root / lib / bootstrap.py @ 80a56f26
History | View | Annotate | Download (42.6 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 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 re |
29 |
import logging |
30 |
import time |
31 |
import tempfile |
32 |
|
33 |
from ganeti.cmdlib import cluster |
34 |
import ganeti.rpc.node as rpc |
35 |
from ganeti import ssh |
36 |
from ganeti import utils |
37 |
from ganeti import errors |
38 |
from ganeti import config |
39 |
from ganeti import constants |
40 |
from ganeti import objects |
41 |
from ganeti import ssconf |
42 |
from ganeti import serializer |
43 |
from ganeti import hypervisor |
44 |
from ganeti.storage import drbd |
45 |
from ganeti.storage import filestorage |
46 |
from ganeti import netutils |
47 |
from ganeti import luxi |
48 |
from ganeti import jstore |
49 |
from ganeti import pathutils |
50 |
from ganeti import runtime |
51 |
|
52 |
|
53 |
# ec_id for InitConfig's temporary reservation manager
|
54 |
_INITCONF_ECID = "initconfig-ecid"
|
55 |
|
56 |
#: After how many seconds daemon must be responsive
|
57 |
_DAEMON_READY_TIMEOUT = 10.0
|
58 |
|
59 |
|
60 |
def _InitSSHSetup(): |
61 |
"""Setup the SSH configuration for the cluster.
|
62 |
|
63 |
This generates a dsa keypair for root, adds the pub key to the
|
64 |
permitted hosts and adds the hostkey to its own known hosts.
|
65 |
|
66 |
"""
|
67 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER) |
68 |
|
69 |
for name in priv_key, pub_key: |
70 |
if os.path.exists(name):
|
71 |
utils.CreateBackup(name) |
72 |
utils.RemoveFile(name) |
73 |
|
74 |
result = utils.RunCmd(["ssh-keygen", "-t", "dsa", |
75 |
"-f", priv_key,
|
76 |
"-q", "-N", ""]) |
77 |
if result.failed:
|
78 |
raise errors.OpExecError("Could not generate ssh keypair, error %s" % |
79 |
result.output) |
80 |
|
81 |
utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) |
82 |
|
83 |
|
84 |
def GenerateHmacKey(file_name): |
85 |
"""Writes a new HMAC key.
|
86 |
|
87 |
@type file_name: str
|
88 |
@param file_name: Path to output file
|
89 |
|
90 |
"""
|
91 |
utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400, |
92 |
backup=True)
|
93 |
|
94 |
|
95 |
# pylint: disable=R0913
|
96 |
def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert, |
97 |
new_confd_hmac_key, new_cds, |
98 |
rapi_cert_pem=None, spice_cert_pem=None, |
99 |
spice_cacert_pem=None, cds=None, |
100 |
nodecert_file=pathutils.NODED_CERT_FILE, |
101 |
rapicert_file=pathutils.RAPI_CERT_FILE, |
102 |
spicecert_file=pathutils.SPICE_CERT_FILE, |
103 |
spicecacert_file=pathutils.SPICE_CACERT_FILE, |
104 |
hmackey_file=pathutils.CONFD_HMAC_KEY, |
105 |
cds_file=pathutils.CLUSTER_DOMAIN_SECRET_FILE): |
106 |
"""Updates the cluster certificates, keys and secrets.
|
107 |
|
108 |
@type new_cluster_cert: bool
|
109 |
@param new_cluster_cert: Whether to generate a new cluster certificate
|
110 |
@type new_rapi_cert: bool
|
111 |
@param new_rapi_cert: Whether to generate a new RAPI certificate
|
112 |
@type new_spice_cert: bool
|
113 |
@param new_spice_cert: Whether to generate a new SPICE certificate
|
114 |
@type new_confd_hmac_key: bool
|
115 |
@param new_confd_hmac_key: Whether to generate a new HMAC key
|
116 |
@type new_cds: bool
|
117 |
@param new_cds: Whether to generate a new cluster domain secret
|
118 |
@type rapi_cert_pem: string
|
119 |
@param rapi_cert_pem: New RAPI certificate in PEM format
|
120 |
@type spice_cert_pem: string
|
121 |
@param spice_cert_pem: New SPICE certificate in PEM format
|
122 |
@type spice_cacert_pem: string
|
123 |
@param spice_cacert_pem: Certificate of the CA that signed the SPICE
|
124 |
certificate, in PEM format
|
125 |
@type cds: string
|
126 |
@param cds: New cluster domain secret
|
127 |
@type nodecert_file: string
|
128 |
@param nodecert_file: optional override of the node cert file path
|
129 |
@type rapicert_file: string
|
130 |
@param rapicert_file: optional override of the rapi cert file path
|
131 |
@type spicecert_file: string
|
132 |
@param spicecert_file: optional override of the spice cert file path
|
133 |
@type spicecacert_file: string
|
134 |
@param spicecacert_file: optional override of the spice CA cert file path
|
135 |
@type hmackey_file: string
|
136 |
@param hmackey_file: optional override of the hmac key file path
|
137 |
|
138 |
"""
|
139 |
# pylint: disable=R0913
|
140 |
# noded SSL certificate
|
141 |
utils.GenerateNewSslCert( |
142 |
new_cluster_cert, nodecert_file, 1,
|
143 |
"Generating new cluster certificate at %s" % nodecert_file)
|
144 |
|
145 |
# confd HMAC key
|
146 |
if new_confd_hmac_key or not os.path.exists(hmackey_file): |
147 |
logging.debug("Writing new confd HMAC key to %s", hmackey_file)
|
148 |
GenerateHmacKey(hmackey_file) |
149 |
|
150 |
if rapi_cert_pem:
|
151 |
# Assume rapi_pem contains a valid PEM-formatted certificate and key
|
152 |
logging.debug("Writing RAPI certificate at %s", rapicert_file)
|
153 |
utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
|
154 |
|
155 |
else:
|
156 |
utils.GenerateNewSslCert( |
157 |
new_rapi_cert, rapicert_file, 1,
|
158 |
"Generating new RAPI certificate at %s" % rapicert_file)
|
159 |
|
160 |
# SPICE
|
161 |
spice_cert_exists = os.path.exists(spicecert_file) |
162 |
spice_cacert_exists = os.path.exists(spicecacert_file) |
163 |
if spice_cert_pem:
|
164 |
# spice_cert_pem implies also spice_cacert_pem
|
165 |
logging.debug("Writing SPICE certificate at %s", spicecert_file)
|
166 |
utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True)
|
167 |
logging.debug("Writing SPICE CA certificate at %s", spicecacert_file)
|
168 |
utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True)
|
169 |
elif new_spice_cert or not spice_cert_exists: |
170 |
if spice_cert_exists:
|
171 |
utils.CreateBackup(spicecert_file) |
172 |
if spice_cacert_exists:
|
173 |
utils.CreateBackup(spicecacert_file) |
174 |
|
175 |
logging.debug("Generating new self-signed SPICE certificate at %s",
|
176 |
spicecert_file) |
177 |
(_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file, 1)
|
178 |
|
179 |
# Self-signed certificate -> the public certificate is also the CA public
|
180 |
# certificate
|
181 |
logging.debug("Writing the public certificate to %s",
|
182 |
spicecert_file) |
183 |
utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem)
|
184 |
|
185 |
# Cluster domain secret
|
186 |
if cds:
|
187 |
logging.debug("Writing cluster domain secret to %s", cds_file)
|
188 |
utils.WriteFile(cds_file, data=cds, backup=True)
|
189 |
|
190 |
elif new_cds or not os.path.exists(cds_file): |
191 |
logging.debug("Generating new cluster domain secret at %s", cds_file)
|
192 |
GenerateHmacKey(cds_file) |
193 |
|
194 |
|
195 |
def _InitGanetiServerSetup(master_name): |
196 |
"""Setup the necessary configuration for the initial node daemon.
