4 # Copyright (C) 2006, 2007, 2008 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
46 """Setup the SSH configuration for the cluster.
48 This generates a dsa keypair for root, adds the pub key to the
49 permitted hosts and adds the hostkey to its own known hosts.
52 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
54 for name in priv_key, pub_key:
55 if os.path.exists(name):
56 utils.CreateBackup(name)
57 utils.RemoveFile(name)
59 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
63 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
66 utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
69 def GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
70 """Generates a self-signed SSL certificate.
73 @param file_name: Path to output file
75 @param validity: Validity for certificate in days
78 (fd, tmp_file_name) = tempfile.mkstemp(dir=os.path.dirname(file_name))
81 # Set permissions before writing key
82 os.chmod(tmp_file_name, 0600)
84 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
85 "-days", str(validity), "-nodes", "-x509",
86 "-keyout", tmp_file_name, "-out", tmp_file_name,
89 raise errors.OpExecError("Could not generate SSL certificate, command"
90 " %s had exitcode %s and error message %s" %
91 (result.cmd, result.exit_code, result.output))
94 os.chmod(tmp_file_name, 0400)
96 os.rename(tmp_file_name, file_name)
98 utils.RemoveFile(tmp_file_name)
103 def GenerateHmacKey(file_name):
104 """Writes a new HMAC key.
107 @param file_name: Path to output file
110 utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400,
114 def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key,
116 nodecert_file=constants.NODED_CERT_FILE,
117 rapicert_file=constants.RAPI_CERT_FILE,
118 hmackey_file=constants.CONFD_HMAC_KEY):
119 """Updates the cluster certificates, keys and secrets.
121 @type new_cluster_cert: bool
122 @param new_cluster_cert: Whether to generate a new cluster certificate
123 @type new_rapi_cert: bool
124 @param new_rapi_cert: Whether to generate a new RAPI certificate
125 @type new_confd_hmac_key: bool
126 @param new_confd_hmac_key: Whether to generate a new HMAC key
127 @type rapi_cert_pem: string
128 @param rapi_cert_pem: New RAPI certificate in PEM format
129 @type nodecert_file: string
130 @param nodecert_file: optional override of the node cert file path
131 @type rapicert_file: string
132 @param rapicert_file: optional override of the rapi cert file path
133 @type hmackey_file: string
134 @param hmackey_file: optional override of the hmac key file path
137 # noded SSL certificate
138 cluster_cert_exists = os.path.exists(nodecert_file)
139 if new_cluster_cert or not cluster_cert_exists:
140 if cluster_cert_exists:
141 utils.CreateBackup(nodecert_file)
143 logging.debug("Generating new cluster certificate at %s", nodecert_file)
144 GenerateSelfSignedSslCert(nodecert_file)
147 if new_confd_hmac_key or not os.path.exists(hmackey_file):
148 logging.debug("Writing new confd HMAC key to %s", hmackey_file)
149 GenerateHmacKey(hmackey_file)
152 rapi_cert_exists = os.path.exists(rapicert_file)
155 # Assume rapi_pem contains a valid PEM-formatted certificate and key
156 logging.debug("Writing RAPI certificate at %s", rapicert_file)
157 utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
159 elif new_rapi_cert or not rapi_cert_exists:
161 utils.CreateBackup(rapicert_file)
163 logging.debug("Generating new RAPI certificate at %s", rapicert_file)
164 GenerateSelfSignedSslCert(rapicert_file)
167 def _InitGanetiServerSetup(master_name):
168 """Setup the necessary configuration for the initial node daemon.
170 This creates the nodepass file containing the shared password for
171 the cluster and also generates the SSL certificate.
174 # Generate cluster secrets
175 GenerateClusterCrypto(True, False, False)
177 result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED])
179 raise errors.OpExecError("Could not start the node daemon, command %s"
180 " had exitcode %s and error %s" %
181 (result.cmd, result.exit_code, result.output))
183 _WaitForNodeDaemon(master_name)
186 def _WaitForNodeDaemon(node_name):
187 """Wait for node daemon to become responsive.
