Implement predictive queue cluster parameter
[ganeti-github.git] / lib / bootstrap.py
1 #
2 #
3
4 # Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 Google Inc.
5 # All rights reserved.
6 #
7 # Redistribution and use in source and binary forms, with or without
8 # modification, are permitted provided that the following conditions are
9 # met:
10 #
11 # 1. Redistributions of source code must retain the above copyright notice,
12 # this list of conditions and the following disclaimer.
13 #
14 # 2. Redistributions in binary form must reproduce the above copyright
15 # notice, this list of conditions and the following disclaimer in the
16 # documentation and/or other materials provided with the distribution.
17 #
18 # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
19 # IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
20 # TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
21 # PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
22 # CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
23 # EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
24 # PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
25 # PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
26 # LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
27 # NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
28 # SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29
30
31 """Functions to bootstrap a new cluster.
32
33 """
34
35 import os
36 import os.path
37 import re
38 import logging
39 import time
40
41 from ganeti.cmdlib import cluster
42 import ganeti.rpc.node as rpc
43 from ganeti import ssh
44 from ganeti import utils
45 from ganeti import errors
46 from ganeti import config
47 from ganeti import constants
48 from ganeti import objects
49 from ganeti import ssconf
50 from ganeti import serializer
51 from ganeti import hypervisor
52 from ganeti.storage import drbd
53 from ganeti.storage import filestorage
54 from ganeti import netutils
55 from ganeti import luxi
56 from ganeti import jstore
57 from ganeti import pathutils
58 from ganeti import runtime
59 from ganeti import vcluster
60
61
62 # ec_id for InitConfig's temporary reservation manager
63 _INITCONF_ECID = "initconfig-ecid"
64
65 #: After how many seconds daemon must be responsive
66 _DAEMON_READY_TIMEOUT = 10.0
67
68
69 def GenerateHmacKey(file_name):
70 """Writes a new HMAC key.
71
72 @type file_name: str
73 @param file_name: Path to output file
74
75 """
76 utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400,
77 backup=True)
78
79
80 # pylint: disable=R0913
81 def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert,
82 new_confd_hmac_key, new_cds, new_client_cert,
83 master_name,
84 rapi_cert_pem=None, spice_cert_pem=None,
85 spice_cacert_pem=None, cds=None,
86 nodecert_file=pathutils.NODED_CERT_FILE,
87 clientcert_file=pathutils.NODED_CLIENT_CERT_FILE,
88 rapicert_file=pathutils.RAPI_CERT_FILE,
89 spicecert_file=pathutils.SPICE_CERT_FILE,
90 spicecacert_file=pathutils.SPICE_CACERT_FILE,
91 hmackey_file=pathutils.CONFD_HMAC_KEY,
92 cds_file=pathutils.CLUSTER_DOMAIN_SECRET_FILE):
93 """Updates the cluster certificates, keys and secrets.
94
95 @type new_cluster_cert: bool
96 @param new_cluster_cert: Whether to generate a new cluster certificate
97 @type new_rapi_cert: bool
98 @param new_rapi_cert: Whether to generate a new RAPI certificate
99 @type new_spice_cert: bool
100 @param new_spice_cert: Whether to generate a new SPICE certificate
101 @type new_confd_hmac_key: bool
102 @param new_confd_hmac_key: Whether to generate a new HMAC key
103 @type new_cds: bool
104 @param new_cds: Whether to generate a new cluster domain secret
105 @type new_client_cert: bool
106 @param new_client_cert: Whether to generate a new client certificate
107 @type master_name: string
108 @param master_name: FQDN of the master node
109 @type rapi_cert_pem: string
110 @param rapi_cert_pem: New RAPI certificate in PEM format
111 @type spice_cert_pem: string
112 @param spice_cert_pem: New SPICE certificate in PEM format
113 @type spice_cacert_pem: string
114 @param spice_cacert_pem: Certificate of the CA that signed the SPICE
115 certificate, in PEM format
116 @type cds: string
117 @param cds: New cluster domain secret
118 @type nodecert_file: string
119 @param nodecert_file: optional override of the node cert file path
120 @type rapicert_file: string
121 @param rapicert_file: optional override of the rapi cert file path
122 @type spicecert_file: string
123 @param spicecert_file: optional override of the spice cert file path
124 @type spicecacert_file: string
125 @param spicecacert_file: optional override of the spice CA cert file path
126 @type hmackey_file: string
127 @param hmackey_file: optional override of the hmac key file path
128
129 """
130 # pylint: disable=R0913
131 # noded SSL certificate
132 utils.GenerateNewSslCert(
133 new_cluster_cert, nodecert_file, 1,
134 "Generating new cluster certificate at %s" % nodecert_file)
135
136 # If the cluster certificate was renewed, the client cert has to be
137 # renewed and resigned.
138 if new_cluster_cert or new_client_cert:
139 utils.GenerateNewClientSslCert(clientcert_file, nodecert_file,
140 master_name)
141
142 # confd HMAC key
143 if new_confd_hmac_key or not os.path.exists(hmackey_file):
144 logging.debug("Writing new confd HMAC key to %s", hmackey_file)
145 GenerateHmacKey(hmackey_file)
146
147 if rapi_cert_pem:
148 # Assume rapi_pem contains a valid PEM-formatted certificate and key
149 logging.debug("Writing RAPI certificate at %s", rapicert_file)
150 utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
151
152 else:
153 utils.GenerateNewSslCert(
154 new_rapi_cert, rapicert_file, 1,
155 "Generating new RAPI certificate at %s" % rapicert_file)
156
157 # SPICE
158 spice_cert_exists = os.path.exists(spicecert_file)
159 spice_cacert_exists = os.path.exists(spicecacert_file)
160 if spice_cert_pem:
161 # spice_cert_pem implies also spice_cacert_pem
162 logging.debug("Writing SPICE certificate at %s", spicecert_file)
163 utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True)
164 logging.debug("Writing SPICE CA certificate at %s", spicecacert_file)
165 utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True)
166 elif new_spice_cert or not spice_cert_exists:
167 if spice_cert_exists:
168 utils.CreateBackup(spicecert_file)
169 if spice_cacert_exists:
170 utils.CreateBackup(spicecacert_file)
171
172 logging.debug("Generating new self-signed SPICE certificate at %s",
173 spicecert_file)
174 (_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file, 1)
175
176 # Self-signed certificate -> the public certificate is also the CA public
177 # certificate
178 logging.debug("Writing the public certificate to %s",
179 spicecert_file)
180 utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem)
181
182 # Cluster domain secret
183 if cds:
184 logging.debug("Writing cluster domain secret to %s", cds_file)
185 utils.WriteFile(cds_file, data=cds, backup=True)
186
187 elif new_cds or not os.path.exists(cds_file):
188 logging.debug("Generating new cluster domain secret at %s", cds_file)
189 GenerateHmacKey(cds_file)
190
191
192 def _InitGanetiServerSetup(master_name, cfg):
193 """Setup the necessary configuration for the initial node daemon.
