root / lib / cmdlib.py @ 7767bbf5
History | View | Annotate | Download (145.6 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007 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 |
"""Module implementing the master-side code."""
|
23 |
|
24 |
# pylint: disable-msg=W0613,W0201
|
25 |
|
26 |
import os |
27 |
import os.path |
28 |
import sha |
29 |
import time |
30 |
import tempfile |
31 |
import re |
32 |
import platform |
33 |
|
34 |
from ganeti import rpc |
35 |
from ganeti import ssh |
36 |
from ganeti import logger |
37 |
from ganeti import utils |
38 |
from ganeti import errors |
39 |
from ganeti import hypervisor |
40 |
from ganeti import config |
41 |
from ganeti import constants |
42 |
from ganeti import objects |
43 |
from ganeti import opcodes |
44 |
from ganeti import ssconf |
45 |
|
46 |
|
47 |
class LogicalUnit(object): |
48 |
"""Logical Unit base class.
|
49 |
|
50 |
Subclasses must follow these rules:
|
51 |
- implement CheckPrereq which also fills in the opcode instance
|
52 |
with all the fields (even if as None)
|
53 |
- implement Exec
|
54 |
- implement BuildHooksEnv
|
55 |
- redefine HPATH and HTYPE
|
56 |
- optionally redefine their run requirements (REQ_CLUSTER,
|
57 |
REQ_MASTER); note that all commands require root permissions
|
58 |
|
59 |
"""
|
60 |
HPATH = None
|
61 |
HTYPE = None
|
62 |
_OP_REQP = [] |
63 |
REQ_CLUSTER = True
|
64 |
REQ_MASTER = True
|
65 |
|
66 |
def __init__(self, processor, op, cfg, sstore): |
67 |
"""Constructor for LogicalUnit.
|
68 |
|
69 |
This needs to be overriden in derived classes in order to check op
|
70 |
validity.
|
71 |
|
72 |
"""
|
73 |
self.proc = processor
|
74 |
self.op = op
|
75 |
self.cfg = cfg
|
76 |
self.sstore = sstore
|
77 |
self.__ssh = None |
78 |
|
79 |
for attr_name in self._OP_REQP: |
80 |
attr_val = getattr(op, attr_name, None) |
81 |
if attr_val is None: |
82 |
raise errors.OpPrereqError("Required parameter '%s' missing" % |
83 |
attr_name) |
84 |
if self.REQ_CLUSTER: |
85 |
if not cfg.IsCluster(): |
86 |
raise errors.OpPrereqError("Cluster not initialized yet," |
87 |
" use 'gnt-cluster init' first.")
|
88 |
if self.REQ_MASTER: |
89 |
master = sstore.GetMasterNode() |
90 |
if master != utils.HostInfo().name:
|
91 |
raise errors.OpPrereqError("Commands must be run on the master" |
92 |
" node %s" % master)
|
93 |
|
94 |
def __GetSSH(self): |
95 |
"""Returns the SshRunner object
|
96 |
|
97 |
"""
|
98 |
if not self.__ssh: |
99 |
self.__ssh = ssh.SshRunner(self.sstore) |
100 |
return self.__ssh |
101 |
|
102 |
ssh = property(fget=__GetSSH)
|
103 |
|
104 |
def CheckPrereq(self): |
105 |
"""Check prerequisites for this LU.
|
106 |
|
107 |
This method should check that the prerequisites for the execution
|
108 |
of this LU are fulfilled. It can do internode communication, but
|
109 |
it should be idempotent - no cluster or system changes are
|
110 |
allowed.
|
111 |
|
112 |
The method should raise errors.OpPrereqError in case something is
|
113 |
not fulfilled. Its return value is ignored.
|
114 |
|
115 |
This method should also update all the parameters of the opcode to
|
116 |
their canonical form; e.g. a short node name must be fully
|
117 |
expanded after this method has successfully completed (so that
|
118 |
hooks, logging, etc. work correctly).
|
119 |
|
120 |
"""
|
121 |
raise NotImplementedError |
122 |
|
123 |
def Exec(self, feedback_fn): |
124 |
"""Execute the LU.
|
125 |
|
126 |
This method should implement the actual work. It should raise
|
127 |
errors.OpExecError for failures that are somewhat dealt with in
|
128 |
code, or expected.
|
129 |
|
130 |
"""
|
131 |
raise NotImplementedError |
132 |
|
133 |
def BuildHooksEnv(self): |
134 |
"""Build hooks environment for this LU.
|
135 |
|
136 |
This method should return a three-node tuple consisting of: a dict
|
137 |
containing the environment that will be used for running the
|
138 |
specific hook for this LU, a list of node names on which the hook
|
139 |
should run before the execution, and a list of node names on which
|
140 |
the hook should run after the execution.
|
141 |
|
142 |
The keys of the dict must not have 'GANETI_' prefixed as this will
|
143 |
be handled in the hooks runner. Also note additional keys will be
|
144 |
added by the hooks runner. If the LU doesn't define any
|
145 |
environment, an empty dict (and not None) should be returned.
|
146 |
|
147 |
As for the node lists, the master should not be included in the
|
148 |
them, as it will be added by the hooks runner in case this LU
|
149 |
requires a cluster to run on (otherwise we don't have a node
|
150 |
list). No nodes should be returned as an empty list (and not
|
151 |
None).
|
152 |
|
153 |
Note that if the HPATH for a LU class is None, this function will
|
154 |
not be called.
|
155 |
|
156 |
"""
|
157 |
raise NotImplementedError |
158 |
|
159 |
|
160 |
class NoHooksLU(LogicalUnit): |
161 |
"""Simple LU which runs no hooks.
|
162 |
|
163 |
This LU is intended as a parent for other LogicalUnits which will
|
164 |
run no hooks, in order to reduce duplicate code.
|
165 |
|
166 |
"""
|
167 |
HPATH = None
|
168 |
HTYPE = None
|
169 |
|
170 |
def BuildHooksEnv(self): |
171 |
"""Build hooks env.
|
172 |
|
173 |
This is a no-op, since we don't run hooks.
|
174 |
|
175 |
"""
|
176 |
return {}, [], []
|
177 |
|
178 |
|
179 |
def _AddHostToEtcHosts(hostname): |
180 |
"""Wrapper around utils.SetEtcHostsEntry.
|
181 |
|
182 |
"""
|
183 |
hi = utils.HostInfo(name=hostname) |
184 |
utils.SetEtcHostsEntry(constants.ETC_HOSTS, hi.ip, hi.name, [hi.ShortName()]) |
185 |
|
186 |
|
187 |
def _RemoveHostFromEtcHosts(hostname): |
188 |
"""Wrapper around utils.RemoveEtcHostsEntry.
|
189 |
|
190 |
"""
|
191 |
hi = utils.HostInfo(name=hostname) |
192 |
utils.RemoveEtcHostsEntry(constants.ETC_HOSTS, hi.name) |
193 |
utils.RemoveEtcHostsEntry(constants.ETC_HOSTS, hi.ShortName()) |
194 |
|
195 |
|
196 |
def _GetWantedNodes(lu, nodes): |
197 |
"""Returns list of checked and expanded node names.
|
198 |
|
199 |
Args:
|
200 |
nodes: List of nodes (strings) or None for all
|
201 |
|
202 |
"""
|
203 |
if not isinstance(nodes, list): |
204 |
raise errors.OpPrereqError("Invalid argument type 'nodes'") |
205 |
|
206 |
if nodes:
|
207 |
wanted = [] |
208 |
|
209 |
for name in nodes: |
210 |
node = lu.cfg.ExpandNodeName(name) |
211 |
if node is None: |
212 |
raise errors.OpPrereqError("No such node name '%s'" % name) |
213 |
wanted.append(node) |
214 |
|
215 |
else:
|
216 |
wanted = lu.cfg.GetNodeList() |
217 |
return utils.NiceSort(wanted)
|
218 |
|
219 |
|
220 |
def _GetWantedInstances(lu, instances): |
221 |
"""Returns list of checked and expanded instance names.
|
222 |
|
223 |
Args:
|
224 |
instances: List of instances (strings) or None for all
|
225 |
|
226 |
"""
|
227 |
if not isinstance(instances, list): |
228 |
raise errors.OpPrereqError("Invalid argument type 'instances'") |
229 |
|
230 |
if instances:
|
231 |
wanted = [] |
232 |
|
233 |
for name in instances: |
234 |
instance = lu.cfg.ExpandInstanceName(name) |
235 |
if instance is None: |
236 |
raise errors.OpPrereqError("No such instance name '%s'" % name) |
237 |
wanted.append(instance) |
238 |
|
239 |
else:
|
240 |
wanted = lu.cfg.GetInstanceList() |
241 |
return utils.NiceSort(wanted)
|
242 |
|
243 |
|
244 |
def _CheckOutputFields(static, dynamic, selected): |
245 |
"""Checks whether all selected fields are valid.
|
246 |
|
247 |
Args:
|
248 |
static: Static fields
|
249 |
dynamic: Dynamic fields
|
250 |
|
251 |
"""
|
252 |
static_fields = frozenset(static)
|
253 |
dynamic_fields = frozenset(dynamic)
|
254 |
|
255 |
all_fields = static_fields | dynamic_fields |
256 |
|
257 |
if not all_fields.issuperset(selected): |
258 |
raise errors.OpPrereqError("Unknown output fields selected: %s" |
259 |
% ",".join(frozenset(selected). |
260 |
difference(all_fields))) |
261 |
|
262 |
|
263 |
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status, |
264 |
memory, vcpus, nics): |
265 |
"""Builds instance related env variables for hooks from single variables.
|
266 |
|
267 |
Args:
|
268 |
secondary_nodes: List of secondary nodes as strings
|
269 |
"""
|
270 |
env = { |
271 |
"OP_TARGET": name,
|
272 |
"INSTANCE_NAME": name,
|
273 |
"INSTANCE_PRIMARY": primary_node,
|
274 |
"INSTANCE_SECONDARIES": " ".join(secondary_nodes), |
275 |
"INSTANCE_OS_TYPE": os_type,
|
276 |
"INSTANCE_STATUS": status,
|
277 |
"INSTANCE_MEMORY": memory,
|
278 |
"INSTANCE_VCPUS": vcpus,
|
279 |
} |
280 |
|
281 |
if nics:
|
282 |
nic_count = len(nics)
|
283 |
for idx, (ip, bridge, mac) in enumerate(nics): |
284 |
if ip is None: |
285 |
ip = ""
|
286 |
env["INSTANCE_NIC%d_IP" % idx] = ip
|
287 |
env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
|
288 |
env["INSTANCE_NIC%d_HWADDR" % idx] = mac
|
289 |
else:
|
290 |
nic_count = 0
|
291 |
|
292 |
env["INSTANCE_NIC_COUNT"] = nic_count
|
293 |
|
294 |
return env
|
295 |
|
296 |
|
297 |
def _BuildInstanceHookEnvByObject(instance, override=None): |
298 |
"""Builds instance related env variables for hooks from an object.
|
299 |
|
300 |
Args:
|
301 |
instance: objects.Instance object of instance
|
302 |
override: dict of values to override
|
303 |
"""
|
304 |
args = { |
305 |
'name': instance.name,
|
306 |
'primary_node': instance.primary_node,
|
307 |
'secondary_nodes': instance.secondary_nodes,
|
308 |
'os_type': instance.os,
|
309 |
'status': instance.os,
|
310 |
'memory': instance.memory,
|
311 |
'vcpus': instance.vcpus,
|
312 |
'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics], |
313 |
} |
314 |
if override:
|
315 |
args.update(override) |
316 |
return _BuildInstanceHookEnv(**args)
|
317 |
|
318 |
|
319 |
def _HasValidVG(vglist, vgname): |
320 |
"""Checks if the volume group list is valid.
|
321 |
|
322 |
A non-None return value means there's an error, and the return value
|
323 |
is the error message.
|
324 |
|
325 |
"""
|
326 |
vgsize = vglist.get(vgname, None)
|
327 |
if vgsize is None: |
328 |
return "volume group '%s' missing" % vgname |
329 |
elif vgsize < 20480: |
330 |
return ("volume group '%s' too small (20480MiB required, %dMib found)" % |
331 |
(vgname, vgsize)) |
332 |
return None |
333 |
|
334 |
|
335 |
def _InitSSHSetup(node): |
336 |
"""Setup the SSH configuration for the cluster.
|
337 |
|
338 |
|
339 |
This generates a dsa keypair for root, adds the pub key to the
|
340 |
permitted hosts and adds the hostkey to its own known hosts.
|
341 |
|
342 |
Args:
|
343 |
node: the name of this host as a fqdn
|
344 |
|
345 |
"""
|
346 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) |
347 |
|
348 |
for name in priv_key, pub_key: |
349 |
if os.path.exists(name):
|
350 |
utils.CreateBackup(name) |
351 |
utils.RemoveFile(name) |
352 |
|
353 |
result = utils.RunCmd(["ssh-keygen", "-t", "dsa", |
354 |
"-f", priv_key,
|
355 |
"-q", "-N", ""]) |
356 |
if result.failed:
|
357 |
raise errors.OpExecError("Could not generate ssh keypair, error %s" % |
358 |
result.output) |
359 |
|
360 |
f = open(pub_key, 'r') |
361 |
try:
|
362 |
utils.AddAuthorizedKey(auth_keys, f.read(8192))
|
363 |
finally:
|
364 |
f.close() |
365 |
|
366 |
|
367 |
def _InitGanetiServerSetup(ss): |
368 |
"""Setup the necessary configuration for the initial node daemon.
|
369 |
|
370 |
This creates the nodepass file containing the shared password for
|
371 |
the cluster and also generates the SSL certificate.
|
372 |
|
373 |
"""
|
374 |
# Create pseudo random password
|
375 |
randpass = sha.new(os.urandom(64)).hexdigest()
|
376 |
# and write it into sstore
|
377 |
ss.SetKey(ss.SS_NODED_PASS, randpass) |
378 |
|
379 |
result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024", |
380 |
"-days", str(365*5), "-nodes", "-x509", |
381 |
"-keyout", constants.SSL_CERT_FILE,
|
382 |
"-out", constants.SSL_CERT_FILE, "-batch"]) |
383 |
if result.failed:
|
384 |
raise errors.OpExecError("could not generate server ssl cert, command" |
385 |
" %s had exitcode %s and error message %s" %
|
386 |
(result.cmd, result.exit_code, result.output)) |
387 |
|
388 |
os.chmod(constants.SSL_CERT_FILE, 0400)
|
389 |
|
390 |
result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
|
391 |
|
392 |
if result.failed:
|
393 |
raise errors.OpExecError("Could not start the node daemon, command %s" |
394 |
" had exitcode %s and error %s" %
|
395 |
(result.cmd, result.exit_code, result.output)) |
396 |
|
397 |
|
398 |
def _CheckInstanceBridgesExist(instance): |
399 |
"""Check that the brigdes needed by an instance exist.
|
400 |
|
401 |
"""
|
402 |
# check bridges existance
|
403 |
brlist = [nic.bridge for nic in instance.nics] |
404 |
if not rpc.call_bridges_exist(instance.primary_node, brlist): |
405 |
raise errors.OpPrereqError("one or more target bridges %s does not" |
406 |
" exist on destination node '%s'" %
|
407 |
(brlist, instance.primary_node)) |
408 |
|
409 |
|
410 |
class LUInitCluster(LogicalUnit): |
411 |
"""Initialise the cluster.
|
412 |
|
413 |
"""
|
414 |
HPATH = "cluster-init"
|
415 |
HTYPE = constants.HTYPE_CLUSTER |
416 |
_OP_REQP = ["cluster_name", "hypervisor_type", "mac_prefix", |
417 |
"def_bridge", "master_netdev", "file_storage_dir"] |
418 |
REQ_CLUSTER = False
|
419 |
|
420 |
def BuildHooksEnv(self): |
421 |
"""Build hooks env.
|
422 |
|
423 |
Notes: Since we don't require a cluster, we must manually add
|
424 |
ourselves in the post-run node list.
|
425 |
|
426 |
"""
|
427 |
env = {"OP_TARGET": self.op.cluster_name} |
428 |
return env, [], [self.hostname.name] |
429 |
|
430 |
def CheckPrereq(self): |
431 |
"""Verify that the passed name is a valid one.
|
432 |
|
433 |
"""
|
434 |
if config.ConfigWriter.IsCluster():
|
435 |
raise errors.OpPrereqError("Cluster is already initialised") |
436 |
|
437 |
if self.op.hypervisor_type == constants.HT_XEN_HVM31: |
438 |
if not os.path.exists(constants.VNC_PASSWORD_FILE): |
439 |
raise errors.OpPrereqError("Please prepare the cluster VNC" |
440 |
"password file %s" %
|
441 |
constants.VNC_PASSWORD_FILE) |
442 |
|
443 |
self.hostname = hostname = utils.HostInfo()
|
444 |
|
445 |
if hostname.ip.startswith("127."): |
446 |
raise errors.OpPrereqError("This host's IP resolves to the private" |
447 |
" range (%s). Please fix DNS or %s." %
|
448 |
(hostname.ip, constants.ETC_HOSTS)) |
449 |
|
450 |
if not utils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT, |
451 |
source=constants.LOCALHOST_IP_ADDRESS): |
452 |
raise errors.OpPrereqError("Inconsistency: this host's name resolves" |
453 |
" to %s,\nbut this ip address does not"
|
454 |
" belong to this host."
|
455 |
" Aborting." % hostname.ip)
|
456 |
|
457 |
self.clustername = clustername = utils.HostInfo(self.op.cluster_name) |
458 |
|
459 |
if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
|
460 |
timeout=5):
|
461 |
raise errors.OpPrereqError("Cluster IP already active. Aborting.") |
462 |
|
463 |
secondary_ip = getattr(self.op, "secondary_ip", None) |
464 |
if secondary_ip and not utils.IsValidIP(secondary_ip): |
465 |
raise errors.OpPrereqError("Invalid secondary ip given") |
466 |
if (secondary_ip and |
467 |
secondary_ip != hostname.ip and
|
468 |
(not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
|
469 |
source=constants.LOCALHOST_IP_ADDRESS))): |
470 |
raise errors.OpPrereqError("You gave %s as secondary IP," |
471 |
" but it does not belong to this host." %
|
472 |
secondary_ip) |
473 |
self.secondary_ip = secondary_ip
|
474 |
|
475 |
if not hasattr(self.op, "vg_name"): |
476 |
self.op.vg_name = None |
477 |
# if vg_name not None, checks if volume group is valid
|
478 |
if self.op.vg_name: |
479 |
vgstatus = _HasValidVG(utils.ListVolumeGroups(), self.op.vg_name)
|
480 |
if vgstatus:
|
481 |
raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" |
482 |
" you are not using lvm" % vgstatus)
|
483 |
|
484 |
self.op.file_storage_dir = os.path.normpath(self.op.file_storage_dir) |
485 |
|
486 |
if not os.path.isabs(self.op.file_storage_dir): |
487 |
raise errors.OpPrereqError("The file storage directory you have is" |
488 |
" not an absolute path.")
|
489 |
|
490 |
if not os.path.exists(self.op.file_storage_dir): |
491 |
try:
|
492 |
os.makedirs(self.op.file_storage_dir, 0750) |
493 |
except OSError, err: |
494 |
raise errors.OpPrereqError("Cannot create file storage directory" |
495 |
" '%s': %s" %
|
496 |
(self.op.file_storage_dir, err))
|
497 |
|
498 |
if not os.path.isdir(self.op.file_storage_dir): |
499 |
raise errors.OpPrereqError("The file storage directory '%s' is not" |
500 |
" a directory." % self.op.file_storage_dir) |
501 |
|
502 |
if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", |
503 |
self.op.mac_prefix):
|
504 |
raise errors.OpPrereqError("Invalid mac prefix given '%s'" % |
505 |
self.op.mac_prefix)
|
506 |
|
507 |
if self.op.hypervisor_type not in constants.HYPER_TYPES: |
508 |
raise errors.OpPrereqError("Invalid hypervisor type given '%s'" % |
509 |
self.op.hypervisor_type)
|
510 |
|
511 |
result = utils.RunCmd(["ip", "link", "show", "dev", self.op.master_netdev]) |
512 |
if result.failed:
|
513 |
raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % |
514 |
(self.op.master_netdev,
|
515 |
result.output.strip())) |
516 |
|
517 |
if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and |
518 |
os.access(constants.NODE_INITD_SCRIPT, os.X_OK)): |
519 |
raise errors.OpPrereqError("Init.d script '%s' missing or not" |
520 |
" executable." % constants.NODE_INITD_SCRIPT)
|
521 |
|
522 |
def Exec(self, feedback_fn): |
523 |
"""Initialize the cluster.
|
524 |
|
525 |
"""
|
526 |
clustername = self.clustername
|
527 |
hostname = self.hostname
|
528 |
|
529 |
# set up the simple store
|
530 |
self.sstore = ss = ssconf.SimpleStore()
|
531 |
ss.SetKey(ss.SS_HYPERVISOR, self.op.hypervisor_type)
|
532 |
ss.SetKey(ss.SS_MASTER_NODE, hostname.name) |
533 |
ss.SetKey(ss.SS_MASTER_IP, clustername.ip) |
534 |
ss.SetKey(ss.SS_MASTER_NETDEV, self.op.master_netdev)
|
535 |
ss.SetKey(ss.SS_CLUSTER_NAME, clustername.name) |
536 |
ss.SetKey(ss.SS_FILE_STORAGE_DIR, self.op.file_storage_dir)
|
537 |
|
538 |
# set up the inter-node password and certificate
|
539 |
_InitGanetiServerSetup(ss) |
540 |
|
541 |
# start the master ip
|
542 |
rpc.call_node_start_master(hostname.name) |
543 |
|
544 |
# set up ssh config and /etc/hosts
|
545 |
f = open(constants.SSH_HOST_RSA_PUB, 'r') |
546 |
try:
|
547 |
sshline = f.read() |
548 |
finally:
|
549 |
f.close() |
550 |
sshkey = sshline.split(" ")[1] |
551 |
|
552 |
_AddHostToEtcHosts(hostname.name) |
553 |
_InitSSHSetup(hostname.name) |
554 |
|
555 |
# init of cluster config file
|
556 |
self.cfg = cfgw = config.ConfigWriter()
|
557 |
cfgw.InitConfig(hostname.name, hostname.ip, self.secondary_ip,
|
558 |
sshkey, self.op.mac_prefix,
|
559 |
self.op.vg_name, self.op.def_bridge) |
560 |
|
561 |
ssh.WriteKnownHostsFile(cfgw, ss, constants.SSH_KNOWN_HOSTS_FILE) |
562 |
|
563 |
|
564 |
class LUDestroyCluster(NoHooksLU): |
565 |
"""Logical unit for destroying the cluster.
