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
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))
82 # Set permissions before writing key
83 os.chmod(tmp_file_name, 0600)
85 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
86 "-days", str(validity), "-nodes", "-x509",
87 "-keyout", tmp_file_name, "-out", tmp_file_name,
90 raise errors.OpExecError("Could not generate SSL certificate, command"
91 " %s had exitcode %s and error message %s" %
92 (result.cmd, result.exit_code, result.output))
95 os.chmod(tmp_file_name, 0400)
97 os.rename(tmp_file_name, file_name)
99 utils.RemoveFile(tmp_file_name)
102 def _InitGanetiServerSetup():
103 """Setup the necessary configuration for the initial node daemon.
105 This creates the nodepass file containing the shared password for
106 the cluster and also generates the SSL certificate.
109 _GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
111 # Don't overwrite existing file
112 if not os.path.exists(constants.RAPI_CERT_FILE):
113 _GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
115 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
118 raise errors.OpExecError("Could not start the node daemon, command %s"
119 " had exitcode %s and error %s" %
120 (result.cmd, result.exit_code, result.output))
123 def InitCluster(cluster_name, mac_prefix, def_bridge,
124 master_netdev, file_storage_dir, candidate_pool_size,
125 secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
126 enabled_hypervisors=None, default_hypervisor=None):
127 """Initialise the cluster.
129 @type candidate_pool_size: int
130 @param candidate_pool_size: master candidate pool size
133 # TODO: complete the docstring
134 if config.ConfigWriter.IsCluster():
135 raise errors.OpPrereqError("Cluster is already initialised")
137 hostname = utils.HostInfo()
139 if hostname.ip.startswith("127."):
140 raise errors.OpPrereqError("This host's IP resolves to the private"
141 " range (%s). Please fix DNS or %s." %
142 (hostname.ip, constants.ETC_HOSTS))
144 if not utils.OwnIpAddress(hostname.ip):
145 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
146 " to %s,\nbut this ip address does not"
147 " belong to this host."
148 " Aborting." % hostname.ip)
150 clustername = utils.HostInfo(cluster_name)
152 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
154 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
157 if not utils.IsValidIP(secondary_ip):
158 raise errors.OpPrereqError("Invalid secondary ip given")
159 if (secondary_ip != hostname.ip and
160 not utils.OwnIpAddress(secondary_ip)):
161 raise errors.OpPrereqError("You gave %s as secondary IP,"
162 " but it does not belong to this host." %
165 secondary_ip = hostname.ip
167 if vg_name is not None:
168 # Check if volume group is valid
169 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
170 constants.MIN_VG_SIZE)
172 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
173 " you are not using lvm" % vgstatus)
175 file_storage_dir = os.path.normpath(file_storage_dir)
177 if not os.path.isabs(file_storage_dir):
178 raise errors.OpPrereqError("The file storage directory you passed is"
179 " not an absolute path.")
181 if not os.path.exists(file_storage_dir):
183 os.makedirs(file_storage_dir, 0750)
185 raise errors.OpPrereqError("Cannot create file storage directory"
187 (file_storage_dir, err))
189 if not os.path.isdir(file_storage_dir):
190 raise errors.OpPrereqError("The file storage directory '%s' is not"
191 " a directory." % file_storage_dir)
193 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
194 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
196 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
198 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
200 result.output.strip()))
202 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
203 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
204 raise errors.OpPrereqError("Init.d script '%s' missing or not"
205 " executable." % constants.NODE_INITD_SCRIPT)
207 utils.CheckBEParams(beparams)
209 # set up the inter-node password and certificate
210 _InitGanetiServerSetup()
212 # set up ssh config and /etc/hosts
213 f = open(constants.SSH_HOST_RSA_PUB, 'r')
218 sshkey = sshline.split(" ")[1]
220 utils.AddHostToEtcHosts(hostname.name)
223 # init of cluster config file
224 cluster_config = objects.Cluster(
226 rsahostkeypub=sshkey,
227 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
228 mac_prefix=mac_prefix,
229 volume_group_name=vg_name,
230 default_bridge=def_bridge,
231 tcpudp_port_pool=set(),
232 master_node=hostname.name,
233 master_ip=clustername.ip,
234 master_netdev=master_netdev,
235 cluster_name=clustername.name,
236 file_storage_dir=file_storage_dir,
237 enabled_hypervisors=enabled_hypervisors,
238 default_hypervisor=default_hypervisor,
239 beparams={constants.BEGR_DEFAULT: beparams},
241 candidate_pool_size=candidate_pool_size,
243 master_node_config = objects.Node(name=hostname.name,
244 primary_ip=hostname.ip,
245 secondary_ip=secondary_ip,
247 master_candidate=True,
251 sscfg = InitConfig(constants.CONFIG_VERSION,
252 cluster_config, master_node_config)
253 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
254 cfg = config.ConfigWriter()
255 cfg.Update(cfg.GetClusterInfo())
257 # start the master ip
258 # TODO: Review rpc call from bootstrap
259 rpc.RpcRunner.call_node_start_master(hostname.name, True)
262 def InitConfig(version, cluster_config, master_node_config,
263 cfg_file=constants.CLUSTER_CONF_FILE):
264 """Create the initial cluster configuration.
266 It will contain the current node, which will also be the master
267 node, and no instances.
270 @param version: configuration version
271 @type cluster_config: L{objects.Cluster}
272 @param cluster_config: cluster configuration
273 @type master_node_config: L{objects.Node}
274 @param master_node_config: master node configuration
275 @type cfg_file: string
276 @param cfg_file: configuration file path
278 @rtype: L{ssconf.SimpleConfigWriter}
279 @returns: initialized config instance
283 master_node_config.name: master_node_config,
286 config_data = objects.ConfigData(version=version,
287 cluster=cluster_config,
291 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
297 def FinalizeClusterDestroy(master):
298 """Execute the last steps of cluster destroy
300 This function shuts down all the daemons, completing the destroy
301 begun in cmdlib.LUDestroyOpcode.
