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 hypervisor
45 """Setup the SSH configuration for the cluster.
47 This generates a dsa keypair for root, adds the pub key to the
48 permitted hosts and adds the hostkey to its own known hosts.
51 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
53 for name in priv_key, pub_key:
54 if os.path.exists(name):
55 utils.CreateBackup(name)
56 utils.RemoveFile(name)
58 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
62 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
65 f = open(pub_key, 'r')
67 utils.AddAuthorizedKey(auth_keys, f.read(8192))
72 def _GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
73 """Generates a self-signed SSL certificate.
76 @param file_name: Path to output file
78 @param validity: Validity for certificate in days
81 (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)
103 def _InitGanetiServerSetup():
104 """Setup the necessary configuration for the initial node daemon.
106 This creates the nodepass file containing the shared password for
107 the cluster and also generates the SSL certificate.
110 _GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
112 # Don't overwrite existing file
113 if not os.path.exists(constants.RAPI_CERT_FILE):
114 _GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
116 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
119 raise errors.OpExecError("Could not start the node daemon, command %s"
120 " had exitcode %s and error %s" %
121 (result.cmd, result.exit_code, result.output))
124 def InitCluster(cluster_name, mac_prefix, def_bridge,
125 master_netdev, file_storage_dir, candidate_pool_size,
126 secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
127 enabled_hypervisors=None, default_hypervisor=None):
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 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
209 # hvparams is a mapping of hypervisor->hvparams dict
210 for hv_name, hv_params in hvparams.iteritems():
211 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
212 hv_class = hypervisor.GetHypervisor(hv_name)
213 hv_class.CheckParameterSyntax(hv_params)
215 # set up the inter-node password and certificate
216 _InitGanetiServerSetup()
218 # set up ssh config and /etc/hosts
219 f = open(constants.SSH_HOST_RSA_PUB, 'r')
224 sshkey = sshline.split(" ")[1]
226 utils.AddHostToEtcHosts(hostname.name)
229 # init of cluster config file
230 cluster_config = objects.Cluster(
232 rsahostkeypub=sshkey,
233 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
234 mac_prefix=mac_prefix,
235 volume_group_name=vg_name,
236 default_bridge=def_bridge,
237 tcpudp_port_pool=set(),
238 master_node=hostname.name,
239 master_ip=clustername.ip,
240 master_netdev=master_netdev,
241 cluster_name=clustername.name,
242 file_storage_dir=file_storage_dir,
243 enabled_hypervisors=enabled_hypervisors,
244 default_hypervisor=default_hypervisor,
245 beparams={constants.BEGR_DEFAULT: beparams},
247 candidate_pool_size=candidate_pool_size,
249 master_node_config = objects.Node(name=hostname.name,
250 primary_ip=hostname.ip,
251 secondary_ip=secondary_ip,
253 master_candidate=True,
254 offline=False, drained=False,
257 sscfg = InitConfig(constants.CONFIG_VERSION,
258 cluster_config, master_node_config)
259 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
260 cfg = config.ConfigWriter()
261 cfg.Update(cfg.GetClusterInfo())
263 # start the master ip
264 # TODO: Review rpc call from bootstrap
265 rpc.RpcRunner.call_node_start_master(hostname.name, True)
268 def InitConfig(version, cluster_config, master_node_config,
269 cfg_file=constants.CLUSTER_CONF_FILE):
270 """Create the initial cluster configuration.
272 It will contain the current node, which will also be the master
273 node, and no instances.
276 @param version: configuration version
277 @type cluster_config: L{objects.Cluster}
278 @param cluster_config: cluster configuration
279 @type master_node_config: L{objects.Node}
280 @param master_node_config: master node configuration
281 @type cfg_file: string
282 @param cfg_file: configuration file path
284 @rtype: L{ssconf.SimpleConfigWriter}
285 @return: initialized config instance
289 master_node_config.name: master_node_config,
292 config_data = objects.ConfigData(version=version,
293 cluster=cluster_config,
297 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
303 def FinalizeClusterDestroy(master):
304 """Execute the last steps of cluster destroy
306 This function shuts down all the daemons, completing the destroy
307 begun in cmdlib.LUDestroyOpcode.
310 result = rpc.RpcRunner.call_node_stop_master(master, True)
311 if result.failed or not result.data:
312 logging.warning("Could not disable the master role")
313 result = rpc.RpcRunner.call_node_leave_cluster(master)
314 if result.failed or not result.data:
315 logging.warning("Could not shutdown the node daemon and cleanup the node")
318 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
319 """Add a node to the cluster.
321 This function must be called before the actual opcode, and will ssh
322 to the remote node, copy the needed files, and start ganeti-noded,
323 allowing the master to do the rest via normal rpc calls.