|
566 |
|
567 |
"""
|
568 |
_OP_REQP = [] |
569 |
|
570 |
def CheckPrereq(self): |
571 |
"""Check prerequisites.
|
572 |
|
573 |
This checks whether the cluster is empty.
|
574 |
|
575 |
Any errors are signalled by raising errors.OpPrereqError.
|
576 |
|
577 |
"""
|
578 |
master = self.sstore.GetMasterNode()
|
579 |
|
580 |
nodelist = self.cfg.GetNodeList()
|
581 |
if len(nodelist) != 1 or nodelist[0] != master: |
582 |
raise errors.OpPrereqError("There are still %d node(s) in" |
583 |
" this cluster." % (len(nodelist) - 1)) |
584 |
instancelist = self.cfg.GetInstanceList()
|
585 |
if instancelist:
|
586 |
raise errors.OpPrereqError("There are still %d instance(s) in" |
587 |
" this cluster." % len(instancelist)) |
588 |
|
589 |
def Exec(self, feedback_fn): |
590 |
"""Destroys the cluster.
|
591 |
|
592 |
"""
|
593 |
master = self.sstore.GetMasterNode()
|
594 |
if not rpc.call_node_stop_master(master): |
595 |
raise errors.OpExecError("Could not disable the master role") |
596 |
priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) |
597 |
utils.CreateBackup(priv_key) |
598 |
utils.CreateBackup(pub_key) |
599 |
rpc.call_node_leave_cluster(master) |
600 |
|
601 |
|
602 |
class LUVerifyCluster(NoHooksLU): |
603 |
"""Verifies the cluster status.
|
604 |
|
605 |
"""
|
606 |
_OP_REQP = [] |
607 |
|
608 |
def _VerifyNode(self, node, file_list, local_cksum, vglist, node_result, |
609 |
remote_version, feedback_fn): |
610 |
"""Run multiple tests against a node.
|
611 |
|
612 |
Test list:
|
613 |
- compares ganeti version
|
614 |
- checks vg existance and size > 20G
|
615 |
- checks config file checksum
|
616 |
- checks ssh to other nodes
|
617 |
|
618 |
Args:
|
619 |
node: name of the node to check
|
620 |
file_list: required list of files
|
621 |
local_cksum: dictionary of local files and their checksums
|
622 |
|
623 |
"""
|
624 |
# compares ganeti version
|
625 |
local_version = constants.PROTOCOL_VERSION |
626 |
if not remote_version: |
627 |
feedback_fn(" - ERROR: connection to %s failed" % (node))
|
628 |
return True |
629 |
|
630 |
if local_version != remote_version:
|
631 |
feedback_fn(" - ERROR: sw version mismatch: master %s, node(%s) %s" %
|
632 |
(local_version, node, remote_version)) |
633 |
return True |
634 |
|
635 |
# checks vg existance and size > 20G
|
636 |
|
637 |
bad = False
|
638 |
if not vglist: |
639 |
feedback_fn(" - ERROR: unable to check volume groups on node %s." %
|
640 |
(node,)) |
641 |
bad = True
|
642 |
else:
|
643 |
vgstatus = _HasValidVG(vglist, self.cfg.GetVGName())
|
644 |
if vgstatus:
|
645 |
feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
|
646 |
bad = True
|
647 |
|
648 |
# checks config file checksum
|
649 |
# checks ssh to any
|
650 |
|
651 |
if 'filelist' not in node_result: |
652 |
bad = True
|
653 |
feedback_fn(" - ERROR: node hasn't returned file checksum data")
|
654 |
else:
|
655 |
remote_cksum = node_result['filelist']
|
656 |
for file_name in file_list: |
657 |
if file_name not in remote_cksum: |
658 |
bad = True
|
659 |
feedback_fn(" - ERROR: file '%s' missing" % file_name)
|
660 |
elif remote_cksum[file_name] != local_cksum[file_name]:
|
661 |
bad = True
|
662 |
feedback_fn(" - ERROR: file '%s' has wrong checksum" % file_name)
|
663 |
|
664 |
if 'nodelist' not in node_result: |
665 |
bad = True
|
666 |
feedback_fn(" - ERROR: node hasn't returned node connectivity data")
|
667 |
else:
|
668 |
if node_result['nodelist']: |
669 |
bad = True
|
670 |
for node in node_result['nodelist']: |
671 |
feedback_fn(" - ERROR: communication with node '%s': %s" %
|
672 |
(node, node_result['nodelist'][node]))
|
673 |
hyp_result = node_result.get('hypervisor', None) |
674 |
if hyp_result is not None: |
675 |
feedback_fn(" - ERROR: hypervisor verify failure: '%s'" % hyp_result)
|
676 |
return bad
|
677 |
|
678 |
def _VerifyInstance(self, instance, node_vol_is, node_instance, feedback_fn): |
679 |
"""Verify an instance.
|
680 |
|
681 |
This function checks to see if the required block devices are
|
682 |
available on the instance's node.
|
683 |
|
684 |
"""
|
685 |
bad = False
|
686 |
|
687 |
instancelist = self.cfg.GetInstanceList()
|
688 |
if not instance in instancelist: |
689 |
feedback_fn(" - ERROR: instance %s not in instance list %s" %
|
690 |
(instance, instancelist)) |
691 |
bad = True
|
692 |
|
693 |
instanceconfig = self.cfg.GetInstanceInfo(instance)
|
694 |
node_current = instanceconfig.primary_node |
695 |
|
696 |
node_vol_should = {} |
697 |
instanceconfig.MapLVsByNode(node_vol_should) |
698 |
|
699 |
for node in node_vol_should: |
700 |
for volume in node_vol_should[node]: |
701 |
if node not in node_vol_is or volume not in node_vol_is[node]: |
702 |
feedback_fn(" - ERROR: volume %s missing on node %s" %
|
703 |
(volume, node)) |
704 |
bad = True
|
705 |
|
706 |
if not instanceconfig.status == 'down': |
707 |
if not instance in node_instance[node_current]: |
708 |
feedback_fn(" - ERROR: instance %s not running on node %s" %
|
709 |
(instance, node_current)) |
710 |
bad = True
|
711 |
|
712 |
for node in node_instance: |
713 |
if (not node == node_current): |
714 |
if instance in node_instance[node]: |
715 |
feedback_fn(" - ERROR: instance %s should not run on node %s" %
|
716 |
(instance, node)) |
717 |
bad = True
|
718 |
|
719 |
return bad
|
720 |
|
721 |
def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn): |
722 |
"""Verify if there are any unknown volumes in the cluster.
|
723 |
|
724 |
The .os, .swap and backup volumes are ignored. All other volumes are
|
725 |
reported as unknown.
|
726 |
|
727 |
"""
|
728 |
bad = False
|
729 |
|
730 |
for node in node_vol_is: |
731 |
for volume in node_vol_is[node]: |
732 |
if node not in node_vol_should or volume not in node_vol_should[node]: |
733 |
feedback_fn(" - ERROR: volume %s on node %s should not exist" %
|
734 |
(volume, node)) |
735 |
bad = True
|
736 |
return bad
|
737 |
|
738 |
def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn): |
739 |
"""Verify the list of running instances.
|
740 |
|
741 |
This checks what instances are running but unknown to the cluster.
|
742 |
|
743 |
"""
|
744 |
bad = False
|
745 |
for node in node_instance: |
746 |
for runninginstance in node_instance[node]: |
747 |
if runninginstance not in instancelist: |
748 |
feedback_fn(" - ERROR: instance %s on node %s should not exist" %
|
749 |
(runninginstance, node)) |
750 |
bad = True
|
751 |
return bad
|
752 |
|
753 |
def CheckPrereq(self): |
754 |
"""Check prerequisites.
|
755 |
|
756 |
This has no prerequisites.
|
757 |
|
758 |
"""
|
759 |
pass
|
760 |
|
761 |
def Exec(self, feedback_fn): |
762 |
"""Verify integrity of cluster, performing various test on nodes.
|
763 |
|
764 |
"""
|
765 |
bad = False
|
766 |
feedback_fn("* Verifying global settings")
|
767 |
for msg in self.cfg.VerifyConfig(): |
768 |
feedback_fn(" - ERROR: %s" % msg)
|
769 |
|
770 |
vg_name = self.cfg.GetVGName()
|
771 |
nodelist = utils.NiceSort(self.cfg.GetNodeList())
|
772 |
instancelist = utils.NiceSort(self.cfg.GetInstanceList())
|
773 |
node_volume = {} |
774 |
node_instance = {} |
775 |
|
776 |
# FIXME: verify OS list
|
777 |
# do local checksums
|
778 |
file_names = list(self.sstore.GetFileList()) |
779 |
file_names.append(constants.SSL_CERT_FILE) |
780 |
file_names.append(constants.CLUSTER_CONF_FILE) |
781 |
local_checksums = utils.FingerprintFiles(file_names) |
782 |
|
783 |
feedback_fn("* Gathering data (%d nodes)" % len(nodelist)) |
784 |
all_volumeinfo = rpc.call_volume_list(nodelist, vg_name) |
785 |
all_instanceinfo = rpc.call_instance_list(nodelist) |
786 |
all_vglist = rpc.call_vg_list(nodelist) |
787 |
node_verify_param = { |
788 |
'filelist': file_names,
|
789 |
'nodelist': nodelist,
|
790 |
'hypervisor': None, |
791 |
} |
792 |
all_nvinfo = rpc.call_node_verify(nodelist, node_verify_param) |
793 |
all_rversion = rpc.call_version(nodelist) |
794 |
|
795 |
for node in nodelist: |
796 |
feedback_fn("* Verifying node %s" % node)
|
797 |
result = self._VerifyNode(node, file_names, local_checksums,
|
798 |
all_vglist[node], all_nvinfo[node], |
799 |
all_rversion[node], feedback_fn) |
800 |
bad = bad or result
|
801 |
|
802 |
# node_volume
|
803 |
volumeinfo = all_volumeinfo[node] |
804 |
|
805 |
if isinstance(volumeinfo, basestring): |
806 |
feedback_fn(" - ERROR: LVM problem on node %s: %s" %
|
807 |
(node, volumeinfo[-400:].encode('string_escape'))) |
808 |
bad = True
|
809 |
node_volume[node] = {} |
810 |
elif not isinstance(volumeinfo, dict): |
811 |
feedback_fn(" - ERROR: connection to %s failed" % (node,))
|
812 |
bad = True
|
813 |
continue
|
814 |
else:
|
815 |
node_volume[node] = volumeinfo |
816 |
|
817 |
# node_instance
|
818 |
nodeinstance = all_instanceinfo[node] |
819 |
if type(nodeinstance) != list: |
820 |
feedback_fn(" - ERROR: connection to %s failed" % (node,))
|
821 |
bad = True
|
822 |
continue
|
823 |
|
824 |
node_instance[node] = nodeinstance |
825 |
|
826 |
node_vol_should = {} |
827 |
|
828 |
for instance in instancelist: |
829 |
feedback_fn("* Verifying instance %s" % instance)
|
830 |
result = self._VerifyInstance(instance, node_volume, node_instance,
|
831 |
feedback_fn) |
832 |
bad = bad or result
|
833 |
|
834 |
inst_config = self.cfg.GetInstanceInfo(instance)
|
835 |
|
836 |
inst_config.MapLVsByNode(node_vol_should) |
837 |
|
838 |
feedback_fn("* Verifying orphan volumes")
|
839 |
result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
|
840 |
feedback_fn) |
841 |
bad = bad or result
|
842 |
|
843 |
feedback_fn("* Verifying remaining instances")
|
844 |
result = self._VerifyOrphanInstances(instancelist, node_instance,
|
845 |
feedback_fn) |
846 |
bad = bad or result
|
847 |
|
848 |
return int(bad) |
849 |
|
850 |
|
851 |
class LUVerifyDisks(NoHooksLU): |
852 |
"""Verifies the cluster disks status.
|
853 |
|
854 |
"""
|
855 |
_OP_REQP = [] |
856 |
|
857 |
def CheckPrereq(self): |
858 |
"""Check prerequisites.
|
859 |
|
860 |
This has no prerequisites.
|
861 |
|
862 |
"""
|
863 |
pass
|
864 |
|
865 |
def Exec(self, feedback_fn): |
866 |
"""Verify integrity of cluster disks.
|
867 |
|
868 |
"""
|
869 |
result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {} |
870 |
|
871 |
vg_name = self.cfg.GetVGName()
|
872 |
nodes = utils.NiceSort(self.cfg.GetNodeList())
|
873 |
instances = [self.cfg.GetInstanceInfo(name)
|
874 |
for name in self.cfg.GetInstanceList()] |
875 |
|
876 |
nv_dict = {} |
877 |
for inst in instances: |
878 |
inst_lvs = {} |
879 |
if (inst.status != "up" or |
880 |
inst.disk_template not in constants.DTS_NET_MIRROR): |
881 |
continue
|
882 |
inst.MapLVsByNode(inst_lvs) |
883 |
# transform { iname: {node: [vol,],},} to {(node, vol): iname}
|
884 |
for node, vol_list in inst_lvs.iteritems(): |
885 |
for vol in vol_list: |
886 |
nv_dict[(node, vol)] = inst |
887 |
|
888 |
if not nv_dict: |
889 |
return result
|
890 |
|
891 |
node_lvs = rpc.call_volume_list(nodes, vg_name) |
892 |
|
893 |
to_act = set()
|
894 |
for node in nodes: |
895 |
# node_volume
|
896 |
lvs = node_lvs[node] |
897 |
|
898 |
if isinstance(lvs, basestring): |
899 |
logger.Info("error enumerating LVs on node %s: %s" % (node, lvs))
|
900 |
res_nlvm[node] = lvs |
901 |
elif not isinstance(lvs, dict): |
902 |
logger.Info("connection to node %s failed or invalid data returned" %
|
903 |
(node,)) |
904 |
res_nodes.append(node) |
905 |
continue
|
906 |
|
907 |
for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems(): |
908 |
inst = nv_dict.pop((node, lv_name), None)
|
909 |
if (not lv_online and inst is not None |
910 |
and inst.name not in res_instances): |
911 |
res_instances.append(inst.name) |
912 |
|
913 |
# any leftover items in nv_dict are missing LVs, let's arrange the
|
914 |
# data better
|
915 |
for key, inst in nv_dict.iteritems(): |
916 |
if inst.name not in res_missing: |
917 |
res_missing[inst.name] = [] |
918 |
res_missing[inst.name].append(key) |
919 |
|
920 |
return result
|
921 |
|
922 |
|
923 |
class LURenameCluster(LogicalUnit): |
924 |
"""Rename the cluster.
|
925 |
|
926 |
"""
|
927 |
HPATH = "cluster-rename"
|
928 |
HTYPE = constants.HTYPE_CLUSTER |
929 |
_OP_REQP = ["name"]
|
930 |
|
931 |
def BuildHooksEnv(self): |
932 |
"""Build hooks env.
|
933 |
|
934 |
"""
|
935 |
env = { |
936 |
"OP_TARGET": self.sstore.GetClusterName(), |
937 |
"NEW_NAME": self.op.name, |
938 |
} |
939 |
mn = self.sstore.GetMasterNode()
|
940 |
return env, [mn], [mn]
|
941 |
|
942 |
def CheckPrereq(self): |
943 |
"""Verify that the passed name is a valid one.
|
944 |
|
945 |
"""
|
946 |
hostname = utils.HostInfo(self.op.name)
|
947 |
|
948 |
new_name = hostname.name |
949 |
self.ip = new_ip = hostname.ip
|
950 |
old_name = self.sstore.GetClusterName()
|
951 |
old_ip = self.sstore.GetMasterIP()
|
952 |
if new_name == old_name and new_ip == old_ip: |
953 |
raise errors.OpPrereqError("Neither the name nor the IP address of the" |
954 |
" cluster has changed")
|
955 |
if new_ip != old_ip:
|
956 |
result = utils.RunCmd(["fping", "-q", new_ip]) |
957 |
if not result.failed: |
958 |
raise errors.OpPrereqError("The given cluster IP address (%s) is" |
959 |
" reachable on the network. Aborting." %
|
960 |
new_ip) |
961 |
|
962 |
self.op.name = new_name
|
963 |
|
964 |
def Exec(self, feedback_fn): |
965 |
"""Rename the cluster.
|
966 |
|
967 |
"""
|
968 |
clustername = self.op.name
|
969 |
ip = self.ip
|
970 |
ss = self.sstore
|
971 |
|
972 |
# shutdown the master IP
|
973 |
master = ss.GetMasterNode() |
974 |
if not rpc.call_node_stop_master(master): |
975 |
raise errors.OpExecError("Could not disable the master role") |
976 |
|
977 |
try:
|
978 |
# modify the sstore
|
979 |
ss.SetKey(ss.SS_MASTER_IP, ip) |
980 |
ss.SetKey(ss.SS_CLUSTER_NAME, clustername) |
981 |
|
982 |
# Distribute updated ss config to all nodes
|
983 |
myself = self.cfg.GetNodeInfo(master)
|
984 |
dist_nodes = self.cfg.GetNodeList()
|
985 |
if myself.name in dist_nodes: |
986 |
dist_nodes.remove(myself.name) |
987 |
|
988 |
logger.Debug("Copying updated ssconf data to all nodes")
|
989 |
for keyname in [ss.SS_CLUSTER_NAME, ss.SS_MASTER_IP]: |
990 |
fname = ss.KeyToFilename(keyname) |
991 |
result = rpc.call_upload_file(dist_nodes, fname) |
992 |
for to_node in dist_nodes: |
993 |
if not result[to_node]: |
994 |
logger.Error("copy of file %s to node %s failed" %
|
995 |
(fname, to_node)) |
996 |
finally:
|
997 |
if not rpc.call_node_start_master(master): |
998 |
logger.Error("Could not re-enable the master role on the master,"
|
999 |
" please restart manually.")
|
1000 |
|
1001 |
|
1002 |
def _RecursiveCheckIfLVMBased(disk): |
1003 |
"""Check if the given disk or its children are lvm-based.
|
1004 |
|
1005 |
Args:
|
1006 |
disk: ganeti.objects.Disk object
|
1007 |
|
1008 |
Returns:
|
1009 |
boolean indicating whether a LD_LV dev_type was found or not
|
1010 |
|
1011 |
"""
|
1012 |
if disk.children:
|
1013 |
for chdisk in disk.children: |
1014 |
if _RecursiveCheckIfLVMBased(chdisk):
|
1015 |
return True |
1016 |
return disk.dev_type == constants.LD_LV
|
1017 |
|
1018 |
|
1019 |
class LUSetClusterParams(LogicalUnit): |
1020 |
"""Change the parameters of the cluster.
|
1021 |
|
1022 |
"""
|
1023 |
HPATH = "cluster-modify"
|
1024 |
HTYPE = constants.HTYPE_CLUSTER |
1025 |
_OP_REQP = [] |
1026 |
|
1027 |
def BuildHooksEnv(self): |
1028 |
"""Build hooks env.
|
1029 |
|
1030 |
"""
|
1031 |
env = { |
1032 |
"OP_TARGET": self.sstore.GetClusterName(), |
1033 |
"NEW_VG_NAME": self.op.vg_name, |
1034 |
} |
1035 |
mn = self.sstore.GetMasterNode()
|
1036 |
return env, [mn], [mn]
|
1037 |
|
1038 |
def CheckPrereq(self): |
1039 |
"""Check prerequisites.
|
1040 |
|
1041 |
This checks whether the given params don't conflict and
|
1042 |
if the given volume group is valid.
|
1043 |
|
1044 |
"""
|
1045 |
if not self.op.vg_name: |
1046 |
instances = [self.cfg.GetInstanceInfo(name)
|
1047 |
for name in self.cfg.GetInstanceList()] |
1048 |
for inst in instances: |
1049 |
for disk in inst.disks: |
1050 |
if _RecursiveCheckIfLVMBased(disk):
|
1051 |
raise errors.OpPrereqError("Cannot disable lvm storage while" |
1052 |
" lvm-based instances exist")
|
1053 |
|
1054 |
# if vg_name not None, checks given volume group on all nodes
|
1055 |
if self.op.vg_name: |
1056 |
node_list = self.cfg.GetNodeList()
|
1057 |
vglist = rpc.call_vg_list(node_list) |
1058 |
for node in node_list: |
1059 |
vgstatus = _HasValidVG(vglist[node], self.op.vg_name)
|
1060 |
if vgstatus:
|
1061 |
raise errors.OpPrereqError("Error on node '%s': %s" % |
1062 |
(node, vgstatus)) |
1063 |
|
1064 |
def Exec(self, feedback_fn): |
1065 |
"""Change the parameters of the cluster.
|
1066 |
|
1067 |
"""
|
1068 |
if self.op.vg_name != self.cfg.GetVGName(): |
1069 |
self.cfg.SetVGName(self.op.vg_name) |
1070 |
else:
|
1071 |
feedback_fn("Cluster LVM configuration already in desired"
|
1072 |
" state, not changing")
|
1073 |
|
1074 |
|
1075 |
def _WaitForSync(cfgw, instance, proc, oneshot=False, unlock=False): |
1076 |
"""Sleep and poll for an instance's disk to sync.