304 result = rpc.RpcRunner.call_node_stop_master(master, True)
305 if result.failed or not result.data:
306 logging.warning("Could not disable the master role")
307 result = rpc.RpcRunner.call_node_leave_cluster(master)
308 if result.failed or not result.data:
309 logging.warning("Could not shutdown the node daemon and cleanup the node")
312 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
313 """Add a node to the cluster.
315 This function must be called before the actual opcode, and will ssh
316 to the remote node, copy the needed files, and start ganeti-noded,
317 allowing the master to do the rest via normal rpc calls.
319 @param cluster_name: the cluster name
320 @param node: the name of the new node
321 @param ssh_key_check: whether to do a strict key check
324 sshrunner = ssh.SshRunner(cluster_name)
326 noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
327 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
329 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
330 # so we use this to detect an invalid certificate; as long as the
331 # cert doesn't contain this, the here-document will be correctly
332 # parsed by the shell sequence below
333 if (re.search('^!EOF\.', noded_cert, re.MULTILINE) or
334 re.search('^!EOF\.', rapi_cert, re.MULTILINE)):
335 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
337 if not noded_cert.endswith("\n"):
339 if not rapi_cert.endswith("\n"):
342 # set up inter-node password and certificate and restarts the node daemon
343 # and then connect with ssh to set password and start ganeti-noded
344 # note that all the below variables are sanitized at this point,
345 # either by being constants or by the checks above
346 mycommand = ("umask 077 && "
347 "cat > '%s' << '!EOF.' && \n"
349 "cat > '%s' << '!EOF.' && \n"
351 "chmod 0400 %s %s && "
353 (constants.SSL_CERT_FILE, noded_cert,
354 constants.RAPI_CERT_FILE, rapi_cert,
355 constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
356 constants.NODE_INITD_SCRIPT))
358 result = sshrunner.Run(node, 'root', mycommand, batch=False,
359 ask_key=ssh_key_check,
360 use_cluster_key=False,
361 strict_host_check=ssh_key_check)
363 raise errors.OpExecError("Remote command on node %s, error: %s,"
365 (node, result.fail_reason, result.output))
368 def MasterFailover():
369 """Failover the master node.
371 This checks that we are not already the master, and will cause the
372 current master to cease being master, and the non-master to become
376 sstore = ssconf.SimpleStore()
378 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
379 node_list = sstore.GetNodeList()
380 mc_list = sstore.GetMasterCandidates()
382 if old_master == new_master:
383 raise errors.OpPrereqError("This commands must be run on the node"
384 " where you want the new master to be."
385 " %s is already the master" %
388 if new_master not in mc_list:
389 mc_no_master = [name for name in mc_list if name != old_master]
390 raise errors.OpPrereqError("This node is not among the nodes marked"
391 " as master candidates. Only these nodes"
392 " can become masters. Current list of"
393 " master candidates is:\n"
394 "%s" % ('\n'.join(mc_no_master)))
396 vote_list = GatherMasterVotes(node_list)
399 voted_master = vote_list[0][0]
400 if voted_master is None:
401 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
403 elif voted_master != old_master:
404 raise errors.OpPrereqError("I have wrong configuration, I believe the"
405 " master is %s but the other nodes voted for"
406 " %s. Please resync the configuration of"
407 " this node." % (old_master, voted_master))
412 logging.info("Setting master to %s, old master: %s", new_master, old_master)
414 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
415 if result.failed or not result.data:
416 logging.error("Could not disable the master role on the old master"
417 " %s, please disable manually", old_master)
419 # Here we have a phase where no master should be running
421 # instantiate a real config writer, as we now know we have the
423 cfg = config.ConfigWriter()
425 cluster_info = cfg.GetClusterInfo()
426 cluster_info.master_node = new_master
427 # this will also regenerate the ssconf files, since we updated the
429 cfg.Update(cluster_info)
431 result = rpc.RpcRunner.call_node_start_master(new_master, True)
432 if result.failed or not result.data:
433 logging.error("Could not start the master role on the new master"
434 " %s, please check", new_master)
440 def GatherMasterVotes(node_list):
441 """Check the agreement on who is the master.
443 This function will return a list of (node, number of votes), ordered
444 by the number of votes. Errors will be denoted by the key 'None'.
446 Note that the sum of votes is the number of nodes this machine
447 knows, whereas the number of entries in the list could be different
448 (if some nodes vote for another master).
450 We remove ourselves from the list since we know that (bugs aside)
451 since we use the same source for configuration information for both
452 backend and boostrap, we'll always vote for ourselves.
454 @type node_list: list
455 @param node_list: the list of nodes to query for master info; the current
456 node wil be removed if it is in the list
458 @return: list of (node, votes)
461 myself = utils.HostInfo().name
463 node_list.remove(myself)
467 # no nodes left (eventually after removing myself)
469 results = rpc.RpcRunner.call_master_info(node_list)
470 if not isinstance(results, dict):
471 # this should not happen (unless internal error in rpc)
472 logging.critical("Can't complete rpc call, aborting master startup")
473 return [(None, len(node_list))]
478 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
479 # here the rpc layer should have already logged errors
480 if None not in votes:
484 master_node = data[2]
485 if master_node not in votes:
486 votes[master_node] = 0
487 votes[master_node] += 1
489 vote_list = [v for v in votes.items()]
490 # sort first on number of votes then on name, since we want None
491 # sorted later if we have the half of the nodes not responding, and
492 # half voting all for the same master
493 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)