4 # Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Functions to bootstrap a new cluster.
33 from ganeti import rpc
34 from ganeti import ssh
35 from ganeti import utils
36 from ganeti import errors
37 from ganeti import config
38 from ganeti import constants
39 from ganeti import objects
40 from ganeti import ssconf
41 from ganeti import serializer
42 from ganeti import hypervisor
43 from ganeti import bdev
44 from ganeti import netutils
45 from ganeti import luxi
46 from ganeti import jstore
47 from ganeti import pathutils
50 # ec_id for InitConfig's temporary reservation manager
51 _INITCONF_ECID = "initconfig-ecid"
53 #: After how many seconds daemon must be responsive
54 _DAEMON_READY_TIMEOUT = 10.0
58 """Setup the SSH configuration for the cluster.
60 This generates a dsa keypair for root, adds the pub key to the
61 permitted hosts and adds the hostkey to its own known hosts.
64 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
66 for name in priv_key, pub_key:
67 if os.path.exists(name):
68 utils.CreateBackup(name)
69 utils.RemoveFile(name)
71 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
75 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
78 utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
81 def GenerateHmacKey(file_name):
82 """Writes a new HMAC key.
85 @param file_name: Path to output file
88 utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400,
92 def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert,
93 new_confd_hmac_key, new_cds,
94 rapi_cert_pem=None, spice_cert_pem=None,
95 spice_cacert_pem=None, cds=None,
96 nodecert_file=pathutils.NODED_CERT_FILE,
97 rapicert_file=pathutils.RAPI_CERT_FILE,
98 spicecert_file=pathutils.SPICE_CERT_FILE,
99 spicecacert_file=pathutils.SPICE_CACERT_FILE,
100 hmackey_file=pathutils.CONFD_HMAC_KEY,
101 cds_file=pathutils.CLUSTER_DOMAIN_SECRET_FILE):
102 """Updates the cluster certificates, keys and secrets.
104 @type new_cluster_cert: bool
105 @param new_cluster_cert: Whether to generate a new cluster certificate
106 @type new_rapi_cert: bool
107 @param new_rapi_cert: Whether to generate a new RAPI certificate
108 @type new_spice_cert: bool
109 @param new_spice_cert: Whether to generate a new SPICE certificate
110 @type new_confd_hmac_key: bool
111 @param new_confd_hmac_key: Whether to generate a new HMAC key
113 @param new_cds: Whether to generate a new cluster domain secret
114 @type rapi_cert_pem: string
115 @param rapi_cert_pem: New RAPI certificate in PEM format
116 @type spice_cert_pem: string
117 @param spice_cert_pem: New SPICE certificate in PEM format
118 @type spice_cacert_pem: string
119 @param spice_cacert_pem: Certificate of the CA that signed the SPICE
120 certificate, in PEM format
122 @param cds: New cluster domain secret
123 @type nodecert_file: string
124 @param nodecert_file: optional override of the node cert file path
125 @type rapicert_file: string
126 @param rapicert_file: optional override of the rapi cert file path
127 @type spicecert_file: string
128 @param spicecert_file: optional override of the spice cert file path
129 @type spicecacert_file: string
130 @param spicecacert_file: optional override of the spice CA cert file path
131 @type hmackey_file: string
132 @param hmackey_file: optional override of the hmac key file path
135 # noded SSL certificate
136 cluster_cert_exists = os.path.exists(nodecert_file)
137 if new_cluster_cert or not cluster_cert_exists:
138 if cluster_cert_exists:
139 utils.CreateBackup(nodecert_file)
141 logging.debug("Generating new cluster certificate at %s", nodecert_file)
142 utils.GenerateSelfSignedSslCert(nodecert_file)
145 if new_confd_hmac_key or not os.path.exists(hmackey_file):