|
1077 |
|
1078 |
"""
|
1079 |
if not instance.disks: |
1080 |
return True |
1081 |
|
1082 |
if not oneshot: |
1083 |
proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
|
1084 |
|
1085 |
node = instance.primary_node |
1086 |
|
1087 |
for dev in instance.disks: |
1088 |
cfgw.SetDiskID(dev, node) |
1089 |
|
1090 |
retries = 0
|
1091 |
while True: |
1092 |
max_time = 0
|
1093 |
done = True
|
1094 |
cumul_degraded = False
|
1095 |
rstats = rpc.call_blockdev_getmirrorstatus(node, instance.disks) |
1096 |
if not rstats: |
1097 |
proc.LogWarning("Can't get any data from node %s" % node)
|
1098 |
retries += 1
|
1099 |
if retries >= 10: |
1100 |
raise errors.RemoteError("Can't contact node %s for mirror data," |
1101 |
" aborting." % node)
|
1102 |
time.sleep(6)
|
1103 |
continue
|
1104 |
retries = 0
|
1105 |
for i in range(len(rstats)): |
1106 |
mstat = rstats[i] |
1107 |
if mstat is None: |
1108 |
proc.LogWarning("Can't compute data for node %s/%s" %
|
1109 |
(node, instance.disks[i].iv_name)) |
1110 |
continue
|
1111 |
# we ignore the ldisk parameter
|
1112 |
perc_done, est_time, is_degraded, _ = mstat |
1113 |
cumul_degraded = cumul_degraded or (is_degraded and perc_done is None) |
1114 |
if perc_done is not None: |
1115 |
done = False
|
1116 |
if est_time is not None: |
1117 |
rem_time = "%d estimated seconds remaining" % est_time
|
1118 |
max_time = est_time |
1119 |
else:
|
1120 |
rem_time = "no time estimate"
|
1121 |
proc.LogInfo("- device %s: %5.2f%% done, %s" %
|
1122 |
(instance.disks[i].iv_name, perc_done, rem_time)) |
1123 |
if done or oneshot: |
1124 |
break
|
1125 |
|
1126 |
if unlock:
|
1127 |
utils.Unlock('cmd')
|
1128 |
try:
|
1129 |
time.sleep(min(60, max_time)) |
1130 |
finally:
|
1131 |
if unlock:
|
1132 |
utils.Lock('cmd')
|
1133 |
|
1134 |
if done:
|
1135 |
proc.LogInfo("Instance %s's disks are in sync." % instance.name)
|
1136 |
return not cumul_degraded |
1137 |
|
1138 |
|
1139 |
def _CheckDiskConsistency(cfgw, dev, node, on_primary, ldisk=False): |
1140 |
"""Check that mirrors are not degraded.
|
1141 |
|
1142 |
The ldisk parameter, if True, will change the test from the
|
1143 |
is_degraded attribute (which represents overall non-ok status for
|
1144 |
the device(s)) to the ldisk (representing the local storage status).
|
1145 |
|
1146 |
"""
|
1147 |
cfgw.SetDiskID(dev, node) |
1148 |
if ldisk:
|
1149 |
idx = 6
|
1150 |
else:
|
1151 |
idx = 5
|
1152 |
|
1153 |
result = True
|
1154 |
if on_primary or dev.AssembleOnSecondary(): |
1155 |
rstats = rpc.call_blockdev_find(node, dev) |
1156 |
if not rstats: |
1157 |
logger.ToStderr("Can't get any data from node %s" % node)
|
1158 |
result = False
|
1159 |
else:
|
1160 |
result = result and (not rstats[idx]) |
1161 |
if dev.children:
|
1162 |
for child in dev.children: |
1163 |
result = result and _CheckDiskConsistency(cfgw, child, node, on_primary)
|
1164 |
|
1165 |
return result
|
1166 |
|
1167 |
|
1168 |
class LUDiagnoseOS(NoHooksLU): |
1169 |
"""Logical unit for OS diagnose/query.
|
1170 |
|
1171 |
"""
|
1172 |
_OP_REQP = [] |
1173 |
|
1174 |
def CheckPrereq(self): |
1175 |
"""Check prerequisites.
|
1176 |
|
1177 |
This always succeeds, since this is a pure query LU.
|
1178 |
|
1179 |
"""
|
1180 |
return
|
1181 |
|
1182 |
def Exec(self, feedback_fn): |
1183 |
"""Compute the list of OSes.
|
1184 |
|
1185 |
"""
|
1186 |
node_list = self.cfg.GetNodeList()
|
1187 |
node_data = rpc.call_os_diagnose(node_list) |
1188 |
if node_data == False: |
1189 |
raise errors.OpExecError("Can't gather the list of OSes") |
1190 |
return node_data
|
1191 |
|
1192 |
|
1193 |
class LURemoveNode(LogicalUnit): |
1194 |
"""Logical unit for removing a node.
|
1195 |
|
1196 |
"""
|
1197 |
HPATH = "node-remove"
|
1198 |
HTYPE = constants.HTYPE_NODE |
1199 |
_OP_REQP = ["node_name"]
|
1200 |
|
1201 |
def BuildHooksEnv(self): |
1202 |
"""Build hooks env.
|
1203 |
|
1204 |
This doesn't run on the target node in the pre phase as a failed
|
1205 |
node would not allows itself to run.
|
1206 |
|
1207 |
"""
|
1208 |
env = { |
1209 |
"OP_TARGET": self.op.node_name, |
1210 |
"NODE_NAME": self.op.node_name, |
1211 |
} |
1212 |
all_nodes = self.cfg.GetNodeList()
|
1213 |
all_nodes.remove(self.op.node_name)
|
1214 |
return env, all_nodes, all_nodes
|
1215 |
|
1216 |
def CheckPrereq(self): |
1217 |
"""Check prerequisites.
|
1218 |
|
1219 |
This checks:
|
1220 |
- the node exists in the configuration
|
1221 |
- it does not have primary or secondary instances
|
1222 |
- it's not the master
|
1223 |
|
1224 |
Any errors are signalled by raising errors.OpPrereqError.
|
1225 |
|
1226 |
"""
|
1227 |
node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name)) |
1228 |
if node is None: |
1229 |
raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name) |
1230 |
|
1231 |
instance_list = self.cfg.GetInstanceList()
|
1232 |
|
1233 |
masternode = self.sstore.GetMasterNode()
|
1234 |
if node.name == masternode:
|
1235 |
raise errors.OpPrereqError("Node is the master node," |
1236 |
" you need to failover first.")
|
1237 |
|
1238 |
for instance_name in instance_list: |
1239 |
instance = self.cfg.GetInstanceInfo(instance_name)
|
1240 |
if node.name == instance.primary_node:
|
1241 |
raise errors.OpPrereqError("Instance %s still running on the node," |
1242 |
" please remove first." % instance_name)
|
1243 |
if node.name in instance.secondary_nodes: |
1244 |
raise errors.OpPrereqError("Instance %s has node as a secondary," |
1245 |
" please remove first." % instance_name)
|
1246 |
self.op.node_name = node.name
|
1247 |
self.node = node
|
1248 |
|
1249 |
def Exec(self, feedback_fn): |
1250 |
"""Removes the node from the cluster.
|
1251 |
|
1252 |
"""
|
1253 |
node = self.node
|
1254 |
logger.Info("stopping the node daemon and removing configs from node %s" %
|
1255 |
node.name) |
1256 |
|
1257 |
rpc.call_node_leave_cluster(node.name) |
1258 |
|
1259 |
self.ssh.Run(node.name, 'root', "%s stop" % constants.NODE_INITD_SCRIPT) |
1260 |
|
1261 |
logger.Info("Removing node %s from config" % node.name)
|
1262 |
|
1263 |
self.cfg.RemoveNode(node.name)
|
1264 |
|
1265 |
_RemoveHostFromEtcHosts(node.name) |
1266 |
|
1267 |
|
1268 |
class LUQueryNodes(NoHooksLU): |
1269 |
"""Logical unit for querying nodes.
|
1270 |
|
1271 |
"""
|
1272 |
_OP_REQP = ["output_fields", "names"] |
1273 |
|
1274 |
def CheckPrereq(self): |
1275 |
"""Check prerequisites.
|
1276 |
|
1277 |
This checks that the fields required are valid output fields.
|
1278 |
|
1279 |
"""
|
1280 |
self.dynamic_fields = frozenset(["dtotal", "dfree", |
1281 |
"mtotal", "mnode", "mfree", |
1282 |
"bootid"])
|
1283 |
|
1284 |
_CheckOutputFields(static=["name", "pinst_cnt", "sinst_cnt", |
1285 |
"pinst_list", "sinst_list", |
1286 |
"pip", "sip"], |
1287 |
dynamic=self.dynamic_fields,
|
1288 |
selected=self.op.output_fields)
|
1289 |
|
1290 |
self.wanted = _GetWantedNodes(self, self.op.names) |
1291 |
|
1292 |
def Exec(self, feedback_fn): |
1293 |
"""Computes the list of nodes and their attributes.
|
1294 |
|
1295 |
"""
|
1296 |
nodenames = self.wanted
|
1297 |
nodelist = [self.cfg.GetNodeInfo(name) for name in nodenames] |
1298 |
|
1299 |
# begin data gathering
|
1300 |
|
1301 |
if self.dynamic_fields.intersection(self.op.output_fields): |
1302 |
live_data = {} |
1303 |
node_data = rpc.call_node_info(nodenames, self.cfg.GetVGName())
|
1304 |
for name in nodenames: |
1305 |
nodeinfo = node_data.get(name, None)
|
1306 |
if nodeinfo:
|
1307 |
live_data[name] = { |
1308 |
"mtotal": utils.TryConvert(int, nodeinfo['memory_total']), |
1309 |
"mnode": utils.TryConvert(int, nodeinfo['memory_dom0']), |
1310 |
"mfree": utils.TryConvert(int, nodeinfo['memory_free']), |
1311 |
"dtotal": utils.TryConvert(int, nodeinfo['vg_size']), |
1312 |
"dfree": utils.TryConvert(int, nodeinfo['vg_free']), |
1313 |
"bootid": nodeinfo['bootid'], |
1314 |
} |
1315 |
else:
|
1316 |
live_data[name] = {} |
1317 |
else:
|
1318 |
live_data = dict.fromkeys(nodenames, {})
|
1319 |
|
1320 |
node_to_primary = dict([(name, set()) for name in nodenames]) |
1321 |
node_to_secondary = dict([(name, set()) for name in nodenames]) |
1322 |
|
1323 |
inst_fields = frozenset(("pinst_cnt", "pinst_list", |
1324 |
"sinst_cnt", "sinst_list")) |
1325 |
if inst_fields & frozenset(self.op.output_fields): |
1326 |
instancelist = self.cfg.GetInstanceList()
|
1327 |
|
1328 |
for instance_name in instancelist: |
1329 |
inst = self.cfg.GetInstanceInfo(instance_name)
|
1330 |
if inst.primary_node in node_to_primary: |
1331 |
node_to_primary[inst.primary_node].add(inst.name) |
1332 |
for secnode in inst.secondary_nodes: |
1333 |
if secnode in node_to_secondary: |
1334 |
node_to_secondary[secnode].add(inst.name) |
1335 |
|
1336 |
# end data gathering
|
1337 |
|
1338 |
output = [] |
1339 |
for node in nodelist: |
1340 |
node_output = [] |
1341 |
for field in self.op.output_fields: |
1342 |
if field == "name": |
1343 |
val = node.name |
1344 |
elif field == "pinst_list": |
1345 |
val = list(node_to_primary[node.name])
|
1346 |
elif field == "sinst_list": |
1347 |
val = list(node_to_secondary[node.name])
|
1348 |
elif field == "pinst_cnt": |
1349 |
val = len(node_to_primary[node.name])
|
1350 |
elif field == "sinst_cnt": |
1351 |
val = len(node_to_secondary[node.name])
|
1352 |
elif field == "pip": |
1353 |
val = node.primary_ip |
1354 |
elif field == "sip": |
1355 |
val = node.secondary_ip |
1356 |
elif field in self.dynamic_fields: |
1357 |
val = live_data[node.name].get(field, None)
|
1358 |
else:
|
1359 |
raise errors.ParameterError(field)
|
1360 |
node_output.append(val) |
1361 |
output.append(node_output) |
1362 |
|
1363 |
return output
|
1364 |
|
1365 |
|
1366 |
class LUQueryNodeVolumes(NoHooksLU): |
1367 |
"""Logical unit for getting volumes on node(s).
|
1368 |
|
1369 |
"""
|
1370 |
_OP_REQP = ["nodes", "output_fields"] |
1371 |
|
1372 |
def CheckPrereq(self): |
1373 |
"""Check prerequisites.
|
1374 |
|
1375 |
This checks that the fields required are valid output fields.
|
1376 |
|
1377 |
"""
|
1378 |
self.nodes = _GetWantedNodes(self, self.op.nodes) |
1379 |
|
1380 |
_CheckOutputFields(static=["node"],
|
1381 |
dynamic=["phys", "vg", "name", "size", "instance"], |
1382 |
selected=self.op.output_fields)
|
1383 |
|
1384 |
|
1385 |
def Exec(self, feedback_fn): |
1386 |
"""Computes the list of nodes and their attributes.
|
1387 |
|
1388 |
"""
|
1389 |
nodenames = self.nodes
|
1390 |
volumes = rpc.call_node_volumes(nodenames) |
1391 |
|
1392 |
ilist = [self.cfg.GetInstanceInfo(iname) for iname |
1393 |
in self.cfg.GetInstanceList()] |
1394 |
|
1395 |
lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist]) |
1396 |
|
1397 |
output = [] |
1398 |
for node in nodenames: |
1399 |
if node not in volumes or not volumes[node]: |
1400 |
continue
|
1401 |
|
1402 |
node_vols = volumes[node][:] |
1403 |
node_vols.sort(key=lambda vol: vol['dev']) |
1404 |
|
1405 |
for vol in node_vols: |
1406 |
node_output = [] |
1407 |
for field in self.op.output_fields: |
1408 |
if field == "node": |
1409 |
val = node |
1410 |
elif field == "phys": |
1411 |
val = vol['dev']
|
1412 |
elif field == "vg": |
1413 |
val = vol['vg']
|
1414 |
elif field == "name": |
1415 |
val = vol['name']
|
1416 |
elif field == "size": |
1417 |
val = int(float(vol['size'])) |
1418 |
elif field == "instance": |
1419 |
for inst in ilist: |
1420 |
if node not in lv_by_node[inst]: |
1421 |
continue
|
1422 |
if vol['name'] in lv_by_node[inst][node]: |
1423 |
val = inst.name |
1424 |
break
|
1425 |
else:
|
1426 |
val = '-'
|
1427 |
else:
|
1428 |
raise errors.ParameterError(field)
|
1429 |
node_output.append(str(val))
|
1430 |
|
1431 |
output.append(node_output) |
1432 |
|
1433 |
return output
|
1434 |
|
1435 |
|
1436 |
class LUAddNode(LogicalUnit): |
1437 |
"""Logical unit for adding node to the cluster.
|
1438 |
|
1439 |
"""
|
1440 |
HPATH = "node-add"
|
1441 |
HTYPE = constants.HTYPE_NODE |
1442 |
_OP_REQP = ["node_name"]
|
1443 |
|
1444 |
def BuildHooksEnv(self): |
1445 |
"""Build hooks env.
|
1446 |
|
1447 |
This will run on all nodes before, and on all nodes + the new node after.
|
1448 |
|
1449 |
"""
|
1450 |
env = { |
1451 |
"OP_TARGET": self.op.node_name, |
1452 |
"NODE_NAME": self.op.node_name, |
1453 |
"NODE_PIP": self.op.primary_ip, |
1454 |
"NODE_SIP": self.op.secondary_ip, |
1455 |
} |
1456 |
nodes_0 = self.cfg.GetNodeList()
|
1457 |
nodes_1 = nodes_0 + [self.op.node_name, ]
|
1458 |
return env, nodes_0, nodes_1
|
1459 |
|
1460 |
def CheckPrereq(self): |
1461 |
"""Check prerequisites.
|
1462 |
|
1463 |
This checks:
|
1464 |
- the new node is not already in the config
|
1465 |
- it is resolvable
|
1466 |
- its parameters (single/dual homed) matches the cluster
|
1467 |
|
1468 |
Any errors are signalled by raising errors.OpPrereqError.
|
1469 |
|
1470 |
"""
|
1471 |
node_name = self.op.node_name
|
1472 |
cfg = self.cfg
|
1473 |
|
1474 |
dns_data = utils.HostInfo(node_name) |
1475 |
|
1476 |
node = dns_data.name |
1477 |
primary_ip = self.op.primary_ip = dns_data.ip
|
1478 |
secondary_ip = getattr(self.op, "secondary_ip", None) |
1479 |
if secondary_ip is None: |
1480 |
secondary_ip = primary_ip |
1481 |
if not utils.IsValidIP(secondary_ip): |
1482 |
raise errors.OpPrereqError("Invalid secondary IP given") |
1483 |
self.op.secondary_ip = secondary_ip
|
1484 |
node_list = cfg.GetNodeList() |
1485 |
if node in node_list: |
1486 |
raise errors.OpPrereqError("Node %s is already in the configuration" |
1487 |
% node) |
1488 |
|
1489 |
for existing_node_name in node_list: |
1490 |
existing_node = cfg.GetNodeInfo(existing_node_name) |
1491 |
if (existing_node.primary_ip == primary_ip or |
1492 |
existing_node.secondary_ip == primary_ip or
|
1493 |
existing_node.primary_ip == secondary_ip or
|
1494 |
existing_node.secondary_ip == secondary_ip): |
1495 |
raise errors.OpPrereqError("New node ip address(es) conflict with" |
1496 |
" existing node %s" % existing_node.name)
|
1497 |
|
1498 |
# check that the type of the node (single versus dual homed) is the
|
1499 |
# same as for the master
|
1500 |
myself = cfg.GetNodeInfo(self.sstore.GetMasterNode())
|
1501 |
master_singlehomed = myself.secondary_ip == myself.primary_ip |
1502 |
newbie_singlehomed = secondary_ip == primary_ip |
1503 |
if master_singlehomed != newbie_singlehomed:
|
1504 |
if master_singlehomed:
|
1505 |
raise errors.OpPrereqError("The master has no private ip but the" |
1506 |
" new node has one")
|
1507 |
else:
|
1508 |
raise errors.OpPrereqError("The master has a private ip but the" |
1509 |
" new node doesn't have one")
|
1510 |
|
1511 |
# checks reachablity
|
1512 |
if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT): |
1513 |
raise errors.OpPrereqError("Node not reachable by ping") |
1514 |
|
1515 |
if not newbie_singlehomed: |
1516 |
# check reachability from my secondary ip to newbie's secondary ip
|
1517 |
if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, |
1518 |
source=myself.secondary_ip): |
1519 |
raise errors.OpPrereqError("Node secondary ip not reachable by TCP" |
1520 |
" based ping to noded port")
|
1521 |
|
1522 |
self.new_node = objects.Node(name=node,
|
1523 |
primary_ip=primary_ip, |
1524 |
secondary_ip=secondary_ip) |
1525 |
|
1526 |
if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31: |
1527 |
if not os.path.exists(constants.VNC_PASSWORD_FILE): |
1528 |
raise errors.OpPrereqError("Cluster VNC password file %s missing" % |
1529 |
constants.VNC_PASSWORD_FILE) |
1530 |
|
1531 |
def Exec(self, feedback_fn): |
1532 |
"""Adds the new node to the cluster.
|
1533 |
|
1534 |
"""
|
1535 |
new_node = self.new_node
|
1536 |
node = new_node.name |
1537 |
|
1538 |
# set up inter-node password and certificate and restarts the node daemon
|
1539 |
gntpass = self.sstore.GetNodeDaemonPassword()
|
1540 |
if not re.match('^[a-zA-Z0-9.]{1,64}$', gntpass): |
1541 |
raise errors.OpExecError("ganeti password corruption detected") |
1542 |
f = open(constants.SSL_CERT_FILE)
|
1543 |
try:
|
1544 |
gntpem = f.read(8192)
|
1545 |
finally:
|
1546 |
f.close() |
1547 |
# in the base64 pem encoding, neither '!' nor '.' are valid chars,
|
1548 |
# so we use this to detect an invalid certificate; as long as the
|
1549 |
# cert doesn't contain this, the here-document will be correctly
|
1550 |
# parsed by the shell sequence below
|
1551 |
if re.search('^!EOF\.', gntpem, re.MULTILINE): |
1552 |
raise errors.OpExecError("invalid PEM encoding in the SSL certificate") |
1553 |
if not gntpem.endswith("\n"): |
1554 |
raise errors.OpExecError("PEM must end with newline") |
1555 |
logger.Info("copy cluster pass to %s and starting the node daemon" % node)
|
1556 |
|
1557 |
# and then connect with ssh to set password and start ganeti-noded
|
1558 |
# note that all the below variables are sanitized at this point,
|
1559 |
# either by being constants or by the checks above
|
1560 |
ss = self.sstore
|
1561 |
mycommand = ("umask 077 && "
|
1562 |
"echo '%s' > '%s' && "
|
1563 |
"cat > '%s' << '!EOF.' && \n"
|
1564 |
"%s!EOF.\n%s restart" %
|
1565 |
(gntpass, ss.KeyToFilename(ss.SS_NODED_PASS), |
1566 |
constants.SSL_CERT_FILE, gntpem, |
1567 |
constants.NODE_INITD_SCRIPT)) |
1568 |
|
1569 |
result = self.ssh.Run(node, 'root', mycommand, batch=False, ask_key=True) |
1570 |
if result.failed:
|
1571 |
raise errors.OpExecError("Remote command on node %s, error: %s," |
1572 |
" output: %s" %
|
1573 |
(node, result.fail_reason, result.output)) |
1574 |
|
1575 |
# check connectivity
|
1576 |
time.sleep(4)
|
1577 |
|
1578 |
result = rpc.call_version([node])[node] |
1579 |
if result:
|
1580 |
if constants.PROTOCOL_VERSION == result:
|
1581 |
logger.Info("communication to node %s fine, sw version %s match" %
|
1582 |
(node, result)) |
1583 |
else:
|
1584 |
raise errors.OpExecError("Version mismatch master version %s," |
1585 |
" node version %s" %
|
1586 |
(constants.PROTOCOL_VERSION, result)) |
1587 |
else:
|
1588 |
raise errors.OpExecError("Cannot get version from the new node") |
1589 |
|
1590 |
# setup ssh on node
|
1591 |
logger.Info("copy ssh key to node %s" % node)
|
1592 |
priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) |
1593 |
keyarray = [] |
1594 |
keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB, |
1595 |
constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB, |
1596 |
priv_key, pub_key] |
1597 |
|
1598 |
for i in keyfiles: |
1599 |
f = open(i, 'r') |
1600 |
try:
|
1601 |
keyarray.append(f.read()) |
1602 |
finally:
|
1603 |
f.close() |
1604 |
|
1605 |
result = rpc.call_node_add(node, keyarray[0], keyarray[1], keyarray[2], |
1606 |
keyarray[3], keyarray[4], keyarray[5]) |
1607 |
|
1608 |
if not result: |
1609 |
raise errors.OpExecError("Cannot transfer ssh keys to the new node") |
1610 |
|
1611 |
# Add node to our /etc/hosts, and add key to known_hosts
|
1612 |
_AddHostToEtcHosts(new_node.name) |
1613 |
|
1614 |
if new_node.secondary_ip != new_node.primary_ip:
|
1615 |
if not rpc.call_node_tcp_ping(new_node.name, |
1616 |
constants.LOCALHOST_IP_ADDRESS, |
1617 |
new_node.secondary_ip, |
1618 |
constants.DEFAULT_NODED_PORT, |
1619 |
10, False): |
1620 |
raise errors.OpExecError("Node claims it doesn't have the secondary ip" |
1621 |
" you gave (%s). Please fix and re-run this"
|
1622 |
" command." % new_node.secondary_ip)
|
1623 |
|
1624 |
success, msg = self.ssh.VerifyNodeHostname(node)
|
1625 |
if not success: |
1626 |
raise errors.OpExecError("Node '%s' claims it has a different hostname" |
1627 |
" than the one the resolver gives: %s."
