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 f = open(pub_key, 'r')
68 utils.AddAuthorizedKey(auth_keys, f.read(8192))
73 def _GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
74 """Generates a self-signed SSL certificate.
77 @param file_name: Path to output file
79 @param validity: Validity for certificate in days
82 (fd, tmp_file_name) = tempfile.mkstemp(dir=os.path.dirname(file_name))
85 # Set permissions before writing key
86 os.chmod(tmp_file_name, 0600)
88 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
89 "-days", str(validity), "-nodes", "-x509",
90 "-keyout", tmp_file_name, "-out", tmp_file_name,
93 raise errors.OpExecError("Could not generate SSL certificate, command"
94 " %s had exitcode %s and error message %s" %
95 (result.cmd, result.exit_code, result.output))
98 os.chmod(tmp_file_name, 0400)
100 os.rename(tmp_file_name, file_name)
102 utils.RemoveFile(tmp_file_name)
107 def _InitGanetiServerSetup():
108 """Setup the necessary configuration for the initial node daemon.
110 This creates the nodepass file containing the shared password for
111 the cluster and also generates the SSL certificate.
114 _GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
116 # Don't overwrite existing file
117 if not os.path.exists(constants.RAPI_CERT_FILE):
118 _GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
120 if not os.path.exists(constants.HMAC_CLUSTER_KEY):
121 utils.WriteFile(constants.HMAC_CLUSTER_KEY,
122 data=utils.GenerateSecret(),
125 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
128 raise errors.OpExecError("Could not start the node daemon, command %s"
129 " had exitcode %s and error %s" %
130 (result.cmd, result.exit_code, result.output))
133 def InitCluster(cluster_name, mac_prefix,
134 master_netdev, file_storage_dir, candidate_pool_size,
135 secondary_ip=None, vg_name=None, beparams=None,
136 nicparams=None, hvparams=None, enabled_hypervisors=None,
137 modify_etc_hosts=True):
138 """Initialise the cluster.
140 @type candidate_pool_size: int
141 @param candidate_pool_size: master candidate pool size
144 # TODO: complete the docstring
145 if config.ConfigWriter.IsCluster():
146 raise errors.OpPrereqError("Cluster is already initialised")
148 if not enabled_hypervisors:
149 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
151 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
153 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
154 " entries: %s" % invalid_hvs)
156 hostname = utils.HostInfo()
158 if hostname.ip.startswith("127."):
159 raise errors.OpPrereqError("This host's IP resolves to the private"
160 " range (%s). Please fix DNS or %s." %
161 (hostname.ip, constants.ETC_HOSTS))
163 if not utils.OwnIpAddress(hostname.ip):
164 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
165 " to %s,\nbut this ip address does not"
166 " belong to this host."
167 " Aborting." % hostname.ip)
169 clustername = utils.HostInfo(cluster_name)
171 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
173 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
176 if not utils.IsValidIP(secondary_ip):
177 raise errors.OpPrereqError("Invalid secondary ip given")
178 if (secondary_ip != hostname.ip and
179 not utils.OwnIpAddress(secondary_ip)):
180 raise errors.OpPrereqError("You gave %s as secondary IP,"
181 " but it does not belong to this host." %
184 secondary_ip = hostname.ip
186 if vg_name is not None:
187 # Check if volume group is valid
188 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
189 constants.MIN_VG_SIZE)
191 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
192 " you are not using lvm" % vgstatus)
194 file_storage_dir = os.path.normpath(file_storage_dir)
196 if not os.path.isabs(file_storage_dir):
197 raise errors.OpPrereqError("The file storage directory you passed is"
198 " not an absolute path.")
