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 hypervisor
44 """Setup the SSH configuration for the cluster.
46 This generates a dsa keypair for root, adds the pub key to the
47 permitted hosts and adds the hostkey to its own known hosts.
50 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
52 for name in priv_key, pub_key:
53 if os.path.exists(name):
54 utils.CreateBackup(name)
55 utils.RemoveFile(name)
57 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
61 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
64 f = open(pub_key, 'r')
66 utils.AddAuthorizedKey(auth_keys, f.read(8192))
71 def _GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
72 """Generates a self-signed SSL certificate.
75 @param file_name: Path to output file
77 @param validity: Validity for certificate in days
80 (fd, tmp_file_name) = tempfile.mkstemp(dir=os.path.dirname(file_name))
83 # Set permissions before writing key
84 os.chmod(tmp_file_name, 0600)
86 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
87 "-days", str(validity), "-nodes", "-x509",
88 "-keyout", tmp_file_name, "-out", tmp_file_name,
91 raise errors.OpExecError("Could not generate SSL certificate, command"
92 " %s had exitcode %s and error message %s" %
93 (result.cmd, result.exit_code, result.output))
96 os.chmod(tmp_file_name, 0400)
98 os.rename(tmp_file_name, file_name)
100 utils.RemoveFile(tmp_file_name)
105 def _InitGanetiServerSetup():
106 """Setup the necessary configuration for the initial node daemon.
108 This creates the nodepass file containing the shared password for
109 the cluster and also generates the SSL certificate.
112 _GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
114 # Don't overwrite existing file
115 if not os.path.exists(constants.RAPI_CERT_FILE):
116 _GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
118 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
121 raise errors.OpExecError("Could not start the node daemon, command %s"
122 " had exitcode %s and error %s" %
123 (result.cmd, result.exit_code, result.output))
126 def InitCluster(cluster_name, mac_prefix, def_bridge,
127 master_netdev, file_storage_dir, candidate_pool_size,
128 secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
129 enabled_hypervisors=None, default_hypervisor=None,
130 modify_etc_hosts=True):
131 """Initialise the cluster.
133 @type candidate_pool_size: int
134 @param candidate_pool_size: master candidate pool size
137 # TODO: complete the docstring
138 if config.ConfigWriter.IsCluster():
139 raise errors.OpPrereqError("Cluster is already initialised")
141 if not enabled_hypervisors:
142 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
144 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
146 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
147 " entries: %s" % invalid_hvs)
149 hostname = utils.HostInfo()
151 if hostname.ip.startswith("127."):
152 raise errors.OpPrereqError("This host's IP resolves to the private"
153 " range (%s). Please fix DNS or %s." %
154 (hostname.ip, constants.ETC_HOSTS))
156 if not utils.OwnIpAddress(hostname.ip):
157 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
158 " to %s,\nbut this ip address does not"
159 " belong to this host."
160 " Aborting." % hostname.ip)
162 clustername = utils.HostInfo(cluster_name)
164 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
166 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
169 if not utils.IsValidIP(secondary_ip):
170 raise errors.OpPrereqError("Invalid secondary ip given")
171 if (secondary_ip != hostname.ip and
172 not utils.OwnIpAddress(secondary_ip)):
173 raise errors.OpPrereqError("You gave %s as secondary IP,"
174 " but it does not belong to this host." %
177 secondary_ip = hostname.ip
179 if vg_name is not None:
180 # Check if volume group is valid
181 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
182 constants.MIN_VG_SIZE)
184 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
185 " you are not using lvm" % vgstatus)
187 file_storage_dir = os.path.normpath(file_storage_dir)
189 if not os.path.isabs(file_storage_dir):
190 raise errors.OpPrereqError("The file storage directory you passed is"
191 " not an absolute path.")