|
1628 |
" Please fix and re-run this command." %
|
1629 |
(node, msg)) |
1630 |
|
1631 |
# Distribute updated /etc/hosts and known_hosts to all nodes,
|
1632 |
# including the node just added
|
1633 |
myself = self.cfg.GetNodeInfo(self.sstore.GetMasterNode()) |
1634 |
dist_nodes = self.cfg.GetNodeList() + [node]
|
1635 |
if myself.name in dist_nodes: |
1636 |
dist_nodes.remove(myself.name) |
1637 |
|
1638 |
logger.Debug("Copying hosts and known_hosts to all nodes")
|
1639 |
for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE): |
1640 |
result = rpc.call_upload_file(dist_nodes, fname) |
1641 |
for to_node in dist_nodes: |
1642 |
if not result[to_node]: |
1643 |
logger.Error("copy of file %s to node %s failed" %
|
1644 |
(fname, to_node)) |
1645 |
|
1646 |
to_copy = ss.GetFileList() |
1647 |
if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31: |
1648 |
to_copy.append(constants.VNC_PASSWORD_FILE) |
1649 |
for fname in to_copy: |
1650 |
if not self.ssh.CopyFileToNode(node, fname): |
1651 |
logger.Error("could not copy file %s to node %s" % (fname, node))
|
1652 |
|
1653 |
logger.Info("adding node %s to cluster.conf" % node)
|
1654 |
self.cfg.AddNode(new_node)
|
1655 |
|
1656 |
|
1657 |
class LUMasterFailover(LogicalUnit): |
1658 |
"""Failover the master node to the current node.
|
1659 |
|
1660 |
This is a special LU in that it must run on a non-master node.
|
1661 |
|
1662 |
"""
|
1663 |
HPATH = "master-failover"
|
1664 |
HTYPE = constants.HTYPE_CLUSTER |
1665 |
REQ_MASTER = False
|
1666 |
_OP_REQP = [] |
1667 |
|
1668 |
def BuildHooksEnv(self): |
1669 |
"""Build hooks env.
|
1670 |
|
1671 |
This will run on the new master only in the pre phase, and on all
|
1672 |
the nodes in the post phase.
|
1673 |
|
1674 |
"""
|
1675 |
env = { |
1676 |
"OP_TARGET": self.new_master, |
1677 |
"NEW_MASTER": self.new_master, |
1678 |
"OLD_MASTER": self.old_master, |
1679 |
} |
1680 |
return env, [self.new_master], self.cfg.GetNodeList() |
1681 |
|
1682 |
def CheckPrereq(self): |
1683 |
"""Check prerequisites.
|
1684 |
|
1685 |
This checks that we are not already the master.
|
1686 |
|
1687 |
"""
|
1688 |
self.new_master = utils.HostInfo().name
|
1689 |
self.old_master = self.sstore.GetMasterNode() |
1690 |
|
1691 |
if self.old_master == self.new_master: |
1692 |
raise errors.OpPrereqError("This commands must be run on the node" |
1693 |
" where you want the new master to be."
|
1694 |
" %s is already the master" %
|
1695 |
self.old_master)
|
1696 |
|
1697 |
def Exec(self, feedback_fn): |
1698 |
"""Failover the master node.
|
1699 |
|
1700 |
This command, when run on a non-master node, will cause the current
|
1701 |
master to cease being master, and the non-master to become new
|
1702 |
master.
|
1703 |
|
1704 |
"""
|
1705 |
#TODO: do not rely on gethostname returning the FQDN
|
1706 |
logger.Info("setting master to %s, old master: %s" %
|
1707 |
(self.new_master, self.old_master)) |
1708 |
|
1709 |
if not rpc.call_node_stop_master(self.old_master): |
1710 |
logger.Error("could disable the master role on the old master"
|
1711 |
" %s, please disable manually" % self.old_master) |
1712 |
|
1713 |
ss = self.sstore
|
1714 |
ss.SetKey(ss.SS_MASTER_NODE, self.new_master)
|
1715 |
if not rpc.call_upload_file(self.cfg.GetNodeList(), |
1716 |
ss.KeyToFilename(ss.SS_MASTER_NODE)): |
1717 |
logger.Error("could not distribute the new simple store master file"
|
1718 |
" to the other nodes, please check.")
|
1719 |
|
1720 |
if not rpc.call_node_start_master(self.new_master): |
1721 |
logger.Error("could not start the master role on the new master"
|
1722 |
" %s, please check" % self.new_master) |
1723 |
feedback_fn("Error in activating the master IP on the new master,"
|
1724 |
" please fix manually.")
|
1725 |
|
1726 |
|
1727 |
|
1728 |
class LUQueryClusterInfo(NoHooksLU): |
1729 |
"""Query cluster configuration.
|
1730 |
|
1731 |
"""
|
1732 |
_OP_REQP = [] |
1733 |
REQ_MASTER = False
|
1734 |
|
1735 |
def CheckPrereq(self): |
1736 |
"""No prerequsites needed for this LU.
|
1737 |
|
1738 |
"""
|
1739 |
pass
|
1740 |
|
1741 |
def Exec(self, feedback_fn): |
1742 |
"""Return cluster config.
|
1743 |
|
1744 |
"""
|
1745 |
result = { |
1746 |
"name": self.sstore.GetClusterName(), |
1747 |
"software_version": constants.RELEASE_VERSION,
|
1748 |
"protocol_version": constants.PROTOCOL_VERSION,
|
1749 |
"config_version": constants.CONFIG_VERSION,
|
1750 |
"os_api_version": constants.OS_API_VERSION,
|
1751 |
"export_version": constants.EXPORT_VERSION,
|
1752 |
"master": self.sstore.GetMasterNode(), |
1753 |
"architecture": (platform.architecture()[0], platform.machine()), |
1754 |
} |
1755 |
|
1756 |
return result
|
1757 |
|
1758 |
|
1759 |
class LUClusterCopyFile(NoHooksLU): |
1760 |
"""Copy file to cluster.
|
1761 |
|
1762 |
"""
|
1763 |
_OP_REQP = ["nodes", "filename"] |
1764 |
|
1765 |
def CheckPrereq(self): |
1766 |
"""Check prerequisites.
|
1767 |
|
1768 |
It should check that the named file exists and that the given list
|
1769 |
of nodes is valid.
|
1770 |
|
1771 |
"""
|
1772 |
if not os.path.exists(self.op.filename): |
1773 |
raise errors.OpPrereqError("No such filename '%s'" % self.op.filename) |
1774 |
|
1775 |
self.nodes = _GetWantedNodes(self, self.op.nodes) |
1776 |
|
1777 |
def Exec(self, feedback_fn): |
1778 |
"""Copy a file from master to some nodes.
|
1779 |
|
1780 |
Args:
|
1781 |
opts - class with options as members
|
1782 |
args - list containing a single element, the file name
|
1783 |
Opts used:
|
1784 |
nodes - list containing the name of target nodes; if empty, all nodes
|
1785 |
|
1786 |
"""
|
1787 |
filename = self.op.filename
|
1788 |
|
1789 |
myname = utils.HostInfo().name |
1790 |
|
1791 |
for node in self.nodes: |
1792 |
if node == myname:
|
1793 |
continue
|
1794 |
if not self.ssh.CopyFileToNode(node, filename): |
1795 |
logger.Error("Copy of file %s to node %s failed" % (filename, node))
|
1796 |
|
1797 |
|
1798 |
class LUDumpClusterConfig(NoHooksLU): |
1799 |
"""Return a text-representation of the cluster-config.
|
1800 |
|
1801 |
"""
|
1802 |
_OP_REQP = [] |
1803 |
|
1804 |
def CheckPrereq(self): |
1805 |
"""No prerequisites.
|
1806 |
|
1807 |
"""
|
1808 |
pass
|
1809 |
|
1810 |
def Exec(self, feedback_fn): |
1811 |
"""Dump a representation of the cluster config to the standard output.
|
1812 |
|
1813 |
"""
|
1814 |
return self.cfg.DumpConfig() |
1815 |
|
1816 |
|
1817 |
class LURunClusterCommand(NoHooksLU): |
1818 |
"""Run a command on some nodes.
|
1819 |
|
1820 |
"""
|
1821 |
_OP_REQP = ["command", "nodes"] |
1822 |
|
1823 |
def CheckPrereq(self): |
1824 |
"""Check prerequisites.
|
1825 |
|
1826 |
It checks that the given list of nodes is valid.
|
1827 |
|
1828 |
"""
|
1829 |
self.nodes = _GetWantedNodes(self, self.op.nodes) |
1830 |
|
1831 |
def Exec(self, feedback_fn): |
1832 |
"""Run a command on some nodes.
|
1833 |
|
1834 |
"""
|
1835 |
data = [] |
1836 |
for node in self.nodes: |
1837 |
result = self.ssh.Run(node, "root", self.op.command) |
1838 |
data.append((node, result.output, result.exit_code)) |
1839 |
|
1840 |
return data
|
1841 |
|
1842 |
|
1843 |
class LUActivateInstanceDisks(NoHooksLU): |
1844 |
"""Bring up an instance's disks.
|
1845 |
|
1846 |
"""
|
1847 |
_OP_REQP = ["instance_name"]
|
1848 |
|
1849 |
def CheckPrereq(self): |
1850 |
"""Check prerequisites.
|
1851 |
|
1852 |
This checks that the instance is in the cluster.
|
1853 |
|
1854 |
"""
|
1855 |
instance = self.cfg.GetInstanceInfo(
|
1856 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
1857 |
if instance is None: |
1858 |
raise errors.OpPrereqError("Instance '%s' not known" % |
1859 |
self.op.instance_name)
|
1860 |
self.instance = instance
|
1861 |
|
1862 |
|
1863 |
def Exec(self, feedback_fn): |
1864 |
"""Activate the disks.
|
1865 |
|
1866 |
"""
|
1867 |
disks_ok, disks_info = _AssembleInstanceDisks(self.instance, self.cfg) |
1868 |
if not disks_ok: |
1869 |
raise errors.OpExecError("Cannot activate block devices") |
1870 |
|
1871 |
return disks_info
|
1872 |
|
1873 |
|
1874 |
def _AssembleInstanceDisks(instance, cfg, ignore_secondaries=False): |
1875 |
"""Prepare the block devices for an instance.
|
1876 |
|
1877 |
This sets up the block devices on all nodes.
|
1878 |
|
1879 |
Args:
|
1880 |
instance: a ganeti.objects.Instance object
|
1881 |
ignore_secondaries: if true, errors on secondary nodes won't result
|
1882 |
in an error return from the function
|
1883 |
|
1884 |
Returns:
|
1885 |
false if the operation failed
|
1886 |
list of (host, instance_visible_name, node_visible_name) if the operation
|
1887 |
suceeded with the mapping from node devices to instance devices
|
1888 |
"""
|
1889 |
device_info = [] |
1890 |
disks_ok = True
|
1891 |
iname = instance.name |
1892 |
# With the two passes mechanism we try to reduce the window of
|
1893 |
# opportunity for the race condition of switching DRBD to primary
|
1894 |
# before handshaking occured, but we do not eliminate it
|
1895 |
|
1896 |
# The proper fix would be to wait (with some limits) until the
|
1897 |
# connection has been made and drbd transitions from WFConnection
|
1898 |
# into any other network-connected state (Connected, SyncTarget,
|
1899 |
# SyncSource, etc.)
|
1900 |
|
1901 |
# 1st pass, assemble on all nodes in secondary mode
|
1902 |
for inst_disk in instance.disks: |
1903 |
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): |
1904 |
cfg.SetDiskID(node_disk, node) |
1905 |
result = rpc.call_blockdev_assemble(node, node_disk, iname, False)
|
1906 |
if not result: |
1907 |
logger.Error("could not prepare block device %s on node %s"
|
1908 |
" (is_primary=False, pass=1)" % (inst_disk.iv_name, node))
|
1909 |
if not ignore_secondaries: |
1910 |
disks_ok = False
|
1911 |
|
1912 |
# FIXME: race condition on drbd migration to primary
|
1913 |
|
1914 |
# 2nd pass, do only the primary node
|
1915 |
for inst_disk in instance.disks: |
1916 |
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): |
1917 |
if node != instance.primary_node:
|
1918 |
continue
|
1919 |
cfg.SetDiskID(node_disk, node) |
1920 |
result = rpc.call_blockdev_assemble(node, node_disk, iname, True)
|
1921 |
if not result: |
1922 |
logger.Error("could not prepare block device %s on node %s"
|
1923 |
" (is_primary=True, pass=2)" % (inst_disk.iv_name, node))
|
1924 |
disks_ok = False
|
1925 |
device_info.append((instance.primary_node, inst_disk.iv_name, result)) |
1926 |
|
1927 |
# leave the disks configured for the primary node
|
1928 |
# this is a workaround that would be fixed better by
|
1929 |
# improving the logical/physical id handling
|
1930 |
for disk in instance.disks: |
1931 |
cfg.SetDiskID(disk, instance.primary_node) |
1932 |
|
1933 |
return disks_ok, device_info
|
1934 |
|
1935 |
|
1936 |
def _StartInstanceDisks(cfg, instance, force): |
1937 |
"""Start the disks of an instance.
|
1938 |
|
1939 |
"""
|
1940 |
disks_ok, dummy = _AssembleInstanceDisks(instance, cfg, |
1941 |
ignore_secondaries=force) |
1942 |
if not disks_ok: |
1943 |
_ShutdownInstanceDisks(instance, cfg) |
1944 |
if force is not None and not force: |
1945 |
logger.Error("If the message above refers to a secondary node,"
|
1946 |
" you can retry the operation using '--force'.")
|
1947 |
raise errors.OpExecError("Disk consistency error") |
1948 |
|
1949 |
|
1950 |
class LUDeactivateInstanceDisks(NoHooksLU): |
1951 |
"""Shutdown an instance's disks.
|
1952 |
|
1953 |
"""
|
1954 |
_OP_REQP = ["instance_name"]
|
1955 |
|
1956 |
def CheckPrereq(self): |
1957 |
"""Check prerequisites.
|
1958 |
|
1959 |
This checks that the instance is in the cluster.
|
1960 |
|
1961 |
"""
|
1962 |
instance = self.cfg.GetInstanceInfo(
|
1963 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
1964 |
if instance is None: |
1965 |
raise errors.OpPrereqError("Instance '%s' not known" % |
1966 |
self.op.instance_name)
|
1967 |
self.instance = instance
|
1968 |
|
1969 |
def Exec(self, feedback_fn): |
1970 |
"""Deactivate the disks
|
1971 |
|
1972 |
"""
|
1973 |
instance = self.instance
|
1974 |
ins_l = rpc.call_instance_list([instance.primary_node]) |
1975 |
ins_l = ins_l[instance.primary_node] |
1976 |
if not type(ins_l) is list: |
1977 |
raise errors.OpExecError("Can't contact node '%s'" % |
1978 |
instance.primary_node) |
1979 |
|
1980 |
if self.instance.name in ins_l: |
1981 |
raise errors.OpExecError("Instance is running, can't shutdown" |
1982 |
" block devices.")
|
1983 |
|
1984 |
_ShutdownInstanceDisks(instance, self.cfg)
|
1985 |
|
1986 |
|
1987 |
def _ShutdownInstanceDisks(instance, cfg, ignore_primary=False): |
1988 |
"""Shutdown block devices of an instance.
|
1989 |
|
1990 |
This does the shutdown on all nodes of the instance.
|
1991 |
|
1992 |
If the ignore_primary is false, errors on the primary node are
|
1993 |
ignored.
|
1994 |
|
1995 |
"""
|
1996 |
result = True
|
1997 |
for disk in instance.disks: |
1998 |
for node, top_disk in disk.ComputeNodeTree(instance.primary_node): |
1999 |
cfg.SetDiskID(top_disk, node) |
2000 |
if not rpc.call_blockdev_shutdown(node, top_disk): |
2001 |
logger.Error("could not shutdown block device %s on node %s" %
|
2002 |
(disk.iv_name, node)) |
2003 |
if not ignore_primary or node != instance.primary_node: |
2004 |
result = False
|
2005 |
return result
|
2006 |
|
2007 |
|
2008 |
def _CheckNodeFreeMemory(cfg, node, reason, requested): |
2009 |
"""Checks if a node has enough free memory.
|
2010 |
|
2011 |
This function check if a given node has the needed amount of free
|
2012 |
memory. In case the node has less memory or we cannot get the
|
2013 |
information from the node, this function raise an OpPrereqError
|
2014 |
exception.
|
2015 |
|
2016 |
Args:
|
2017 |
- cfg: a ConfigWriter instance
|
2018 |
- node: the node name
|
2019 |
- reason: string to use in the error message
|
2020 |
- requested: the amount of memory in MiB
|
2021 |
|
2022 |
"""
|
2023 |
nodeinfo = rpc.call_node_info([node], cfg.GetVGName()) |
2024 |
if not nodeinfo or not isinstance(nodeinfo, dict): |
2025 |
raise errors.OpPrereqError("Could not contact node %s for resource" |
2026 |
" information" % (node,))
|
2027 |
|
2028 |
free_mem = nodeinfo[node].get('memory_free')
|
2029 |
if not isinstance(free_mem, int): |
2030 |
raise errors.OpPrereqError("Can't compute free memory on node %s, result" |
2031 |
" was '%s'" % (node, free_mem))
|
2032 |
if requested > free_mem:
|
2033 |
raise errors.OpPrereqError("Not enough memory on node %s for %s:" |
2034 |
" needed %s MiB, available %s MiB" %
|
2035 |
(node, reason, requested, free_mem)) |
2036 |
|
2037 |
|
2038 |
class LUStartupInstance(LogicalUnit): |
2039 |
"""Starts an instance.
|
2040 |
|
2041 |
"""
|
2042 |
HPATH = "instance-start"
|
2043 |
HTYPE = constants.HTYPE_INSTANCE |
2044 |
_OP_REQP = ["instance_name", "force"] |
2045 |
|
2046 |
def BuildHooksEnv(self): |
2047 |
"""Build hooks env.
|
2048 |
|
2049 |
This runs on master, primary and secondary nodes of the instance.
|
2050 |
|
2051 |
"""
|
2052 |
env = { |
2053 |
"FORCE": self.op.force, |
2054 |
} |
2055 |
env.update(_BuildInstanceHookEnvByObject(self.instance))
|
2056 |
nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + |
2057 |
list(self.instance.secondary_nodes)) |
2058 |
return env, nl, nl
|
2059 |
|
2060 |
def CheckPrereq(self): |
2061 |
"""Check prerequisites.
|
2062 |
|
2063 |
This checks that the instance is in the cluster.
|
2064 |
|
2065 |
"""
|
2066 |
instance = self.cfg.GetInstanceInfo(
|
2067 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2068 |
if instance is None: |
2069 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2070 |
self.op.instance_name)
|
2071 |
|
2072 |
# check bridges existance
|
2073 |
_CheckInstanceBridgesExist(instance) |
2074 |
|
2075 |
_CheckNodeFreeMemory(self.cfg, instance.primary_node,
|
2076 |
"starting instance %s" % instance.name,
|
2077 |
instance.memory) |
2078 |
|
2079 |
self.instance = instance
|
2080 |
self.op.instance_name = instance.name
|
2081 |
|
2082 |
def Exec(self, feedback_fn): |
2083 |
"""Start the instance.
|
2084 |
|
2085 |
"""
|
2086 |
instance = self.instance
|
2087 |
force = self.op.force
|
2088 |
extra_args = getattr(self.op, "extra_args", "") |
2089 |
|
2090 |
self.cfg.MarkInstanceUp(instance.name)
|
2091 |
|
2092 |
node_current = instance.primary_node |
2093 |
|
2094 |
_StartInstanceDisks(self.cfg, instance, force)
|
2095 |
|
2096 |
if not rpc.call_instance_start(node_current, instance, extra_args): |
2097 |
_ShutdownInstanceDisks(instance, self.cfg)
|
2098 |
raise errors.OpExecError("Could not start instance") |
2099 |
|
2100 |
|
2101 |
class LURebootInstance(LogicalUnit): |
2102 |
"""Reboot an instance.
|
2103 |
|
2104 |
"""
|
2105 |
HPATH = "instance-reboot"
|
2106 |
HTYPE = constants.HTYPE_INSTANCE |
2107 |
_OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"] |
2108 |
|
2109 |
def BuildHooksEnv(self): |
2110 |
"""Build hooks env.
|
2111 |
|
2112 |
This runs on master, primary and secondary nodes of the instance.
|
2113 |
|
2114 |
"""
|
2115 |
env = { |
2116 |
"IGNORE_SECONDARIES": self.op.ignore_secondaries, |
2117 |
} |
2118 |
env.update(_BuildInstanceHookEnvByObject(self.instance))
|
2119 |
nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + |
2120 |
list(self.instance.secondary_nodes)) |
2121 |
return env, nl, nl
|
2122 |
|
2123 |
def CheckPrereq(self): |
2124 |
"""Check prerequisites.
