)
 
   """
-  myself = utils.HostInfo().name
+  myself = netutils.Hostname.GetSysName()
   try:
   try:
-    node_list.remove(myself)
+    node_names.remove(myself)
   except ValueError:
     pass
   except ValueError:
     pass
-  if not node_list:
+  if not node_names:
     # no nodes left (eventually after removing myself)
     return []
     # no nodes left (eventually after removing myself)
     return []
-  results = rpc.RpcRunner.call_master_info(node_list)
+  results = rpc.BootstrapRunner().call_master_info(node_names)
   if not isinstance(results, dict):
     # this should not happen (unless internal error in rpc)
     logging.critical("Can't complete rpc call, aborting master startup")
   if not isinstance(results, dict):
     # this should not happen (unless internal error in rpc)
     logging.critical("Can't complete rpc call, aborting master startup")
-    return [(None, len(node_list))]
+    return [(None, len(node_names))]
   votes = {}
   votes = {}
-  for node in results:
-    nres = results[node]
-    data = nres.data
-    if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
-      # here the rpc layer should have already logged errors
+  for node_name in results:
+    nres = results[node_name]
+    data = nres.payload
+    msg = nres.fail_msg
+    fail = False
+    if msg:
+      logging.warning("Error contacting node %s: %s", node_name, msg)
+      fail = True
+    # for now we accept both length 3, 4 and 5 (data[3] is primary ip version
+    # and data[4] is the master netmask)
+    elif not isinstance(data, (tuple, list)) or len(data) < 3:
+      logging.warning("Invalid data received from node %s: %s",
+                      node_name, data)
+      fail = True
+    if fail:
       if None not in votes:
         votes[None] = 0
       votes[None] += 1
       if None not in votes:
         votes[None] = 0
       votes[None] += 1