190 def _CheckNodeDaemon():
191 result = rpc.RpcRunner.call_version([node_name])[node_name]
193 raise utils.RetryAgain()
196 utils.Retry(_CheckNodeDaemon, 1.0, 10.0)
197 except utils.RetryTimeout:
198 raise errors.OpExecError("Node daemon on %s didn't answer queries within"
199 " 10 seconds" % node_name)
202 def _InitFileStorage(file_storage_dir):
203 """Initialize if needed the file storage.
205 @param file_storage_dir: the user-supplied value
206 @return: either empty string (if file storage was disabled at build
207 time) or the normalized path to the storage directory
210 if not constants.ENABLE_FILE_STORAGE:
213 file_storage_dir = os.path.normpath(file_storage_dir)
215 if not os.path.isabs(file_storage_dir):
216 raise errors.OpPrereqError("The file storage directory you passed is"
217 " not an absolute path.", errors.ECODE_INVAL)
219 if not os.path.exists(file_storage_dir):
221 os.makedirs(file_storage_dir, 0750)
223 raise errors.OpPrereqError("Cannot create file storage directory"
224 " '%s': %s" % (file_storage_dir, err),
225 errors.ECODE_ENVIRON)
227 if not os.path.isdir(file_storage_dir):
228 raise errors.OpPrereqError("The file storage directory '%s' is not"
229 " a directory." % file_storage_dir,
230 errors.ECODE_ENVIRON)
231 return file_storage_dir
234 def InitCluster(cluster_name, mac_prefix,
235 master_netdev, file_storage_dir, candidate_pool_size,
236 secondary_ip=None, vg_name=None, beparams=None,
237 nicparams=None, hvparams=None, enabled_hypervisors=None,
238 modify_etc_hosts=True, modify_ssh_setup=True,
239 maintain_node_health=False,
241 """Initialise the cluster.
243 @type candidate_pool_size: int
244 @param candidate_pool_size: master candidate pool size
247 # TODO: complete the docstring
248 if config.ConfigWriter.IsCluster():
249 raise errors.OpPrereqError("Cluster is already initialised",
252 if not enabled_hypervisors:
253 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
254 " least one member", errors.ECODE_INVAL)
255 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
257 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
258 " entries: %s" % invalid_hvs,
261 hostname = utils.GetHostInfo()
263 if hostname.ip.startswith("127."):
264 raise errors.OpPrereqError("This host's IP resolves to the private"
265 " range (%s). Please fix DNS or %s." %
266 (hostname.ip, constants.ETC_HOSTS),
267 errors.ECODE_ENVIRON)
269 if not utils.OwnIpAddress(hostname.ip):
270 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
271 " to %s,\nbut this ip address does not"
272 " belong to this host. Aborting." %
273 hostname.ip, errors.ECODE_ENVIRON)
275 clustername = utils.GetHostInfo(utils.HostInfo.NormalizeName(cluster_name))
277 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
279 raise errors.OpPrereqError("Cluster IP already active. Aborting.",
280 errors.ECODE_NOTUNIQUE)
283 if not utils.IsValidIP(secondary_ip):
284 raise errors.OpPrereqError("Invalid secondary ip given",
286 if (secondary_ip != hostname.ip and
287 not utils.OwnIpAddress(secondary_ip)):
288 raise errors.OpPrereqError("You gave %s as secondary IP,"
289 " but it does not belong to this host." %
290 secondary_ip, errors.ECODE_ENVIRON)
292 secondary_ip = hostname.ip
294 if vg_name is not None:
295 # Check if volume group is valid
296 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
297 constants.MIN_VG_SIZE)
299 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
300 " you are not using lvm" % vgstatus,
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 the inter-node password and certificate
329 _InitGanetiServerSetup(hostname.name)
331 # set up ssh config and /etc/hosts
332 sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB)
333 sshkey = sshline.split(" ")[1]
336 utils.AddHostToEtcHosts(hostname.name)
343 # init of cluster config file
344 cluster_config = objects.Cluster(
346 rsahostkeypub=sshkey,
347 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
348 mac_prefix=mac_prefix,
349 volume_group_name=vg_name,
350 tcpudp_port_pool=set(),
351 master_node=hostname.name,
352 master_ip=clustername.ip,
353 master_netdev=master_netdev,
354 cluster_name=clustername.name,
355 file_storage_dir=file_storage_dir,
356 enabled_hypervisors=enabled_hypervisors,
357 beparams={constants.PP_DEFAULT: beparams},
358 nicparams={constants.PP_DEFAULT: nicparams},
360 candidate_pool_size=candidate_pool_size,
361 modify_etc_hosts=modify_etc_hosts,
362 modify_ssh_setup=modify_ssh_setup,
366 uuid=utils.NewUUID(),
367 maintain_node_health=maintain_node_health,
369 master_node_config = objects.Node(name=hostname.name,
370 primary_ip=hostname.ip,
371 secondary_ip=secondary_ip,
373 master_candidate=True,
374 offline=False, drained=False,
376 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
377 cfg = config.ConfigWriter()
378 ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
379 cfg.Update(cfg.GetClusterInfo(), logging.error)
381 # start the master ip
382 # TODO: Review rpc call from bootstrap
383 # TODO: Warn on failed start master
384 rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
387 def InitConfig(version, cluster_config, master_node_config,
388 cfg_file=constants.CLUSTER_CONF_FILE):
389 """Create the initial cluster configuration.