194
195 This creates the nodepass file containing the shared password for
196 the cluster, generates the SSL certificate and starts the node daemon.
197
198 @type master_name: str
199 @param master_name: Name of the master node
200 @type cfg: ConfigWriter
201 @param cfg: the configuration writer
202
203 """
204 # Generate cluster secrets
205 GenerateClusterCrypto(True, False, False, False, False, False, master_name)
206
207 # Add the master's SSL certificate digest to the configuration.
208 master_uuid = cfg.GetMasterNode()
209 master_digest = utils.GetCertificateDigest()
210 cfg.AddNodeToCandidateCerts(master_uuid, master_digest)
211 cfg.Update(cfg.GetClusterInfo(), logging.error)
212 ssconf.WriteSsconfFiles(cfg.GetSsconfValues())
213
214 if not os.path.exists(
215 os.path.join(pathutils.DATA_DIR,
216 "%s%s" % (constants.SSCONF_FILEPREFIX,
217 constants.SS_MASTER_CANDIDATES_CERTS))):
218 raise errors.OpExecError("Ssconf file for master candidate certificates"
219 " was not written.")
220
221 if not os.path.exists(pathutils.NODED_CERT_FILE):
222 raise errors.OpExecError("The server certficate was not created properly.")
223
224 if not os.path.exists(pathutils.NODED_CLIENT_CERT_FILE):
225 raise errors.OpExecError("The client certificate was not created"
226 " properly.")
227
228 # set up the inter-node password and certificate
229 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start", constants.NODED])
230 if result.failed:
231 raise errors.OpExecError("Could not start the node daemon, command %s"
232 " had exitcode %s and error %s" %
233 (result.cmd, result.exit_code, result.output))
234
235 _WaitForNodeDaemon(master_name)
236
237
238 def _WaitForNodeDaemon(node_name):
239 """Wait for node daemon to become responsive.
240
241 """
242 def _CheckNodeDaemon():
243 # Pylint bug <http://www.logilab.org/ticket/35642>
244 # pylint: disable=E1101
245 result = rpc.BootstrapRunner().call_version([node_name])[node_name]
246 if result.fail_msg:
247 raise utils.RetryAgain()
248
249 try:
250 utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT)
251 except utils.RetryTimeout:
252 raise errors.OpExecError("Node daemon on %s didn't answer queries within"
253 " %s seconds" % (node_name, _DAEMON_READY_TIMEOUT))
254
255
256 def _WaitForMasterDaemon():
257 """Wait for master daemon to become responsive.
258
259 """
260 def _CheckMasterDaemon():
261 try:
262 cl = luxi.Client()
263 (cluster_name, ) = cl.QueryConfigValues(["cluster_name"])
264 except Exception:
265 raise utils.RetryAgain()
266
267 logging.debug("Received cluster name %s from master", cluster_name)
268
269 try:
270 utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT)
271 except utils.RetryTimeout:
272 raise errors.OpExecError("Master daemon didn't answer queries within"
273 " %s seconds" % _DAEMON_READY_TIMEOUT)
274
275
276 def _WaitForSshDaemon(hostname, port):
277 """Wait for SSH daemon to become responsive.
278
279 """
280 family = ssconf.SimpleStore().GetPrimaryIPFamily()
281 hostip = netutils.GetHostname(name=hostname, family=family).ip
282
283 def _CheckSshDaemon():
284 if netutils.TcpPing(hostip, port, timeout=1.0, live_port_needed=True):
285 logging.debug("SSH daemon on %s:%s (IP address %s) has become"
286 " responsive", hostname, port, hostip)
287 else:
288 raise utils.RetryAgain()
289
290 try:
291 utils.Retry(_CheckSshDaemon, 1.0, _DAEMON_READY_TIMEOUT)
292 except utils.RetryTimeout:
293 raise errors.OpExecError("SSH daemon on %s:%s (IP address %s) didn't"
294 " become responsive within %s seconds" %
295 (hostname, port, hostip, _DAEMON_READY_TIMEOUT))
296
297
298 def _InitFileStorageDir(file_storage_dir):
299 """Initialize if needed the file storage.
300
301 @param file_storage_dir: the user-supplied value
302 @return: either empty string (if file storage was disabled at build
303 time) or the normalized path to the storage directory
304
305 """
306 file_storage_dir = os.path.normpath(file_storage_dir)
307
308 if not os.path.isabs(file_storage_dir):
309 raise errors.OpPrereqError("File storage directory '%s' is not an absolute"
310 " path" % file_storage_dir, errors.ECODE_INVAL)
311
312 if not os.path.exists(file_storage_dir):
313 try:
314 os.makedirs(file_storage_dir, 0750)
315 except OSError, err:
316 raise errors.OpPrereqError("Cannot create file storage directory"
317 " '%s': %s" % (file_storage_dir, err),
318 errors.ECODE_ENVIRON)
319
320 if not os.path.isdir(file_storage_dir):
321 raise errors.OpPrereqError("The file storage directory '%s' is not"
322 " a directory." % file_storage_dir,
323 errors.ECODE_ENVIRON)
324
325 return file_storage_dir
326
327
328 def _PrepareFileBasedStorage(
329 enabled_disk_templates, file_storage_dir,
330 default_dir, file_disk_template, _storage_path_acceptance_fn,
331 init_fn=_InitFileStorageDir, acceptance_fn=None):
332 """Checks if a file-base storage type is enabled and inits the dir.