193 if not os.path.exists(file_storage_dir):
195 os.makedirs(file_storage_dir, 0750)
197 raise errors.OpPrereqError("Cannot create file storage directory"
199 (file_storage_dir, err))
201 if not os.path.isdir(file_storage_dir):
202 raise errors.OpPrereqError("The file storage directory '%s' is not"
203 " a directory." % file_storage_dir)
205 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
206 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
208 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
210 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
212 result.output.strip()))
214 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
215 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
216 raise errors.OpPrereqError("Init.d script '%s' missing or not"
217 " executable." % constants.NODE_INITD_SCRIPT)
219 dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
220 utils.EnsureDirs(dirs)
222 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
223 # hvparams is a mapping of hypervisor->hvparams dict
224 for hv_name, hv_params in hvparams.iteritems():
225 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
226 hv_class = hypervisor.GetHypervisor(hv_name)
227 hv_class.CheckParameterSyntax(hv_params)
229 # set up the inter-node password and certificate
230 _InitGanetiServerSetup()
232 # set up ssh config and /etc/hosts
233 f = open(constants.SSH_HOST_RSA_PUB, 'r')
238 sshkey = sshline.split(" ")[1]
241 utils.AddHostToEtcHosts(hostname.name)
245 # init of cluster config file
246 cluster_config = objects.Cluster(
248 rsahostkeypub=sshkey,
249 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
250 mac_prefix=mac_prefix,
251 volume_group_name=vg_name,
252 default_bridge=def_bridge,
253 tcpudp_port_pool=set(),
254 master_node=hostname.name,
255 master_ip=clustername.ip,
256 master_netdev=master_netdev,
257 cluster_name=clustername.name,
258 file_storage_dir=file_storage_dir,
259 enabled_hypervisors=enabled_hypervisors,
260 default_hypervisor=default_hypervisor,
261 beparams={constants.BEGR_DEFAULT: beparams},
263 candidate_pool_size=candidate_pool_size,
264 modify_etc_hosts=opts.modify_etc_hosts,
266 master_node_config = objects.Node(name=hostname.name,
267 primary_ip=hostname.ip,
268 secondary_ip=secondary_ip,
270 master_candidate=True,
271 offline=False, drained=False,
274 sscfg = InitConfig(constants.CONFIG_VERSION,
275 cluster_config, master_node_config)
276 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
277 cfg = config.ConfigWriter()
278 cfg.Update(cfg.GetClusterInfo())
280 # start the master ip
281 # TODO: Review rpc call from bootstrap
282 rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
285 def InitConfig(version, cluster_config, master_node_config,
286 cfg_file=constants.CLUSTER_CONF_FILE):
287 """Create the initial cluster configuration.
289 It will contain the current node, which will also be the master
290 node, and no instances.
293 @param version: configuration version
294 @type cluster_config: L{objects.Cluster}
295 @param cluster_config: cluster configuration
296 @type master_node_config: L{objects.Node}
297 @param master_node_config: master node configuration
298 @type cfg_file: string
299 @param cfg_file: configuration file path
301 @rtype: L{ssconf.SimpleConfigWriter}
302 @return: initialized config instance
306 master_node_config.name: master_node_config,
309 config_data = objects.ConfigData(version=version,
310 cluster=cluster_config,
314 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
320 def FinalizeClusterDestroy(master):
321 """Execute the last steps of cluster destroy
323 This function shuts down all the daemons, completing the destroy
324 begun in cmdlib.LUDestroyOpcode.
327 result = rpc.RpcRunner.call_node_stop_master(master, True)
328 if result.failed or not result.data:
329 logging.warning("Could not disable the master role")
330 result = rpc.RpcRunner.call_node_leave_cluster(master)
331 if result.failed or not result.data:
332 logging.warning("Could not shutdown the node daemon and cleanup the node")
335 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
336 """Add a node to the cluster.
338 This function must be called before the actual opcode, and will ssh
339 to the remote node, copy the needed files, and start ganeti-noded,
340 allowing the master to do the rest via normal rpc calls.
342 @param cluster_name: the cluster name
343 @param node: the name of the new node
344 @param ssh_key_check: whether to do a strict key check
347 sshrunner = ssh.SshRunner(cluster_name)
349 noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
350 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
352 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
353 # so we use this to detect an invalid certificate; as long as the
354 # cert doesn't contain this, the here-document will be correctly
355 # parsed by the shell sequence below
356 if (re.search('^!EOF\.', noded_cert, re.MULTILINE) or
357 re.search('^!EOF\.', rapi_cert, re.MULTILINE)):
358 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
360 if not noded_cert.endswith("\n"):
362 if not rapi_cert.endswith("\n"):
365 # set up inter-node password and certificate and restarts the node daemon
366 # and then connect with ssh to set password and start ganeti-noded
367 # note that all the below variables are sanitized at this point,
368 # either by being constants or by the checks above
369 mycommand = ("umask 077 && "
370 "cat > '%s' << '!EOF.' && \n"
372 "cat > '%s' << '!EOF.' && \n"
374 "chmod 0400 %s %s && "
376 (constants.SSL_CERT_FILE, noded_cert,
377 constants.RAPI_CERT_FILE, rapi_cert,
378 constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
379 constants.NODE_INITD_SCRIPT))
381 result = sshrunner.Run(node, 'root', mycommand, batch=False,
382 ask_key=ssh_key_check,
383 use_cluster_key=False,
384 strict_host_check=ssh_key_check)
386 raise errors.OpExecError("Remote command on node %s, error: %s,"
388 (node, result.fail_reason, result.output))
391 def MasterFailover(no_voting=False):
392 """Failover the master node.