146 logging.debug("Writing new confd HMAC key to %s", hmackey_file)
147 GenerateHmacKey(hmackey_file)
150 rapi_cert_exists = os.path.exists(rapicert_file)
153 # Assume rapi_pem contains a valid PEM-formatted certificate and key
154 logging.debug("Writing RAPI certificate at %s", rapicert_file)
155 utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
157 elif new_rapi_cert or not rapi_cert_exists:
159 utils.CreateBackup(rapicert_file)
161 logging.debug("Generating new RAPI certificate at %s", rapicert_file)
162 utils.GenerateSelfSignedSslCert(rapicert_file)
165 spice_cert_exists = os.path.exists(spicecert_file)
166 spice_cacert_exists = os.path.exists(spicecacert_file)
168 # spice_cert_pem implies also spice_cacert_pem
169 logging.debug("Writing SPICE certificate at %s", spicecert_file)
170 utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True)
171 logging.debug("Writing SPICE CA certificate at %s", spicecacert_file)
172 utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True)
173 elif new_spice_cert or not spice_cert_exists:
174 if spice_cert_exists:
175 utils.CreateBackup(spicecert_file)
176 if spice_cacert_exists:
177 utils.CreateBackup(spicecacert_file)
179 logging.debug("Generating new self-signed SPICE certificate at %s",
181 (_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file)
183 # Self-signed certificate -> the public certificate is also the CA public
185 logging.debug("Writing the public certificate to %s",
187 utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem)
189 # Cluster domain secret
191 logging.debug("Writing cluster domain secret to %s", cds_file)
192 utils.WriteFile(cds_file, data=cds, backup=True)
194 elif new_cds or not os.path.exists(cds_file):
195 logging.debug("Generating new cluster domain secret at %s", cds_file)
196 GenerateHmacKey(cds_file)
199 def _InitGanetiServerSetup(master_name):
200 """Setup the necessary configuration for the initial node daemon.
202 This creates the nodepass file containing the shared password for
203 the cluster, generates the SSL certificate and starts the node daemon.
205 @type master_name: str
206 @param master_name: Name of the master node
209 # Generate cluster secrets
210 GenerateClusterCrypto(True, False, False, False, False)
212 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start", constants.NODED])
214 raise errors.OpExecError("Could not start the node daemon, command %s"
215 " had exitcode %s and error %s" %
216 (result.cmd, result.exit_code, result.output))
218 _WaitForNodeDaemon(master_name)
221 def _WaitForNodeDaemon(node_name):
222 """Wait for node daemon to become responsive.
225 def _CheckNodeDaemon():
226 # Pylint bug <http://www.logilab.org/ticket/35642>
227 # pylint: disable=E1101
228 result = rpc.BootstrapRunner().call_version([node_name])[node_name]
230 raise utils.RetryAgain()
233 utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT)
234 except utils.RetryTimeout:
235 raise errors.OpExecError("Node daemon on %s didn't answer queries within"
236 " %s seconds" % (node_name, _DAEMON_READY_TIMEOUT))
239 def _WaitForMasterDaemon():
240 """Wait for master daemon to become responsive.
243 def _CheckMasterDaemon():
246 (cluster_name, ) = cl.QueryConfigValues(["cluster_name"])
248 raise utils.RetryAgain()
250 logging.debug("Received cluster name %s from master", cluster_name)
253 utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT)
254 except utils.RetryTimeout:
255 raise errors.OpExecError("Master daemon didn't answer queries within"
256 " %s seconds" % _DAEMON_READY_TIMEOUT)
259 def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose,
260 use_cluster_key, ask_key, strict_host_check, data):
261 """Runs a command to configure something on a remote machine.