391 It will contain the current node, which will also be the master
392 node, and no instances.
395 @param version: configuration version
396 @type cluster_config: L{objects.Cluster}
397 @param cluster_config: cluster configuration
398 @type master_node_config: L{objects.Node}
399 @param master_node_config: master node configuration
400 @type cfg_file: string
401 @param cfg_file: configuration file path
405 master_node_config.name: master_node_config,
409 config_data = objects.ConfigData(version=version,
410 cluster=cluster_config,
414 ctime=now, mtime=now)
415 utils.WriteFile(cfg_file,
416 data=serializer.Dump(config_data.ToDict()),
420 def FinalizeClusterDestroy(master):
421 """Execute the last steps of cluster destroy
423 This function shuts down all the daemons, completing the destroy
424 begun in cmdlib.LUDestroyOpcode.
427 cfg = config.ConfigWriter()
428 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
429 result = rpc.RpcRunner.call_node_stop_master(master, True)
430 msg = result.fail_msg
432 logging.warning("Could not disable the master role: %s", msg)
433 result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup)
434 msg = result.fail_msg
436 logging.warning("Could not shutdown the node daemon and cleanup"
437 " the node: %s", msg)
440 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
441 """Add a node to the cluster.
443 This function must be called before the actual opcode, and will ssh
444 to the remote node, copy the needed files, and start ganeti-noded,
445 allowing the master to do the rest via normal rpc calls.
447 @param cluster_name: the cluster name
448 @param node: the name of the new node
449 @param ssh_key_check: whether to do a strict key check
452 sshrunner = ssh.SshRunner(cluster_name)
454 noded_cert = utils.ReadFile(constants.NODED_CERT_FILE)
455 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
456 confd_hmac_key = utils.ReadFile(constants.CONFD_HMAC_KEY)
458 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
459 # so we use this to detect an invalid certificate; as long as the
460 # cert doesn't contain this, the here-document will be correctly
461 # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
462 # so the same restrictions apply.
463 for content in (noded_cert, rapi_cert, confd_hmac_key):
464 if re.search('^!EOF\.', content, re.MULTILINE):
465 raise errors.OpExecError("invalid SSL certificate or HMAC key")
467 if not noded_cert.endswith("\n"):
469 if not rapi_cert.endswith("\n"):
471 if not confd_hmac_key.endswith("\n"):
472 confd_hmac_key += "\n"
474 # set up inter-node password and certificate and restarts the node daemon
475 # and then connect with ssh to set password and start ganeti-noded
476 # note that all the below variables are sanitized at this point,
477 # either by being constants or by the checks above
478 mycommand = ("umask 077 && "
479 "cat > '%s' << '!EOF.' && \n"
481 "cat > '%s' << '!EOF.' && \n"
483 "cat > '%s' << '!EOF.' && \n"
485 "chmod 0400 %s %s %s && "
487 (constants.NODED_CERT_FILE, noded_cert,
488 constants.RAPI_CERT_FILE, rapi_cert,
489 constants.CONFD_HMAC_KEY, confd_hmac_key,
490 constants.NODED_CERT_FILE, constants.RAPI_CERT_FILE,
491 constants.CONFD_HMAC_KEY,
492 constants.DAEMON_UTIL, constants.NODED))
494 result = sshrunner.Run(node, 'root', mycommand, batch=False,
495 ask_key=ssh_key_check,
496 use_cluster_key=False,
497 strict_host_check=ssh_key_check)
499 raise errors.OpExecError("Remote command on node %s, error: %s,"
501 (node, result.fail_reason, result.output))
503 _WaitForNodeDaemon(node)
506 def MasterFailover(no_voting=False):
507 """Failover the master node.
