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.
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
45 """Setup the SSH configuration for the cluster.
47 This generates a dsa keypair for root, adds the pub key to the
48 permitted hosts and adds the hostkey to its own known hosts.
51 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
53 for name in priv_key, pub_key:
54 if os.path.exists(name):
55 utils.CreateBackup(name)
56 utils.RemoveFile(name)
58 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
62 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
65 utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
68 def GenerateHmacKey(file_name):
69 """Writes a new HMAC key.
72 @param file_name: Path to output file
75 utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400,
79 def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key,
80 new_cds, rapi_cert_pem=None, cds=None):
81 """Updates the cluster certificates, keys and secrets.
83 @type new_cluster_cert: bool
84 @param new_cluster_cert: Whether to generate a new cluster certificate
85 @type new_rapi_cert: bool
86 @param new_rapi_cert: Whether to generate a new RAPI certificate
87 @type new_confd_hmac_key: bool
88 @param new_confd_hmac_key: Whether to generate a new HMAC key
90 @param new_cds: Whether to generate a new cluster domain secret
91 @type rapi_cert_pem: string
92 @param rapi_cert_pem: New RAPI certificate in PEM format
94 @param cds: New cluster domain secret
97 # noded SSL certificate
98 cluster_cert_exists = os.path.exists(constants.NODED_CERT_FILE)
99 if new_cluster_cert or not cluster_cert_exists:
100 if cluster_cert_exists:
101 utils.CreateBackup(constants.NODED_CERT_FILE)
103 logging.debug("Generating new cluster certificate at %s",
104 constants.NODED_CERT_FILE)
105 utils.GenerateSelfSignedSslCert(constants.NODED_CERT_FILE)
108 if new_confd_hmac_key or not os.path.exists(constants.CONFD_HMAC_KEY):
109 logging.debug("Writing new confd HMAC key to %s", constants.CONFD_HMAC_KEY)
110 GenerateHmacKey(constants.CONFD_HMAC_KEY)
113 rapi_cert_exists = os.path.exists(constants.RAPI_CERT_FILE)
116 # Assume rapi_pem contains a valid PEM-formatted certificate and key
117 logging.debug("Writing RAPI certificate at %s",
118 constants.RAPI_CERT_FILE)
119 utils.WriteFile(constants.RAPI_CERT_FILE, data=rapi_cert_pem, backup=True)
121 elif new_rapi_cert or not rapi_cert_exists:
123 utils.CreateBackup(constants.RAPI_CERT_FILE)
125 logging.debug("Generating new RAPI certificate at %s",
126 constants.RAPI_CERT_FILE)
127 utils.GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
129 # Cluster domain secret
131 logging.debug("Writing cluster domain secret to %s",
132 constants.CLUSTER_DOMAIN_SECRET_FILE)
133 utils.WriteFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
134 data=cds, backup=True)
136 elif new_cds or not os.path.exists(constants.CLUSTER_DOMAIN_SECRET_FILE):
137 logging.debug("Generating new cluster domain secret at %s",
138 constants.CLUSTER_DOMAIN_SECRET_FILE)
139 GenerateHmacKey(constants.CLUSTER_DOMAIN_SECRET_FILE)
142 def _InitGanetiServerSetup(master_name):
143 """Setup the necessary configuration for the initial node daemon.
145 This creates the nodepass file containing the shared password for
146 the cluster and also generates the SSL certificate.
149 # Generate cluster secrets
150 GenerateClusterCrypto(True, False, False, False)
152 result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED])
154 raise errors.OpExecError("Could not start the node daemon, command %s"
155 " had exitcode %s and error %s" %
156 (result.cmd, result.exit_code, result.output))
158 _WaitForNodeDaemon(master_name)
161 def _WaitForNodeDaemon(node_name):
162 """Wait for node daemon to become responsive.
165 def _CheckNodeDaemon():
166 result = rpc.RpcRunner.call_version([node_name])[node_name]
168 raise utils.RetryAgain()
171 utils.Retry(_CheckNodeDaemon, 1.0, 10.0)
172 except utils.RetryTimeout:
173 raise errors.OpExecError("Node daemon on %s didn't answer queries within"
174 " 10 seconds" % node_name)
177 def InitCluster(cluster_name, mac_prefix,
178 master_netdev, file_storage_dir, candidate_pool_size,
179 secondary_ip=None, vg_name=None, beparams=None,
180 nicparams=None, hvparams=None, enabled_hypervisors=None,
181 modify_etc_hosts=True, modify_ssh_setup=True):
182 """Initialise the cluster.