263 @type cluster_name: string
264 @param cluster_name: Cluster name
266 @param node: Node name
267 @type basecmd: string
268 @param basecmd: Base command (path on the remote machine)
270 @param debug: Enable debug output
272 @param verbose: Enable verbose output
273 @type use_cluster_key: bool
274 @param use_cluster_key: See L{ssh.SshRunner.BuildCmd}
276 @param ask_key: See L{ssh.SshRunner.BuildCmd}
277 @type strict_host_check: bool
278 @param strict_host_check: See L{ssh.SshRunner.BuildCmd}
279 @param data: JSON-serializable input data for script (passed to stdin)
284 # Pass --debug/--verbose to the external script if set on our invocation
286 cmd.append("--debug")
289 cmd.append("--verbose")
291 family = ssconf.SimpleStore().GetPrimaryIPFamily()
292 srun = ssh.SshRunner(cluster_name,
293 ipv6=(family == netutils.IP6Address.family))
294 scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER,
295 utils.ShellQuoteArgs(cmd),
296 batch=False, ask_key=ask_key, quiet=False,
297 strict_host_check=strict_host_check,
298 use_cluster_key=use_cluster_key)
300 tempfh = tempfile.TemporaryFile()
302 tempfh.write(serializer.DumpJson(data))
305 result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh)
310 raise errors.OpExecError("Command '%s' failed: %s" %
311 (result.cmd, result.fail_reason))
314 def _InitFileStorage(file_storage_dir):
315 """Initialize if needed the file storage.
317 @param file_storage_dir: the user-supplied value
318 @return: either empty string (if file storage was disabled at build
319 time) or the normalized path to the storage directory
322 file_storage_dir = os.path.normpath(file_storage_dir)
324 if not os.path.isabs(file_storage_dir):
325 raise errors.OpPrereqError("File storage directory '%s' is not an absolute"
326 " path" % file_storage_dir, errors.ECODE_INVAL)
328 if not os.path.exists(file_storage_dir):
330 os.makedirs(file_storage_dir, 0750)
332 raise errors.OpPrereqError("Cannot create file storage directory"
333 " '%s': %s" % (file_storage_dir, err),
334 errors.ECODE_ENVIRON)
336 if not os.path.isdir(file_storage_dir):
337 raise errors.OpPrereqError("The file storage directory '%s' is not"
338 " a directory." % file_storage_dir,
339 errors.ECODE_ENVIRON)
340 return file_storage_dir
343 def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914
344 master_netmask, master_netdev, file_storage_dir,
345 shared_file_storage_dir, candidate_pool_size, secondary_ip=None,
346 vg_name=None, beparams=None, nicparams=None, ndparams=None,
347 hvparams=None, diskparams=None, enabled_hypervisors=None,
348 modify_etc_hosts=True, modify_ssh_setup=True,
349 maintain_node_health=False, drbd_helper=None, uid_pool=None,
350 default_iallocator=None, primary_ip_version=None, ipolicy=None,
351 prealloc_wipe_disks=False, use_external_mip_script=False,
352 hv_state=None, disk_state=None):
353 """Initialise the cluster.
