Newer
Older
# Copyright (C) 2010 Google Inc.
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
# General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
# 02110-1301, USA.
"""Tool to merge two or more clusters together.
The clusters have to run the same version of Ganeti!
"""
# pylint: disable-msg=C0103
# C0103: Invalid name cluster-merge
import logging
import os
import optparse
import shutil
import sys
import tempfile
from ganeti import cli
from ganeti import config
from ganeti import constants
from ganeti import errors
from ganeti import ssh
from ganeti import utils
_GROUPS_MERGE = "merge"
_GROUPS_RENAME = "rename"
_CLUSTERMERGE_ECID = "clustermerge-ecid"
PAUSE_PERIOD_OPT = cli.cli_option("-p", "--watcher-pause-period", default=1800,
action="store", type="int",
dest="pause_period",
help=("Amount of time in seconds watcher"
" should be suspended from running"))
GROUPS_OPT = cli.cli_option("--groups", default=None, metavar="STRATEGY",
choices=(_GROUPS_MERGE, _GROUPS_RENAME), dest="groups",
help=("How to handle groups that have the"
" same name (One of: %s/%s)" %
(_GROUPS_MERGE, _GROUPS_RENAME)))
@param unflattened_list: A list of unflattened list objects.
@return: A flattened list
flattened_list.append(item)
return flattened_list
class MergerData(object):
"""Container class to hold data used for merger.
"""
def __init__(self, cluster, key_path, nodes, instances, config_path=None):
"""Initialize the container.
@param cluster: The name of the cluster
@param key_path: Path to the ssh private key used for authentication
@param nodes: List of nodes in the merging cluster
@param instances: List of instances running on merging cluster
@param config_path: Path to the merging cluster config
"""
self.cluster = cluster
self.key_path = key_path
self.nodes = nodes
self.instances = instances
self.config_path = config_path
class Merger(object):
"""Handling the merge.
"""
def __init__(self, clusters, pause_period, groups):
"""Initialize object with sane defaults and infos required.
@param clusters: The list of clusters to merge in
@param pause_period: The time watcher shall be disabled for
@param groups: How to handle group conflicts
"""
self.merger_data = []
self.clusters = clusters
self.pause_period = pause_period
self.work_dir = tempfile.mkdtemp(suffix="cluster-merger")
(self.cluster_name, ) = cli.GetClient().QueryConfigValues(["cluster_name"])
self.ssh_runner = ssh.SshRunner(self.cluster_name)
def Setup(self):
"""Sets up our end so we can do the merger.
This method is setting us up as a preparation for the merger.
It makes the initial contact and gathers information needed.
@raise errors.RemoteError: for errors in communication/grabbing
"""
(remote_path, _, _) = ssh.GetUserFiles("root")
if self.cluster_name in self.clusters:
raise errors.CommandError("Cannot merge cluster %s with itself" %
self.cluster_name)
# Fetch remotes private key
for cluster in self.clusters:
result = self._RunCmd(cluster, "cat %s" % remote_path, batch=False,
ask_key=False)
if result.failed:
raise errors.RemoteError("There was an error while grabbing ssh private"
" key from %s. Fail reason: %s; output: %s" %
(cluster, result.fail_reason, result.output))
key_path = utils.PathJoin(self.work_dir, cluster)
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
utils.WriteFile(key_path, mode=0600, data=result.stdout)
result = self._RunCmd(cluster, "gnt-node list -o name --no-header",
private_key=key_path)
if result.failed:
raise errors.RemoteError("Unable to retrieve list of nodes from %s."
" Fail reason: %s; output: %s" %
(cluster, result.fail_reason, result.output))
nodes = result.stdout.splitlines()
result = self._RunCmd(cluster, "gnt-instance list -o name --no-header",
private_key=key_path)
if result.failed:
raise errors.RemoteError("Unable to retrieve list of instances from"
" %s. Fail reason: %s; output: %s" %
(cluster, result.fail_reason, result.output))
instances = result.stdout.splitlines()
self.merger_data.append(MergerData(cluster, key_path, nodes, instances))
def _PrepareAuthorizedKeys(self):
"""Prepare the authorized_keys on every merging node.