|
197 |
|
198 |
This creates the nodepass file containing the shared password for
|
199 |
the cluster, generates the SSL certificate and starts the node daemon.
|
200 |
|
201 |
@type master_name: str
|
202 |
@param master_name: Name of the master node
|
203 |
|
204 |
"""
|
205 |
# Generate cluster secrets
|
206 |
GenerateClusterCrypto(True, False, False, False, False) |
207 |
|
208 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "start", constants.NODED])
|
209 |
if result.failed:
|
210 |
raise errors.OpExecError("Could not start the node daemon, command %s" |
211 |
" had exitcode %s and error %s" %
|
212 |
(result.cmd, result.exit_code, result.output)) |
213 |
|
214 |
_WaitForNodeDaemon(master_name) |
215 |
|
216 |
|
217 |
def _WaitForNodeDaemon(node_name): |
218 |
"""Wait for node daemon to become responsive.
|
219 |
|
220 |
"""
|
221 |
def _CheckNodeDaemon(): |
222 |
# Pylint bug <http://www.logilab.org/ticket/35642>
|
223 |
# pylint: disable=E1101
|
224 |
result = rpc.BootstrapRunner().call_version([node_name])[node_name] |
225 |
if result.fail_msg:
|
226 |
raise utils.RetryAgain()
|
227 |
|
228 |
try:
|
229 |
utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT)
|
230 |
except utils.RetryTimeout:
|
231 |
raise errors.OpExecError("Node daemon on %s didn't answer queries within" |
232 |
" %s seconds" % (node_name, _DAEMON_READY_TIMEOUT))
|
233 |
|
234 |
|
235 |
def _WaitForMasterDaemon(): |
236 |
"""Wait for master daemon to become responsive.
|
237 |
|
238 |
"""
|
239 |
def _CheckMasterDaemon(): |
240 |
try:
|
241 |
cl = luxi.Client() |
242 |
(cluster_name, ) = cl.QueryConfigValues(["cluster_name"])
|
243 |
except Exception: |
244 |
raise utils.RetryAgain()
|
245 |
|
246 |
logging.debug("Received cluster name %s from master", cluster_name)
|
247 |
|
248 |
try:
|
249 |
utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT)
|
250 |
except utils.RetryTimeout:
|
251 |
raise errors.OpExecError("Master daemon didn't answer queries within" |
252 |
" %s seconds" % _DAEMON_READY_TIMEOUT)
|
253 |
|
254 |
|
255 |
def _WaitForSshDaemon(hostname, port, family): |
256 |
"""Wait for SSH daemon to become responsive.
|
257 |
|
258 |
"""
|
259 |
hostip = netutils.GetHostname(name=hostname, family=family).ip |
260 |
|
261 |
def _CheckSshDaemon(): |
262 |
if netutils.TcpPing(hostip, port, timeout=1.0, live_port_needed=True): |
263 |
logging.debug("SSH daemon on %s:%s (IP address %s) has become"
|
264 |
" responsive", hostname, port, hostip)
|
265 |
else:
|
266 |
raise utils.RetryAgain()
|
267 |
|
268 |
try:
|
269 |
utils.Retry(_CheckSshDaemon, 1.0, _DAEMON_READY_TIMEOUT)
|
270 |
except utils.RetryTimeout:
|
271 |
raise errors.OpExecError("SSH daemon on %s:%s (IP address %s) didn't" |
272 |
" become responsive within %s seconds" %
|
273 |
(hostname, port, hostip, _DAEMON_READY_TIMEOUT)) |
274 |
|
275 |
|
276 |
def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose, |
277 |
use_cluster_key, ask_key, strict_host_check, |
278 |
port, data): |
279 |
"""Runs a command to configure something on a remote machine.
|
280 |
|
281 |
@type cluster_name: string
|
282 |
@param cluster_name: Cluster name
|
283 |
@type node: string
|
284 |
@param node: Node name
|
285 |
@type basecmd: string
|
286 |
@param basecmd: Base command (path on the remote machine)
|
287 |
@type debug: bool
|
288 |
@param debug: Enable debug output
|
289 |
@type verbose: bool
|
290 |
@param verbose: Enable verbose output
|
291 |
@type use_cluster_key: bool
|
292 |
@param use_cluster_key: See L{ssh.SshRunner.BuildCmd}
|
293 |
@type ask_key: bool
|
294 |
@param ask_key: See L{ssh.SshRunner.BuildCmd}
|
295 |
@type strict_host_check: bool
|
296 |
@param strict_host_check: See L{ssh.SshRunner.BuildCmd}
|
297 |
@type port: int
|
298 |
@param port: The SSH port of the remote machine or None for the default
|
299 |
@param data: JSON-serializable input data for script (passed to stdin)
|
300 |
|
301 |
"""
|
302 |
cmd = [basecmd] |
303 |
|
304 |
# Pass --debug/--verbose to the external script if set on our invocation
|
305 |
if debug:
|
306 |
cmd.append("--debug")
|
307 |
|
308 |
if verbose:
|
309 |
cmd.append("--verbose")
|
310 |
|
311 |
if port is None: |
312 |
port = netutils.GetDaemonPort(constants.SSH) |
313 |
|
314 |
family = ssconf.SimpleStore().GetPrimaryIPFamily() |
315 |
srun = ssh.SshRunner(cluster_name, |
316 |
ipv6=(family == netutils.IP6Address.family)) |
317 |
scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER, |
318 |
utils.ShellQuoteArgs(cmd), |
319 |
batch=False, ask_key=ask_key, quiet=False, |
320 |
strict_host_check=strict_host_check, |
321 |
use_cluster_key=use_cluster_key, |
322 |
port=port) |
323 |
|
324 |
tempfh = tempfile.TemporaryFile() |
325 |
try:
|
326 |
tempfh.write(serializer.DumpJson(data)) |
327 |
tempfh.seek(0)
|
328 |
|
329 |
result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh)
|
330 |
finally:
|
331 |
tempfh.close() |
332 |
|
333 |
if result.failed:
|
334 |
raise errors.OpExecError("Command '%s' failed: %s" % |
335 |
(result.cmd, result.fail_reason)) |
336 |
|
337 |
_WaitForSshDaemon(node, port, family) |
338 |
|
339 |
|
340 |
def _InitFileStorageDir(file_storage_dir): |
341 |
"""Initialize if needed the file storage.