355 @type candidate_pool_size: int
356 @param candidate_pool_size: master candidate pool size
359 # TODO: complete the docstring
360 if config.ConfigWriter.IsCluster():
361 raise errors.OpPrereqError("Cluster is already initialised",
364 if not enabled_hypervisors:
365 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
366 " least one member", errors.ECODE_INVAL)
367 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
369 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
370 " entries: %s" % invalid_hvs,
374 ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version)
375 except errors.ProgrammerError:
376 raise errors.OpPrereqError("Invalid primary ip version: %d." %
377 primary_ip_version, errors.ECODE_INVAL)
379 hostname = netutils.GetHostname(family=ipcls.family)
380 if not ipcls.IsValid(hostname.ip):
381 raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d"
382 " address." % (hostname.ip, primary_ip_version),
385 if ipcls.IsLoopback(hostname.ip):
386 raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback"
387 " address. Please fix DNS or %s." %
388 (hostname.ip, pathutils.ETC_HOSTS),
389 errors.ECODE_ENVIRON)
391 if not ipcls.Own(hostname.ip):
392 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
393 " to %s,\nbut this ip address does not"
394 " belong to this host" %
395 hostname.ip, errors.ECODE_ENVIRON)
397 clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family)
399 if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5):
400 raise errors.OpPrereqError("Cluster IP already active",
401 errors.ECODE_NOTUNIQUE)
404 if primary_ip_version == constants.IP6_VERSION:
405 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
406 " IPv4 address must be given as secondary",
408 secondary_ip = hostname.ip
410 if not netutils.IP4Address.IsValid(secondary_ip):
411 raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid"
412 " IPv4 address." % secondary_ip,
415 if not netutils.IP4Address.Own(secondary_ip):
416 raise errors.OpPrereqError("You gave %s as secondary IP,"
417 " but it does not belong to this host." %
418 secondary_ip, errors.ECODE_ENVIRON)
420 if master_netmask is not None:
421 if not ipcls.ValidateNetmask(master_netmask):
422 raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " %
423 (master_netmask, primary_ip_version),
426 master_netmask = ipcls.iplen
428 if vg_name is not None:
429 # Check if volume group is valid
430 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
431 constants.MIN_VG_SIZE)
433 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
434 " you are not using lvm" % vgstatus,
437 if drbd_helper is not None:
439 curr_helper = bdev.BaseDRBD.GetUsermodeHelper()
440 except errors.BlockDeviceError, err:
441 raise errors.OpPrereqError("Error while checking drbd helper"
442 " (specify --no-drbd-storage if you are not"
443 " using drbd): %s" % str(err),
444 errors.ECODE_ENVIRON)
445 if drbd_helper != curr_helper:
446 raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s"
447 " is the current helper" % (drbd_helper,
451 if constants.ENABLE_FILE_STORAGE:
452 file_storage_dir = _InitFileStorage(file_storage_dir)
454 file_storage_dir = ""
456 if constants.ENABLE_SHARED_FILE_STORAGE:
457 shared_file_storage_dir = _InitFileStorage(shared_file_storage_dir)
459 shared_file_storage_dir = ""
461 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
462 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
465 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
467 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
469 result.output.strip()), errors.ECODE_INVAL)
471 dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)]
472 utils.EnsureDirs(dirs)
474 objects.UpgradeBeParams(beparams)
475 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
476 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
478 objects.NIC.CheckParameterSyntax(nicparams)
480 full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy)
482 if ndparams is not None:
483 utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES)
485 ndparams = dict(constants.NDC_DEFAULTS)
487 # This is ugly, as we modify the dict itself
488 # FIXME: Make utils.ForceDictType pure functional or write a wrapper
491 for hvname, hvs_data in hv_state.items():
492 utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES)
493 hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data)
495 hv_state = dict((hvname, constants.HVST_DEFAULTS)
496 for hvname in enabled_hypervisors)
498 # FIXME: disk_state has no default values yet
500 for storage, ds_data in disk_state.items():
501 if storage not in constants.DS_VALID_TYPES:
502 raise errors.OpPrereqError("Invalid storage type in disk state: %s" %
503 storage, errors.ECODE_INVAL)
504 for ds_name, state in ds_data.items():
505 utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES)
506 ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state)
508 # hvparams is a mapping of hypervisor->hvparams dict
