4 # Copyright (C) 2006, 2007, 2008, 2010 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.
32 from ganeti import rpc
33 from ganeti import ssh
34 from ganeti import utils
35 from ganeti import errors
36 from ganeti import config
37 from ganeti import constants
38 from ganeti import objects
39 from ganeti import ssconf
40 from ganeti import serializer
41 from ganeti import hypervisor
42 from ganeti import bdev
43 from ganeti import netutils
44 from ganeti import backend
48 """Setup the SSH configuration for the cluster.
50 This generates a dsa keypair for root, adds the pub key to the
51 permitted hosts and adds the hostkey to its own known hosts.
54 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
56 for name in priv_key, pub_key:
57 if os.path.exists(name):
58 utils.CreateBackup(name)
59 utils.RemoveFile(name)
61 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
65 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
68 utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
71 def GenerateHmacKey(file_name):
72 """Writes a new HMAC key.
75 @param file_name: Path to output file
78 utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400,
82 def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key,
83 new_cds, rapi_cert_pem=None, cds=None,
84 nodecert_file=constants.NODED_CERT_FILE,
85 rapicert_file=constants.RAPI_CERT_FILE,
86 hmackey_file=constants.CONFD_HMAC_KEY,
87 cds_file=constants.CLUSTER_DOMAIN_SECRET_FILE):
88 """Updates the cluster certificates, keys and secrets.
90 @type new_cluster_cert: bool
91 @param new_cluster_cert: Whether to generate a new cluster certificate
92 @type new_rapi_cert: bool
93 @param new_rapi_cert: Whether to generate a new RAPI certificate
94 @type new_confd_hmac_key: bool
95 @param new_confd_hmac_key: Whether to generate a new HMAC key
97 @param new_cds: Whether to generate a new cluster domain secret
98 @type rapi_cert_pem: string
99 @param rapi_cert_pem: New RAPI certificate in PEM format
101 @param cds: New cluster domain secret
102 @type nodecert_file: string
103 @param nodecert_file: optional override of the node cert file path
104 @type rapicert_file: string
105 @param rapicert_file: optional override of the rapi cert file path
106 @type hmackey_file: string
107 @param hmackey_file: optional override of the hmac key file path
110 # noded SSL certificate
111 cluster_cert_exists = os.path.exists(nodecert_file)
112 if new_cluster_cert or not cluster_cert_exists:
113 if cluster_cert_exists:
114 utils.CreateBackup(nodecert_file)
116 logging.debug("Generating new cluster certificate at %s", nodecert_file)
117 utils.GenerateSelfSignedSslCert(nodecert_file)
120 if new_confd_hmac_key or not os.path.exists(hmackey_file):
121 logging.debug("Writing new confd HMAC key to %s", hmackey_file)
122 GenerateHmacKey(hmackey_file)
125 rapi_cert_exists = os.path.exists(rapicert_file)
128 # Assume rapi_pem contains a valid PEM-formatted certificate and key
129 logging.debug("Writing RAPI certificate at %s", rapicert_file)
130 utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
132 elif new_rapi_cert or not rapi_cert_exists:
134 utils.CreateBackup(rapicert_file)
136 logging.debug("Generating new RAPI certificate at %s", rapicert_file)
137 utils.GenerateSelfSignedSslCert(rapicert_file)
139 # Cluster domain secret
141 logging.debug("Writing cluster domain secret to %s", cds_file)
142 utils.WriteFile(cds_file, data=cds, backup=True)
144 elif new_cds or not os.path.exists(cds_file):
145 logging.debug("Generating new cluster domain secret at %s", cds_file)
146 GenerateHmacKey(cds_file)
149 def _InitGanetiServerSetup(master_name):
150 """Setup the necessary configuration for the initial node daemon.
152 This creates the nodepass file containing the shared password for
153 the cluster, generates the SSL certificate and starts the node daemon.