184 @type candidate_pool_size: int
185 @param candidate_pool_size: master candidate pool size
188 # TODO: complete the docstring
189 if config.ConfigWriter.IsCluster():
190 raise errors.OpPrereqError("Cluster is already initialised",
193 if not enabled_hypervisors:
194 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
195 " least one member", errors.ECODE_INVAL)
196 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
198 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
199 " entries: %s" % invalid_hvs,
202 hostname = utils.GetHostInfo()
204 if hostname.ip.startswith("127."):
205 raise errors.OpPrereqError("This host's IP resolves to the private"
206 " range (%s). Please fix DNS or %s." %
207 (hostname.ip, constants.ETC_HOSTS),
208 errors.ECODE_ENVIRON)
210 if not utils.OwnIpAddress(hostname.ip):
211 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
212 " to %s,\nbut this ip address does not"
213 " belong to this host. Aborting." %
214 hostname.ip, errors.ECODE_ENVIRON)
216 clustername = utils.GetHostInfo(utils.HostInfo.NormalizeName(cluster_name))
218 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
220 raise errors.OpPrereqError("Cluster IP already active. Aborting.",
221 errors.ECODE_NOTUNIQUE)
224 if not utils.IsValidIP(secondary_ip):
225 raise errors.OpPrereqError("Invalid secondary ip given",
227 if (secondary_ip != hostname.ip and
228 not utils.OwnIpAddress(secondary_ip)):
229 raise errors.OpPrereqError("You gave %s as secondary IP,"
230 " but it does not belong to this host." %
231 secondary_ip, errors.ECODE_ENVIRON)
233 secondary_ip = hostname.ip
235 if vg_name is not None:
236 # Check if volume group is valid
237 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
238 constants.MIN_VG_SIZE)
240 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
241 " you are not using lvm" % vgstatus,
244 file_storage_dir = os.path.normpath(file_storage_dir)
246 if not os.path.isabs(file_storage_dir):
247 raise errors.OpPrereqError("The file storage directory you passed is"
248 " not an absolute path.", errors.ECODE_INVAL)
250 if not os.path.exists(file_storage_dir):
252 os.makedirs(file_storage_dir, 0750)
254 raise errors.OpPrereqError("Cannot create file storage directory"
255 " '%s': %s" % (file_storage_dir, err),
256 errors.ECODE_ENVIRON)
258 if not os.path.isdir(file_storage_dir):
259 raise errors.OpPrereqError("The file storage directory '%s' is not"
260 " a directory." % file_storage_dir,
261 errors.ECODE_ENVIRON)
263 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
264 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
267 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
269 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
271 result.output.strip()), errors.ECODE_INVAL)
273 dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
274 utils.EnsureDirs(dirs)
276 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
277 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
278 objects.NIC.CheckParameterSyntax(nicparams)
280 # hvparams is a mapping of hypervisor->hvparams dict
281 for hv_name, hv_params in hvparams.iteritems():
282 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
283 hv_class = hypervisor.GetHypervisor(hv_name)
284 hv_class.CheckParameterSyntax(hv_params)
286 # set up the inter-node password and certificate
287 _InitGanetiServerSetup(hostname.name)
289 # set up ssh config and /etc/hosts
290 sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB)
291 sshkey = sshline.split(" ")[1]
294 utils.AddHostToEtcHosts(hostname.name)
301 # init of cluster config file
302 cluster_config = objects.Cluster(
304 rsahostkeypub=sshkey,
305 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
306 mac_prefix=mac_prefix,
307 volume_group_name=vg_name,
308 tcpudp_port_pool=set(),
309 master_node=hostname.name,
310 master_ip=clustername.ip,
311 master_netdev=master_netdev,
312 cluster_name=clustername.name,
313 file_storage_dir=file_storage_dir,
314 enabled_hypervisors=enabled_hypervisors,
315 beparams={constants.PP_DEFAULT: beparams},
316 nicparams={constants.PP_DEFAULT: nicparams},
318 candidate_pool_size=candidate_pool_size,
319 modify_etc_hosts=modify_etc_hosts,
320 modify_ssh_setup=modify_ssh_setup,
323 uuid=utils.NewUUID(),
325 master_node_config = objects.Node(name=hostname.name,
326 primary_ip=hostname.ip,
327 secondary_ip=secondary_ip,
329 master_candidate=True,
330 offline=False, drained=False,
332 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
333 cfg = config.ConfigWriter()
334 ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
335 cfg.Update(cfg.GetClusterInfo(), logging.error)
337 # start the master ip
338 # TODO: Review rpc call from bootstrap
339 # TODO: Warn on failed start master
340 rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
343 def InitConfig(version, cluster_config, master_node_config,
344 cfg_file=constants.CLUSTER_CONF_FILE):
345 """Create the initial cluster configuration.