|
342 |
|
343 |
@param file_storage_dir: the user-supplied value
|
344 |
@return: either empty string (if file storage was disabled at build
|
345 |
time) or the normalized path to the storage directory
|
346 |
|
347 |
"""
|
348 |
file_storage_dir = os.path.normpath(file_storage_dir) |
349 |
|
350 |
if not os.path.isabs(file_storage_dir): |
351 |
raise errors.OpPrereqError("File storage directory '%s' is not an absolute" |
352 |
" path" % file_storage_dir, errors.ECODE_INVAL)
|
353 |
|
354 |
if not os.path.exists(file_storage_dir): |
355 |
try:
|
356 |
os.makedirs(file_storage_dir, 0750)
|
357 |
except OSError, err: |
358 |
raise errors.OpPrereqError("Cannot create file storage directory" |
359 |
" '%s': %s" % (file_storage_dir, err),
|
360 |
errors.ECODE_ENVIRON) |
361 |
|
362 |
if not os.path.isdir(file_storage_dir): |
363 |
raise errors.OpPrereqError("The file storage directory '%s' is not" |
364 |
" a directory." % file_storage_dir,
|
365 |
errors.ECODE_ENVIRON) |
366 |
|
367 |
return file_storage_dir
|
368 |
|
369 |
|
370 |
def _PrepareFileBasedStorage( |
371 |
enabled_disk_templates, file_storage_dir, |
372 |
default_dir, file_disk_template, |
373 |
init_fn=_InitFileStorageDir, acceptance_fn=None):
|
374 |
"""Checks if a file-base storage type is enabled and inits the dir.
|
375 |
|
376 |
@type enabled_disk_templates: list of string
|
377 |
@param enabled_disk_templates: list of enabled disk templates
|
378 |
@type file_storage_dir: string
|
379 |
@param file_storage_dir: the file storage directory
|
380 |
@type default_dir: string
|
381 |
@param default_dir: default file storage directory when C{file_storage_dir}
|
382 |
is 'None'
|
383 |
@type file_disk_template: string
|
384 |
@param file_disk_template: a disk template whose storage type is 'ST_FILE' or
|
385 |
'ST_SHARED_FILE'
|
386 |
@rtype: string
|
387 |
@returns: the name of the actual file storage directory
|
388 |
|
389 |
"""
|
390 |
assert (file_disk_template in utils.storage.GetDiskTemplatesOfStorageTypes( |
391 |
constants.ST_FILE, constants.ST_SHARED_FILE |
392 |
)) |
393 |
|
394 |
if file_storage_dir is None: |
395 |
file_storage_dir = default_dir |
396 |
if not acceptance_fn: |
397 |
acceptance_fn = \ |
398 |
lambda path: filestorage.CheckFileStoragePathAcceptance(
|
399 |
path, exact_match_ok=True)
|
400 |
|
401 |
cluster.CheckFileStoragePathVsEnabledDiskTemplates( |
402 |
logging.warning, file_storage_dir, enabled_disk_templates) |
403 |
|
404 |
file_storage_enabled = file_disk_template in enabled_disk_templates
|
405 |
if file_storage_enabled:
|
406 |
try:
|
407 |
acceptance_fn(file_storage_dir) |
408 |
except errors.FileStoragePathError as e: |
409 |
raise errors.OpPrereqError(str(e)) |
410 |
result_file_storage_dir = init_fn(file_storage_dir) |
411 |
else:
|
412 |
result_file_storage_dir = file_storage_dir |
413 |
return result_file_storage_dir
|
414 |
|
415 |
|
416 |
def _PrepareFileStorage( |
417 |
enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir, |
418 |
acceptance_fn=None):
|
419 |
"""Checks if file storage is enabled and inits the dir.
|
420 |
|
421 |
@see: C{_PrepareFileBasedStorage}
|
422 |
|
423 |
"""
|
424 |
return _PrepareFileBasedStorage(
|
425 |
enabled_disk_templates, file_storage_dir, |
426 |
pathutils.DEFAULT_FILE_STORAGE_DIR, constants.DT_FILE, |
427 |
init_fn=init_fn, acceptance_fn=acceptance_fn) |
428 |
|
429 |
|
430 |
def _PrepareSharedFileStorage( |
431 |
enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir, |
432 |
acceptance_fn=None):
|
433 |
"""Checks if shared file storage is enabled and inits the dir.
|
434 |
|
435 |
@see: C{_PrepareFileBasedStorage}
|
436 |
|
437 |
"""
|
438 |
return _PrepareFileBasedStorage(
|
439 |
enabled_disk_templates, file_storage_dir, |
440 |
pathutils.DEFAULT_SHARED_FILE_STORAGE_DIR, constants.DT_SHARED_FILE, |
441 |
init_fn=init_fn, acceptance_fn=acceptance_fn) |
442 |
|
443 |
|
444 |
def _PrepareGlusterStorage( |
445 |
enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir, |
446 |
acceptance_fn=None):
|
447 |
"""Checks if gluster storage is enabled and inits the dir.
|
448 |
|
449 |
@see: C{_PrepareFileBasedStorage}
|
450 |
|
451 |
"""
|
452 |
return _PrepareFileBasedStorage(
|
453 |
enabled_disk_templates, file_storage_dir, |
454 |
pathutils.DEFAULT_GLUSTER_STORAGE_DIR, constants.DT_GLUSTER, |
455 |
init_fn=init_fn, acceptance_fn=acceptance_fn) |
456 |
|
457 |
|
458 |
def _InitCheckEnabledDiskTemplates(enabled_disk_templates): |
459 |
"""Checks the sanity of the enabled disk templates.
