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))
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 modify_etc_hosts=True):
128 """Initialise the cluster.
130 @type candidate_pool_size: int
131 @param candidate_pool_size: master candidate pool size
134 # TODO: complete the docstring
135 if config.ConfigWriter.IsCluster():
136 raise errors.OpPrereqError("Cluster is already initialised")
138 hostname = utils.HostInfo()
140 if hostname.ip.startswith("127."):
141 raise errors.OpPrereqError("This host's IP resolves to the private"
142 " range (%s). Please fix DNS or %s." %
143 (hostname.ip, constants.ETC_HOSTS))
145 if not utils.OwnIpAddress(hostname.ip):
146 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
147 " to %s,\nbut this ip address does not"
148 " belong to this host."
149 " Aborting." % hostname.ip)
151 clustername = utils.HostInfo(cluster_name)
153 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
155 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
158 if not utils.IsValidIP(secondary_ip):
159 raise errors.OpPrereqError("Invalid secondary ip given")
160 if (secondary_ip != hostname.ip and
161 not utils.OwnIpAddress(secondary_ip)):
162 raise errors.OpPrereqError("You gave %s as secondary IP,"
163 " but it does not belong to this host." %
166 secondary_ip = hostname.ip
168 if vg_name is not None:
169 # Check if volume group is valid
170 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
171 constants.MIN_VG_SIZE)
173 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
174 " you are not using lvm" % vgstatus)
176 file_storage_dir = os.path.normpath(file_storage_dir)
178 if not os.path.isabs(file_storage_dir):
179 raise errors.OpPrereqError("The file storage directory you passed is"
180 " not an absolute path.")
182 if not os.path.exists(file_storage_dir):
184 os.makedirs(file_storage_dir, 0750)
186 raise errors.OpPrereqError("Cannot create file storage directory"
188 (file_storage_dir, err))
190 if not os.path.isdir(file_storage_dir):
191 raise errors.OpPrereqError("The file storage directory '%s' is not"
192 " a directory." % file_storage_dir)
194 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
195 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
197 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
199 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
201 result.output.strip()))
203 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
204 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
205 raise errors.OpPrereqError("Init.d script '%s' missing or not"
206 " executable." % constants.NODE_INITD_SCRIPT)
208 dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
209 utils.EnsureDirs(dirs)
211 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
212 # hvparams is a mapping of hypervisor->hvparams dict
213 for hv_name, hv_params in hvparams.iteritems():
214 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
215 hv_class = hypervisor.GetHypervisor(hv_name)
216 hv_class.CheckParameterSyntax(hv_params)
218 # set up the inter-node password and certificate
219 _InitGanetiServerSetup()
221 # set up ssh config and /etc/hosts
222 f = open(constants.SSH_HOST_RSA_PUB, 'r')
227 sshkey = sshline.split(" ")[1]
230 utils.AddHostToEtcHosts(hostname.name)
234 # init of cluster config file
235 cluster_config = objects.Cluster(
237 rsahostkeypub=sshkey,
238 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
239 mac_prefix=mac_prefix,
240 volume_group_name=vg_name,
241 default_bridge=def_bridge,
242 tcpudp_port_pool=set(),
243 master_node=hostname.name,
244 master_ip=clustername.ip,
245 master_netdev=master_netdev,
246 cluster_name=clustername.name,
247 file_storage_dir=file_storage_dir,
248 enabled_hypervisors=enabled_hypervisors,
249 default_hypervisor=default_hypervisor,
250 beparams={constants.BEGR_DEFAULT: beparams},
252 candidate_pool_size=candidate_pool_size,
253 modify_etc_hosts=opts.modify_etc_hosts,
255 master_node_config = objects.Node(name=hostname.name,
256 primary_ip=hostname.ip,
257 secondary_ip=secondary_ip,
259 master_candidate=True,
260 offline=False, drained=False,
263 sscfg = InitConfig(constants.CONFIG_VERSION,
264 cluster_config, master_node_config)
265 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
266 cfg = config.ConfigWriter()
267 cfg.Update(cfg.GetClusterInfo())
269 # start the master ip
270 # TODO: Review rpc call from bootstrap
271 rpc.RpcRunner.call_node_start_master(hostname.name, True)
274 def InitConfig(version, cluster_config, master_node_config,
275 cfg_file=constants.CLUSTER_CONF_FILE):
276 """Create the initial cluster configuration.
278 It will contain the current node, which will also be the master
279 node, and no instances.
282 @param version: configuration version
283 @type cluster_config: L{objects.Cluster}
284 @param cluster_config: cluster configuration
285 @type master_node_config: L{objects.Node}
286 @param master_node_config: master node configuration
287 @type cfg_file: string
288 @param cfg_file: configuration file path
290 @rtype: L{ssconf.SimpleConfigWriter}
291 @return: initialized config instance
295 master_node_config.name: master_node_config,
298 config_data = objects.ConfigData(version=version,
299 cluster=cluster_config,
303 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
309 def FinalizeClusterDestroy(master):
310 """Execute the last steps of cluster destroy
312 This function shuts down all the daemons, completing the destroy
313 begun in cmdlib.LUDestroyOpcode.
316 result = rpc.RpcRunner.call_node_stop_master(master, True)
317 if result.failed or not result.data:
318 logging.warning("Could not disable the master role")
319 result = rpc.RpcRunner.call_node_leave_cluster(master)
320 if result.failed or not result.data:
321 logging.warning("Could not shutdown the node daemon and cleanup the node")
324 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
325 """Add a node to the cluster.