394 This checks that we are not already the master, and will cause the
395 current master to cease being master, and the non-master to become
398 @type no_voting: boolean
399 @param no_voting: force the operation without remote nodes agreement
403 sstore = ssconf.SimpleStore()
405 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
406 node_list = sstore.GetNodeList()
407 mc_list = sstore.GetMasterCandidates()
409 if old_master == new_master:
410 raise errors.OpPrereqError("This commands must be run on the node"
411 " where you want the new master to be."
412 " %s is already the master" %
415 if new_master not in mc_list:
416 mc_no_master = [name for name in mc_list if name != old_master]
417 raise errors.OpPrereqError("This node is not among the nodes marked"
418 " as master candidates. Only these nodes"
419 " can become masters. Current list of"
420 " master candidates is:\n"
421 "%s" % ('\n'.join(mc_no_master)))
424 vote_list = GatherMasterVotes(node_list)
427 voted_master = vote_list[0][0]
428 if voted_master is None:
429 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
431 elif voted_master != old_master:
432 raise errors.OpPrereqError("I have a wrong configuration, I believe"
433 " the master is %s but the other nodes"
434 " voted %s. Please resync the configuration"
436 (old_master, voted_master))
441 logging.info("Setting master to %s, old master: %s", new_master, old_master)
443 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
444 if result.failed or not result.data:
445 logging.error("Could not disable the master role on the old master"
446 " %s, please disable manually", old_master)
448 # Here we have a phase where no master should be running
450 # instantiate a real config writer, as we now know we have the
452 cfg = config.ConfigWriter()
454 cluster_info = cfg.GetClusterInfo()
455 cluster_info.master_node = new_master
456 # this will also regenerate the ssconf files, since we updated the
458 cfg.Update(cluster_info)
460 result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
461 if result.failed or not result.data:
462 logging.error("Could not start the master role on the new master"
463 " %s, please check", new_master)
470 """Returns the current master node.
472 This is a separate function in bootstrap since it's needed by
473 gnt-cluster, and instead of importing directly ssconf, it's better
474 to abstract it in bootstrap, where we do use ssconf in other
478 sstore = ssconf.SimpleStore()
480 old_master, _ = ssconf.GetMasterAndMyself(sstore)
485 def GatherMasterVotes(node_list):
486 """Check the agreement on who is the master.
488 This function will return a list of (node, number of votes), ordered
489 by the number of votes. Errors will be denoted by the key 'None'.
491 Note that the sum of votes is the number of nodes this machine
492 knows, whereas the number of entries in the list could be different
493 (if some nodes vote for another master).
495 We remove ourselves from the list since we know that (bugs aside)
496 since we use the same source for configuration information for both
497 backend and boostrap, we'll always vote for ourselves.
499 @type node_list: list
500 @param node_list: the list of nodes to query for master info; the current
501 node will be removed if it is in the list
503 @return: list of (node, votes)
506 myself = utils.HostInfo().name
508 node_list.remove(myself)
512 # no nodes left (eventually after removing myself)
514 results = rpc.RpcRunner.call_master_info(node_list)
515 if not isinstance(results, dict):
516 # this should not happen (unless internal error in rpc)
517 logging.critical("Can't complete rpc call, aborting master startup")
518 return [(None, len(node_list))]
523 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
524 # here the rpc layer should have already logged errors
525 if None not in votes:
529 master_node = data[2]
530 if master_node not in votes:
531 votes[master_node] = 0
532 votes[master_node] += 1
534 vote_list = [v for v in votes.items()]
535 # sort first on number of votes then on name, since we want None
536 # sorted later if we have the half of the nodes not responding, and
537 # half voting all for the same master
538 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)