347 It will contain the current node, which will also be the master
348 node, and no instances.
351 @param version: configuration version
352 @type cluster_config: L{objects.Cluster}
353 @param cluster_config: cluster configuration
354 @type master_node_config: L{objects.Node}
355 @param master_node_config: master node configuration
356 @type cfg_file: string
357 @param cfg_file: configuration file path
361 master_node_config.name: master_node_config,
365 config_data = objects.ConfigData(version=version,
366 cluster=cluster_config,
370 ctime=now, mtime=now)
371 utils.WriteFile(cfg_file,
372 data=serializer.Dump(config_data.ToDict()),
376 def FinalizeClusterDestroy(master):
377 """Execute the last steps of cluster destroy
379 This function shuts down all the daemons, completing the destroy
380 begun in cmdlib.LUDestroyOpcode.
383 cfg = config.ConfigWriter()
384 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
385 result = rpc.RpcRunner.call_node_stop_master(master, True)
386 msg = result.fail_msg
388 logging.warning("Could not disable the master role: %s", msg)
389 result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup)
390 msg = result.fail_msg
392 logging.warning("Could not shutdown the node daemon and cleanup"
393 " the node: %s", msg)
396 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
397 """Add a node to the cluster.
399 This function must be called before the actual opcode, and will ssh
400 to the remote node, copy the needed files, and start ganeti-noded,
401 allowing the master to do the rest via normal rpc calls.
403 @param cluster_name: the cluster name
404 @param node: the name of the new node
405 @param ssh_key_check: whether to do a strict key check
408 sshrunner = ssh.SshRunner(cluster_name)
410 noded_cert = utils.ReadFile(constants.NODED_CERT_FILE)
411 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
412 confd_hmac_key = utils.ReadFile(constants.CONFD_HMAC_KEY)
414 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
415 # so we use this to detect an invalid certificate; as long as the
416 # cert doesn't contain this, the here-document will be correctly
417 # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
418 # so the same restrictions apply.
419 for content in (noded_cert, rapi_cert, confd_hmac_key):
420 if re.search('^!EOF\.', content, re.MULTILINE):
421 raise errors.OpExecError("invalid SSL certificate or HMAC key")
423 if not noded_cert.endswith("\n"):
425 if not rapi_cert.endswith("\n"):
427 if not confd_hmac_key.endswith("\n"):
428 confd_hmac_key += "\n"
430 # set up inter-node password and certificate and restarts the node daemon
431 # and then connect with ssh to set password and start ganeti-noded
432 # note that all the below variables are sanitized at this point,
433 # either by being constants or by the checks above
434 # TODO: Could this command exceed a shell's maximum command length?
435 mycommand = ("umask 077 && "
436 "cat > '%s' << '!EOF.' && \n"
438 "cat > '%s' << '!EOF.' && \n"
440 "cat > '%s' << '!EOF.' && \n"
442 "chmod 0400 %s %s %s && "
444 (constants.NODED_CERT_FILE, noded_cert,
445 constants.RAPI_CERT_FILE, rapi_cert,
446 constants.CONFD_HMAC_KEY, confd_hmac_key,
447 constants.NODED_CERT_FILE, constants.RAPI_CERT_FILE,
448 constants.CONFD_HMAC_KEY,
449 constants.DAEMON_UTIL, constants.NODED))
451 result = sshrunner.Run(node, 'root', mycommand, batch=False,
452 ask_key=ssh_key_check,
453 use_cluster_key=False,
454 strict_host_check=ssh_key_check)
456 raise errors.OpExecError("Remote command on node %s, error: %s,"
458 (node, result.fail_reason, result.output))
460 _WaitForNodeDaemon(node)
463 def MasterFailover(no_voting=False):
464 """Failover the master node.
