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.storage import drbd
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 _WaitForSshDaemon(hostname, port, family):
260 """Wait for SSH daemon to become responsive.
263 hostip = netutils.GetHostname(name=hostname, family=family).ip
265 def _CheckSshDaemon():
266 if netutils.TcpPing(hostip, port, timeout=1.0, live_port_needed=True):
267 logging.debug("SSH daemon on %s:%s (IP address %s) has become"
268 " responsive", hostname, port, hostip)
270 raise utils.RetryAgain()
273 utils.Retry(_CheckSshDaemon, 1.0, _DAEMON_READY_TIMEOUT)
274 except utils.RetryTimeout:
275 raise errors.OpExecError("SSH daemon on %s:%s (IP address %s) didn't"
276 " become responsive within %s seconds" %
277 (hostname, port, hostip, _DAEMON_READY_TIMEOUT))
280 def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose,
281 use_cluster_key, ask_key, strict_host_check, data):
282 """Runs a command to configure something on a remote machine.
284 @type cluster_name: string
285 @param cluster_name: Cluster name
287 @param node: Node name
288 @type basecmd: string
289 @param basecmd: Base command (path on the remote machine)
291 @param debug: Enable debug output
293 @param verbose: Enable verbose output
294 @type use_cluster_key: bool
295 @param use_cluster_key: See L{ssh.SshRunner.BuildCmd}
297 @param ask_key: See L{ssh.SshRunner.BuildCmd}
298 @type strict_host_check: bool
299 @param strict_host_check: See L{ssh.SshRunner.BuildCmd}
300 @param data: JSON-serializable input data for script (passed to stdin)
305 # Pass --debug/--verbose to the external script if set on our invocation
307 cmd.append("--debug")
310 cmd.append("--verbose")
312 family = ssconf.SimpleStore().GetPrimaryIPFamily()
313 srun = ssh.SshRunner(cluster_name,
314 ipv6=(family == netutils.IP6Address.family))
315 scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER,
316 utils.ShellQuoteArgs(cmd),
317 batch=False, ask_key=ask_key, quiet=False,
318 strict_host_check=strict_host_check,
319 use_cluster_key=use_cluster_key)
321 tempfh = tempfile.TemporaryFile()
323 tempfh.write(serializer.DumpJson(data))
326 result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh)
331 raise errors.OpExecError("Command '%s' failed: %s" %
332 (result.cmd, result.fail_reason))
334 _WaitForSshDaemon(node, netutils.GetDaemonPort(constants.SSH), family)
337 def _InitFileStorage(file_storage_dir):
338 """Initialize if needed the file storage.
340 @param file_storage_dir: the user-supplied value
341 @return: either empty string (if file storage was disabled at build
342 time) or the normalized path to the storage directory
345 file_storage_dir = os.path.normpath(file_storage_dir)
347 if not os.path.isabs(file_storage_dir):
348 raise errors.OpPrereqError("File storage directory '%s' is not an absolute"
349 " path" % file_storage_dir, errors.ECODE_INVAL)
351 if not os.path.exists(file_storage_dir):
353 os.makedirs(file_storage_dir, 0750)
355 raise errors.OpPrereqError("Cannot create file storage directory"
356 " '%s': %s" % (file_storage_dir, err),
357 errors.ECODE_ENVIRON)
359 if not os.path.isdir(file_storage_dir):
360 raise errors.OpPrereqError("The file storage directory '%s' is not"
361 " a directory." % file_storage_dir,
362 errors.ECODE_ENVIRON)
363 return file_storage_dir
366 def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914
367 master_netmask, master_netdev, file_storage_dir,
368 shared_file_storage_dir, candidate_pool_size, secondary_ip=None,
369 vg_name=None, beparams=None, nicparams=None, ndparams=None,
370 hvparams=None, diskparams=None, enabled_hypervisors=None,
371 modify_etc_hosts=True, modify_ssh_setup=True,
372 maintain_node_health=False, drbd_helper=None, uid_pool=None,
373 default_iallocator=None, primary_ip_version=None, ipolicy=None,
374 prealloc_wipe_disks=False, use_external_mip_script=False,
375 hv_state=None, disk_state=None, enabled_disk_templates=None):
376 """Initialise the cluster.