|
460 |
|
461 |
"""
|
462 |
if not enabled_disk_templates: |
463 |
raise errors.OpPrereqError("Enabled disk templates list must contain at" |
464 |
" least one member", errors.ECODE_INVAL)
|
465 |
invalid_disk_templates = \ |
466 |
set(enabled_disk_templates) - constants.DISK_TEMPLATES
|
467 |
if invalid_disk_templates:
|
468 |
raise errors.OpPrereqError("Enabled disk templates list contains invalid" |
469 |
" entries: %s" % invalid_disk_templates,
|
470 |
errors.ECODE_INVAL) |
471 |
|
472 |
|
473 |
def _RestrictIpolicyToEnabledDiskTemplates(ipolicy, enabled_disk_templates): |
474 |
"""Restricts the ipolicy's disk templates to the enabled ones.
|
475 |
|
476 |
This function clears the ipolicy's list of allowed disk templates from the
|
477 |
ones that are not enabled by the cluster.
|
478 |
|
479 |
@type ipolicy: dict
|
480 |
@param ipolicy: the instance policy
|
481 |
@type enabled_disk_templates: list of string
|
482 |
@param enabled_disk_templates: the list of cluster-wide enabled disk
|
483 |
templates
|
484 |
|
485 |
"""
|
486 |
assert constants.IPOLICY_DTS in ipolicy |
487 |
allowed_disk_templates = ipolicy[constants.IPOLICY_DTS] |
488 |
restricted_disk_templates = list(set(allowed_disk_templates) |
489 |
.intersection(set(enabled_disk_templates)))
|
490 |
ipolicy[constants.IPOLICY_DTS] = restricted_disk_templates |
491 |
|
492 |
|
493 |
def _InitCheckDrbdHelper(drbd_helper, drbd_enabled): |
494 |
"""Checks the DRBD usermode helper.
|
495 |
|
496 |
@type drbd_helper: string
|
497 |
@param drbd_helper: name of the DRBD usermode helper that the system should
|
498 |
use
|
499 |
|
500 |
"""
|
501 |
if not drbd_enabled: |
502 |
return
|
503 |
|
504 |
if drbd_helper is not None: |
505 |
try:
|
506 |
curr_helper = drbd.DRBD8.GetUsermodeHelper() |
507 |
except errors.BlockDeviceError, err:
|
508 |
raise errors.OpPrereqError("Error while checking drbd helper" |
509 |
" (disable drbd with --enabled-disk-templates"
|
510 |
" if you are not using drbd): %s" % str(err), |
511 |
errors.ECODE_ENVIRON) |
512 |
if drbd_helper != curr_helper:
|
513 |
raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s" |
514 |
" is the current helper" % (drbd_helper,
|
515 |
curr_helper), |
516 |
errors.ECODE_INVAL) |
517 |
|
518 |
|
519 |
def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914 |
520 |
master_netmask, master_netdev, file_storage_dir, |
521 |
shared_file_storage_dir, gluster_storage_dir, |
522 |
candidate_pool_size, secondary_ip=None,
|
523 |
vg_name=None, beparams=None, nicparams=None, ndparams=None, |
524 |
hvparams=None, diskparams=None, enabled_hypervisors=None, |
525 |
modify_etc_hosts=True, modify_ssh_setup=True, |
526 |
maintain_node_health=False, drbd_helper=None, uid_pool=None, |
527 |
default_iallocator=None, default_iallocator_params=None, |
528 |
primary_ip_version=None, ipolicy=None, |
529 |
prealloc_wipe_disks=False, use_external_mip_script=False, |
530 |
hv_state=None, disk_state=None, enabled_disk_templates=None): |
531 |
"""Initialise the cluster.
|
532 |
|
533 |
@type candidate_pool_size: int
|
534 |
@param candidate_pool_size: master candidate pool size
|
535 |
@type enabled_disk_templates: list of string
|
536 |
@param enabled_disk_templates: list of disk_templates to be used in this
|
537 |
cluster
|
538 |
|
539 |
"""
|
540 |
# TODO: complete the docstring
|
541 |
if config.ConfigWriter.IsCluster():
|
542 |
raise errors.OpPrereqError("Cluster is already initialised", |
543 |
errors.ECODE_STATE) |
544 |
|
545 |
if not enabled_hypervisors: |
546 |
raise errors.OpPrereqError("Enabled hypervisors list must contain at" |
547 |
" least one member", errors.ECODE_INVAL)
|
548 |
invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
|
549 |
if invalid_hvs:
|
550 |
raise errors.OpPrereqError("Enabled hypervisors contains invalid" |
551 |
" entries: %s" % invalid_hvs,
|
552 |
errors.ECODE_INVAL) |
553 |
|
554 |
_InitCheckEnabledDiskTemplates(enabled_disk_templates) |
555 |
|
556 |
try:
|
557 |
ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version) |
558 |
except errors.ProgrammerError:
|
559 |
raise errors.OpPrereqError("Invalid primary ip version: %d." % |
560 |
primary_ip_version, errors.ECODE_INVAL) |
561 |
|
562 |
hostname = netutils.GetHostname(family=ipcls.family) |
563 |
if not ipcls.IsValid(hostname.ip): |
564 |
raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d" |
565 |
" address." % (hostname.ip, primary_ip_version),
|
566 |
errors.ECODE_INVAL) |
567 |
|
568 |
if ipcls.IsLoopback(hostname.ip):
|
569 |
raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback" |
570 |
" address. Please fix DNS or %s." %
|
571 |
(hostname.ip, pathutils.ETC_HOSTS), |
572 |
errors.ECODE_ENVIRON) |
573 |
|
574 |
if not ipcls.Own(hostname.ip): |
575 |
raise errors.OpPrereqError("Inconsistency: this host's name resolves" |
576 |
" to %s,\nbut this ip address does not"
|
577 |
" belong to this host" %
|
578 |
hostname.ip, errors.ECODE_ENVIRON) |
579 |
|
580 |
clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family) |
581 |
|
582 |
if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5): |
583 |
raise errors.OpPrereqError("Cluster IP already active", |
584 |
errors.ECODE_NOTUNIQUE) |
585 |
|
586 |
if not secondary_ip: |
587 |
if primary_ip_version == constants.IP6_VERSION:
|
588 |
raise errors.OpPrereqError("When using a IPv6 primary address, a valid" |
589 |
" IPv4 address must be given as secondary",
|
590 |
errors.ECODE_INVAL) |
591 |
secondary_ip = hostname.ip |
592 |
|
593 |
if not netutils.IP4Address.IsValid(secondary_ip): |
594 |
raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid" |
595 |
" IPv4 address." % secondary_ip,
|
596 |
errors.ECODE_INVAL) |
