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)
113 def _InitGanetiServerSetup(master_name):
114 """Setup the necessary configuration for the initial node daemon.
116 This creates the nodepass file containing the shared password for
117 the cluster and also generates the SSL certificate.
120 GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
122 # Don't overwrite existing file
123 if not os.path.exists(constants.RAPI_CERT_FILE):
124 GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
126 if not os.path.exists(constants.HMAC_CLUSTER_KEY):
127 GenerateHmacKey(constants.HMAC_CLUSTER_KEY)
129 result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED])
131 raise errors.OpExecError("Could not start the node daemon, command %s"
132 " had exitcode %s and error %s" %
133 (result.cmd, result.exit_code, result.output))
135 # Wait for node daemon to become responsive
136 def _CheckNodeDaemon():
137 result = rpc.RpcRunner.call_version([master_name])[master_name]
139 raise utils.RetryAgain()
142 utils.Retry(_CheckNodeDaemon, 1.0, 10.0)
143 except utils.RetryTimeout:
144 raise errors.OpExecError("Node daemon didn't answer queries within"
147 def InitCluster(cluster_name, mac_prefix,
148 master_netdev, file_storage_dir, candidate_pool_size,
149 secondary_ip=None, vg_name=None, beparams=None,
150 nicparams=None, hvparams=None, enabled_hypervisors=None,
151 modify_etc_hosts=True, modify_ssh_setup=True):
152 """Initialise the cluster.
154 @type candidate_pool_size: int
155 @param candidate_pool_size: master candidate pool size
158 # TODO: complete the docstring
159 if config.ConfigWriter.IsCluster():
160 raise errors.OpPrereqError("Cluster is already initialised",
163 if not enabled_hypervisors:
164 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
165 " least one member", errors.ECODE_INVAL)
166 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
168 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
169 " entries: %s" % invalid_hvs,
172 hostname = utils.GetHostInfo()
174 if hostname.ip.startswith("127."):
175 raise errors.OpPrereqError("This host's IP resolves to the private"
176 " range (%s). Please fix DNS or %s." %
177 (hostname.ip, constants.ETC_HOSTS),
178 errors.ECODE_ENVIRON)
180 if not utils.OwnIpAddress(hostname.ip):
181 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
182 " to %s,\nbut this ip address does not"
183 " belong to this host. Aborting." %
184 hostname.ip, errors.ECODE_ENVIRON)
186 clustername = utils.GetHostInfo(cluster_name)
188 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
190 raise errors.OpPrereqError("Cluster IP already active. Aborting.",
191 errors.ECODE_NOTUNIQUE)
194 if not utils.IsValidIP(secondary_ip):
195 raise errors.OpPrereqError("Invalid secondary ip given",
197 if (secondary_ip != hostname.ip and
198 not utils.OwnIpAddress(secondary_ip)):
199 raise errors.OpPrereqError("You gave %s as secondary IP,"
200 " but it does not belong to this host." %
201 secondary_ip, errors.ECODE_ENVIRON)
203 secondary_ip = hostname.ip
205 if vg_name is not None:
206 # Check if volume group is valid
207 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
208 constants.MIN_VG_SIZE)
210 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
211 " you are not using lvm" % vgstatus,
214 file_storage_dir = os.path.normpath(file_storage_dir)
216 if not os.path.isabs(file_storage_dir):
217 raise errors.OpPrereqError("The file storage directory you passed is"
218 " not an absolute path.", errors.ECODE_INVAL)
220 if not os.path.exists(file_storage_dir):
222 os.makedirs(file_storage_dir, 0750)
224 raise errors.OpPrereqError("Cannot create file storage directory"
225 " '%s': %s" % (file_storage_dir, err),
226 errors.ECODE_ENVIRON)
228 if not os.path.isdir(file_storage_dir):
229 raise errors.OpPrereqError("The file storage directory '%s' is not"
230 " a directory." % file_storage_dir,
231 errors.ECODE_ENVIRON)
233 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
234 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
237 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
239 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
241 result.output.strip()), errors.ECODE_INVAL)
243 dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
244 utils.EnsureDirs(dirs)
246 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
247 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
248 objects.NIC.CheckParameterSyntax(nicparams)
250 # hvparams is a mapping of hypervisor->hvparams dict
251 for hv_name, hv_params in hvparams.iteritems():
252 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
253 hv_class = hypervisor.GetHypervisor(hv_name)
254 hv_class.CheckParameterSyntax(hv_params)
256 # set up the inter-node password and certificate
257 _InitGanetiServerSetup(hostname.name)
259 # set up ssh config and /etc/hosts
260 sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB)
261 sshkey = sshline.split(" ")[1]
264 utils.AddHostToEtcHosts(hostname.name)