327 This function must be called before the actual opcode, and will ssh
328 to the remote node, copy the needed files, and start ganeti-noded,
329 allowing the master to do the rest via normal rpc calls.
331 @param cluster_name: the cluster name
332 @param node: the name of the new node
333 @param ssh_key_check: whether to do a strict key check
336 sshrunner = ssh.SshRunner(cluster_name)
338 noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
339 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
341 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
342 # so we use this to detect an invalid certificate; as long as the
343 # cert doesn't contain this, the here-document will be correctly
344 # parsed by the shell sequence below
345 if (re.search('^!EOF\.', noded_cert, re.MULTILINE) or
346 re.search('^!EOF\.', rapi_cert, re.MULTILINE)):
347 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
349 if not noded_cert.endswith("\n"):
351 if not rapi_cert.endswith("\n"):
354 # set up inter-node password and certificate and restarts the node daemon
355 # and then connect with ssh to set password and start ganeti-noded
356 # note that all the below variables are sanitized at this point,
357 # either by being constants or by the checks above
358 mycommand = ("umask 077 && "
359 "cat > '%s' << '!EOF.' && \n"
361 "cat > '%s' << '!EOF.' && \n"
363 "chmod 0400 %s %s && "
365 (constants.SSL_CERT_FILE, noded_cert,
366 constants.RAPI_CERT_FILE, rapi_cert,
367 constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
368 constants.NODE_INITD_SCRIPT))
370 result = sshrunner.Run(node, 'root', mycommand, batch=False,
371 ask_key=ssh_key_check,
372 use_cluster_key=False,
373 strict_host_check=ssh_key_check)
375 raise errors.OpExecError("Remote command on node %s, error: %s,"
377 (node, result.fail_reason, result.output))
380 def MasterFailover():
381 """Failover the master node.
383 This checks that we are not already the master, and will cause the
384 current master to cease being master, and the non-master to become
388 sstore = ssconf.SimpleStore()
390 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
391 node_list = sstore.GetNodeList()
392 mc_list = sstore.GetMasterCandidates()
394 if old_master == new_master:
395 raise errors.OpPrereqError("This commands must be run on the node"
396 " where you want the new master to be."
397 " %s is already the master" %
400 if new_master not in mc_list:
401 mc_no_master = [name for name in mc_list if name != old_master]
402 raise errors.OpPrereqError("This node is not among the nodes marked"
403 " as master candidates. Only these nodes"
404 " can become masters. Current list of"
405 " master candidates is:\n"
406 "%s" % ('\n'.join(mc_no_master)))
408 vote_list = GatherMasterVotes(node_list)
411 voted_master = vote_list[0][0]
412 if voted_master is None:
413 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
415 elif voted_master != old_master:
416 raise errors.OpPrereqError("I have wrong configuration, I believe the"
417 " master is %s but the other nodes voted for"
418 " %s. Please resync the configuration of"
419 " this node." % (old_master, voted_master))
424 logging.info("Setting master to %s, old master: %s", new_master, old_master)
426 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
427 if result.failed or not result.data:
428 logging.error("Could not disable the master role on the old master"
429 " %s, please disable manually", old_master)
431 # Here we have a phase where no master should be running
433 # instantiate a real config writer, as we now know we have the
435 cfg = config.ConfigWriter()
437 cluster_info = cfg.GetClusterInfo()
438 cluster_info.master_node = new_master
439 # this will also regenerate the ssconf files, since we updated the
441 cfg.Update(cluster_info)
443 result = rpc.RpcRunner.call_node_start_master(new_master, True)
444 if result.failed or not result.data:
445 logging.error("Could not start the master role on the new master"
446 " %s, please check", new_master)
453 """Returns the current master node.
455 This is a separate function in bootstrap since it's needed by
456 gnt-cluster, and instead of importing directly ssconf, it's better
457 to abstract it in bootstrap, where we do use ssconf in other
461 sstore = ssconf.SimpleStore()
463 old_master, _ = ssconf.GetMasterAndMyself(sstore)
468 def GatherMasterVotes(node_list):
469 """Check the agreement on who is the master.
471 This function will return a list of (node, number of votes), ordered
472 by the number of votes. Errors will be denoted by the key 'None'.
474 Note that the sum of votes is the number of nodes this machine
475 knows, whereas the number of entries in the list could be different
476 (if some nodes vote for another master).
478 We remove ourselves from the list since we know that (bugs aside)
479 since we use the same source for configuration information for both
480 backend and boostrap, we'll always vote for ourselves.
482 @type node_list: list
483 @param node_list: the list of nodes to query for master info; the current
484 node wil be removed if it is in the list
486 @return: list of (node, votes)
489 myself = utils.HostInfo().name
491 node_list.remove(myself)
495 # no nodes left (eventually after removing myself)
497 results = rpc.RpcRunner.call_master_info(node_list)
498 if not isinstance(results, dict):
499 # this should not happen (unless internal error in rpc)
500 logging.critical("Can't complete rpc call, aborting master startup")
501 return [(None, len(node_list))]
506 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
507 # here the rpc layer should have already logged errors
508 if None not in votes:
512 master_node = data[2]
513 if master_node not in votes:
514 votes[master_node] = 0
515 votes[master_node] += 1
517 vote_list = [v for v in votes.items()]
518 # sort first on number of votes then on name, since we want None
519 # sorted later if we have the half of the nodes not responding, and
520 # half voting all for the same master
521 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)