597 |
|
598 |
if not netutils.IP4Address.Own(secondary_ip): |
599 |
raise errors.OpPrereqError("You gave %s as secondary IP," |
600 |
" but it does not belong to this host." %
|
601 |
secondary_ip, errors.ECODE_ENVIRON) |
602 |
|
603 |
if master_netmask is not None: |
604 |
if not ipcls.ValidateNetmask(master_netmask): |
605 |
raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " % |
606 |
(master_netmask, primary_ip_version), |
607 |
errors.ECODE_INVAL) |
608 |
else:
|
609 |
master_netmask = ipcls.iplen |
610 |
|
611 |
if vg_name:
|
612 |
# Check if volume group is valid
|
613 |
vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, |
614 |
constants.MIN_VG_SIZE) |
615 |
if vgstatus:
|
616 |
raise errors.OpPrereqError("Error: %s" % vgstatus, errors.ECODE_INVAL) |
617 |
|
618 |
drbd_enabled = constants.DT_DRBD8 in enabled_disk_templates
|
619 |
_InitCheckDrbdHelper(drbd_helper, drbd_enabled) |
620 |
|
621 |
logging.debug("Stopping daemons (if any are running)")
|
622 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"])
|
623 |
if result.failed:
|
624 |
raise errors.OpExecError("Could not stop daemons, command %s" |
625 |
" had exitcode %s and error '%s'" %
|
626 |
(result.cmd, result.exit_code, result.output)) |
627 |
|
628 |
file_storage_dir = _PrepareFileStorage(enabled_disk_templates, |
629 |
file_storage_dir) |
630 |
shared_file_storage_dir = _PrepareSharedFileStorage(enabled_disk_templates, |
631 |
shared_file_storage_dir) |
632 |
|
633 |
if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): |
634 |
raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix, |
635 |
errors.ECODE_INVAL) |
636 |
|
637 |
if not nicparams.get('mode', None) == constants.NIC_MODE_OVS: |
638 |
# Do not do this check if mode=openvswitch, since the openvswitch is not
|
639 |
# created yet
|
640 |
result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) |
641 |
if result.failed:
|
642 |
raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % |
643 |
(master_netdev, |
644 |
result.output.strip()), errors.ECODE_INVAL) |
645 |
|
646 |
dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)] |
647 |
utils.EnsureDirs(dirs) |
648 |
|
649 |
objects.UpgradeBeParams(beparams) |
650 |
utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES) |
651 |
utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES) |
652 |
|
653 |
objects.NIC.CheckParameterSyntax(nicparams) |
654 |
|
655 |
full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy) |
656 |
_RestrictIpolicyToEnabledDiskTemplates(full_ipolicy, enabled_disk_templates) |
657 |
|
658 |
if ndparams is not None: |
659 |
utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES) |
660 |
else:
|
661 |
ndparams = dict(constants.NDC_DEFAULTS)
|
662 |
|
663 |
# This is ugly, as we modify the dict itself
|
664 |
# FIXME: Make utils.ForceDictType pure functional or write a wrapper
|
665 |
# around it
|
666 |
if hv_state:
|
667 |
for hvname, hvs_data in hv_state.items(): |
668 |
utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES) |
669 |
hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data) |
670 |
else:
|
671 |
hv_state = dict((hvname, constants.HVST_DEFAULTS)
|
672 |
for hvname in enabled_hypervisors) |
673 |
|
674 |
# FIXME: disk_state has no default values yet
|
675 |
if disk_state:
|
676 |
for storage, ds_data in disk_state.items(): |
677 |
if storage not in constants.DS_VALID_TYPES: |
678 |
raise errors.OpPrereqError("Invalid storage type in disk state: %s" % |
679 |
storage, errors.ECODE_INVAL) |
680 |
for ds_name, state in ds_data.items(): |
681 |
utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES) |
682 |
ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state) |
683 |
|
684 |
# hvparams is a mapping of hypervisor->hvparams dict
|
685 |
for hv_name, hv_params in hvparams.iteritems(): |
686 |
utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) |
687 |
hv_class = hypervisor.GetHypervisor(hv_name) |
688 |
hv_class.CheckParameterSyntax(hv_params) |
689 |
|
690 |
# diskparams is a mapping of disk-template->diskparams dict
|
691 |
for template, dt_params in diskparams.items(): |
692 |
param_keys = set(dt_params.keys())
|
693 |
default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys())
|
694 |
if not (param_keys <= default_param_keys): |
695 |
unknown_params = param_keys - default_param_keys |
696 |
raise errors.OpPrereqError("Invalid parameters for disk template %s:" |
697 |
" %s" % (template,
|
698 |
utils.CommaJoin(unknown_params)), |
699 |
errors.ECODE_INVAL) |
700 |
utils.ForceDictType(dt_params, constants.DISK_DT_TYPES) |
701 |
if template == constants.DT_DRBD8 and vg_name is not None: |
702 |
# The default METAVG value is equal to the VG name set at init time,
|
703 |
# if provided
|
704 |
dt_params[constants.DRBD_DEFAULT_METAVG] = vg_name |
705 |
|
706 |
try:
|
707 |
utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS) |
708 |
except errors.OpPrereqError, err:
|
709 |
raise errors.OpPrereqError("While verify diskparam options: %s" % err, |
710 |
errors.ECODE_INVAL) |
711 |
|
712 |
# set up ssh config and /etc/hosts
|
713 |
rsa_sshkey = ""
|
714 |
dsa_sshkey = ""
|
715 |
if os.path.isfile(pathutils.SSH_HOST_RSA_PUB):
|
716 |
sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB) |
717 |
rsa_sshkey = sshline.split(" ")[1] |
718 |
if os.path.isfile(pathutils.SSH_HOST_DSA_PUB):
|
719 |
sshline = utils.ReadFile(pathutils.SSH_HOST_DSA_PUB) |
720 |
dsa_sshkey = sshline.split(" ")[1] |
721 |
if not rsa_sshkey and not dsa_sshkey: |
722 |
raise errors.OpPrereqError("Failed to find SSH public keys", |
723 |
errors.ECODE_ENVIRON) |
724 |
|
725 |
if modify_etc_hosts:
|
726 |