378 @type candidate_pool_size: int
379 @param candidate_pool_size: master candidate pool size
380 @type enabled_disk_templates: list of string
381 @param enabled_disk_templates: list of disk_templates to be used in this
385 # TODO: complete the docstring
386 if config.ConfigWriter.IsCluster():
387 raise errors.OpPrereqError("Cluster is already initialised",
390 if not enabled_hypervisors:
391 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
392 " least one member", errors.ECODE_INVAL)
393 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
395 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
396 " entries: %s" % invalid_hvs,
399 if not enabled_disk_templates:
400 raise errors.OpPrereqError("Enabled disk templates list must contain at"
401 " least one member", errors.ECODE_INVAL)
402 invalid_disk_templates = \
403 set(enabled_disk_templates) - constants.DISK_TEMPLATES
404 if invalid_disk_templates:
405 raise errors.OpPrereqError("Enabled disk templates list contains invalid"
406 " entries: %s" % invalid_disk_templates,
410 ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version)
411 except errors.ProgrammerError:
412 raise errors.OpPrereqError("Invalid primary ip version: %d." %
413 primary_ip_version, errors.ECODE_INVAL)
415 hostname = netutils.GetHostname(family=ipcls.family)
416 if not ipcls.IsValid(hostname.ip):
417 raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d"
418 " address." % (hostname.ip, primary_ip_version),
421 if ipcls.IsLoopback(hostname.ip):
422 raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback"
423 " address. Please fix DNS or %s." %
424 (hostname.ip, pathutils.ETC_HOSTS),
425 errors.ECODE_ENVIRON)
427 if not ipcls.Own(hostname.ip):
428 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
429 " to %s,\nbut this ip address does not"
430 " belong to this host" %
431 hostname.ip, errors.ECODE_ENVIRON)
433 clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family)
435 if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5):
436 raise errors.OpPrereqError("Cluster IP already active",
437 errors.ECODE_NOTUNIQUE)
440 if primary_ip_version == constants.IP6_VERSION:
441 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
442 " IPv4 address must be given as secondary",
444 secondary_ip = hostname.ip
446 if not netutils.IP4Address.IsValid(secondary_ip):
447 raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid"
448 " IPv4 address." % secondary_ip,
451 if not netutils.IP4Address.Own(secondary_ip):
452 raise errors.OpPrereqError("You gave %s as secondary IP,"
453 " but it does not belong to this host." %
454 secondary_ip, errors.ECODE_ENVIRON)
456 if master_netmask is not None:
457 if not ipcls.ValidateNetmask(master_netmask):
458 raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " %
459 (master_netmask, primary_ip_version),
462 master_netmask = ipcls.iplen
465 # Check if volume group is valid
466 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
467 constants.MIN_VG_SIZE)
469 raise errors.OpPrereqError("Error: %s" % vgstatus, errors.ECODE_INVAL)
471 if drbd_helper is not None:
473 curr_helper = drbd.DRBD8.GetUsermodeHelper()
474 except errors.BlockDeviceError, err:
475 raise errors.OpPrereqError("Error while checking drbd helper"
476 " (specify --no-drbd-storage if you are not"
477 " using drbd): %s" % str(err),
478 errors.ECODE_ENVIRON)
479 if drbd_helper != curr_helper:
480 raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s"
481 " is the current helper" % (drbd_helper,
485 logging.debug("Stopping daemons (if any are running)")
486 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"])
488 raise errors.OpExecError("Could not stop daemons, command %s"
489 " had exitcode %s and error '%s'" %
490 (result.cmd, result.exit_code, result.output))
492 if constants.ENABLE_FILE_STORAGE:
493 file_storage_dir = _InitFileStorage(file_storage_dir)
495 file_storage_dir = ""
497 if constants.ENABLE_SHARED_FILE_STORAGE:
498 shared_file_storage_dir = _InitFileStorage(shared_file_storage_dir)
500 shared_file_storage_dir = ""
502 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
503 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
506 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
508 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
510 result.output.strip()), errors.ECODE_INVAL)
512 dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)]
513 utils.EnsureDirs(dirs)
515 objects.UpgradeBeParams(beparams)
516 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
517 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
519 objects.NIC.CheckParameterSyntax(nicparams)
521 full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy)
523 if ndparams is not None:
524 utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES)
526 ndparams = dict(constants.NDC_DEFAULTS)