|
2125 |
|
2126 |
This checks that the instance is in the cluster.
|
2127 |
|
2128 |
"""
|
2129 |
instance = self.cfg.GetInstanceInfo(
|
2130 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2131 |
if instance is None: |
2132 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2133 |
self.op.instance_name)
|
2134 |
|
2135 |
# check bridges existance
|
2136 |
_CheckInstanceBridgesExist(instance) |
2137 |
|
2138 |
self.instance = instance
|
2139 |
self.op.instance_name = instance.name
|
2140 |
|
2141 |
def Exec(self, feedback_fn): |
2142 |
"""Reboot the instance.
|
2143 |
|
2144 |
"""
|
2145 |
instance = self.instance
|
2146 |
ignore_secondaries = self.op.ignore_secondaries
|
2147 |
reboot_type = self.op.reboot_type
|
2148 |
extra_args = getattr(self.op, "extra_args", "") |
2149 |
|
2150 |
node_current = instance.primary_node |
2151 |
|
2152 |
if reboot_type not in [constants.INSTANCE_REBOOT_SOFT, |
2153 |
constants.INSTANCE_REBOOT_HARD, |
2154 |
constants.INSTANCE_REBOOT_FULL]: |
2155 |
raise errors.ParameterError("reboot type not in [%s, %s, %s]" % |
2156 |
(constants.INSTANCE_REBOOT_SOFT, |
2157 |
constants.INSTANCE_REBOOT_HARD, |
2158 |
constants.INSTANCE_REBOOT_FULL)) |
2159 |
|
2160 |
if reboot_type in [constants.INSTANCE_REBOOT_SOFT, |
2161 |
constants.INSTANCE_REBOOT_HARD]: |
2162 |
if not rpc.call_instance_reboot(node_current, instance, |
2163 |
reboot_type, extra_args): |
2164 |
raise errors.OpExecError("Could not reboot instance") |
2165 |
else:
|
2166 |
if not rpc.call_instance_shutdown(node_current, instance): |
2167 |
raise errors.OpExecError("could not shutdown instance for full reboot") |
2168 |
_ShutdownInstanceDisks(instance, self.cfg)
|
2169 |
_StartInstanceDisks(self.cfg, instance, ignore_secondaries)
|
2170 |
if not rpc.call_instance_start(node_current, instance, extra_args): |
2171 |
_ShutdownInstanceDisks(instance, self.cfg)
|
2172 |
raise errors.OpExecError("Could not start instance for full reboot") |
2173 |
|
2174 |
self.cfg.MarkInstanceUp(instance.name)
|
2175 |
|
2176 |
|
2177 |
class LUShutdownInstance(LogicalUnit): |
2178 |
"""Shutdown an instance.
|
2179 |
|
2180 |
"""
|
2181 |
HPATH = "instance-stop"
|
2182 |
HTYPE = constants.HTYPE_INSTANCE |
2183 |
_OP_REQP = ["instance_name"]
|
2184 |
|
2185 |
def BuildHooksEnv(self): |
2186 |
"""Build hooks env.
|
2187 |
|
2188 |
This runs on master, primary and secondary nodes of the instance.
|
2189 |
|
2190 |
"""
|
2191 |
env = _BuildInstanceHookEnvByObject(self.instance)
|
2192 |
nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + |
2193 |
list(self.instance.secondary_nodes)) |
2194 |
return env, nl, nl
|
2195 |
|
2196 |
def CheckPrereq(self): |
2197 |
"""Check prerequisites.
|
2198 |
|
2199 |
This checks that the instance is in the cluster.
|
2200 |
|
2201 |
"""
|
2202 |
instance = self.cfg.GetInstanceInfo(
|
2203 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2204 |
if instance is None: |
2205 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2206 |
self.op.instance_name)
|
2207 |
self.instance = instance
|
2208 |
|
2209 |
def Exec(self, feedback_fn): |
2210 |
"""Shutdown the instance.
|
2211 |
|
2212 |
"""
|
2213 |
instance = self.instance
|
2214 |
node_current = instance.primary_node |
2215 |
self.cfg.MarkInstanceDown(instance.name)
|
2216 |
if not rpc.call_instance_shutdown(node_current, instance): |
2217 |
logger.Error("could not shutdown instance")
|
2218 |
|
2219 |
_ShutdownInstanceDisks(instance, self.cfg)
|
2220 |
|
2221 |
|
2222 |
class LUReinstallInstance(LogicalUnit): |
2223 |
"""Reinstall an instance.
|
2224 |
|
2225 |
"""
|
2226 |
HPATH = "instance-reinstall"
|
2227 |
HTYPE = constants.HTYPE_INSTANCE |
2228 |
_OP_REQP = ["instance_name"]
|
2229 |
|
2230 |
def BuildHooksEnv(self): |
2231 |
"""Build hooks env.
|
2232 |
|
2233 |
This runs on master, primary and secondary nodes of the instance.
|
2234 |
|
2235 |
"""
|
2236 |
env = _BuildInstanceHookEnvByObject(self.instance)
|
2237 |
nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + |
2238 |
list(self.instance.secondary_nodes)) |
2239 |
return env, nl, nl
|
2240 |
|
2241 |
def CheckPrereq(self): |
2242 |
"""Check prerequisites.
|
2243 |
|
2244 |
This checks that the instance is in the cluster and is not running.
|
2245 |
|
2246 |
"""
|
2247 |
instance = self.cfg.GetInstanceInfo(
|
2248 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2249 |
if instance is None: |
2250 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2251 |
self.op.instance_name)
|
2252 |
if instance.disk_template == constants.DT_DISKLESS:
|
2253 |
raise errors.OpPrereqError("Instance '%s' has no disks" % |
2254 |
self.op.instance_name)
|
2255 |
if instance.status != "down": |
2256 |
raise errors.OpPrereqError("Instance '%s' is marked to be up" % |
2257 |
self.op.instance_name)
|
2258 |
remote_info = rpc.call_instance_info(instance.primary_node, instance.name) |
2259 |
if remote_info:
|
2260 |
raise errors.OpPrereqError("Instance '%s' is running on the node %s" % |
2261 |
(self.op.instance_name,
|
2262 |
instance.primary_node)) |
2263 |
|
2264 |
self.op.os_type = getattr(self.op, "os_type", None) |
2265 |
if self.op.os_type is not None: |
2266 |
# OS verification
|
2267 |
pnode = self.cfg.GetNodeInfo(
|
2268 |
self.cfg.ExpandNodeName(instance.primary_node))
|
2269 |
if pnode is None: |
2270 |
raise errors.OpPrereqError("Primary node '%s' is unknown" % |
2271 |
self.op.pnode)
|
2272 |
os_obj = rpc.call_os_get(pnode.name, self.op.os_type)
|
2273 |
if not os_obj: |
2274 |
raise errors.OpPrereqError("OS '%s' not in supported OS list for" |
2275 |
" primary node" % self.op.os_type) |
2276 |
|
2277 |
self.instance = instance
|
2278 |
|
2279 |
def Exec(self, feedback_fn): |
2280 |
"""Reinstall the instance.
|
2281 |
|
2282 |
"""
|
2283 |
inst = self.instance
|
2284 |
|
2285 |
if self.op.os_type is not None: |
2286 |
feedback_fn("Changing OS to '%s'..." % self.op.os_type) |
2287 |
inst.os = self.op.os_type
|
2288 |
self.cfg.AddInstance(inst)
|
2289 |
|
2290 |
_StartInstanceDisks(self.cfg, inst, None) |
2291 |
try:
|
2292 |
feedback_fn("Running the instance OS create scripts...")
|
2293 |
if not rpc.call_instance_os_add(inst.primary_node, inst, "sda", "sdb"): |
2294 |
raise errors.OpExecError("Could not install OS for instance %s" |
2295 |
" on node %s" %
|
2296 |
(inst.name, inst.primary_node)) |
2297 |
finally:
|
2298 |
_ShutdownInstanceDisks(inst, self.cfg)
|
2299 |
|
2300 |
|
2301 |
class LURenameInstance(LogicalUnit): |
2302 |
"""Rename an instance.
|
2303 |
|
2304 |
"""
|
2305 |
HPATH = "instance-rename"
|
2306 |
HTYPE = constants.HTYPE_INSTANCE |
2307 |
_OP_REQP = ["instance_name", "new_name"] |
2308 |
|
2309 |
def BuildHooksEnv(self): |
2310 |
"""Build hooks env.
|
2311 |
|
2312 |
This runs on master, primary and secondary nodes of the instance.
|
2313 |
|
2314 |
"""
|
2315 |
env = _BuildInstanceHookEnvByObject(self.instance)
|
2316 |
env["INSTANCE_NEW_NAME"] = self.op.new_name |
2317 |
nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + |
2318 |
list(self.instance.secondary_nodes)) |
2319 |
return env, nl, nl
|
2320 |
|
2321 |
def CheckPrereq(self): |
2322 |
"""Check prerequisites.
|
2323 |
|
2324 |
This checks that the instance is in the cluster and is not running.
|
2325 |
|
2326 |
"""
|
2327 |
instance = self.cfg.GetInstanceInfo(
|
2328 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2329 |
if instance is None: |
2330 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2331 |
self.op.instance_name)
|
2332 |
if instance.status != "down": |
2333 |
raise errors.OpPrereqError("Instance '%s' is marked to be up" % |
2334 |
self.op.instance_name)
|
2335 |
remote_info = rpc.call_instance_info(instance.primary_node, instance.name) |
2336 |
if remote_info:
|
2337 |
raise errors.OpPrereqError("Instance '%s' is running on the node %s" % |
2338 |
(self.op.instance_name,
|
2339 |
instance.primary_node)) |
2340 |
self.instance = instance
|
2341 |
|
2342 |
# new name verification
|
2343 |
name_info = utils.HostInfo(self.op.new_name)
|
2344 |
|
2345 |
self.op.new_name = new_name = name_info.name
|
2346 |
instance_list = self.cfg.GetInstanceList()
|
2347 |
if new_name in instance_list: |
2348 |
raise errors.OpPrereqError("Instance '%s' is already in the cluster" % |
2349 |
instance_name) |
2350 |
|
2351 |
if not getattr(self.op, "ignore_ip", False): |
2352 |
command = ["fping", "-q", name_info.ip] |
2353 |
result = utils.RunCmd(command) |
2354 |
if not result.failed: |
2355 |
raise errors.OpPrereqError("IP %s of instance %s already in use" % |
2356 |
(name_info.ip, new_name)) |
2357 |
|
2358 |
|
2359 |
def Exec(self, feedback_fn): |
2360 |
"""Reinstall the instance.
|
2361 |
|
2362 |
"""
|
2363 |
inst = self.instance
|
2364 |
old_name = inst.name |
2365 |
|
2366 |
self.cfg.RenameInstance(inst.name, self.op.new_name) |
2367 |
|
2368 |
# re-read the instance from the configuration after rename
|
2369 |
inst = self.cfg.GetInstanceInfo(self.op.new_name) |
2370 |
|
2371 |
_StartInstanceDisks(self.cfg, inst, None) |
2372 |
try:
|
2373 |
if not rpc.call_instance_run_rename(inst.primary_node, inst, old_name, |
2374 |
"sda", "sdb"): |
2375 |
msg = ("Could run OS rename script for instance %s on node %s (but the"
|
2376 |
" instance has been renamed in Ganeti)" %
|
2377 |
(inst.name, inst.primary_node)) |
2378 |
logger.Error(msg) |
2379 |
finally:
|
2380 |
_ShutdownInstanceDisks(inst, self.cfg)
|
2381 |
|
2382 |
|
2383 |
class LURemoveInstance(LogicalUnit): |
2384 |
"""Remove an instance.
|
2385 |
|
2386 |
"""
|
2387 |
HPATH = "instance-remove"
|
2388 |
HTYPE = constants.HTYPE_INSTANCE |
2389 |
_OP_REQP = ["instance_name"]
|
2390 |
|
2391 |
def BuildHooksEnv(self): |
2392 |
"""Build hooks env.
|
2393 |
|
2394 |
This runs on master, primary and secondary nodes of the instance.
|
2395 |
|
2396 |
"""
|
2397 |
env = _BuildInstanceHookEnvByObject(self.instance)
|
2398 |
nl = [self.sstore.GetMasterNode()]
|
2399 |
return env, nl, nl
|
2400 |
|
2401 |
def CheckPrereq(self): |
2402 |
"""Check prerequisites.
|
2403 |
|
2404 |
This checks that the instance is in the cluster.
|
2405 |
|
2406 |
"""
|
2407 |
instance = self.cfg.GetInstanceInfo(
|
2408 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2409 |
if instance is None: |
2410 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2411 |
self.op.instance_name)
|
2412 |
self.instance = instance
|
2413 |
|
2414 |
def Exec(self, feedback_fn): |
2415 |
"""Remove the instance.
|
2416 |
|
2417 |
"""
|
2418 |
instance = self.instance
|
2419 |
logger.Info("shutting down instance %s on node %s" %
|
2420 |
(instance.name, instance.primary_node)) |
2421 |
|
2422 |
if not rpc.call_instance_shutdown(instance.primary_node, instance): |
2423 |
if self.op.ignore_failures: |
2424 |
feedback_fn("Warning: can't shutdown instance")
|
2425 |
else:
|
2426 |
raise errors.OpExecError("Could not shutdown instance %s on node %s" % |
2427 |
(instance.name, instance.primary_node)) |
2428 |
|
2429 |
logger.Info("removing block devices for instance %s" % instance.name)
|
2430 |
|
2431 |
if not _RemoveDisks(instance, self.cfg): |
2432 |
if self.op.ignore_failures: |
2433 |
feedback_fn("Warning: can't remove instance's disks")
|
2434 |
else:
|
2435 |
raise errors.OpExecError("Can't remove instance's disks") |
2436 |
|
2437 |
logger.Info("removing instance %s out of cluster config" % instance.name)
|
2438 |
|
2439 |
self.cfg.RemoveInstance(instance.name)
|
2440 |
|
2441 |
|
2442 |
class LUQueryInstances(NoHooksLU): |
2443 |
"""Logical unit for querying instances.
|
2444 |
|
2445 |
"""
|
2446 |
_OP_REQP = ["output_fields", "names"] |
2447 |
|
2448 |
def CheckPrereq(self): |
2449 |
"""Check prerequisites.
|
2450 |
|
2451 |
This checks that the fields required are valid output fields.
|
2452 |
|
2453 |
"""
|
2454 |
self.dynamic_fields = frozenset(["oper_state", "oper_ram", "status"]) |
2455 |
_CheckOutputFields(static=["name", "os", "pnode", "snodes", |
2456 |
"admin_state", "admin_ram", |
2457 |
"disk_template", "ip", "mac", "bridge", |
2458 |
"sda_size", "sdb_size", "vcpus"], |
2459 |
dynamic=self.dynamic_fields,
|
2460 |
selected=self.op.output_fields)
|
2461 |
|
2462 |
self.wanted = _GetWantedInstances(self, self.op.names) |
2463 |
|
2464 |
def Exec(self, feedback_fn): |
2465 |
"""Computes the list of nodes and their attributes.
|
2466 |
|
2467 |
"""
|
2468 |
instance_names = self.wanted
|
2469 |
instance_list = [self.cfg.GetInstanceInfo(iname) for iname |
2470 |
in instance_names]
|
2471 |
|
2472 |
# begin data gathering
|
2473 |
|
2474 |
nodes = frozenset([inst.primary_node for inst in instance_list]) |
2475 |
|
2476 |
bad_nodes = [] |
2477 |
if self.dynamic_fields.intersection(self.op.output_fields): |
2478 |
live_data = {} |
2479 |
node_data = rpc.call_all_instances_info(nodes) |
2480 |
for name in nodes: |
2481 |
result = node_data[name] |
2482 |
if result:
|
2483 |
live_data.update(result) |
2484 |
elif result == False: |
2485 |
bad_nodes.append(name) |
2486 |
# else no instance is alive
|
2487 |
else:
|
2488 |
live_data = dict([(name, {}) for name in instance_names]) |
2489 |
|
2490 |
# end data gathering
|
2491 |
|
2492 |
output = [] |
2493 |
for instance in instance_list: |
2494 |
iout = [] |
2495 |
for field in self.op.output_fields: |
2496 |
if field == "name": |
2497 |
val = instance.name |
2498 |
elif field == "os": |
2499 |
val = instance.os |
2500 |
elif field == "pnode": |
2501 |
val = instance.primary_node |
2502 |
elif field == "snodes": |
2503 |
val = list(instance.secondary_nodes)
|
2504 |
elif field == "admin_state": |
2505 |
val = (instance.status != "down")
|
2506 |
elif field == "oper_state": |
2507 |
if instance.primary_node in bad_nodes: |
2508 |
val = None
|
2509 |
else:
|
2510 |
val = bool(live_data.get(instance.name))
|
2511 |
elif field == "status": |
2512 |
if instance.primary_node in bad_nodes: |
2513 |
val = "ERROR_nodedown"
|
2514 |
else:
|
2515 |
running = bool(live_data.get(instance.name))
|
2516 |
if running:
|
2517 |
if instance.status != "down": |
2518 |
val = "running"
|
2519 |
else:
|
2520 |
val = "ERROR_up"
|
2521 |
else:
|
2522 |
if instance.status != "down": |
2523 |
val = "ERROR_down"
|
2524 |
else:
|
2525 |
val = "ADMIN_down"
|
2526 |
elif field == "admin_ram": |
2527 |
val = instance.memory |
2528 |
elif field == "oper_ram": |
2529 |
if instance.primary_node in bad_nodes: |
2530 |
val = None
|
2531 |
elif instance.name in live_data: |
2532 |
val = live_data[instance.name].get("memory", "?") |
2533 |
else:
|
2534 |
val = "-"
|
2535 |
elif field == "disk_template": |
2536 |
val = instance.disk_template |
2537 |
elif field == "ip": |
2538 |
val = instance.nics[0].ip
|
2539 |
elif field == "bridge": |
2540 |
val = instance.nics[0].bridge
|
2541 |
elif field == "mac": |
2542 |
val = instance.nics[0].mac
|
2543 |
elif field == "sda_size" or field == "sdb_size": |
2544 |
disk = instance.FindDisk(field[:3])
|
2545 |
if disk is None: |
2546 |
val = None
|
2547 |
else:
|
2548 |
val = disk.size |
2549 |
elif field == "vcpus": |
2550 |
val = instance.vcpus |
2551 |
else:
|
2552 |
raise errors.ParameterError(field)
|
2553 |
iout.append(val) |
2554 |
output.append(iout) |
2555 |
|
2556 |
return output
|
2557 |
|
2558 |
|
2559 |
class LUFailoverInstance(LogicalUnit): |
2560 |
"""Failover an instance.
|
2561 |
|
2562 |
"""
|
2563 |
HPATH = "instance-failover"
|
2564 |
HTYPE = constants.HTYPE_INSTANCE |
2565 |
_OP_REQP = ["instance_name", "ignore_consistency"] |
2566 |
|
2567 |
def BuildHooksEnv(self): |
2568 |
"""Build hooks env.
|
2569 |
|
2570 |
This runs on master, primary and secondary nodes of the instance.
|
2571 |
|
2572 |
"""
|
2573 |
env = { |
2574 |
"IGNORE_CONSISTENCY": self.op.ignore_consistency, |
2575 |
} |
2576 |
env.update(_BuildInstanceHookEnvByObject(self.instance))
|
2577 |
nl = [self.sstore.GetMasterNode()] + list(self.instance.secondary_nodes) |
2578 |
return env, nl, nl
|
2579 |
|
2580 |
def CheckPrereq(self): |
2581 |
"""Check prerequisites.
|
2582 |
|
2583 |
This checks that the instance is in the cluster.
|
2584 |
|
2585 |
"""
|
2586 |
instance = self.cfg.GetInstanceInfo(
|
2587 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
2588 |
if instance is None: |
2589 |
raise errors.OpPrereqError("Instance '%s' not known" % |
2590 |
self.op.instance_name)
|
2591 |
|
2592 |
if instance.disk_template not in constants.DTS_NET_MIRROR: |
2593 |
raise errors.OpPrereqError("Instance's disk layout is not" |
2594 |
" network mirrored, cannot failover.")
|
2595 |
|
2596 |
secondary_nodes = instance.secondary_nodes |
2597 |
if not secondary_nodes: |
2598 |
raise errors.ProgrammerError("no secondary node but using " |
2599 |
"DT_REMOTE_RAID1 template")
|
2600 |
|
2601 |
target_node = secondary_nodes[0]
|
2602 |
# check memory requirements on the secondary node
|
2603 |
_CheckNodeFreeMemory(self.cfg, target_node, "failing over instance %s" % |
2604 |
instance.name, instance.memory) |
2605 |
|
2606 |
# check bridge existance
|
2607 |
brlist = [nic.bridge for nic in instance.nics] |
2608 |
if not rpc.call_bridges_exist(target_node, brlist): |
2609 |
raise errors.OpPrereqError("One or more target bridges %s does not" |
2610 |
" exist on destination node '%s'" %
|
2611 |
(brlist, target_node)) |
2612 |
|
2613 |
self.instance = instance
|
2614 |
|
2615 |
def Exec(self, feedback_fn): |
2616 |
"""Failover an instance.
|
2617 |
|
2618 |
The failover is done by shutting it down on its present node and
|
2619 |
starting it on the secondary.