This method add our public key to remotes authorized_key for further
communication.
"""
(_, pub_key_file, auth_keys) = ssh.GetUserFiles("root")
pub_key = utils.ReadFile(pub_key_file)
for data in self.merger_data:
for node in data.nodes:
result = self._RunCmd(node, ("cat >> %s << '!EOF.'\n%s!EOF.\n" %
(auth_keys, pub_key)),
private_key=data.key_path)
if result.failed:
raise errors.RemoteError("Unable to add our public key to %s in %s."
" Fail reason: %s; output: %s" %
(node, data.cluster, result.fail_reason,
result.output))
def _RunCmd(self, hostname, command, user="root", use_cluster_key=False,
strict_host_check=False, private_key=None, batch=True,
ask_key=False):
"""Wrapping SshRunner.Run with default parameters.
For explanation of parameters see L{ganeti.ssh.SshRunner.Run}.
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
"""
return self.ssh_runner.Run(hostname=hostname, command=command, user=user,
use_cluster_key=use_cluster_key,
strict_host_check=strict_host_check,
private_key=private_key, batch=batch,
ask_key=ask_key)
def _StopMergingInstances(self):
"""Stop instances on merging clusters.
"""
for cluster in self.clusters:
result = self._RunCmd(cluster, "gnt-instance shutdown --all"
" --force-multiple")
if result.failed:
raise errors.RemoteError("Unable to stop instances on %s."
" Fail reason: %s; output: %s" %
(cluster, result.fail_reason, result.output))
def _DisableWatcher(self):
"""Disable watch on all merging clusters, including ourself.
"""
for cluster in ["localhost"] + self.clusters:
result = self._RunCmd(cluster, "gnt-cluster watcher pause %d" %
self.pause_period)
if result.failed:
raise errors.RemoteError("Unable to pause watcher on %s."
" Fail reason: %s; output: %s" %
(cluster, result.fail_reason, result.output))
def _StopDaemons(self):
"""Stop all daemons on merging nodes.
"""
cmd = "%s stop-all" % constants.DAEMON_UTIL
for data in self.merger_data:
for node in data.nodes:
result = self._RunCmd(node, cmd)
if result.failed:
raise errors.RemoteError("Unable to stop daemons on %s."
" Fail reason: %s; output: %s." %
(node, result.fail_reason, result.output))
def _FetchRemoteConfig(self):
"""Fetches and stores remote cluster config from the master.
This step is needed before we can merge the config.
"""
for data in self.merger_data:
result = self._RunCmd(data.cluster, "cat %s" %
constants.CLUSTER_CONF_FILE)
if result.failed:
raise errors.RemoteError("Unable to retrieve remote config on %s."
" Fail reason: %s; output %s" %
(data.cluster, result.fail_reason,
result.output))
data.config_path = utils.PathJoin(self.work_dir, "%s_config.data" %
data.cluster)
utils.WriteFile(data.config_path, data=result.stdout)
# R0201: Method could be a function
def _KillMasterDaemon(self): # pylint: disable-msg=R0201
"""Kills the local master daemon.
@raise errors.CommandError: If unable to kill
"""
result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
if result.failed:
raise errors.CommandError("Unable to stop master daemons."
" Fail reason: %s; output: %s" %
(result.fail_reason, result.output))
def _MergeConfig(self):
"""Merges all foreign config into our own config.