333
334 @type enabled_disk_templates: list of string
335 @param enabled_disk_templates: list of enabled disk templates
336 @type file_storage_dir: string
337 @param file_storage_dir: the file storage directory
338 @type default_dir: string
339 @param default_dir: default file storage directory when C{file_storage_dir}
340 is 'None'
341 @type file_disk_template: string
342 @param file_disk_template: a disk template whose storage type is 'ST_FILE',
343 'ST_SHARED_FILE' or 'ST_GLUSTER'
344 @type _storage_path_acceptance_fn: function
345 @param _storage_path_acceptance_fn: checks whether the given file-based
346 storage directory is acceptable
347 @see: C{cluster.CheckFileBasedStoragePathVsEnabledDiskTemplates} for details
348
349 @rtype: string
350 @returns: the name of the actual file storage directory
351
352 """
353 assert (file_disk_template in utils.storage.GetDiskTemplatesOfStorageTypes(
354 constants.ST_FILE, constants.ST_SHARED_FILE, constants.ST_GLUSTER
355 ))
356
357 file_storage_enabled = file_disk_template in enabled_disk_templates
358
359 if file_storage_dir is None:
360 if file_storage_enabled:
361 file_storage_dir = default_dir
362 else:
363 file_storage_dir = ""
364
365 if not acceptance_fn:
366 acceptance_fn = \
367 lambda path: filestorage.CheckFileStoragePathAcceptance(
368 path, exact_match_ok=True)
369
370 _storage_path_acceptance_fn(logging.warning, file_storage_dir,
371 enabled_disk_templates)
372
373 if file_storage_enabled:
374 try:
375 acceptance_fn(file_storage_dir)
376 except errors.FileStoragePathError as e:
377 raise errors.OpPrereqError(str(e))
378 result_file_storage_dir = init_fn(file_storage_dir)
379 else:
380 result_file_storage_dir = file_storage_dir
381 return result_file_storage_dir
382
383
384 def _PrepareFileStorage(
385 enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir,
386 acceptance_fn=None):
387 """Checks if file storage is enabled and inits the dir.
388
389 @see: C{_PrepareFileBasedStorage}
390
391 """
392 return _PrepareFileBasedStorage(
393 enabled_disk_templates, file_storage_dir,
394 pathutils.DEFAULT_FILE_STORAGE_DIR, constants.DT_FILE,
395 cluster.CheckFileStoragePathVsEnabledDiskTemplates,
396 init_fn=init_fn, acceptance_fn=acceptance_fn)
397
398
399 def _PrepareSharedFileStorage(
400 enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir,
401 acceptance_fn=None):
402 """Checks if shared file storage is enabled and inits the dir.
403
404 @see: C{_PrepareFileBasedStorage}
405
406 """
407 return _PrepareFileBasedStorage(
408 enabled_disk_templates, file_storage_dir,
409 pathutils.DEFAULT_SHARED_FILE_STORAGE_DIR, constants.DT_SHARED_FILE,
410 cluster.CheckSharedFileStoragePathVsEnabledDiskTemplates,
411 init_fn=init_fn, acceptance_fn=acceptance_fn)
412
413
414 def _PrepareGlusterStorage(
415 enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir,
416 acceptance_fn=None):
417 """Checks if gluster storage is enabled and inits the dir.
418
419 @see: C{_PrepareFileBasedStorage}
420
421 """
422 return _PrepareFileBasedStorage(
423 enabled_disk_templates, file_storage_dir,
424 pathutils.DEFAULT_GLUSTER_STORAGE_DIR, constants.DT_GLUSTER,
425 cluster.CheckGlusterStoragePathVsEnabledDiskTemplates,
426 init_fn=init_fn, acceptance_fn=acceptance_fn)
427
428
429 def _InitCheckEnabledDiskTemplates(enabled_disk_templates):
430 """Checks the sanity of the enabled disk templates.
431
432 """
433 if not enabled_disk_templates:
434 raise errors.OpPrereqError("Enabled disk templates list must contain at"
435 " least one member", errors.ECODE_INVAL)
436 invalid_disk_templates = \
437 set(enabled_disk_templates) - constants.DISK_TEMPLATES
438 if invalid_disk_templates:
439 raise errors.OpPrereqError("Enabled disk templates list contains invalid"
440 " entries: %s" % invalid_disk_templates,
441 errors.ECODE_INVAL)
442
443
444 def _RestrictIpolicyToEnabledDiskTemplates(ipolicy, enabled_disk_templates):
445 """Restricts the ipolicy's disk templates to the enabled ones.
446
447 This function clears the ipolicy's list of allowed disk templates from the
448 ones that are not enabled by the cluster.
449
450 @type ipolicy: dict
451 @param ipolicy: the instance policy
452 @type enabled_disk_templates: list of string
453 @param enabled_disk_templates: the list of cluster-wide enabled disk
454 templates
455
456 """
457 assert constants.IPOLICY_DTS in ipolicy
458 allowed_disk_templates = ipolicy[constants.IPOLICY_DTS]
459 restricted_disk_templates = list(set(allowed_disk_templates)
460 .intersection(set(enabled_disk_templates)))
461 ipolicy[constants.IPOLICY_DTS] = restricted_disk_templates
462
463
464 def _InitCheckDrbdHelper(drbd_helper, drbd_enabled):
465 """Checks the DRBD usermode helper.