155 @type master_name: str
156 @param master_name: Name of the master node
159 # Generate cluster secrets
160 GenerateClusterCrypto(True, False, False, False)
162 result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED])
164 raise errors.OpExecError("Could not start the node daemon, command %s"
165 " had exitcode %s and error %s" %
166 (result.cmd, result.exit_code, result.output))
168 _WaitForNodeDaemon(master_name)
171 def _WaitForNodeDaemon(node_name):
172 """Wait for node daemon to become responsive.
175 def _CheckNodeDaemon():
176 result = rpc.RpcRunner.call_version([node_name])[node_name]
178 raise utils.RetryAgain()
181 utils.Retry(_CheckNodeDaemon, 1.0, 10.0)
182 except utils.RetryTimeout:
183 raise errors.OpExecError("Node daemon on %s didn't answer queries within"
184 " 10 seconds" % node_name)
187 def _InitFileStorage(file_storage_dir):
188 """Initialize if needed the file storage.
190 @param file_storage_dir: the user-supplied value
191 @return: either empty string (if file storage was disabled at build
192 time) or the normalized path to the storage directory
195 if not constants.ENABLE_FILE_STORAGE:
198 file_storage_dir = os.path.normpath(file_storage_dir)
200 if not os.path.isabs(file_storage_dir):
201 raise errors.OpPrereqError("The file storage directory you passed is"
202 " not an absolute path.", errors.ECODE_INVAL)
204 if not os.path.exists(file_storage_dir):
206 os.makedirs(file_storage_dir, 0750)
208 raise errors.OpPrereqError("Cannot create file storage directory"
209 " '%s': %s" % (file_storage_dir, err),
210 errors.ECODE_ENVIRON)
212 if not os.path.isdir(file_storage_dir):
213 raise errors.OpPrereqError("The file storage directory '%s' is not"
214 " a directory." % file_storage_dir,
215 errors.ECODE_ENVIRON)
216 return file_storage_dir
219 #pylint: disable-msg=R0913
220 def InitCluster(cluster_name, mac_prefix,
221 master_netdev, file_storage_dir, candidate_pool_size,
222 secondary_ip=None, vg_name=None, beparams=None,
223 nicparams=None, hvparams=None, enabled_hypervisors=None,
224 modify_etc_hosts=True, modify_ssh_setup=True,
225 maintain_node_health=False, drbd_helper=None,
226 uid_pool=None, default_iallocator=None):
227 """Initialise the cluster.
229 @type candidate_pool_size: int
230 @param candidate_pool_size: master candidate pool size
233 # TODO: complete the docstring
234 if config.ConfigWriter.IsCluster():
235 raise errors.OpPrereqError("Cluster is already initialised",
238 if not enabled_hypervisors:
239 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
240 " least one member", errors.ECODE_INVAL)
241 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
243 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
244 " entries: %s" % invalid_hvs,
247 hostname = netutils.GetHostname()
249 if netutils.IP4Address.IsLoopback(hostname.ip):
250 raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback"
251 " address. Please fix DNS or %s." %
252 (hostname.ip, constants.ETC_HOSTS),
253 errors.ECODE_ENVIRON)
255 if not netutils.IPAddress.Own(hostname.ip):
256 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
257 " to %s,\nbut this ip address does not"
258 " belong to this host. Aborting." %
259 hostname.ip, errors.ECODE_ENVIRON)
261 clustername = netutils.GetHostname(name=cluster_name)
263 if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
265 raise errors.OpPrereqError("Cluster IP already active. Aborting.",
266 errors.ECODE_NOTUNIQUE)
269 if not netutils.IP4Address.IsValid(secondary_ip):
270 raise errors.OpPrereqError("Invalid secondary ip given",
272 if (secondary_ip != hostname.ip and
273 not netutils.IPAddress.Own(secondary_ip)):
274 raise errors.OpPrereqError("You gave %s as secondary IP,"
275 " but it does not belong to this host." %
276 secondary_ip, errors.ECODE_ENVIRON)