200 if not os.path.exists(file_storage_dir):
202 os.makedirs(file_storage_dir, 0750)
204 raise errors.OpPrereqError("Cannot create file storage directory"
206 (file_storage_dir, err))
208 if not os.path.isdir(file_storage_dir):
209 raise errors.OpPrereqError("The file storage directory '%s' is not"
210 " a directory." % file_storage_dir)
212 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
213 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
215 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
217 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
219 result.output.strip()))
221 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
222 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
223 raise errors.OpPrereqError("Init.d script '%s' missing or not"
224 " executable." % constants.NODE_INITD_SCRIPT)
226 dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
227 utils.EnsureDirs(dirs)
229 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
230 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
231 objects.NIC.CheckParameterSyntax(nicparams)
233 # hvparams is a mapping of hypervisor->hvparams dict
234 for hv_name, hv_params in hvparams.iteritems():
235 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
236 hv_class = hypervisor.GetHypervisor(hv_name)
237 hv_class.CheckParameterSyntax(hv_params)
239 # set up the inter-node password and certificate
240 _InitGanetiServerSetup()
242 # set up ssh config and /etc/hosts
243 f = open(constants.SSH_HOST_RSA_PUB, 'r')
248 sshkey = sshline.split(" ")[1]
251 utils.AddHostToEtcHosts(hostname.name)
255 # init of cluster config file
256 cluster_config = objects.Cluster(
258 rsahostkeypub=sshkey,
259 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
260 mac_prefix=mac_prefix,
261 volume_group_name=vg_name,
262 tcpudp_port_pool=set(),
263 master_node=hostname.name,
264 master_ip=clustername.ip,
265 master_netdev=master_netdev,
266 cluster_name=clustername.name,
267 file_storage_dir=file_storage_dir,
268 enabled_hypervisors=enabled_hypervisors,
269 beparams={constants.PP_DEFAULT: beparams},
270 nicparams={constants.PP_DEFAULT: nicparams},
272 candidate_pool_size=candidate_pool_size,
273 modify_etc_hosts=modify_etc_hosts,
275 master_node_config = objects.Node(name=hostname.name,
276 primary_ip=hostname.ip,
277 secondary_ip=secondary_ip,
279 master_candidate=True,
280 offline=False, drained=False,
282 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
283 cfg = config.ConfigWriter()
284 ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
285 cfg.Update(cfg.GetClusterInfo())
287 # start the master ip
288 # TODO: Review rpc call from bootstrap
289 # TODO: Warn on failed start master
290 rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
293 def InitConfig(version, cluster_config, master_node_config,
294 cfg_file=constants.CLUSTER_CONF_FILE):
295 """Create the initial cluster configuration.
297 It will contain the current node, which will also be the master
298 node, and no instances.
301 @param version: configuration version
302 @type cluster_config: L{objects.Cluster}
303 @param cluster_config: cluster configuration
304 @type master_node_config: L{objects.Node}
305 @param master_node_config: master node configuration
306 @type cfg_file: string
307 @param cfg_file: configuration file path
311 master_node_config.name: master_node_config,
315 config_data = objects.ConfigData(version=version,
316 cluster=cluster_config,
320 ctime=now, mtime=now)
321 utils.WriteFile(cfg_file,
322 data=serializer.Dump(config_data.ToDict()),
326 def FinalizeClusterDestroy(master):
327 """Execute the last steps of cluster destroy
329 This function shuts down all the daemons, completing the destroy
330 begun in cmdlib.LUDestroyOpcode.
333 result = rpc.RpcRunner.call_node_stop_master(master, True)
334 msg = result.RemoteFailMsg()
336 logging.warning("Could not disable the master role: %s" % msg)
337 result = rpc.RpcRunner.call_node_leave_cluster(master)
338 msg = result.RemoteFailMsg()
340 logging.warning("Could not shutdown the node daemon and cleanup"
341 " the node: %s", msg)
344 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
345 """Add a node to the cluster.
347 This function must be called before the actual opcode, and will ssh
348 to the remote node, copy the needed files, and start ganeti-noded,
349 allowing the master to do the rest via normal rpc calls.
351 @param cluster_name: the cluster name
352 @param node: the name of the new node
353 @param ssh_key_check: whether to do a strict key check
356 sshrunner = ssh.SshRunner(cluster_name)
358 noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
359 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
361 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
362 # so we use this to detect an invalid certificate; as long as the
363 # cert doesn't contain this, the here-document will be correctly
364 # parsed by the shell sequence below
365 if (re.search('^!EOF\.', noded_cert, re.MULTILINE) or
366 re.search('^!EOF\.', rapi_cert, re.MULTILINE)):
367 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
369 if not noded_cert.endswith("\n"):
371 if not rapi_cert.endswith("\n"):
374 # set up inter-node password and certificate and restarts the node daemon
375 # and then connect with ssh to set password and start ganeti-noded
376 # note that all the below variables are sanitized at this point,
377 # either by being constants or by the checks above
378 mycommand = ("umask 077 && "
379 "cat > '%s' << '!EOF.' && \n"
381 "cat > '%s' << '!EOF.' && \n"
383 "chmod 0400 %s %s && "
385 (constants.SSL_CERT_FILE, noded_cert,
386 constants.RAPI_CERT_FILE, rapi_cert,
387 constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
388 constants.NODE_INITD_SCRIPT))
390 result = sshrunner.Run(node, 'root', mycommand, batch=False,
391 ask_key=ssh_key_check,
392 use_cluster_key=False,
393 strict_host_check=ssh_key_check)
395 raise errors.OpExecError("Remote command on node %s, error: %s,"
397 (node, result.fail_reason, result.output))
400 def MasterFailover(no_voting=False):
401 """Failover the master node.
