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
43 """Setup the SSH configuration for the cluster.
45 This generates a dsa keypair for root, adds the pub key to the
46 permitted hosts and adds the hostkey to its own known hosts.
49 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
51 for name in priv_key, pub_key:
52 if os.path.exists(name):
53 utils.CreateBackup(name)
54 utils.RemoveFile(name)
56 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
60 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
63 f = open(pub_key, 'r')
65 utils.AddAuthorizedKey(auth_keys, f.read(8192))
70 def _GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
71 """Generates a self-signed SSL certificate.
74 @param file_name: Path to output file
76 @param validity: Validity for certificate in days
79 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
80 "-days", str(validity), "-nodes", "-x509",
81 "-keyout", file_name, "-out", file_name, "-batch"])
83 raise errors.OpExecError("Could not generate SSL certificate, command"
84 " %s had exitcode %s and error message %s" %
85 (result.cmd, result.exit_code, result.output))
87 os.chmod(file_name, 0400)
90 def _InitGanetiServerSetup():
91 """Setup the necessary configuration for the initial node daemon.
93 This creates the nodepass file containing the shared password for
94 the cluster and also generates the SSL certificate.
97 _GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
99 # Don't overwrite existing file
100 if not os.path.exists(constants.RAPI_CERT_FILE):
101 _GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
103 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
106 raise errors.OpExecError("Could not start the node daemon, command %s"
107 " had exitcode %s and error %s" %
108 (result.cmd, result.exit_code, result.output))
111 def InitCluster(cluster_name, mac_prefix, def_bridge,
112 master_netdev, file_storage_dir, candidate_pool_size,
113 secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
114 enabled_hypervisors=None, default_hypervisor=None):
115 """Initialise the cluster.
117 @type candidate_pool_size: int
118 @param candidate_pool_size: master candidate pool size
121 # TODO: complete the docstring
122 if config.ConfigWriter.IsCluster():
123 raise errors.OpPrereqError("Cluster is already initialised")
125 hostname = utils.HostInfo()
127 if hostname.ip.startswith("127."):
128 raise errors.OpPrereqError("This host's IP resolves to the private"
129 " range (%s). Please fix DNS or %s." %
130 (hostname.ip, constants.ETC_HOSTS))
132 if not utils.OwnIpAddress(hostname.ip):
133 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
134 " to %s,\nbut this ip address does not"
135 " belong to this host."
136 " Aborting." % hostname.ip)
138 clustername = utils.HostInfo(cluster_name)
140 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
142 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
145 if not utils.IsValidIP(secondary_ip):
146 raise errors.OpPrereqError("Invalid secondary ip given")
147 if (secondary_ip != hostname.ip and
148 not utils.OwnIpAddress(secondary_ip)):
149 raise errors.OpPrereqError("You gave %s as secondary IP,"
150 " but it does not belong to this host." %
153 secondary_ip = hostname.ip
155 if vg_name is not None:
156 # Check if volume group is valid
157 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
158 constants.MIN_VG_SIZE)
160 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
161 " you are not using lvm" % vgstatus)
163 file_storage_dir = os.path.normpath(file_storage_dir)
165 if not os.path.isabs(file_storage_dir):
166 raise errors.OpPrereqError("The file storage directory you passed is"
167 " not an absolute path.")