utils.AddHostToEtcHosts(hostname.name, hostname.ip) |
727 |
|
728 |
if modify_ssh_setup:
|
729 |
_InitSSHSetup() |
730 |
|
731 |
if default_iallocator is not None: |
732 |
alloc_script = utils.FindFile(default_iallocator, |
733 |
constants.IALLOCATOR_SEARCH_PATH, |
734 |
os.path.isfile) |
735 |
if alloc_script is None: |
736 |
raise errors.OpPrereqError("Invalid default iallocator script '%s'" |
737 |
" specified" % default_iallocator,
|
738 |
errors.ECODE_INVAL) |
739 |
else:
|
740 |
# default to htools
|
741 |
if utils.FindFile(constants.IALLOC_HAIL,
|
742 |
constants.IALLOCATOR_SEARCH_PATH, |
743 |
os.path.isfile): |
744 |
default_iallocator = constants.IALLOC_HAIL |
745 |
|
746 |
# check if we have all the users we need
|
747 |
try:
|
748 |
runtime.GetEnts() |
749 |
except errors.ConfigurationError, err:
|
750 |
raise errors.OpPrereqError("Required system user/group missing: %s" % |
751 |
err, errors.ECODE_ENVIRON) |
752 |
|
753 |
candidate_certs = {} |
754 |
|
755 |
now = time.time() |
756 |
|
757 |
# init of cluster config file
|
758 |
cluster_config = objects.Cluster( |
759 |
serial_no=1,
|
760 |
rsahostkeypub=rsa_sshkey, |
761 |
dsahostkeypub=dsa_sshkey, |
762 |
highest_used_port=(constants.FIRST_DRBD_PORT - 1),
|
763 |
mac_prefix=mac_prefix, |
764 |
volume_group_name=vg_name, |
765 |
tcpudp_port_pool=set(),
|
766 |
master_ip=clustername.ip, |
767 |
master_netmask=master_netmask, |
768 |
master_netdev=master_netdev, |
769 |
cluster_name=clustername.name, |
770 |
file_storage_dir=file_storage_dir, |
771 |
shared_file_storage_dir=shared_file_storage_dir, |
772 |
gluster_storage_dir=gluster_storage_dir, |
773 |
enabled_hypervisors=enabled_hypervisors, |
774 |
beparams={constants.PP_DEFAULT: beparams}, |
775 |
nicparams={constants.PP_DEFAULT: nicparams}, |
776 |
ndparams=ndparams, |
777 |
hvparams=hvparams, |
778 |
diskparams=diskparams, |
779 |
candidate_pool_size=candidate_pool_size, |
780 |
modify_etc_hosts=modify_etc_hosts, |
781 |
modify_ssh_setup=modify_ssh_setup, |
782 |
uid_pool=uid_pool, |
783 |
ctime=now, |
784 |
mtime=now, |
785 |
maintain_node_health=maintain_node_health, |
786 |
drbd_usermode_helper=drbd_helper, |
787 |
default_iallocator=default_iallocator, |
788 |
default_iallocator_params=default_iallocator_params, |
789 |
primary_ip_family=ipcls.family, |
790 |
prealloc_wipe_disks=prealloc_wipe_disks, |
791 |
use_external_mip_script=use_external_mip_script, |
792 |
ipolicy=full_ipolicy, |
793 |
hv_state_static=hv_state, |
794 |
disk_state_static=disk_state, |
795 |
enabled_disk_templates=enabled_disk_templates, |
796 |
candidate_certs=candidate_certs, |
797 |
osparams={}, |
798 |
osparams_private_cluster={} |
799 |
) |
800 |
master_node_config = objects.Node(name=hostname.name, |
801 |
primary_ip=hostname.ip, |
802 |
secondary_ip=secondary_ip, |
803 |
serial_no=1,
|
804 |
master_candidate=True,
|
805 |
offline=False, drained=False, |
806 |
ctime=now, mtime=now, |
807 |
) |
808 |
InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) |
809 |
cfg = config.ConfigWriter(offline=True)
|
810 |
ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE) |
811 |
cfg.Update(cfg.GetClusterInfo(), logging.error) |
812 |
ssconf.WriteSsconfFiles(cfg.GetSsconfValues()) |
813 |
|
814 |
# set up the inter-node password and certificate
|
815 |
_InitGanetiServerSetup(hostname.name) |
816 |
|
817 |
logging.debug("Starting daemons")
|
818 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"])
|
819 |
if result.failed:
|
820 |
raise errors.OpExecError("Could not start daemons, command %s" |
821 |
" had exitcode %s and error %s" %
|
822 |
(result.cmd, result.exit_code, result.output)) |
823 |
|
824 |
_WaitForMasterDaemon() |
825 |
|
826 |
|
827 |
def InitConfig(version, cluster_config, master_node_config, |
828 |
cfg_file=pathutils.CLUSTER_CONF_FILE): |
829 |
"""Create the initial cluster configuration.
|
830 |
|
831 |
It will contain the current node, which will also be the master
|
832 |
node, and no instances.
|
833 |
|
834 |
@type version: int
|
835 |
@param version: configuration version
|
836 |
@type cluster_config: L{objects.Cluster}
|
837 |
@param cluster_config: cluster configuration
|
838 |
@type master_node_config: L{objects.Node}
|
839 |
@param master_node_config: master node configuration
|
840 |
@type cfg_file: string
|
841 |
@param cfg_file: configuration file path
|
842 |
|
843 |
"""
|
844 |
uuid_generator = config.TemporaryReservationManager() |
845 |
cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID, |
846 |
_INITCONF_ECID) |
847 |
master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID, |
848 |
_INITCONF_ECID) |
849 |
cluster_config.master_node = master_node_config.uuid |
850 |
nodes = { |
851 |
master_node_config.uuid: master_node_config, |
852 |
} |
853 |
default_nodegroup = objects.NodeGroup( |
854 |
uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID), |
855 |
name=constants.INITIAL_NODE_GROUP_NAME, |
856 |
members=[master_node_config.uuid], |
857 |
diskparams={}, |
858 |
) |
859 |
nodegroups = { |
860 |
default_nodegroup.uuid: default_nodegroup, |
861 |
} |
862 |
now = time.time() |
863 |
config_data = objects.ConfigData(version=version, |
864 |
cluster=cluster_config, |
865 |
nodegroups=nodegroups, |
866 |
nodes=nodes, |
867 |
instances={}, |
868 |
networks={}, |
869 |
serial_no=1,
|
870 |
ctime=now, mtime=now) |
871 |
utils.WriteFile(cfg_file, |
872 |
data=serializer.Dump(config_data.ToDict()), |
873 |
mode=0600)
|
874 |
|
875 |
|
876 |
def FinalizeClusterDestroy(master_uuid): |
877 |
"""Execute the last steps of cluster destroy
|
878 |
|
879 |
This function shuts down all the daemons, completing the destroy
|
880 |
begun in cmdlib.LUDestroyOpcode.