466 This checks that we are not already the master, and will cause the
467 current master to cease being master, and the non-master to become
470 @type no_voting: boolean
471 @param no_voting: force the operation without remote nodes agreement
475 sstore = ssconf.SimpleStore()
477 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
478 node_list = sstore.GetNodeList()
479 mc_list = sstore.GetMasterCandidates()
481 if old_master == new_master:
482 raise errors.OpPrereqError("This commands must be run on the node"
483 " where you want the new master to be."
484 " %s is already the master" %
485 old_master, errors.ECODE_INVAL)
487 if new_master not in mc_list:
488 mc_no_master = [name for name in mc_list if name != old_master]
489 raise errors.OpPrereqError("This node is not among the nodes marked"
490 " as master candidates. Only these nodes"
491 " can become masters. Current list of"
492 " master candidates is:\n"
493 "%s" % ('\n'.join(mc_no_master)),
497 vote_list = GatherMasterVotes(node_list)
500 voted_master = vote_list[0][0]
501 if voted_master is None:
502 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
503 " not respond.", errors.ECODE_ENVIRON)
504 elif voted_master != old_master:
505 raise errors.OpPrereqError("I have a wrong configuration, I believe"
506 " the master is %s but the other nodes"
507 " voted %s. Please resync the configuration"
509 (old_master, voted_master),
515 logging.info("Setting master to %s, old master: %s", new_master, old_master)
517 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
518 msg = result.fail_msg
520 logging.error("Could not disable the master role on the old master"
521 " %s, please disable manually: %s", old_master, msg)
523 # Here we have a phase where no master should be running
525 # instantiate a real config writer, as we now know we have the
527 cfg = config.ConfigWriter()
529 cluster_info = cfg.GetClusterInfo()
530 cluster_info.master_node = new_master
531 # this will also regenerate the ssconf files, since we updated the
533 cfg.Update(cluster_info, logging.error)
535 result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
536 msg = result.fail_msg
538 logging.error("Could not start the master role on the new master"
539 " %s, please check: %s", new_master, msg)
546 """Returns the current master node.
548 This is a separate function in bootstrap since it's needed by
549 gnt-cluster, and instead of importing directly ssconf, it's better
550 to abstract it in bootstrap, where we do use ssconf in other
554 sstore = ssconf.SimpleStore()
556 old_master, _ = ssconf.GetMasterAndMyself(sstore)
561 def GatherMasterVotes(node_list):
562 """Check the agreement on who is the master.
564 This function will return a list of (node, number of votes), ordered
565 by the number of votes. Errors will be denoted by the key 'None'.
567 Note that the sum of votes is the number of nodes this machine
568 knows, whereas the number of entries in the list could be different
569 (if some nodes vote for another master).
571 We remove ourselves from the list since we know that (bugs aside)
572 since we use the same source for configuration information for both
573 backend and boostrap, we'll always vote for ourselves.
575 @type node_list: list
576 @param node_list: the list of nodes to query for master info; the current
577 node will be removed if it is in the list
579 @return: list of (node, votes)
582 myself = utils.HostInfo().name
584 node_list.remove(myself)
588 # no nodes left (eventually after removing myself)
590 results = rpc.RpcRunner.call_master_info(node_list)
591 if not isinstance(results, dict):
592 # this should not happen (unless internal error in rpc)
593 logging.critical("Can't complete rpc call, aborting master startup")
594 return [(None, len(node_list))]
602 logging.warning("Error contacting node %s: %s", node, msg)
604 elif not isinstance(data, (tuple, list)) or len(data) < 3:
605 logging.warning("Invalid data received from node %s: %s", node, data)
608 if None not in votes:
612 master_node = data[2]
613 if master_node not in votes:
614 votes[master_node] = 0
615 votes[master_node] += 1
617 vote_list = [v for v in votes.items()]
618 # sort first on number of votes then on name, since we want None
619 # sorted later if we have the half of the nodes not responding, and
620 # half voting all for the same master
621 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)