528 # This is ugly, as we modify the dict itself
529 # FIXME: Make utils.ForceDictType pure functional or write a wrapper
532 for hvname, hvs_data in hv_state.items():
533 utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES)
534 hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data)
536 hv_state = dict((hvname, constants.HVST_DEFAULTS)
537 for hvname in enabled_hypervisors)
539 # FIXME: disk_state has no default values yet
541 for storage, ds_data in disk_state.items():
542 if storage not in constants.DS_VALID_TYPES:
543 raise errors.OpPrereqError("Invalid storage type in disk state: %s" %
544 storage, errors.ECODE_INVAL)
545 for ds_name, state in ds_data.items():
546 utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES)
547 ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state)
549 # hvparams is a mapping of hypervisor->hvparams dict
550 for hv_name, hv_params in hvparams.iteritems():
551 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
552 hv_class = hypervisor.GetHypervisor(hv_name)
553 hv_class.CheckParameterSyntax(hv_params)
555 # diskparams is a mapping of disk-template->diskparams dict
556 for template, dt_params in diskparams.items():
557 param_keys = set(dt_params.keys())
558 default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys())
559 if not (param_keys <= default_param_keys):
560 unknown_params = param_keys - default_param_keys
561 raise errors.OpPrereqError("Invalid parameters for disk template %s:"
563 utils.CommaJoin(unknown_params)),
565 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
566 if template == constants.DT_DRBD8 and vg_name is not None:
567 # The default METAVG value is equal to the VG name set at init time,
569 dt_params[constants.DRBD_DEFAULT_METAVG] = vg_name
572 utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS)
573 except errors.OpPrereqError, err:
574 raise errors.OpPrereqError("While verify diskparam options: %s" % err,
577 # set up ssh config and /etc/hosts
578 sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB)
579 sshkey = sshline.split(" ")[1]
582 utils.AddHostToEtcHosts(hostname.name, hostname.ip)
587 if default_iallocator is not None:
588 alloc_script = utils.FindFile(default_iallocator,
589 constants.IALLOCATOR_SEARCH_PATH,
591 if alloc_script is None:
592 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
593 " specified" % default_iallocator,
595 elif constants.HTOOLS:
596 # htools was enabled at build-time, we default to it
597 if utils.FindFile(constants.IALLOC_HAIL,
598 constants.IALLOCATOR_SEARCH_PATH,
600 default_iallocator = constants.IALLOC_HAIL
604 # init of cluster config file
605 cluster_config = objects.Cluster(
607 rsahostkeypub=sshkey,
608 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
609 mac_prefix=mac_prefix,
610 volume_group_name=vg_name,
611 tcpudp_port_pool=set(),
612 master_ip=clustername.ip,
613 master_netmask=master_netmask,
614 master_netdev=master_netdev,
615 cluster_name=clustername.name,
616 file_storage_dir=file_storage_dir,
617 shared_file_storage_dir=shared_file_storage_dir,
618 enabled_hypervisors=enabled_hypervisors,
619 beparams={constants.PP_DEFAULT: beparams},
620 nicparams={constants.PP_DEFAULT: nicparams},
623 diskparams=diskparams,
624 candidate_pool_size=candidate_pool_size,
625 modify_etc_hosts=modify_etc_hosts,
626 modify_ssh_setup=modify_ssh_setup,
630 maintain_node_health=maintain_node_health,
631 drbd_usermode_helper=drbd_helper,
632 default_iallocator=default_iallocator,
633 primary_ip_family=ipcls.family,
634 prealloc_wipe_disks=prealloc_wipe_disks,
635 use_external_mip_script=use_external_mip_script,
636 ipolicy=full_ipolicy,
637 hv_state_static=hv_state,
638 disk_state_static=disk_state,
639 enabled_disk_templates=enabled_disk_templates,
641 master_node_config = objects.Node(name=hostname.name,
642 primary_ip=hostname.ip,
643 secondary_ip=secondary_ip,
645 master_candidate=True,
646 offline=False, drained=False,
647 ctime=now, mtime=now,
649 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
650 cfg = config.ConfigWriter(offline=True)
651 ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE)
652 cfg.Update(cfg.GetClusterInfo(), logging.error)
653 ssconf.WriteSsconfFiles(cfg.GetSsconfValues())
655 # set up the inter-node password and certificate
656 _InitGanetiServerSetup(hostname.name)
658 logging.debug("Starting daemons")
659 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"])
661 raise errors.OpExecError("Could not start daemons, command %s"
662 " had exitcode %s and error %s" %
663 (result.cmd, result.exit_code, result.output))
665 _WaitForMasterDaemon()
668 def InitConfig(version, cluster_config, master_node_config,
669 cfg_file=pathutils.CLUSTER_CONF_FILE):
670 """Create the initial cluster configuration.
672 It will contain the current node, which will also be the master
673 node, and no instances.