|
881 |
|
882 |
"""
|
883 |
cfg = config.ConfigWriter() |
884 |
modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup |
885 |
runner = rpc.BootstrapRunner() |
886 |
|
887 |
master_name = cfg.GetNodeName(master_uuid) |
888 |
|
889 |
master_params = cfg.GetMasterNetworkParameters() |
890 |
master_params.uuid = master_uuid |
891 |
ems = cfg.GetUseExternalMipScript() |
892 |
result = runner.call_node_deactivate_master_ip(master_name, master_params, |
893 |
ems) |
894 |
|
895 |
msg = result.fail_msg |
896 |
if msg:
|
897 |
logging.warning("Could not disable the master IP: %s", msg)
|
898 |
|
899 |
result = runner.call_node_stop_master(master_name) |
900 |
msg = result.fail_msg |
901 |
if msg:
|
902 |
logging.warning("Could not disable the master role: %s", msg)
|
903 |
|
904 |
result = runner.call_node_leave_cluster(master_name, modify_ssh_setup) |
905 |
msg = result.fail_msg |
906 |
if msg:
|
907 |
logging.warning("Could not shutdown the node daemon and cleanup"
|
908 |
" the node: %s", msg)
|
909 |
|
910 |
|
911 |
def SetupNodeDaemon(opts, cluster_name, node, ssh_port): |
912 |
"""Add a node to the cluster.
|
913 |
|
914 |
This function must be called before the actual opcode, and will ssh
|
915 |
to the remote node, copy the needed files, and start ganeti-noded,
|
916 |
allowing the master to do the rest via normal rpc calls.
|
917 |
|
918 |
@param cluster_name: the cluster name
|
919 |
@param node: the name of the new node
|
920 |
@param ssh_port: the SSH port of the new node
|
921 |
|
922 |
"""
|
923 |
data = { |
924 |
constants.NDS_CLUSTER_NAME: cluster_name, |
925 |
constants.NDS_NODE_DAEMON_CERTIFICATE: |
926 |
utils.ReadFile(pathutils.NODED_CERT_FILE), |
927 |
constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(), |
928 |
constants.NDS_START_NODE_DAEMON: True,
|
929 |
} |
930 |
|
931 |
RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP, |
932 |
opts.debug, opts.verbose, |
933 |
True, opts.ssh_key_check, opts.ssh_key_check,
|
934 |
ssh_port, data) |
935 |
|
936 |
_WaitForNodeDaemon(node) |
937 |
|
938 |
|
939 |
def MasterFailover(no_voting=False): |
940 |
"""Failover the master node.
|
941 |
|
942 |
This checks that we are not already the master, and will cause the
|
943 |
current master to cease being master, and the non-master to become
|
944 |
new master.
|
945 |
|
946 |
@type no_voting: boolean
|
947 |
@param no_voting: force the operation without remote nodes agreement
|
948 |
(dangerous)
|
949 |
|
950 |
"""
|
951 |
sstore = ssconf.SimpleStore() |
952 |
|
953 |
old_master, new_master = ssconf.GetMasterAndMyself(sstore) |
954 |
node_names = sstore.GetNodeList() |
955 |
mc_list = sstore.GetMasterCandidates() |
956 |
|
957 |
if old_master == new_master:
|
958 |
raise errors.OpPrereqError("This commands must be run on the node" |
959 |
" where you want the new master to be."
|
960 |
" %s is already the master" %
|
961 |
old_master, errors.ECODE_INVAL) |
962 |
|
963 |
if new_master not in mc_list: |
964 |
mc_no_master = [name for name in mc_list if name != old_master] |
965 |
raise errors.OpPrereqError("This node is not among the nodes marked" |
966 |
" as master candidates. Only these nodes"
|
967 |
" can become masters. Current list of"
|
968 |
" master candidates is:\n"
|
969 |
"%s" % ("\n".join(mc_no_master)), |
970 |
errors.ECODE_STATE) |
971 |
|
972 |
if not no_voting: |
973 |
vote_list = GatherMasterVotes(node_names) |
974 |
|
975 |
if vote_list:
|
976 |
voted_master = vote_list[0][0] |
977 |
if voted_master is None: |
978 |
raise errors.OpPrereqError("Cluster is inconsistent, most nodes did" |
979 |
" not respond.", errors.ECODE_ENVIRON)
|
980 |
elif voted_master != old_master:
|
981 |
raise errors.OpPrereqError("I have a wrong configuration, I believe" |
982 |
" the master is %s but the other nodes"
|
983 |
" voted %s. Please resync the configuration"
|
984 |
" of this node." %
|
985 |
(old_master, voted_master), |
986 |
errors.ECODE_STATE) |
987 |
# end checks
|
988 |
|
989 |
rcode = 0
|
990 |
|
991 |
logging.info("Setting master to %s, old master: %s", new_master, old_master)
|
992 |
|
993 |
try:
|
994 |
# Start WConfd so that we can access the configuration
|
995 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, |
996 |
"start", constants.WCONFD, "--force-node"]) |
997 |
if result.failed:
|
998 |
raise errors.OpPrereqError("Could not start the configuration daemon," |
999 |
" command %s had exitcode %s and error %s" %
|
1000 |
(result.cmd, result.exit_code, result.output), |
1001 |
errors.ECODE_NOENT) |
1002 |
|
1003 |
# instantiate a real config writer, as we now know we have the
|
1004 |
# configuration data
|
1005 |
cfg = config.ConfigWriter(accept_foreign=True)
|
1006 |
|
1007 |
old_master_node = cfg.GetNodeInfoByName(old_master) |
1008 |
if old_master_node is None: |
1009 |
raise errors.OpPrereqError("Could not find old master node '%s' in" |
1010 |
" cluster configuration." % old_master,
|
1011 |
errors.ECODE_NOENT) |
1012 |
|
1013 |
cluster_info = cfg.GetClusterInfo() |
1014 |
new_master_node = cfg.GetNodeInfoByName(new_master) |
1015 |
if new_master_node is None: |
1016 |
raise errors.OpPrereqError("Could not find new master node '%s' in" |
1017 |
" cluster configuration." % new_master,
|
1018 |
errors.ECODE_NOENT) |
1019 |
|
1020 |
cluster_info.master_node = new_master_node.uuid |
1021 |
# this will also regenerate the ssconf files, since we updated the
|
1022 |
# cluster info
|
1023 |
cfg.Update(cluster_info, logging.error) |
1024 |
except errors.ConfigurationError, err:
|
