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
41 def _InitSSHSetup(node):
42 """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 node: the name of this host as a fqdn
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 _InitGanetiServerSetup():
74 """Setup the necessary configuration for the initial node daemon.
76 This creates the nodepass file containing the shared password for
77 the cluster and also generates the SSL certificate.
80 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
81 "-days", str(365*5), "-nodes", "-x509",
82 "-keyout", constants.SSL_CERT_FILE,
83 "-out", constants.SSL_CERT_FILE, "-batch"])
85 raise errors.OpExecError("could not generate server ssl cert, command"
86 " %s had exitcode %s and error message %s" %
87 (result.cmd, result.exit_code, result.output))
89 os.chmod(constants.SSL_CERT_FILE, 0400)
91 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
94 raise errors.OpExecError("Could not start the node daemon, command %s"
95 " had exitcode %s and error %s" %
96 (result.cmd, result.exit_code, result.output))
99 def InitCluster(cluster_name, mac_prefix, def_bridge,
100 master_netdev, file_storage_dir, candidate_pool_size,
101 secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
102 enabled_hypervisors=None, default_hypervisor=None):
103 """Initialise the cluster.
105 @type candidate_pool_size: int
106 @param candidate_pool_size: master candidate pool size
109 # TODO: complete the docstring
110 if config.ConfigWriter.IsCluster():
111 raise errors.OpPrereqError("Cluster is already initialised")
113 hostname = utils.HostInfo()
115 if hostname.ip.startswith("127."):
116 raise errors.OpPrereqError("This host's IP resolves to the private"
117 " range (%s). Please fix DNS or %s." %
118 (hostname.ip, constants.ETC_HOSTS))
120 if not utils.OwnIpAddress(hostname.ip):
121 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
122 " to %s,\nbut this ip address does not"
123 " belong to this host."
124 " Aborting." % hostname.ip)
126 clustername = utils.HostInfo(cluster_name)
128 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
130 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
133 if not utils.IsValidIP(secondary_ip):
134 raise errors.OpPrereqError("Invalid secondary ip given")
135 if (secondary_ip != hostname.ip and
136 not utils.OwnIpAddress(secondary_ip)):
137 raise errors.OpPrereqError("You gave %s as secondary IP,"
138 " but it does not belong to this host." %
141 secondary_ip = hostname.ip
143 if vg_name is not None:
144 # Check if volume group is valid
145 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
146 constants.MIN_VG_SIZE)
148 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
149 " you are not using lvm" % vgstatus)
151 file_storage_dir = os.path.normpath(file_storage_dir)
153 if not os.path.isabs(file_storage_dir):
154 raise errors.OpPrereqError("The file storage directory you passed is"
155 " not an absolute path.")
157 if not os.path.exists(file_storage_dir):
159 os.makedirs(file_storage_dir, 0750)
161 raise errors.OpPrereqError("Cannot create file storage directory"
163 (file_storage_dir, err))
165 if not os.path.isdir(file_storage_dir):
166 raise errors.OpPrereqError("The file storage directory '%s' is not"
167 " a directory." % file_storage_dir)
169 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
170 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
172 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
174 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
176 result.output.strip()))
178 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
179 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
180 raise errors.OpPrereqError("Init.d script '%s' missing or not"
181 " executable." % constants.NODE_INITD_SCRIPT)
183 utils.CheckBEParams(beparams)
185 # set up the inter-node password and certificate
186 _InitGanetiServerSetup()
188 # set up ssh config and /etc/hosts
189 f = open(constants.SSH_HOST_RSA_PUB, 'r')
194 sshkey = sshline.split(" ")[1]
196 utils.AddHostToEtcHosts(hostname.name)
197 _InitSSHSetup(hostname.name)
199 # init of cluster config file
200 cluster_config = objects.Cluster(
202 rsahostkeypub=sshkey,
203 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
204 mac_prefix=mac_prefix,
205 volume_group_name=vg_name,
206 default_bridge=def_bridge,
207 tcpudp_port_pool=set(),
208 master_node=hostname.name,
209 master_ip=clustername.ip,
210 master_netdev=master_netdev,
211 cluster_name=clustername.name,
212 file_storage_dir=file_storage_dir,
213 enabled_hypervisors=enabled_hypervisors,
214 default_hypervisor=default_hypervisor,
215 beparams={constants.BEGR_DEFAULT: beparams},
217 candidate_pool_size=candidate_pool_size,
219 master_node_config = objects.Node(name=hostname.name,
220 primary_ip=hostname.ip,
221 secondary_ip=secondary_ip,
223 master_candidate=True,
226 sscfg = InitConfig(constants.CONFIG_VERSION,
227 cluster_config, master_node_config)
228 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
229 cfg = config.ConfigWriter()
230 cfg.Update(cfg.GetClusterInfo())
232 # start the master ip
233 # TODO: Review rpc call from bootstrap
234 rpc.RpcRunner.call_node_start_master(hostname.name, True)
237 def InitConfig(version, cluster_config, master_node_config,
238 cfg_file=constants.CLUSTER_CONF_FILE):
239 """Create the initial cluster configuration.
241 It will contain the current node, which will also be the master
242 node, and no instances.
245 @param version: Configuration version
246 @type cluster_config: objects.Cluster
247 @param cluster_config: Cluster configuration
248 @type master_node_config: objects.Node
249 @param master_node_config: Master node configuration
250 @type file_name: string
251 @param file_name: Configuration file path
253 @rtype: ssconf.SimpleConfigWriter
254 @returns: Initialized config instance
258 master_node_config.name: master_node_config,
261 config_data = objects.ConfigData(version=version,
262 cluster=cluster_config,
266 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
272 def FinalizeClusterDestroy(master):
273 """Execute the last steps of cluster destroy
275 This function shuts down all the daemons, completing the destroy
276 begun in cmdlib.LUDestroyOpcode.