169 if not os.path.exists(file_storage_dir):
171 os.makedirs(file_storage_dir, 0750)
173 raise errors.OpPrereqError("Cannot create file storage directory"
175 (file_storage_dir, err))
177 if not os.path.isdir(file_storage_dir):
178 raise errors.OpPrereqError("The file storage directory '%s' is not"
179 " a directory." % file_storage_dir)
181 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
182 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
184 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
186 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
188 result.output.strip()))
190 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
191 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
192 raise errors.OpPrereqError("Init.d script '%s' missing or not"
193 " executable." % constants.NODE_INITD_SCRIPT)
195 utils.CheckBEParams(beparams)
197 # set up the inter-node password and certificate
198 _InitGanetiServerSetup()
200 # set up ssh config and /etc/hosts
201 f = open(constants.SSH_HOST_RSA_PUB, 'r')
206 sshkey = sshline.split(" ")[1]
208 utils.AddHostToEtcHosts(hostname.name)
211 # init of cluster config file
212 cluster_config = objects.Cluster(
214 rsahostkeypub=sshkey,
215 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
216 mac_prefix=mac_prefix,
217 volume_group_name=vg_name,
218 default_bridge=def_bridge,
219 tcpudp_port_pool=set(),
220 master_node=hostname.name,
221 master_ip=clustername.ip,
222 master_netdev=master_netdev,
223 cluster_name=clustername.name,
224 file_storage_dir=file_storage_dir,
225 enabled_hypervisors=enabled_hypervisors,
226 default_hypervisor=default_hypervisor,
227 beparams={constants.BEGR_DEFAULT: beparams},
229 candidate_pool_size=candidate_pool_size,
231 master_node_config = objects.Node(name=hostname.name,
232 primary_ip=hostname.ip,
233 secondary_ip=secondary_ip,
235 master_candidate=True,
239 sscfg = InitConfig(constants.CONFIG_VERSION,
240 cluster_config, master_node_config)
241 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
242 cfg = config.ConfigWriter()
243 cfg.Update(cfg.GetClusterInfo())
245 # start the master ip
246 # TODO: Review rpc call from bootstrap
247 rpc.RpcRunner.call_node_start_master(hostname.name, True)
250 def InitConfig(version, cluster_config, master_node_config,
251 cfg_file=constants.CLUSTER_CONF_FILE):
252 """Create the initial cluster configuration.
254 It will contain the current node, which will also be the master
255 node, and no instances.
258 @param version: configuration version
259 @type cluster_config: L{objects.Cluster}
260 @param cluster_config: cluster configuration
261 @type master_node_config: L{objects.Node}
262 @param master_node_config: master node configuration
263 @type cfg_file: string
264 @param cfg_file: configuration file path
266 @rtype: L{ssconf.SimpleConfigWriter}
267 @returns: initialized config instance
271 master_node_config.name: master_node_config,
274 config_data = objects.ConfigData(version=version,
275 cluster=cluster_config,
279 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
285 def FinalizeClusterDestroy(master):
286 """Execute the last steps of cluster destroy
288 This function shuts down all the daemons, completing the destroy
289 begun in cmdlib.LUDestroyOpcode.
292 result = rpc.RpcRunner.call_node_stop_master(master, True)
293 if result.failed or not result.data:
294 logging.warning("Could not disable the master role")
295 result = rpc.RpcRunner.call_node_leave_cluster(master)
296 if result.failed or not result.data:
297 logging.warning("Could not shutdown the node daemon and cleanup the node")
300 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
301 """Add a node to the cluster.
303 This function must be called before the actual opcode, and will ssh
304 to the remote node, copy the needed files, and start ganeti-noded,
305 allowing the master to do the rest via normal rpc calls.
307 @param cluster_name: the cluster name
308 @param node: the name of the new node
309 @param ssh_key_check: whether to do a strict key check
312 sshrunner = ssh.SshRunner(cluster_name)
313 gntpem = utils.ReadFile(constants.SSL_CERT_FILE)
314 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
315 # so we use this to detect an invalid certificate; as long as the
316 # cert doesn't contain this, the here-document will be correctly
317 # parsed by the shell sequence below
318 if re.search('^!EOF\.', gntpem, re.MULTILINE):
319 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
320 if not gntpem.endswith("\n"):
321 raise errors.OpExecError("PEM must end with newline")
323 # set up inter-node password and certificate and restarts the node daemon
324 # and then connect with ssh to set password and start ganeti-noded
325 # note that all the below variables are sanitized at this point,
326 # either by being constants or by the checks above
327 mycommand = ("umask 077 && "
328 "cat > '%s' << '!EOF.' && \n"
329 "%s!EOF.\n%s restart" %
330 (constants.SSL_CERT_FILE, gntpem,
331 constants.NODE_INITD_SCRIPT))
333 result = sshrunner.Run(node, 'root', mycommand, batch=False,
334 ask_key=ssh_key_check,
335 use_cluster_key=False,
336 strict_host_check=ssh_key_check)
338 raise errors.OpExecError("Remote command on node %s, error: %s,"
340 (node, result.fail_reason, result.output))
343 def MasterFailover():
344 """Failover the master node.