466
467 @type drbd_helper: string
468 @param drbd_helper: name of the DRBD usermode helper that the system should
469 use
470
471 """
472 if not drbd_enabled:
473 return
474
475 if drbd_helper is not None:
476 try:
477 curr_helper = drbd.DRBD8.GetUsermodeHelper()
478 except errors.BlockDeviceError, err:
479 raise errors.OpPrereqError("Error while checking drbd helper"
480 " (disable drbd with --enabled-disk-templates"
481 " if you are not using drbd): %s" % str(err),
482 errors.ECODE_ENVIRON)
483 if drbd_helper != curr_helper:
484 raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s"
485 " is the current helper" % (drbd_helper,
486 curr_helper),
487 errors.ECODE_INVAL)
488
489
490 def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914
491 master_netmask, master_netdev, file_storage_dir,
492 shared_file_storage_dir, gluster_storage_dir,
493 candidate_pool_size, ssh_key_type, ssh_key_bits,
494 secondary_ip=None, vg_name=None, beparams=None, nicparams=None,
495 ndparams=None, hvparams=None, diskparams=None,
496 enabled_hypervisors=None, modify_etc_hosts=True,
497 modify_ssh_setup=True, maintain_node_health=False,
498 drbd_helper=None, uid_pool=None, default_iallocator=None,
499 default_iallocator_params=None, primary_ip_version=None,
500 ipolicy=None, prealloc_wipe_disks=False,
501 use_external_mip_script=False, hv_state=None, disk_state=None,
502 enabled_disk_templates=None, install_image=None,
503 zeroing_image=None, compression_tools=None,
504 enabled_user_shutdown=False, enabled_predictive_queue=True):
505 """Initialise the cluster.
506
507 @type candidate_pool_size: int
508 @param candidate_pool_size: master candidate pool size
509
510 @type enabled_disk_templates: list of string
511 @param enabled_disk_templates: list of disk_templates to be used in this
512 cluster
513
514 @type enabled_user_shutdown: bool
515 @param enabled_user_shutdown: whether user shutdown is enabled cluster
516 wide
517
518 """
519 # TODO: complete the docstring
520 if config.ConfigWriter.IsCluster():
521 raise errors.OpPrereqError("Cluster is already initialised",
522 errors.ECODE_STATE)
523
524 data_dir = vcluster.AddNodePrefix(pathutils.DATA_DIR)
525 queue_dir = vcluster.AddNodePrefix(pathutils.QUEUE_DIR)
526 archive_dir = vcluster.AddNodePrefix(pathutils.JOB_QUEUE_ARCHIVE_DIR)
527 for ddir in [queue_dir, data_dir, archive_dir]:
528 if os.path.isdir(ddir):
529 for entry in os.listdir(ddir):
530 if not os.path.isdir(os.path.join(ddir, entry)):
531 raise errors.OpPrereqError(
532 "%s contains non-directory entries like %s. Remove left-overs of an"
533 " old cluster before initialising a new one" % (ddir, entry),
534 errors.ECODE_STATE)
535
536 if not enabled_hypervisors:
537 raise errors.OpPrereqError("Enabled hypervisors list must contain at"
538 " least one member", errors.ECODE_INVAL)
539 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
540 if invalid_hvs:
541 raise errors.OpPrereqError("Enabled hypervisors contains invalid"
542 " entries: %s" % invalid_hvs,
543 errors.ECODE_INVAL)
544
545 _InitCheckEnabledDiskTemplates(enabled_disk_templates)
546
547 try:
548 ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version)
549 except errors.ProgrammerError:
550 raise errors.OpPrereqError("Invalid primary ip version: %d." %
551 primary_ip_version, errors.ECODE_INVAL)
552
553 hostname = netutils.GetHostname(family=ipcls.family)
554 if not ipcls.IsValid(hostname.ip):
555 raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d"
556 " address." % (hostname.ip, primary_ip_version),
557 errors.ECODE_INVAL)
558
559 if ipcls.IsLoopback(hostname.ip):
560 raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback"
561 " address. Please fix DNS or %s." %
562 (hostname.ip, pathutils.ETC_HOSTS),
563 errors.ECODE_ENVIRON)
564
565 if not ipcls.Own(hostname.ip):
566 raise errors.OpPrereqError("Inconsistency: this host's name resolves"
567 " to %s,\nbut this ip address does not"
568 " belong to this host" %
569 hostname.ip, errors.ECODE_ENVIRON)
570
571 clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family)
572
573 if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5):
574 raise errors.OpPrereqError("Cluster IP already active",
575 errors.ECODE_NOTUNIQUE)
576
577 if not secondary_ip:
578 if primary_ip_version == constants.IP6_VERSION:
579 raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
580 " IPv4 address must be given as secondary",
581 errors.ECODE_INVAL)
582 secondary_ip = hostname.ip
583
584 if not netutils.IP4Address.IsValid(secondary_ip):
585 raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid"
586 " IPv4 address." % secondary_ip,
587 errors.ECODE_INVAL)
588
589 if not netutils.IP4Address.Own(secondary_ip):
590 raise errors.OpPrereqError("You gave %s as secondary IP,"
591 " but it does not belong to this host." %
592 secondary_ip, errors.ECODE_ENVIRON)
593
594 if master_netmask is not None:
595 if not ipcls.ValidateNetmask(master_netmask):
596 raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " %
597 (master_netmask, primary_ip_version),
598 errors.ECODE_INVAL)
599 else:
600 master_netmask = ipcls.iplen
601
602 if vg_name:
603 # Check if volume group is valid
604 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name,
605 constants.MIN_VG_SIZE)
606 if vgstatus:
607 raise errors.OpPrereqError("Error: %s" % vgstatus, errors.ECODE_INVAL)
608
609 drbd_enabled = constants.DT_DRBD8 in enabled_disk_templates
610 _InitCheckDrbdHelper(drbd_helper, drbd_enabled)
611
612 logging.debug("Stopping daemons (if any are running)")
613 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"])
614 if result.failed:
