root / lib / bootstrap.py @ e38220e4
History | View | Annotate | Download (16 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008 Google Inc.
|
5 |
#
|
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.
|
10 |
#
|
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.
|
15 |
#
|
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
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
|
22 |
"""Functions to bootstrap a new cluster.
|
23 |
|
24 |
"""
|
25 |
|
26 |
import os |
27 |
import os.path |
28 |
import sha |
29 |
import re |
30 |
import logging |
31 |
|
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 |
40 |
|
41 |
|
42 |
def _InitSSHSetup(): |
43 |
"""Setup the SSH configuration for the cluster.
|
44 |
|
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.
|
47 |
|
48 |
"""
|
49 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) |
50 |
|
51 |
for name in priv_key, pub_key: |
52 |
if os.path.exists(name):
|
53 |
utils.CreateBackup(name) |
54 |
utils.RemoveFile(name) |
55 |
|
56 |
result = utils.RunCmd(["ssh-keygen", "-t", "dsa", |
57 |
"-f", priv_key,
|
58 |
"-q", "-N", ""]) |
59 |
if result.failed:
|
60 |
raise errors.OpExecError("Could not generate ssh keypair, error %s" % |
61 |
result.output) |
62 |
|
63 |
f = open(pub_key, 'r') |
64 |
try:
|
65 |
utils.AddAuthorizedKey(auth_keys, f.read(8192))
|
66 |
finally:
|
67 |
f.close() |
68 |
|
69 |
|
70 |
def _InitGanetiServerSetup(): |
71 |
"""Setup the necessary configuration for the initial node daemon.
|
72 |
|
73 |
This creates the nodepass file containing the shared password for
|
74 |
the cluster and also generates the SSL certificate.
|
75 |
|
76 |
"""
|
77 |
result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024", |
78 |
"-days", str(365*5), "-nodes", "-x509", |
79 |
"-keyout", constants.SSL_CERT_FILE,
|
80 |
"-out", constants.SSL_CERT_FILE, "-batch"]) |
81 |
if result.failed:
|
82 |
raise errors.OpExecError("could not generate server ssl cert, command" |
83 |
" %s had exitcode %s and error message %s" %
|
84 |
(result.cmd, result.exit_code, result.output)) |
85 |
|
86 |
os.chmod(constants.SSL_CERT_FILE, 0400)
|
87 |
|
88 |
result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
|
89 |
|
90 |
if result.failed:
|
91 |
raise errors.OpExecError("Could not start the node daemon, command %s" |
92 |
" had exitcode %s and error %s" %
|
93 |
(result.cmd, result.exit_code, result.output)) |
94 |
|
95 |
|
96 |
def InitCluster(cluster_name, mac_prefix, def_bridge, |
97 |
master_netdev, file_storage_dir, candidate_pool_size, |
98 |
secondary_ip=None, vg_name=None, beparams=None, hvparams=None, |
99 |
enabled_hypervisors=None, default_hypervisor=None): |
100 |
"""Initialise the cluster.
|
101 |
|
102 |
@type candidate_pool_size: int
|
103 |
@param candidate_pool_size: master candidate pool size
|
104 |
|
105 |
"""
|
106 |
# TODO: complete the docstring
|
107 |
if config.ConfigWriter.IsCluster():
|
108 |
raise errors.OpPrereqError("Cluster is already initialised") |
109 |
|
110 |
hostname = utils.HostInfo() |
111 |
|
112 |
if hostname.ip.startswith("127."): |
113 |
raise errors.OpPrereqError("This host's IP resolves to the private" |
114 |
" range (%s). Please fix DNS or %s." %
|
115 |
(hostname.ip, constants.ETC_HOSTS)) |
116 |
|
117 |
if not utils.OwnIpAddress(hostname.ip): |
118 |
raise errors.OpPrereqError("Inconsistency: this host's name resolves" |
119 |
" to %s,\nbut this ip address does not"
|
120 |
" belong to this host."
|
121 |
" Aborting." % hostname.ip)
|
122 |
|
123 |
clustername = utils.HostInfo(cluster_name) |
124 |
|
125 |
if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
|
126 |
timeout=5):
|
127 |
raise errors.OpPrereqError("Cluster IP already active. Aborting.") |
128 |
|
129 |
if secondary_ip:
|
130 |
if not utils.IsValidIP(secondary_ip): |
131 |
raise errors.OpPrereqError("Invalid secondary ip given") |
132 |
if (secondary_ip != hostname.ip and |
133 |
not utils.OwnIpAddress(secondary_ip)):
|
134 |
raise errors.OpPrereqError("You gave %s as secondary IP," |
135 |
" but it does not belong to this host." %
|
136 |
secondary_ip) |
137 |
else:
|
138 |
secondary_ip = hostname.ip |
139 |
|
140 |
if vg_name is not None: |
141 |
# Check if volume group is valid
|
142 |
vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, |
143 |
constants.MIN_VG_SIZE) |
144 |
if vgstatus:
|
145 |
raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" |
146 |
" you are not using lvm" % vgstatus)
|
147 |
|
148 |
file_storage_dir = os.path.normpath(file_storage_dir) |
149 |
|
150 |
if not os.path.isabs(file_storage_dir): |
151 |
raise errors.OpPrereqError("The file storage directory you passed is" |
152 |
" not an absolute path.")