676 @param version: configuration version
677 @type cluster_config: L{objects.Cluster}
678 @param cluster_config: cluster configuration
679 @type master_node_config: L{objects.Node}
680 @param master_node_config: master node configuration
681 @type cfg_file: string
682 @param cfg_file: configuration file path
685 uuid_generator = config.TemporaryReservationManager()
686 cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID,
688 master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID,
690 cluster_config.master_node = master_node_config.uuid
692 master_node_config.uuid: master_node_config,
694 default_nodegroup = objects.NodeGroup(
695 uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID),
696 name=constants.INITIAL_NODE_GROUP_NAME,
697 members=[master_node_config.uuid],
701 default_nodegroup.uuid: default_nodegroup,
704 config_data = objects.ConfigData(version=version,
705 cluster=cluster_config,
706 nodegroups=nodegroups,
711 ctime=now, mtime=now)
712 utils.WriteFile(cfg_file,
713 data=serializer.Dump(config_data.ToDict()),
717 def FinalizeClusterDestroy(master_uuid):
718 """Execute the last steps of cluster destroy
720 This function shuts down all the daemons, completing the destroy
721 begun in cmdlib.LUDestroyOpcode.
724 cfg = config.ConfigWriter()
725 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
726 runner = rpc.BootstrapRunner()
728 master_name = cfg.GetNodeName(master_uuid)
730 master_params = cfg.GetMasterNetworkParameters()
731 master_params.uuid = master_uuid
732 ems = cfg.GetUseExternalMipScript()
733 result = runner.call_node_deactivate_master_ip(master_name, master_params,
736 msg = result.fail_msg
738 logging.warning("Could not disable the master IP: %s", msg)
740 result = runner.call_node_stop_master(master_name)
741 msg = result.fail_msg
743 logging.warning("Could not disable the master role: %s", msg)
745 result = runner.call_node_leave_cluster(master_name, modify_ssh_setup)
746 msg = result.fail_msg
748 logging.warning("Could not shutdown the node daemon and cleanup"
749 " the node: %s", msg)
752 def SetupNodeDaemon(opts, cluster_name, node):
753 """Add a node to the cluster.
755 This function must be called before the actual opcode, and will ssh
756 to the remote node, copy the needed files, and start ganeti-noded,
757 allowing the master to do the rest via normal rpc calls.
759 @param cluster_name: the cluster name
760 @param node: the name of the new node
764 constants.NDS_CLUSTER_NAME: cluster_name,
765 constants.NDS_NODE_DAEMON_CERTIFICATE:
766 utils.ReadFile(pathutils.NODED_CERT_FILE),
767 constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(),
768 constants.NDS_START_NODE_DAEMON: True,
771 RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP,
772 opts.debug, opts.verbose,
773 True, opts.ssh_key_check, opts.ssh_key_check, data)
775 _WaitForNodeDaemon(node)
778 def MasterFailover(no_voting=False):
779 """Failover the master node.
781 This checks that we are not already the master, and will cause the
782 current master to cease being master, and the non-master to become
785 @type no_voting: boolean
786 @param no_voting: force the operation without remote nodes agreement
790 sstore = ssconf.SimpleStore()
792 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
793 node_names = sstore.GetNodeList()
794 mc_list = sstore.GetMasterCandidates()
796 if old_master == new_master:
797 raise errors.OpPrereqError("This commands must be run on the node"
798 " where you want the new master to be."