615 raise errors.OpExecError("Could not stop daemons, command %s"
616 " had exitcode %s and error '%s'" %
617 (result.cmd, result.exit_code, result.output))
618
619 file_storage_dir = _PrepareFileStorage(enabled_disk_templates,
620 file_storage_dir)
621 shared_file_storage_dir = _PrepareSharedFileStorage(enabled_disk_templates,
622 shared_file_storage_dir)
623 gluster_storage_dir = _PrepareGlusterStorage(enabled_disk_templates,
624 gluster_storage_dir)
625
626 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
627 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
628 errors.ECODE_INVAL)
629
630 if not nicparams.get('mode', None) == constants.NIC_MODE_OVS:
631 # Do not do this check if mode=openvswitch, since the openvswitch is not
632 # created yet
633 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
634 if result.failed:
635 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
636 (master_netdev,
637 result.output.strip()), errors.ECODE_INVAL)
638
639 dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)]
640 utils.EnsureDirs(dirs)
641
642 objects.UpgradeBeParams(beparams)
643 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
644 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
645
646 objects.NIC.CheckParameterSyntax(nicparams)
647
648 full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy)
649 _RestrictIpolicyToEnabledDiskTemplates(full_ipolicy, enabled_disk_templates)
650
651 if ndparams is not None:
652 utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES)
653 else:
654 ndparams = dict(constants.NDC_DEFAULTS)
655
656 # This is ugly, as we modify the dict itself
657 # FIXME: Make utils.ForceDictType pure functional or write a wrapper
658 # around it
659 if hv_state:
660 for hvname, hvs_data in hv_state.items():
661 utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES)
662 hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data)
663 else:
664 hv_state = dict((hvname, constants.HVST_DEFAULTS)
665 for hvname in enabled_hypervisors)
666
667 # FIXME: disk_state has no default values yet
668 if disk_state:
669 for storage, ds_data in disk_state.items():
670 if storage not in constants.DS_VALID_TYPES:
671 raise errors.OpPrereqError("Invalid storage type in disk state: %s" %
672 storage, errors.ECODE_INVAL)
673 for ds_name, state in ds_data.items():
674 utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES)
675 ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state)
676
677 # hvparams is a mapping of hypervisor->hvparams dict
678 for hv_name, hv_params in hvparams.iteritems():
679 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
680 hv_class = hypervisor.GetHypervisor(hv_name)
681 hv_class.CheckParameterSyntax(hv_params)
682
683 # diskparams is a mapping of disk-template->diskparams dict
684 for template, dt_params in diskparams.items():
685 param_keys = set(dt_params.keys())
686 default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys())
687 if param_keys > default_param_keys:
688 unknown_params = param_keys - default_param_keys
689 raise errors.OpPrereqError("Invalid parameters for disk template %s:"
690 " %s" % (template,
691 utils.CommaJoin(unknown_params)),
692 errors.ECODE_INVAL)
693 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES)
694 if template == constants.DT_DRBD8 and vg_name is not None:
695 # The default METAVG value is equal to the VG name set at init time,
696 # if provided
697 dt_params[constants.DRBD_DEFAULT_METAVG] = vg_name
698
699 try:
700 utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS)
701 except errors.OpPrereqError, err:
702 raise errors.OpPrereqError("While verify diskparam options: %s" % err,
703 errors.ECODE_INVAL)
704
705 # set up ssh config and /etc/hosts
706 rsa_sshkey = ""
707 dsa_sshkey = ""
708 if os.path.isfile(pathutils.SSH_HOST_RSA_PUB):
709 sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB)
710 rsa_sshkey = sshline.split(" ")[1]
711 if os.path.isfile(pathutils.SSH_HOST_DSA_PUB):
712 sshline = utils.ReadFile(pathutils.SSH_HOST_DSA_PUB)
713 dsa_sshkey = sshline.split(" ")[1]
714 if not rsa_sshkey and not dsa_sshkey:
715 raise errors.OpPrereqError("Failed to find SSH public keys",
716 errors.ECODE_ENVIRON)
717
718 if modify_etc_hosts:
719 utils.AddHostToEtcHosts(hostname.name, hostname.ip)
720
721 if modify_ssh_setup:
722 ssh.InitSSHSetup(ssh_key_type, ssh_key_bits)
723
724 if default_iallocator is not None:
725 alloc_script = utils.FindFile(default_iallocator,
726 constants.IALLOCATOR_SEARCH_PATH,
727 os.path.isfile)
728 if alloc_script is None:
729 raise errors.OpPrereqError("Invalid default iallocator script '%s'"
730 " specified" % default_iallocator,
731 errors.ECODE_INVAL)
732 else:
733 # default to htools
734 if utils.FindFile(constants.IALLOC_HAIL,
735 constants.IALLOCATOR_SEARCH_PATH,
736 os.path.isfile):
737 default_iallocator = constants.IALLOC_HAIL
738
739 # check if we have all the users we need
740 try:
741 runtime.GetEnts()
742 except errors.ConfigurationError, err:
743 raise errors.OpPrereqError("Required system user/group missing: %s" %
744 err, errors.ECODE_ENVIRON)
745
746 candidate_certs = {}
747
748 now = time.time()
749
750 if compression_tools is not None:
751 cluster.CheckCompressionTools(compression_tools)
752
753 initial_dc_config = dict(active=True,
754 interval=int(constants.MOND_TIME_INTERVAL * 1e6))
755 data_collectors = dict(
756 (name, initial_dc_config.copy())
757 for name in constants.DATA_COLLECTOR_NAMES)
758
759 # init of cluster config file
760 cluster_config = objects.Cluster(
761 serial_no=1,
762 rsahostkeypub=rsa_sshkey,
763 dsahostkeypub=dsa_sshkey,
764 highest_used_port=(constants.FIRST_DRBD_PORT - 1),