509 for hv_name, hv_params in hvparams.iteritems():
510 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
511 hv_class = hypervisor.GetHypervisor(hv_name)
512 hv_class.CheckParameterSyntax(hv_params)
514 # diskparams is a mapping of disk-template->diskparams dict
515 for template, dt_params in diskparams.items():
516 param_keys = set(dt_params.keys())
517 default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys())
518 if not (param_keys <= default_param_keys):
519 unknown_params = param_keys - default_param_keys
520 raise errors.OpPrereqError("Invalid parameters for disk template %s:"
522 utils.CommaJoin(unknown_params)),
524 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
526 utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS)
527 except errors.OpPrereqError, err:
528 raise errors.OpPrereqError("While verify diskparam options: %s" % err,
531 # set up ssh config and /etc/hosts
532 sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB)
533 sshkey = sshline.split(" ")[1]
536 utils.AddHostToEtcHosts(hostname.name, hostname.ip)
541 if default_iallocator is not None:
542 alloc_script = utils.FindFile(default_iallocator,
543 constants.IALLOCATOR_SEARCH_PATH,
545 if alloc_script is None:
546 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
547 " specified" % default_iallocator,
549 elif constants.HTOOLS:
550 # htools was enabled at build-time, we default to it
551 if utils.FindFile(constants.IALLOC_HAIL,
552 constants.IALLOCATOR_SEARCH_PATH,
554 default_iallocator = constants.IALLOC_HAIL
558 # init of cluster config file
559 cluster_config = objects.Cluster(
561 rsahostkeypub=sshkey,
562 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
563 mac_prefix=mac_prefix,
564 volume_group_name=vg_name,
565 tcpudp_port_pool=set(),
566 master_node=hostname.name,
567 master_ip=clustername.ip,
568 master_netmask=master_netmask,
569 master_netdev=master_netdev,
570 cluster_name=clustername.name,
571 file_storage_dir=file_storage_dir,
572 shared_file_storage_dir=shared_file_storage_dir,
573 enabled_hypervisors=enabled_hypervisors,
574 beparams={constants.PP_DEFAULT: beparams},
575 nicparams={constants.PP_DEFAULT: nicparams},
578 diskparams=diskparams,
579 candidate_pool_size=candidate_pool_size,
580 modify_etc_hosts=modify_etc_hosts,
581 modify_ssh_setup=modify_ssh_setup,
585 maintain_node_health=maintain_node_health,
586 drbd_usermode_helper=drbd_helper,
587 default_iallocator=default_iallocator,
588 primary_ip_family=ipcls.family,
589 prealloc_wipe_disks=prealloc_wipe_disks,
590 use_external_mip_script=use_external_mip_script,
591 ipolicy=full_ipolicy,
592 hv_state_static=hv_state,
593 disk_state_static=disk_state,
595 master_node_config = objects.Node(name=hostname.name,
596 primary_ip=hostname.ip,
597 secondary_ip=secondary_ip,
599 master_candidate=True,
600 offline=False, drained=False,
601 ctime=now, mtime=now,
603 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
604 cfg = config.ConfigWriter(offline=True)
605 ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE)
606 cfg.Update(cfg.GetClusterInfo(), logging.error)
607 ssconf.WriteSsconfFiles(cfg.GetSsconfValues())
609 # set up the inter-node password and certificate
610 _InitGanetiServerSetup(hostname.name)
612 logging.debug("Starting daemons")
613 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"])
615 raise errors.OpExecError("Could not start daemons, command %s"
616 " had exitcode %s and error %s" %
617 (result.cmd, result.exit_code, result.output))
619 _WaitForMasterDaemon()
622 def InitConfig(version, cluster_config, master_node_config,
623 cfg_file=pathutils.CLUSTER_CONF_FILE):
624 """Create the initial cluster configuration.
626 It will contain the current node, which will also be the master
627 node, and no instances.
630 @param version: configuration version
631 @type cluster_config: L{objects.Cluster}
632 @param cluster_config: cluster configuration
633 @type master_node_config: L{objects.Node}
634 @param master_node_config: master node configuration
635 @type cfg_file: string
636 @param cfg_file: configuration file path
639 uuid_generator = config.TemporaryReservationManager()
640 cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID,
642 master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID,
645 master_node_config.name: master_node_config,
647 default_nodegroup = objects.NodeGroup(
648 uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID),
649 name=constants.INITIAL_NODE_GROUP_NAME,
650 members=[master_node_config.name],
654 default_nodegroup.uuid: default_nodegroup,
657 config_data = objects.ConfigData(version=version,
658 cluster=cluster_config,
659 nodegroups=nodegroups,
664 ctime=now, mtime=now)
665 utils.WriteFile(cfg_file,
666 data=serializer.Dump(config_data.ToDict()),
670 def FinalizeClusterDestroy(master):
671 """Execute the last steps of cluster destroy
673 This function shuts down all the daemons, completing the destroy
674 begun in cmdlib.LUDestroyOpcode.