"""
my_config = config.ConfigWriter(offline=True)
fake_ec_id = 0 # Needs to be uniq over the whole config merge
for data in self.merger_data:
other_config = config.ConfigWriter(data.config_path, accept_foreign=True)
self._MergeNodeGroups(my_config, other_config)
for node in other_config.GetNodeList():
node_info = other_config.GetNodeInfo(node)
my_config.AddNode(node_info, _CLUSTERMERGE_ECID + str(fake_ec_id))
fake_ec_id += 1
for instance in other_config.GetInstanceList():
instance_info = other_config.GetInstanceInfo(instance)
# Update the DRBD port assignments
# This is a little bit hackish
for dsk in instance_info.disks:
if dsk.dev_type in constants.LDS_DRBD:
port = my_config.AllocatePort()
logical_id = list(dsk.logical_id)
logical_id[2] = port
dsk.logical_id = tuple(logical_id)
physical_id = list(dsk.physical_id)
physical_id[1] = physical_id[3] = port
dsk.physical_id = tuple(physical_id)
my_config.AddInstance(instance_info,
_CLUSTERMERGE_ECID + str(fake_ec_id))
# R0201: Method could be a function
def _MergeNodeGroups(self, my_config, other_config):
"""Adds foreign node groups
ConfigWriter.AddNodeGroup takes care of making sure there are no conflicts.
"""
# pylint: disable-msg=R0201
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
logging.info("Node group conflict strategy: %s" % self.groups)
my_grps = my_config.GetAllNodeGroupsInfo().values()
other_grps = other_config.GetAllNodeGroupsInfo().values()
# Check for node group naming conflicts:
conflicts = []
for other_grp in other_grps:
for my_grp in my_grps:
if other_grp.name == my_grp.name:
conflicts.append(other_grp)
if conflicts:
conflict_names = utils.CommaJoin([g.name for g in conflicts])
logging.info("Node groups in both local and remote cluster: %s" %
conflict_names)
# User hasn't specified how to handle conflicts
if not self.groups:
raise errors.CommandError("The following node group(s) are in both"
" clusters, and no merge strategy has been"
" supplied (see the --groups option): %s" %
conflict_names)
# User wants to rename conflicts
if self.groups == _GROUPS_RENAME:
for grp in conflicts:
new_name = "%s-%s" % (grp.name, other_config.GetClusterName())
logging.info("Renaming remote node group from %s to %s"
" to resolve conflict" % (grp.name, new_name))
grp.name = new_name
for grp in other_grps:
#TODO: handle node group conflicts
my_config.AddNodeGroup(grp, _CLUSTERMERGE_ECID)
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
# R0201: Method could be a function
def _StartMasterDaemon(self, no_vote=False): # pylint: disable-msg=R0201
"""Starts the local master daemon.
@param no_vote: Should the masterd started without voting? default: False
@raise errors.CommandError: If unable to start daemon.
"""
env = {}
if no_vote:
env["EXTRA_MASTERD_ARGS"] = "--no-voting --yes-do-it"
result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
if result.failed:
raise errors.CommandError("Couldn't start ganeti master."
" Fail reason: %s; output: %s" %
(result.fail_reason, result.output))
def _ReaddMergedNodesAndRedist(self):
"""Readds all merging nodes and make sure their config is up-to-date.
@raise errors.CommandError: If anything fails.
"""
for data in self.merger_data:
for node in data.nodes:
result = utils.RunCmd(["gnt-node", "add", "--readd",
"--no-ssh-key-check", "--force-join", node])
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
if result.failed:
raise errors.CommandError("Couldn't readd node %s. Fail reason: %s;"
" output: %s" % (node, result.fail_reason,
result.output))
result = utils.RunCmd(["gnt-cluster", "redist-conf"])
if result.failed:
raise errors.CommandError("Redistribution failed. Fail reason: %s;"
" output: %s" % (result.fail_reason,
result.output))
# R0201: Method could be a function
def _StartupAllInstances(self): # pylint: disable-msg=R0201
"""Starts up all instances (locally).
@raise errors.CommandError: If unable to start clusters
"""
result = utils.RunCmd(["gnt-instance", "startup", "--all",
"--force-multiple"])
if result.failed:
raise errors.CommandError("Unable to start all instances."
" Fail reason: %s; output: %s" %
(result.fail_reason, result.output))
# R0201: Method could be a function
def _VerifyCluster(self): # pylint: disable-msg=R0201
"""Runs gnt-cluster verify to verify the health.