765 mac_prefix=mac_prefix,
766 volume_group_name=vg_name,
767 tcpudp_port_pool=set(),
768 master_ip=clustername.ip,
769 master_netmask=master_netmask,
770 master_netdev=master_netdev,
771 cluster_name=clustername.name,
772 file_storage_dir=file_storage_dir,
773 shared_file_storage_dir=shared_file_storage_dir,
774 gluster_storage_dir=gluster_storage_dir,
775 enabled_hypervisors=enabled_hypervisors,
776 beparams={constants.PP_DEFAULT: beparams},
777 nicparams={constants.PP_DEFAULT: nicparams},
778 ndparams=ndparams,
779 hvparams=hvparams,
780 diskparams=diskparams,
781 candidate_pool_size=candidate_pool_size,
782 modify_etc_hosts=modify_etc_hosts,
783 modify_ssh_setup=modify_ssh_setup,
784 uid_pool=uid_pool,
785 ctime=now,
786 mtime=now,
787 maintain_node_health=maintain_node_health,
788 data_collectors=data_collectors,
789 drbd_usermode_helper=drbd_helper,
790 default_iallocator=default_iallocator,
791 default_iallocator_params=default_iallocator_params,
792 primary_ip_family=ipcls.family,
793 prealloc_wipe_disks=prealloc_wipe_disks,
794 use_external_mip_script=use_external_mip_script,
795 ipolicy=full_ipolicy,
796 hv_state_static=hv_state,
797 disk_state_static=disk_state,
798 enabled_disk_templates=enabled_disk_templates,
799 candidate_certs=candidate_certs,
800 osparams={},
801 osparams_private_cluster={},
802 install_image=install_image,
803 zeroing_image=zeroing_image,
804 compression_tools=compression_tools,
805 enabled_user_shutdown=enabled_user_shutdown,
806 ssh_key_type=ssh_key_type,
807 ssh_key_bits=ssh_key_bits,
808 enabled_predictive_queue=enabled_predictive_queue,
809 )
810 master_node_config = objects.Node(name=hostname.name,
811 primary_ip=hostname.ip,
812 secondary_ip=secondary_ip,
813 serial_no=1,
814 master_candidate=True,
815 offline=False, drained=False,
816 ctime=now, mtime=now,
817 )
818 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
819 cfg = config.ConfigWriter(offline=True)
820 ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE)
821 cfg.Update(cfg.GetClusterInfo(), logging.error)
822 ssconf.WriteSsconfFiles(cfg.GetSsconfValues())
823
824 master_uuid = cfg.GetMasterNode()
825 if modify_ssh_setup:
826 ssh.InitPubKeyFile(master_uuid, ssh_key_type)
827 # set up the inter-node password and certificate
828 _InitGanetiServerSetup(hostname.name, cfg)
829
830 logging.debug("Starting daemons")
831 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"])
832 if result.failed:
833 raise errors.OpExecError("Could not start daemons, command %s"
834 " had exitcode %s and error %s" %
835 (result.cmd, result.exit_code, result.output))
836
837 _WaitForMasterDaemon()
838
839
840 def InitConfig(version, cluster_config, master_node_config,
841 cfg_file=pathutils.CLUSTER_CONF_FILE):
842 """Create the initial cluster configuration.
843
844 It will contain the current node, which will also be the master
845 node, and no instances.
846
847 @type version: int
848 @param version: configuration version
849 @type cluster_config: L{objects.Cluster}
850 @param cluster_config: cluster configuration
851 @type master_node_config: L{objects.Node}
852 @param master_node_config: master node configuration
853 @type cfg_file: string
854 @param cfg_file: configuration file path
855
856 """
857 uuid_generator = config.TemporaryReservationManager()
858 cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID,
859 _INITCONF_ECID)
860 master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID,
861 _INITCONF_ECID)
862 cluster_config.master_node = master_node_config.uuid
863 nodes = {
864 master_node_config.uuid: master_node_config,
865 }
866 default_nodegroup = objects.NodeGroup(
867 uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID),
868 name=constants.INITIAL_NODE_GROUP_NAME,
869 members=[master_node_config.uuid],
870 diskparams={},
871 )
872 nodegroups = {
873 default_nodegroup.uuid: default_nodegroup,
874 }
875 now = time.time()
876 maintenance = objects.Maintenance(serial_no=1, ctime=now, mtime=now)
877 config_data = objects.ConfigData(version=version,
878 cluster=cluster_config,
879 nodegroups=nodegroups,
880 nodes=nodes,
881 instances={},
882 networks={},
883 disks={},
884 filters={},
885 maintenance=maintenance,
886 serial_no=1,
887 ctime=now, mtime=now)
888 utils.WriteFile(cfg_file,
889 data=serializer.Dump(config_data.ToDict()),
890 mode=0600)
891
892
893 def FinalizeClusterDestroy(master_uuid):
894 """Execute the last steps of cluster destroy
895
896 This function shuts down all the daemons, completing the destroy
897 begun in cmdlib.LUDestroyOpcode.
898
899 """
900 livelock = utils.livelock.LiveLock("bootstrap_destroy")
901 cfg = config.GetConfig(None, livelock)
902 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
903 runner = rpc.BootstrapRunner()
904
905 master_name = cfg.GetNodeName(master_uuid)
906
907 master_params = cfg.GetMasterNetworkParameters()
908 master_params.uuid = master_uuid
909 ems = cfg.GetUseExternalMipScript()
910 result = runner.call_node_deactivate_master_ip(master_name, master_params,
911 ems)
912
913 msg = result.fail_msg
914 if msg:
915 logging.warning("Could not disable the master IP: %s", msg)
916
917 result = runner.call_node_stop_master(master_name)
918 msg = result.fail_msg
919 if msg:
920 logging.warning("Could not disable the master role: %s", msg)
921
922 result = runner.call_node_leave_cluster(master_name, modify_ssh_setup)
923 msg = result.fail_msg
924 if msg:
925 logging.warning("Could not shutdown the node daemon and cleanup"
926 " the node: %s", msg)
927
928
929 def SetupNodeDaemon(opts, cluster_name, node, ssh_port):
930 """Add a node to the cluster.