279 result = rpc.RpcRunner.call_node_stop_master(master, True)
280 if result.failed or not result.data:
281 logging.warning("Could not disable the master role")
282 result = rpc.RpcRunner.call_node_leave_cluster(master)
283 if result.failed or not result.data:
284 logging.warning("Could not shutdown the node daemon and cleanup the node")
287 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
288 """Add a node to the cluster.
290 This function must be called before the actual opcode, and will ssh
291 to the remote node, copy the needed files, and start ganeti-noded,
292 allowing the master to do the rest via normal rpc calls.
294 @param cluster_name: the cluster name
295 @param node: the name of the new node
296 @param ssh_key_check: whether to do a strict key check
299 sshrunner = ssh.SshRunner(cluster_name)
300 gntpem = utils.ReadFile(constants.SSL_CERT_FILE)
301 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
302 # so we use this to detect an invalid certificate; as long as the
303 # cert doesn't contain this, the here-document will be correctly
304 # parsed by the shell sequence below
305 if re.search('^!EOF\.', gntpem, re.MULTILINE):
306 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
307 if not gntpem.endswith("\n"):
308 raise errors.OpExecError("PEM must end with newline")
310 # set up inter-node password and certificate and restarts the node daemon
311 # and then connect with ssh to set password and start ganeti-noded
312 # note that all the below variables are sanitized at this point,
313 # either by being constants or by the checks above
314 mycommand = ("umask 077 && "
315 "cat > '%s' << '!EOF.' && \n"
316 "%s!EOF.\n%s restart" %
317 (constants.SSL_CERT_FILE, gntpem,
318 constants.NODE_INITD_SCRIPT))
320 result = sshrunner.Run(node, 'root', mycommand, batch=False,
321 ask_key=ssh_key_check,
322 use_cluster_key=False,
323 strict_host_check=ssh_key_check)
325 raise errors.OpExecError("Remote command on node %s, error: %s,"
327 (node, result.fail_reason, result.output))
330 def MasterFailover():
331 """Failover the master node.
333 This checks that we are not already the master, and will cause the
334 current master to cease being master, and the non-master to become
338 sstore = ssconf.SimpleStore()
340 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
341 node_list = sstore.GetNodeList()
342 mc_list = sstore.GetMasterCandidates()
344 if old_master == new_master:
345 raise errors.OpPrereqError("This commands must be run on the node"
346 " where you want the new master to be."
347 " %s is already the master" %
350 if new_master not in mc_list:
351 mc_no_master = [name for name in mc_list if name != old_master]
352 raise errors.OpPrereqError("This node is not among the nodes marked"
353 " as master candidates. Only these nodes"
354 " can become masters. Current list of"
355 " master candidates is:\n"
356 "%s" % ('\n'.join(mc_no_master)))
358 vote_list = GatherMasterVotes(node_list)
361 voted_master = vote_list[0][0]
362 if voted_master is None:
363 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
365 elif voted_master != old_master:
366 raise errors.OpPrereqError("I have wrong configuration, I believe the"
367 " master is %s but the other nodes voted for"
368 " %s. Please resync the configuration of"
369 " this node." % (old_master, voted_master))
374 logging.info("Setting master to %s, old master: %s", new_master, old_master)
376 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
377 if result.failed or not result.data:
378 logging.error("Could not disable the master role on the old master"
379 " %s, please disable manually", old_master)
381 # Here we have a phase where no master should be running
383 # instantiate a real config writer, as we now know we have the
385 cfg = config.ConfigWriter()
387 cluster_info = cfg.GetClusterInfo()
388 cluster_info.master_node = new_master
389 # this will also regenerate the ssconf files, since we updated the
391 cfg.Update(cluster_info)
393 result = rpc.RpcRunner.call_node_start_master(new_master, True)
394 if result.failed or not result.data:
395 logging.error("Could not start the master role on the new master"
396 " %s, please check", new_master)
402 def GatherMasterVotes(node_list):
403 """Check the agreement on who is the master.
405 This function will return a list of (node, number of votes), ordered
406 by the number of votes. Errors will be denoted by the key 'None'.
408 Note that the sum of votes is the number of nodes this machine
409 knows, whereas the number of entries in the list could be different
410 (if some nodes vote for another master).
412 We remove ourselves from the list since we know that (bugs aside)
413 since we use the same source for configuration information for both
414 backend and boostrap, we'll always vote for ourselves.
416 @type node_list: list
417 @param node_list: the list of nodes to query for master info; the current
418 node wil be removed if it is in the list
420 @return: list of (node, votes)
423 myself = utils.HostInfo().name
425 node_list.remove(myself)
429 # no nodes left (eventually after removing myself)
431 results = rpc.RpcRunner.call_master_info(node_list)
432 if not isinstance(results, dict):
433 # this should not happen (unless internal error in rpc)
434 logging.critical("Can't complete rpc call, aborting master startup")
435 return [(None, len(node_list))]
436 positive = negative = 0
442 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
443 # here the rpc layer should have already logged errors
444 if None not in votes:
448 master_node = data[2]
449 if master_node not in votes:
450 votes[master_node] = 0
451 votes[master_node] += 1
453 vote_list = [v for v in votes.items()]
454 # sort first on number of votes then on name, since we want None
455 # sorted later if we have the half of the nodes not responding, and
456 # half voting all for the same master
457 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)