@raise errors.ProgrammError: If cluster fails on verification
"""
result = utils.RunCmd(["gnt-cluster", "verify"])
if result.failed:
raise errors.CommandError("Verification of cluster failed."
" Fail reason: %s; output: %s" %
(result.fail_reason, result.output))
def Merge(self):
"""Does the actual merge.
It runs all the steps in the right order and updates the user about steps
taken. Also it keeps track of rollback_steps to undo everything.
"""
rbsteps = []
try:
logging.info("Pre cluster verification")
self._VerifyCluster()
logging.info("Prepare authorized_keys")
rbsteps.append("Remove our key from authorized_keys on nodes:"
" %(nodes)s")
self._PrepareAuthorizedKeys()
rbsteps.append("Start all instances again on the merging"
" clusters: %(clusters)s")
logging.info("Stopping merging instances (takes a while)")
self._StopMergingInstances()
logging.info("Disable watcher")
self._DisableWatcher()
logging.info("Stop daemons on merging nodes")
self._StopDaemons()
logging.info("Merging config")
self._FetchRemoteConfig()
logging.info("Stopping master daemon")
self._KillMasterDaemon()
rbsteps.append("Restore %s from another master candidate"
" and restart master daemon" %
constants.CLUSTER_CONF_FILE)
self._MergeConfig()
self._StartMasterDaemon(no_vote=True)
# Point of no return, delete rbsteps
del rbsteps[:]
logging.warning("We are at the point of no return. Merge can not easily"
" be undone after this point.")
logging.info("Readd nodes")
self._ReaddMergedNodesAndRedist()
logging.info("Merge done, restart master daemon normally")
self._KillMasterDaemon()
self._StartMasterDaemon()
logging.info("Starting instances again")
self._StartupAllInstances()
logging.info("Post cluster verification")
self._VerifyCluster()
except errors.GenericError, e:
logging.exception(e)
if rbsteps:
nodes = Flatten([data.nodes for data in self.merger_data])
info = {
"clusters": self.clusters,
"nodes": nodes,
}
logging.critical("In order to rollback do the following:")
for step in rbsteps:
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
else:
logging.critical("Nothing to rollback.")
# TODO: Keep track of steps done for a flawless resume?
def Cleanup(self):
"""Clean up our environment.
This cleans up remote private keys and configs and after that
deletes the temporary directory.
"""
shutil.rmtree(self.work_dir)
def SetupLogging(options):
"""Setting up logging infrastructure.
@param options: Parsed command line options
"""
formatter = logging.Formatter("%(asctime)s: %(levelname)s %(message)s")
stderr_handler = logging.StreamHandler()
stderr_handler.setFormatter(formatter)
if options.debug:
stderr_handler.setLevel(logging.NOTSET)
elif options.verbose:
stderr_handler.setLevel(logging.INFO)
else:
stderr_handler.setLevel(logging.ERROR)
root_logger = logging.getLogger("")
root_logger.setLevel(logging.NOTSET)
root_logger.addHandler(stderr_handler)
def main():
"""Main routine.
"""
program = os.path.basename(sys.argv[0])
parser = optparse.OptionParser(usage=("%%prog [--debug|--verbose]"
" [--watcher-pause-period SECONDS]"
" [--groups [%s|%s]]"
" <cluster> [<cluster...>]" %
(_GROUPS_MERGE, _GROUPS_RENAME)),
prog=program)
parser.add_option(cli.DEBUG_OPT)
parser.add_option(cli.VERBOSE_OPT)
parser.add_option(PAUSE_PERIOD_OPT)
(options, args) = parser.parse_args()
SetupLogging(options)
if not args:
parser.error("No clusters specified")
cluster_merger = Merger(utils.UniqueSequence(args), options.pause_period,
options.groups)
try:
try:
cluster_merger.Setup()
cluster_merger.Merge()
except errors.GenericError, e:
logging.exception(e)
return constants.EXIT_FAILURE
finally:
cluster_merger.Cleanup()
return constants.EXIT_SUCCESS
if __name__ == "__main__":
sys.exit(main())