|
153 |
|
154 |
if not os.path.exists(file_storage_dir): |
155 |
try:
|
156 |
os.makedirs(file_storage_dir, 0750)
|
157 |
except OSError, err: |
158 |
raise errors.OpPrereqError("Cannot create file storage directory" |
159 |
" '%s': %s" %
|
160 |
(file_storage_dir, err)) |
161 |
|
162 |
if not os.path.isdir(file_storage_dir): |
163 |
raise errors.OpPrereqError("The file storage directory '%s' is not" |
164 |
" a directory." % file_storage_dir)
|
165 |
|
166 |
if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): |
167 |
raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix) |
168 |
|
169 |
result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) |
170 |
if result.failed:
|
171 |
raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % |
172 |
(master_netdev, |
173 |
result.output.strip())) |
174 |
|
175 |
if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and |
176 |
os.access(constants.NODE_INITD_SCRIPT, os.X_OK)): |
177 |
raise errors.OpPrereqError("Init.d script '%s' missing or not" |
178 |
" executable." % constants.NODE_INITD_SCRIPT)
|
179 |
|
180 |
utils.CheckBEParams(beparams) |
181 |
|
182 |
# set up the inter-node password and certificate
|
183 |
_InitGanetiServerSetup() |
184 |
|
185 |
# set up ssh config and /etc/hosts
|
186 |
f = open(constants.SSH_HOST_RSA_PUB, 'r') |
187 |
try:
|
188 |
sshline = f.read() |
189 |
finally:
|
190 |
f.close() |
191 |
sshkey = sshline.split(" ")[1] |
192 |
|
193 |
utils.AddHostToEtcHosts(hostname.name) |
194 |
_InitSSHSetup() |
195 |
|
196 |
# init of cluster config file
|
197 |
cluster_config = objects.Cluster( |
198 |
serial_no=1,
|
199 |
rsahostkeypub=sshkey, |
200 |
highest_used_port=(constants.FIRST_DRBD_PORT - 1),
|
201 |
mac_prefix=mac_prefix, |
202 |
volume_group_name=vg_name, |
203 |
default_bridge=def_bridge, |
204 |
tcpudp_port_pool=set(),
|
205 |
master_node=hostname.name, |
206 |
master_ip=clustername.ip, |
207 |
master_netdev=master_netdev, |
208 |
cluster_name=clustername.name, |
209 |
file_storage_dir=file_storage_dir, |
210 |
enabled_hypervisors=enabled_hypervisors, |
211 |
default_hypervisor=default_hypervisor, |
212 |
beparams={constants.BEGR_DEFAULT: beparams}, |
213 |
hvparams=hvparams, |
214 |
candidate_pool_size=candidate_pool_size, |
215 |
) |
216 |
master_node_config = objects.Node(name=hostname.name, |
217 |
primary_ip=hostname.ip, |
218 |
secondary_ip=secondary_ip, |
219 |
serial_no=1,
|
220 |
master_candidate=True,
|
221 |
offline=False,
|
222 |
) |
223 |
|
224 |
sscfg = InitConfig(constants.CONFIG_VERSION, |
225 |
cluster_config, master_node_config) |
226 |
ssh.WriteKnownHostsFile(sscfg, constants.SSH_KNOWN_HOSTS_FILE) |
227 |
cfg = config.ConfigWriter() |
228 |
cfg.Update(cfg.GetClusterInfo()) |
229 |
|
230 |
# start the master ip
|
231 |
# TODO: Review rpc call from bootstrap
|
232 |
rpc.RpcRunner.call_node_start_master(hostname.name, True)
|
233 |
|
234 |
|
235 |
def InitConfig(version, cluster_config, master_node_config, |
236 |
cfg_file=constants.CLUSTER_CONF_FILE): |
237 |
"""Create the initial cluster configuration.
|
238 |
|
239 |
It will contain the current node, which will also be the master
|
240 |
node, and no instances.
