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
42 """Setup the SSH configuration for the cluster.
44 This generates a dsa keypair for root, adds the pub key to the
45 permitted hosts and adds the hostkey to its own known hosts.
48 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
50 for name in priv_key, pub_key:
51 if os.path.exists(name):
52 utils.CreateBackup(name)
53 utils.RemoveFile(name)
55 result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
59 raise errors.OpExecError("Could not generate ssh keypair, error %s" %
62 f = open(pub_key, 'r')
64 utils.AddAuthorizedKey(auth_keys, f.read(8192))
69 def _InitGanetiServerSetup():
70 """Setup the necessary configuration for the initial node daemon.
72 This creates the nodepass file containing the shared password for
73 the cluster and also generates the SSL certificate.
76 result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
77 "-days", str(365*5), "-nodes", "-x509",
78 "-keyout", constants.SSL_CERT_FILE,
79 "-out", constants.SSL_CERT_FILE, "-batch"])
81 raise errors.OpExecError("could not generate server ssl cert, command"
82 " %s had exitcode %s and error message %s" %
83 (result.cmd, result.exit_code, result.output))
85 os.chmod(constants.SSL_CERT_FILE, 0400)
87 result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
90 raise errors.OpExecError("Could not start the node daemon, command %s"
91 " had exitcode %s and error %s" %
92 (result.cmd, result.exit_code, result.output))
95 def InitCluster(cluster_name, mac_prefix, def_bridge,
96 master_netdev, file_storage_dir, candidate_pool_size,
97 secondary_ip=None, vg_name=None, beparams=None, hvparams=None,
98 enabled_hypervisors=None, default_hypervisor=None):
99 """Initialise the cluster.
101 @type candidate_pool_size: int
102 @param candidate_pool_size: master candidate pool size
105 # TODO: complete the docstring
106 if config.ConfigWriter.IsCluster():
107 raise errors.OpPrereqError("Cluster is already initialised")
109 hostname = utils.HostInfo()
111 if hostname.ip.startswith("127."):
112 raise errors.OpPrereqError("This host's IP resolves to the private"
113 " range (%s). Please fix DNS or %s." %
114 (hostname.ip, constants.ETC_HOSTS))
116 if not utils.OwnIpAddress(hostname.ip):
117 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
118 " to %s,\nbut this ip address does not"
119 " belong to this host."
120 " Aborting." % hostname.ip)
122 clustername = utils.HostInfo(cluster_name)
124 if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
126 raise errors.OpPrereqError("Cluster IP already active. Aborting.")
129 if not utils.IsValidIP(secondary_ip):
130 raise errors.OpPrereqError("Invalid secondary ip given")
131 if (secondary_ip != hostname.ip and
132 not utils.OwnIpAddress(secondary_ip)):
133 raise errors.OpPrereqError("You gave %s as secondary IP,"
134 " but it does not belong to this host." %
137 secondary_ip = hostname.ip
139 if vg_name is not None:
140 # Check if volume group is valid
141 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
142 constants.MIN_VG_SIZE)
144 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
145 " you are not using lvm" % vgstatus)
147 file_storage_dir = os.path.normpath(file_storage_dir)
149 if not os.path.isabs(file_storage_dir):
150 raise errors.OpPrereqError("The file storage directory you passed is"
151 " not an absolute path.")