278 secondary_ip = hostname.ip
280 if vg_name is not None:
281 # Check if volume group is valid
282 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
283 constants.MIN_VG_SIZE)
285 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
286 " you are not using lvm" % vgstatus,
289 if drbd_helper is not None:
291 curr_helper = bdev.BaseDRBD.GetUsermodeHelper()
292 except errors.BlockDeviceError, err:
293 raise errors.OpPrereqError("Error while checking drbd helper"
294 " (specify --no-drbd-storage if you are not"
295 " using drbd): %s" % str(err),
296 errors.ECODE_ENVIRON)
297 if drbd_helper != curr_helper:
298 raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s"
299 " is the current helper" % (drbd_helper,
303 file_storage_dir = _InitFileStorage(file_storage_dir)
305 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
306 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
309 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
311 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
313 result.output.strip()), errors.ECODE_INVAL)
315 dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
316 utils.EnsureDirs(dirs)
318 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
319 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
320 objects.NIC.CheckParameterSyntax(nicparams)
322 # hvparams is a mapping of hypervisor->hvparams dict
323 for hv_name, hv_params in hvparams.iteritems():
324 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
325 hv_class = hypervisor.GetHypervisor(hv_name)
326 hv_class.CheckParameterSyntax(hv_params)
328 # set up ssh config and /etc/hosts
329 sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB)
330 sshkey = sshline.split(" ")[1]
333 utils.AddHostToEtcHosts(hostname.name)
338 if default_iallocator is not None:
339 alloc_script = utils.FindFile(default_iallocator,
340 constants.IALLOCATOR_SEARCH_PATH,
342 if alloc_script is None:
343 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
344 " specified" % default_iallocator,
349 # init of cluster config file
350 cluster_config = objects.Cluster(
352 rsahostkeypub=sshkey,
353 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
354 mac_prefix=mac_prefix,
355 volume_group_name=vg_name,
356 tcpudp_port_pool=set(),
357 master_node=hostname.name,
358 master_ip=clustername.ip,
359 master_netdev=master_netdev,
360 cluster_name=clustername.name,
361 file_storage_dir=file_storage_dir,
362 enabled_hypervisors=enabled_hypervisors,
363 beparams={constants.PP_DEFAULT: beparams},
364 nicparams={constants.PP_DEFAULT: nicparams},
366 candidate_pool_size=candidate_pool_size,
367 modify_etc_hosts=modify_etc_hosts,
368 modify_ssh_setup=modify_ssh_setup,
372 uuid=utils.NewUUID(),
373 maintain_node_health=maintain_node_health,
374 drbd_usermode_helper=drbd_helper,
375 default_iallocator=default_iallocator,
377 master_node_config = objects.Node(name=hostname.name,
378 primary_ip=hostname.ip,
379 secondary_ip=secondary_ip,
381 master_candidate=True,
382 offline=False, drained=False,
384 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
385 cfg = config.ConfigWriter(offline=True)
386 ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
387 cfg.Update(cfg.GetClusterInfo(), logging.error)
388 backend.WriteSsconfFiles(cfg.GetSsconfValues())
391 # set up the inter-node password and certificate
392 _InitGanetiServerSetup(hostname.name)
394 # start the master ip
395 # TODO: Review rpc call from bootstrap
396 # TODO: Warn on failed start master
397 rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
400 def InitConfig(version, cluster_config, master_node_config,
401 cfg_file=constants.CLUSTER_CONF_FILE):
402 """Create the initial cluster configuration.
404 It will contain the current node, which will also be the master
405 node, and no instances.