509 This checks that we are not already the master, and will cause the
510 current master to cease being master, and the non-master to become
513 @type no_voting: boolean
514 @param no_voting: force the operation without remote nodes agreement
518 sstore = ssconf.SimpleStore()
520 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
521 node_list = sstore.GetNodeList()
522 mc_list = sstore.GetMasterCandidates()
524 if old_master == new_master:
525 raise errors.OpPrereqError("This commands must be run on the node"
526 " where you want the new master to be."
527 " %s is already the master" %
528 old_master, errors.ECODE_INVAL)
530 if new_master not in mc_list:
531 mc_no_master = [name for name in mc_list if name != old_master]
532 raise errors.OpPrereqError("This node is not among the nodes marked"
533 " as master candidates. Only these nodes"
534 " can become masters. Current list of"
535 " master candidates is:\n"
536 "%s" % ('\n'.join(mc_no_master)),
540 vote_list = GatherMasterVotes(node_list)
543 voted_master = vote_list[0][0]
544 if voted_master is None:
545 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
546 " not respond.", errors.ECODE_ENVIRON)
547 elif voted_master != old_master:
548 raise errors.OpPrereqError("I have a wrong configuration, I believe"
549 " the master is %s but the other nodes"
550 " voted %s. Please resync the configuration"
552 (old_master, voted_master),
558 logging.info("Setting master to %s, old master: %s", new_master, old_master)
560 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
561 msg = result.fail_msg
563 logging.error("Could not disable the master role on the old master"
564 " %s, please disable manually: %s", old_master, msg)
566 # Here we have a phase where no master should be running
568 # instantiate a real config writer, as we now know we have the
570 cfg = config.ConfigWriter()
572 cluster_info = cfg.GetClusterInfo()
573 cluster_info.master_node = new_master
574 # this will also regenerate the ssconf files, since we updated the
576 cfg.Update(cluster_info, logging.error)
578 result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
579 msg = result.fail_msg
581 logging.error("Could not start the master role on the new master"
582 " %s, please check: %s", new_master, msg)
589 """Returns the current master node.
591 This is a separate function in bootstrap since it's needed by
592 gnt-cluster, and instead of importing directly ssconf, it's better
593 to abstract it in bootstrap, where we do use ssconf in other
597 sstore = ssconf.SimpleStore()
599 old_master, _ = ssconf.GetMasterAndMyself(sstore)
604 def GatherMasterVotes(node_list):
605 """Check the agreement on who is the master.
607 This function will return a list of (node, number of votes), ordered
608 by the number of votes. Errors will be denoted by the key 'None'.
610 Note that the sum of votes is the number of nodes this machine
611 knows, whereas the number of entries in the list could be different
612 (if some nodes vote for another master).
614 We remove ourselves from the list since we know that (bugs aside)
615 since we use the same source for configuration information for both
616 backend and boostrap, we'll always vote for ourselves.
618 @type node_list: list
619 @param node_list: the list of nodes to query for master info; the current
620 node will be removed if it is in the list
622 @return: list of (node, votes)
625 myself = utils.HostInfo().name
627 node_list.remove(myself)
631 # no nodes left (eventually after removing myself)
633 results = rpc.RpcRunner.call_master_info(node_list)
634 if not isinstance(results, dict):
635 # this should not happen (unless internal error in rpc)
636 logging.critical("Can't complete rpc call, aborting master startup")
637 return [(None, len(node_list))]
645 logging.warning("Error contacting node %s: %s", node, msg)
647 elif not isinstance(data, (tuple, list)) or len(data) < 3:
648 logging.warning("Invalid data received from node %s: %s", node, data)
651 if None not in votes:
655 master_node = data[2]
656 if master_node not in votes:
657 votes[master_node] = 0
658 votes[master_node] += 1
660 vote_list = [v for v in votes.items()]
661 # sort first on number of votes then on name, since we want None
662 # sorted later if we have the half of the nodes not responding, and
663 # half voting all for the same master
664 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)