271 # init of cluster config file
272 cluster_config = objects.Cluster(
274 rsahostkeypub=sshkey,
275 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
276 mac_prefix=mac_prefix,
277 volume_group_name=vg_name,
278 tcpudp_port_pool=set(),
279 master_node=hostname.name,
280 master_ip=clustername.ip,
281 master_netdev=master_netdev,
282 cluster_name=clustername.name,
283 file_storage_dir=file_storage_dir,
284 enabled_hypervisors=enabled_hypervisors,
285 beparams={constants.PP_DEFAULT: beparams},
286 nicparams={constants.PP_DEFAULT: nicparams},
288 candidate_pool_size=candidate_pool_size,
289 modify_etc_hosts=modify_etc_hosts,
290 modify_ssh_setup=modify_ssh_setup,
293 uuid=utils.NewUUID(),
295 master_node_config = objects.Node(name=hostname.name,
296 primary_ip=hostname.ip,
297 secondary_ip=secondary_ip,
299 master_candidate=True,
300 offline=False, drained=False,
302 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
303 cfg = config.ConfigWriter()
304 ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
305 cfg.Update(cfg.GetClusterInfo(), logging.error)
307 # start the master ip
308 # TODO: Review rpc call from bootstrap
309 # TODO: Warn on failed start master
310 rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
313 def InitConfig(version, cluster_config, master_node_config,
314 cfg_file=constants.CLUSTER_CONF_FILE):
315 """Create the initial cluster configuration.
317 It will contain the current node, which will also be the master
318 node, and no instances.
321 @param version: configuration version
322 @type cluster_config: L{objects.Cluster}
323 @param cluster_config: cluster configuration
324 @type master_node_config: L{objects.Node}
325 @param master_node_config: master node configuration
326 @type cfg_file: string
327 @param cfg_file: configuration file path
331 master_node_config.name: master_node_config,
335 config_data = objects.ConfigData(version=version,
336 cluster=cluster_config,
340 ctime=now, mtime=now)
341 utils.WriteFile(cfg_file,
342 data=serializer.Dump(config_data.ToDict()),
346 def FinalizeClusterDestroy(master):
347 """Execute the last steps of cluster destroy
349 This function shuts down all the daemons, completing the destroy
350 begun in cmdlib.LUDestroyOpcode.
353 cfg = config.ConfigWriter()
354 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
355 result = rpc.RpcRunner.call_node_stop_master(master, True)
356 msg = result.fail_msg
358 logging.warning("Could not disable the master role: %s", msg)
359 result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup)
360 msg = result.fail_msg
362 logging.warning("Could not shutdown the node daemon and cleanup"
363 " the node: %s", msg)
366 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
367 """Add a node to the cluster.
369 This function must be called before the actual opcode, and will ssh
370 to the remote node, copy the needed files, and start ganeti-noded,
371 allowing the master to do the rest via normal rpc calls.
373 @param cluster_name: the cluster name
374 @param node: the name of the new node
375 @param ssh_key_check: whether to do a strict key check
378 sshrunner = ssh.SshRunner(cluster_name)
380 noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
381 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
382 hmac_key = utils.ReadFile(constants.HMAC_CLUSTER_KEY)
384 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
385 # so we use this to detect an invalid certificate; as long as the
386 # cert doesn't contain this, the here-document will be correctly
387 # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
388 # so the same restrictions apply.
389 for content in (noded_cert, rapi_cert, hmac_key):
390 if re.search('^!EOF\.', content, re.MULTILINE):
391 raise errors.OpExecError("invalid SSL certificate or HMAC key")
393 if not noded_cert.endswith("\n"):
395 if not rapi_cert.endswith("\n"):
397 if not hmac_key.endswith("\n"):
400 # set up inter-node password and certificate and restarts the node daemon
401 # and then connect with ssh to set password and start ganeti-noded
402 # note that all the below variables are sanitized at this point,
403 # either by being constants or by the checks above
404 mycommand = ("umask 077 && "
405 "cat > '%s' << '!EOF.' && \n"
407 "cat > '%s' << '!EOF.' && \n"
409 "cat > '%s' << '!EOF.' && \n"
411 "chmod 0400 %s %s %s && "
413 (constants.SSL_CERT_FILE, noded_cert,
414 constants.RAPI_CERT_FILE, rapi_cert,
415 constants.HMAC_CLUSTER_KEY, hmac_key,
416 constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
417 constants.HMAC_CLUSTER_KEY,
418 constants.DAEMON_UTIL, constants.NODED))
420 result = sshrunner.Run(node, 'root', mycommand, batch=False,
421 ask_key=ssh_key_check,
422 use_cluster_key=False,
423 strict_host_check=ssh_key_check)
425 raise errors.OpExecError("Remote command on node %s, error: %s,"
427 (node, result.fail_reason, result.output))
430 def MasterFailover(no_voting=False):
431 """Failover the master node.