325 @param cluster_name: the cluster name
326 @param node: the name of the new node
327 @param ssh_key_check: whether to do a strict key check
330 sshrunner = ssh.SshRunner(cluster_name)
332 noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
333 rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
335 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
336 # so we use this to detect an invalid certificate; as long as the
337 # cert doesn't contain this, the here-document will be correctly
338 # parsed by the shell sequence below
339 if (re.search('^!EOF\.', noded_cert, re.MULTILINE) or
340 re.search('^!EOF\.', rapi_cert, re.MULTILINE)):
341 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
343 if not noded_cert.endswith("\n"):
345 if not rapi_cert.endswith("\n"):
348 # set up inter-node password and certificate and restarts the node daemon
349 # and then connect with ssh to set password and start ganeti-noded
350 # note that all the below variables are sanitized at this point,
351 # either by being constants or by the checks above
352 mycommand = ("umask 077 && "
353 "cat > '%s' << '!EOF.' && \n"
355 "cat > '%s' << '!EOF.' && \n"
357 "chmod 0400 %s %s && "
359 (constants.SSL_CERT_FILE, noded_cert,
360 constants.RAPI_CERT_FILE, rapi_cert,
361 constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
362 constants.NODE_INITD_SCRIPT))
364 result = sshrunner.Run(node, 'root', mycommand, batch=False,
365 ask_key=ssh_key_check,
366 use_cluster_key=False,
367 strict_host_check=ssh_key_check)
369 raise errors.OpExecError("Remote command on node %s, error: %s,"
371 (node, result.fail_reason, result.output))
374 def MasterFailover():
375 """Failover the master node.
377 This checks that we are not already the master, and will cause the
378 current master to cease being master, and the non-master to become
382 sstore = ssconf.SimpleStore()
384 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
385 node_list = sstore.GetNodeList()
386 mc_list = sstore.GetMasterCandidates()
388 if old_master == new_master:
389 raise errors.OpPrereqError("This commands must be run on the node"
390 " where you want the new master to be."
391 " %s is already the master" %
394 if new_master not in mc_list:
395 mc_no_master = [name for name in mc_list if name != old_master]
396 raise errors.OpPrereqError("This node is not among the nodes marked"
397 " as master candidates. Only these nodes"
398 " can become masters. Current list of"
399 " master candidates is:\n"
400 "%s" % ('\n'.join(mc_no_master)))
402 vote_list = GatherMasterVotes(node_list)
405 voted_master = vote_list[0][0]
406 if voted_master is None:
407 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
409 elif voted_master != old_master:
410 raise errors.OpPrereqError("I have wrong configuration, I believe the"
411 " master is %s but the other nodes voted for"
412 " %s. Please resync the configuration of"
413 " this node." % (old_master, voted_master))
418 logging.info("Setting master to %s, old master: %s", new_master, old_master)
420 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
421 if result.failed or not result.data:
422 logging.error("Could not disable the master role on the old master"
423 " %s, please disable manually", old_master)
425 # Here we have a phase where no master should be running
427 # instantiate a real config writer, as we now know we have the
429 cfg = config.ConfigWriter()
431 cluster_info = cfg.GetClusterInfo()
432 cluster_info.master_node = new_master
433 # this will also regenerate the ssconf files, since we updated the
435 cfg.Update(cluster_info)
437 result = rpc.RpcRunner.call_node_start_master(new_master, True)
438 if result.failed or not result.data:
439 logging.error("Could not start the master role on the new master"
440 " %s, please check", new_master)
446 def GatherMasterVotes(node_list):
447 """Check the agreement on who is the master.
449 This function will return a list of (node, number of votes), ordered
450 by the number of votes. Errors will be denoted by the key 'None'.
452 Note that the sum of votes is the number of nodes this machine
453 knows, whereas the number of entries in the list could be different
454 (if some nodes vote for another master).
456 We remove ourselves from the list since we know that (bugs aside)
457 since we use the same source for configuration information for both
458 backend and boostrap, we'll always vote for ourselves.
460 @type node_list: list
461 @param node_list: the list of nodes to query for master info; the current
462 node wil be removed if it is in the list
464 @return: list of (node, votes)
467 myself = utils.HostInfo().name
469 node_list.remove(myself)
473 # no nodes left (eventually after removing myself)
475 results = rpc.RpcRunner.call_master_info(node_list)
476 if not isinstance(results, dict):
477 # this should not happen (unless internal error in rpc)
478 logging.critical("Can't complete rpc call, aborting master startup")
479 return [(None, len(node_list))]
484 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
485 # here the rpc layer should have already logged errors
486 if None not in votes:
490 master_node = data[2]
491 if master_node not in votes:
492 votes[master_node] = 0
493 votes[master_node] += 1
495 vote_list = [v for v in votes.items()]
496 # sort first on number of votes then on name, since we want None
497 # sorted later if we have the half of the nodes not responding, and
498 # half voting all for the same master
499 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)