|
2620 |
|
2621 |
"""
|
2622 |
instance = self.instance
|
2623 |
|
2624 |
source_node = instance.primary_node |
2625 |
target_node = instance.secondary_nodes[0]
|
2626 |
|
2627 |
feedback_fn("* checking disk consistency between source and target")
|
2628 |
for dev in instance.disks: |
2629 |
# for remote_raid1, these are md over drbd
|
2630 |
if not _CheckDiskConsistency(self.cfg, dev, target_node, False): |
2631 |
if not self.op.ignore_consistency: |
2632 |
raise errors.OpExecError("Disk %s is degraded on target node," |
2633 |
" aborting failover." % dev.iv_name)
|
2634 |
|
2635 |
feedback_fn("* shutting down instance on source node")
|
2636 |
logger.Info("Shutting down instance %s on node %s" %
|
2637 |
(instance.name, source_node)) |
2638 |
|
2639 |
if not rpc.call_instance_shutdown(source_node, instance): |
2640 |
if self.op.ignore_consistency: |
2641 |
logger.Error("Could not shutdown instance %s on node %s. Proceeding"
|
2642 |
" anyway. Please make sure node %s is down" %
|
2643 |
(instance.name, source_node, source_node)) |
2644 |
else:
|
2645 |
raise errors.OpExecError("Could not shutdown instance %s on node %s" % |
2646 |
(instance.name, source_node)) |
2647 |
|
2648 |
feedback_fn("* deactivating the instance's disks on source node")
|
2649 |
if not _ShutdownInstanceDisks(instance, self.cfg, ignore_primary=True): |
2650 |
raise errors.OpExecError("Can't shut down the instance's disks.") |
2651 |
|
2652 |
instance.primary_node = target_node |
2653 |
# distribute new instance config to the other nodes
|
2654 |
self.cfg.AddInstance(instance)
|
2655 |
|
2656 |
feedback_fn("* activating the instance's disks on target node")
|
2657 |
logger.Info("Starting instance %s on node %s" %
|
2658 |
(instance.name, target_node)) |
2659 |
|
2660 |
disks_ok, dummy = _AssembleInstanceDisks(instance, self.cfg,
|
2661 |
ignore_secondaries=True)
|
2662 |
if not disks_ok: |
2663 |
_ShutdownInstanceDisks(instance, self.cfg)
|
2664 |
raise errors.OpExecError("Can't activate the instance's disks") |
2665 |
|
2666 |
feedback_fn("* starting the instance on the target node")
|
2667 |
if not rpc.call_instance_start(target_node, instance, None): |
2668 |
_ShutdownInstanceDisks(instance, self.cfg)
|
2669 |
raise errors.OpExecError("Could not start instance %s on node %s." % |
2670 |
(instance.name, target_node)) |
2671 |
|
2672 |
|
2673 |
def _CreateBlockDevOnPrimary(cfg, node, instance, device, info): |
2674 |
"""Create a tree of block devices on the primary node.
|
2675 |
|
2676 |
This always creates all devices.
|
2677 |
|
2678 |
"""
|
2679 |
if device.children:
|
2680 |
for child in device.children: |
2681 |
if not _CreateBlockDevOnPrimary(cfg, node, instance, child, info): |
2682 |
return False |
2683 |
|
2684 |
cfg.SetDiskID(device, node) |
2685 |
new_id = rpc.call_blockdev_create(node, device, device.size, |
2686 |
instance.name, True, info)
|
2687 |
if not new_id: |
2688 |
return False |
2689 |
if device.physical_id is None: |
2690 |
device.physical_id = new_id |
2691 |
return True |
2692 |
|
2693 |
|
2694 |
def _CreateBlockDevOnSecondary(cfg, node, instance, device, force, info): |
2695 |
"""Create a tree of block devices on a secondary node.
|
2696 |
|
2697 |
If this device type has to be created on secondaries, create it and
|
2698 |
all its children.
|
2699 |
|
2700 |
If not, just recurse to children keeping the same 'force' value.
|
2701 |
|
2702 |
"""
|
2703 |
if device.CreateOnSecondary():
|
2704 |
force = True
|
2705 |
if device.children:
|
2706 |
for child in device.children: |
2707 |
if not _CreateBlockDevOnSecondary(cfg, node, instance, |
2708 |
child, force, info): |
2709 |
return False |
2710 |
|
2711 |
if not force: |
2712 |
return True |
2713 |
cfg.SetDiskID(device, node) |
2714 |
new_id = rpc.call_blockdev_create(node, device, device.size, |
2715 |
instance.name, False, info)
|
2716 |
if not new_id: |
2717 |
return False |
2718 |
if device.physical_id is None: |
2719 |
device.physical_id = new_id |
2720 |
return True |
2721 |
|
2722 |
|
2723 |
def _GenerateUniqueNames(cfg, exts): |
2724 |
"""Generate a suitable LV name.
|
2725 |
|
2726 |
This will generate a logical volume name for the given instance.
|
2727 |
|
2728 |
"""
|
2729 |
results = [] |
2730 |
for val in exts: |
2731 |
new_id = cfg.GenerateUniqueID() |
2732 |
results.append("%s%s" % (new_id, val))
|
2733 |
return results
|
2734 |
|
2735 |
|
2736 |
def _GenerateMDDRBDBranch(cfg, primary, secondary, size, names): |
2737 |
"""Generate a drbd device complete with its children.
|
2738 |
|
2739 |
"""
|
2740 |
port = cfg.AllocatePort() |
2741 |
vgname = cfg.GetVGName() |
2742 |
dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, |
2743 |
logical_id=(vgname, names[0]))
|
2744 |
dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
|
2745 |
logical_id=(vgname, names[1]))
|
2746 |
drbd_dev = objects.Disk(dev_type=constants.LD_DRBD7, size=size, |
2747 |
logical_id = (primary, secondary, port), |
2748 |
children = [dev_data, dev_meta]) |
2749 |
return drbd_dev
|
2750 |
|
2751 |
|
2752 |
def _GenerateDRBD8Branch(cfg, primary, secondary, size, names, iv_name): |
2753 |
"""Generate a drbd8 device complete with its children.
|
2754 |
|
2755 |
"""
|
2756 |
port = cfg.AllocatePort() |
2757 |
vgname = cfg.GetVGName() |
2758 |
dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, |
2759 |
logical_id=(vgname, names[0]))
|
2760 |
dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
|
2761 |
logical_id=(vgname, names[1]))
|
2762 |
drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size, |
2763 |
logical_id = (primary, secondary, port), |
2764 |
children = [dev_data, dev_meta], |
2765 |
iv_name=iv_name) |
2766 |
return drbd_dev
|
2767 |
|
2768 |
|
2769 |
def _GenerateDiskTemplate(cfg, template_name, |
2770 |
instance_name, primary_node, |
2771 |
secondary_nodes, disk_sz, swap_sz): |
2772 |
"""Generate the entire disk layout for a given template type.
|
2773 |
|
2774 |
"""
|
2775 |
#TODO: compute space requirements
|
2776 |
|
2777 |
vgname = cfg.GetVGName() |
2778 |
if template_name == constants.DT_DISKLESS:
|
2779 |
disks = [] |
2780 |
elif template_name == constants.DT_PLAIN:
|
2781 |
if len(secondary_nodes) != 0: |
2782 |
raise errors.ProgrammerError("Wrong template configuration") |
2783 |
|
2784 |
names = _GenerateUniqueNames(cfg, [".sda", ".sdb"]) |
2785 |
sda_dev = objects.Disk(dev_type=constants.LD_LV, size=disk_sz, |
2786 |
logical_id=(vgname, names[0]),
|
2787 |
iv_name = "sda")
|
2788 |
sdb_dev = objects.Disk(dev_type=constants.LD_LV, size=swap_sz, |
2789 |
logical_id=(vgname, names[1]),
|
2790 |
iv_name = "sdb")
|
2791 |
disks = [sda_dev, sdb_dev] |
2792 |
elif template_name == constants.DT_DRBD8:
|
2793 |
if len(secondary_nodes) != 1: |
2794 |
raise errors.ProgrammerError("Wrong template configuration") |
2795 |
remote_node = secondary_nodes[0]
|
2796 |
names = _GenerateUniqueNames(cfg, [".sda_data", ".sda_meta", |
2797 |
".sdb_data", ".sdb_meta"]) |
2798 |
drbd_sda_dev = _GenerateDRBD8Branch(cfg, primary_node, remote_node, |
2799 |
disk_sz, names[0:2], "sda") |
2800 |
drbd_sdb_dev = _GenerateDRBD8Branch(cfg, primary_node, remote_node, |
2801 |
swap_sz, names[2:4], "sdb") |
2802 |
disks = [drbd_sda_dev, drbd_sdb_dev] |
2803 |
else:
|
2804 |
raise errors.ProgrammerError("Invalid disk template '%s'" % template_name) |
2805 |
return disks
|
2806 |
|
2807 |
|
2808 |
def _GetInstanceInfoText(instance): |
2809 |
"""Compute that text that should be added to the disk's metadata.
|
2810 |
|
2811 |
"""
|
2812 |
return "originstname+%s" % instance.name |
2813 |
|
2814 |
|
2815 |
def _CreateDisks(cfg, instance): |
2816 |
"""Create all disks for an instance.
|
2817 |
|
2818 |
This abstracts away some work from AddInstance.
|
2819 |
|
2820 |
Args:
|
2821 |
instance: the instance object
|
2822 |
|
2823 |
Returns:
|
2824 |
True or False showing the success of the creation process
|
2825 |
|
2826 |
"""
|
2827 |
info = _GetInstanceInfoText(instance) |
2828 |
|
2829 |
for device in instance.disks: |
2830 |
logger.Info("creating volume %s for instance %s" %
|
2831 |
(device.iv_name, instance.name)) |
2832 |
#HARDCODE
|
2833 |
for secondary_node in instance.secondary_nodes: |
2834 |
if not _CreateBlockDevOnSecondary(cfg, secondary_node, instance, |
2835 |
device, False, info):
|
2836 |
logger.Error("failed to create volume %s (%s) on secondary node %s!" %
|
2837 |
(device.iv_name, device, secondary_node)) |
2838 |
return False |
2839 |
#HARDCODE
|
2840 |
if not _CreateBlockDevOnPrimary(cfg, instance.primary_node, |
2841 |
instance, device, info): |
2842 |
logger.Error("failed to create volume %s on primary!" %
|
2843 |
device.iv_name) |
2844 |
return False |
2845 |
return True |
2846 |
|
2847 |
|
2848 |
def _RemoveDisks(instance, cfg): |
2849 |
"""Remove all disks for an instance.
|
2850 |
|
2851 |
This abstracts away some work from `AddInstance()` and
|
2852 |
`RemoveInstance()`. Note that in case some of the devices couldn't
|
2853 |
be removed, the removal will continue with the other ones (compare
|
2854 |
with `_CreateDisks()`).
|
2855 |
|
2856 |
Args:
|
2857 |
instance: the instance object
|
2858 |
|
2859 |
Returns:
|
2860 |
True or False showing the success of the removal proces
|
2861 |
|
2862 |
"""
|
2863 |
logger.Info("removing block devices for instance %s" % instance.name)
|
2864 |
|
2865 |
result = True
|
2866 |
for device in instance.disks: |
2867 |
for node, disk in device.ComputeNodeTree(instance.primary_node): |
2868 |
cfg.SetDiskID(disk, node) |
2869 |
if not rpc.call_blockdev_remove(node, disk): |
2870 |
logger.Error("could not remove block device %s on node %s,"
|
2871 |
" continuing anyway" %
|
2872 |
(device.iv_name, node)) |
2873 |
result = False
|
2874 |
return result
|
2875 |
|
2876 |
|
2877 |
class LUCreateInstance(LogicalUnit): |
2878 |
"""Create an instance.
|
2879 |
|
2880 |
"""
|
2881 |
HPATH = "instance-add"
|
2882 |
HTYPE = constants.HTYPE_INSTANCE |
2883 |
_OP_REQP = ["instance_name", "mem_size", "disk_size", "pnode", |
2884 |
"disk_template", "swap_size", "mode", "start", "vcpus", |
2885 |
"wait_for_sync", "ip_check", "mac"] |
2886 |
|
2887 |
def BuildHooksEnv(self): |
2888 |
"""Build hooks env.
|
2889 |
|
2890 |
This runs on master, primary and secondary nodes of the instance.
|
2891 |
|
2892 |
"""
|
2893 |
env = { |
2894 |
"INSTANCE_DISK_TEMPLATE": self.op.disk_template, |
2895 |
"INSTANCE_DISK_SIZE": self.op.disk_size, |
2896 |
"INSTANCE_SWAP_SIZE": self.op.swap_size, |
2897 |
"INSTANCE_ADD_MODE": self.op.mode, |
2898 |
} |
2899 |
if self.op.mode == constants.INSTANCE_IMPORT: |
2900 |
env["INSTANCE_SRC_NODE"] = self.op.src_node |
2901 |
env["INSTANCE_SRC_PATH"] = self.op.src_path |
2902 |
env["INSTANCE_SRC_IMAGE"] = self.src_image |
2903 |
|
2904 |
env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
|
2905 |
primary_node=self.op.pnode,
|
2906 |
secondary_nodes=self.secondaries,
|
2907 |
status=self.instance_status,
|
2908 |
os_type=self.op.os_type,
|
2909 |
memory=self.op.mem_size,
|
2910 |
vcpus=self.op.vcpus,
|
2911 |
nics=[(self.inst_ip, self.op.bridge, self.op.mac)], |
2912 |
)) |
2913 |
|
2914 |
nl = ([self.sstore.GetMasterNode(), self.op.pnode] + |
2915 |
self.secondaries)
|
2916 |
return env, nl, nl
|
2917 |
|
2918 |
|
2919 |
def CheckPrereq(self): |
2920 |
"""Check prerequisites.
|
2921 |
|
2922 |
"""
|
2923 |
for attr in ["kernel_path", "initrd_path", "hvm_boot_order"]: |
2924 |
if not hasattr(self.op, attr): |
2925 |
setattr(self.op, attr, None) |
2926 |
|
2927 |
if self.op.mode not in (constants.INSTANCE_CREATE, |
2928 |
constants.INSTANCE_IMPORT): |
2929 |
raise errors.OpPrereqError("Invalid instance creation mode '%s'" % |
2930 |
self.op.mode)
|
2931 |
|
2932 |
if (not self.cfg.GetVGName() and |
2933 |
self.op.disk_template not in constants.DTS_NOT_LVM): |
2934 |
raise errors.OpPrereqError("Cluster does not support lvm-based" |
2935 |
" instances")
|
2936 |
|
2937 |
if self.op.mode == constants.INSTANCE_IMPORT: |
2938 |
src_node = getattr(self.op, "src_node", None) |
2939 |
src_path = getattr(self.op, "src_path", None) |
2940 |
if src_node is None or src_path is None: |
2941 |
raise errors.OpPrereqError("Importing an instance requires source" |
2942 |
" node and path options")
|
2943 |
src_node_full = self.cfg.ExpandNodeName(src_node)
|
2944 |
if src_node_full is None: |
2945 |
raise errors.OpPrereqError("Unknown source node '%s'" % src_node) |
2946 |
self.op.src_node = src_node = src_node_full
|
2947 |
|
2948 |
if not os.path.isabs(src_path): |
2949 |
raise errors.OpPrereqError("The source path must be absolute") |
2950 |
|
2951 |
export_info = rpc.call_export_info(src_node, src_path) |
2952 |
|
2953 |
if not export_info: |
2954 |
raise errors.OpPrereqError("No export found in dir %s" % src_path) |
2955 |
|
2956 |
if not export_info.has_section(constants.INISECT_EXP): |
2957 |
raise errors.ProgrammerError("Corrupted export config") |
2958 |
|
2959 |
ei_version = export_info.get(constants.INISECT_EXP, 'version')
|
2960 |
if (int(ei_version) != constants.EXPORT_VERSION): |
2961 |
raise errors.OpPrereqError("Wrong export version %s (wanted %d)" % |
2962 |
(ei_version, constants.EXPORT_VERSION)) |
2963 |
|
2964 |
if int(export_info.get(constants.INISECT_INS, 'disk_count')) > 1: |
2965 |
raise errors.OpPrereqError("Can't import instance with more than" |
2966 |
" one data disk")
|
2967 |
|
2968 |
# FIXME: are the old os-es, disk sizes, etc. useful?
|
2969 |
self.op.os_type = export_info.get(constants.INISECT_EXP, 'os') |
2970 |
diskimage = os.path.join(src_path, export_info.get(constants.INISECT_INS, |
2971 |
'disk0_dump'))
|
2972 |
self.src_image = diskimage
|
2973 |
else: # INSTANCE_CREATE |
2974 |
if getattr(self.op, "os_type", None) is None: |
2975 |
raise errors.OpPrereqError("No guest OS specified") |
2976 |
|
2977 |
# check primary node
|
2978 |
pnode = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.pnode)) |
2979 |
if pnode is None: |
2980 |
raise errors.OpPrereqError("Primary node '%s' is unknown" % |
2981 |
self.op.pnode)
|
2982 |
self.op.pnode = pnode.name
|
2983 |
self.pnode = pnode
|
2984 |
self.secondaries = []
|
2985 |
# disk template and mirror node verification
|
2986 |
if self.op.disk_template not in constants.DISK_TEMPLATES: |
2987 |
raise errors.OpPrereqError("Invalid disk template name") |
2988 |
|
2989 |
if self.op.disk_template in constants.DTS_NET_MIRROR: |
2990 |
if getattr(self.op, "snode", None) is None: |
2991 |
raise errors.OpPrereqError("The networked disk templates need" |
2992 |
" a mirror node")
|
2993 |
|
2994 |
snode_name = self.cfg.ExpandNodeName(self.op.snode) |
2995 |
if snode_name is None: |
2996 |
raise errors.OpPrereqError("Unknown secondary node '%s'" % |
2997 |
self.op.snode)
|
2998 |
elif snode_name == pnode.name:
|
2999 |
raise errors.OpPrereqError("The secondary node cannot be" |
3000 |
" the primary node.")
|
3001 |
self.secondaries.append(snode_name)
|
3002 |
|
3003 |
# Required free disk space as a function of disk and swap space
|
3004 |
req_size_dict = { |
3005 |
constants.DT_DISKLESS: None,
|
3006 |
constants.DT_PLAIN: self.op.disk_size + self.op.swap_size, |
3007 |
# 256 MB are added for drbd metadata, 128MB for each drbd device
|
3008 |
constants.DT_DRBD8: self.op.disk_size + self.op.swap_size + 256, |
3009 |
} |
3010 |
|
3011 |
if self.op.disk_template not in req_size_dict: |
3012 |
raise errors.ProgrammerError("Disk template '%s' size requirement" |
3013 |
" is unknown" % self.op.disk_template) |
3014 |
|
3015 |
req_size = req_size_dict[self.op.disk_template]
|
3016 |
|
3017 |
# Check lv size requirements
|
3018 |
if req_size is not None: |
3019 |
nodenames = [pnode.name] + self.secondaries
|
3020 |
nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName())
|
3021 |
for node in nodenames: |
3022 |
info = nodeinfo.get(node, None)
|
3023 |
if not info: |
3024 |
raise errors.OpPrereqError("Cannot get current information" |
3025 |
" from node '%s'" % nodeinfo)
|
3026 |
vg_free = info.get('vg_free', None) |
3027 |
if not isinstance(vg_free, int): |
3028 |
raise errors.OpPrereqError("Can't compute free disk space on" |
3029 |
" node %s" % node)
|
3030 |
if req_size > info['vg_free']: |
3031 |
raise errors.OpPrereqError("Not enough disk space on target node %s." |
3032 |
" %d MB available, %d MB required" %
|
3033 |
(node, info['vg_free'], req_size))
|
3034 |
|
3035 |
# os verification
|
3036 |
os_obj = rpc.call_os_get(pnode.name, self.op.os_type)
|
3037 |
if not os_obj: |
3038 |
raise errors.OpPrereqError("OS '%s' not in supported os list for" |
3039 |
" primary node" % self.op.os_type) |
3040 |
|
3041 |
if self.op.kernel_path == constants.VALUE_NONE: |
3042 |
raise errors.OpPrereqError("Can't set instance kernel to none") |
3043 |
|
3044 |
# instance verification
|
3045 |
hostname1 = utils.HostInfo(self.op.instance_name)
|
3046 |
|
3047 |
self.op.instance_name = instance_name = hostname1.name
|
3048 |
instance_list = self.cfg.GetInstanceList()
|
3049 |
if instance_name in instance_list: |
3050 |
raise errors.OpPrereqError("Instance '%s' is already in the cluster" % |
3051 |
instance_name) |
3052 |
|
3053 |
ip = getattr(self.op, "ip", None) |
3054 |
if ip is None or ip.lower() == "none": |
3055 |
inst_ip = None
|
3056 |
elif ip.lower() == "auto": |
3057 |
inst_ip = hostname1.ip |
3058 |
else:
|
3059 |
if not utils.IsValidIP(ip): |
3060 |
raise errors.OpPrereqError("given IP address '%s' doesn't look" |
3061 |
" like a valid IP" % ip)
|
3062 |
inst_ip = ip |
3063 |
self.inst_ip = inst_ip
|
3064 |
|
3065 |
if self.op.start and not self.op.ip_check: |
3066 |
raise errors.OpPrereqError("Cannot ignore IP address conflicts when" |
3067 |
" adding an instance in start mode")
|
3068 |
|
3069 |
if self.op.ip_check: |
3070 |
if utils.TcpPing(hostname1.ip, constants.DEFAULT_NODED_PORT):
|
3071 |
raise errors.OpPrereqError("IP %s of instance %s already in use" % |
3072 |
(hostname1.ip, instance_name)) |
3073 |
|
3074 |
# MAC address verification
|
3075 |
if self.op.mac != "auto": |
3076 |
if not utils.IsValidMac(self.op.mac.lower()): |
3077 |
raise errors.OpPrereqError("invalid MAC address specified: %s" % |
3078 |
self.op.mac)
|
3079 |
|
3080 |
# bridge verification
|
3081 |
bridge = getattr(self.op, "bridge", None) |
3082 |
if bridge is None: |
3083 |
self.op.bridge = self.cfg.GetDefBridge() |
3084 |
else:
|
3085 |
self.op.bridge = bridge
|
3086 |
|
3087 |
if not rpc.call_bridges_exist(self.pnode.name, [self.op.bridge]): |
3088 |
raise errors.OpPrereqError("target bridge '%s' does not exist on" |
3089 |
" destination node '%s'" %
|
3090 |
(self.op.bridge, pnode.name))
|
3091 |
|
3092 |
# boot order verification
|
3093 |
if self.op.hvm_boot_order is not None: |
3094 |
if len(self.op.hvm_boot_order.strip("acdn")) != 0: |
3095 |
raise errors.OpPrereqError("invalid boot order specified," |
3096 |
" must be one or more of [acdn]")
|
3097 |
|
3098 |
if self.op.start: |
3099 |
self.instance_status = 'up' |
3100 |
else:
|
3101 |
self.instance_status = 'down' |
3102 |
|
3103 |
def Exec(self, feedback_fn): |
3104 |
"""Create and add the instance to the cluster.