931
932 This function must be called before the actual opcode, and will ssh
933 to the remote node, copy the needed files, and start ganeti-noded,
934 allowing the master to do the rest via normal rpc calls.
935
936 @param cluster_name: the cluster name
937 @param node: the name of the new node
938 @param ssh_port: the SSH port of the new node
939
940 """
941 data = {
942 constants.NDS_CLUSTER_NAME: cluster_name,
943 constants.NDS_NODE_DAEMON_CERTIFICATE:
944 utils.ReadFile(pathutils.NODED_CERT_FILE),
945 constants.NDS_HMAC:
946 utils.ReadFile(pathutils.CONFD_HMAC_KEY),
947 constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(),
948 constants.NDS_START_NODE_DAEMON: True,
949 constants.NDS_NODE_NAME: node,
950 }
951
952 ssh.RunSshCmdWithStdin(cluster_name, node, pathutils.NODE_DAEMON_SETUP,
953 ssh_port, data,
954 debug=opts.debug, verbose=opts.verbose,
955 use_cluster_key=True, ask_key=opts.ssh_key_check,
956 strict_host_check=opts.ssh_key_check,
957 ensure_version=True)
958
959 _WaitForSshDaemon(node, ssh_port)
960 _WaitForNodeDaemon(node)
961
962
963 def MasterFailover(no_voting=False):
964 """Failover the master node.
965
966 This checks that we are not already the master, and will cause the
967 current master to cease being master, and the non-master to become
968 new master.
969
970 Note: The call to MasterFailover from lib/client/gnt_cluster.py checks that
971 a majority of nodes are healthy and responding before calling this. If this
972 function is called from somewhere else, the caller should also verify that a
973 majority of nodes are healthy.
974
975 @type no_voting: boolean
976 @param no_voting: force the operation without remote nodes agreement
977 (dangerous)
978
979 @returns: the pair of an exit code and warnings to display
980 """
981 sstore = ssconf.SimpleStore()
982
983 old_master, new_master = ssconf.GetMasterAndMyself(sstore)
984 node_names = sstore.GetNodeList()
985 mc_list = sstore.GetMasterCandidates()
986
987 if old_master == new_master:
988 raise errors.OpPrereqError("This commands must be run on the node"
989 " where you want the new master to be."
990 " %s is already the master" %
991 old_master, errors.ECODE_INVAL)
992
993 if new_master not in mc_list:
994 mc_no_master = [name for name in mc_list if name != old_master]
995 raise errors.OpPrereqError("This node is not among the nodes marked"
996 " as master candidates. Only these nodes"
997 " can become masters. Current list of"
998 " master candidates is:\n"
999 "%s" % ("\n".join(mc_no_master)),
1000 errors.ECODE_STATE)
1001
1002 if not no_voting:
1003 vote_list = _GatherMasterVotes(node_names)
1004 if vote_list:
1005 voted_master = vote_list[0][0]
1006 if voted_master != old_master:
1007 raise errors.OpPrereqError("I have a wrong configuration, I believe"
1008 " the master is %s but the other nodes"
1009 " voted %s. Please resync the configuration"
1010 " of this node." %
1011 (old_master, voted_master),
1012 errors.ECODE_STATE)
1013 # end checks
1014
1015 rcode = 0
1016 warnings = []
1017
1018 logging.info("Setting master to %s, old master: %s", new_master, old_master)
1019
1020 try:
1021 # Forcefully start WConfd so that we can access the configuration
1022 result = utils.RunCmd([pathutils.DAEMON_UTIL,
1023 "start", constants.WCONFD, "--force-node",
1024 "--no-voting", "--yes-do-it"])
1025 if result.failed:
1026 raise errors.OpPrereqError("Could not start the configuration daemon,"
1027 " command %s had exitcode %s and error %s" %
1028 (result.cmd, result.exit_code, result.output),
1029 errors.ECODE_NOENT)
1030
1031 # instantiate a real config writer, as we now know we have the
1032 # configuration data
1033 livelock = utils.livelock.LiveLock("bootstrap_failover")
1034 cfg = config.GetConfig(None, livelock, accept_foreign=True)
1035
1036 old_master_node = cfg.GetNodeInfoByName(old_master)
1037 if old_master_node is None:
1038 raise errors.OpPrereqError("Could not find old master node '%s' in"
1039 " cluster configuration." % old_master,
1040 errors.ECODE_NOENT)
1041
1042 cluster_info = cfg.GetClusterInfo()
1043 new_master_node = cfg.GetNodeInfoByName(new_master)
1044 if new_master_node is None:
1045 raise errors.OpPrereqError("Could not find new master node '%s' in"
1046 " cluster configuration." % new_master,
1047 errors.ECODE_NOENT)
1048
1049 cluster_info.master_node = new_master_node.uuid
1050 # this will also regenerate the ssconf files, since we updated the
1051 # cluster info
1052 cfg.Update(cluster_info, logging.error)
1053
1054 # if cfg.Update worked, then it means the old master daemon won't be
1055 # able now to write its own config file (we rely on locking in both
1056 # backend.UploadFile() and ConfigWriter._Write(); hence the next
1057 # step is to kill the old master
1058
1059 logging.info("Stopping the master daemon on node %s", old_master)
1060
1061 runner = rpc.BootstrapRunner()
1062 master_params = cfg.GetMasterNetworkParameters()
1063 master_params.uuid = old_master_node.uuid
1064 ems = cfg.GetUseExternalMipScript()
1065 result = runner.call_node_deactivate_master_ip(old_master,
1066 master_params, ems)
1067
1068 msg = result.fail_msg
1069 if msg:
1070 warning = "Could not disable the master IP: %s" % (msg,)
1071 logging.warning("%s", warning)
1072 warnings.append(warning)
1073
1074 result = runner.call_node_stop_master(old_master)
1075 msg = result.fail_msg
1076 if msg:
1077 warning = ("Could not disable the master role on the old master"
1078 " %s, please disable manually: %s" % (old_master, msg))
1079 logging.error("%s", warning)
1080 warnings.append(warning)
1081 except errors.ConfigurationError, err:
1082 logging.error("Error while trying to set the new master: %s",
1083 str(err))
1084 return 1, warnings
1085 finally:
1086 # stop WConfd again:
1087 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.WCONFD])
1088 if result.failed:
1089 warning = ("Could not stop the configuration daemon,"
1090 " command %s had exitcode %s and error %s"
1091 % (result.cmd, result.exit_code, result.output))
1092 logging.error("%s", warning)
1093 rcode = 1
1094
1095 logging.info("Checking master IP non-reachability...")