677 cfg = config.ConfigWriter()
678 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
679 runner = rpc.BootstrapRunner()
681 master_params = cfg.GetMasterNetworkParameters()
682 master_params.name = master
683 ems = cfg.GetUseExternalMipScript()
684 result = runner.call_node_deactivate_master_ip(master_params.name,
687 msg = result.fail_msg
689 logging.warning("Could not disable the master IP: %s", msg)
691 result = runner.call_node_stop_master(master)
692 msg = result.fail_msg
694 logging.warning("Could not disable the master role: %s", msg)
696 result = runner.call_node_leave_cluster(master, modify_ssh_setup)
697 msg = result.fail_msg
699 logging.warning("Could not shutdown the node daemon and cleanup"
700 " the node: %s", msg)
703 def SetupNodeDaemon(opts, cluster_name, node):
704 """Add a node to the cluster.
706 This function must be called before the actual opcode, and will ssh
707 to the remote node, copy the needed files, and start ganeti-noded,
708 allowing the master to do the rest via normal rpc calls.
710 @param cluster_name: the cluster name
711 @param node: the name of the new node
715 constants.NDS_CLUSTER_NAME: cluster_name,
716 constants.NDS_NODE_DAEMON_CERTIFICATE:
717 utils.ReadFile(pathutils.NODED_CERT_FILE),
718 constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(),
719 constants.NDS_START_NODE_DAEMON: True,
722 RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP,
723 opts.debug, opts.verbose,
724 True, opts.ssh_key_check, opts.ssh_key_check, data)
726 _WaitForNodeDaemon(node)
729 def MasterFailover(no_voting=False):
730 """Failover the master node.
732 This checks that we are not already the master, and will cause the
733 current master to cease being master, and the non-master to become
736 @type no_voting: boolean
737 @param no_voting: force the operation without remote nodes agreement
741 sstore = ssconf.SimpleStore()
743 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
744 node_list = sstore.GetNodeList()
745 mc_list = sstore.GetMasterCandidates()
747 if old_master == new_master:
748 raise errors.OpPrereqError("This commands must be run on the node"
749 " where you want the new master to be."
750 " %s is already the master" %
751 old_master, errors.ECODE_INVAL)
753 if new_master not in mc_list:
754 mc_no_master = [name for name in mc_list if name != old_master]
755 raise errors.OpPrereqError("This node is not among the nodes marked"
756 " as master candidates. Only these nodes"
757 " can become masters. Current list of"
758 " master candidates is:\n"
759 "%s" % ("\n".join(mc_no_master)),
763 vote_list = GatherMasterVotes(node_list)
766 voted_master = vote_list[0][0]
767 if voted_master is None:
768 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
769 " not respond.", errors.ECODE_ENVIRON)
770 elif voted_master != old_master:
771 raise errors.OpPrereqError("I have a wrong configuration, I believe"
772 " the master is %s but the other nodes"
773 " voted %s. Please resync the configuration"
775 (old_master, voted_master),
781 logging.info("Setting master to %s, old master: %s", new_master, old_master)
784 # instantiate a real config writer, as we now know we have the
786 cfg = config.ConfigWriter(accept_foreign=True)
788 cluster_info = cfg.GetClusterInfo()
789 cluster_info.master_node = new_master
790 # this will also regenerate the ssconf files, since we updated the
792 cfg.Update(cluster_info, logging.error)
793 except errors.ConfigurationError, err:
794 logging.error("Error while trying to set the new master: %s",
798 # if cfg.Update worked, then it means the old master daemon won't be
799 # able now to write its own config file (we rely on locking in both
800 # backend.UploadFile() and ConfigWriter._Write(); hence the next
801 # step is to kill the old master
803 logging.info("Stopping the master daemon on node %s", old_master)
805 runner = rpc.BootstrapRunner()
806 master_params = cfg.GetMasterNetworkParameters()
807 master_params.name = old_master
808 ems = cfg.GetUseExternalMipScript()
809 result = runner.call_node_deactivate_master_ip(master_params.name,
812 msg = result.fail_msg
814 logging.warning("Could not disable the master IP: %s", msg)
816 result = runner.call_node_stop_master(old_master)
817 msg = result.fail_msg
819 logging.error("Could not disable the master role on the old master"
820 " %s, please disable manually: %s", old_master, msg)
822 logging.info("Checking master IP non-reachability...")