|
3105 |
|
3106 |
"""
|
3107 |
instance = self.op.instance_name
|
3108 |
pnode_name = self.pnode.name
|
3109 |
|
3110 |
if self.op.mac == "auto": |
3111 |
mac_address = self.cfg.GenerateMAC()
|
3112 |
else:
|
3113 |
mac_address = self.op.mac
|
3114 |
|
3115 |
nic = objects.NIC(bridge=self.op.bridge, mac=mac_address)
|
3116 |
if self.inst_ip is not None: |
3117 |
nic.ip = self.inst_ip
|
3118 |
|
3119 |
ht_kind = self.sstore.GetHypervisorType()
|
3120 |
if ht_kind in constants.HTS_REQ_PORT: |
3121 |
network_port = self.cfg.AllocatePort()
|
3122 |
else:
|
3123 |
network_port = None
|
3124 |
|
3125 |
disks = _GenerateDiskTemplate(self.cfg,
|
3126 |
self.op.disk_template,
|
3127 |
instance, pnode_name, |
3128 |
self.secondaries, self.op.disk_size, |
3129 |
self.op.swap_size)
|
3130 |
|
3131 |
iobj = objects.Instance(name=instance, os=self.op.os_type,
|
3132 |
primary_node=pnode_name, |
3133 |
memory=self.op.mem_size,
|
3134 |
vcpus=self.op.vcpus,
|
3135 |
nics=[nic], disks=disks, |
3136 |
disk_template=self.op.disk_template,
|
3137 |
status=self.instance_status,
|
3138 |
network_port=network_port, |
3139 |
kernel_path=self.op.kernel_path,
|
3140 |
initrd_path=self.op.initrd_path,
|
3141 |
hvm_boot_order=self.op.hvm_boot_order,
|
3142 |
) |
3143 |
|
3144 |
feedback_fn("* creating instance disks...")
|
3145 |
if not _CreateDisks(self.cfg, iobj): |
3146 |
_RemoveDisks(iobj, self.cfg)
|
3147 |
raise errors.OpExecError("Device creation failed, reverting...") |
3148 |
|
3149 |
feedback_fn("adding instance %s to cluster config" % instance)
|
3150 |
|
3151 |
self.cfg.AddInstance(iobj)
|
3152 |
|
3153 |
if self.op.wait_for_sync: |
3154 |
disk_abort = not _WaitForSync(self.cfg, iobj, self.proc) |
3155 |
elif iobj.disk_template in constants.DTS_NET_MIRROR: |
3156 |
# make sure the disks are not degraded (still sync-ing is ok)
|
3157 |
time.sleep(15)
|
3158 |
feedback_fn("* checking mirrors status")
|
3159 |
disk_abort = not _WaitForSync(self.cfg, iobj, self.proc, oneshot=True) |
3160 |
else:
|
3161 |
disk_abort = False
|
3162 |
|
3163 |
if disk_abort:
|
3164 |
_RemoveDisks(iobj, self.cfg)
|
3165 |
self.cfg.RemoveInstance(iobj.name)
|
3166 |
raise errors.OpExecError("There are some degraded disks for" |
3167 |
" this instance")
|
3168 |
|
3169 |
feedback_fn("creating os for instance %s on node %s" %
|
3170 |
(instance, pnode_name)) |
3171 |
|
3172 |
if iobj.disk_template != constants.DT_DISKLESS:
|
3173 |
if self.op.mode == constants.INSTANCE_CREATE: |
3174 |
feedback_fn("* running the instance OS create scripts...")
|
3175 |
if not rpc.call_instance_os_add(pnode_name, iobj, "sda", "sdb"): |
3176 |
raise errors.OpExecError("could not add os for instance %s" |
3177 |
" on node %s" %
|
3178 |
(instance, pnode_name)) |
3179 |
|
3180 |
elif self.op.mode == constants.INSTANCE_IMPORT: |
3181 |
feedback_fn("* running the instance OS import scripts...")
|
3182 |
src_node = self.op.src_node
|
3183 |
src_image = self.src_image
|
3184 |
if not rpc.call_instance_os_import(pnode_name, iobj, "sda", "sdb", |
3185 |
src_node, src_image): |
3186 |
raise errors.OpExecError("Could not import os for instance" |
3187 |
" %s on node %s" %
|
3188 |
(instance, pnode_name)) |
3189 |
else:
|
3190 |
# also checked in the prereq part
|
3191 |
raise errors.ProgrammerError("Unknown OS initialization mode '%s'" |
3192 |
% self.op.mode)
|
3193 |
|
3194 |
if self.op.start: |
3195 |
logger.Info("starting instance %s on node %s" % (instance, pnode_name))
|
3196 |
feedback_fn("* starting instance...")
|
3197 |
if not rpc.call_instance_start(pnode_name, iobj, None): |
3198 |
raise errors.OpExecError("Could not start instance") |
3199 |
|
3200 |
|
3201 |
class LUConnectConsole(NoHooksLU): |
3202 |
"""Connect to an instance's console.
|
3203 |
|
3204 |
This is somewhat special in that it returns the command line that
|
3205 |
you need to run on the master node in order to connect to the
|
3206 |
console.
|
3207 |
|
3208 |
"""
|
3209 |
_OP_REQP = ["instance_name"]
|
3210 |
|
3211 |
def CheckPrereq(self): |
3212 |
"""Check prerequisites.
|
3213 |
|
3214 |
This checks that the instance is in the cluster.
|
3215 |
|
3216 |
"""
|
3217 |
instance = self.cfg.GetInstanceInfo(
|
3218 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
3219 |
if instance is None: |
3220 |
raise errors.OpPrereqError("Instance '%s' not known" % |
3221 |
self.op.instance_name)
|
3222 |
self.instance = instance
|
3223 |
|
3224 |
def Exec(self, feedback_fn): |
3225 |
"""Connect to the console of an instance
|
3226 |
|
3227 |
"""
|
3228 |
instance = self.instance
|
3229 |
node = instance.primary_node |
3230 |
|
3231 |
node_insts = rpc.call_instance_list([node])[node] |
3232 |
if node_insts is False: |
3233 |
raise errors.OpExecError("Can't connect to node %s." % node) |
3234 |
|
3235 |
if instance.name not in node_insts: |
3236 |
raise errors.OpExecError("Instance %s is not running." % instance.name) |
3237 |
|
3238 |
logger.Debug("connecting to console of %s on %s" % (instance.name, node))
|
3239 |
|
3240 |
hyper = hypervisor.GetHypervisor() |
3241 |
console_cmd = hyper.GetShellCommandForConsole(instance) |
3242 |
|
3243 |
# build ssh cmdline
|
3244 |
return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True) |
3245 |
|
3246 |
|
3247 |
class LUReplaceDisks(LogicalUnit): |
3248 |
"""Replace the disks of an instance.
|
3249 |
|
3250 |
"""
|
3251 |
HPATH = "mirrors-replace"
|
3252 |
HTYPE = constants.HTYPE_INSTANCE |
3253 |
_OP_REQP = ["instance_name", "mode", "disks"] |
3254 |
|
3255 |
def BuildHooksEnv(self): |
3256 |
"""Build hooks env.
|
3257 |
|
3258 |
This runs on the master, the primary and all the secondaries.
|
3259 |
|
3260 |
"""
|
3261 |
env = { |
3262 |
"MODE": self.op.mode, |
3263 |
"NEW_SECONDARY": self.op.remote_node, |
3264 |
"OLD_SECONDARY": self.instance.secondary_nodes[0], |
3265 |
} |
3266 |
env.update(_BuildInstanceHookEnvByObject(self.instance))
|
3267 |
nl = [ |
3268 |
self.sstore.GetMasterNode(),
|
3269 |
self.instance.primary_node,
|
3270 |
] |
3271 |
if self.op.remote_node is not None: |
3272 |
nl.append(self.op.remote_node)
|
3273 |
return env, nl, nl
|
3274 |
|
3275 |
def CheckPrereq(self): |
3276 |
"""Check prerequisites.
|
3277 |
|
3278 |
This checks that the instance is in the cluster.
|
3279 |
|
3280 |
"""
|
3281 |
instance = self.cfg.GetInstanceInfo(
|
3282 |
self.cfg.ExpandInstanceName(self.op.instance_name)) |
3283 |
if instance is None: |
3284 |
raise errors.OpPrereqError("Instance '%s' not known" % |
3285 |
self.op.instance_name)
|
3286 |
self.instance = instance
|
3287 |
self.op.instance_name = instance.name
|
3288 |
|
3289 |
if instance.disk_template not in constants.DTS_NET_MIRROR: |
3290 |
raise errors.OpPrereqError("Instance's disk layout is not" |
3291 |
" network mirrored.")
|
3292 |
|
3293 |
if len(instance.secondary_nodes) != 1: |
3294 |
raise errors.OpPrereqError("The instance has a strange layout," |
3295 |
" expected one secondary but found %d" %
|
3296 |
len(instance.secondary_nodes))
|
3297 |
|
3298 |
self.sec_node = instance.secondary_nodes[0] |
3299 |
|
3300 |
remote_node = getattr(self.op, "remote_node", None) |
3301 |
if remote_node is not None: |
3302 |
remote_node = self.cfg.ExpandNodeName(remote_node)
|
3303 |
if remote_node is None: |
3304 |
raise errors.OpPrereqError("Node '%s' not known" % |
3305 |
self.op.remote_node)
|
3306 |
self.remote_node_info = self.cfg.GetNodeInfo(remote_node) |
3307 |
else:
|
3308 |
self.remote_node_info = None |
3309 |
if remote_node == instance.primary_node:
|
3310 |
raise errors.OpPrereqError("The specified node is the primary node of" |
3311 |
" the instance.")
|
3312 |
elif remote_node == self.sec_node: |
3313 |
if self.op.mode == constants.REPLACE_DISK_SEC: |
3314 |
# this is for DRBD8, where we can't execute the same mode of
|
3315 |
# replacement as for drbd7 (no different port allocated)
|
3316 |
raise errors.OpPrereqError("Same secondary given, cannot execute" |
3317 |
" replacement")
|
3318 |
# the user gave the current secondary, switch to
|
3319 |
# 'no-replace-secondary' mode for drbd7
|
3320 |
remote_node = None
|
3321 |
if (instance.disk_template == constants.DT_REMOTE_RAID1 and |
3322 |
self.op.mode != constants.REPLACE_DISK_ALL):
|
3323 |
raise errors.OpPrereqError("Template 'remote_raid1' only allows all" |
3324 |
" disks replacement, not individual ones")
|
3325 |
if instance.disk_template == constants.DT_DRBD8:
|
3326 |
if (self.op.mode == constants.REPLACE_DISK_ALL and |
3327 |
remote_node is not None): |
3328 |
# switch to replace secondary mode
|
3329 |
self.op.mode = constants.REPLACE_DISK_SEC
|
3330 |
|
3331 |
if self.op.mode == constants.REPLACE_DISK_ALL: |
3332 |
raise errors.OpPrereqError("Template 'drbd' only allows primary or" |
3333 |
" secondary disk replacement, not"
|
3334 |
" both at once")
|
3335 |
elif self.op.mode == constants.REPLACE_DISK_PRI: |
3336 |
if remote_node is not None: |
3337 |
raise errors.OpPrereqError("Template 'drbd' does not allow changing" |
3338 |
" the secondary while doing a primary"
|
3339 |
" node disk replacement")
|
3340 |
self.tgt_node = instance.primary_node
|
3341 |
self.oth_node = instance.secondary_nodes[0] |
3342 |
elif self.op.mode == constants.REPLACE_DISK_SEC: |
3343 |
self.new_node = remote_node # this can be None, in which case |
3344 |
# we don't change the secondary
|
3345 |
self.tgt_node = instance.secondary_nodes[0] |
3346 |
self.oth_node = instance.primary_node
|
3347 |
else:
|
3348 |
raise errors.ProgrammerError("Unhandled disk replace mode") |
3349 |
|
3350 |
for name in self.op.disks: |
3351 |
if instance.FindDisk(name) is None: |
3352 |
raise errors.OpPrereqError("Disk '%s' not found for instance '%s'" % |
3353 |
(name, instance.name)) |
3354 |
self.op.remote_node = remote_node
|
3355 |
|
3356 |
def _ExecRR1(self, feedback_fn): |
3357 |
"""Replace the disks of an instance.
|
3358 |
|
3359 |
"""
|
3360 |
instance = self.instance
|
3361 |
iv_names = {} |
3362 |
# start of work
|
3363 |
if self.op.remote_node is None: |
3364 |
remote_node = self.sec_node
|
3365 |
else:
|
3366 |
remote_node = self.op.remote_node
|
3367 |
cfg = self.cfg
|
3368 |
for dev in instance.disks: |
3369 |
size = dev.size |
3370 |
lv_names = [".%s_%s" % (dev.iv_name, suf) for suf in ["data", "meta"]] |
3371 |
names = _GenerateUniqueNames(cfg, lv_names) |
3372 |
new_drbd = _GenerateMDDRBDBranch(cfg, instance.primary_node, |
3373 |
remote_node, size, names) |
3374 |
iv_names[dev.iv_name] = (dev, dev.children[0], new_drbd)
|
3375 |
logger.Info("adding new mirror component on secondary for %s" %
|
3376 |
dev.iv_name) |
3377 |
#HARDCODE
|
3378 |
if not _CreateBlockDevOnSecondary(cfg, remote_node, instance, |
3379 |
new_drbd, False,
|
3380 |
_GetInstanceInfoText(instance)): |
3381 |
raise errors.OpExecError("Failed to create new component on secondary" |
3382 |
" node %s. Full abort, cleanup manually!" %
|
3383 |
remote_node) |
3384 |
|
3385 |
logger.Info("adding new mirror component on primary")
|
3386 |
#HARDCODE
|
3387 |
if not _CreateBlockDevOnPrimary(cfg, instance.primary_node, |
3388 |
instance, new_drbd, |
3389 |
_GetInstanceInfoText(instance)): |
3390 |
# remove secondary dev
|
3391 |
cfg.SetDiskID(new_drbd, remote_node) |
3392 |
rpc.call_blockdev_remove(remote_node, new_drbd) |
3393 |
raise errors.OpExecError("Failed to create volume on primary!" |
3394 |
" Full abort, cleanup manually!!")
|
3395 |
|
3396 |
# the device exists now
|
3397 |
# call the primary node to add the mirror to md
|
3398 |
logger.Info("adding new mirror component to md")
|
3399 |
if not rpc.call_blockdev_addchildren(instance.primary_node, dev, |
3400 |
[new_drbd]): |
3401 |
logger.Error("Can't add mirror compoment to md!")
|
3402 |
cfg.SetDiskID(new_drbd, remote_node) |
3403 |
if not rpc.call_blockdev_remove(remote_node, new_drbd): |
3404 |
logger.Error("Can't rollback on secondary")
|
3405 |
cfg.SetDiskID(new_drbd, instance.primary_node) |
3406 |
if not rpc.call_blockdev_remove(instance.primary_node, new_drbd): |
3407 |
logger.Error("Can't rollback on primary")
|
3408 |
raise errors.OpExecError("Full abort, cleanup manually!!") |
3409 |
|
3410 |
dev.children.append(new_drbd) |
3411 |
cfg.AddInstance(instance) |
3412 |
|
3413 |
# this can fail as the old devices are degraded and _WaitForSync
|
3414 |
# does a combined result over all disks, so we don't check its
|
3415 |
# return value
|
3416 |
_WaitForSync(cfg, instance, self.proc, unlock=True) |
3417 |
|
3418 |
# so check manually all the devices
|
3419 |
for name in iv_names: |
3420 |
dev, child, new_drbd = iv_names[name] |
3421 |
cfg.SetDiskID(dev, instance.primary_node) |
3422 |
is_degr = rpc.call_blockdev_find(instance.primary_node, dev)[5]
|
3423 |
if is_degr:
|
3424 |
raise errors.OpExecError("MD device %s is degraded!" % name) |
3425 |
cfg.SetDiskID(new_drbd, instance.primary_node) |
3426 |
is_degr = rpc.call_blockdev_find(instance.primary_node, new_drbd)[5]
|
3427 |
if is_degr:
|
3428 |
raise errors.OpExecError("New drbd device %s is degraded!" % name) |
3429 |
|
3430 |
for name in iv_names: |
3431 |
dev, child, new_drbd = iv_names[name] |
3432 |
logger.Info("remove mirror %s component" % name)
|
3433 |
cfg.SetDiskID(dev, instance.primary_node) |
3434 |
if not rpc.call_blockdev_removechildren(instance.primary_node, |
3435 |
dev, [child]): |
3436 |
logger.Error("Can't remove child from mirror, aborting"
|
3437 |
" *this device cleanup*.\nYou need to cleanup manually!!")
|
3438 |
continue
|
3439 |
|
3440 |
for node in child.logical_id[:2]: |
3441 |
logger.Info("remove child device on %s" % node)
|
3442 |
cfg.SetDiskID(child, node) |
3443 |
if not rpc.call_blockdev_remove(node, child): |
3444 |
logger.Error("Warning: failed to remove device from node %s,"
|
3445 |
" continuing operation." % node)
|
3446 |
|
3447 |
dev.children.remove(child) |
3448 |
|
3449 |
cfg.AddInstance(instance) |
3450 |
|
3451 |
def _ExecD8DiskOnly(self, feedback_fn): |
3452 |
"""Replace a disk on the primary or secondary for dbrd8.
|
3453 |
|
3454 |
The algorithm for replace is quite complicated:
|
3455 |
- for each disk to be replaced:
|
3456 |
- create new LVs on the target node with unique names
|
3457 |
- detach old LVs from the drbd device
|
3458 |
- rename old LVs to name_replaced.<time_t>
|
3459 |
- rename new LVs to old LVs
|
3460 |
- attach the new LVs (with the old names now) to the drbd device
|
3461 |
- wait for sync across all devices
|
3462 |
- for each modified disk:
|
3463 |
- remove old LVs (which have the name name_replaces.<time_t>)
|
3464 |
|
3465 |
Failures are not very well handled.