1096
1097 master_ip = sstore.GetMasterIP()
1098 total_timeout = 30
1099
1100 # Here we have a phase where no master should be running
1101 def _check_ip(expected):
1102 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT) != expected:
1103 raise utils.RetryAgain()
1104
1105 try:
1106 utils.Retry(_check_ip, (1, 1.5, 5), total_timeout, args=[False])
1107 except utils.RetryTimeout:
1108 warning = ("The master IP is still reachable after %s seconds,"
1109 " continuing but activating the master IP on the current"
1110 " node will probably fail" % total_timeout)
1111 logging.warning("%s", warning)
1112 warnings.append(warning)
1113 rcode = 1
1114
1115 if jstore.CheckDrainFlag():
1116 logging.info("Undraining job queue")
1117 jstore.SetDrainFlag(False)
1118
1119 logging.info("Starting the master daemons on the new master")
1120
1121 result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master,
1122 no_voting)
1123 msg = result.fail_msg
1124 if msg:
1125 logging.error("Could not start the master role on the new master"
1126 " %s, please check: %s", new_master, msg)
1127 rcode = 1
1128
1129 # Finally verify that the new master managed to set up the master IP
1130 # and warn if it didn't.
1131 try:
1132 utils.Retry(_check_ip, (1, 1.5, 5), total_timeout, args=[True])
1133 except utils.RetryTimeout:
1134 warning = ("The master IP did not come up within %s seconds; the"
1135 " cluster should still be working and reachable via %s,"
1136 " but not via the master IP address"
1137 % (total_timeout, new_master))
1138 logging.warning("%s", warning)
1139 warnings.append(warning)
1140 rcode = 1
1141
1142 logging.info("Master failed over from %s to %s", old_master, new_master)
1143 return rcode, warnings
1144
1145
1146 def GetMaster():
1147 """Returns the current master node.
1148
1149 This is a separate function in bootstrap since it's needed by
1150 gnt-cluster, and instead of importing directly ssconf, it's better
1151 to abstract it in bootstrap, where we do use ssconf in other
1152 functions too.
1153
1154 """
1155 sstore = ssconf.SimpleStore()
1156
1157 old_master, _ = ssconf.GetMasterAndMyself(sstore)
1158
1159 return old_master
1160
1161
1162 def _GatherMasterVotes(node_names):
1163 """Check the agreement on who is the master.
1164
1165 This function will return a list of (node, number of votes), ordered
1166 by the number of votes.
1167
1168 Note that the sum of votes is the number of nodes this machine
1169 knows, whereas the number of entries in the list could be different
1170 (if some nodes vote for another master).
1171
1172 @type node_names: list
1173 @param node_names: the list of nodes to query for master info
1174 @rtype: list
1175 @return: list of (node, votes)
1176
1177 """
1178 if not node_names:
1179 # no nodes
1180 return []
1181 results = rpc.BootstrapRunner().call_master_node_name(node_names)
1182 if not isinstance(results, dict):
1183 # this should not happen (unless internal error in rpc)
1184 logging.critical("Can't complete rpc call, aborting master startup")
1185 return [(None, len(node_names))]
1186
1187 votes = {}
1188 for (node_name, nres) in results.iteritems():
1189 msg = nres.fail_msg
1190 if msg:
1191 logging.warning("Error contacting node %s: %s", node_name, msg)
1192 continue
1193 node = nres.payload
1194 if not node:
1195 logging.warning(('Expected a Node, encountered a None. Skipping this'
1196 ' voting result.'))
1197 if node not in votes:
1198 votes[node] = 1
1199 else:
1200 votes[node] += 1
1201
1202 vote_list = votes.items()
1203 vote_list.sort(key=lambda x: x[1], reverse=True)
1204 return vote_list
1205
1206
1207 def MajorityHealthy(ignore_offline_nodes=False):
1208 """Check if the majority of nodes is healthy
1209
1210 Gather master votes from all nodes known to this node;
1211 return True if a strict majority of nodes is reachable and
1212 has some opinion on which node is master. Note that this will
1213 not guarantee any node to win an election but it ensures that
1214 a standard master-failover is still possible.
1215
1216 @return: tuple of (boolean, [str]); the first is if a majority of nodes are
1217 healthy, the second is a list of the node names that are not considered
1218 healthy.
1219 """
1220 if ignore_offline_nodes:
1221 node_names = ssconf.SimpleStore().GetOnlineNodeList()
1222 else:
1223 node_names = ssconf.SimpleStore().GetNodeList()
1224
1225 node_count = len(node_names)
1226 vote_list = _GatherMasterVotes(node_names)
1227
1228 if not vote_list:
1229 logging.warning(('Voting list was None; cannot determine if a majority of '
1230 'nodes are healthy'))
1231 return (False, node_names)
1232
1233 total_votes = sum([count for (node, count) in vote_list if node is not None])
1234 majority_healthy = 2 * total_votes > node_count
1235
1236 # The list of nodes that did not vote is calculated to provide useful
1237 # debugging information to the client.
1238 voting_nodes = [node for (node, _) in vote_list]
1239 nonvoting_nodes = [node for node in node_names if node not in voting_nodes]
1240
1241 logging.info("Total %d nodes, %d votes: %s", node_count, total_votes,
1242 vote_list)
1243
1244 return (majority_healthy, nonvoting_nodes)