433 This checks that we are not already the master, and will cause the
434 current master to cease being master, and the non-master to become
437 @type no_voting: boolean
438 @param no_voting: force the operation without remote nodes agreement
442 sstore = ssconf.SimpleStore()
444 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
445 node_list = sstore.GetNodeList()
446 mc_list = sstore.GetMasterCandidates()
448 if old_master == new_master:
449 raise errors.OpPrereqError("This commands must be run on the node"
450 " where you want the new master to be."
451 " %s is already the master" %
452 old_master, errors.ECODE_INVAL)
454 if new_master not in mc_list:
455 mc_no_master = [name for name in mc_list if name != old_master]
456 raise errors.OpPrereqError("This node is not among the nodes marked"
457 " as master candidates. Only these nodes"
458 " can become masters. Current list of"
459 " master candidates is:\n"
460 "%s" % ('\n'.join(mc_no_master)),
464 vote_list = GatherMasterVotes(node_list)
467 voted_master = vote_list[0][0]
468 if voted_master is None:
469 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
470 " not respond.", errors.ECODE_ENVIRON)
471 elif voted_master != old_master:
472 raise errors.OpPrereqError("I have a wrong configuration, I believe"
473 " the master is %s but the other nodes"
474 " voted %s. Please resync the configuration"
476 (old_master, voted_master),
482 logging.info("Setting master to %s, old master: %s", new_master, old_master)
484 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
485 msg = result.fail_msg
487 logging.error("Could not disable the master role on the old master"
488 " %s, please disable manually: %s", old_master, msg)
490 # Here we have a phase where no master should be running
492 # instantiate a real config writer, as we now know we have the
494 cfg = config.ConfigWriter()
496 cluster_info = cfg.GetClusterInfo()
497 cluster_info.master_node = new_master
498 # this will also regenerate the ssconf files, since we updated the
500 cfg.Update(cluster_info, logging.error)
502 result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
503 msg = result.fail_msg
505 logging.error("Could not start the master role on the new master"
506 " %s, please check: %s", new_master, msg)
513 """Returns the current master node.
515 This is a separate function in bootstrap since it's needed by
516 gnt-cluster, and instead of importing directly ssconf, it's better
517 to abstract it in bootstrap, where we do use ssconf in other
521 sstore = ssconf.SimpleStore()
523 old_master, _ = ssconf.GetMasterAndMyself(sstore)
528 def GatherMasterVotes(node_list):
529 """Check the agreement on who is the master.
531 This function will return a list of (node, number of votes), ordered
532 by the number of votes. Errors will be denoted by the key 'None'.
534 Note that the sum of votes is the number of nodes this machine
535 knows, whereas the number of entries in the list could be different
536 (if some nodes vote for another master).
538 We remove ourselves from the list since we know that (bugs aside)
539 since we use the same source for configuration information for both
540 backend and boostrap, we'll always vote for ourselves.
542 @type node_list: list
543 @param node_list: the list of nodes to query for master info; the current
544 node will be removed if it is in the list
546 @return: list of (node, votes)
549 myself = utils.HostInfo().name
551 node_list.remove(myself)
555 # no nodes left (eventually after removing myself)
557 results = rpc.RpcRunner.call_master_info(node_list)
558 if not isinstance(results, dict):
559 # this should not happen (unless internal error in rpc)
560 logging.critical("Can't complete rpc call, aborting master startup")
561 return [(None, len(node_list))]
569 logging.warning("Error contacting node %s: %s", node, msg)
571 elif not isinstance(data, (tuple, list)) or len(data) < 3:
572 logging.warning("Invalid data received from node %s: %s", node, data)
575 if None not in votes:
579 master_node = data[2]
580 if master_node not in votes:
581 votes[master_node] = 0
582 votes[master_node] += 1
584 vote_list = [v for v in votes.items()]
585 # sort first on number of votes then on name, since we want None
586 # sorted later if we have the half of the nodes not responding, and
587 # half voting all for the same master
588 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)