824 master_ip = sstore.GetMasterIP()
827 # Here we have a phase where no master should be running
829 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
830 raise utils.RetryAgain()
833 utils.Retry(_check_ip, (1, 1.5, 5), total_timeout)
834 except utils.RetryTimeout:
835 logging.warning("The master IP is still reachable after %s seconds,"
836 " continuing but activating the master on the current"
837 " node will probably fail", total_timeout)
839 if jstore.CheckDrainFlag():
840 logging.info("Undraining job queue")
841 jstore.SetDrainFlag(False)
843 logging.info("Starting the master daemons on the new master")
845 result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master,
847 msg = result.fail_msg
849 logging.error("Could not start the master role on the new master"
850 " %s, please check: %s", new_master, msg)
853 logging.info("Master failed over from %s to %s", old_master, new_master)
858 """Returns the current master node.
860 This is a separate function in bootstrap since it's needed by
861 gnt-cluster, and instead of importing directly ssconf, it's better
862 to abstract it in bootstrap, where we do use ssconf in other
866 sstore = ssconf.SimpleStore()
868 old_master, _ = ssconf.GetMasterAndMyself(sstore)
873 def GatherMasterVotes(node_list):
874 """Check the agreement on who is the master.
876 This function will return a list of (node, number of votes), ordered
877 by the number of votes. Errors will be denoted by the key 'None'.
879 Note that the sum of votes is the number of nodes this machine
880 knows, whereas the number of entries in the list could be different
881 (if some nodes vote for another master).
883 We remove ourselves from the list since we know that (bugs aside)
884 since we use the same source for configuration information for both
885 backend and boostrap, we'll always vote for ourselves.
887 @type node_list: list
888 @param node_list: the list of nodes to query for master info; the current
889 node will be removed if it is in the list
891 @return: list of (node, votes)
894 myself = netutils.Hostname.GetSysName()
896 node_list.remove(myself)
900 # no nodes left (eventually after removing myself)
902 results = rpc.BootstrapRunner().call_master_info(node_list)
903 if not isinstance(results, dict):
904 # this should not happen (unless internal error in rpc)
905 logging.critical("Can't complete rpc call, aborting master startup")
906 return [(None, len(node_list))]
914 logging.warning("Error contacting node %s: %s", node, msg)
916 # for now we accept both length 3, 4 and 5 (data[3] is primary ip version
917 # and data[4] is the master netmask)
918 elif not isinstance(data, (tuple, list)) or len(data) < 3:
919 logging.warning("Invalid data received from node %s: %s", node, data)
922 if None not in votes:
926 master_node = data[2]
927 if master_node not in votes:
928 votes[master_node] = 0
929 votes[master_node] += 1
931 vote_list = [v for v in votes.items()]
932 # sort first on number of votes then on name, since we want None
933 # sorted later if we have the half of the nodes not responding, and
934 # half voting all for the same master
935 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)