346 This checks that we are not already the master, and will cause the
347 current master to cease being master, and the non-master to become
351 sstore = ssconf.SimpleStore()
353 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
354 node_list = sstore.GetNodeList()
355 mc_list = sstore.GetMasterCandidates()
357 if old_master == new_master:
358 raise errors.OpPrereqError("This commands must be run on the node"
359 " where you want the new master to be."
360 " %s is already the master" %
363 if new_master not in mc_list:
364 mc_no_master = [name for name in mc_list if name != old_master]
365 raise errors.OpPrereqError("This node is not among the nodes marked"
366 " as master candidates. Only these nodes"
367 " can become masters. Current list of"
368 " master candidates is:\n"
369 "%s" % ('\n'.join(mc_no_master)))
371 vote_list = GatherMasterVotes(node_list)
374 voted_master = vote_list[0][0]
375 if voted_master is None:
376 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
378 elif voted_master != old_master:
379 raise errors.OpPrereqError("I have wrong configuration, I believe the"
380 " master is %s but the other nodes voted for"
381 " %s. Please resync the configuration of"
382 " this node." % (old_master, voted_master))
387 logging.info("Setting master to %s, old master: %s", new_master, old_master)
389 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
390 if result.failed or not result.data:
391 logging.error("Could not disable the master role on the old master"
392 " %s, please disable manually", old_master)
394 # Here we have a phase where no master should be running
396 # instantiate a real config writer, as we now know we have the
398 cfg = config.ConfigWriter()
400 cluster_info = cfg.GetClusterInfo()
401 cluster_info.master_node = new_master
402 # this will also regenerate the ssconf files, since we updated the
404 cfg.Update(cluster_info)
406 result = rpc.RpcRunner.call_node_start_master(new_master, True)
407 if result.failed or not result.data:
408 logging.error("Could not start the master role on the new master"
409 " %s, please check", new_master)
415 def GatherMasterVotes(node_list):
416 """Check the agreement on who is the master.
418 This function will return a list of (node, number of votes), ordered
419 by the number of votes. Errors will be denoted by the key 'None'.
421 Note that the sum of votes is the number of nodes this machine
422 knows, whereas the number of entries in the list could be different
423 (if some nodes vote for another master).
425 We remove ourselves from the list since we know that (bugs aside)
426 since we use the same source for configuration information for both
427 backend and boostrap, we'll always vote for ourselves.
429 @type node_list: list
430 @param node_list: the list of nodes to query for master info; the current
431 node wil be removed if it is in the list
433 @return: list of (node, votes)
436 myself = utils.HostInfo().name
438 node_list.remove(myself)
442 # no nodes left (eventually after removing myself)
444 results = rpc.RpcRunner.call_master_info(node_list)
445 if not isinstance(results, dict):
446 # this should not happen (unless internal error in rpc)
447 logging.critical("Can't complete rpc call, aborting master startup")
448 return [(None, len(node_list))]
453 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
454 # here the rpc layer should have already logged errors
455 if None not in votes:
459 master_node = data[2]
460 if master_node not in votes:
461 votes[master_node] = 0
462 votes[master_node] += 1
464 vote_list = [v for v in votes.items()]
465 # sort first on number of votes then on name, since we want None
466 # sorted later if we have the half of the nodes not responding, and
467 # half voting all for the same master
468 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)