408 @param version: configuration version
409 @type cluster_config: L{objects.Cluster}
410 @param cluster_config: cluster configuration
411 @type master_node_config: L{objects.Node}
412 @param master_node_config: master node configuration
413 @type cfg_file: string
414 @param cfg_file: configuration file path
418 master_node_config.name: master_node_config,
422 config_data = objects.ConfigData(version=version,
423 cluster=cluster_config,
427 ctime=now, mtime=now)
428 utils.WriteFile(cfg_file,
429 data=serializer.Dump(config_data.ToDict()),
433 def FinalizeClusterDestroy(master):
434 """Execute the last steps of cluster destroy
436 This function shuts down all the daemons, completing the destroy
437 begun in cmdlib.LUDestroyOpcode.
440 cfg = config.ConfigWriter()
441 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
442 result = rpc.RpcRunner.call_node_stop_master(master, True)
443 msg = result.fail_msg
445 logging.warning("Could not disable the master role: %s", msg)
446 result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup)
447 msg = result.fail_msg
449 logging.warning("Could not shutdown the node daemon and cleanup"
450 " the node: %s", msg)
453 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
454 """Add a node to the cluster.
456 This function must be called before the actual opcode, and will ssh
457 to the remote node, copy the needed files, and start ganeti-noded,
458 allowing the master to do the rest via normal rpc calls.
460 @param cluster_name: the cluster name
461 @param node: the name of the new node
462 @param ssh_key_check: whether to do a strict key check
465 sshrunner = ssh.SshRunner(cluster_name)
467 noded_cert = utils.ReadFile(constants.NODED_CERT_FILE)
468 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
469 confd_hmac_key = utils.ReadFile(constants.CONFD_HMAC_KEY)
471 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
472 # so we use this to detect an invalid certificate; as long as the
473 # cert doesn't contain this, the here-document will be correctly
474 # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
475 # so the same restrictions apply.
476 for content in (noded_cert, rapi_cert, confd_hmac_key):
477 if re.search('^!EOF\.', content, re.MULTILINE):
478 raise errors.OpExecError("invalid SSL certificate or HMAC key")
480 if not noded_cert.endswith("\n"):
482 if not rapi_cert.endswith("\n"):
484 if not confd_hmac_key.endswith("\n"):
485 confd_hmac_key += "\n"
487 # set up inter-node password and certificate and restarts the node daemon
488 # and then connect with ssh to set password and start ganeti-noded
489 # note that all the below variables are sanitized at this point,
490 # either by being constants or by the checks above
491 # TODO: Could this command exceed a shell's maximum command length?
492 mycommand = ("umask 077 && "
493 "cat > '%s' << '!EOF.' && \n"
495 "cat > '%s' << '!EOF.' && \n"
497 "cat > '%s' << '!EOF.' && \n"
499 "chmod 0400 %s %s %s && "
501 (constants.NODED_CERT_FILE, noded_cert,
502 constants.RAPI_CERT_FILE, rapi_cert,
503 constants.CONFD_HMAC_KEY, confd_hmac_key,
504 constants.NODED_CERT_FILE, constants.RAPI_CERT_FILE,
505 constants.CONFD_HMAC_KEY,
506 constants.DAEMON_UTIL, constants.NODED))
508 result = sshrunner.Run(node, 'root', mycommand, batch=False,
509 ask_key=ssh_key_check,
510 use_cluster_key=False,
511 strict_host_check=ssh_key_check)
513 raise errors.OpExecError("Remote command on node %s, error: %s,"
515 (node, result.fail_reason, result.output))
517 _WaitForNodeDaemon(node)
520 def MasterFailover(no_voting=False):
521 """Failover the master node.
523 This checks that we are not already the master, and will cause the
524 current master to cease being master, and the non-master to become
527 @type no_voting: boolean
528 @param no_voting: force the operation without remote nodes agreement
532 sstore = ssconf.SimpleStore()
534 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
535 node_list = sstore.GetNodeList()
536 mc_list = sstore.GetMasterCandidates()
538 if old_master == new_master:
539 raise errors.OpPrereqError("This commands must be run on the node"
540 " where you want the new master to be."