153 if not os.path.exists(file_storage_dir):
155 os.makedirs(file_storage_dir, 0750)
157 raise errors.OpPrereqError("Cannot create file storage directory"
159 (file_storage_dir, err))
161 if not os.path.isdir(file_storage_dir):
162 raise errors.OpPrereqError("The file storage directory '%s' is not"
163 " a directory." % file_storage_dir)
165 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
166 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
168 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
170 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
172 result.output.strip()))
174 if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
175 os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
176 raise errors.OpPrereqError("Init.d script '%s' missing or not"
177 " executable." % constants.NODE_INITD_SCRIPT)
179 utils.CheckBEParams(beparams)
181 # set up the inter-node password and certificate
182 _InitGanetiServerSetup()
184 # set up ssh config and /etc/hosts
185 f = open(constants.SSH_HOST_RSA_PUB, 'r')
190 sshkey = sshline.split(" ")[1]
192 utils.AddHostToEtcHosts(hostname.name)
195 # init of cluster config file
196 cluster_config = objects.Cluster(
198 rsahostkeypub=sshkey,
199 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
200 mac_prefix=mac_prefix,
201 volume_group_name=vg_name,
202 default_bridge=def_bridge,
203 tcpudp_port_pool=set(),
204 master_node=hostname.name,
205 master_ip=clustername.ip,
206 master_netdev=master_netdev,
207 cluster_name=clustername.name,
208 file_storage_dir=file_storage_dir,
209 enabled_hypervisors=enabled_hypervisors,
210 default_hypervisor=default_hypervisor,
211 beparams={constants.BEGR_DEFAULT: beparams},
213 candidate_pool_size=candidate_pool_size,
215 master_node_config = objects.Node(name=hostname.name,
216 primary_ip=hostname.ip,
217 secondary_ip=secondary_ip,
219 master_candidate=True,
223 sscfg = InitConfig(constants.CONFIG_VERSION,
224 cluster_config, master_node_config)
225 ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE)
226 cfg = config.ConfigWriter()
227 cfg.Update(cfg.GetClusterInfo())
229 # start the master ip
230 # TODO: Review rpc call from bootstrap
231 rpc.RpcRunner.call_node_start_master(hostname.name, True)
234 def InitConfig(version, cluster_config, master_node_config,
235 cfg_file=constants.CLUSTER_CONF_FILE):
236 """Create the initial cluster configuration.
238 It will contain the current node, which will also be the master
239 node, and no instances.
242 @param version: configuration version
243 @type cluster_config: L{objects.Cluster}
244 @param cluster_config: cluster configuration
245 @type master_node_config: L{objects.Node}
246 @param master_node_config: master node configuration
247 @type cfg_file: string
248 @param cfg_file: configuration file path
250 @rtype: L{ssconf.SimpleConfigWriter}
251 @returns: initialized config instance
255 master_node_config.name: master_node_config,
258 config_data = objects.ConfigData(version=version,
259 cluster=cluster_config,
263 cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
269 def FinalizeClusterDestroy(master):
270 """Execute the last steps of cluster destroy
272 This function shuts down all the daemons, completing the destroy
273 begun in cmdlib.LUDestroyOpcode.
276 result = rpc.RpcRunner.call_node_stop_master(master, True)
277 if result.failed or not result.data:
278 logging.warning("Could not disable the master role")
279 result = rpc.RpcRunner.call_node_leave_cluster(master)
280 if result.failed or not result.data:
281 logging.warning("Could not shutdown the node daemon and cleanup the node")
284 def SetupNodeDaemon(cluster_name, node, ssh_key_check):
285 """Add a node to the cluster.
287 This function must be called before the actual opcode, and will ssh
288 to the remote node, copy the needed files, and start ganeti-noded,
289 allowing the master to do the rest via normal rpc calls.
291 @param cluster_name: the cluster name
292 @param node: the name of the new node
293 @param ssh_key_check: whether to do a strict key check
296 sshrunner = ssh.SshRunner(cluster_name)
297 gntpem = utils.ReadFile(constants.SSL_CERT_FILE)
298 # in the base64 pem encoding, neither '!' nor '.' are valid chars,
299 # so we use this to detect an invalid certificate; as long as the
300 # cert doesn't contain this, the here-document will be correctly
301 # parsed by the shell sequence below
302 if re.search('^!EOF\.', gntpem, re.MULTILINE):
303 raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
304 if not gntpem.endswith("\n"):
305 raise errors.OpExecError("PEM must end with newline")
307 # set up inter-node password and certificate and restarts the node daemon
308 # and then connect with ssh to set password and start ganeti-noded
309 # note that all the below variables are sanitized at this point,
310 # either by being constants or by the checks above
311 mycommand = ("umask 077 && "
312 "cat > '%s' << '!EOF.' && \n"
313 "%s!EOF.\n%s restart" %
314 (constants.SSL_CERT_FILE, gntpem,
315 constants.NODE_INITD_SCRIPT))
317 result = sshrunner.Run(node, 'root', mycommand, batch=False,
318 ask_key=ssh_key_check,
319 use_cluster_key=False,
320 strict_host_check=ssh_key_check)
322 raise errors.OpExecError("Remote command on node %s, error: %s,"
324 (node, result.fail_reason, result.output))
327 def MasterFailover():
328 """Failover the master node.