799 " %s is already the master" %
800 old_master, errors.ECODE_INVAL)
802 if new_master not in mc_list:
803 mc_no_master = [name for name in mc_list if name != old_master]
804 raise errors.OpPrereqError("This node is not among the nodes marked"
805 " as master candidates. Only these nodes"
806 " can become masters. Current list of"
807 " master candidates is:\n"
808 "%s" % ("\n".join(mc_no_master)),
812 vote_list = GatherMasterVotes(node_names)
815 voted_master = vote_list[0][0]
816 if voted_master is None:
817 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
818 " not respond.", errors.ECODE_ENVIRON)
819 elif voted_master != old_master:
820 raise errors.OpPrereqError("I have a wrong configuration, I believe"
821 " the master is %s but the other nodes"
822 " voted %s. Please resync the configuration"
824 (old_master, voted_master),
830 logging.info("Setting master to %s, old master: %s", new_master, old_master)
833 # instantiate a real config writer, as we now know we have the
835 cfg = config.ConfigWriter(accept_foreign=True)
837 old_master_node = cfg.GetNodeInfoByName(old_master)
838 if old_master_node is None:
839 raise errors.OpPrereqError("Could not find old master node '%s' in"
840 " cluster configuration." % old_master,
843 cluster_info = cfg.GetClusterInfo()
844 new_master_node = cfg.GetNodeInfoByName(new_master)
845 if new_master_node is None:
846 raise errors.OpPrereqError("Could not find new master node '%s' in"
847 " cluster configuration." % new_master,
850 cluster_info.master_node = new_master_node.uuid
851 # this will also regenerate the ssconf files, since we updated the
853 cfg.Update(cluster_info, logging.error)
854 except errors.ConfigurationError, err:
855 logging.error("Error while trying to set the new master: %s",
859 # if cfg.Update worked, then it means the old master daemon won't be
860 # able now to write its own config file (we rely on locking in both
861 # backend.UploadFile() and ConfigWriter._Write(); hence the next
862 # step is to kill the old master
864 logging.info("Stopping the master daemon on node %s", old_master)
866 runner = rpc.BootstrapRunner()
867 master_params = cfg.GetMasterNetworkParameters()
868 master_params.uuid = old_master_node.uuid
869 ems = cfg.GetUseExternalMipScript()
870 result = runner.call_node_deactivate_master_ip(old_master,
873 msg = result.fail_msg
875 logging.warning("Could not disable the master IP: %s", msg)
877 result = runner.call_node_stop_master(old_master)
878 msg = result.fail_msg
880 logging.error("Could not disable the master role on the old master"
881 " %s, please disable manually: %s", old_master, msg)
883 logging.info("Checking master IP non-reachability...")
885 master_ip = sstore.GetMasterIP()
888 # Here we have a phase where no master should be running
890 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
891 raise utils.RetryAgain()
894 utils.Retry(_check_ip, (1, 1.5, 5), total_timeout)
895 except utils.RetryTimeout:
896 logging.warning("The master IP is still reachable after %s seconds,"
897 " continuing but activating the master on the current"
898 " node will probably fail", total_timeout)
900 if jstore.CheckDrainFlag():
901 logging.info("Undraining job queue")
902 jstore.SetDrainFlag(False)
904 logging.info("Starting the master daemons on the new master")
906 result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master,
908 msg = result.fail_msg
910 logging.error("Could not start the master role on the new master"
911 " %s, please check: %s", new_master, msg)
914 logging.info("Master failed over from %s to %s", old_master, new_master)
919 """Returns the current master node.
921 This is a separate function in bootstrap since it's needed by
922 gnt-cluster, and instead of importing directly ssconf, it's better
923 to abstract it in bootstrap, where we do use ssconf in other
927 sstore = ssconf.SimpleStore()
929 old_master, _ = ssconf.GetMasterAndMyself(sstore)
934 def GatherMasterVotes(node_names):
935 """Check the agreement on who is the master.
937 This function will return a list of (node, number of votes), ordered
938 by the number of votes. Errors will be denoted by the key 'None'.
940 Note that the sum of votes is the number of nodes this machine
941 knows, whereas the number of entries in the list could be different
942 (if some nodes vote for another master).
944 We remove ourselves from the list since we know that (bugs aside)
945 since we use the same source for configuration information for both
946 backend and boostrap, we'll always vote for ourselves.
948 @type node_names: list
949 @param node_names: the list of nodes to query for master info; the current
950 node will be removed if it is in the list
952 @return: list of (node, votes)
955 myself = netutils.Hostname.GetSysName()
957 node_names.remove(myself)
961 # no nodes left (eventually after removing myself)
963 results = rpc.BootstrapRunner().call_master_info(node_names)
964 if not isinstance(results, dict):
965 # this should not happen (unless internal error in rpc)
966 logging.critical("Can't complete rpc call, aborting master startup")
967 return [(None, len(node_names))]
969 for node_name in results:
970 nres = results[node_name]
975 logging.warning("Error contacting node %s: %s", node_name, msg)
977 # for now we accept both length 3, 4 and 5 (data[3] is primary ip version
978 # and data[4] is the master netmask)
979 elif not isinstance(data, (tuple, list)) or len(data) < 3:
980 logging.warning("Invalid data received from node %s: %s",
984 if None not in votes:
988 master_node = data[2]
989 if master_node not in votes:
990 votes[master_node] = 0
991 votes[master_node] += 1
993 vote_list = [v for v in votes.items()]
994 # sort first on number of votes then on name, since we want None
995 # sorted later if we have the half of the nodes not responding, and
996 # half voting all for the same master
997 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)