|
241 |
|
242 |
@type version: int
|
243 |
@param version: configuration version
|
244 |
@type cluster_config: L{objects.Cluster}
|
245 |
@param cluster_config: cluster configuration
|
246 |
@type master_node_config: L{objects.Node}
|
247 |
@param master_node_config: master node configuration
|
248 |
@type cfg_file: string
|
249 |
@param cfg_file: configuration file path
|
250 |
|
251 |
@rtype: L{ssconf.SimpleConfigWriter}
|
252 |
@returns: initialized config instance
|
253 |
|
254 |
"""
|
255 |
nodes = { |
256 |
master_node_config.name: master_node_config, |
257 |
} |
258 |
|
259 |
config_data = objects.ConfigData(version=version, |
260 |
cluster=cluster_config, |
261 |
nodes=nodes, |
262 |
instances={}, |
263 |
serial_no=1)
|
264 |
cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file) |
265 |
cfg.Save() |
266 |
|
267 |
return cfg
|
268 |
|
269 |
|
270 |
def FinalizeClusterDestroy(master): |
271 |
"""Execute the last steps of cluster destroy
|
272 |
|
273 |
This function shuts down all the daemons, completing the destroy
|
274 |
begun in cmdlib.LUDestroyOpcode.
|
275 |
|
276 |
"""
|
277 |
result = rpc.RpcRunner.call_node_stop_master(master, True)
|
278 |
if result.failed or not result.data: |
279 |
logging.warning("Could not disable the master role")
|
280 |
result = rpc.RpcRunner.call_node_leave_cluster(master) |
281 |
if result.failed or not result.data: |
282 |
logging.warning("Could not shutdown the node daemon and cleanup the node")
|
283 |
|
284 |
|
285 |
def SetupNodeDaemon(cluster_name, node, ssh_key_check): |
286 |
"""Add a node to the cluster.
|
287 |
|
288 |
This function must be called before the actual opcode, and will ssh
|
289 |
to the remote node, copy the needed files, and start ganeti-noded,
|
290 |
allowing the master to do the rest via normal rpc calls.
|
291 |
|
292 |
@param cluster_name: the cluster name
|
293 |
@param node: the name of the new node
|
294 |
@param ssh_key_check: whether to do a strict key check
|
295 |
|
296 |
"""
|
297 |
sshrunner = ssh.SshRunner(cluster_name) |
298 |
gntpem = utils.ReadFile(constants.SSL_CERT_FILE) |
299 |
# in the base64 pem encoding, neither '!' nor '.' are valid chars,
|
300 |
# so we use this to detect an invalid certificate; as long as the
|
301 |
# cert doesn't contain this, the here-document will be correctly
|
302 |
# parsed by the shell sequence below
|
303 |
if re.search('^!EOF\.', gntpem, re.MULTILINE): |
304 |
raise errors.OpExecError("invalid PEM encoding in the SSL certificate") |
305 |
if not gntpem.endswith("\n"): |
306 |
raise errors.OpExecError("PEM must end with newline") |
307 |
|
308 |
# set up inter-node password and certificate and restarts the node daemon
|
309 |
# and then connect with ssh to set password and start ganeti-noded
|
310 |
# note that all the below variables are sanitized at this point,
|
311 |
# either by being constants or by the checks above
|
312 |
mycommand = ("umask 077 && "
|
313 |
"cat > '%s' << '!EOF.' && \n"
|
314 |
"%s!EOF.\n%s restart" %
|
315 |
(constants.SSL_CERT_FILE, gntpem, |
316 |
constants.NODE_INITD_SCRIPT)) |
317 |
|
318 |
result = sshrunner.Run(node, 'root', mycommand, batch=False, |
319 |
ask_key=ssh_key_check, |
320 |
use_cluster_key=False,
|
321 |
strict_host_check=ssh_key_check) |
322 |
if result.failed:
|
323 |
raise errors.OpExecError("Remote command on node %s, error: %s," |
324 |
" output: %s" %
|
325 |
(node, result.fail_reason, result.output)) |
326 |
|
327 |
|
328 |
def MasterFailover(): |
329 |
"""Failover the master node.
|
330 |
|
331 |
This checks that we are not already the master, and will cause the
|
332 |
current master to cease being master, and the non-master to become
|
333 |
new master.
|
334 |
|
335 |
"""
|
336 |
sstore = ssconf.SimpleStore() |
337 |
|
338 |
old_master, new_master = ssconf.GetMasterAndMyself(sstore) |
339 |
node_list = sstore.GetNodeList() |
340 |
mc_list = sstore.GetMasterCandidates() |
341 |
|
342 |
if old_master == new_master:
|
343 |
raise errors.OpPrereqError("This commands must be run on the node" |
344 |
" where you want the new master to be."