|
3466 |
|
3467 |
"""
|
3468 |
steps_total = 6
|
3469 |
warning, info = (self.proc.LogWarning, self.proc.LogInfo) |
3470 |
instance = self.instance
|
3471 |
iv_names = {} |
3472 |
vgname = self.cfg.GetVGName()
|
3473 |
# start of work
|
3474 |
cfg = self.cfg
|
3475 |
tgt_node = self.tgt_node
|
3476 |
oth_node = self.oth_node
|
3477 |
|
3478 |
# Step: check device activation
|
3479 |
self.proc.LogStep(1, steps_total, "check device existence") |
3480 |
info("checking volume groups")
|
3481 |
my_vg = cfg.GetVGName() |
3482 |
results = rpc.call_vg_list([oth_node, tgt_node]) |
3483 |
if not results: |
3484 |
raise errors.OpExecError("Can't list volume groups on the nodes") |
3485 |
for node in oth_node, tgt_node: |
3486 |
res = results.get(node, False)
|
3487 |
if not res or my_vg not in res: |
3488 |
raise errors.OpExecError("Volume group '%s' not found on %s" % |
3489 |
(my_vg, node)) |
3490 |
for dev in instance.disks: |
3491 |
if not dev.iv_name in self.op.disks: |
3492 |
continue
|
3493 |
for node in tgt_node, oth_node: |
3494 |
info("checking %s on %s" % (dev.iv_name, node))
|
3495 |
cfg.SetDiskID(dev, node) |
3496 |
if not rpc.call_blockdev_find(node, dev): |
3497 |
raise errors.OpExecError("Can't find device %s on node %s" % |
3498 |
(dev.iv_name, node)) |
3499 |
|
3500 |
# Step: check other node consistency
|
3501 |
self.proc.LogStep(2, steps_total, "check peer consistency") |
3502 |
for dev in instance.disks: |
3503 |
if not dev.iv_name in self.op.disks: |
3504 |
continue
|
3505 |
info("checking %s consistency on %s" % (dev.iv_name, oth_node))
|
3506 |
if not _CheckDiskConsistency(self.cfg, dev, oth_node, |
3507 |
oth_node==instance.primary_node): |
3508 |
raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe" |
3509 |
" to replace disks on this node (%s)" %
|
3510 |
(oth_node, tgt_node)) |
3511 |
|
3512 |
# Step: create new storage
|
3513 |
self.proc.LogStep(3, steps_total, "allocate new storage") |
3514 |
for dev in instance.disks: |
3515 |
if not dev.iv_name in self.op.disks: |
3516 |
continue
|
3517 |
size = dev.size |
3518 |
cfg.SetDiskID(dev, tgt_node) |
3519 |
lv_names = [".%s_%s" % (dev.iv_name, suf) for suf in ["data", "meta"]] |
3520 |
names = _GenerateUniqueNames(cfg, lv_names) |
3521 |
lv_data = objects.Disk(dev_type=constants.LD_LV, size=size, |
3522 |
logical_id=(vgname, names[0]))
|
3523 |
lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
|
3524 |
logical_id=(vgname, names[1]))
|
3525 |
new_lvs = [lv_data, lv_meta] |
3526 |
old_lvs = dev.children |
3527 |
iv_names[dev.iv_name] = (dev, old_lvs, new_lvs) |
3528 |
info("creating new local storage on %s for %s" %
|
3529 |
(tgt_node, dev.iv_name)) |
3530 |
# since we *always* want to create this LV, we use the
|
3531 |
# _Create...OnPrimary (which forces the creation), even if we
|
3532 |
# are talking about the secondary node
|
3533 |
for new_lv in new_lvs: |
3534 |
if not _CreateBlockDevOnPrimary(cfg, tgt_node, instance, new_lv, |
3535 |
_GetInstanceInfoText(instance)): |
3536 |
raise errors.OpExecError("Failed to create new LV named '%s' on" |
3537 |
" node '%s'" %
|
3538 |
(new_lv.logical_id[1], tgt_node))
|
3539 |
|
3540 |
# Step: for each lv, detach+rename*2+attach
|
3541 |
self.proc.LogStep(4, steps_total, "change drbd configuration") |
3542 |
for dev, old_lvs, new_lvs in iv_names.itervalues(): |
3543 |
info("detaching %s drbd from local storage" % dev.iv_name)
|
3544 |
if not rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs): |
3545 |
raise errors.OpExecError("Can't detach drbd from local storage on node" |
3546 |
" %s for device %s" % (tgt_node, dev.iv_name))
|
3547 |
#dev.children = []
|
3548 |
#cfg.Update(instance)
|
3549 |
|
3550 |
# ok, we created the new LVs, so now we know we have the needed
|
3551 |
# storage; as such, we proceed on the target node to rename
|
3552 |
# old_lv to _old, and new_lv to old_lv; note that we rename LVs
|
3553 |
# using the assumption that logical_id == physical_id (which in
|
3554 |
# turn is the unique_id on that node)
|
3555 |
|
3556 |
# FIXME(iustin): use a better name for the replaced LVs
|
3557 |
temp_suffix = int(time.time())
|
3558 |
ren_fn = lambda d, suff: (d.physical_id[0], |
3559 |
d.physical_id[1] + "_replaced-%s" % suff) |
3560 |
# build the rename list based on what LVs exist on the node
|
3561 |
rlist = [] |
3562 |
for to_ren in old_lvs: |
3563 |
find_res = rpc.call_blockdev_find(tgt_node, to_ren) |
3564 |
if find_res is not None: # device exists |
3565 |
rlist.append((to_ren, ren_fn(to_ren, temp_suffix))) |
3566 |
|
3567 |
info("renaming the old LVs on the target node")
|
3568 |
if not rpc.call_blockdev_rename(tgt_node, rlist): |
3569 |
raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node) |
3570 |
# now we rename the new LVs to the old LVs
|
3571 |
info("renaming the new LVs on the target node")
|
3572 |
rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)] |
3573 |
if not rpc.call_blockdev_rename(tgt_node, rlist): |
3574 |
raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node) |
3575 |
|
3576 |
for old, new in zip(old_lvs, new_lvs): |
3577 |
new.logical_id = old.logical_id |
3578 |
cfg.SetDiskID(new, tgt_node) |
3579 |
|
3580 |
for disk in old_lvs: |
3581 |
disk.logical_id = ren_fn(disk, temp_suffix) |
3582 |
cfg.SetDiskID(disk, tgt_node) |
3583 |
|
3584 |
# now that the new lvs have the old name, we can add them to the device
|
3585 |
info("adding new mirror component on %s" % tgt_node)
|
3586 |
if not rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs): |
3587 |
for new_lv in new_lvs: |
3588 |
if not rpc.call_blockdev_remove(tgt_node, new_lv): |
3589 |
warning("Can't rollback device %s", hint="manually cleanup unused" |
3590 |
" logical volumes")
|
3591 |
raise errors.OpExecError("Can't add local storage to drbd") |
3592 |
|
3593 |
dev.children = new_lvs |
3594 |
cfg.Update(instance) |
3595 |
|
3596 |
# Step: wait for sync
|
3597 |
|
3598 |
# this can fail as the old devices are degraded and _WaitForSync
|
3599 |
# does a combined result over all disks, so we don't check its
|
3600 |
# return value
|
3601 |
self.proc.LogStep(5, steps_total, "sync devices") |
3602 |
_WaitForSync(cfg, instance, self.proc, unlock=True) |
3603 |
|
3604 |
# so check manually all the devices
|
3605 |
for name, (dev, old_lvs, new_lvs) in iv_names.iteritems(): |
3606 |
cfg.SetDiskID(dev, instance.primary_node) |
3607 |
is_degr = rpc.call_blockdev_find(instance.primary_node, dev)[5]
|
3608 |
if is_degr:
|
3609 |
raise errors.OpExecError("DRBD device %s is degraded!" % name) |
3610 |
|
3611 |
# Step: remove old storage
|
3612 |
self.proc.LogStep(6, steps_total, "removing old storage") |
3613 |
for name, (dev, old_lvs, new_lvs) in iv_names.iteritems(): |
3614 |
info("remove logical volumes for %s" % name)
|
3615 |
for lv in old_lvs: |
3616 |
cfg.SetDiskID(lv, tgt_node) |
3617 |
if not rpc.call_blockdev_remove(tgt_node, lv): |
3618 |
warning("Can't remove old LV", hint="manually remove unused LVs") |
3619 |
continue
|
3620 |
|
3621 |
def _ExecD8Secondary(self, feedback_fn): |
3622 |
"""Replace the secondary node for drbd8.
|
3623 |
|
3624 |
The algorithm for replace is quite complicated:
|
3625 |
- for all disks of the instance:
|
3626 |
- create new LVs on the new node with same names
|
3627 |
- shutdown the drbd device on the old secondary
|
3628 |
- disconnect the drbd network on the primary
|
3629 |
- create the drbd device on the new secondary
|
3630 |
- network attach the drbd on the primary, using an artifice:
|
3631 |
the drbd code for Attach() will connect to the network if it
|
3632 |
finds a device which is connected to the good local disks but
|
3633 |
not network enabled
|
3634 |
- wait for sync across all devices
|
3635 |
- remove all disks from the old secondary
|
3636 |
|
3637 |
Failures are not very well handled.
|
3638 |
|
3639 |
"""
|
3640 |
steps_total = 6
|
3641 |
warning, info = (self.proc.LogWarning, self.proc.LogInfo) |
3642 |
instance = self.instance
|
3643 |
iv_names = {} |
3644 |
vgname = self.cfg.GetVGName()
|
3645 |
# start of work
|
3646 |
cfg = self.cfg
|
3647 |
old_node = self.tgt_node
|
3648 |
new_node = self.new_node
|
3649 |
pri_node = instance.primary_node |
3650 |
|
3651 |
# Step: check device activation
|
3652 |
self.proc.LogStep(1, steps_total, "check device existence") |
3653 |
info("checking volume groups")
|
3654 |
my_vg = cfg.GetVGName() |
3655 |
results = rpc.call_vg_list([pri_node, new_node]) |
3656 |
if not results: |
3657 |
raise errors.OpExecError("Can't list volume groups on the nodes") |
3658 |
for node in pri_node, new_node: |
3659 |
res = results.get(node, False)
|
3660 |
if not res or my_vg not in res: |
3661 |
raise errors.OpExecError("Volume group '%s' not found on %s" % |
3662 |
(my_vg, node)) |
3663 |
for dev in instance.disks: |
3664 |
if not dev.iv_name in self.op.disks: |
3665 |
continue
|
3666 |
info("checking %s on %s" % (dev.iv_name, pri_node))
|
3667 |
cfg.SetDiskID(dev, pri_node) |
3668 |
if not rpc.call_blockdev_find(pri_node, dev): |
3669 |
raise errors.OpExecError("Can't find device %s on node %s" % |
3670 |
(dev.iv_name, pri_node)) |
3671 |
|
3672 |
# Step: check other node consistency
|
3673 |
self.proc.LogStep(2, steps_total, "check peer consistency") |
3674 |
for dev in instance.disks: |
3675 |
if not dev.iv_name in self.op.disks: |
3676 |
continue
|
3677 |
info("checking %s consistency on %s" % (dev.iv_name, pri_node))
|
3678 |
if not _CheckDiskConsistency(self.cfg, dev, pri_node, True, ldisk=True): |
3679 |
raise errors.OpExecError("Primary node (%s) has degraded storage," |
3680 |
" unsafe to replace the secondary" %
|
3681 |
pri_node) |
3682 |
|
3683 |
# Step: create new storage
|
3684 |
self.proc.LogStep(3, steps_total, "allocate new storage") |
3685 |
for dev in instance.disks: |
3686 |
size = dev.size |
3687 |
info("adding new local storage on %s for %s" % (new_node, dev.iv_name))
|
3688 |
# since we *always* want to create this LV, we use the
|
3689 |
# _Create...OnPrimary (which forces the creation), even if we
|
3690 |
# are talking about the secondary node
|
3691 |
for new_lv in dev.children: |
3692 |
if not _CreateBlockDevOnPrimary(cfg, new_node, instance, new_lv, |
3693 |
_GetInstanceInfoText(instance)): |
3694 |
raise errors.OpExecError("Failed to create new LV named '%s' on" |
3695 |
" node '%s'" %
|
3696 |
(new_lv.logical_id[1], new_node))
|
3697 |
|
3698 |
iv_names[dev.iv_name] = (dev, dev.children) |
3699 |
|
3700 |
self.proc.LogStep(4, steps_total, "changing drbd configuration") |
3701 |
for dev in instance.disks: |
3702 |
size = dev.size |
3703 |
info("activating a new drbd on %s for %s" % (new_node, dev.iv_name))
|
3704 |
# create new devices on new_node
|
3705 |
new_drbd = objects.Disk(dev_type=constants.LD_DRBD8, |
3706 |
logical_id=(pri_node, new_node, |
3707 |
dev.logical_id[2]),
|
3708 |
children=dev.children) |
3709 |
if not _CreateBlockDevOnSecondary(cfg, new_node, instance, |
3710 |
new_drbd, False,
|
3711 |
_GetInstanceInfoText(instance)): |
3712 |
raise errors.OpExecError("Failed to create new DRBD on" |
3713 |
" node '%s'" % new_node)
|
3714 |
|
3715 |
for dev in instance.disks: |
3716 |
# we have new devices, shutdown the drbd on the old secondary
|
3717 |
info("shutting down drbd for %s on old node" % dev.iv_name)
|
3718 |
cfg.SetDiskID(dev, old_node) |
3719 |
if not rpc.call_blockdev_shutdown(old_node, dev): |
3720 |
warning("Failed to shutdown drbd for %s on old node" % dev.iv_name,
|
3721 |
hint="Please cleanup this device manually as soon as possible")
|
3722 |
|
3723 |
info("detaching primary drbds from the network (=> standalone)")
|
3724 |
done = 0
|
3725 |
for dev in instance.disks: |
3726 |
cfg.SetDiskID(dev, pri_node) |
3727 |
# set the physical (unique in bdev terms) id to None, meaning
|
3728 |
# detach from network
|
3729 |
dev.physical_id = (None,) * len(dev.physical_id) |
3730 |
# and 'find' the device, which will 'fix' it to match the
|
3731 |
# standalone state
|
3732 |
if rpc.call_blockdev_find(pri_node, dev):
|
3733 |
done += 1
|
3734 |
else:
|
3735 |
warning("Failed to detach drbd %s from network, unusual case" %
|
3736 |
dev.iv_name) |
3737 |
|
3738 |
if not done: |
3739 |
# no detaches succeeded (very unlikely)
|
3740 |
raise errors.OpExecError("Can't detach at least one DRBD from old node") |
3741 |
|
3742 |
# if we managed to detach at least one, we update all the disks of
|
3743 |
# the instance to point to the new secondary
|
3744 |
info("updating instance configuration")
|
3745 |
for dev in instance.disks: |
3746 |
dev.logical_id = (pri_node, new_node) + dev.logical_id[2:]
|
3747 |
cfg.SetDiskID(dev, pri_node) |
3748 |
cfg.Update(instance) |
3749 |
|
3750 |
# and now perform the drbd attach
|
3751 |
info("attaching primary drbds to new secondary (standalone => connected)")
|
3752 |
failures = [] |
3753 |
for dev in instance.disks: |
3754 |
info("attaching primary drbd for %s to new secondary node" % dev.iv_name)
|
3755 |
# since the attach is smart, it's enough to 'find' the device,
|
3756 |
# it will automatically activate the network, if the physical_id
|
3757 |
# is correct
|
3758 |
cfg.SetDiskID(dev, pri_node) |
3759 |
if not rpc.call_blockdev_find(pri_node, dev): |
3760 |
warning("can't attach drbd %s to new secondary!" % dev.iv_name,
|
3761 |
"please do a gnt-instance info to see the status of disks")
|
3762 |
|
3763 |
# this can fail as the old devices are degraded and _WaitForSync
|
3764 |
# does a combined result over all disks, so we don't check its
|
3765 |
# return value
|
3766 |
self.proc.LogStep(5, steps_total, "sync devices") |
3767 |
_WaitForSync(cfg, instance, self.proc, unlock=True) |
3768 |
|
3769 |
# so check manually all the devices
|
3770 |
for name, (dev, old_lvs) in iv_names.iteritems(): |
3771 |
cfg.SetDiskID(dev, pri_node) |
3772 |
is_degr = rpc.call_blockdev_find(pri_node, dev)[5]
|
3773 |
if is_degr:
|
3774 |
raise errors.OpExecError("DRBD device %s is degraded!" % name) |
3775 |
|
3776 |
self.proc.LogStep(6, steps_total, "removing old storage") |
3777 |
for name, (dev, old_lvs) in iv_names.iteritems(): |
3778 |
info("remove logical volumes for %s" % name)
|
3779 |
for lv in old_lvs: |
3780 |
cfg.SetDiskID(lv, old_node) |
3781 |
if not rpc.call_blockdev_remove(old_node, lv): |
3782 |
warning("Can't remove LV on old secondary",
|
3783 |
hint="Cleanup stale volumes by hand")
|
3784 |
|
3785 |
def Exec(self, feedback_fn): |
3786 |
"""Execute disk replacement.
|
3787 |
|
3788 |
This dispatches the disk replacement to the appropriate handler.
|
3789 |
|
3790 |
"""
|
3791 |
instance = self.instance
|
3792 |
if instance.disk_template == constants.DT_REMOTE_RAID1:
|
3793 |
fn = self._ExecRR1
|
3794 |
elif instance.disk_template == constants.DT_DRBD8:
|
3795 |
if self.op.remote_node is None: |
3796 |
fn = self._ExecD8DiskOnly
|
3797 |
else:
|
3798 |
fn = self._ExecD8Secondary
|
3799 |
else:
|
3800 |
raise errors.ProgrammerError("Unhandled disk replacement case") |
3801 |
return fn(feedback_fn)
|
3802 |
|
3803 |
|
3804 |
class LUQueryInstanceData(NoHooksLU): |
3805 |
"""Query runtime instance data.
|
3806 |
|
3807 |
"""
|
3808 |
_OP_REQP = ["instances"]
|
3809 |
|
3810 |
def CheckPrereq(self): |
3811 |
"""Check prerequisites.
|
3812 |
|
3813 |
This only checks the optional instance list against the existing names.
|
3814 |
|
3815 |
"""
|
3816 |
if not isinstance(self.op.instances, list): |
3817 |
raise errors.OpPrereqError("Invalid argument type 'instances'") |
3818 |
if self.op.instances: |
3819 |
self.wanted_instances = []
|
3820 |
names = self.op.instances
|
3821 |
for name in names: |
3822 |
instance = self.cfg.GetInstanceInfo(self.cfg.ExpandInstanceName(name)) |
3823 |
if instance is None: |
3824 |
raise errors.OpPrereqError("No such instance name '%s'" % name) |
3825 |
self.wanted_instances.append(instance)
|
3826 |
else:
|
3827 |
self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name |
3828 |
in self.cfg.GetInstanceList()] |
3829 |
return
|
3830 |
|
3831 |
|
3832 |
def _ComputeDiskStatus(self, instance, snode, dev): |
3833 |
"""Compute block device status.
|
3834 |
|
3835 |
"""
|
3836 |
self.cfg.SetDiskID(dev, instance.primary_node)
|
3837 |
dev_pstatus = rpc.call_blockdev_find(instance.primary_node, dev) |
3838 |
if dev.dev_type in constants.LDS_DRBD: |
3839 |
# we change the snode then (otherwise we use the one passed in)
|
3840 |
if dev.logical_id[0] == instance.primary_node: |
3841 |
snode = dev.logical_id[1]
|
3842 |
else:
|
3843 |
snode = dev.logical_id[0]
|
3844 |
|
3845 |
if snode:
|
3846 |
self.cfg.SetDiskID(dev, snode)
|
3847 |
dev_sstatus = rpc.call_blockdev_find(snode, dev) |
3848 |
else:
|
3849 |
dev_sstatus = None
|
3850 |
|
3851 |
if dev.children:
|
3852 |
dev_children = [self._ComputeDiskStatus(instance, snode, child)
|
3853 |
for child in dev.children] |
3854 |
else:
|
3855 |
dev_children = [] |
3856 |
|
3857 |
data = { |
3858 |
"iv_name": dev.iv_name,
|
3859 |
"dev_type": dev.dev_type,
|
3860 |
"logical_id": dev.logical_id,
|
3861 |
"physical_id": dev.physical_id,
|
3862 |
"pstatus": dev_pstatus,
|
3863 |
"sstatus": dev_sstatus,
|
3864 |
"children": dev_children,
|
3865 |
} |
3866 |
|
3867 |
return data
|
3868 |
|
3869 |
def Exec(self, feedback_fn): |
3870 |
"""Gather and return data"""
|
3871 |
result = {} |
3872 |
for instance in self.wanted_instances: |
3873 |
remote_info = rpc.call_instance_info(instance.primary_node, |
3874 |
instance.name) |
3875 |
if remote_info and "state" in remote_info: |
3876 |
remote_state = "up"
|
3877 |
else:
|
3878 |
remote_state = "down"
|
3879 |
if instance.status == "down": |
3880 |
config_state = "down"
|
3881 |
else:
|
3882 |
config_state = "up"
|
3883 |
|
3884 |
disks = [self._ComputeDiskStatus(instance, None, device) |
3885 |
for device in instance.disks] |
3886 |
|
3887 |
idict = { |
3888 |
"name": instance.name,
|
3889 |
"config_state": config_state,
|
3890 |
"run_state": remote_state,
|
3891 |
"pnode": instance.primary_node,
|
3892 |
"snodes": instance.secondary_nodes,
|
3893 |
"os": instance.os,
|
3894 |
"memory": instance.memory,
|
3895 |
"nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics], |
3896 |
"disks": disks,
|
3897 |
"network_port": instance.network_port,
|
3898 |
"vcpus": instance.vcpus,
|
3899 |
"kernel_path": instance.kernel_path,
|
3900 |
"initrd_path": instance.initrd_path,
|
3901 |
"hvm_boot_order": instance.hvm_boot_order,
|
3902 |
} |
3903 |
|
3904 |
result[instance.name] = idict |
3905 |
|
3906 |
return result
|
3907 |
|
3908 |
|
3909 |
class LUSetInstanceParams(LogicalUnit): |
3910 |
"""Modifies an instances's parameters.
|
3911 |
|
3912 |
"""
|
3913 |
HPATH = "instance-modify"
|
3914 |
HTYPE = constants.HTYPE_INSTANCE |
3915 |
_OP_REQP = ["instance_name"]
|
3916 |
|
3917 |
def BuildHooksEnv(self): |
3918 |
"""Build hooks env.
|
3919 |
|
3920 |
This runs on the master, primary and secondaries.
|
3921 |
|
3922 |
"""
|
3923 |
args = dict()
|
3924 |
if self.mem: |
3925 |
args['memory'] = self.mem |
3926 |
if self.vcpus: |
3927 |
args['vcpus'] = self.vcpus |
3928 |
if self.do_ip or self.do_bridge or self.mac: |
3929 |
if self.do_ip: |
3930 |
ip = self.ip
|
3931 |
else:
|
3932 |
ip = self.instance.nics[0].ip |
3933 |
if self.bridge: |
3934 |
bridge = self.bridge
|
3935 |
else:
|
3936 |
bridge = self.instance.nics[0].bridge |
3937 |
if self.mac: |
3938 |
mac = self.mac
|
3939 |
else:
|
3940 |
mac = self.instance.nics[0].mac |
3941 |
args['nics'] = [(ip, bridge, mac)]
|
3942 |
env = _BuildInstanceHookEnvByObject(self.instance, override=args)
|
3943 |
nl = [self.sstore.GetMasterNode(),
|
3944 |
self.instance.primary_node] + list(self.instance.secondary_nodes) |
3945 |
return env, nl, nl
|
3946 |
|
3947 |
def CheckPrereq(self): |
3948 |
"""Check prerequisites.
|
3949 |
|
3950 |
This only checks the instance list against the existing names.
|
3951 |
|
3952 |
"""
|
3953 |
self.mem = getattr(self.op, "mem", None) |
3954 |
self.vcpus = getattr(self.op, "vcpus", None) |
3955 |
self.ip = getattr(self.op, "ip", None) |
3956 |
self.mac = getattr(self.op, "mac", None) |
3957 |
self.bridge = getattr(self.op, "bridge", None) |
3958 |
self.kernel_path = getattr(self.op, "kernel_path", None) |
3959 |
self.initrd_path = getattr(self.op, "initrd_path", None) |
3960 |
self.hvm_boot_order = getattr(self.op, "hvm_boot_order", None) |
3961 |
all_params = [self.mem, self.vcpus, self.ip, self.bridge, self.mac, |
3962 |
self.kernel_path, self.initrd_path, self.hvm_boot_order] |
3963 |
if all_params.count(None) == len(all_params): |
3964 |
raise errors.OpPrereqError("No changes submitted") |
3965 |
if self.mem is not None: |
3966 |
try:
|
3967 |
self.mem = int(self.mem) |
3968 |
except ValueError, err: |
3969 |
raise errors.OpPrereqError("Invalid memory size: %s" % str(err)) |
3970 |
if self.vcpus is not None: |
3971 |
try:
|
3972 |
self.vcpus = int(self.vcpus) |
3973 |
except ValueError, err: |
3974 |
raise errors.OpPrereqError("Invalid vcpus number: %s" % str(err)) |
3975 |
if self.ip is not None: |
3976 |
self.do_ip = True |
3977 |
if self.ip.lower() == "none": |
3978 |
self.ip = None |
3979 |
else:
|
3980 |
if not utils.IsValidIP(self.ip): |
3981 |
raise errors.OpPrereqError("Invalid IP address '%s'." % self.ip) |
3982 |
else:
|
3983 |
self.do_ip = False |
3984 |
self.do_bridge = (self.bridge is not None) |
3985 |
if self.mac is not None: |
3986 |
if self.cfg.IsMacInUse(self.mac): |
3987 |
raise errors.OpPrereqError('MAC address %s already in use in cluster' % |
3988 |
|