403 This checks that we are not already the master, and will cause the
404 current master to cease being master, and the non-master to become
407 @type no_voting: boolean
408 @param no_voting: force the operation without remote nodes agreement
412 sstore = ssconf.SimpleStore()
414 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
415 node_list = sstore.GetNodeList()
416 mc_list = sstore.GetMasterCandidates()
418 if old_master == new_master:
419 raise errors.OpPrereqError("This commands must be run on the node"
420 " where you want the new master to be."
421 " %s is already the master" %
424 if new_master not in mc_list:
425 mc_no_master = [name for name in mc_list if name != old_master]
426 raise errors.OpPrereqError("This node is not among the nodes marked"
427 " as master candidates. Only these nodes"
428 " can become masters. Current list of"
429 " master candidates is:\n"
430 "%s" % ('\n'.join(mc_no_master)))
433 vote_list = GatherMasterVotes(node_list)
436 voted_master = vote_list[0][0]
437 if voted_master is None:
438 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
440 elif voted_master != old_master:
441 raise errors.OpPrereqError("I have a wrong configuration, I believe"
442 " the master is %s but the other nodes"
443 " voted %s. Please resync the configuration"
445 (old_master, voted_master))
450 logging.info("Setting master to %s, old master: %s", new_master, old_master)
452 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
453 msg = result.RemoteFailMsg()
455 logging.error("Could not disable the master role on the old master"
456 " %s, please disable manually: %s", old_master, msg)
458 # Here we have a phase where no master should be running
460 # instantiate a real config writer, as we now know we have the
462 cfg = config.ConfigWriter()
464 cluster_info = cfg.GetClusterInfo()
465 cluster_info.master_node = new_master
466 # this will also regenerate the ssconf files, since we updated the
468 cfg.Update(cluster_info)
470 result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
471 msg = result.RemoteFailMsg()
473 logging.error("Could not start the master role on the new master"
474 " %s, please check: %s", new_master, msg)
481 """Returns the current master node.
483 This is a separate function in bootstrap since it's needed by
484 gnt-cluster, and instead of importing directly ssconf, it's better
485 to abstract it in bootstrap, where we do use ssconf in other
489 sstore = ssconf.SimpleStore()
491 old_master, _ = ssconf.GetMasterAndMyself(sstore)
496 def GatherMasterVotes(node_list):
497 """Check the agreement on who is the master.
499 This function will return a list of (node, number of votes), ordered
500 by the number of votes. Errors will be denoted by the key 'None'.
502 Note that the sum of votes is the number of nodes this machine
503 knows, whereas the number of entries in the list could be different
504 (if some nodes vote for another master).
506 We remove ourselves from the list since we know that (bugs aside)
507 since we use the same source for configuration information for both
508 backend and boostrap, we'll always vote for ourselves.
510 @type node_list: list
511 @param node_list: the list of nodes to query for master info; the current
512 node will be removed if it is in the list
514 @return: list of (node, votes)
517 myself = utils.HostInfo().name
519 node_list.remove(myself)
523 # no nodes left (eventually after removing myself)
525 results = rpc.RpcRunner.call_master_info(node_list)
526 if not isinstance(results, dict):
527 # this should not happen (unless internal error in rpc)
528 logging.critical("Can't complete rpc call, aborting master startup")
529 return [(None, len(node_list))]
534 msg = nres.RemoteFailMsg()
537 logging.warning("Error contacting node %s: %s", node, msg)
539 elif not isinstance(data, (tuple, list)) or len(data) < 3:
540 logging.warning("Invalid data received from node %s: %s", node, data)
543 if None not in votes:
547 master_node = data[2]
548 if master_node not in votes:
549 votes[master_node] = 0
550 votes[master_node] += 1
552 vote_list = [v for v in votes.items()]
553 # sort first on number of votes then on name, since we want None
554 # sorted later if we have the half of the nodes not responding, and
555 # half voting all for the same master
556 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)