|
345 |
" %s is already the master" %
|
346 |
old_master) |
347 |
|
348 |
if new_master not in mc_list: |
349 |
mc_no_master = [name for name in mc_list if name != old_master] |
350 |
raise errors.OpPrereqError("This node is not among the nodes marked" |
351 |
" as master candidates. Only these nodes"
|
352 |
" can become masters. Current list of"
|
353 |
" master candidates is:\n"
|
354 |
"%s" % ('\n'.join(mc_no_master))) |
355 |
|
356 |
vote_list = GatherMasterVotes(node_list) |
357 |
|
358 |
if vote_list:
|
359 |
voted_master = vote_list[0][0] |
360 |
if voted_master is None: |
361 |
raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not" |
362 |
" respond.")
|
363 |
elif voted_master != old_master:
|
364 |
raise errors.OpPrereqError("I have wrong configuration, I believe the" |
365 |
" master is %s but the other nodes voted for"
|
366 |
" %s. Please resync the configuration of"
|
367 |
" this node." % (old_master, voted_master))
|
368 |
# end checks
|
369 |
|
370 |
rcode = 0
|
371 |
|
372 |
logging.info("Setting master to %s, old master: %s", new_master, old_master)
|
373 |
|
374 |
result = rpc.RpcRunner.call_node_stop_master(old_master, True)
|
375 |
if result.failed or not result.data: |
376 |
logging.error("Could not disable the master role on the old master"
|
377 |
" %s, please disable manually", old_master)
|
378 |
|
379 |
# Here we have a phase where no master should be running
|
380 |
|
381 |
# instantiate a real config writer, as we now know we have the
|
382 |
# configuration data
|
383 |
cfg = config.ConfigWriter() |
384 |
|
385 |
cluster_info = cfg.GetClusterInfo() |
386 |
cluster_info.master_node = new_master |
387 |
# this will also regenerate the ssconf files, since we updated the
|
388 |
# cluster info
|
389 |
cfg.Update(cluster_info) |
390 |
|
391 |
result = rpc.RpcRunner.call_node_start_master(new_master, True)
|
392 |
if result.failed or not result.data: |
393 |
logging.error("Could not start the master role on the new master"
|
394 |
" %s, please check", new_master)
|
395 |
rcode = 1
|
396 |
|
397 |
return rcode
|
398 |
|
399 |
|
400 |
def GatherMasterVotes(node_list): |
401 |
"""Check the agreement on who is the master.
|
402 |
|
403 |
This function will return a list of (node, number of votes), ordered
|
404 |
by the number of votes. Errors will be denoted by the key 'None'.
|
405 |
|
406 |
Note that the sum of votes is the number of nodes this machine
|
407 |
knows, whereas the number of entries in the list could be different
|
408 |
(if some nodes vote for another master).
|
409 |
|
410 |
We remove ourselves from the list since we know that (bugs aside)
|
411 |
since we use the same source for configuration information for both
|
412 |
backend and boostrap, we'll always vote for ourselves.
|
413 |
|
414 |
@type node_list: list
|
415 |
@param node_list: the list of nodes to query for master info; the current
|
416 |
node wil be removed if it is in the list
|
417 |
@rtype: list
|
418 |
@return: list of (node, votes)
|
419 |
|
420 |
"""
|
421 |
myself = utils.HostInfo().name |
422 |
try:
|
423 |
node_list.remove(myself) |
424 |
except ValueError: |
425 |
pass
|
426 |
if not node_list: |
427 |
# no nodes left (eventually after removing myself)
|
428 |
return []
|
429 |
results = rpc.RpcRunner.call_master_info(node_list) |
430 |
if not isinstance(results, dict): |
431 |
# this should not happen (unless internal error in rpc)
|
432 |
logging.critical("Can't complete rpc call, aborting master startup")
|
433 |
return [(None, len(node_list))] |
434 |
votes = {} |
435 |
for node in results: |
436 |
nres = results[node] |
437 |
data = nres.data |
438 |
if nres.failed or not isinstance(data, (tuple, list)) or len(data) < 3: |
439 |
# here the rpc layer should have already logged errors
|
440 |
if None not in votes: |
441 |
votes[None] = 0 |
442 |
votes[None] += 1 |
443 |
continue
|
444 |
master_node = data[2]
|
445 |
if master_node not in votes: |
446 |
votes[master_node] = 0
|
447 |
votes[master_node] += 1
|
448 |
|
449 |
vote_list = [v for v in votes.items()] |
450 |
# sort first on number of votes then on name, since we want None
|
451 |
# sorted later if we have the half of the nodes not responding, and
|
452 |
# half voting all for the same master
|
453 |
vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True) |
454 |
|
455 |
return vote_list
|