1025 |
logging.error("Error while trying to set the new master: %s",
|
1026 |
str(err))
|
1027 |
return 1 |
1028 |
finally:
|
1029 |
# stop WConfd again:
|
1030 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.WCONFD])
|
1031 |
if result.failed:
|
1032 |
logging.error("Could not stop the configuration daemon,"
|
1033 |
" command %s had exitcode %s and error %s",
|
1034 |
result.cmd, result.exit_code, result.output) |
1035 |
|
1036 |
# if cfg.Update worked, then it means the old master daemon won't be
|
1037 |
# able now to write its own config file (we rely on locking in both
|
1038 |
# backend.UploadFile() and ConfigWriter._Write(); hence the next
|
1039 |
# step is to kill the old master
|
1040 |
|
1041 |
logging.info("Stopping the master daemon on node %s", old_master)
|
1042 |
|
1043 |
runner = rpc.BootstrapRunner() |
1044 |
master_params = cfg.GetMasterNetworkParameters() |
1045 |
master_params.uuid = old_master_node.uuid |
1046 |
ems = cfg.GetUseExternalMipScript() |
1047 |
result = runner.call_node_deactivate_master_ip(old_master, |
1048 |
master_params, ems) |
1049 |
|
1050 |
msg = result.fail_msg |
1051 |
if msg:
|
1052 |
logging.warning("Could not disable the master IP: %s", msg)
|
1053 |
|
1054 |
result = runner.call_node_stop_master(old_master) |
1055 |
msg = result.fail_msg |
1056 |
if msg:
|
1057 |
logging.error("Could not disable the master role on the old master"
|
1058 |
" %s, please disable manually: %s", old_master, msg)
|
1059 |
|
1060 |
logging.info("Checking master IP non-reachability...")
|
1061 |
|
1062 |
master_ip = sstore.GetMasterIP() |
1063 |
total_timeout = 30
|
1064 |
|
1065 |
# Here we have a phase where no master should be running
|
1066 |
def _check_ip(): |
1067 |
if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
|
1068 |
raise utils.RetryAgain()
|
1069 |
|
1070 |
try:
|
1071 |
utils.Retry(_check_ip, (1, 1.5, 5), total_timeout) |
1072 |
except utils.RetryTimeout:
|
1073 |
logging.warning("The master IP is still reachable after %s seconds,"
|
1074 |
" continuing but activating the master on the current"
|
1075 |
" node will probably fail", total_timeout)
|
1076 |
|
1077 |
if jstore.CheckDrainFlag():
|
1078 |
logging.info("Undraining job queue")
|
1079 |
jstore.SetDrainFlag(False)
|
1080 |
|
1081 |
logging.info("Starting the master daemons on the new master")
|
1082 |
|
1083 |
result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master, |
1084 |
no_voting) |
1085 |
msg = result.fail_msg |
1086 |
if msg:
|
1087 |
logging.error("Could not start the master role on the new master"
|
1088 |
" %s, please check: %s", new_master, msg)
|
1089 |
rcode = 1
|
1090 |
|
1091 |
logging.info("Master failed over from %s to %s", old_master, new_master)
|
1092 |
return rcode
|
1093 |
|
1094 |
|
1095 |
def GetMaster(): |
1096 |
"""Returns the current master node.
|
1097 |
|
1098 |
This is a separate function in bootstrap since it's needed by
|
1099 |
gnt-cluster, and instead of importing directly ssconf, it's better
|
1100 |
to abstract it in bootstrap, where we do use ssconf in other
|
1101 |
functions too.
|
1102 |
|
1103 |
"""
|
1104 |
sstore = ssconf.SimpleStore() |
1105 |
|
1106 |
old_master, _ = ssconf.GetMasterAndMyself(sstore) |
1107 |
|
1108 |
return old_master
|
1109 |
|
1110 |
|
1111 |
def GatherMasterVotes(node_names): |
1112 |
"""Check the agreement on who is the master.
|
1113 |
|
1114 |
This function will return a list of (node, number of votes), ordered
|
1115 |
by the number of votes. Errors will be denoted by the key 'None'.
|
1116 |
|
1117 |
Note that the sum of votes is the number of nodes this machine
|
1118 |
knows, whereas the number of entries in the list could be different
|
1119 |
(if some nodes vote for another master).
|
1120 |
|
1121 |
We remove ourselves from the list since we know that (bugs aside)
|
1122 |
since we use the same source for configuration information for both
|
1123 |
backend and boostrap, we'll always vote for ourselves.
|
1124 |
|
1125 |
@type node_names: list
|
1126 |
@param node_names: the list of nodes to query for master info; the current
|
1127 |
node will be removed if it is in the list
|
1128 |
@rtype: list
|
1129 |
@return: list of (node, votes)
|
1130 |
|
1131 |
"""
|
1132 |
myself = netutils.Hostname.GetSysName() |
1133 |
try:
|
1134 |
node_names.remove(myself) |
1135 |
except ValueError: |
1136 |
pass
|
1137 |
if not node_names: |
1138 |
# no nodes left (eventually after removing myself)
|
1139 |
return []
|
1140 |
results = rpc.BootstrapRunner().call_master_node_name(node_names) |
1141 |
if not isinstance(results, dict): |
1142 |
# this should not happen (unless internal error in rpc)
|
1143 |
logging.critical("Can't complete rpc call, aborting master startup")
|
1144 |
return [(None, len(node_names))] |
1145 |
votes = {} |
1146 |
for node_name in results: |
1147 |
nres = results[node_name] |
1148 |
msg = nres.fail_msg |
1149 |
|
1150 |
if msg:
|
1151 |
logging.warning("Error contacting node %s: %s", node_name, msg)
|
1152 |
node = None
|
1153 |
else:
|
1154 |
node = nres.payload |
1155 |
|
1156 |
if node not in votes: |
1157 |
votes[node] = 1
|
1158 |
else:
|
1159 |
votes[node] += 1
|
1160 |
|
1161 |
vote_list = [v for v in votes.items()] |
1162 |
# sort first on number of votes then on name, since we want None
|
1163 |
# sorted later if we have the half of the nodes not responding, and
|
1164 |
# half voting all for the same master
|
1165 |
vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True) |
1166 |
|
1167 |
return vote_list
|