541 " %s is already the master" %
542 old_master, errors.ECODE_INVAL)
544 if new_master not in mc_list:
545 mc_no_master = [name for name in mc_list if name != old_master]
546 raise errors.OpPrereqError("This node is not among the nodes marked"
547 " as master candidates. Only these nodes"
548 " can become masters. Current list of"
549 " master candidates is:\n"
550 "%s" % ('\n'.join(mc_no_master)),
554 vote_list = GatherMasterVotes(node_list)
557 voted_master = vote_list[0][0]
558 if voted_master is None:
559 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
560 " not respond.", errors.ECODE_ENVIRON)
561 elif voted_master != old_master:
562 raise errors.OpPrereqError("I have a wrong configuration, I believe"
563 " the master is %s but the other nodes"
564 " voted %s. Please resync the configuration"
566 (old_master, voted_master),
572 logging.info("Setting master to %s, old master: %s", new_master, old_master)
574 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
575 msg = result.fail_msg
577 logging.error("Could not disable the master role on the old master"
578 " %s, please disable manually: %s", old_master, msg)
580 master_ip = sstore.GetMasterIP()
582 # Here we have a phase where no master should be running
584 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
585 raise utils.RetryAgain()
588 utils.Retry(_check_ip, (1, 1.5, 5), total_timeout)
589 except utils.RetryTimeout:
590 logging.warning("The master IP is still reachable after %s seconds,"
591 " continuing but activating the master on the current"
592 " node will probably fail", total_timeout)
594 # instantiate a real config writer, as we now know we have the
596 cfg = config.ConfigWriter()
598 cluster_info = cfg.GetClusterInfo()
599 cluster_info.master_node = new_master
600 # this will also regenerate the ssconf files, since we updated the
602 cfg.Update(cluster_info, logging.error)
604 result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
605 msg = result.fail_msg
607 logging.error("Could not start the master role on the new master"
608 " %s, please check: %s", new_master, msg)
615 """Returns the current master node.
617 This is a separate function in bootstrap since it's needed by
618 gnt-cluster, and instead of importing directly ssconf, it's better
619 to abstract it in bootstrap, where we do use ssconf in other
623 sstore = ssconf.SimpleStore()
625 old_master, _ = ssconf.GetMasterAndMyself(sstore)
630 def GatherMasterVotes(node_list):
631 """Check the agreement on who is the master.
633 This function will return a list of (node, number of votes), ordered
634 by the number of votes. Errors will be denoted by the key 'None'.
636 Note that the sum of votes is the number of nodes this machine
637 knows, whereas the number of entries in the list could be different
638 (if some nodes vote for another master).
640 We remove ourselves from the list since we know that (bugs aside)
641 since we use the same source for configuration information for both
642 backend and boostrap, we'll always vote for ourselves.
644 @type node_list: list
645 @param node_list: the list of nodes to query for master info; the current
646 node will be removed if it is in the list
648 @return: list of (node, votes)
651 myself = netutils.Hostname.GetSysName()
653 node_list.remove(myself)
657 # no nodes left (eventually after removing myself)
659 results = rpc.RpcRunner.call_master_info(node_list)
660 if not isinstance(results, dict):
661 # this should not happen (unless internal error in rpc)
662 logging.critical("Can't complete rpc call, aborting master startup")
663 return [(None, len(node_list))]
671 logging.warning("Error contacting node %s: %s", node, msg)
673 elif not isinstance(data, (tuple, list)) or len(data) < 3:
674 logging.warning("Invalid data received from node %s: %s", node, data)
677 if None not in votes:
681 master_node = data[2]
682 if master_node not in votes:
683 votes[master_node] = 0
684 votes[master_node] += 1
686 vote_list = [v for v in votes.items()]
687 # sort first on number of votes then on name, since we want None
688 # sorted later if we have the half of the nodes not responding, and
689 # half voting all for the same master
690 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)