330 This checks that we are not already the master, and will cause the
331 current master to cease being master, and the non-master to become
335 sstore = ssconf.SimpleStore()
337 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
338 node_list = sstore.GetNodeList()
339 mc_list = sstore.GetMasterCandidates()
341 if old_master == new_master:
342 raise errors.OpPrereqError("This commands must be run on the node"
343 " where you want the new master to be."
344 " %s is already the master" %
347 if new_master not in mc_list:
348 mc_no_master = [name for name in mc_list if name != old_master]
349 raise errors.OpPrereqError("This node is not among the nodes marked"
350 " as master candidates. Only these nodes"
351 " can become masters. Current list of"
352 " master candidates is:\n"
353 "%s" % ('\n'.join(mc_no_master)))
355 vote_list = GatherMasterVotes(node_list)
358 voted_master = vote_list[0][0]
359 if voted_master is None:
360 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
362 elif voted_master != old_master:
363 raise errors.OpPrereqError("I have wrong configuration, I believe the"
364 " master is %s but the other nodes voted for"
365 " %s. Please resync the configuration of"
366 " this node." % (old_master, voted_master))
371 logging.info("Setting master to %s, old master: %s", new_master, old_master)
373 result = rpc.RpcRunner.call_node_stop_master(old_master, True)
374 if result.failed or not result.data:
375 logging.error("Could not disable the master role on the old master"
376 " %s, please disable manually", old_master)
378 # Here we have a phase where no master should be running
380 # instantiate a real config writer, as we now know we have the
382 cfg = config.ConfigWriter()
384 cluster_info = cfg.GetClusterInfo()
385 cluster_info.master_node = new_master
386 # this will also regenerate the ssconf files, since we updated the
388 cfg.Update(cluster_info)
390 result = rpc.RpcRunner.call_node_start_master(new_master, True)
391 if result.failed or not result.data:
392 logging.error("Could not start the master role on the new master"
393 " %s, please check", new_master)
399 def GatherMasterVotes(node_list):
400 """Check the agreement on who is the master.
402 This function will return a list of (node, number of votes), ordered
403 by the number of votes. Errors will be denoted by the key 'None'.
405 Note that the sum of votes is the number of nodes this machine
406 knows, whereas the number of entries in the list could be different
407 (if some nodes vote for another master).
409 We remove ourselves from the list since we know that (bugs aside)
410 since we use the same source for configuration information for both
411 backend and boostrap, we'll always vote for ourselves.
413 @type node_list: list
414 @param node_list: the list of nodes to query for master info; the current
415 node wil be removed if it is in the list
417 @return: list of (node, votes)
420 myself = utils.HostInfo().name
422 node_list.remove(myself)
426 # no nodes left (eventually after removing myself)
428 results = rpc.RpcRunner.call_master_info(node_list)
429 if not isinstance(results, dict):
430 # this should not happen (unless internal error in rpc)
431 logging.critical("Can't complete rpc call, aborting master startup")
432 return [(None, len(node_list))]
437 if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3:
438 # here the rpc layer should have already logged errors
439 if None not in votes:
443 master_node = data[2]
444 if master_node not in votes:
445 votes[master_node] = 0
446 votes[master_node] += 1
448 vote_list = [v for v in votes.items()]
449 # sort first on number of votes then on name, since we want None
450 # sorted later if we have the half of the nodes not responding, and
451 # half voting all for the same master
452 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True)